Final Technical Support Document
Fuel Economy Labeling of Motor Vehicle
Revisions to Improve Calculation of Fuel
Economy Estimates
&EPA
United States
Environmental Protection
Agency
Office of Transportation and Air Quality
EPA420-R-06-017
December 2006
-------
Final Technical Support Document
Fuel Economy Labeling of Motor Vehicles:
Revisions to Improve Calculation of Fuel
Economy Estimates
Assessment and Standards Division
and
Certification and Innovative Strategies Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
v>EPA
United States EPA420-R-06-017
Environmental Protection December 2006
Agency
-------
Table of Contents
LIST OF TABLES m
LIST OF FIGURES v
LIST OF ACRONYMS vi
CHAPTER I: EXECUTIVE SUMMARY 1
CHAPTER II: CURRENT AND PROPOSED LABEL VALUES COMPARED TO ONROAD
ESTIMATES 7
A. ONROAD FUEL ECONOMY ESTIMATES DURING TYPICAL OPERATION 7
7. ORNL "YourMPG"Program 7
2. DOE FreedomCar Program 9
3. Strategic Visions New Vehicle Survey 12
4. Kansas City Instrumented Vehicle Study 13
B. FUEL ECONOMY ESTIMATES BY INDEPENDENT ORGANIZATIONS 15
1. Consumer Reports Estimates of Onroad Fuel Economy 15
2. AAA Estimates of Onroad Fuel Economy 20
3. Edmunds 22
C. FLEET-WIDE ESTIMATES OF ONROAD FUEL ECONOMY 25
D. OVERALL COMPARISON OF HYBRID FUEL ECONOMY 28
CHAPTER II REFERENCES 32
CHAPTER III: DOCUMENTATION OF FINAL APPROACH FOR ESTIMATING ON-ROAD FUEL
ECONOMY 33
A. VEHICLE SPECIFIC S-CYCLE METHOD FOR ESTIMATING ON-ROAD FUEL ECONOMY FROM DYNAMOMETER
TESTS 34
7. Start Fuel Use 36
a. Start Fuel 36
b. Trip Length 44
c. Formula for Start Fuel Use 50
2. Running Fuel Use at 75°F Without Air Conditioning 51
a. On-Road Driving Patterns 51
b. Representative Mix of Dynamometer Driving Cycles 61
3. Effect of Air Conditioning on Fuel Economy 69
4. Effect of Cold Ambient Temperatures on Running Fuel Use 78
5. Adjustment Factor for Non-Dynamometer Effects 81
6. 5-Cycle Fuel Economy Formulae 92
a. 5-Cycle Fuel Economy Formulae 92
b. Alternative 5-cycle Highway Fuel Economy Formula 97
B. DERIVATION OF THE MPG-BASED APPROACH 100
C. VARIABILITY IN ONROAD FUEL ECONOMY 106
D. IMP ACT OF THE S-CYCLE AND MPG-BASED FORMULAE ON FUEL ECONOMY LABELS 110
E. SENSITIVITIES AND UNCERTAINTIES IN THE S-CYCLE FUEL ECONOMY FORMULAE 113
7. Start Fuel Use 114
2. Running Fuel Use At 75°F. 118
a. Alternative Definition of US06 City and Highway Bags 118
b. Elimination of Three Highest Speed Freeway Cycles in Draft MOVES2004 122
c. Alternative Fuel Rates and Number of VSPBins 123
d. Kansas City VSP Distributions and Fuel Rates 125
e. California Chase Car Studies 127
f Alternative Splits of City/Highway Driving 128
g. Complete Cycles 132
3. Air Conditioning Effects 134
4. Cold Temperature Running Fuel Use 135
CHAPTER III REFERENCES 138
-------
CHAPTER IV: ECONOMIC IMPACTS 142
A. TESTING AND FACILITIES BURDEN 142
7. Test Volume 142
a. Testing Burden for MY 2008 through 2010 142
b. Testing Burden for MY 2011 and After 143
2. Facilities Burden 146
3. Startup Burden 147
4. Summary 149
B. IMPACT ON CONFIRMATORY TESTING OF VEHICLES 151
C. CHANGES TO LABEL FORMAT AND CONTENT 152
D. CERTIFICATION FEES 152
APPENDIX A: EPA KANSAS CITY TEST PROGRAM 153
A. QUALITY ASSURANCE 154
B. ONROAD FUEL ECONOMY 155
C. RECENT DRIVING ACTIVITY IN KANSAS CITY AND CALIFORNIA 156
D. EVALUATION OF 5-CYCLE APPROACH TO FUEL ECONOMY ESTIMATION 161
11
-------
List of Tables
TABLE I-l AGGREGATE COSTS 6
TABLE II.A-1. YouRMPG VERSUS CURRENT EPA LABEL FUEL ECONOMY 8
TABLE II.A-2. FREEDOMCAR HYBRID FLEET CUMULATIVE VERSUS EPA COMPOSITE LABEL FUEL ECONOMY ... 10
TABLE II.B-1. CONSUMER REPORTS AND CURRENT EPA AND MPG-BASED FUEL ECONOMY: 303 VEHICLES 16
TABLE II.B-2. CR AND CURRENT EPA, S-CYCLE AND MPG-BASED FUEL ECONOMY: 70 VEHICLES 16
TABLE II.B-3. COMPARISON OF CONSUMER REPORTS AND EPA FUEL ECONOMY VALUES FOR HYBRIDS 18
TABLE II.B-4. AAA AND CURRENT EPA, S-CYCLE AND MPG-BASED FUEL ECONOMY ESTIMATES 20
TABLE II.B-5. 2003-2006 EDMUNDS LONG-TERM TEST VEHICLES 23
TABLE II.B-6. EDMUNDS LONG-TERM TEST VEHICLES COMPARED TO EPA COMBINED MPG ESTIMATES 24
TABLE II.C-1. FHWA-BASED ESTIMATE OF ONROAD FUEL ECONOMY 26
TABLE II.C-2. BREAKDOWN OF VMT AND FUEL USE BY 4-WHEEL, 2-AxLE TRUCKS 27
TABLE II.D-1. ONROAD HYBRID FUEL ECONOMY VERSUS EPA LABEL ESTIMATES (MPG) 30
TABLE II.D-2. ONROAD HYBRID FUEL ECONOMY ESTIMATES VERSUS EPA LABEL ESTIMATES (MPG) 31
TABLE III. A-l. KEY FEATURES OF THE FIVE CURRENT EMISSION AND FUEL ECONOMY TESTS 34
TABLE III.A-2. DISTRIBUTION OF STARTS BY HOUR OF THE DAY (IN PERCENT) 40
TABLE III. A-3. BREAKDOWN OF ANNUAL VMT BY MONTH 41
TABLE III. A-4. DISTRIBUTION OF STARTS BY SOAK TIME: THREE HOURS DURING WEEKDAYS 42
TABLE III.A-5. ESTIMATION OF DAILY AVERAGE OVERNIGHT SOAK EQUIVALENT 43
TABLE III.A-6. TRIP AND START RELATED INFORMATION IN DRAFT MOVES2004 44
TABLE III. A-7. ESTIMATES OF IN-USE AVERAGE TRIP LENGTH 45
TABLE III.A-8. INVENTORY DRIVING CYCLES IN DRAFT MOVES2004 49
TABLE III.A-9. VSP-SPEED BINS IN DRAFT 2004MOVES 54
TABLE III.A-10. EXPANDED SET OF 26 VSP-SPEED BINS 55
TABLE III.A-11. VSP FREQUENCY DISTRIBUTIONS FOR ONROAD DRIVING CYCLES IN MOVES 56
TABLE III.A-12. VSP FREQUENCY DISTRIBUTIONS FOR ONROAD DRIVING CYCLES IN MOVES 57
TABLE III. A-13. DISTRIBUTION OF ONROAD DRIVING PATTERNS: DRAFT MOVES2004 58
TABLE III. A-14. VSP DISTRIBUTIONS FOR U.S. DRIVING (% OF TIME) 59
TABLE III.A-15. DRIVING CHARACTERISTICS OF THE CURRENT DYNAMOMETER TESTS 62
TABLE III. A-16. SPLIT OF US06 CYCLE INTO CITY AND HIGHWAY PORTIONS 63
TABLE III. A-17. VSP DISTRIBUTIONS FOR DYNAMOMETER CYCLES (% OF TIME) 64
TABLE III. A-18. 26-BiN VSP FUEL RATES (GRAM PER SECOND) 66
TABLE III.A-19. BEST-FIT COMBINATIONS OF DYNAMOMETER CYCLES 67
TABLE III. A-20. COEFFICIENTS FOR A/C COMPRESSOR Us AGE EQUATIONS 72
TABLE III.A-21. INCREASED FUEL USE DUE TO AIR CONDITIONING AS A FUNCTION OF VEHICLE SPEED 75
TABLE III. A-22. WARMED UP FUEL USE VERSUS TEMPERATURE: HONDA DATA 79
TABLE III. A-23. NHTSA ONROAD TIRE PRESSURE SURVEY 83
TABLE III. A-24. EFFECT OF WIND ANGLE ON VEHICLE DRAG COEFFICIENT 86
TABLE III. A-25. FREQUENCY OF WIND SPEEDS IN THE U.S 87
TABLE III. A-26. EFFECT OF ROAD ROUGHNESS ON ONROAD FUEL ECONOMY: 1977 88
TABLE III. A-27. MAPPING OF ROADWAY SURFACES 89
TABLE III. A-28. EFFECT OF NON-DYNAMOMETER FACTORS ON ONROAD FUEL ECONOMY 90
TABLE III.B-1. RATIO OF FTP BAG TO CYCLE FUEL CONSUMPTION 100
TABLE III.B-2. EFFECT OF HEATER/DEFROSTER USE ON COLD FTP FUEL USE 102
TABLE III.D-1. CURRENT AND S-CYCLE LABEL FUEL ECONOMIES BY MODEL TYPE Ill
TABLE III.D-2. CURRENT AND S-CYCLE LABEL FUEL ECONOMY BY PROPULSION SYSTEM Ill
TABLE III.D-3. EFFECT OF MPG-BASED FORMULAE ON CITY AND HIGHWAY FUEL ECONOMY 112
TABLE II.D-4. EFFECT OF NEW METHODS ON FUEL ECONOMY ESTIMATES FOR MAJOR MANUFACTURERS 113
TABLE III.D-5. EFFECT OF VARIOUS FACTORS ON 5-CYCLE FUEL ECONOMY 113
TABLE III.E-1. SENSITIVITY OF HYBRID START FUEL USE TO AMBIENT TEMPERATURE 115
TABLE III.E-2. SPLIT OF US06 CYCLE INTO CITY AND HIGHWAY PORTIONS 119
TABLE III.E-3. VSP DISTRIBUTIONS FOR US06 CITY AND HIGHWAY BAGS (% OF TIME) 120
TABLE III.E-4. BAG/CYCLE COMBINATIONS FOR CITY AND HIGHWAY DRIVING: ALTERNATIVE US06 SPLITS 121
TABLE III.E-5. AVERAGE 5-CYCLE FUEL ECONOMY: ALTERNATIVE US06 SPLITS 121
in
-------
TABLE III.E-6. BAG/CYCLE COMBINATIONS FOR HIGHWAY DRIVING: HIGH SPEED FREEWAY CYCLES 123
TABLE III.E-7. BAG/CYCLE COMBINATIONS FOR CITY AND HIGHWAY DRIVING: ALTERNATIVE FUEL RATES 124
TABLE III.E-8. KANSAS CITY VSP DISTRIBUTIONS AND FUEL RATES 125
TABLE III.E-9. BAG/CYCLE COMBINATIONS FOR CITY AND HIGHWAY DRIVING: KANSAS CITY 126
TABLE III.E-10. CALIFORNIA URBAN AND RURAL VSP DISTRIBUTIONS 127
TABLE III.E-11. BAG/CYCLE COMBINATIONS FOR CITY AND HIGHWAY DRIVING: CALIFORNIA 128
TABLE III.E-12. VSP DISTRIBUTIONS FOR U.S. DRIVING WITH ALTERNATIVE DEFINITION OF CITY DRIVING (% OF
TIME) 130
TABLE III.E-13. CYCLE COMBINATIONS FOR CITY AND HIGHWAY DRIVING: REVISED CITY/HIGHWAY SPLIT 131
TABLE III.E-14. S-CYCLE FUEL ECONOMY VALUES: EFFECT OF THE DEFINITION OF CITY DRIVING (MPG) 132
TABLE III.E-15. BAG/CYCLE COMBINATIONS FOR COMPLETE CYCLE ALTERNATIVES 133
TABLE III.E-16. EFFECT OF USING WHOLE CYCLES ON 5-CYCLE FUEL ECONOMY VALUES (MPG) 133
TABLE IV.A-1. ESTIMATED COST PER TEST: PROPOSED AND FINAL 145
TABLE IV. A-2. ESTIMATED INCREASE IN NUMBER OF TESTS FOR MODEL YEARS 2008-2010: PROPOSED AND FINAL
145
TABLE IV.A-3. ESTIMATED INCREASE IN NUMBER OF TESTS FOR MODEL YEARS 2011 AND LATER: PROPOSED... 145
TABLE IV.A-4. ESTIMATED INCREASE IN NUMBER OF TESTS FOR MODEL YEARS 2011 AND LATER: FINAL 146
TABLE IV.A-5. ESTIMATED FACILITY COSTS PROPOSED 147
TABLE IV.A-6. ESTIMATED FACILITY COSTS: FINAL 147
TABLE IV.A-7. ESTIMATED STARTUP COSTS PROPOSED 149
TABLE IV.A-8. ESTIMATED STARTUP COSTS: FINAL 149
TABLE IV.A-9. ESTIMATED TOTAL COSTS PROPOSED 150
TABLE IV.A-10. ESTIMATED TOTAL COSTS: FINAL 150
TABLE IV.A-11. ESTIMATED TOTAL HOURS PROPOSED 150
TABLE IV.A-12. ESTIMATED TOTAL HOURS: FINAL 151
TABLE A-l. PROCESSING OF RAW DATA OBTAINED IN KANSAS CITY 155
TABLE A-2. CYCLE COMBINATIONS FOR THE MITSUBISHI MONTERO SPORT 164
TABLE A-3. ONROAD AND PREDICTED FUEL ECONOMY: KANSAS CITY TEST PROGRAM 166
TABLE A-4. ON-ROAD AND MODELED FUEL ECONOMIES USING VEHICLE-SPECIFIC CYCLE WEIGHTS (MPG) .. 168
TABLE A-5. COMPARISON OF CYCLE FUEL ECONOMY 169
IV
-------
List of Figures
FIGURE II-1. DISTRIBUTION OF ONROAD FUEL ECONOMY ESTIMATES - STRATEGIC VISION, 2006 12
FIGURE II-2. COMPARISON ONROAD TO CURRENT LABEL ECONOMY: KANSAS CITY 14
FIGURE III-l. SPEED-ACCELERATION FREQUENCY DISTRIBUTION: KANSAS CITY Vs. TEST CYCLES 52
FIGURE III-2. SPEED-ACCELERATION FREQUENCY DISTRIBUTION: URBAN CALIFORNIA Vs. TEST CYCLES 53
FIGURE III-3. KANSAS CITY AND CALIFORNIA VSP FREQUENCY DISTRIBUTIONS vs. MOVES 60
FIGURE III-4. VSP FREQUENCY DISTRIBUTIONS IN KANSAS CITY: HYBRIDS vs. NON-HYBRIDS 61
FIGURE III-5. HEAT INDEX vs. TEMPERATURE AND HUMIDITY 71
FIGURE III-6. AIR CONDITIONING USE IN PHOENIX 73
FIGURE III-7. COMPRESSOR ENGAGEMENT AS A FUNCTION OF AMBIENT TEMPERATURE 73
FIGURE III-8. 5-CYCLE CITY VERSUS FTP FUEL CONSUMPTION 104
FIGURE III-9. 5-CYCLE HIGHWAY VERSUS HFET FUEL CONSUMPTION 104
FIGURE 111-10. MPG-BASED CITY FUEL ECONOMY 105
FIGURE III-l 1. MPG-BASED HIGHWAY FUEL ECONOMY 105
FIGURE 111-12. ONROAD FE VERSUS PRE-1984 EPA CITY LABEL FOR CITY DRIVEN CARS 106
FIGURE 111-13. VARIABILITY IN ONROAD FE 110
FIGURE A-l. COMPARISON ONROAD TO CURRENT LABEL ECONOMY: KANSAS CITY 156
FIGURE A-2. SPEED-ACCELERATION FREQUENCY DISTRIBUTION: KANSAS CITY Vs. TEST CYCLES 157
FIGURE A-3. SPEED-ACCELERATION FREQUENCY DISTRIBUTION: URBAN CALIFORNIA Vs. TEST CYCLES 158
FIGURE A-4. KANSAS CITY AND CALIFORNIA VSP FREQUENCY DISTRIBUTIONS vs. MOVES 159
FIGURE A-5. VSP FREQUENCY DISTRIBUTIONS IN KANSAS CITY: HYBRIDS vs. NON-HYBRIDS 160
-------
List of Acronyms
AAA American Automobile Association
ADFE analytically derived fuel economy
AVTA Advanced Vehicle Testing Activity
CAFE Corporate Average Fuel Economy
CO Carbon Monoxide
CO2 Carbon Dioxide
CR Consumer Reports
CRC Coordinating Research Council
DOE Department of Energy
DOT Department of Transportation
EIA Energy Information Administration
EPA Environmental Protection Agency
FHWA Federal Highway Administration
FTP Federal Test Procedure
GPS geographical positioning system
HC Hydrocarbon
HFET Highway Fuel Economy Test
INL Idaho National Laboratory
LOT light-duty truck
LDV light-duty vehicle
LOS level of service (volume of traffic)
MOVES Motor Vehicle Emission Inventory System
mpg miles per gallon
MY model year
NHTS National Household Travel Survey
NHTSA National Highway Traffic Safety Administration
NOx Oxides of Nitrogen
NREL National Renewable Energy Laboratory
OAP Office of Atmospheric Programs
ORNL Oak Ridge National Laboratory
PEMS Portable Emissions Measurement System
PERE Physical Emission Rate Estimator
RVP Reid Vapor Pressure
SAFD speed-acceleration frequency distribution
SFTP Supplemental Federal Test Procedure
SUV sport utility vehicle
TRLHP tractive road load horsepower
TSD Technical Support Document
UDDS Urban Dynamometer Driving Schedule
VMT vehicle miles traveled
VSP vehicle specific power
VI
-------
Chapter I: Executive Summary
The EPA fuel economy estimates have appeared on the window stickers of all new cars
and light trucks since the late 1970's and are well-recognized by consumers. The fuel economy
estimates essentially serve two purposes: to provide consumers with a basis on which to
compare the fuel economy of different vehicles, and to provide consumers with a reasonable
estimate of the range of fuel economy they can expect to achieve. While the estimates
historically have been a valuable tool for comparison shopping purposes, attention has been
focused recently on how closely the EPA estimates approximate consumers' real-world fuel
economy experience.
We are making changes to EPA's fuel economy test methods to bring the estimates closer
to the fuel economy consumers are achieving in the real-world. We believe these estimates will
provide car buyers with useful information when comparing the fuel economy of different
vehicles. It is important to emphasize that fuel economy varies from driver to driver for a wide
variety of reasons, such as different driving styles, climates, traffic patterns, use of accessories,
loads, weather, and vehicle maintenance. Even different drivers of the same vehicle will
experience different fuel economy as these and other factors vary. Therefore, it is impossible to
design a "perfect" fuel economy test that will provide accurate real-world fuel economy
estimates for every consumer. With any estimate, there will always be consumers that get better
or worse actual fuel economy. The EPA estimates are meant to be a general guideline for
consumers, particularly to compare the relative fuel economy of one vehicle to another.
Nevertheless, we do believe that today's new fuel economy test methods will do a better job of
giving consumers a more accurate estimate of the fuel economy they can achieve in the real-
world.
It is essential that our fuel economy estimates continue to be derived from controlled,
repeatable, laboratory tests. However, the inputs to our estimates are based on data from actual
real-world driving behavior and conditions. Because the test is controlled and repeatable, an
EPA fuel economy test result can be used for comparison of different vehicle models and types.
EPA and manufacturers test over 1,250 vehicle models annually and every test is run under
identical conditions and under a precise driver's trace, which assures that the result will be the
same for an individual vehicle model no matter when and where the laboratory test is performed.
Variations in temperature, road grade, driving patterns, and other variables do not impact the
result of the test. While such external conditions impact fuel economy on a trip-to-trip basis,
they do not change the laboratory test result. Therefore, a repeatable test provides a level playing
field for all vehicles, which is essential for comparing the fuel economy of one vehicle to
another. Finally, EPA must preserve the ability to confirm the values achieved by the
manufacturers' testing, and this can only be achieved with a highly repeatable test or set of tests.
No other fuel economy test program provides the level of repeatability as the EPA program.
However, the EPA fuel economy test methods need to reflect real world conditions as
well as being a repeatable test. While some consumer groups have issued their own fuel
economy numbers based on on-road driving, this approach introduces a wide number of
variables - different drivers, driving patterns, weather conditions, temperatures, etc. - that make
repeatability impossible. Our new fuel economy test methods are more representative of real-
-------
world conditions than the current fuel economy tests - yet we retain our practice of relying on
controlled, repeatable, laboratory tests.
The methods used today for calculating the city and highway mpg estimates were
established in the 1970's, and were adjusted in the mid-1980's. Since these adjustments were
made, America's driving behavior has changed. In the past 20 years, speed limits have increased
and vehicles have been designed for higher power - as a result, Americans are driving faster and
more aggressively than ever before. Vehicle technology has changed markedly, and many more
vehicles are equipped with energy-consuming accessories like air conditioning. These and other
factors are not accounted for in the current test procedures used to determine the city and
highway mpg estimates. Our analyses indicate that if these factors were better accounted for, the
city and highway fuel economy label estimates would be generally lower and closer to the
average real-world experience of consumers.
A fundamental issue with today's fuel economy estimates is that the underlying test
procedures do not fully represent real-world driving conditions. Some of the key limitations are
that the highway test has a top speed of only 60 miles per hour, both the city and highway tests
are run at mild climatic conditions (75°F), both tests have mild acceleration rates, and neither test
is run with the use of accessories, such as air conditioning. However, since the time of the last
fuel economy labeling revisions in the mid-1980's, EPA has established several additional test
procedures, used for emissions compliance purposes, which capture a much broader range of
real-world driving conditions. Specifically, these emissions test cycles capture the effects of
higher speeds, more aggressive driving (i.e., higher acceleration rates), the use of air
conditioning at higher ambient temperatures, and colder temperature operation. Our analysis
indicates that these factors can have a significant impact on fuel economy, and that the impacts
can vary widely across different vehicles.
We are now requiring that three additional emission tests, already used by manufacturers,
will be utilized to derive more accurate fuel economy estimates. These three test procedures
encompass a much broader range of real-world driving, as they incorporate the effects of higher
speeds, more rapid accelerations, air conditioning use, and cold temperatures. Our new approach
will utilize these additional emission tests, together with the current two fuel economy tests, so
that our fuel economy test methods reflect a much broader range of driving conditions.
Our final rule revises the test methods by which the city and highway fuel economy
estimates are calculated. We are replacing the current method of adjusting the city (FTP) test
result downward by 10% and the highway (HFET) test result downward by 22%. Instead, we are
finalizing a new approach that incorporates additional test methods that address factors that
impact fuel economy, but are missing from today's tests - specifically, higher speeds, more
aggressive driving (higher acceleration rates), the use of air conditioning, and the effect of cold
temperature. The new test methods will bring into the fuel economy estimates the test results
from the five emissions tests in place today: FTP, HFET, US06, SC03, and Cold FTP.a Thus,
a The US06 test is designed to represent high speed highway driving and aggressive (i.e., rapid accelerations and
decelerations) urban driving. The SC03 test is designed to represent the impact of air conditioner operation at high
temperatures. The Cold FTP, which is conducted at 20°F, is designed to reflect the impact of cold temperatures.
-------
we refer to this as the "5-cycle" method. Under this new method, rather than basing the city mpg
estimate solely on the adjusted FTP test result, and the highway mpg estimate solely on the
adjusted HFET test result, each estimate will be based on a "composite" calculation of all five
tests, weighting each appropriately to arrive at new city and highway mpg estimates. The new
city and highway estimates will each be calculated according to separate city and highway "5-
cycle" formulae that are based on fuel economy results over these five tests. The conditions
represented by each test will be "weighted" according to how much they occur over average real-
world city or highway driving. For example, we have derived weightings to represent driving
cycle effects, trip length, air conditioner compressor-on usage, and operation over various
temperatures. The derivation of this methodology and the relevant weighting factors and
formulae are the principal subject of this Final Technical Support Document.
We also are finalizing an additional downward adjustment to fuel economy estimates
within the 5-cycle method. We put in place a downward adjustment to account for effects that
cannot be replicated on the dynamometer. There are many factors that affect fuel economy that
are not accounted for in any of our existing test cycles. These include road grade, wind, tire
pressure, heavier loads, hills, snow/ice, effects of ethanol in gasoline, and others. We are
finalizing a 9.5% downward adjustment to account for these effects. The detailed technical basis
for this adjustment factor is contained in section III. A. 5 of this Final Technical Support
Document.
Because the 5-cycle method is inherently vehicle-specific, the difference between today's
values and the new fuel economy estimates could vary widely from vehicle to vehicle. Our new
approach will result in city fuel economy estimates that are between 8 to 15 percent lower than
today's labels for the majority of conventional vehicles. The city mpg estimates for the
manufacturers of most vehicles will drop by about 12 percent on average relative to today's
estimates. For vehicles that achieve generally better fuel economy, such as gasoline-electric
hybrid vehicles, the new city estimates will be about 20 to 30 percent lower than today's labels.
The new highway fuel economy estimates will be 5 to 15 percent lower for the majority of
vehicles, including most hybrids. The highway mpg estimates for the manufacturers of most
vehicles will drop on average by about 8 percent, with estimates for most hybrid vehicles
dropping by 10 to 20 percent relative to today's estimates.
In Chapter II of this Final Technical Support Document, we compare current EPA label
fuel economy values, as well as the proposed 5-cycle and mpg-based values, to several
independent estimates of onroad fuel economy. The independent estimates fall into several
general categories, depending on the type of data involved. One type of estimate involves the
measurement or estimate of onroad fuel economy of vehicles in typical operation. Two
examples in this category are the U.S. Department of Energy (DOE) FreedomCar program and
the DOE Your MPG program. A second type of estimate involves onroad measurement of fuel
economy according to some established protocol by an independent organization. Examples in
this category are fuel economy estimates developed by Consumer Report, Edmunds, and AAA.
A third type of estimate involves broad estimates of national fuel consumption and national
VMT and the development of fleet-wide fuel economy estimates. Examples in this category are
fleet-wide fuel economy estimates developed by Federal Highway Administration (FHWA) and
the Energy Information Administration (EIA).
-------
These estimates and studies often suggest that there is a shortfall between the EPA
estimates and real-world fuel economy. For example, Consumer Reports derives city, highway,
and overall fuel economy estimates, and their methods clearly demonstrate the large degree of
variation across vehicles. While their city fuel economy estimates fall on average below the
EPA label values, their highway estimates are, on average, higher than the EPA label values.
Consumer Reports' overall fuel economy estimates range from 27% below to 20% above the
EPA overall rating. The Automobile Association of America (AAA) likewise publishes the fuel
economy results they achieve in their annual auto guide for new cars and trucks. In their 2004
auto guide, about half of their estimates were below the EPA combined city/highway value, and
about one half were above the EPA city/highway combined value. Their estimates ranged from
40% lower than EPA's to 22% higher, again reflecting a great deal of vehicle-to-vehicle
variation.
Each of these studies differs in its test methods, driving cycles, sampling of vehicles, and
methods of measuring fuel economy. There are strengths and weaknesses of each study, which
we discuss further in this Technical Support Document. Collectively, these studies and data
indicate there are many cases where real-world fuel economy falls below the EPA estimates.
The studies also indicate that real-world fuel economy varies significantly depending on the
conditions under which it is evaluated. Nevertheless, taken as a whole, these studies reflect a
wide range of real-world driving conditions, and show that fuel economy can be much lower
than EPA's estimates if more real-world conditions are considered. Where possible, we also
compare the results of these studies with the new label values that would result from the 5-cycle
and mpg-based methods, and we found that in virtually every case the 5-cycle method resulted in
fuel economy values that were significantly closer to these other estimates than the existing
labels.
In Chapter III of this Final Technical Support Document we describe the development of
the vehicle specific 5-cycle and mpg-based methods. We also evaluate the range and variability
of onroad fuel economy experienced by drivers of the same vehicle and we develop an
adjustment factor that accounts for fuel economy impacts not reproducible on the dynamometer
or in the testing laboratory. We describe the final vehicle specific 5-cycle formulae and the final
mpg-based formulae, followed by a discussion of how the current city and highway fuel
economy values would change under the two methods. Finally, we evaluate the sensitivities and
uncertainties in the vehicle specific 5-cycle formulae.
We describe how these different elements of our fuel economy model are developed and
assembled from the test data. We develop methodologies for estimating the following:
• Fuel use related to engine start-up, or start fuel use;
• Fuel use once the engine is warmed up at 75°F (with no air conditioner operation);
• Fuel use due to air conditioner use;
• Fuel use once the engine is warmed up at colder temperatures; and
• Factors that affect onroad fuel economy but which are not addressed by any of the five
dynamometer cycles.
-------
We describe the derivation of the mpg-based approach - a simplified method that will be
an interim option in the first three years of the program and an available option under certain
circumstances in subsequent years. The 5-cycle fuel economy formulae assume that fuel
economy estimates are available for specific vehicles for all five dynamometer cycles and their
respective bags of emission measurements. As discussed in the preamble to the final rule, these
estimates may be based on fuel economy measurements, or on estimates based on test results
from a similar vehicle. A simplified approach to implementing the 5-cycle formulae is to apply
these formulae to test results on recent model vehicles and develop correlations between the 5-
cycle city and highway fuel economy estimates for these vehicles and their fuel economy over
the FTP and HFET, respectively. This simplified approach is referred to as the mpg-based
approach, since the resultant label adjustment will vary depending on the measured fuel economy
(i.e., mpg) of a vehicle over the FTP and HFET tests, and will not require any additional tests.
Following the detailed discussion of the 5-cycle and mpg-based approaches, we present
the actual formulae and an assessment of the impact our new approaches will have on fuel
economy label values. The impact of today's final rule on city and highway fuel economy label
values was assessed using the same database of 615 late model year vehicles used to develop the
mpg-based adjustments. Use of the 5-cycle formulae will reduce both current city and highway
fuel economy label values. For conventional vehicles, city and highway fuel economy values
would be reduced an average of 11% and 8%, respectively. For higher than average fuel
economy vehicles, the reduction in city fuel economy will be slightly higher, while for lower
than average fuel economy vehicles, the reduction in city fuel economy will be slightly lower.
The change in highway fuel economy is essentially independent of current highway fuel
economy.
The impact on hybrid vehicles will be significantly greater for city fuel economy,
averaging a 22% reduction. However, the reduction in highway fuel economy will be similar,
but toward the higher end of the range as for conventional gasoline-fueled vehicles. The impacts
of the 5-cycle formulae on the single diesel vehicle in the database are very similar to those for
conventional gasoline fueled vehicles.
In Chapter IV we detail our estimates of the cost impacts of our new regulation. For
model years 2008 through 2010, manufacturers may use the mpg-based calculation for the five-
cycle fuel economy values or they may conduct voluntary 5-cycle testing. For model years 2011
and after, if the five-cycle city and highway fuel economy values for an emission data vehicle
group are not more than 4% and 5% below the mpg-based regression line, respectively, then all
the vehicle configurations represented by the emission data vehicle (e.g., all vehicles within the
vehicle test group) could continue to use the mpg-based approach. Vehicles within a test group
falling more than 5% below the tolerance band for highway fuel economy values would be
required to conduct US06 tests; those falling more than 4% below the city fuel economy
tolerance band would be required to conduct SC03, US06, and Cold FTP tests. In addition, we
expect that some of these vehicles falling below the tolerance levels may be eligible to estimate
fuel economy for a given test through the application of analytically derived fuel economy
(ADFE) values. Some data is currently available for vehicles that have conducted all five tests;
based on this data, EPA has estimated the number of vehicles for which additional testing would
be required because they fall below the 4 and 5% tolerance bands.
-------
We prepared a range of burden estimates for this analysis, estimating minimum and
maximum cost scenarios. These low and high estimates are intended to be our estimate of the
outer boundaries of the likely testing and information costs. Aggregate annual costs are
estimated to be between $1.4 and $1.7 million. A complete discussion of how these costs were
estimated is in Chapter IV of this Technical Support Document.
Table 1-1
Aggregate Costs
Cost Element
Test Volume
Facilities
Startup
TOTAL
MY 2008 through MY 2010
Minimum
$0
$0
$659,000
$659,000
Maximum
$0
$0
$748,000
$748,000
MY 20 11 and After
Minimum
$343,000
$375,000
$659,000
$1,377,000
Maximum
$424,000
$560,000
$748,000
$1,731,000
-------
Chapter II: Current and Proposed Label Values Compared to
Onroad Estimates
A. Onroad Fuel Economy Estimates During Typical Operation
In the 1984 label adjustment rule, EPA was able to compare fleetwide estimates of a
variety of city and highway fuel economy label options to a number of independent estimates of
onroad fleet fuel economy. In the late 1970's and early 1980's, EPA and several auto
manufacturers had collected onroad fuel economy estimates from tens of thousands of drivers
which could be compared to the EPA city and highway fuel economy labels. EPA primarily
used the driver-based fuel economy estimates to develop the current 10% and 22% adjustments
to fuel economy over the FTP and HFET, respectively.
It is not possible to repeat this type of comparison today, as the auto manufacturers no
longer conduct the extensive monitoring of fuel economy that was performed in the late 1970's
and early 1980's. At the same time, we have discovered three new sources of similar
information. One, Oak Ridge National Laboratory (ORNL) has recently begun a program where
drivers can submit their own fuel economy measurements via the Internet. This program is
referred to as "YourMPG." Two, DOE has also been operating an extensive hybrid
demonstration project for a few years as part of their Freedom Car project. This program
carefully monitors both VMT and fuel consumption, so accurate fuel economy estimates are
available for a number of hybrid vehicles. Three, a private survey firm, Strategic Visions,
performs two surveys of new vehicle purchasers a year to assess consumer satisfaction. The
survey includes questions regarding the fuel economy being achieved to date. We have
purchased the Strategic Visions survey results for model years 2004-06. The results of our
analysis to date are discussed below.
In addition to these three programs, EPA conducted it own testing of vehicle fuel
economy in Kansas City, in conjunction with cooperative efforts between EPA and the
Coordinating Research Council (CRC). The state of California, in conjunction with automobile
manufacturers and others have been obtaining vehicle operational data via chase car studies. All
of these studies are discussed below.
1. ORNL "YourMPG" Program
The ORNL YourMPG data are similar in nature to the much larger databases analyzed
for the 1985 label adjustment rule. Drivers measure their own fuel economy and provide a
perceived split of their driving into city and highway categories. The strength of this type of data
is the fact that the vehicle is being operated by the owner or regular driver in typical use. The
weaknesses are the unknown representativeness of the sample, the unknown nature of the
technique used by the owner/driver to measure fuel economy and the unknown time period over
which fuel economy is generally assessed (e.g., a couple of tanks full or the past year). In the
particular case of the ORNL database, its current size is still small (8180 estimates of fuel
economy for 4092 vehicles) compared to those available in 1985, though it is growing daily.
-------
We compared the fuel economy estimates submitted to the ORNL website with each
vehicle's fuel economy label. We combined the city and highway labels using each driver's
estimate of the percentage which was city and highway. If a driver did not provide an estimate
of the breakdown of their driving pattern, we assumed that their driving was 55% city and 45%
highway with respect to the current label values and 43% city and 57% highway with respect to
the mpg-based label values. We conducted separate comparisons for conventional gasoline
vehicles, conventional vehicles with relatively high fuel economy, hybrids and diesels. The
results are shown below.
Table II.A-1. YourMPG Versus Current EPA Label Fuel Economy
Vehicle
Type
Conventional
Gasoline
High MPG
Conventional
Gasoline*
Hybrid
Gasoline
Diesel
No. of
Estimates
7330
680
520
221
YourMPG
23.8
35.1
43.2
41.8
Current
Label
24.1
35.8
47.1
40.1
Difference
-1.4%
-1.7%
-8.2%
4.3%
MPG-Based
Label
21.7
31.6
40.5
35.3
Difference
9.1%
11.2%
6.3%
18.3%
* Combined EPA Label fuel economy value of 32 mpg or greater, representing about the top 10% fuel
economy conventional vehicles.
As can be seen, diesels appear to perform the best with respect to their label fuel economy,
outperforming the label by 4.3%. Conventional gasoline vehicles come very close to meeting
their label, falling short by only 1.4%. Conventional vehicles with relatively high combined fuel
economy (here assumed to be 32 mpg or more, representing the top 10% of conventional
vehicles in terms of fuel economy) performed only slightly worse, falling short by 1.7%.
Hybrids fall short by a much larger margin, 8.2%. Thus, the greater shortfall seen with hybrids
appears to be more related to hybrid technology than to simply high levels of fuel economy.
With respect to the mpg-based label values, diesels still perform the best of the four types
of vehicles, now exceeding their label values by 18%.b Those conventional vehicles with
relatively high fuel economy fall next, followed by the typical conventional vehicle and hybrids.
Thus, the YourMPG estimates indicate that hybrid performance differs from that of conventional
vehicles, including those with high fuel economy.
There is a larger apparent difference between the mpg-based label values and the current label values in Table
II. A.2 than the 6% average impact of the mpg-based approach on current label values cited elsewhere in this Final
Technical Support Document. This occurs because when working with the YourMPG estimates, we are using the
driver's estimate of city/highway driving breakdown in all cases. When speaking of combined label values,
however, we use a 55/45 breakdown for the current label values and a 43/57 breakdown for the mpg-based values.
The lower city driving weight in the mpg-based formula increases the combined value relative to that for the current
label values and reduces the difference between the two approaches.
-------
The YourMPG database also provides us with an estimate of drivers' perception of the
type of driving which they perform. On average, they estimated that 43.2% of their driving was
city driving and 56.8% was highway. These two figures are essentially identical to the weights
developed from the Draft MOVES2004 depiction of onroad driving and embedded in the 5-cycle
combined fuel economy formula. This is encouraging. However, this change from the
traditional 55/45 city/highway split of onroad driving also causes some unusual relationships
between onroad and label fuel economy as depicted in Table II. A-1 above.
Focusing on conventional vehicles, which comprise the great majority of the database,
Table II.A-1 shows that people's onroad fuel economy is only 1.4% lower than the current label
values indicates and 9.1% higher than the mpg-based label values would have been. If a 55/45
split would have been used with the current label values, onroad fuel economy would have been
1-2% higher than the label values for these vehicles. This differs dramatically from the onroad
fuel economy shortfall indicated by the FHWA estimates of onroad fuel economy discussed in
Section II.C below. In other words, the YourMPG database indicates that onroad fuel economy
is exceeding the standard 55/45 label fuel economy, while FHWA indicates the opposite.
Examining the YourMPG database further, we found that the average 55/45 label value in the
database was 25.4 mpg. Per MOBILE6.2, the average 55/45 label value of the onroad vehicle
fleet is 21.2 mpg. Thus, people submitting their onroad fuel economy estimates to the YourMPG
database drive more fuel efficient vehicles than the average vehicle. This could indicate that
those participating in the program have a greater interest in reducing fuel consumption than the
average driver, but this cannot be known for certain. However, if true, this would explain the
difference seen between the YourMPG database and the FHWA fleetwide estimates.
2. DOE FreedomCar Program
The Department of Energy has overseen the real world operation of a number of electric
hybrid vehicles for a period of years. The Advanced Vehicle Testing Activity (AVTA),
conducted jointly by the Idaho National Laboratory (INL) and the National Renewable Energy
Laboratory (NREL), has been benchmarking hybrid electric vehicle performance as part of the
FreedomCAR & Vehicle Technologies Program. The strength of the FreedomCAR program
testing of hybrid vehicles lies in the fact that the vehicles are operated on the road over long term
periods similar to what consumer-purchased vehicles experience, albeit often in commercial
applications. Over a million miles of operation have been assessed and careful fuel consumption
and mileage records are kept. The weaknesses are that some of the vehicles are in commercial
use (e.g., company pool vehicles) for accelerated mileage accumulation and that the vehicles are
operated exclusively in the Southwest, mainly in Phoenix, Arizona and surrounding areas.
Nevertheless, the vehicles are operated just as any other vehicle would be in that application and
the vehicles are subject to all of the environmental and roadway factors which affect the fuel
economy of typical vehicles, such as winds, rough roads, hills, traffic congestion, etc. Because
of the limited geographic area of the program, the vehicles are more likely to experience hot
temperatures and air conditioning use than cold temperatures.
The vehicles' operators report mileage and fuel usage to FreedomCAR which posts the
monthly and cumulative fuel economy of each electric hybrid fleet on a monthly schedule.
-------
Therefore, seasonal changes in fuel economy can be observed. The results of the fleets are
shown in Table II. A-2.
Table II.A-2. FreedomCAR Hybrid Fleet Cumulative Versus EPA Composite Label Fuel
Economy
Vehicle
2001 Honda Insight
2002 Toyota Prius
2003 Honda Civic
2004 Toyota Prius
2004 Chevrolet
Silverado 2wd
2004 Chevrolet
Silverado 4wd
2005 Ford Escape
2wd
2005 Ford Escape
4wd
2005 Honda Accord
2005 Lexus RX400h
2006 Toyota
Highlander
Average
Accumulated
Mileage
417,000
458,000
378,000
186,000
48,000
53,000
70,000
78,000
158,000
67,000
69,000
180,000
Fleet
Size
6
6
4
2
1
1
1
1
2
2
2
2.5
Fuel Economy (mpg)
Onroad
45.2
41.0
37.6
44.9
17.7
17.9
28.6
27.0
27.8
24.8
24.7
30.7
EPA Composite Label *
Current
61.0
48.6
46.3
54.6
18.8
16.9
33.6
29.9
32.3
28.1
28.1
37.0
5-Cycle
50.4
37.9
44.1
—
15.1
—
23.6
25.9
24.0
24.0
31.6
MPG-
Based
51.5
39.7
45.5
—
15.2
—
26.1
28.3
24.4
24.4
32.9
Difference (%)
Current
35%
19%
23%
22%
6%
-6%
17%
11%
16%
13%
14%
16%
5-Cycle
12%
1%
-2%
-16%
-13%
-7%
-3%
-3%
-3%
MPG-
Based
14%
6%
1%
-15%
-3%
2%
-2%
-1%
1%
* Current combined is a 55/45 weighting of city/highway fuel economy. Combined 5-cycle and mpg-based fuel
economy is a 43/57 weighting of city/highway fuel economy. All label values from EPA certification database.
Current combined label fuel economy values shown will not match official label values due to differences in vehicle
configurations. The FreedomCAR fleet information as reported thru August 2006.
As can be seen, EPA's current label formulae over-estimate the onroad fuel economy achieved
by all but one of the hybrid vehicle fleets. It should be noted that the values for current
combined fuel economy are those from EPA's certification database and are not the official label
values. The official label values are even higher due to differences between the worse case
vehicles tested over the Supplemental FTP cycles and the average vehicle sold. The largest
shortfall was 35% for the Honda Insights. The Chevrolet Silverado was the only model which
exceeded the current label value of the test vehicle in our certification database. This is likely
related to the fact that its hybrid design includes limited fuel economy targeted features. Except
for the Chevrolet Silverado, the onroad fuel economy for each fleet never exceeded either the
city or highway fuel economy label. This indicates that regardless of whether the vehicles were
driven predominantly in city or highway driving modes, other real world factors reduced onroad
fuel economy beyond that captured in the FTP and HFET and the current 10% and 22%
adjustment factors.
10
-------
Table II.A-2 also presents combined fuel economy values using the final 5-cycle and
mpg-based formulae for those vehicles for which we have 5-cycle fuel economy data. The final
combined 5-cycle label values exceed onroad fuel economy for two out of eight models, while
the final mpg-based values do so for four out of eight models. The average of the differences is
very small in both cases. On average, the combined 5-cycle value is 3% lower than those
measured onroad. However, as mentioned above, the specific vehicles in our 5-cycle database
tend to be worse case. For example, the current official label values exceed those shown in
Table II.A-2 by 3%. If we increased the combined 5-cycle values commensurately, they would
match the onroad values on average. Thus, while both of the final approaches do a much more
reasonable job at predicting the onroad fuel economy achieved in the DOE FreedomCar program
than the current label formulae, the final 5-cycle formulae appear to be particularly accurate
when compared to the FreedomCar experience.
The close match between the final 5-cycle formulae and the FreedomCar experience is
somewhat fortuitous, as the climate where the vehicles were primarily driven is not typical of
most of the U.S. The FreedomCar program focuses on the southwest U.S. There, air
conditioning use is much higher than average, while cold temperature operation is much lower
than for the U.S. on average. While both factors reduce fuel economy, they do so to different
extents. Colder temperatures have a much larger impact on national average, 5-cycle fuel
economy than air conditioning. In projecting 5-cycle fuel economy values for individual
vehicles, we have had to estimate the impact of heater-defroster operation on fuel economy
during the cold FTP. For conventional vehicles, the effect is likely very small (i.e., less than
2%). However, for a couple of hybrids tested, the effect was much larger. The impact of heater-
defroster operation is likely to vary significantly across individual hybrids, but without data, we
cannot anticipate this variability. Overall, basing this impact on the two hybrids tested reduced
the combined 5-cycle fuel economy of the hybrids in our certification database by 3%. Clearly
this change is not relevant in areas like Phoenix. This simply indicates the limitations involved
in very direct comparisons of vehicle test programs and 5-cycle fuel economy estimates.
Overall, the fact that both the final mpg-based equations and 5-cycle formulae yield fuel
economy label values quite close to the FreedomCar findings is very encouraging and the best
that one could hope for without fine-tuning the 5-cycle formulae to exactly match the driving
activity and conditions of the FreedomCar vehicles.
When analyzing monthly reported fuel economy, large seasonal fluctuations in fuel
economy were observed on most of the hybrid fleets. The seasonal fluctuations are especially
noticeable on the fleets that had been in service for over one year. The fuel economy during the
hot and often humid summer weather months when heavy air conditioning usage could be
expected was as much as 15 mpg lower than observed fuel economy during mild Phoenix area
winter months. Fuel economy over the SC03 air conditioning test for the three hybrids with the
highest rated fuel economy shown in Table II.A-2 (Prius, Insight and Civic) tends to be 15-20
mpg lower than that over the FTP. No cold weather operation similar to northern states or the
Cold FTP (20°F) was reported which would likely have resulted in further shortfalls.
11
-------
3. Strategic Visions New Vehicle Survey
Strategic Visions surveys roughly 100,000 purchasers of new cars and light trucks each
year. The survey recipients are selected randomly from among the purchasers of each vehicle
model and therefore represent all regions of the country. Some models are more heavily
surveyed than others, particularly models which have just been introduced or significantly
redesigned. Therefore, the results should be assessed on a model by model basis and not
averaged across models before averaged within each model.
The survey asks recipients to write down the fuel economy which they are currently
achieving. About half of the recipients respond to the request for fuel economy information.
Thus, about 50,000 estimates of onroad fuel economy are received each year. The strengths of
this survey are the large number of estimates and the fact that the survey recipients are randomly
selected. The weaknesses are the unknown source of each consumer's fuel economy estimate
and the survey response rate for this question (i.e., only 50%). Still the fact that a wide range of
models are surveyed with each model having a number of independent estimates allows very
direct comparison to the current, 5-cycle and mpg-based label values on a model by model basis.
EPA purchased the Strategic Visions survey results for the 2004-06 model years. In
preparing the data for analysis, we noticed a peculiarity. The frequency of consumers' estimated
fuel economy for fuel economy values being a multiple of five were much higher than those with
other values. Figure II-l shows the distribution of estimated city fuel economy for 2006 model
year vehicles.
Figure II-l. Distribution of Onroad Fuel Economy Estimates - Strategic Vision, 2006
"1 O°/
1 Z 70
mo/.
I U 70
80/1
70
60/1
70
40/1
70
20/1
70
no/
— "1
/
-^ys/
\
i
Arrows indicate
multiples of 5
1
/ IA fi~~
i
— Vs*
U 70 ' ' '
0 10
V
v A r
*vv
\ \
20 30
City FE (MPG)
. — •
*^^S
\
40
50
As can be seen from Figure II-l, each of the frequencies of fuel economy values ending
in 5 or 0 is much higher than those of nearby values. By comparing the difference between the
12
-------
frequencies of fuel economy values of a multiple of five with those just above and below that
value, we estimate that there is an excess response of 13% for fuel economy values of a multiple
of five compared to other values. This implies that 13% of the respondents were only estimating
their onroad fuel economy to within ±5 mpg. We found the same effect with highway fuel
economy estimates. Thus, it is very likely that the same respondents rounded both their city and
highway fuel economy estimates to a factor of five.
It is unknown whether this tendency biases the estimated onroad fuel economy upward or
downward. However, the presence of such rough estimates significantly reduces the value of
this database to distinguish between two sets of fuel economy estimates. This is particularly true
for comparisons between the mpg-based and 5-cycle formulae, which often only differ by 1-2
mpg. We are working with Strategic Visions to better identify the method used by respondents
to estimate their onroad fuel economy so that we can focus on those who actually kept mileage
and fuel usage records. This should also avoid those respondents who estimate, versus measure,
their actual fuel economy. We plan to analyze this improved data over the next couple of years.
4. Kansas City Instrumented Vehicle Study
During 2004-2005, EPA in association with the Coordinating Research Council, DOT
and DOE, recruited and tested over 600 privately owned passenger vehicles in the Kansas City
area. The vehicles included an assortment of compact cars, mid-size cars, pick-ups and SUVs
from a variety of manufacturers. The program was split into 3 rounds (1, 1.5 and 2), each
consisting of 120-300 vehicles. In all three rounds, vehicles were recruited randomly from lists
of vehicle registrations in the Kansas City area. Care was taken to ensure that the sample were
random with respect to the geographic location of the owner and socio-economic status. In
rounds 1 and 2, the desired sample of vehicles was stratified into four groups of model years,
with emphasis on older vehicles1'2. The primary purpose of Rounds 1 and 2 was the
quantification of particulate emissions, particularly those from high emitters. In Round 1.5, only
2001 and later model year vehicles were sampled. (Details about the design and performance of
Round 1.5 are described the study's final report.3) The primary purpose of Round 1.5 was the
measurement of onroad fuel economy from vehicles for which we could estimate 5-cycle fuel
economy. This meant that we had to have fuel economy estimates over all five cycles for these
vehicles (i.e., that the vehicle had to be certified to the Supplemental FTP standards). These
standards began phasing in with the 2001 model year.
Only a few of the vehicles tested in Rounds 1 and 2 were instrumented with a Portable
Emissions Measurement System (PEMS) and tested in the hands of their owners. As these
vehicles ranged in model year from 1968 to 2005, very few of the vehicles tested in Rounds 1
and 2 had been certified to the Supplemental FTP standards. However, all of the vehicles tested
in Round 1.5 were instrumented with PEMS and had their fuel economy measured while being
driven in normal use by their owners. The round 1.5 vehicle fleet consisted of approximately
120 vehicles, including over 30 hybrid electric vehicles. The PEMS measures driving activity, as
well as second-by-second mass emissions of CO2, CO, HC, and NOx for roughly 24 hours while
the owners of the vehicles are utilizing their vehicles on the road under normal, real-world
conditions.
13
-------
Total fuel consumption for each vehicle was determined from the carbon balance of the
CC>2, HC, and CO emissions. (Details of the test program and the methods used to process the
data obtained are described in Appendix A of this Final Technical Support Document.) Each
vehicle's fuel was sampled and tested for density and weight percent carbon. The total distance
of driving was determined by summing vehicle speed and multiplying by total time of operation.
This total distance traveled was then divided by total fuel consumption to determine onroad fuel
economy.
EPA city and highway label fuel economy values were obtained from EPA mileage
guides. The test vehicles were matched to those tested in Kansas City to the closest degree
possible. Figure 11-2 compares the measured fuel economy to the 55/45 composite label fuel
economy from Round 1.5 (newer vehicles). We segregated the vehicles into two groups:
conventional gasoline-fueled vehicles and hybrids. A linear regression with no constant of the
conventional vehicles showed nearly one-to-one correlation, with a slope of 1.006. The
correlation was also quite good (r-squared value of 0.77). The largest difference was only 6
mpg, or about 30%. Thus, the onroad fuel economy data indicate no offset from the current EPA
label values on average.
Figure II-2. Comparison Onroad to Current Label Economy: Kansas City
The correlation of hybrid data shows much more scatter. This is partially explained by
the fact that only three hybrid models were tested, a number of Toyota Prius and Honda Civic
vehicles and one Honda Insight. The range of fuel economy label values for these three vehicles
is very small, 48-56 mpg, plus one vehicle at 64 mpg. With the high degree of variability in
measured onroad fuel economy, it is not surprising that the correlation coefficient was small.
On average, hybrid fuel economy was 11% less than the composite EPA label values.
The average onroad fuel economy of the Toyota Prius vehicles was closer to their composite
label values than those for the two Honda models. On average, the onroad fuel economy of the
14
-------
hybrids tested varied more than the conventional vehicles. This could be due to hybrids' greater
sensitivity to operating conditions which can either take full advantage of the hybrid technology
or essentially nullify it. The fact that many vehicles started out testing with a hot start likely
biased onroad fuel economy upwards to some degree. Thus, the actual shortfalls found would
have been greater to some degree if testing had begun with a cold start.
We also performed a regression of onroad fuel consumption per mile versus the inverse
of the current fuel economy value for hybrid vehicles, as Honda suggested in their comments
(see Section 5.5 of the Response to Comments document). First, we found that the intercept was
not statistically significant (p-value of 0.684). Thus, we performed a new regression with an
intercept of zero. We found an r-squared value of 0.18, which is not much different than that for
the regression of fuel economy. The slope of the regression was 1.135, indicating that the
hybrids consumed 13.5% more fuel than predicted by the inverse of their label values. More
importantly, this slope had a p-value of 10"31, indicating that it was extremely unlikely to be zero.
The 95% confidence interval for the slope ranged from 1.09 to 1.18. Thus, on average, the data
collected in Kansas City indicate that the hybrid vehicles tested did not perform as well as the
conventional vehicles compared to their current fuel economy label values.
B. Fuel Economy Estimates by Independent Organizations
Several consumer organizations perform their own fuel economy assessments. Of these,
the American Automobile Association (AAA) and Consumer Reports (CR) have tested the
greatest number of vehicles. The relative strengths of this testing include the fact that the
vehicles are tested on actual roads, usually in traffic and under real environmental conditions.
The primary weaknesses of this testing include:
1) The fact that the drivers or driving patterns involved are not typically published, so
they may or may not be representative of average U.S. drivers or driving,
2) Vehicles are tested throughout the year, so some vehicles are tested in hot weather and
other in cold weather and some under moderate conditions, and
3) In some cases, the actual test procedures used to measure the volume of fuel consumed
during the test are not described, leaving some doubt as to their accuracy. Still, because of the
public interest in these estimates, we believed that they should be considered here. We will
begin with an analysis of the Consumer Reports estimates, followed by those of Edmund's and
AAA.
1. Consumer Reports Estimates of Onroad Fuel Economy
Consumer Reports published their fuel economy estimates for 303 2000-2006 model year
vehicles. They publish both EPA's current city, highway and combined fuel economy estimates,
as well as their own city, highway and combined fuel economy estimates. Therefore, we can
compare EPA's current label values to those of CR for all 303 vehicles. As the mpg-based
formulae only require knowledge of fuel economy over the FTP and FIFET, we can apply these
formulae to the EPA city and highway fuel economy values presented by CR (after removing the
current label adjustments of 10% and 22%) and calculate mpg-based fuel economy values for all
303 vehicles. We were also able to match 70 of these vehicles with those in our 5-cycle fuel
15
-------
economy database.0 Thus, for these 70 vehicles, we were able to calculate 5-cycle fuel economy
values.
We made two sets of comparisons. One set included all 303 vehicles. The other set
included only 70 vehicles. The results of the first comparison are shown in Table II.B-1.
Table II.B-1.
Consumer Reports and Current EPA and MPG-Based Fuel Economy:
303 Vehicles
City
Highway
Combined
Consumer Reports
MPG
14.2
29.3
20.7
Current EPA Label
MPG
20.4
26.9
22.9
Difference *
-30%
9%
-9%
MPG-Based
MPG
18.0
24.7
21.2
Difference
-21%
19%
-3%
* Consumer Reports fuel economy compared to EPA label value.
As can be seen, the CR city fuel economy values are well below both the current label or mpg-
based label values (21% to 30%). The reverse is true for highway fuel economy. The CR
estimate of combined fuel economy is 9% lower on average than the 55/45 composite of the
current EPA city and highway label values. However, the CR estimate of combined fuel
economy is only 3% lower on average than the 43/57 composite of the mpg-based city and
highway fuel economy values. Thus, there is a much better match up between the composite
mpg-based fuel economy and the CR combined fuel economy than with current label values.
Table II.B-2 presents the same comparisons, except that it includes the 5-cycle estimates
and only includes the 70 matched vehicles.
Table II.B-2. CR and Current EPA, 5-Cycle and MPG-Based Fuel Economy: 70
Vehicles
City
Highway
Combined
Consumer
Reports
MPG
14.3
29.3
20.6
Current EPA Label
MPG
20.4
26.4
22.7
Difference*
-30%
11%
-9%
5-Cycle
MPG
18.0
24.3
21.0
Difference
-21%
21%
-2%
MPG-Based
MPG
17.8
24.1
20.9
Difference
-20%
22%
-2%
* Consumer Reports fuel economy compared to EPA label value.
As can be seen, the comparisons between the CR, current EPA and mpg-based fuel economies
are very similar to those in Table II.B-1. On average across 70 vehicles, the CR combined fuel
economy estimates differ from the current, mpg-based and 5-cycle combined fuel economy
In the Draft Technical Support Document, we identified 151 vehicles which were both tested by Consumer
Reports and in our certification database. However, many of these matching vehicles were not from the same model
year.
16
-------
values 9%, 2%, and 2%. The standard deviations of the percentage differs for individual models
provides an indication of the consistency in the offset, or in the ability of the various label
approaches to predict relative fuel economy differences between models. Across 70 vehicles, the
standard deviation of the percentage differences between the CR combined fuel economy
estimates and the current, mpg-based and 5-cycle combined fuel economy values are 6.5%,
6.6%, and 6.3%. These standard deviations are very similar. The 5-cycle label values provide
slightly better estimates of relative vehicle fuel economy than the other two label approaches. As
mentioned above, the CR test procedures do not include cold starts or air conditioning operation.
As these are two important features of the 5-cycle formulae, much of the potential improvement
associated with the 5-cycle approach is not reflected in CR's fuel economy estimates.
Of particular interest here are the fuel economy values for hybrid vehicles. For hybrids,
the 5-cycle and mpg-based formulae often give different results. The 303 vehicles tested by
Consumer Reports include six hybrid vehicles. We have 5-cycle fuel economy estimates for four
of these vehicles, all except the 2001 Prius and 2000 Insight. A comparison of the various fuel
economy estimates for the five hybrids values are shown in the Table II.B-3. To make the
comparison between the three label approaches as equitable as possible, we show based the
current label values on the FTP and FIFET fuel economy values in the 5-cycle fuel economy
database and not those shown in the CR report.
17
-------
Table II.B-3. Comparison of Consumer Reports and EPA Fuel Economy Values for
Hybrids
Consumer
Reports
MPG
Current EPA Label
MPG
Difference
5-Cycle Fuel Economy
MPG
Difference
MPG-Based Fuel
Economy
MPG
Difference
City Fuel Economy
Escape
Accord
Civic
2005 Prius
Average
22
18
26
35
25
32
30
47
57
41
-31%
-39%
-44%
-38%
-38%
27
26
38
45
34
-20%
-29%
-32%
-23%
-26%
23
22
34
45
31
-4%
-18%
-24%
-22%
-17%
Highway Fuel Economy
Escape
Accord
Civic
2005 Prius
Average
29
37
45
50
40
28
36
46
50
40
5%
2%
-2%
0%
1%
25
33
41
45
36
14%
12%
9%
12%
12%
24
30
39
45
35
20%
25%
15%
10%
17%
Combined Fuel Economy
Escape
Accord
Civic
2005 Prius
Average
26
25
36
44
33
30
32
46
54
41
-13%
-23%
-22%
-18%
-19%
26
29
40
45
33
-1%
-15%
-10%
-2%
-7%
24
26
37
45
0
10%
-3%
-2%
-2%
1%
A lot of information is presented in Table II.B-3. We will focus first on the results for all
five hybrids averaged together, indicated in bold in the table, starting with the city, then the
highway, then the combined fuel economy estimates.
Starting with the city values, the CR city fuel economy estimates average 38% less than
the current EPA city label values. This is greater than for the average vehicle, where the
difference was 30%. The differences are smaller for the mpg-based and 5-cycle city values (26%
and 17%, respectively). While the 26% difference for the mpg-based approach is greater than
that for the average vehicle (21%), the 17% difference for the 5-cycle approach is less than that
for the average vehicle (20%). This indicates that the 5-cycle formula for city driving is likely
reflecting factors which are included in CR's city test protocol and which are not included in the
FTP, nor a constant 10% adjustment factor. In contrast, the current and mpg-based label
approaches do not. While the mpg-based formula for city driving produces fuel economy
estimates more closely resembling those of CR than the current label values, the mpg-based city
18
-------
formula does not pick up factors which are apparently unique to hybrids which are included in
CR's test procedure.
With respect to highway fuel economy, as described earlier for conventional vehicles, the
CR highway fuel economy estimates tend to be much higher than all the label approaches (as
opposed to the CR city fuel economy estimates, which tend to be much lower than all the label
approaches). However, except for this fundamental shift, the relative performance of the three
label approaches with respect to CR highway fuel economy estimates is the same for city and
highway fuel economy. The current and mpg-based approaches predict greater relative fuel
economy benefits for hybrids that the CR highway testing is not finding. In contrast, the benefits
of hybrid technology on highway fuel economy as indicated by the 5-cycle formulae are
reflected in the CR testing.
Again, the story is similar for the combined label values. The current and mpg-based
label approaches average 27% higher than CR's combined values for the five hybrids. This is
more than twice the difference for the average vehicle, where the difference was only 11%.
Again, this difference in combined values indicates that the current EPA label formulae are
granting some relative benefits to hybrid vehicles which are not reflected in CR's combined test
protocol. The 5-cycle combined values average 5% higher than the CR combined values, which
is only 2% more than the 3% difference found for 70 vehicles. This indicates that the 5-cycle
formula on a combined driving basis is only granting a very slight relative benefit to hybrid
vehicles compared to CR's combined test protocol. In contrast, the mpg-based equations appear
to be granting hybrids a greater relative benefit compared to the CR test protocols. The mpg-
based combined fuel economy averages 8% higher than the CR values for the five hybrids, while
only 2% for the 70 and 303 vehicle fleets.
Fourth, moving to the comparison of combined fuel economy for individual hybrid
vehicles, the differences between the three "EPA" estimates and the CR estimates tend to be
consistent in percentage terms, with the exception of the Escape and to some extent, the Accord.
The differences for these two hybrids are not surprising. The other hybrids have very high fuel
economy values compared to conventional vehicles. Thus, they essentially "set" the mpg-based
equations for their range of fuel economy. The fuel economies of the Escape and the Accord fall
within the range of conventional vehicle fuel economy. Here, the mpg-based equations are "set"
by the more numerous conventional vehicle data.
Overall, fuel economy estimates based on the 5-cycle and mpg-based formulae both
match the CR test results more closely than the current label values for conventional vehicles,
which dominate the 70 and 303 vehicle samples. In addition, hybrid fuel economy estimates
based on the 5-cycle formulae more closely match those of CR compared to either the current
label or mpg-based formulae. The CR estimates do not necessarily match those of the average
driver. Their driving cycles, in particular, are only generally described. Still, they represent an
ostensibly consistent set of estimates. The CR test procedures find a lower benefit to hybrid
technology than the FTP and HFET indicate. The 5-cycle formulae perform similarly, though to
a slightly less extent (i.e., 2%).
19
-------
Of additional interest is whether the performance of hybrids is due to their relatively high
fuel economy or to their hybrid technology. To shed some light on this question, we compared
the mpg-based label values to the Consumer Reports fuel economy values for all conventional
vehicles, the top ten percentile (in terms of combined current label values) of all conventional
vehicles and hybrids. As described above in Table II.B-1, the mpg-based combined label values
averaged 2% higher than the combined fuel economy measured by Consumer Reports. The
mpg-based combined label values for the top 10 percentile of conventional vehicles matched the
combined fuel economy measured by Consumer Reports, those performing better than
conventional vehicles as a whole, at least in terms of Consumer Reports fuel economy
measurements. However, as shown in Table II.B-3 above, the mpg-based combined label values
for hybrids averaged 8% higher than the combined fuel economy measured by Consumer
Reports. Thus, hybrids performed worse in the Consumer Reports testing than the average
conventional vehicles and worse to a slighter greater degree than those conventional vehicles
with the highest fuel economy. Thus, based on the Consumer Reports testing, the differential
performance of hybrids appears to be more related to their hybrid technology than to their
relatively high fuel economy. The same relationships hold for the 5-cycle fuel economy values.
2. AAA Estimates of Onroad Fuel Economy
The American Automobile Association (AAA) also develops its own fuel economy
estimates. In their 2004 report, AAA presented their test results and the EPA label values for
163 models. AAA only presents a single fuel economy estimate, which we understand to be a
composite of city and highway operation.
Overall, AAA found a higher overall fuel economy than the current combined EPA label
value for 85 models, and lower fuel economy for 73 models. On average, the AAA fuel
economy estimates were 1.5% lower than the current combined EPA label values. We
calculated an mpg-based 43/57 combined fuel economy using the mpg-based equations. (FTP
and HFET fuel economy values were back calculated from the current EPA fuel economy label
values.) On average, the AAA fuel economy estimates were 5.3% higher than the mpg-based
combined fuel economy values. Table II.B-4 shows these comparisons.
Table II.B-4. AAA and Current EPA, 5-Cycle and MPG-Based Fuel Economy Estimates
163 vehicles
61 vehicles
AAA
MPG
21.7
23.2
Current
EPA
Label
MPG
22.1
23.4
Difference*
-1.5%
-0.4%
5-
Cycle
MPG
N/A
21.7
Difference
N/A
6.7%
MPG-
Based
MPG
20.6
21.9
Difference
5.3%
6.1%
We were able to match 61 out of the 163 AAA-tested vehicles with similar vehicles in
our 5-cycle certification database. This is lower than the 98 models which we matched in the
analysis described in the Draft Technical Support Document. The lower figure is due to the use
of a more stringent criterion that the vehicles match in terms of model year. (We assumed that
all of the vehicles tested by AAA were 2004 model year vehicles.) As was the case with the
Consumer Reports data, the FTP and HFET fuel economies for the certification vehicles in our
20
-------
5-cycle fuel economy database were generally lower than those tested by AAA. Thus, we
adjusted the 5-cycle city and highway fuel economy values using the ratio of the FTP and HFET
fuel economy values, respectively, from our certification database and the AAA database. On
average, the AAA estimates were 6.1% higher than the combined, 5-cycle fuel economy values
for these 61 vehicles. On average, the AAA estimates were 6.7% higher than the combined
mpg-based fuel economy values for these 61 vehicles, just slightly greater than with the 5-cycle
formulae. While not shown in the table, the standard deviation of the percentage differences is
9% for all three EPA label approaches. Thus, none of the three EPA label approaches stands out
with respect to their ability to predict relative onroad fuel economy as measured by AAA.
The AAA fuel economy values include two hybrids, a Prius and an Insight. Based on the
official EPA label values, the AAA estimates were 6.6% lower than the current EPA composite
fuel economy values for these two vehicles, 5% lower than for the average vehicle. Thus, the
current label adjustments are indicating greater fuel economy improvement due to hybrid
technology than AAA is finding during their testing.
In contrast, the AAA estimates average 10.6% and 12.0% higher than the mpg-based and
5-cycle combined fuel economy values for these two vehicles, respectively. These differences
are about 6% and 4% greater difference than for the average vehicle, respectively. Thus, the 5-
cycle and mpg-based formulae are indicating less fuel economy improvement due to hybrid
technology than AAA is finding during their testing. Thus, the AAA testing is finding hybrid-
related benefits somewhere in between those indicated by the current label formulae and those
indicated by the mpg-based and 5-cycle formulae. This is in contrast to the Consumer Reports
estimates, where their testing indicated hybrid benefits more consistent with those indicated by
the 5-cycle formulae.
The consistency between the mpg-based equations and 5-cycle formulae with respect to
the AAA hybrid testing is due to the fact that AAA only tested the two hybrids with the highest
fuel economy label values. In this range of fuel economy, both the 5-cycle formulae and the
mpg-based equations predict very similar values for hybrids. Consumer Reports, on the other
hand, tested four hybrids, two of which have much lower fuel economy values. In this range, the
mpg-based equations are dominated by conventional vehicles and the 5-cycle values for hybrids
tend to fall below the mpg-based lines.
The main reason for the difference, however, is that AAA found much higher fuel
economy values for the two hybrids which both organizations tested than Consumer Reports.
Consumer Reports found that a 2000 Insight and 2004 Prius achieved combined fuel economy
values of 51 mpg and 41 mpg, respectively. AAA found that a 2004 Insight and 2004 Prius
achieved combined fuel economy values of 58 mpg and 52 mpg, respectively. In terms of fuel
consumption, AAA found 17% less fuel use for these two hybrids than that found by Consumer
Reports. We developed an analogous estimate for conventional vehicles by comparing the
difference in fuel economy found by each organization for all the vehicles tested relative to the
current and mpg-based label formulae. For conventional vehicles, in terms of fuel consumption,
AAA found 8% less fuel use for these two hybrids than that found by Consumer Reports (.
Thus, the different test procedures used by the two organizations are finding much different
benefits of hybrid technology relative to conventional vehicles. This reinforces the observation
21
-------
that the capability of hybrid technology is sensitive to how and where the vehicles are operated
(e.g., colder temperatures, trip length, driving pattern, etc.). The comparison of the fuel economy
values predicted by the various label approaches to the onroad fuel economy estimated by
Consumer Reports, AAA and other organizations is investigated further in Section II.D. below.
As part of their comments on the rule, AAA provided onroad and dynamometer fuel
economy estimates for 42 additional vehicles. The average fuel economy recorded by the
owners was 25.6 mpg, while the average of the current combined EPA label values was 29.6
mpg, Thus, this additional data indicates a 15% shortfall in onroad fuel economy relative to the
current EPA label values. In addition, AAA tested 17 of these vehicles over the FTP, HFET, a
hot start US06 and a US06 test with a cold start. On average, current EPA label values based the
AAA FTP and FIFET testing was very consistent with those implied by the vehicle labels. The
average of the vehicles' combined city/highway label values (26.9 mpg) and those based on the
AAA testing (26.8 mpg) differed by only 0.1 mpg. The average onroad fuel economy of these
vehicles as estimated by the owners was 23.5 mpg. This indicated an onroad fuel economy
shortfall of 12%, or slightly smaller than that for the complete set of 41 vehicles. The average
cold start and hot start US06 fuel economy values for these vehicles were 23.1 and 25.0 mpg,
respectively, bracketing the owners' experience. In contrast, the owner's fuel economy was
lower than either the current city or highway label values for these 17 vehicles. AAA
commented that the US06 test appeared to be a better predictor of onroad fuel economy than
either the current city or highway label values and encouraged EPA to move forward with its
proposed 5-cycle formulae, which included fuel economy measured over the US06 test.
3. Edmunds
The on-line car journal Edmunds.com measures fuel economy on new cars they evaluate
for reviews. Edmunds reviews and road tests cars in a variety of ways, but the most relevant data
come from their "long-term tests." For these tests they purchase or lease vehicles directly from
dealers and keep the vehicle for 1-2 years, generally accumulating as much as 30,000 miles of
experience with the vehicles from several different reviewers. Edmunds reports the best, worst,
and average fuel economy achieved during their long-term use of the vehicle. We reviewed
Edmunds data from 40 model year 2003-2006 long-term test vehicles and compared their
average to the EPA "combined" city/highway fuel economy value. On average, the Edmunds
reviewers achieved fuel economy about 14% lower than the current EPA combined label value.
Hybrid vehicles performed even more poorly; the four included in the recent Edmunds long-term
tests on average fell 24% below the current EPA combined label value. The data from Edmunds
and current EPA City, Highway, and Combined label values are shown in Table II.B-5 below.
As can be seen in this table, the average fuel economy achieved by Edmunds reviewers is
frequently lower than the current EPA City estimate - in fact this occurs in more than half of the
vehicles. And only in a minority of cases (8) does the best achieved by Edmunds exceed the
EPA highway estimate, supporting the belief of many that the current highway fuel economy
label value is a near best-case estimate.
22
-------
Table II.B-5. 2003-2006 Edmunds Long-term Test Vehicles
MY
2006
2006
2006
2006
2005
2005
2005
2005
2005
2005
2005
2005
2005
2005
2005
2005
2005
2005
2005
2005
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2003
2003
2003
2003
2003
2003
2003
2003
MFR
Honda
Lexus
Mitsubishi
VW
Audi
BMW
Chevrolet
Dodge
Ford
Ford
Ford
Honda
Honda
Kia
Landrover
Nissan
Scion
Subaru
Toyota
Volvo
Acura
Chevrolet
Chrysler
Ford
CMC
Mazda
Mitsubishi
Nissan
Nissan
Toyota
Toyota
Volvo
Honda
Honda
Infiniti
Lexus
Mazda
Mitsubishi
Nissan
Subaru
Model
Ridgeline
RX 400h
Eclipse GT
Jetta
A4 2.0T
X3
Cobalt
Magnum
Escape Hybrid
GT
Mustang GT
Accord Hybrid
Odyssey
Spectra
LR3
Frontier 4x4 Nismo
tc
Legacy GT
Solara
S40
TL
Malibu
Pacifica
F-150
Canyon
RX-8
Endeavor
Quest
Titan
Prius
Sienna
XC90
Accord
Pilot EX
G35 Coupe
SC430
Mazda6
Outlander
350Z
Forester
Edmunds.com
Best
16.1
27.9
19.2
21.0
27.7
21.6
27.3
24.4
29.9
17.5
17.8
35.0
22.7
28.9
27.6
18.2
31.6
25.2
21.4
25.3
25.2
30.8
20.7
17.7
21.1
22.5
25.0
22.9
15.7
45.2
19.1
19.6
30.0
25.6
23.2
18.3
26.4
26.8
25.7
27.7
Worst
14.3
22.6
14.6
15.0
16.2
16.9
20.2
11.8
19.0
14.8
13.0
14.9
15.3
17.5
12.6
12.7
14.7
15.4
14.3
19.6
18.1
16.8
9.9
9.9
13.7
12.0
9.0
11.7
10.7
31.4
12.5
15.1
14.5
12.6
13.1
15.4
14.6
13.8
13.1
15.0
(mpg)
Avg.
14.9
25.4
16.5
17.7
22.3
17.2
24.9
17.3
23.0
16.0
16.5
23.4
18.5
23.6
16.3
14.6
21.6
20.3
18.3
22.4
22.8
22.6
15.8
13.3
17.3
17.5
16.9
17.8
13.4
41.0
15.7
17.7
24.1
18.3
15.4
15.9
21.5
19.5
19.0
21.6
Current
City
16.0
31.0
18.0
22.0
22.0
16.0
24.0
17.0
36.0
13.0
17.0
29.0
20.0
25.0
14.0
15.0
22.0
19.0
20.0
20.0
20.0
23.0
17.0
14.0
18.0
18.0
17.0
19.0
14.0
60.0
19.0
15.0
24.0
17.0
20.0
18.0
20.0
20.0
20.0
21.0
EPA
Hwy
21.0
27.0
27.0
30.0
30.0
23.0
32.0
25.0
31.0
21.0
25.0
37.0
28.0
34.0
18.0
20.0
29.0
25.0
29.0
27.0
28.0
32.0
22.0
18.0
23.0
24.0
21.0
26.0
18.0
51.0
27.0
20.0
33.0
22.0
27.0
23.0
27.0
25.0
26.0
26.0
Label (mpg)
Combined
17.9
29.1
21.2
25.0
25.0
18.5
27.0
19.9
33.6
15.7
19.9
32.1
23.0
28.4
15.6
16.9
24.7
21.3
23.2
22.6
23.0
26.3
18.9
15.6
20.0
20.3
18.6
21.6
15.6
55.6
21.9
16.9
27.4
18.9
22.6
20.0
22.6
22.0
22.3
23.0
Table II.B-6 illustrates the differences between the Edmunds average values and the
current EPA label, the MPG-specific values, and, for the hybrids, the 5-cycle values.
23
-------
Table II.B-6. Edmunds Long-term Test Vehicles Compared to EPA Combined MPG
Estimates
MY MFR Model
2006 Honda Ridgeline
2006 Lexus RX 400h
2006 Mitsubishi Eclipse GT
2006 VW Jetta
2005 Audi A4 2.0T
2005 BMW X3
2005 Chevrolet Cobalt
2005 Dodge Magnum
2005 Ford Escape Hybrid
2005 Ford GT
2005 Ford Mustang GT
2005 Honda Accord Hybrid
2005 Honda Odyssey
2005 Kia Spectra
2005 Landrover LR3
2005 Nissan Frontier 4x4 Nismo
2005 Scion tC
2005 Subaru Legacy GT
2005 Toyota Solara
2005 Volvo S40
2004 Acura TL
2004 Chevrolet Malibu
2004 Chrysler Pacifica
2004 Ford F-150
2004 CMC Canyon
2004 Mazda RX-8
2004 Mitsubishi Endeavor
2004 Nissan Quest
2004 Nissan Titan
2004 Toyota Prius
2004 Toyota Sienna
2004 Volvo XC90
2003 Honda Accord
2003 Honda Pilot EX
2003 Infiniti G35 Coupe
2003 Lexus SC 430
2003 Mazda Mazda6
2003 Mitsubishi Outlander
2003 Nissan 350Z
2003 Subaru Forester
Average
Edmunds
Avg.
15
25
17
18
22
17
25
17
23
16
17
23
19
24
16
15
22
20
18
22
23
23
16
13
17
18
17
18
13
41
16
18
24
18
15
16
22
20
19
22
19
EPA Combined Label
(mpg)
MPG- 5-
Current Based Cycle
18 16
29 25 26
21 20 20
25 23 24
25 23
19 17
27 25 25
20 18
34 29 27
16 15 15
20 18 19
32 29 26
23 21 20
28 26 26
16 14
17 16 15
25 22 24
21 19 19
23 21
23 21 22
23 21 23
26 24 24
19 17
16 14 13
20 18
20 19 19
19 17 17
22 20 19
16 14
56 46 47
22 20 20
17 16 16
27 25 26
19 18 18
23 21 22
20 19 19
23 21 22
22 20
22 21 22
23 21
23 21
Difference: Edmunds
Vs. EPA (%)
MPG- 5-
Current Based Cycle
-17% -9%
-13% 1% -1%
-22% -16% -19%
-29% -22% -28%
-11% -2%
-7% 0%
-8% 1 % -2%
-13% -6%
-31% -20% -15%
2% 8% 4%
-17% -10% -12%
-27% -19% -10%
-19% -12% -9%
-17% -9% -8%
5% 14%
-14% -6% -6%
-12% -4% -8%
-5% 4% 7%
-21% -15%
-1% 8% 1%
-1% 8% 0%
-14% -6% -6%
-17% -9%
-15% -7% 1%
-13% -5%
-14% -6% -7%
-9% 0% -3%
-18% -10% -9%
-14% -6%
-26% -12% -12%
-28% -22% -22%
5% 14% 7%
-12% -4% -7%
-3% 3% 4%
-32% -27% -29%
-21% -15% -18%
-5% 3% -2%
-11% -5%
-15% -8% -13%
-6% 1%
-14% -6% -8%
24
-------
As can be seen from Table II.B-6, both the mpg-based and 5-cycle formulae still produce
fuel economy label values higher than those developed by Edmunds. However, the differences
are smaller than with the current label formulae. The difference for the mpg-based equations is
8%, while that for the 5-cycle formulae is 7% (the Edmunds estimates being the lower of the two
in both cases). While not shown, the standard deviation of the percentage difference is 12% for
the mpg-based equations and 9% for the 5-cycle formulae. Thus, the relative differences in
vehicles' fuel economy are slightly better indicated by the 5-cycle formulae.
The vehicles which Edmunds tested include four hybrids. The Edmunds fuel economy
measurements are on average 24%, 13%, and 9% lower than the current, mpg-based and 5-cycle
fuel economy label values, respectively. These percentage differences are 5-6% higher than
those for the 40 vehicles tested on average for the current and mpg-based labels, but only 2%
higher than that for the 5-cycle approach. Thus, the current and mpg-based approaches are
indicating hybrid-related benefits which the Edmunds testing is not confirming. However, the
relative benefit of hybrid technology as indicated by Edmunds testing and the 5-cycle approach
are very similar. In this sense, the Edmunds testing is more consistent with that of Consumer
Reports, versus AAA.
In contrast, the current, mpg-based and 5-cycle fuel economy label values for the 10%
conventional vehicles with the highest fuel economy average 18%, 8%, and 8%, respectively.
These latter differences are very similar to those for the average conventional vehicle. The same
findings hold for the 20% of conventional vehicles with the highest fuel economy. Thus, the
Edmunds testing is finding something about hybrids which is not occurring with either typical or
high fuel economy conventional vehicles.
C. Fleet-wide Estimates of Onroad Fuel Economy
We begin with a comparison of the 5-cycle fuel economy values with the fleetwide fuel
economy estimates developed by FHWA. Because we do not have fuel economy data for all
vehicles over all 5 dynamometer cycles, and therefore cannot develop a 5-cycle fuel economy
estimate for the current onroad fleet directly, this comparison requires a two step process.
The first step in this process compares fleetwide fuel economy estimates based on EPA's
current fuel economy labels to the FHWA estimate of onroad fuel economy. The second step in
this process is to compare combined city-highway fuel economy using the 5-cycle formulae to
that using the current EPA city and highway label procedures. This comparison is performed for
vehicles for which we have 5-cycle fuel economy data. We will assume that this relationship
also applies to those vehicles for which we do not have 5-cycle data.
In the NPRM, we added a third step which evaluated changes in FTP and HFET test
procedures which accompanied the implementation of the US06 and SC03 testing requirements.
We estimated that these changes had a positive impact on the fuel economy values measured
during these tests. However, as discussed in the Response to Comments document, we now
believe that these changes had a neutral impact on fuel economy. Thus, this third step is no
longer needed.
25
-------
Overall, the difference between 5-cycle fuel economy and FHWA onroad fuel economy
is the combination of the percentage differences from the two comparisons:
1) Current EPA label fuel economy to FHWA onroad fuel economy, and
2) 5-cycle fuel economy to current EPA label fuel economy (without 10% road load
adjustment).
FFIWA publishes fleet-wide estimates of onroad fuel economy for cars and light trucks in
their annual Highway Statistics publication.4 We will focus on the combined estimates for cars
and light trucks here, since various states use different criteria to distinguish between the two
vehicle classes. At the same time, the criteria used to distinguish between cars plus light trucks
and other vehicles are very consistent.
Table II.C-1 presents the FHWA estimates of vehicle miles traveled (VMT), fuel
consumption and onroad fuel economy for passenger cars and 4-tire, 2-wheel trucks.
Table II.C-1. FHWA-Based Estimate of Onroad Fuel Economy
Passenger cars
2 axle 4 tire Trucks
Light trucks
Passenger cars and
light trucks
Year
VMT (million miles)
Fuel Use (thousand gallons)
MPG (mpg)
VMT (million miles)
Fuel Use (thousand gallons)
MPG (mpg)
VMT (million miles)
Fuel Use (thousand gallons)
MPG (mpg)
VMT (million miles)
Fuel Use (thousand gallons)
MPG (mpg)
2003
1,672,079
75,455
22.2
984,094
60,758
16.2
908,712
55,271
16.4
2,580,791
130,726
19.7
2004
1,704,982
76,007
22.4
1,014,342
62,626
16.2
936,643
56,970
16.4
2,641,625
132,976
19.9
The FHWA category of 4-tire, 2-wheel trucks includes some vehicles which EPA classifies as
heavy-duty vehicles. We have adjusted the FHWA estimates upward to provide a more direct
comparison. This adjustment is based on a study performed by Oak Ridge National Laboratory
(ORNL).5 ORNL estimated both VMT and fuel use for several categories of light trucks. Class
1 and 2a trucks fall into EPA's definition of light-duty trucks, while Class 2b trucks do not.
Together, the three classes of trucks are approximately equivalent to FHWA's 4-wheel, 2-axle
truck class. The results of this study are shown in Table II.C-2.
26
-------
Table II.C-2. Breakdown of VMT and Fuel Use by 4-Wheel, 2-Axle
Class 1
Class 2a
Class 2b
Total
Class l+2a
VMT (billions)
672.7
251.9
76.7
1,001
92.3%
Fuel use (billion gallons)
37.4
18
5.5
60.9
91.0%
As can be seen, ORNL estimated that 92.3% of the VMT and 91.0% of the fuel use by 4-wheel,
2-axle trucks was by vehicles falling into EPA's light-duty truck category (and which are labeled
for fuel economy). Therefore, we adjusted FHWA's VMT and fuel use estimates for 4-wheel, 2-
axle trucks by these percentages to convert them to values applicable to EPA's light-duty truck
class. These adjusted values and the resulting onroad fuel economy are shown in Table II.C-1.
We then added the VMT and fuel use by passenger cars and EPA light trucks together and
calculated an overall fuel economy for the two vehicle classes. These values are also shown at
the bottom of Table II.C-1. The result is that the FHWA-based estimate of fleet-wide onroad
fuel economy for cars and EPA light trucks is 16.4 mpg for 2003 and 2004. This is nearly 20%
lower than the onroad fuel economy for light trucks presented in the NPRM analysis. The
difference is due to the use of more recent figures from FHWA which considers fuel economy
data indicating lower fuel economy from light trucks than previously estimated.
We then used the EPA MOBILE6.2 in-use emission model to calculate fleet-wide
average EPA combined fuel economy label values for these two years. MOBILE6.2 estimates
fuel economy using a sales-weighted average of the combined EPA city/highway label values for
each model year of cars and light trucks. MOBILE6.2 then estimates an average fuel economy
for the onroad fleet by weighting the fuel economy values for each model year by the fraction of
vehicles on the road from each model and their typical annual mileage (which decreases with
age). Thus, MOBILE6.2 is an ideal tool for estimating the EPA label fuel economy using the
current label formulae for the onroad vehicle fleet in any particular calendar year.
For 2003, MOBILE6.2 estimates average passenger car and light truck fuel economy of
24.0 mpg and 17.3 mpg, respectively. For 2004, MOBILE6.2 estimates average passenger car
and light truck fuel economy of 24.0 mpg and 17.4 mpg, respectively. We weighted the fuel
economy values for cars and light trucks together using their respective VMT from Table II.C-1.
The result were overall average label fuel economy values of 21.1 mpg for 2003 and 21.2 mpg
for 2004. Thus, for 2003 and 2004, the FHWA-based onroad fuel economy was 6.5% and 6.1%
lower than the current combined EPA label value, respectively. Thus, the result of the first step
in this process is an indication that the current labeling formulae could be over-estimating onroad
fuel economy by 6-7%.
Moving to the second step, in Tables III.E-1 and III.E-2 shown in Section III.E below, we
present city and highway fuel economy label values using both current and 5-cycle formulae for
615 2003-2006 model year vehicles. The FHWA estimates apply to all driving, both city and
highway. Therefore, we are primarily interested in combined city-highway fuel economy values.
Also, we are using FHWA estimates for the 2003 and 2004 calendar years, as these are the most
recent available. The number of hybrid vehicles on the road was very low during this timeframe,
27
-------
much lower than the 2% level present in our certification database. Therefore, we will only use
the 5-cycle fuel economy estimates for the 601 non-hybrid vehicles in our database. There is no
need to perform this comparison separately for the mpg-based formulae, since the average fuel
economy from the 5-cycle and mpg-based formulae are identical for non-hybrid vehicles.
The combined fuel economy using the current label formulae is a 55/45 harmonic
weighting of the current city and highway fuel economy labels. The average combined fuel
economy using the current EPA label values for these 601 vehicles is 20.9 mpg. For the final 5-
cycle formulae, combined fuel economy is a 43/57 harmonic weighting of the 5-cycle city and
highway fuel economies. This city/highway split for the 5-cycle fuel economies is based on:
1) the assumption that driving generally less than 45 mph is city driving and that above
45 mph is highway driving, and
2) the description of onroad driving patterns contained in MOVES.
The mathematical formula for converting the 5-cycle city and highway fuel economy values into
an estimate of average onroad fuel economy is as follows:
Aver age onroad fuel economy = -.—
0.43 0.57
^ 5 - cycle City FE j ^ 5 - cycle Highway FE j
The average combined 5-cycle fuel economy using this formula for the 601 conventional
vehicles is 19.6 mpg, which is 6.2% lower than that based on the current label values. This is the
result of the second step in the process.
Overall, then, the current label values over-estimate onroad fuel economy per FHWA
(with some adjustments by EPA) by 6-6.5%, while the 5-cycle formulae decrease current label
values (of the 2002-2003 fleet) by 6.2%. Thus, the final 5-cycle formulae should move the
combined fuel economy label values to within a few tenths of a percent of a comparable estimate
of fleetwide fuel economy using FHWA techniques. This should not be surprising, since the
value of the factor in the 5-cycle formulae representing factors not represented in dynamometer
tests was set to match onroad fuel economy as estimated by FHWA (see Section III. A.5 below).
D. Overall Comparison of Hybrid Fuel Economy
When comparing onroad fuel economy to EPA estimates, it is often appropriate to focus
on the EPA combined fuel economy, as it is not possible to determine whether a particular
vehicle's driving was city-like or highway-like. Overall, the 5-cycle formulae predict a
combined fuel economy about 6% less than the current combined fuel economy label for our 615
vehicle certification fuel economy database.
These relationships hold for the complete 615 vehicle database, as well as the 601
conventional vehicles which dominate the database. However, the effect of the 5-cycle formulae
on the combined fuel economy of hybrids is much more significant. Overall, the 5-cycle
formulae predict a combined fuel economy about 18% less than the current combined fuel
economy label for the 14 hybrids in our certification fuel economy database. Thus, the 5-cycle
28
-------
formulae reduce hybrid fuel economy roughly 12% more than non-hybrids compared to today's
labels. Two thirds of this difference occurs in city fuel economy. It is primarily due to a greater
impact of running fuel use at colder temperatures and inclusion of US06 city driving in the 5-
cycle formulae. This difference appears to be related more to hybrid technology itself, as
opposed to just high fuel economy values. For example, the average impact of the 5-cycle
formulae on the combined fuel economy of the top 10% of conventional vehicles in terms of
current combined fuel economy is 8%. This is only 2% greater than that for all conventional
vehicles and well below that for hybrids.
Some care must be taken in relating to this difference between the impact of the 5-cycle
formulae on conventional and hybrid vehicles. First and foremost, the impact of the 5-cycle
formulae on label fuel economy involves a number of projections which may not be accurate for
individual vehicles. For example, we currently lack measured fuel economy values over the
US06 city and highway bags. Thus, these must be estimated. We are using one relationship for
conventional vehicles and another relationship for hybrids. The former is based on the testing of
over 100 vehicles, which shows fairly consistent results. The latter is based on the testing of two
vehicles. The relative performance of other hybrid vehicles over the two bags of the US06 test
could differ significantly from our current estimate. Similarly, we have included the effect of
turning on the heater or defroster during the cold FTP, as this will be required in the future.
However, the data currently available on fuel economy over the cold FTP do not include this
factor. Our projected impact for conventional vehicles is relatively small. Thus, variability
between vehicles is likely even smaller. However, the testing of two hybrids showed large
impacts on Bag 1 fuel economy at 20°F. Other hybrids may show different impacts.
Second, manufacturers may improve their hybrid designs in the future to reduce the
impact of colder temperatures, air conditioning operation, etc. While we believe that
manufacturers have always been concerned about fuel efficiency under realistic conditions, the
fact that fuel economy labels have been based solely on operation at 75°F has inevitably focused
their attention especially on this type of operation. Hybrid technology is still relatively new and
improvements are constantly being made. With the 5-cycle formulae, we expect that these
improvements will affect vehicle operation over a much wider set of in-use conditions than
might have been the case with the current label formulae.
Nonetheless, the question still arises: is this greater reduction in combined fuel economy
of today's hybrids appropriate? In this section, we compile all the onroad fuel economy
measurements and consumer organization estimates in one place and compare them to combined
fuel economy estimated using the current, 5-cycle and mpg-based formulae.
Table II.D-1 compares three sets of onroad fuel economy measurements to both current
and final EPA combined fuel economy label values. Combined fuel economy using the current
label formulae weights city fuel economy 55% and highway fuel economy 45%. Combined fuel
economy using the 5-cycle and mpg-based label formulae weighs city fuel economy by 43% and
highway fuel economy by 57%. It should be noted that the values for current combined fuel
economy are those from EPA's certification database in order to be directly comparable to the 5-
cycle and mpg-based values. However, they are not the current official label values. The
29
-------
official label values average about 3% higher due to differences between the worse case vehicles
tested over the Supplemental FTP cycles and the average vehicle sold.
Table II.D-1. Onroad Hybrid Fuel Economy Versus EPA Label Estimates (mpg)
2001 Honda Insight
2003 Honda Civic
2004 Toyota Prius
2004 Chevrolet Silverado 4wd
2005 Ford Escape 4wd
2005 Honda Accord
2005 Lexus RX400h
EPA Combined Fuel Economy
Current
61
46
54
16
30
32
28
MPG-Based
51
40
45
15
26
29
24
5-Cycle
50
37-38
45
15
24
26
24
Onroad Fuel Economy Measurements
DOE
FreedomCar
45
38
45
18
27
28
25
YourMPG
66
46
48
15
30
31
25
EPA Kansas
City
47
40
50
—
—
—
—
It is difficult to draw any definite conclusions from the above data due to its scatter. The
onroad fuel economy estimates from the YourMPG database tend to be significantly higher than
those from the DOE FreedomCar project. The Kansas City measurements tend to fall in
between, except for the Prius. Label fuel economy values based on all three approaches can
come close to one of the three onroad fuel economy estimates, even for the same vehicle.
For example, the current EPA label formulae appear to significantly over-estimate fuel
economy as measured in the FreedomCar program, except for the Chevrolet Silverado. This
vehicle has the least hybrid capability with respect to improved fuel economy of the vehicles
listed in Table II-D.l. The mpg-based and 5-cycle label formulae tend to perform equally well
with respect to the DOE FreedomCar program
In contrast, the current label formulae provide reasonable estimates of onroad fuel
economy for five of the seven hybrids based on the YourMPG database. The mpg-based and 5-
cycle label formulae do so for four of the seven hybrids based on the YourMPG database. As
would be expected, the sets of five and four hybrids tend not to overlap.
Finally, the current label formulae over-estimate onroad fuel economy as measured in the
Kansas City test program. The mpg-based and 5-cycle label formulae tend to provide reasonable
estimates of onroad fuel economy as measured in Kansas City on average, over-estimating fuel
economy for one of the three matched hybrids and under-estimating fuel economy for one
hybrid.
Table II.D-2 compares onroad fuel economy estimates by consumer organizations to EPA
estimates.
30
-------
Table II.D-2. Onroad Hybrid Fuel Economy Estimates Versus EPA Label Estimates
2001 Honda Insight
2003 Honda Civic
2004 Toyota Prius
2005 Ford Escape 4wd
2005 Honda Accord
2005 Lexus RX400h
EPA Combined Fuel Economy
Current
61
46
54
30
32
28
MPG-Based
51
40
45
26
29
24
5-Cycle
50
37-38
45
24
26
24
Onroad Fuel Economy Estimates
Consumer Reports
51
36
44
26
25
—
Edmunds
—
—
41
23
23.4
25.4
AAA
58
—
52
—
—
—
Again, there is significant scatter in the data. Of the three consumer organizations,
Edmunds predicts the lowest fuel economy. The Edmunds fuel economy values are lower than
all three sets of EPA fuel economy estimates. Since the 5-cycle formulae produce the lowest
estimates on label fuel economy, the 5-cycle formulae come closest to matching the Edmunds'
fuel economy values.
The 5-cycle fuel economy values match those of Consumer Reports very closely,
differing by at most 2 mpg for any individual vehicle. The mpg-based estimates are higher for
three of the five matching hybrids. The current label values exceed those found by Consumer
Reports significantly in all cases.
The two AAA estimates are higher than the 5-cycle and mpg-based estimates, but lower
than (but closer to) the current EPA estimates. Overall, the 5-cycle formulae tend to match the
findings of the three consumer organizations more closely than the other two label approaches.
However, there is significant scatter in the data. Over time, we will continue to assess how our
5-cycle estimates compare with those of other studies. In particular, it would be useful to have
both vehicle activity and fuel economy data, coupled with environmental conditions in order to
more precisely verify the ability of the five cycles to estimate onroad fuel economy under the full
range of ambient conditions.
31
-------
Chapter II References
1. Eastern Research Group. Kansas City PM Characterization Study: Round 1 Testing Report.
ERG No. 0133.18.001.001, March 7, 2005.
2. Eastern Research Group. Kansas City Field Testing: Round 2 Testing Final Progress Report.
ERG No. 0133.18.005.003, April 15, 2005.
3. Eastern Research Group. Late Model Vehicle Emissions and Fuel Economy Characterization
Study: Addendum to the Kansas City Exhaust Characterization Study-Draft Report. ERG No.
0133.18.004.001, September 26, 2005.
4. U.S. Department of Transportation, Federal Highway Administration. Highway Statistics
2004. See Table VM-1. Website: http://www.fhwa.dot.gov/policy/ohim/hs04/htm/vml.htm.
5. Davis, S. C. and S. W. Diegel. Transportation Energy Data Book: Edition 24. Oak Ridge
National Laboratory, prepared for the U.S. Department of Energy, Report No. ORNL-6973,
December 2004, Table 4.3. Website: http://cta.ornl.gov/data/download24.shtml.
32
-------
Chapter III: Documentation of Final Approach for Estimating
On-Road Fuel Economy
The current fuel economy label values utilize measured fuel economy over city (FTP)
and highway (HFET) driving cycles and adjust these values downward by 10 and 22%,
respectively, to account for a variety of factors not addressed in EPA's vehicle test procedures.
These factors include differences between the way vehicles are driven on the road and over the
test cycles, air conditioning use, widely varying ambient temperature and humidity, widely
varying trip lengths, wind, precipitation, rough road conditions, hills, etc.
The purpose of the new formulae for city and highway fuel economies is to better
account for three of these factors: 1) on-road driving patterns (i.e., vehicle speeds and
accelerations), 2) air conditioning, and 3) colder temperatures. Vehicles are often driven more
aggressively and at higher speeds than is represented in the FTP and HFET tests, which have
maximum speeds of 55-60 mph and maximum acceleration rates of 3.2-3.3 mph per second. The
incorporation of measured fuel economy over the US06 test cycle into the fuel economy label
values makes the label values more realistic, as this cycle includes speeds up to 80 mph and
acceleration rates of 8.4 mph per second.
Drivers often use air conditioning in warm, humid conditions, while the air conditioner is
turned off during the FTP and FIFET tests. The incorporation of measured fuel economy over
the SC03 test cycle into the fuel economy label values reflects the added fuel needed to operate
the air conditioning system. The SC03 test is performed at 95 degrees Fahrenheit (°F) with
simulated solar heating.
Vehicles are also often driven at temperatures below 75°F, at which the FTP and HFET
tests are performed. The incorporation of measured fuel economy over the cold temperature FTP
test into the fuel economy label values reflects the additional fuel needed to start up a cold
engine at colder temperatures.
We developed two methods for incorporating these three factors into our onroad fuel
economy predictions. The first method, termed vehicle specific 5-cycle, combines the fuel
economy over all five dynamometer cycles in a vehicle specific manner (i.e., as much
information as possible reflects the fuel economy performance of the specific vehicle being
examined). The second method, termed mpg-based, utilizes fuel economy estimates based on
the vehicle specific 5-cycle method for a large number of vehicles and develops generic
adjustment factors as a function of the vehicle's FTP and HFET fuel economy values.
In Section III. A, we describe the development of the vehicle specific 5-cycle method. In
Section III.B, we describe the mpg-based method. In Section III.C, we evaluate the range and
variability of onroad fuel economy experienced by drivers of the same vehicle. In section HID,
we describe how the current city and highway fuel economy values would change under the two
final methods. Finally, in section III.E, we evaluate the sensitivities and uncertainties in the
vehicle specific 5-cycle formulae.
33
-------
A. Vehicle Specific 5-Cycle Method for Estimating On-Road Fuel
Economy from Dynamometer Tests
The city and highway fuel economy label values are intended to give the vehicle
purchaser an idea of the fuel economy that they should expect to achieve while driving the
vehicle during normal use. As such, the label values are intended to take into account the effect
of seasonal and geographical variations on automotive fuel economy, as well as the different
driving habits of individual drivers. Due to variations in climate and the way various drivers
drive their vehicle, no one set of fuel economy values can accurately predict any individual
drivers' actual fuel economy while driving. However, the set of fuel economy values should be
able to predict on-road fuel economy for the average driver under average environmental
conditions. The goal of the analysis presented below is to use the available fuel economy
information developed during emission and fuel economy testing to predict on-road fuel
economy under these "average" conditions to the greatest extent possible. This onroad average
fuel economy can then be adjusted to represent more worse-case conditions, such as 25th or 10th
percentile estimates (i.e., fuel economy levels achieved by 75% or 90% of all drivers,
respectively), which is discussed in Section III.C below.
As described in the preamble to this final rule, we chose to base the fuel economy label
values on vehicle emission and fuel economy tests which are already being performed. This
minimizes the costs associated with the final rule, as described below in Chapter IV. The five
current emission and fuel economy tests and their key aspects are described below in Table
III.A-1.
Table III.A-1. Key Features of the Five Current Emission and Fuel Economy Tests
Test
FTP
HFET
US06
SC03
Cold FTP
Driving
Low speed
Mid-speed
Aggressive; low
and high speed
Low speed
Low speed
Ambient
Temperature
75°F
75°F
75°F
95°F
20°F
Engine Condition
at Start
Cold and hot
Hot
Hot
Hot
Cold and hot
Accessories
None
None
None
A/Con
None
We have highlighted in bold the distinctive features of the five current vehicle tests. The
FTP, HFET and US06 are all performed at an ambient temperature of 75°F. Each test consists of
a distinctive driving pattern. In addition, the FTP test consists of three distinct measurements,
called bags because the emissions produced during each portion of the test are literally collected
in separate plastic "bags". Bags 1 and 3 consist of the exact same driving pattern, while Bag 2
consists of a different pattern. Thus, fuel economy measurements at 75°F are available for four
distinct driving cycles.
Bag 1 begins with an engine start after the vehicle has been sitting with the engine off for
at least 12 hours, representing an overnight soak (soak refers to the time during which a vehicle
sits with its engine off). This is referred to as a "cold" start. In this case, the cold start occurs at
34
-------
75°F, which may not seem not very "cold". However, the engine is at the same temperature as
the ambient air, so it is referred to as being "cold". After sitting this long, the engine has
essentially fully cooled off and sitting longer has no effect on either emissions or fuel economy.
Bag 3 begins with an engine start after the vehicle has been sitting with the engine off for only
10 minutes, representing a short stop to refuel, a short shopping trip, etc. This is referred to as a
"hot" start, since the engine is still essentially at its fully warmed up temperature. In-use, vehicle
soaks between starts occur anywhere between a few seconds to a few days. The hot and cold
starts are intended to bracket this distribution of in-use soak times and ensure that vehicle
manufacturers design their emission control systems to work efficiently under a wide range of
operation. Practically speaking, very little fuel is necessary to warm up the vehicle after it has
been sitting with the engine off for only 10 minutes. Thus, the difference between fuel use over
Bags 1 and 3 is that associated with a cold start. As indicated in Table III.A-1, this estimate is
available at both 75 and 20°F.
The SC03 test is the only test performed with the air conditioning system operational.
Therefore, its results are used to augment the fuel economy from the five driving pattern tests for
the fuel needed to operate air conditioning. The cold FTP is the only test performed at a
temperature below 75°F. Therefore, its results are used to represent the additional fuel needed to
warm up an engine after a cold start, as well as any fuel needed to operate a warmed up engine,
at colder temperatures.
Conceptually, our approach to modeling fuel economy can be depicted as follows:
Onroad fuel economy =
Warmed up fuel economy (from Bags 2 and 3 of the FTP, HFET and US06 cycles),
decreased by the need to warm up a cold engine (from Bags 1 and 3 of the FTP and cold FTP),
decreased by use of the air conditioner (from SC03), and
decreased due to operation at colder temperatures (from the cold FTP).
Actually, since cold starts, air conditioning and colder temperatures simply add fuel use over and
above that needed for warmed up driving, it is more straightforward to model these effects in
terms of fuel consumption (e.g., gallons of fuel used per mile) than in terms of fuel economy. In
terms of fuel consumption, the above equation looks like this:
Onroad fuel economy = I/ Onroad fuel consumption
Onroad fuel consumption =
Warmed up fuel consumption (from Bags 2 and 3 of the FTP, FIFET and US06 cycles),
plus the extra fuel associated with
1) warming up a cold engine (from Bags 1 and 3 of the FTP and the cold FTP),
2) use of the air conditioner (from SC03), and
3) operation at colder temperatures (from the cold FTP).
The remainder of this section is broken up into 6 pieces. The first, section III.A.I,
develops a methodology for estimating fuel use related to engine start-up, or start fuel use. The
second, Section III. A.2, develops a methodology for estimating fuel use once the engine is
35
-------
warmed up at 75°F, assuming that the air conditioner is not turned on. The third, Section III. A.3,
develops a methodology for estimating fuel use due to air conditioner use. The fourth, Section
III. A.4, develops a methodology for estimating fuel use once the engine is warmed up at colder
temperatures. This fourth section also presents a combination of the results from the previous
two sections into overall formulae for running fuel use during city, highway and composite
driving. The fifth, Section III. A.5, evaluates the impact of factors that affect onroad fuel
economy but which are not addressed by any of the five dynamometer cycles. The sixth, Section
III. A.6, presents for final 5-cycle city and highway fuel economy formulae.
1. Start Fuel Use
We estimate the fuel needed to start and warm up the engine separately from fuel used to
operate the engine after start-up, or running fuel use primarily to be able to estimate fuel
economy for trips of various lengths. The longer the trip, the less significant is start fuel use.
This is consistent with the approach taken in EPA emission models, such as MOBILE6.2 and
MOVES. We estimate the volume of fuel needed to start and warm up an engine first. We then
estimate average trip lengths for both city and highway driving. Finally, we combine the two
estimates to develop a formula for start fuel use for typical trip lengths during city and highway
driving.
a. Start Fuel
For a specific vehicle, the fuel needed to warm up the engine depends primarily on two
factors:
1) The ambient temperature at which the vehicle has been sitting, and
2) The length of time that the vehicle has been sitting since it was last used
(commonly referred to as soak time).
Emissions during engine start up have been studied for some time.1 Most recently,
estimates of start fuel use as a function of ambient temperature were made for use in EPA's new
emission inventory model, MOVES (MOtor Vehicle Emission inventory Siystem).d For
MOVES, EPA analyzed start fuel use from 580 gasoline fueled vehicles.2 In this analysis, the
start fuel use measured was the difference between fuel use during Bag 1 of the FTP and Bag 3
of the FTP. The only difference between these two bags is the time prior to the test that the
engine has been turned off, or "soaking." Prior to Bag 1, the engine has been off for 12 hours or
more. This start is commonly referred to as a cold start. Prior to Bag 3, the engine has been off
for only 10 minutes. This start is commonly referred to as a hot start. Since start fuel use during
a hot start is much lower than that during a cold start, the difference between start fuel use for a
cold start and a hot start is usually assumed to be that of the cold start.
The resulting relationship between cold start fuel use at other ambient temperatures
relative to that at 75°F (a typical temperature for the standard FTP test) is as follows:
d A draft of MOVES2004 was released for public comment on Dec. 31, 2004.
36
-------
Equation 1
Cold Start Fuel Use Re lative to that at 75 F =
1 - (0.01971 x (AmbientTemperature - 75)) + (o.000219 x (AmbientTemperature - 75)2)
At 75°F, the nominal temperature of the FTP test, this formula yields a value of 1.0. At 20°F, for
example, the temperature of the cold temperature FTP, it yields a value of 2.75. This means that
the volume of fuel needed to start and warm up an engine and drivetrain is 2.75 times as great
after a 12 hour soak at 20°F as it is at 75°F. These relationships assume that the vehicle had been
sitting with the engine turned off for the same amount of time at each temperature.
It should be noted that none of the 580 vehicles tested incorporated hybrid technology.
Thus, the application of Equation 1 to hybrids involves greater uncertainty than for other
gasoline vehicles. This issue is addressed in detail in Section III.E below.
EPA also analyzed start fuel use for diesel vehicles. Relevant data were only available
for 66 vehicles, or roughly one-tenth the number of gasoline vehicles. Based on these data, EPA
found that start fuel use for diesels was roughly 44% that of gasoline vehicles. Thus, the
relationship between cold start fuel use at other ambient temperatures relative to that at 75°F for
diesels is as follows:
Equation 2
Diesel Cold Start Fuel Use Re lative to that at 15 F =
1 - (0.00867 x (AmbientTemperature - 75)) + (o.000096 x (AmbientTemperature - 75)2)
For a typical diesel vehicle, start fuel use after a cold start at 20°F is only 1.77 times that at 75°F.
Moving to the issue of soak time prior to engine start up, the Draft MOVES2004 model
does not yet include estimates for the effect of soak time on start fuel use. Therefore, we
obtained a relationship between start fuel use and ambient temperature which was developed by
the California Air Resources Board for use in their emission inventory model, EMFAC2000.3
These relationships were based on the testing of 238 vehicles. EPA utilizes the results of this
study in our current emission model, MOBILE6.2, to estimate the effect of soak time on
regulated emissions (VOC, CO, NOx) during start-up. The equation for fuel use versus soak
time (in minutes) relative to the fuel use after a 12 hour soak is as follows:
For soaks of 90 minutes or less:
Equations
StartFuelUse^ = 0.00433672 x SoakTime - 0.000002393 x SoakTime
37
-------
For soaks greater than 90 minutes:
Equation 4
StartFuelUsex = 0.25889542 + 0.0014848 x SoakTime - 0.0000006364 x SoakTime2
As is done in EMFAC2000 and MOBILE6.2, we assumed that these relationships are
independent of ambient temperature.
In order obtain the combined effect of ambient temperature and soak time, we combined
the above equations multiplicatively, as follows:
For gasoline vehicles:
For soaks of 90 minutes or less:
Equation 5
StartFuelUsex = [o.00433672 x SoakTime - 0.000002393 x SoakTime2 ] x
[l - 0.01971 x (AmbientTemperature -15)+ 0.000219 x (AmbientTemperature - 15)2 J
For soaks greater than 90 minutes:
Equation 6
StartFuelUsex = [o.25889542 + 0.0014848 x SoakTime - 0.0000006364 x SoakTime2 ]x
[l - 0.01971 x (AmbientTemperature -15)+ 0.000219 x (AmbientTemperature -15)2\
For diesel vehicles:
For soaks of 90 minutes or less:
Equation 7
StartFuelUsex = [o.00433672 x SoakTime - 0.000002393 x SoakTime2 ] x
[l - 0.00867 x (AmbientTemperature -15)+ 0.000096 x (AmbientTemperature -15)2 \
For soaks greater than 90 minutes:
Equation 8
StartFuelUsex = [o.25889542 + 0.0014848 x SoakTime - 0.0000006364 x SoakTime2 ]x
[l - 0.00867 x (AmbientTemperature -15}+ 0.000096 x (AmbientTemperature -15}2 \
All of the above equations estimate start fuel use in terms of the fraction of start fuel use
following an overnight soak at 75°F, which are the conditions of the "cold start" contained in the
FTP. We will use these equations to estimate the relative start fuel use for the range of starting
conditions occurring throughout the nation. We will then sum up these start fuel volumes and
estimate the average start fuel use for an engine start in the U.S. Then, we will estimate the
38
-------
combination of start fuel volumes measured in the FTP tests performed at 20 and 75°F which is
equal to the national average start fuel use.
The "hot" and "cold" starts contained in the standard and cold temperature FTP tests
occur after 10 minute and 12 hour or longer hour soaks, respectively. With the hot start, the
engine was fully warmed up prior to the test and turned off for 10 minutes prior to being turned
on again at the start of the hot start portion of the FTP. With the cold start, the vehicle has been
sitting with the engine off for at least 12 hours in a room at the temperature specified.
Equations 3 and 4 relating the effect of soak time on start fuel use indicate that the start
fuel use after a 10 minute soak is only 4% of that after a 12 hour soak. Equation 1 relating the
effect of temperature on start fuel use from gasoline vehicles indicates that start fuel use at 20°F
is 2.75 times that at 75°F. Combining these effects, the start fuel use after a 10 minutes soak at
20°F is about 11% (0.04 * 2.75) of the start fuel use following a 12 hour soak at 75°F. Thus, the
start fuel use after the hot starts of both standard and cold temperature FTP tests are relatively
small compared to that of a cold start at 75°F. The US06, SC03 and HFET tests begin with a hot
start at 75°F, as is the case with Bag 3 of the FTP. Hereafter, we ignore any start fuel use
included in these three tests due to their hot start. (Bag 2 of the FTP does not begin with an
engine start. The sampling equipment simply switches the emissions from Bag 1 to Bag 2 during
a vehicle idle at second 505 of the test.)
In order to estimate start fuel use throughout the U.S. under average ambient conditions,
we need estimates of the soak times for typical vehicle operation, as well as the ambient
temperature at start up. The amount of time a vehicle has sat prior to start up varies dramatically
depending on the time of day at which it is started. For example, for vehicles started up at 6 am,
nearly all have sat overnight. However, for vehicles started at noon, most have been driven in
the past 4-5 hours. Ambient temperatures vary significantly during the day, so it is more
accurate to evaluate start fuel use by hour of the day rather than simply at the daily average
temperature. Ambient temperatures also vary dramatically across the U.S., as does the
distribution of vehicle miles traveled (VMT). Therefore, we combined estimates of vehicle starts
and prior soak times by hour of the day with estimates of ambient temperature and VMT by
county in order to reflect the effects of both soak time and ambient temperature on start fuel use.
We obtained estimates of each of these input parameters from EPA's MOBLE6.2 and
MOVES2004 emission models. The Draft MOVES2004 model includes estimates of ambient
temperature by hour of the day for each month of the year for each county in the U.S.4 These
estimates were obtained from the National Weather Service and represent 30-year averages. The
Draft MOVES2004 model also includes estimates of vehicle miles traveled (VMT) by vehicle
type for every county in the U.S. during 2002.5 We assumed that the distribution of engine starts
across counties was the same as that for VMT (i.e., that trip length was the same across the U.S.).
We used these estimates to determine the percentage of total U.S. VMT by cars (LDVs) and light
trucks (LDTs) occurring in each county (excluding Puerto Rico and the Virgin Islands).
MOBILE6.2 includes estimates of the frequency distributions of vehicle soak times prior
to vehicle start-up by time of day, as well as the frequency distribution of vehicle starts by hour
of the day.6 Table III.A-2 presents the distribution of starts by the hour of the day. MOBILE6.2
39
-------
includes separate estimates for weekdays and weekends. Since we are primarily concerned with
fuel use on an annual average basis, we combined the two sets of estimates by weighting the
percentage of vehicle starts occurring after specific lengths of soak at each hour of the day for
weekdays by five-sevenths and those for weekends by two-sevenths and added the two sets of
percentages. This is indicated below.
Table III.A-2. Distribution of Starts by Hour of the Day (in percent)
Hour of the
Day
6 am
7 am
8 am
9 am
10am
11 am
Noon
1 pm
2pm
3 pm
4 pm
5 pm
6 pm
7 pm - 5 am
Weekday
2.04
5.54
6.02
4.73
5.16
6.72
8.07
7.30
8.04
8.98
8.41
7.73
6.02
15.24
Weekend
0.91
1.93
3.10
6.45
6.91
7.97
10.16
7.26
8.89
7.36
8.02
7.11
6.15
17.78
Average Day
1.72
4.51
5.19
5.22
5.66
7.08
8.67
7.29
8.28
8.52
8.30
7.55
6.05
15.97
The relative amounts of VMT by month of the year were taken from the Draft MOVE2004
model and are shown in Table III. A.-3 below.7 MOVES includes estimates for both non leap
years and leap years. We averaged the two estimates in a 3:1 ratio to develop estimates for a
typical year.
40
-------
Table III.A-3. Breakdown of Annual VMT by Month
Month
January
February
March
April
May
June
July
August
September
October
November
December
Non Leap
Year
0.0731
0.0697
0.0817
0.0823
0.0875
0.0883
0.0923
0.0934
0.0847
0.0865
0.0802
0.0802
Leap Year
0.0729
0.0720
0.0815
0.0821
0.0873
0.0881
0.0921
0.0932
0.0845
0.0863
0.0800
0.0800
Average Year
0.0731
0.0703
0.0817
0.0823
0.0875
0.0883
0.0923
0.0934
0.0847
0.0865
0.0802
0.0802
We obtained our estimate of total VMT by LDVs and LDTs by county from that used in
the Draft 2002 Mobile National Emission Inventory.5 We assumed that trip length is
independent of the season of the year.
We first estimated the effect of soak time on start fuel use by hour of the day. The first
step in this procedure was to estimate the percentage of vehicle starts occurring after specific
lengths of soak at each hour of the day. As is the case for vehicle starts by time of day,
MOBILE6.2 includes estimates of soak times for weekdays and weekends. Again, we weighted
the percentage of vehicle starts occurring after specific lengths of soak at each hour of the day
for weekdays by five-sevenths and those for weekends by two-sevenths and added the two sets
of percentages. MOBILE6.2 tracks 69 distinct intervals of soak time for 14 "hours" of the day
(starts between 9 pm and 5:59 am are combined into a single "hour"). Thus, the specific
estimates are too extensive for presentation here. However, to provide an indication of how soak
times are distributed throughout the day, Table III.A-4 presents the distribution of starts by soak
time for weekdays for several aggregated soak time intervals.
41
-------
Table III.A-4. Distribution of Starts by Soak Time: Three Hours
During Weekdays
Soak Time (minutes)
0-20
21-40
41-60
61-120
121-240
241-720
720+
7:00 AM
25.7%
2.2%
1.7%
0.3%
0.3%
27.1%
42.7%
Noon
48.3%
12.0%
8.6%
7.4%
8.8%
8.4%
6.5%
5:OOPM
43.4%
12.2%
6.4%
10.2%
7.4%
18.9%
1.4%
The second step in this procedure is to weight the relative start fuel use for each soak
time interval by the percentage of starts occurring after that range of soak times. The result is a
percentage of an overnight soak equivalent for each hour of the day. These estimates are shown
in the second column of Table III. A-5 for each hour of the day. For example, between 6 and 7
a.m., each start uses a volume of fuel equivalent to 68.3% of that associated with a start
following an overnight soak. As can be seen, these estimates ranged from a low of 0.25 of an
overnight soak at 2 pm to a high of 0.68 of an overnight soak at 6 am. This follows common
sense, as most vehicles being started at 6 am in the morning have sat overnight, while many
vehicles being started in the middle of the afternoon have been used in the past few minutes or
hours. We assume that these estimates are independent of temperature, because the variation in
temperature during any particular hour of the day is relatively small.
42
-------
Table III.A-5. Estimation of Daily Average Overnight Soak Equivalent
Hour of the Day
6 a.m.
7a.m.
8 a.m
9 a.m.
10 a.m
11 a.m.
12 Noon
I p.m.
2p.m.
3 p.m.
4 p.m.
5 p.m.
6p.m.
7p.m - 5 a.m.
Sum
Overnight Soak, Start
Fuel Use Equivalent
68.3%
68.5%
51.5%
41.1%
34.2%
29.1%
25.5%
25.8%
25.1%
26.9%
25.4%
27.8%
26.9%
35.5%
Distribution of Starts by
Hour
1.7%
4.5%
5.2%
5.2%
5.7%
7.1%
8.7%
7.3%
8.3%
8.5%
8.3%
7.5%
6.1%
16.0%
Product of Columns 2
and 3
1.2%
3.1%
2.7%
2.1%
1.9%
2.1%
2.2%
1.9%
2.1%
2.3%
2.1%
2.1%
1.6%
5.7%
33.0%
The distribution of starts by hour of the day is also shown in Table III.A-5. By the weighting the
overnight soak, start fuel use equivalents by the percentage of starts for each hour of the day, we
can calculate the average overnight soak, start fuel use equivalent for each start throughout the
day. As shown by the sum of the last column in Table III.A-5, the average start occurring
throughout the day uses a volume of fuel equal to 33% of that following an overnight soak. In
comparison, the cold start in the FTP has a weighting of 43%.
While this analysis produces a reasonable estimate of the relative number of hot and cold
start equivalents during real world driving, it ignores the effect of ambient temperature, which
varies throughout the day, as well as between seasons. In order to estimate start fuel use across
the nation throughout each day and throughout the year, we estimated the start fuel use for each
hour of the day by month for each county in the U.S. and then weighted each estimate by the
relative number of starts occurring in each hour of the day and by the relative amount VMT in
each month and county. Finally we summed the weighted start fuel use estimates across all
hours of the days, months and counties to determine the average start fuel use in terms of a cold
start at 75°F.
The national average start fuel use resulting from this process was 0.4665 of an overnight
soak at 75°F for gasoline vehicles and 0.4137 for diesels. We can simulate these average start
fuel use estimates with a variety of combinations of hot and cold starts at 20°F and 75°F. In
order to select a single combination, we used the estimate described in Table III.A-5 above that
the daily average start fuel use at a constant temperature is 33.0% of that of a cold start (i.e., that
following an overnight soak at that temperature). Assuming that start fuel use after a hot start
43
-------
(10 minute soak) is negligible and using equations (5) and (6) for relative start fuel use as a
function of ambient temperature and soak time, we determined that a 24% weighting of a cold
start at 20°F and a 76% weighting of a cold start at 75°F, both multiplied by 33%, has the same
start fuel usage as 46.65% of a cold start at 75°F. Thus, for gasoline vehicles, we can simulate
the entire distribution of vehicle starts throughout the nation annually by summing the excess
start fuel from the cold start for the FTP at 20°F times 0.0904 (24% * 33.0%) and that for the
cold start for the FTP at 75°F times 0.2394 (76% * 33.0%).
For diesels, the appropriate weight for the cold start at 20°F is the same as that for
gasoline vehicles, 24%. The weight is the same, because the shape of the curves of cold start
fuel use for both temperature and soak time for diesels is the same as that for gasoline vehicles.
Thus, for diesels, start fuel use under national average conditions is the excess start fuel from the
cold start for the FTP at 20°F times 0.0904 plus the cold start for the FTP at 75°F times 0.2394.
As mentioned above, cold start fuel use is the difference in total fuel use in Bags 1 and 3 of the
FTP, either at 75°F or 20°F. As also mentioned above, we will evaluate the appropriateness of
applying these weights to hybrid vehicles in Section III.E below.
b. Trip Length
The previous section estimated start fuel use in terms of total fuel use per start. In this
section, we address the frequency of starts per mile of typical on-road driving. The inverse of
starts per miles is trip length. The FTP implicitly includes one engine start for every 7.5 miles of
driving (i.e., an average trip length of 7.5 miles). This was the average trip length estimated for
Los Angeles in 1969. We have updated this estimate using several sources of information.
First, the Draft MOVES2004 model contains an estimate of average trip length in-use.8
These estimates are shown in Table III.A-6.
Table III.A-6. Trip and Start Related Information in Draft MOVES2004
Vehicle Class
Passenger Cars
Light Trucks <
6000 pounds
Light Trucks >
6000 pounds
Starts per Day
Weekday
7.28
8.06
8.06
Weekend
5.41
5.68
5.68
Average Day
6.75
7.38
7.38
Miles per
Day
29.48
35.29
34.08
Average
Trip
Length
4.37
4.78
4.62
VMT: 2003
(billion miles)
1,661
670
251
The estimates of starts and miles per day are based on the results of instrumenting 168
vehicles in Baltimore and Spokane in 1992. The estimates of starts per day for an average day
were again determined by weighting the starts per day for weekdays by five-sevenths and for
weekends by two-sevenths and summing. The estimates of miles per day were also taken from
the Draft MOVES2004 model.9 Average trip lengths (i.e., miles per start) were determined by
dividing miles per day for each vehicle class by the number of starts per day for the average day.
44
-------
The number of miles per start across all three vehicle classes was determined by dividing
total VMT by the total number of starts. From FHWA Highway Statistics 2003, total VMT for
LDVs in 2003 was 1,661 billion miles.10 Total VMT for 2-axle, 4 wheel trucks in 2003 was 998
billion miles. This latter vehicle class includes trucks over 8500 pounds gross vehicle weight
rating (GVWR), which are not included within EPA's definition of LDTs. Based on an analysis
by Oak Ridge National Laboratory, 92.34% of the VMT by 2-axle, 4-wheel trucks was by
vehicles below 8500 pounds GVWR, or LDTs.n Thus, LOT VMT in 2003 is estimated as 923
billion miles. Based on this same study, 72.8% of LDT VMT is by LDTls and LDT2s and
27.2% by LDT3s and LDT4s. Using this breakdown, we estimate LDT1 and LDT2 VMT in
2003 as 670 billion miles and LDT3 and LDT4 VMT as 251 billion miles. Total VMT is then
2,585 billion miles. The total number of starts was estimated by dividing the total VMT for each
class by the number of miles per start for each class and then summing. The result was one start
for every 4.49 miles, or 0.223 starts per mile.
As mentioned above, the estimates of starts and miles per day came from vehicles
operating in the Baltimore and Spokane areas. Therefore, most of this operation was likely
urban (not to be confused with "city" when defining driving for fuel economy labeling
purposes). These studies were performed along with several others in the early 1990's when
EPA was developing the Supplemental FTP rule, which developed and implemented the US06
and SC03 test cycles.12 In addition to Baltimore and Spokane, vehicle operational information
was obtained in Atlanta and Los Angeles. The data obtained in Baltimore and Spokane received
primary focus, as the vehicles were recruited from centralized inspection and maintenance
stations and the study involved instrumented vehicles. The data from Atlanta involved
instrumented vehicles, but the vehicles were recruited via phone contact. While the initial
vehicle selection was random, the rate of people declining to participate was higher than at the
centralized inspection and maintenance stations. Thus, there is slightly more concern that the
final vehicle sample in Atlanta may not be as representative of onroad driving as those in
Baltimore and Spokane. The findings from these studies are shown in Table III.A-7.
Table III.A-7. Estimates of In-Use Average Trip Length
Location
Baltimore - Exeter
Baltimore - Rossville
Baltimore - Combined
Spokane
Atlanta
Los Angeles
Average Trip Length (miles)
Instrumented Vehicle Studies
4.0
5.9
4.9
3.6
6.0
Not Available
Chase Car Studies
Not Available
Not Available
7.5
5.8
Not Available
7.8
As can be seen from Table III.A-7, there is some variability in average trip length
depending on where people reside. People living in the central city section of Baltimore and in
the less populated city of Spokane take much shorter trips than those living in the outskirts of
Baltimore and the more sprawling city of Atlanta.
45
-------
Table III.A-7 also shows estimates of average trip lengths from chase car studies
conducted at the same time as the instrumented vehicle studies. While chase car studies are
ideally suited to study travel on selected roadways, their weakness is that they are not designed
to follow a vehicle from trip start to trip end. Thus, they utilize other information in order to
piece together trip start and end distances and thus, trip length. As can be seen from Table III. A-
7, estimates of average trip length from chase car studies tend to exceed those of instrumented
vehicle studies. The latter clearly yield a more accurate estimate, due to the instrument's ability
to accurately and precisely determine if the engine is running or not.6 However, consideration of
the chase car studies is useful in that it brings in an additional city, Los Angeles. Based on the
relative trip lengths from the chase car studies, it appears that trips in Los Angeles are just
slightly longer than in Baltimore and probably shorter than those in Atlanta. Thus, or
urban/suburban dwellers, it appears that the average trip length is somewhere between 4 and 6
miles and could be close to 5 miles plus or minus a half mile or so. This estimate includes all
trips by these people, those involving both city and highway driving.
No instrumented vehicle studies have been performed in rural areas. Chase car studies
have recently been performed in several rural California areas. However, chase car studies do
not obtain trip based information, as they do not follow the same vehicle from start to stop. Still,
it is likely that trips by rural dwellers are longer than those living in cities.
A second estimate of average trip length in the U.S. is available from the National
Household Travel Survey (NHTS).13 The NHTS was performed in 2001 by the Department of
Transportation and statistically surveyed approximately 26,000 households in the U.S. This
survey represents the sixth in a series of surveys dating back to 1969. (The name of the survey
has changed a few times and the precise survey methods have varied to some degree.) NHTS
found that the average trip taken using a personal vehicle in the U.S. was 9.8 miles long. While
the average trip length was relatively constant from 1969 to 1990, it has increased roughly one
mile since 1990.14 Thus, the estimates shown in Table III.A-7 above could under-estimate trip
length today to some degree.
As noted in the NHTS, the 9.8 mile estimate excludes very long trips, such as those taken
on vacations, as well as commercial trips, such as those by taxi cabs, police officers, municipal
workers, etc.. These excluded trips could be both shorter and longer than those covered by the
survey. However, the number of commercial vehicle operators far exceeds those on vacation at
any given time. Thus, the exclusions likely bias the survey results upwards (i.e., result in an
over-estimation of national average trip length for all light-duty vehicles and trucks).
As the case with the instrumented vehicle studies, the average trip length of 9.8 miles
includes all driving, both city and highway oriented. NHTS does not attempt to split driving into
city and highway categories. However, unlike the instrumented vehicle studies, which studied
the driving patterns of urban/suburban dwellers, NHTS is intended to cover the entire U.S.
e This is often not the case for recent model hybrid vehicles, whose engines shut off frequently during city
driving. However, as the engine is usually turned off for only seconds or minutes, the added fuel due to the
perceived engine off is negligible both on the road and per our modeling. This issue did not affect the studies being
cited here, as there were no hybrids on the road in the early 1990's.
46
-------
population, both urban and rural. Per the Federal Highway Administration, 60-63% of car and
light truck VMT is urban and suburban and 37-40% is rural. Assuming that urban dwellers drive
primarily in urban areas and rural dwellers do likewise, a mile 5 average trip length for 60-63%
of all VMT translates into an 8 mile trip length for all driving, even if rural dwellers never shut
off their engines. In other words, there appears to be an inconsistency between the two sets of
estimates, with the instrumented vehicle studies yielding a shorter trip length than that which can
be consistent with the estimate from NHTS. An increase in average trip length of roughly one
mile since the time of the instrumented vehicle studies could explain part of this difference.
However, some difference still remains. Obviously, the average trip length for rural driving is
finite, not infinite.
Based on the NHTS survey questionnaire, we believe that the survey could miss brief
stops between the primary trip origination and destination points (e.g., those at gas stations or
convenience stores), as well as extremely short trips (e.g., moving a vehicle out of a driveway to
allow another vehicle to exit, moving from one parking place at a shopping center to another, or
from one shopping center to another just across the street). Using trip information from
instrumented vehicles in Baltimore and Spokane (described in more detail below), about 28% of
all trips fall into one of these two categories. In the NPRM, we assumed that the NHTS survey
missed 28% of the trips, reducing the average trip length 7.66 miles (9.8 miles divided by 1.28).
Since the time of the NPRM, EPA obtained more recent data on the operation of vehicles
in Atlanta from Georgia Tech.15 Georgia Tech has been instrumenting vehicles in the Atlanta
area for some time and gathering operational data. As with the previous instrumented vehicle
studies, this type of study captures all vehicle trips no matter how short. The major difference
between this more recent work and the previous studies is the amount of data being collected.
To date, Georgia Tech has collected data on over 620,000 vehicle trips. In contrast, the previous
instrumented vehicle studies involved about a week's worth of operation for less than a hundred
vehicles per city. Thus, the recent data represents almost two orders of magnitude more data
than was evaluated in all of the instrumented vehicle studies described in Table III.A-7 above,
albeit from one urban area and one of the most sprawling at that.
The average trip length found for these 620,000 trips around Atlanta was 7.25 miles.16
The average number of trips per day was 4.62. This means that the vehicles surveyed in Atlanta
drove 33 miles per day, which is very close to the national average of 34 miles per day (per
MOVES/MOBILE6.2). This is 1.2 miles, or 20% greater than the average trip length of 6.0
miles found in the early 1990's (see Table III.A-7). The series of NHTS have also found a
general increase in trip length nationally of about 1.0 mile between 1990 and 2001.17 Thus, it is
likely that the average trip length in the other cities shown in Table III.A-7 have also increased,
as well, since the early 1990's when the data was collected. If the average trip length for urban
driving was slightly more than 5 miles in the early 1990's, then it is likely more than 6 miles
today.
A comparison of the trip lengths resulting from the instrumented vehicle studies and the
chase car studies shown in Table III.A-7 indicates that the latter clearly over-estimate trip length.
This is likely because chase car studies do not focus on entire vehicle trips, but on vehicle
operation along a specified highway segment. However, there is considerable consistency
47
-------
between the relative trip lengths in Baltimore and Spokane based on the instrumented vehicle
data and the chase car data. Extrapolating this to Los Angeles, average trip length in Los
Angeles appears to be slightly longer than that in Baltimore and considerable longer than that in
Spokane. This is not unexpected, given the relative sizes of these cities and their geographical
constraints. Including the Atlanta and implied Los Angeles estimates with those from Baltimore
and Spokane would produce an average trip length for urban driving of considerably more than
4.49 miles. An overall estimate closer to 5.0 miles would appear to be reasonable for the early
1990's timeframe, given Spokane's relatively unique combination of size and geographical
constraints.
Given the growth in average trip length indicated by the two Atlanta studies and the
series of NHTS, we estimate that average urban trip length today is roughly 6.2 miles. Assuming
an average trip length for urban driving of 6.2 miles, an urban VMT fraction of 0.61, and an
overall national average trip length of 9.8 miles from NHTS, the average trip length of highway
driving is 135 miles. While not impossible, this is still likely too high. Thus, it still appears
likely that the NHTS is missing trips of very short length or combining trips with very short
engine off times, though not to the degree assumed in the NPRM analysis. Assuming that the
NHTS misses 11% of all trips for these reasons reduces the average trip length for all driving to
8.7 miles and that for rural driving to about 24 miles. This 24 mile figure for an average rural
trip is much more reasonable than 135 miles. Therefore, we will update our estimate the national
average trip length to be 8.7 miles, from the estimate of 7.7 miles in the NPRM.
This estimate includes all driving, both city and highway oriented. Start fuel use must be
split between city and highway driving. Neither MOBILE6.2 nor Draft MOVES2004 includes a
direct estimate of the split between city and highway driving as defined for fuel economy
labeling purposes. However, such a split can be derived from the driving patterns contained in
Draft MOVES2004. This derivation is described next.
Table III. A-8 presents the 14 driving cycles which are used in Draft MOVES2004 to
estimate on-road emissions, along with each cycle's average speed.18 We ran the Draft
MOVES2004 model for the entire nation using national default inputs and determined the
percentage of driving time which LDVs and LDTs spend in each type of driving (i.e., driving
cycle). These percentages are shown in the third column of Table III. A-8.
48
-------
Table III.A-8. Inventory Driving Cycles in Draft MOVES2004
Cycle
Low Speed
New York City
LOS EF Non-Freeway
LOS CD Non-Freeway
LOS AB Non-Freeway
LOS G Freeway
LOS F Freeway
LOS E Freeway
LOS D Freeway
LOS AC Freeway
High Speed Freeway 1
High Speed Freeway 2
High Speed Freeway 3
Freeway Ramp
Avg. Speed
(mph)
2.5
7.1
11.6
19.2
24.8
13.1
18.6
30.5
52.9
59.7
63.2
68.2
76
34.6
Time Spent During
On-Road Driving
5.2%
5.8%
16.0%
8.7%
6.2%
0.1%
0.5%
24.7%
17.4%
7.3%
1.9%
2.5%
2.0%
1.7%
Assumed City
Percentage
100
100
100
100
100
100
100
100
0
0
0
0
0
44.9
We assigned driving spent in each driving cycle to either city or highway driving, based
on its average speed. If the cycle's average speed was less than 45 mph, we assumed that it
represented city driving. If the cycle's average speed was greater than 45 mph, we assumed that
it represented highway driving. The only exception was driving on freeway ramps. Since
freeway driving occurs in both city and highway modes, we assumed that driving on ramps also
occurs in both modes. No information exists concerning the use of ramps to access freeways
with respect the average speed of the freeway driving at the time. Therefore, we assigned ramp
driving to city and highway driving simply in proportion to the percentage of time spent in city
freeway driving and highway freeway driving. Therefore, the city percentage of ramp driving
was the driving percentages for the LOS E, F, and G freeway cycles (25.3%) divided by the
driving percentages for all eight freeway cycles (56.3%), or 44.9%.
In order to calculate the percentage of VMT occurring during city and highway driving,
we multiplied the average speed of each cycle by its driving time percentage. We then summed
this product of speed and percentage time across all fourteen cycles (31.77 mph). We then
multiplied each cycle's product of speed and percentage time by its city percentage and summed
again (13.51 mph). The percentage of VMT occurring as city driving is the ratio of these two
numbers (13.51/31.77), or 42.6%. We repeated this procedure using the percentage of highway
driving (equal to 100% minus the percentage of city driving). The sum of the product of speed,
percentage time and highway percentage was 18.26 mph. Thus, the percentage of highway
driving is the ratio of 18.26 to 31.77, or 57.4% of VMT.
This city/highway VMT split of 43/57 is quite different from the current 55/45 split.
Some analysts have suggested that the 55/45 split should be updated to reflect an even higher
city fraction, such as 60/40 to 63/37. The current 55/45 split and the suggested updates are based
49
-------
on FHWA estimates of the urban and rural fractions of national VMT. This assumes that city
driving is equivalent to urban driving and highway driving is equivalent to rural driving. There
is some merit to this approach for the current city and highway label fuel economy estimates, as
the original Los Angeles Road Route No. 4 (the basis for the LA-4 driving cycle included in the
FTP) includes some freeway driving which reaches 55 mph. Most of the driving used to develop
the HFET occurred on uncongested rural highways.
Generally, we can expect trips dominated by low speed driving to usually involve more
starts per mile than high speed trips. The current highway driving cycle, the HFET, contains no
cold starts (i.e., infinite trip length). Of course, the 22% adjustment factor applied to the FIFET
fuel economy to calculate the highway fuel economy label value could account for some cold
starts, but no specific trip length is specified. Thus, for simplicity purposes, we assigned an
average trip length of 60 miles to highway driving. This is approximate. However, once trip
length is over 30-40 miles, start fuel use has a very small effect on fuel economy. Still, a finite
trip length for highway driving helps the public to relate to the way highway fuel economy is
estimated. Back calculating using the city and highway VMT split of 43/57 and an overall trip
length of 8.7 miles, the average trip length for city driving is 4.1 miles/ This estimate is 17%
longer than the 3.5 mile city trip length estimated in the NPRM.
By itself, this increase in trip length increases 5-cycle city fuel economy for the 615
vehicles in our certification database by 1%. Highway fuel economy remains unchanged, since
the average trip length remains at 60 miles. However, this higher 5-cycle city fuel economy will
indirectly cause the non-dynamometer adjustment factor to be increased by roughly 0.5% (see
Section III.A.5). This factor applies to both city and highway fuel economy. Thus, the net effect
of increasing the average trip length of city driving is about a 0.5% increase in 5-cycle city fuel
economy and a 0.5% decrease in highway fuel economy.
As it will be useful later in this section, using the information shown in Table III.A-8, we
calculated the average speed of city and highway driving. The average speed of city driving is
the sum of the product of speed, percentage time and city driving percentage (13.51 mph),
divided by the total percentage of city driving (68.0%), or 19.9 mph. The average speed of
highway driving is the sum of the product of speed, percentage time and highway driving
percentage (18.26 mph), divided by the total percentage of city driving (32.0%), or 57.1 mph.
The average speed of all driving represented in Draft MOVES2004 is 31.8 mph.
c. Formula for Start Fuel Use
The total fuel usage in either Bag 1 or Bag 3 can be determined by dividing the number
of miles of driving during these portions of the test (3.59 miles for either bag) by the fuel
economy measured during that bag. Thus, the equation for start fuel use at either 20°F or 75°F is
as follows:
f Given 100 miles of total VMT, 42.6 miles consists of city driving and 57.4 miles of highway driving. At
3.5 miles per city trip, this means 12.14 city trips. At 60 miles per highway trip, this means 0.96 highway trips.
The total number of trips is therefore 13.1, for an overall trip length of 7.66 miles (100/13.1).
50
-------
StartFuel =3.6x1
X
Bag\FEx Bag3FEx
where x is either 20°F or 75°F.
The equation for start fuel use in terms of gallons per mile is:
For City Fuel Economy:
e, ^n< „ .,\ „„ ((0.76xStartFuel75)+(0.24
StartFC (gallons per mile) = 0.33 x - ——-
For Highway Fuel Economy:
e, ^t n .j\ „„ ((0.16xStartFuel15)+(0.24xStartFuel20)~]
StartFC (gallons per mile) = 0.33 x —— —
V 60 )
2. Running Fuel Use at 75°F Without Air Conditioning
Running fuel use depends primarily on how the vehicle is driven, particularly the
distribution of speed and power. In this section, we develop a description of onroad driving from
the Draft MOVES2004 model and then represent this onroad driving using the available
dynamometer tests.
a. On-Road Driving Patterns
On-road driving patterns have been studied for the purpose of emissions since at least the
late 1960's. The driving cycle contained in the FTP, the LA-4, was developed from following
typical vehicle operation over a particular road route in Los Angeles in the late 1960's. The
FIFET was based on instrumented vehicles operated over a rural road route outside of Ann
Arbor, Michigan in the late 1970's. The US06 and SC03 cycles, among others, were developed
in the early 1990's to augment the LA-4 and HFET cycles. These cycles were based on
instrumented vehicles monitored during normal operation in Baltimore, Spokane and Atlanta,
which were already discussed in the section on start fuel use above.
Based on the driving data obtained in Baltimore, Spokane and Atlanta, EPA developed
three cycles which together did a reasonable job of representing the complete breadth of urban
driving: SC03, REP05, and REM01. The SC03 cycle represented driving immediately following
engine start-up. REP05 represented high speed and aggressive driving. REM01 represented all
other driving. Ignoring changes in driving habits since the early 1990's, on-road fuel economy
(at least in urban areas) could be reasonably represented by the fuel economies measured over
these three cycles. However, tests over the REP05 and REM01 cycles are not regularly
performed during certification. Performing these tests would entail additional testing costs. The
SC03 test is only performed with the air conditioning on, so it too would need to be re-performed
with the air conditioning turned off. We are primarily interested here in approaches to estimating
51
-------
on-road fuel economy using the current dynamometer tests. Thus, using REP05, SC03 and
REM01 to project onroad fuel economy is not a practical option.
The US06 cycle was developed as a concentrated version of the REP05 cycle. Thus, it
can conceivably be used in lieu of REP05. The REM01 cycle is similar to the LA-4 cycle which
comprises the driving in the FTP. SC03 is only performed with the air conditioning turned on.
Thus, it cannot be used to estimate fuel use without air conditioning. SC03 includes higher rates
of acceleration than the FTP. However, the US06 cycle includes some low speed driving with
higher acceleration rates. Thus, with correct weighting, the US06 test and the warmed up portion
of the FTP test should be able to represent much or most of the driving contained in the SC03,
REP05 and REM01 cycles.
Our recent testing of driving in Kansas City indicates the need to include the US06
driving pattern in our estimation of fuel economy. This test program and our processing of the
data collected are described in detail in Appendix A. We grouped the vehicle operation
monitored in Kansas City into combinations of vehicle speed and acceleration and compared it to
similar combinations represented in the FTP, HFET and US06 cycles. The breadth or envelope
of operation over the FTP and FIFET cycles is shown as the innermost area (colored in dark blue
in the following figure). The envelope of vehicle operation monitored in Kansas City which
exceeds that of the FTP and FIFET is shown in purple. Finally, the envelope of the US06 cycle
is shown where it exceeds that of the vehicle operation monitored in Kansas City.
Figure III-l. Speed-Acceleration Frequency Distribution: Kansas City Vs. Test Cycles
0 I 5 I 10 I 15 I 20 I 25 I 30 I 35 I 40 I 45 I 50 I 55 I 60 I 65 I 70 I 75 I 80 I 85 I 90
Driving frequency covered by FTP/HFET style driving
Driving frequency covered by Kansas City Real-World Driving
Driving frequency covered by US06 style driving
KC activity cut off < 0.1 %
18% of KC driving outside FTP/HFET driving envelope
0.6% of KC driving is outside US06 driving envelope
90% of US06 driving is within the 0.1% KC boundary
Overall, 18% of the onroad driving activity (time based) in Kansas City fell outside of the
FTP/FIFET envelope. This corresponds to 33% in VMT terms. As can be seen, most of this
operation which exceeds the FTP/HFET envelope has either a higher rate of acceleration or
52
-------
higher vehicle speed. However, only 0.6% of the Kansas City operation fell outside the US06
(0.4%oftheVMT).
The SAFD envelopes for hybrid and conventional vehicles did not differ significantly
from each other. However, the percentages of driving in the various bins did vary, as will
become more evident below when we evaluate the VSP frequency distributions for the two types
of vehicles.
Recent chase car studies of driving in California show even more operation outside of the
FTP/HFET envelope. (The details of these test programs are also discussed in Appendix A.)
The combinations of speed and acceleration for the monitored California urban driving are
shown in the following figure.
Figure III-2. Speed-Acceleration Frequency Distribution: Urban California Vs. Test
Cycles
SPEED BIN
(mph)
15
20 | 25
30
35 | 40
45
50
55 | 60
65
70 | 75
80
85
90
KEY
Driving frequency covered by FTP/HFET style driving
Driving frequency covered by real-world California urban style drivin
Driving frequency covered by US06
Driving frequency covered by real-world CA but NOT US06
20% of CA URBAN driving is outside FTP/HFET driving envelope
1.2% of CA URBAN driving is outside US06 driving envelope
41% of CA RURAL driving is outside FTP/HFET driving envelope
3% of CA RURAL driving is outside US06 driving envelope
As can be seen, the breadth of California urban driving which exceeds the FTP/HFET
envelope is much more expansive than that found in Kansas City. Onroad driving includes much
higher speeds and both higher and lower rates of acceleration. Overall, 20% of California urban
driving lies outside the FTP/HFET envelope (34% of the VMT). Just over 1% fell outside the
US06 envelope (1.3% of the VMT). Rural driving was found to be more aggressive than urban
driving. While not shown in the above figure, 3% of the rural driving was outside the US06
envelope.
Both the MOBILE6.2 and Draft MOVES2004 models contain estimates of on-road
driving. MOBILE6.2 describes driving in a more traditional fashion by assigning VMT to
53
-------
various driving cycles, like those listed in Table III. A-8 above. This is very acceptable for
estimating emission inventories. However, it is not straightforward to convert driving over such
cycles to driving over the five available dynamometer cycles or bags. Average speed and power
are available for each cycle. However, with only two degrees of freedom, one cannot determine
weightings for five dynamometer cycles. A more detailed distribution of speed and power is
necessary. This brings us to the approach taken in Draft MOVES2004, which is more amenable
to the task faced here.
Draft MOVES2004 takes a much different approach to describing on-road driving than
previous EPA emission inventory models, for example, MOBILE6.2. While starting with whole
driving cycles, like those listed in Table III. A-8 above, it goes further by breaking down vehicle
operation on a second by second basis into 17 categories or bins. One bin (Bin 0) contains
significant decelerations. Another bin (Bin 1) contains idling operation. The other 15 bins
contain briefer modest decelerations, cruising operation and accelerations. The 15 bins are
broken down into three sets of bins by vehicle speed: Bins 11-16 contain operation at 1-25 mph,
Bins 21-26 contain operation at 25-50 mph, Bins 33-36 contain operation at 51 mph or faster.
These three sets of bins are further sub-divided according to the power required of the engine
divided by vehicle mass. This ratio is termed vehicle specific power, or VSP, and has the units
of kilowatt per megagram (kW/Mg). The VSP bins are described in Table III.A-9.
Table III.A-9. VSP-Speed Bins in Draft 2004MOVES
Bin Label
MOVES
0
1
11
12
13
14
15
16
21
22
23
24
25
26
33
35
36
Vehicle Speed (mph)
Deceleration
Idle
1-25
25-50
>50
Vehicle Specific
Power (kW/Mg)
—
—
<0
0-3
3-6
6-9
9-12
>12
<0
0-3
3-6
6-9
9-12
>12
<6
6-12
>12
54
-------
The three bins with open ended power levels (bins 16, 26, and 36) have lower limits of 12
kW/Mg. Power during onroad driving can exceed 50 kW/Mg. Nearly 35% of the US06 cycle
falls into bin 36. Therefore, it is useful to split these bins up further into smaller power
increments. This is being considered for the next version of the MOVES model.
Here, we expanded the set of 17 VSP bins by splitting bins 16, 26, and 36 into four bins,
for a net increase of 9 bins. The three speed ranges stay the same. However, instead of bin x6
(i.e., 16, 26, and 36) including all power levels above 12 kW/Mg, bin x6, x7, and x8 will all
include a range in power of 3 kW/Mg, while bin x9 will be open ended. The expanded set of
VSP bins is shown along with that from Draft MOVES2004 in Table III. A-10.
Table III.A-10. Expanded Set of 26 VSP-Speed Bins
Bin Label
MOVES
0
1
11
12
13
14
15
16
21
22
23
24
25
26
33
35
36
Expanded Set
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
Vehicle Speed
(mph)
Deceleration
Idle
1-25
25-50
>50
Vehicle Specific
Power (kW/Mg)
—
—
<0
0-3
3-6
6-9
9-12
12-15
15-18
18-21
>21
<0
0-3
3-6
6-9
9-12
12-15
15-18
18-21
>21
<6
6-12
12-15
15-18
18-21
>21
55
-------
Now that the VSP concept has been introduced, it will be easier to understand how Draft
MOVES2004 describes on-road driving, which is accomplished in three steps. First, a VSP
frequency distribution is developed for each of the fourteen inventory cycles listed in Table
III.A-8 above. Most of these inventory cycles were developed from the on-road driving data
obtained in Baltimore, Spokane and Atlanta in the early 1980's. The three highest speed freeway
cycles have been added more recently to represent driving at speeds higher than those typical of
this earlier timeframe, since the highest speed limit during the time of these studies was 55 mph.
Each VSP distribution shows the percentage of time spent driving in each of the 26 VSP bins.
Different vehicles will produce slightly different VSP distributions. However, because VSP is
defined as the ratio of required power to vehicle weight, the differences in VSP across various
vehicles are relatively small for a given driving pattern. Draft MOVES2004 includes VSP
distributions for typical LDVs and LDTs. We have combined these two sets of VSP
distributions using a mix of 50% cars and 50% light trucks, which represents the split of onroad
VMT in calendar year 2004 from MOBILE6.2. These LDV/LDT weighted VSP distributions
are shown in Tables III.A-11 and III. A-12. (LOS stands for level of service, or volume of traffic.
LOS A involves the least volume of traffic. LOS G is the most congested.)
Table III.A-11. VSP Frequency Distributions for Onroad driving Cycles in MOVES
VSP
Bin
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
Low
Speed
2.2%
49.7%
15.3%
32.6%
0.3%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
New York
City
1 1 .0%
41 .7%
13.1%
17.7%
7.9%
2.5%
2.2%
0.8%
0.3%
0.5%
0.0%
0.2%
0.4%
1 .2%
0.2%
0.2%
0.2%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
LOSEF
Non-Freeway
14.3%
33.9%
7.5%
13.9%
6.0%
3.3%
2.5%
1.1%
0.0%
0.0%
0.0%
4.1%
2.1%
3.9%
2.7%
2.5%
0.9%
1 .2%
0.0%
0.2%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
LOS CD
Non-Freeway
14.5%
22.4%
7.3%
7.5%
4.3%
4.4%
2.4%
0.8%
0.2%
0.3%
0.0%
7.7%
7.5%
5.5%
6.0%
4.1%
3.1%
1 .3%
0.5%
0.2%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
LOSAB
Non-Freeway
13.0%
15.1%
3.8%
6.4%
4.2%
2.8%
2.9%
0.9%
0.3%
0.1%
0.0%
8.7%
7.1%
10.3%
8.3%
3.8%
2.4%
0.9%
0.8%
0.3%
2.8%
3.2%
1 .6%
0.4%
0.0%
0.0%
LOSG
Freeway
8.5%
4.4%
29.2%
34.0%
1 1 .4%
3.8%
1 .3%
0.0%
0.0%
0.0%
0.0%
0.5%
0.8%
1 .8%
2.1%
2.1%
0.3%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
LOSF
Freeway
1 1 .3%
3.6%
16.5%
23.5%
1 1 .5%
4.1%
1 .2%
1 .8%
0.5%
0.0%
0.0%
5.3%
5.4%
4.1%
3.2%
3.7%
2.3%
1 .2%
0.5%
0.3%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
56
-------
Table III.A-12. VSP Frequency Distributions for Onroad driving Cycles in MOVES
VSP Bin
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
LOSE
Freeway
1 1 .2%
1 .8%
7.8%
10.5%
7.5%
4.6%
2.1%
0.1%
0.0%
0.0%
0.0%
7.6%
1 1 .0%
5.3%
5.4%
3.8%
3.4%
0.5%
0.7%
0.4%
4.2%
5.8%
2.1%
2.5%
1 .2%
0.4%
LOSD
Freeway
5.2%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
4.7%
4.9%
5.4%
4.8%
4.3%
3.9%
1 .3%
0.5%
1 .0%
17.4%
18.1%
9.1%
7.2%
5.8%
6.4%
LOS AC
Freeway
1 .4%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.2%
0.5%
0.6%
0.5%
0.5%
0.5%
0.5%
0.1%
0.8%
23.2%
30.2%
10.7%
12.8%
9.1%
8.4%
High Speed
Freeway 1
2.1%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
18.5%
29.9%
13.9%
12.0%
10.5%
12.9%
High Speed
Freeway 2
3.3%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
9.6%
22.2%
15.0%
18.0%
12.2%
19.6%
High Speed
Freeway 3
0.9%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
7.4%
14.9%
7.3%
1 1 .9%
18.6%
39.0%
Freeway
Ramp
10.5%
6.8%
4.5%
4.1%
4.5%
1 .9%
1.1%
0.4%
0.8%
0.0%
0.0%
8.3%
4.4%
4.1%
4.2%
6.4%
5.3%
4.1%
3.3%
5.5%
9.2%
3.2%
2.6%
0.8%
2.1%
2.0%
Second, a distribution of average speeds by hour of the day and road segment for the area
of interest are developed. In urban areas, these estimates come from travel demand models for
five specific urban areas (Ada County, ID, Charlotte, NC, Chicago, IL, Houston, TX, and New
York, NY).19 Travel in other urban areas is assigned to one of the five modeled areas. In rural
areas, these estimates come from chase car data recently obtained in California. Based on
average speed and roadway type, travel is assigned to two of the fourteen inventory cycles (listed
in Table III.A-8 above). The two cycles selected are those which have average speeds which
most closely match the average speed of that roadway type during that hour of the day. Cycles
are chosen which also represent driving on the same roadway type, if they are available.8 The
weighting of the two cycles are determined so that the average speed in-use is matched exactly.
Then, the VSP distributions of these two cycles are combined using these same weightings.
Third, these combinations of VSP distributions for the various cycles are aggregated over
the geographical and temporal intervals of interest, weighting each by the amount of driving
8 Sometimes, higher speed driving on non-freeways in rural areas exceeds the average speed of the
available non-freeway driving cycles. In this case, freeway cycles are used.
57
-------
occurring over each roadway segment and time interval. Here, we are concerned with national,
annual average driving. Thus, Draft MOVES2004 was run for the nation as a whole for an entire
year and the distribution of driving over the various inventory cycles was output.
Table III.A-13 shows the resulting distribution of on-road driving into the 14 inventory
cycles. Three distributions are shown. The first distribution represents all U.S. driving, which
was the direct output from Draft MOVES2004. The second represents city driving. In this case,
the all U.S. driving percentages were multiplied by the city percentages shown in Table III.A-8
above and normalized to sum to 100%. The third shows the distribution for highway driving,
using 100% minus the city percentages shown in Table III.A-8 above. Also shown is the average
speed for each cycle. Weighting the average speed of each cycle by its percentage of driving
time also yields the average speed of that type of driving (e.g., city). These speeds are shown in
the last row of the table.
Table III.A-13. Distribution of Onroad Driving Patterns: Draft MOVES2004
Inventory
Cycle
Low Speed 1
New York City
LOS EF Non-Freeway
LOS CD Non-Freeway
LOS AB Non-Freeway
LOS G Freeway
LOS F Freeway
LOS E Freeway
LOS D Freeway
LOS AC Freeway
High Speed Freeway 1
High Speed Freeway 2
High Speed Freeway 3
Freeway Ramp
Average Speed (mph)
Average Speed
(mph)
2.5
7.1
11.6
19.2
24.8
13.1
18.6
30.5
52.9
59.7
63.2
68.2
76.0
34.6
—
Distribution of Driving Time
All Driving
5.2%
5.8%
16.0%
8.7%
6.2%
0.1%
0.5%
24.7%
17.4%
7.3%
1.9%
2.5%
2.0%
1.7%
19.9
City Driving
7.7%
8.6%
23.5%
12.8%
9.1%
0.2%
0.7%
36.3%
0.0%
0.0%
0.0%
0.0%
0.0%
1.1%
57.1
Highway Driving
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
54.5%
22.8%
5.8%
7.7%
6.4%
2.9%
31.8
The average speed of city driving, 19.9 mph, is just slightly higher than that of the FTP,
19.6 mph. The average speed of highway driving is well above that of both the HFET and US06
cycles, but slightly below the highway portion of the US06 cycle. As described in section III. A. 1
above, according to this methodology, the percentage of national VMT that is like city driving is
42.6% and that which is like highway driving is 57.4%. As also mentioned in section III.A.I
above, we evaluate two alternatives which assign portions of the driving over the LOS D
Freeway to city driving in section III.E.2 below. These two options increase the city percentage
of national VMT to 50% and 55%, respectively.
58
-------
We then weighted the VSP distributions of each inventory cycle (from Table III. A-11 and
III.A-12) by the percentage of driving represented by each cycle (from Table III.A-13). This
produced VSP distributions for all U.S. driving, city driving and highway driving. These three
VSP distributions are shown in Table III.A-14.
Table III.A-14. VSP Distributions for U.S. Driving (% of time)
VSP Bin
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
City
1 1 .8%
20.4%
8.3%
13.0%
6.0%
3.5%
2.2%
0.6%
0.1%
0.1%
0.0%
5.5%
6.2%
4.7%
4.2%
3.0%
2.2%
0.8%
0.4%
0.3%
1 .9%
2.3%
1.1%
0.9%
0.5%
0.2%
Highway
3.9%
0.2%
0.1%
0.1%
0.1%
0.1%
0.0%
0.0%
0.0%
0.0%
0.0%
2.8%
2.9%
3.2%
2.9%
2.6%
2.4%
0.9%
0.5%
0.9%
16.7%
21 .0%
10.1%
9.8%
8.0%
10.7%
All U.S.
9.2%
13.9%
5.7%
8.8%
4.1%
2.4%
1 .5%
0.4%
0.1%
0.1%
0.0%
4.6%
5.2%
4.2%
3.8%
2.9%
2.2%
0.9%
0.4%
0.5%
6.6%
8.3%
3.9%
3.7%
2.9%
3.6%
These VSP distributions represent national average city and highway driving in the U.S.
The next step is to determine what combination of the five dynamometer cycles best matches
each VSP distribution.
The following figure compares the VSP distribution of vehicle operation monitored in
Kansas City to that in Draft MOVES2004.
59
-------
Figure III-3. Kansas City and California VSP Frequency Distributions vs. MOVES
0 16 -^——^^
0 14
019
0-1
-------
Figure III-4. VSP Frequency Distributions in Kansas City: Hybrids vs. Non-Hybrids
0 2 -i^^^— ^^^~
0 18
0 16
0 14
019
-------
Table III.A-l5. Driving Characteristics of the Current Dynamometer Tests
Cycle
FTP (Bags 2 and 3)
FTP: Bag 3
FTP: Bag 2
HFET
US06
US06: City Bag
US06: Highway Bag
SC03 (run with air conditioning on)
Cold Temperature FTP (same driving cycle as FTP)
Average Speed
(mph)
19.6
25.6
16.1
48.2
48.0
21.7
61.2
21.5
19.6
Average Power*
(mph2 per second)
40.9
53.6
33.8
34.9
104.3
152.9
78.2
49.2
40.9
* Power defined as velocity times the change in velocity per second during cruise or accelerations. Power is set
equal to zero during decelerations and not considered in the determination of average power.
The FTP and the cold temperature FTP both involve the same driving cycle, just at
different ambient temperatures. Thus, their average speeds and power are identical, both for the
total cycle and for each bag of emissions measured. The FTP and SC03 involve distinct, but
similar driving cycles. Both are low speed cycles having similar average speeds and power
levels. As the SC03 test is only run with the air conditioning on and all the other tests are run
with air conditioning off, it is not possible to isolate the effect of the driving cycle differences
between the FTP and SC03 tests directly. Thus, this leaves five distinct driving patterns which
can be used to represent typical U.S. driving: Bag 2 of the FTP, Bag 3 of the FTP, HFET, City
Bag of US06 and Highway Bag of US06.
As shown in Table III. A-15, both Bags 2 and 3 of the FTP are low speed cycles, but their
average power requirements differ by a factor of 1.7. As will be seen below, it may be useful to
consider each bag separately in simulating typical city and highway driving.
The current US06 test currently consists of 600 seconds of driving and the emissions are
collected in one bag (i.e., one single collection of pollutants emitted during the test). Thus, the
fuel economy is measured over the entire cycle. The US06 driving cycle consists of 5 hills, or 5
driving segments which begin and end with the vehicle at idle.11 The first hill of the cycle peaks
at 44.2 mph, while the last three hills peak at 28.5, 28.0 and 51.6 mph. All of these hills are also
relatively short in duration. These hills are indicative of city like driving. The second and third
hills peak at 70.6 and 80.3 mph, which are more indicative of highway driving. The second hill
is relatively short (roughly 90 seconds), while the third hill comprises most of the US06 test
(roughly 360 seconds).
h A "hill" within a driving cycle is a segment of driving which starts and finishes with the vehicle at rest
(zero speed). The term hill comes from the view of a driving trace where vehicle speed is plotted versus time. As
the vehicle accelerates, its speed increases, causing this trace to climb up a hill. Then as the vehicle decelerates, it
proceeds down the hill.
62
-------
As discussed in the preamble to the proposed rule, we are proposing that two separate
emission measurements be made during the US06 cycle to better estimate city and highway fuel
economy. To best distinguish between city and highway like driving, we would group hills 1, 4,
and 5 into a city bag and hills 2 and 3 into a highway bag. However, the total length of time of
hill 1 is quite short, only 45 seconds. In some cases, emissions can be collected into a bag at the
beginning of a test (i.e., hill 1), emissions then collected into a second bag (i.e., hills 2 and 3),
and then emissions collected into the first bag again (i.e., hills 4 and 5). In this case, there should
be sufficient emission sample in each of the two bags to measure accurately with today's
equipment. However, in some cases, a manufacturer might have to collect the sample in three
bags, with the first bag only containing the emissions from the first hill. We are concerned that
this is not sufficient time to generate enough emissions to measure accurately. Including hill 2 in
the first bag triples the driving time and ensures the ability to accurately measure emissions and
fuel use. Thus, we are proposing to place the second hill into the city portion of the cycle,
essentially separating the measurement of emissions during the third hill from the other four
hills. Bag 1 would consist of hills 1 and 2, Bag 2 would consist of hill 3 and Bag 3 would consist
of hills 4, and 5. While this incorporates some highway like driving into the "city" portion of the
segregated US06 test, as will be shown below, this is still a much improved segregation of city
and highway like driving than the US06 cycle as a whole.
For example, even with the second hill included in the city portion, the average speed of
the city portion of US06 is only 27.7 mph. The average speed of the highway portion of US06 is
61.2 mph. The average speed of the entire cycle is 48.0 mph. Thus, separating the cycle in this
way creates two dramatically different driving cycles, each of which falls much more clearly into
our definitions of city and highway driving than the US06 cycle as a whole. To avoid any
confusion with the bags of the FTP, we will refer to the city and highway portions of the US06
cycle as US06 city and US06 highway. Overall, seconds 0-131 and 496-600 of the cycle would
comprise the city bag and seconds 132-495 would comprise the highway bag. The description of
the hills within US06 and their designation is summarized in the table below.
Table III.A-16. Split of US06 Cycle into City and Highway Portions
Hill
1
2
O
4
5
Portion of Driving Cycle (cumulative seconds)
0-43
44-131
132-495
496-563
564-600
Maximum Speed (mph)
44.2
70.7
80.3
29.8
51.6
Designation
City
City
Highway
City
City
We evaluate the impact of a more ideal separation of US06 into city and highway driving in
section III.E below. There, the US06 highway bag contains hills 2 and 3 and the US06 city bag
contains hills 1, 4 and 5.
With the split of US06 into two bags, we have available fuel economy estimates for five
distinct driving patterns:
1) Bags 1 and 3 of the FTP;
2) Bag 2 of the FTP;
63
-------
3) HFET;
4) City portion of US06; and
5) Highway portion of US06.
We propose to combine the results of these five tests to represent typical city and highway
driving patterns.
The VSP distributions for the four complete dynamometer cycles plus individual bags of
the FTP and US06 cycles are shown in Table III.A-17. As was the case for the Draft
MOVES2004 inventory cycles, these VSP distributions represent a 50-50 mix of cars and light
trucks.
Table III.A-17. VSP Distributions for Dynamometer Cycles (% of time)
BinID
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
LA4
12.0%
18.6%
6.4%
10.6%
8.0%
5.0%
2.3%
0.7%
0.1%
0.0%
0.0%
4.4%
10.9%
9.5%
2.9%
1 .5%
0.7%
0.2%
0.3%
0.2%
1 .6%
2.6%
1.1%
0.1%
0.2%
0.0%
HFET
3.5%
0.7%
0.0%
0.1%
0.3%
0.1%
0.5%
0.1%
0.0%
0.0%
0.0%
4.1%
3.6%
12.4%
17.3%
7.5%
2.5%
0.4%
0.3%
0.0%
6.1%
32.3%
6.0%
2.0%
0.1%
0.0%
US06
16.8%
7.5%
0.7%
0.8%
0.3%
0.5%
0.2%
1 .0%
0.8%
0.5%
1 .7%
1 .4%
0.7%
0.8%
0.1%
0.4%
0.2%
0.7%
1 .2%
4.5%
7.8%
14.1%
8.6%
10.3%
6.6%
12.0%
US06
City
32.6%
14.3%
1 .3%
1 .7%
0.9%
1 .3%
0.4%
2.2%
2.2%
1 .3%
3.5%
3.3%
1 .5%
1 .7%
0.2%
1.1%
0.4%
1 .5%
2.4%
9.1%
4.6%
3.7%
0.4%
0.9%
0.9%
6.5%
US06
Hwy
7.1%
2.5%
0.3%
0.3%
0.0%
0.0%
0.0%
0.3%
0.0%
0.0%
0.5%
0.3%
0.1%
0.1%
0.0%
0.0%
0.0%
0.1%
0.4%
1 .6%
9.9%
20.7%
13.8%
16.2%
10.2%
15.5%
SC03
10.5%
19.7%
3.7%
9.2%
5.0%
2.4%
2.0%
2.1%
0.0%
0.0%
0.0%
7.7%
9.6%
10.3%
6.6%
3.9%
4.0%
0.0%
0.0%
0.0%
1 .8%
0.6%
0.5%
0.4%
0.0%
0.2%
Bag 3
13.3%
18.8%
2.8%
4.4%
3.0%
2.4%
2.9%
1 .7%
0.4%
0.0%
0.0%
4.6%
7.3%
12.4%
4.5%
2.7%
2.0%
0.6%
0.9%
0.6%
4.3%
7.1%
2.9%
0.3%
0.5%
0.0%
Bag 2
1 1 .2%
18.5%
8.5%
14.3%
10.9%
6.5%
1 .9%
0.2%
0.0%
0.0%
0.0%
4.3%
13.0%
7.9%
2.0%
0.8%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
We then performed a set of linear regressions of the VSP distributions of the various
dynamometer cycles against the city and highway VSP distributions. To maximize the ability of
the dynamometer cycles to predict on-road fuel economy, we weighted the squared error in each
VSP bin by its average rate of fuel consumption in each bin. We could not use the Draft
MOVES2004 fuel rates directly, since they only exist for 17 VSP bins. Given this, we
64
-------
developed several sets of 26-bin VSP fuel rates, as each approach had its relative strengths and
weaknesses.
One set of fuel rates was based on the results of an EPA study performed on 15 cars in
2001.20 This study instrumented 15 passenger cars with portable emission measurement devices
and measured vehicle activity and fuel economy on a second by second basis. The strength of
this study is that it consisted of onroad testing, though the selection of vehicles and drivers were
not necessarily representative of those in the U.S. The weakness is that only cars were tested; no
light trucks.
Another set of fuel rates was taken from a recent testing program conducted in Kansas
City. This program is described in detail in Appendix A. We used the average fuel rates for the
63 non-hybrid vehicles tested in the program, 30 cars and 33 light trucks.
A final set of fuel rates was developed by extrapolating the Draft MOVES2004 fuel rates
(average of a 50/50 mix of cars and light trucks). The fuel rates for the x5 bins and those of
lower power were takes directly from those in Draft MOVES2004. The x6, x7, x8 and x9 bins
were determined by multiplying the x5 bin fuel rates from Draft MOVES2004 by the ratios of
the fuel rates in the higher power bins to the fuel rates of the analogous x5 bins from the EPA 15
car study.
These four alternative sets of fuel rates are shown in Table III.A-18.
65
-------
Table III.A-18.26-Bin VSP Fuel Rates (gram per second)
VSP Bin
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
EPA 15 Car
0.295
0.286
0.382
0.610
0.911
1.223
1.516
1.897
2.161
2.591
3.447
0.457
0.637
0.857
1.040
1.379
1.709
2.053
2.346
3.686
1.004
1.430
1.813
1.985
2.163
3.315
Kansas City
0.412
0.384
0.454
0.641
1.122
1.406
1.715
2.006
2.172
2.358
2.296
0.593
0.833
0.986
1.180
1.352
1.631
1.917
2.272
2.424
1.069
1.486
1.753
1.937
1.948
2.309
Extrapolated MOVES
EPA 15 Car
0.391
0.326
0.509
0.660
0.978
1.260
1.536
1.797
1.945
2.112
2.056
0.659
0.766
0.971
1.264
1.629
1.965
2.310
2.737
2.921
1.001
1.573
1.856
2.051
2.061
2.444
Kansas City
0.391
0.326
0.509
0.660
0.978
1.260
1.536
1.921
2.189
2.624
3.491
0.659
0.766
0.971
1.264
1.629
2.019
2.426
2.772
4.356
1.001
1.573
1.994
2.183
2.379
3.646
We selected the fuel rates from the Kansas City test program over the other two sets of
fuel rates. The Kansas City fuel rates are based on a mix of cars and light trucks which were
selected randomly from vehicle registrations of 2001 and later model year vehicles. The vehicles
were also driven by their owners in normal operation. The 15 car test program only included
cars, no light trucks. Also, the operators of the vehicles were EPA employees or contractors who
were aware of the purpose of the test program. The fuel rates from Draft MOVES2004 include
data from a large number of cars and light trucks. However, the extrapolation of the fuel rates
for the 12 highest power bins is based on data from the 15 cars. Thus, the fuel rates from the
Kansas City program are the most balanced and representative of actual onroad operation of the
three sets of fuel rates.
We performed the regressions using the regression function in Excel. The intercept was
set to zero. This function does not provide for weighting the residuals. Thus, we incorporated
this weighting by multiplying the VSP frequencies in both Tables III.A-12 and III.A-15 by the
square root of the fuel rate for each VSP bin from the middle column of Table III.A-18. We also
66
-------
performed the regressions in a stepwise fashion. For both city and highway driving, we started
with all five dynamometer driving cycles in the regression. Any driving cycle with a negative
coefficient was dropped from the regression. If two or more cycles had negative coefficients, the
cycle (of this group of cycles with negative coefficients) with the lowest p-value (most
statistically significant) was dropped first. The regression was then rerun. Once all cycles with
negative coefficients were deleted, we continued to drop the cycle with the highest p-value until
the adjusted r-squared value began to decrease. We then selected the set of cycles and their
coefficients which produced the greatest adjusted r-squared value. The final set of coefficients
generally did not sum to 1.0. Therefore, we normalized them to sum to 1.0. The results of the
final regressions for city and highway driving are shown in the first two columns of numbers in
Table III. A-19 below. Cycle coefficients were rounded to the nearest percentage.
Table III.A-19. Best-Fit Combinations of Dynamometer Cycles
Cycles
Coefficients
Bag 3 FTP
Bag 2 FTP
HFET
US06 City
US06 Hwy
Adjusted R-Squared
Time Basis
City
32%
60%
0%
8%
0%
0.6927
Highway
0%
0%
25%
0%
75%
0.8649
All
30%
38%
6%
0%
26%
0.6034
Mileage Basis
City
41%
48%
0%
11%
0%
N/A
Highway
0%
0%
21%
0%
79%
N/A
All
24%
19%
9%
0%
48%
N/A
A number of conclusions can be drawn from Table III.A-19. First, as might be expected,
the two higher speed cycles are absent from the MOVES description of city driving; vice versa
for highway driving. Second, the relative weighting for Bags 2 and 3 of the FTP in the
description of city driving are similar, but not identical to those inherent in the FTP. Third, FTP
driving is indicative of the great majority of the MOVES representation of city driving, nearly
90%. Fourth, the modeling indicates a strong preference to split the low and high speed driving
of US06 into city and highway driving, respectively. In contrast, HFET driving is only
indicative of 21% of the MOVES representation of highway driving. Even more interesting is
the fact that the percentage of average U.S. driving as a whole which is represented by the HFET
is only 9%. Based on the results shown in Table III.A-19, the HFET is the least representative
and presumably contributes the least amount of predictive information regarding onroad fuel
economy, of the three complete cycles. Of the individual bags, the US06 city bag is the least
representative of driving overall. However, separating this low speed aggressive driving out of
US06 appears to help focus the contribution of high speed, aggressive driving over the highway
and overall, given the large contributions assigned to the US06 highway bag.
Because VSP is defined on a second by second basis, the VSP distributions used in the
above regression analyses are on a time basis, not a mileage basis. For example, 60% of the time
spent city driving is like that of Bag 2 of the FTP. However, because this Bag's driving is the
slowest of all the cycles, only 48% of the mileage spent city driving is like Bag 2. Since fuel
economy is most commonly reported on a mileage basis, and not on a time basis, it is useful to
convert the cycle combinations shown in Table III.A-19 above to a mileage basis. This is done
67
-------
by multiplying the percentages shown in Table III. A- 19 by the average speed of the
dynamometer cycle or bag presented in Table III.A-15 above.
The changes in the cycle weights for city and highway are relatively small, because the
average speed of the cycles which predominate the city and highway weights all have similar
(e.g., either low or high) speeds. Most of the percentages on a time basis change significantly
when converted to a mileage basis for all U.S. driving, as every cycle other than US06 city has
either much lower or much higher average speed than the overall U.S. average.
In their comments, Honda expressed concern over the high percentage of highway
driving represented by the US06 highway bag given their impression of the extreme nature of the
US06 highway driving cycle. As discussed in detail in the Response to Comments document, we
used second by second fuel economy values of 80 vehicles tested over the FTP and US06 tests to
estimate fuel economy over the HFET, US06 highway and onroad highway driving per the Draft
MOVES2004 model. We found that the 5-cycle 79%/21% weighting of the US06 and HFET
fuel consumption values predicted onroad highway fuel consumption much more accurately on
an individual vehicle basis than the HFET fuel consumption alone (analogous to the current
highway label formula). Thus, Honda's concern does not appear to be a real problem with the 5-
cycle formulae.
Practically, the fuel economies of the various cycles or bags are combined harmonically,
as is done today with the combination of city and highway fuel economies to estimate the 55/45
composite fuel economy. Mathematically, the formulae for running fuel use without
consideration of air conditioning or cold temperature are:
City Driving:
n • r^ ( °'48 '
RunmngFC = - + - +
{Bag275FEj {Bag375FEj
1 (
+
j {
Highway Driving:
J> • T7^ ( °'21 1 . ( °'79 1
RunmngFC = - + -
{HFETFEJ {USQ6 Highway FEJ
All U.S. Driving:
„ • r^ ( °'24 1 ( °'19 1 ( °-09 1 ( °'48 1
RunmngFC = - + - + - + -
{Bag315FEj {Bag215FEj {HFET FE ) (US06 Highway FE )
For the NPRM, we developed a number of alternatives to these combinations. These
alternatives were developed using: 1) different estimates of fuel consumption as a function of
VSP, 2) whole cycles instead of bags, 3) a different split of the US06 test into city and highway
portions, etc. We received very little comment on these alternatives. No new vehicle activity
68
-------
data has become available since the NPRM. Thus, we are retaining the cycle combinations
shown in Table III.A-19 in the final 5-cycle formulae.
Hybrid vehicles are required to be tested over a full 4-bag FTP. Some manufacturers
prefer to perform this test as a two bag test, with Bags 1 and 2 being combined into a single bag
and Bags 3 and 4 being similarly combined. We developed cycle weighting factors for this case
using the same methodology as described in Section III.A.2 of the Draft Technical Support
Document to the NPRM. The result was a weighting factor of 0.90 for Bag 2 of a 2-Bag FTP
and 0.10 for the US06 City Bag. These factors are only slightly different from a simple
combination of the weighting factors for Bags 2 and 3 of the FTP (i.e., 89%) and US06 city
shown in Table III.A-19 above.
Practically, the fuel economies of the various cycles or bags are combined harmonically,
as is done today with the combination of city and highway fuel economies to estimate the 55/45
composite fuel economy. Mathematically, the formulae for running fuel use without
consideration of air conditioning or cold temperature are:
City Driving:
n . ^ ( 0.48 ^ ( 0.41 ^ 0.11
RunnmgFC = - + - +
{Bag275FEj {Bag375FEj
Highway Driving:
*> • vr ( °-21 1 . ( °-79
RunnmgFC = - + -
[HFETFEJ [usoe Highway FE)
For hybrids tested over a four bag FTP, the fuel consumption measured over Bag 4 can
be substituted for Bag 2 in the city driving equation. The equation for hybrids tested over a two
bag FTP is as follows:
City Driving (Two-Bag FTP):
RunningFC =
0.90
Bag215FE
0.10
\
US 06 City FE)
3. Effect of Air Conditioning on Fuel Economy
The performance of emission controls while the air conditioning system is operating is
assessed via the SC03 test. The SC03 test begins with a hot start (i.e., the engine has been turned
off for 10 minutes after having been fully warmed up prior to engine shutdown). The test cell is
at 95°F and 40% relative humidity, with a solar load of 850 Watts per square meter on the
vehicle. The vehicle is also pre-heated at this solar load for 10 minutes prior to the test, so the air
69
-------
conditioning compressor is generally engaged throughout the entire test. The driving pattern of
the SC03 test is designed to represent driving performed immediately after vehicle start-up, so it
is a relatively low speed cycle.21'22 The driving pattern contained in the SC03 test is similar to
that of the FTP, but not identical.
We estimate the impact of air conditioning operation on fuel economy based on the
difference in fuel use over the SC03 and Bags 2 and 3 of the FTP. The most significant
difference between these two tests from the perspective of fuel economy is the operation of the
air conditioning system. However, differences in the driving pattern between the two tests
should also be considered. Also, the air conditioning system is not always on in-use, so this
needs to be accounted for. In addition, when the air conditioning system is on in-use, the
temperature and relative humidity are not always 95°F and 40%, respectively. Temperature, in
particular, can affect the load applied by the compressor on the engine. Thus, the effect of
different ambient conditions should be considered, as well. Finally, the air conditioning
compressor can also be engaged when the defroster is turned on. Each of these factors will be
assessed below, starting with the difference in driving pattern.
Using the same methodology for modeling fuel use described above, we determined the
combination of Bags 2 and 3 of the FTP and the US06 city cycle which matches the fuel use over
the SC03 cycle with the air conditioning turned off. The adjusted r-squared value was higher
without the US06 city bag than with it included. Therefore, we excluded this cycle from the
final cycle combination. Overall, a combination of 39% of Bag 2 and 61% of Bag 3 on a
mileage basis best represents the speed and power distribution of SC03. Thus, we propose to
estimate the incremental fuel use due to the operation of the air conditioner at 95°F and 40%
relative humidity at an average speed of 21.5 mph as the difference in fuel consumption
measured over the SC03 versus this combination of fuel consumption over Bags 2 and 3 of the
standard FTP. The following equation depicts this mathematically:
Excess fueluse due to air conditioning at 95 F =
(Fueleconomy over the SC03 test) ( 0.39 ^ ( 0.61
(FueleconomyoverBag2) j \ (Fuel economy over Bag 3)
The next factor to address is that of compressor operation. The length of the SC03 test is
10 minutes. Since the vehicle has been sitting at 95°F for some time and has been under a solar
load of 850 Watts per square meter for 10 minutes, the air conditioning compressor is usually
engaged throughout the test. We assume here that the air conditioning compressor is engaged
during 100% of the SC03 test. However, this estimate could be too high. The effect of a lower
estimate will be evaluated in section III.E below.
This is not the case in-use. The air conditioning compressor generally cycles on and off
depending on the ambient temperature, humidity, solar load and the length of time that the
vehicle has been operating. The greater the temperature and humidity, the more often drivers
70
-------
turn the air conditioning on. The greater the temperature and humidity, the more frequent the
compressor needs to operate in order to keep the cabin at a comfortable temperature; the shorter
the trip, the more relevant any solar loading of the vehicle. Thus, it is appropriate to adjust the
excess fuel use determined from the SC03 test to account for the times when the air conditioning
compressor is not engaged.
The Draft MOVES2004 model contains an algorithm which estimates the percentage of
time which the compressor is engaged as a function of heat index. Heat index is a complex
combination of ambient temperature and humidity which was developed to predict degrees of
personal comfort.23 Heat index is used by the National Weather Service to quantify discomfort
caused by the combined effects of temperature and relative humidity. Figure III-5 is reproduced
from the MOBILE6 report and shows how heat index varies with both temperature and humidity.
Figure III-5. Heat Index vs. Temperature and Humidity
160-
£ 140-
8
"F 120-
to
^ 100-
80-
60
7
+ ***.*
*'''\
_ A • *'* m ^
•^^ ^^^*
^0f
^p
^^ 'r
*" ^
t "
••^
«t^ *- - - -^^
^ *'^^^
,
/*
f
y
5 80 85 90 95 100 1(
Rel Hum
80%
60%
40%
D5
Temperature (F)
Note: Heat Index values based on shady conditions
Lines represent curva fit of tabular data
The Draft MOVES2004 algorithm of compressor on fraction versus heat index was
developed from the direct measurement of air conditioning operation of over 1000 trips by 20
vehicles in Phoenix, Arizona during the summer and fall of 1992.24 The algorithm considers
both the frequency that the system is turned on by the driver and the frequency that the
compressor is engaged once the system is turned on. The algorithm is of the form:
AIC compressor on fraction = A + B x Heatlndex + C x Heatlndex2
71
-------
The coefficients vary depending on time of day, basically causing the predicted compressor use
to increase when the sun rises higher in the sky for a given level of heat index. The coefficients
are shown in Table III. A-20 below.
Table III.A-20. Coefficients for A/C Compressor Usage Equations
Heat
Index
<65
65-74
74-76
76-96
96-101
101-104
104-110
>110
7-10 am, 5-9 pm
A
0
-4.101
-2.930
-2.930
-3.632
-3.632
-3.632
B
0
0.0864
0.0591
0.0591
0.0725
0.0725
0.0725
C
0
-0.000367
-0.000213
-0.000213
-0.000276
-0.000276
-0.000276
1 1 am - 4 pm
A
0
-4.101
-4.101
-5.307
-5.307
-5.307
-3.632
B
0
0.0864
0.0864
0.1140
0.1140
0.1140
0.0725
C
0
-0.00037
-0.00037
-0.00052
-0.00052
-0.00052
-0.00028
10 pm - 6 am
A
0
0.000
-1.257
-1.257
-1.257
-3.632
-3.632
B
0
0.0000
0.0068
0.0068
0.0068
0.0725
0.0725
C
0
0.000000
0.000143
0.000143
0.000143
-0.000276
-0.000276
A/C compressor is engaged 100% of the time
Since emissions and fuel economy are affected by the operation of the air conditioning
compressor and not simply whether the switch is turned on or off, the MOBILE6 analysis did not
develop analogous correlations for drivers turning their air conditioning systems on as a function
of heat index. However, whether the switch was turned on or off was recorded during the test
program. In order to provide a point of comparison with other studies, which have focused on
the frequency that the switch is turned on, we estimated this parameter, as well. Using the data
collected during the 1992 study in Phoenix,25 we performed a regression of the average
percentage of time during each trip that the A/C system was turned on against the ambient
temperature at the time of the trip. For 5 degree F intervals, we calculated the average
percentage of time that the air conditioning system was turned on and that the air conditioning
compressor was engaged. These data are plotted in Figure III-6.
72
-------
Figure III-6. Air Conditioning Use in Phoenix
A/C Use in Phoenix
1 nn°/
I UU /O
RD%
ou /o
RD%
ou /o
4D%
HU /o
onox.
^u /o
n%
^*-77i
-------
We then performed a simple regression of these percentages as a function of temperature. The
result was:
Compressor engagement (fraction) = 1.0535 In (ambient temperature (F)) - 3.9981
We then divided the MOBILE6 estimate of compressor on fraction by this estimate in order to
convert the fraction of time that the compressor is on to the fraction of time that the air
conditioning system is on.
We combined the estimates of air conditioning system on and compressor on fractions
with 30-year average meteorological conditions for each hour of the day for every month of the
year for each county in the U.S. to estimate the percentage of driving time during which the
system and compressor were engaged under those conditions. We then weighted these
percentages by the relative driving time occurring during each hour of the day and month in each
county to obtain an estimate of the overall percentage of the time which air conditioning
compressors are engaged in the U.S. From this, we estimate that, on average across the nation
and throughout the year, the air conditioning system is turned on 23.9% of the time and the
compressor is engaged 15.2% of the time.
We then adjusted this latter percentage to account for reduced compressor loads at
temperatures less than 95°F and higher loads above 95°F.26 Ed Nam, at the University of
Michigan, developed a model of air conditioning load on the engine as a function of
temperature.27 From Figure 4 of this paper, we derived the following equation of compressor
torque in foot-pounds versus temperature in degrees F.
Compressor torque = 1.70 + 0.084 * Ambient Temperature (°F)
At the temperature of the SC03 test, 95°F, compressor torque is 9.68 foot-pounds. Therefore, the
estimated torque at a specific ambient temperature was divided by 9.68 and multiplied by the
estimated compressor on fraction from Draft MOVES2004 in order to derive a compressor on
fraction which was consistent with the conditions of the SC03 test. This adjusted formula for
compressor on fraction was again applied to each county in the U.S., accounting for diurnal and
seasonal temperature and driving differences. Adjusting for load, the compressor is on in-use
13.3% of the time, versus 15.2% without adjustment for load. Thus, the average load in-use is
87.5% of the load experienced during the SC03 test at 95°F.
Finally, the impact of air conditioning on fuel economy varies with vehicle driving
pattern. Most air conditioning compressors are belt-driven by the engine. The efficiency of both
the engine and compressor will vary with engine speed and load. This variation is difficult to
model, as the speed and load of engines in various vehicles will vary dramatically based on the
vehicle's drivetrain design, even over the same driving cycle. Lacking specific information on
each vehicle's air conditioning system design and how engine speed and load affect it efficiency,
we assume that the efficiency of the engine and air conditioning compressor implied in the SC03
test applies to other types of driving, as well. However, a more basic effect related to driving
pattern is that the faster a vehicle is moving, the shorter the amount of time that the vehicle needs
to be cooled while it travels a specific distance. Other factors being equal, this reduces the
74
-------
amount of energy needed to cool the vehicle per mile of travel. Therefore, for a specific set of
ambient conditions, we assume that the impact of air conditioning on fuel use is constant with
driving time (i.e., fuel use in terms of gallons per hour is constant). This means that the excess
fuel use due to operating the air conditioner varies inversely proportional to vehicle speed. In
other words, at low vehicle speeds, like that of the SC03 test, excess fuel use is relatively high on
a per mile basis. At high vehicle speeds, like that of highway driving, the excess fuel use due to
operating the air condition is relatively low on a per mile basis.
We confirmed this assumption by testing five vehicles over a variety of test cycles at
EPA's Ann Arbor laboratory with both the air conditioning turned on and off. These tests were
not run in an environmental chamber which simulates numerous aspects of higher temperature
conditions, like humidity and solar load. Vehicles were tested as they normally required for an
FTP at 75°F, except that the ambient temperature of the test cell was either higher or lower.
Since the primary purpose of the test program was to determine the relative load of the
compressor on the engine and the relative effect on fuel economy, this simplified test procedure
was sufficient. A full report of this test program is contained in the docket.28 The data are
summarized in Table III.A-21.
Table III.A-21. Increased Fuel Use Due to Air Conditioning as a Function of Vehicle
Speed
60°F
75°F
95°F
Absolute increase in fuel consumption (gallons per 100 miles)
FTP Bag 3
FTP Bag 2
HFET
SC03
0.46
0.70
0.27
—
0.81
1.11
0.41
0.87
1.01
1.41
0.50
—
Absolute increase in fuel consumption adjusted to 21.5 mph (gallons per 100 miles)
FTP Bag 3
FTP Bag 2
HFET
SC03
0.55
0.52
0.60
—
0.97
0.83
0.91
0.87
1.20
1.06
1.13
—
The upper half of Table III.A-21 shows the increased fuel use as directly measured over
the four cycles. The lower half multiplies these fuel increases by the ratio of the average speed
of SC03 to the average speed of the cycle tested. As can be seen in the upper half of the table,
the increase in fuel use varies by roughly a factor of three across the 3-4 cycles at each ambient
temperature. The cycle with the highest vehicle speed, HFET, showed the smallest increase at
all three temperatures. Bag 2 of the FTP, with the lowest average speed, showed the largest
increase at all three temperatures. The fuel increases shown in the lower half of Table III.A-21
show much lower variability. Fuel increases now vary by less than 20% from lowest to highest
at any given temperature. These data convincingly confirm the assumption that the increased
fuel needed to operate the air conditioning system is roughly constant with time, as opposed to
mileage.
75
-------
Third, the air conditioning compressor also can be engaged in defroster mode to
dehumidify the air and keep the windshield from fogging up. Due to the fact that the defroster
tends to be operated at lower ambient temperatures than the air conditioner, the load on the
engine is generally lower. We do not have a direct estimate of the frequency that the defroster is
turned on, nor the frequency that the compressor is engaged during defroster mode. No study
analogous to that performed in Phoenix has yet been performed. However, a recent study by the
National Renewable Energy Laboratory (NREL) and the Office of Atmospheric Programs (OAP)
within EPA estimated the percentage of time that people turned on the defroster, as well as the
air conditioning system, while driving.29'30
This study uses a personal comfort model to predict when a driver would be likely to turn
on the air conditioner. The greater the combination of ambient temperature and humidity, the
more uncomfortably hot a person would feel. The more the model predicted that a person would
feel uncomfortably hot, the greater the estimated likelihood that the driver would turn on the air
conditioner. A simpler model was used for defroster use. If the ambient temperature is between
35 and 55°F and the relative humidity is greater than 80%, then the driver is assumed to turn on
the defroster.
The NREL-OAP studies presented estimates of air conditioning use and air conditioning
use plus defroster use by state, but not for the nation as a whole. We combined these state-
specific estimates with the VMT estimates described above (aggregated by state) in order to
estimate national averages. In the NPRM, we estimated that the national average air
conditioning use was 22.9% and combined air conditioning plus defroster use was 33.5%.
However, comments provided by NREL indicate that their more recent estimates of air
conditioning and air conditioning plus defroster use are 28.1% and 32.6%, respectively. Since
the conditions for air conditioning and defroster use do not overlap, national average defroster
use, according to this model is 4.5% (versus the 10.6% figure estimated in the NPRM). Thus,
the NREL-OAP model estimates slightly higher average air conditioning use nationwide (28.1%)
than MOBILE6.2 (23.9%). These two estimates are still remarkably close given the differences
in methodology. It is also not surprising that the MOBILE6.2 estimate would be the lower of the
two, given that the NREL-OAP model does not account for people choosing alternatives to air
conditioning use under hot-humid conditions, such as putting a convertible top down, driving
with the windows open, or inoperative air conditioning systems.
The NREL-OAP study also performed vehicle modeling in order to estimate the impact
of air conditioning and defroster load on vehicle fuel economy. They assumed an average
temperature of 81°F for air conditioning use and 61°F for defroster use. (This defroster
temperature exceeds the range of defroster use and appears to consider higher under the hood
temperatures. However, the air conditioning temperature appears reasonable without such an
adjustment.) The also used the FTP as their driving cycle. Assuming a mix of 65% cars and
35% trucks, the load of the air conditioner increased fuel consumption 19.8%, while that for the
defroster was 4.1%. Thus, the impact of the defroster on fuel consumption is 20.7% of the
impact of air conditioning. This includes the impact of a lower ambient temperature and the
periodic cycling of the compressor on and off. NREL-OAP's modeling of the air conditioner is
comparable to our estimate of 13.3% air conditioning use adjusted for compressor load. Thus,
the 13.3% estimate can then be scaled based on the results of the NREL-OAP study.
76
-------
Two scaling factors are necessary. One, defroster use is 13.8% of that air conditioning,
based on the NREL-OAP results in both cases for consistency (4.5%/32.6%). (Use of our own
estimate for air conditioning (23.9%) would increase this percentage by 36%, but we believe that
it would be more appropriate to use estimates from the same source in this case.) Two, when
turned on, defroster use has 20.7% of the fuel economy impact as air conditioning. Combining
these two factors (20.7% times 13.8%) produces an overall scaling factor of 2.9%. Applying this
to our estimate of 13.3% for the compressor on percentage in terms of the ambient conditions of
the SC03 test produces an analogous estimate of 0.4% for defroster use. Combined air
conditioning and defroster use would be 13.7% (13.3% + 0.4%), or 3% higher than that of air
conditioning alone. This is a very small impact. We decided not to include the impact of
defroster use in the 5-cycle formulae at this time for two reasons. One, no vehicle studies have
yet been performed to confirm the projection that drivers actually turn on the defroster as
assumed by NREL-OAP. Two, the ambient conditions existing during defroster use differ
dramatically from those of the SC03 test.
Based on these three assumptions, the impact of air conditioning on running fuel use is
estimated as 13.3% of the difference between fuel use per mile over the SC03 and a combination
of Bag 2 and Bag 3 tests times 21.5 mph and divided by the average speed of either city or
highway driving. Based on the descriptions of city and highway driving from Draft
MOVES2004, the average speeds are 19.9 mph and 57.2 mph, respectively. Thus, the excess
fuel use due to air conditioning operation is as follows.
For city driving:
Excess fuel use due to air conditioning =
0.
21.5
x x
19.9
1
(Fuel economy ^
[over the SC03 test)
0.39
(Fuel economy}
0.61
(Fuel economy}
For highway driving:
77
-------
Excess fuel use due to air conditioning =
O.i33x —x
57.2
1
( Fuel economy
\over the SC03 test ^
( \
0.39
(Fuel economy^
\over Bag 2 J
+
( \
0.61
(Fuel economy}
i I over 5a^ 3 J
We received several other comments on this methodology. However, as discussed in the
Response to Comments document, these comments did not lead us to revise the above equations.
Thus, the 5-cycle formulae will continue to use the above relationships to estimate the impact of
air conditioning use on city and highway fuel economy.
4. Effect of Cold Ambient Temperatures on Running Fuel Use
Finally, we added the impact of colder ambient temperatures on running fuel use. As was
done for start fuel use, we base our estimate of the impact of colder ambient temperatures on
running fuel use by comparing the fuel use over the standard and cold temperature FTP tests. At
75°F, engine, drivetrain and tires are generally assumed to be fully warmed up by the end of Bag
1. Thus, Bag 2 of the FTP at 75°F includes only fully warmed up operation. As was discussed
above, the start fuel use associated with a 10 minute soak at 75°F is very small and generally
considered to be negligible. Thus, Bag 3 at 75°F can also be considered to consist of essentially
fully warmed up vehicle operation.
As discussed in detail in the Draft Technical Support Document to the NPRM, it is not
clear that vehicles are fully warmed up during Bags 2 and 3 at 20°F. However, as described
above, the average city driving trip is only 4.1 miles, well below that of the FTP (7.5 miles).
Thus, Bags 2 and 3 of the cold FTP provide a reasonable estimate of warmed up driving during
city-like driving (i.e., the vehicle is warmed up to the extent that it typically reaches during short
trips). The effect of cold temperature on fuel use during city driving can be estimated from the
difference in fuel use over Bags 2 and 3 of the FTP at 20°F and that at 75°F.
However, we could not make the same conclusion for either longer highway-like driving
trips, nor conclude that the effect of colder temperatures would be the same at higher vehicle
speeds. Based on a number of studies which investigated the impact of colder temperatures on
fuel economy, we estimated that running fuel use at 20°F at higher vehicle speeds would be
roughly 4% higher than that at 75°F.
When we determined the appropriate weighting factors for running fuel use at 20°F and
75°F in the NPRM, we assumed that running fuel use increased linearly with temperature below
75°F. We received one comment which challenged this assumption. As discussed in Section
78
-------
5.2.4 of the Response to Comment document, we evaluated the running fuel use over Bags 2 and
3 of the FTP of several conventional Honda vehicles at 20, 50 and 75°F to determine if running
fuel use did in fact change linearly with a decrease in temperature from 75°F. These data are
summarized in Table III.A-22.
Table III.A-22. Warmed Up Fuel Use Versus Temperature: Honda Data
Civic
Element
Accord (L4)
Accord (V6)
MDX
TSX
Odyssey
RSXPRB
Average
Civic
Element
Accord (L4)
Accord (V6)
MDX
TSX
Odyssey
RSXPRB
Average
Civic
Element
Accord (L4)
Accord (V6)
MDX
TSX
Odyssey
RSXPRB
Average
Bag 2 mpg
Bag 3 mpg
Bag 2+3 Fuel Consumption (gal/mi)
75°F
38.1
23.2
25.6
21.7
17.2
23.0
20.9
24.0
41.9
25.3
30.0
26.1
20.2
27.4
24.5
28.5
0.0251
0.0414
0.0363
0.0424
0.0540
0.0401
0.0445
0.0386
0.0403
50°F
35.3
22.2
23.4
20.8
16.8
22.2
19.1
22.7
39.1
24.6
28.4
26.1
20.3
26.6
22.6
28
0.0270
0.0429
0.0391
0.0434
0.0546
0.0415
0.0485
0.0401
0.0421
20°F
29.2
19.2
20.6
19.2
15.2
20.1
17.5
20.7
33.9
21.5
25.2
23.7
17.9
24.6
21.4
25.3
0.0320
0.0494
0.0443
0.0473
0.0610
0.0454
0.0522
0.0441
0.0470
These data indicate that running fuel use at 50°F averaged 0.0421 gallon per mile, or 0.0018
gallon per mile higher than that at 75°F. Running fuel use at 20°F averaged 0.0470 gallon per
mile, or 0.0067 gallon per mile higher than that at 75°F. Thus, fuel use increased faster below
50°F than from 75 to 50°F. On average the increase in running fuel use at 50°F was 27% that at
79
-------
20°F. The same relationship was true when four hybrids (two Honda vehicles and two Toyota
vehicles) were added to the database.
We used this relationship to evaluate the impact of ambient temperature on running fuel
use throughout the U.S. using the same methodology that was used in conjunction with the
assumption of linearity performed for the NPRM. The result was a weighting factor for running
fuel use at 20°F of 0.18, versus the weighting factor of 0.30 developed for the NPRM. Thus, we
have reduced the weighting factor for running fuel use at 20°F in the 5-cycle formulae for both
city and highway fuel economy to 0.18. This is shown in the equations below.
For city driving:
Excess fuel use due to colder temperatures =
0.18x
0.5 0.5 1 [ 0.41 0.48 0.11 }
Bag 2 20 FE Bag 3 20 FE ) ^ Bag375 FE Bag2^ FE US06 City FE )
For highway driving:
Excess fuel use due to colder temperatures =
0.18 x 0.04 x running fuel use without air conditioning at 75 F =
0.21 0.79
0.18x0.04x
HFETFE US06 Highway FE
Combining the estimates of running fuel use at 75°F with the air conditioning turned off
with the estimate of excess fuel use of running the air conditioning system and the estimate of
fuel use due to colder ambient temperatures produces the following formulae for running fuel
use:
80
-------
For city driving:
Running Fuel Use =
0.82 x
0.48
0.41
0.11
Bag275 FE Bag375 FE US06City FE
0.18x
0.5
0.5
Bag2 20 FE Bag3 20 FE
+ 0.133x1.083:
1
0.61
0.39
SC03 FE ( Bag315 FE Bag215 FE
For highway driving:
Running Fuel Use =
(l.007)>
0.79
0.21
US06 Highway FE HFET FE
+ 0.133x0.377x
1
0.61
0.39
SC03FE \Bag375FE Bag275FE
5. Adjustment Factor for Non-Dynamometer Effects
There are a large number of factors which affect vehicle fuel economy which are not
addressed by any of the five current dynamometers tests. These include roadway roughness,
road grade (hills), large vehicle loads (e.g., trailers, cargo, multiple passengers), wind,
precipitation, to name just a few. Even when a factor is addressed by a dynamometer test, such
as driving pattern or air conditioning, the factor is only approximately measured, as all realistic
driving patterns cannot possibly be included in a test having a reasonable length of time. Nor
can all the possible ambient conditions affecting air conditioner operation be tested. Thus, any
estimate of in-use fuel economy derived from the five dynamometer tests is necessarily
approximate.
It would be possible to use the formulae described in Section III.B to directly estimate the
fuel economy label values. These fuel economy values would provide drivers with an indication
of the relative fuel economy which they should expect to achieve in-use, at least the best estimate
that the current five dynamometer tests can provide. This would provide vehicle purchasers with
information as to which vehicle would provide greater or lesser fuel economy than another
vehicle. However, as discussed in section I, many vehicle owners expect to achieve the fuel
economy label values when they drive. Often, they do this to determine if their vehicle is
operating properly. Thus, it would be advantageous to such vehicle operators if the fuel economy
label values accounted for all factors affecting fuel economy and not just those addressed by the
dynamometer tests. This is the rationale for the 90% and 78% adjustment factors which are
currently applied to the measured FTP and FIFET fuel economies when determining the city and
81
-------
highway label values. This cannot be done in a vehicle specific manner, since no such estimates
are available without additional dynamometer tests and additional vehicle testing. However, it is
possible to account for these factors on a fleet wide basis, as was done with the current 90 and
78% adjustment factors.
It is possible to estimate the effect of each of the untested factors mentioned above and
then add them up. This was done as part of the 1984 label adjustment rule. That study estimated
that the net effect of all the non-dynamometer factors was roughly 30%.31 However, at that time,
these factors included the impact of vehicle speed, acceleration, colder temperature and air
conditioning, in addition to the factors described above. With the inclusion of the fuel
consumption over the US06, SC03 and cold FTP tests in the 5-cycle formulae, the net effect of
these factors should be much smaller. One caution is that the comparisons of dynamometer
measured fuel economy and onroad fuel economy performed for the 1984 rule tended to find a
smaller shortfall than the total of all the individual non-dynamometer factors. The reason for this
was not clear. In other words, the individual factors which might have been over-estimated
could not be determined. In this section, we will update the impact of several of these factors
and develop a revised estimate of the overall impact of non-dynamometer factors on onroad fuel
economy relative to an estimate based on the 5-cycle formulae.
The easiest factor to quantify is likely that related to fuel energy density. EPA's test fuels
do not contain any oxygen. However, commercial gasoline can contain either of two
oxygenates, methyl tertiary butyl ether (MTBE) and ethanol. Future levels of MTBE use are
uncertain because of water contamination issues related to leaking underground storage tanks.
However, the recently passed Energy Policy Act of 2005 (EPA2005) guarantees a certain level
of ethanol use. For example, in 2008, EPA2005 requires the use of 5.4 billion gallons of
renewable fuel, the vast majority of which is expected to be ethanol blended into gasoline. DOE
projects that total gasoline consumption will be 146 billion gallons in 2008. Thus, ethanol would
represent 3.7% of gasoline by volume. Ethanol contains roughly 33% less energy per gallon, so
on average, commercial gasoline in 2008 will contain 1.2% less energy per gallon than it would
if it were not oxygenated. Engine efficiency is unaffected by fuel energy content in this range.
Thus, reducing the energy content of gasoline by 1.2% will reduce volumetric fuel economy by
the same 1.2%. This is not reflected in EPA dynamometer testing.
Currently, dynamometer testing performed for the state of California is done using a fuel
containing 2% oxygen by weight. This fuel contains approximately 2% less energy per gallon
than EPA's test fuel. In many cases, dynamometer tests performed for California can also be
used in EPA certification. Thus, it would be appropriate to divide the measured fuel economy of
any test performed using an oxygenated California test fuel by one minus the oxygen content of
that fuel by weight, usually 2%. However, with the passage of EPA2005, gasoline sold in
Federal reformulated gasoline areas in California will no longer be required to contain two
weight percent oxygen. Thus, California may no longer require its test fuels to contain oxygen.
In any case, measured fuel economy using an oxygenated fuel should be adjusted to reflect the
energy content of EPA's non-oxygenated test fuel. At the same time, the average driver in 2008
will achieve 1.2% lower fuel economy than they would if they were using a non-oxygenated
fuel.
82
-------
The volume of renewable fuel required under EPA2005 increases in later years, reaching
7.5 billion gallons in 2012. Assuming this is all ethanol and using DOE's projected gasoline
volume of 147 billion gallons, average commercial gasoline would contain 1.5% less energy than
EPA test fuel. Thus, in the early years of the revised fuel economy labels, difference in fuel
quality will cause onroad fuel economy to be 1.2-1.5% less than that measured on the
dynamometer.
Another factor which has recently been studied in some detail is tire pressure. NHTSA
recently promulgated a regulation requiring car and light truck manufacturers to install tire
pressure monitoring systems in future vehicles. In preparation for this rule, NHTSA conducted
a survey of the tire pressure of in-use vehicles in February 2001.32 Tire pressures were
measured on over 11,500 vehicles at 24 locations throughout the U.S. The results are
summarized in the Table III.A-23. NHTSA presented data for each of the four tires separately
(i.e., front, driver's side tire).33 We averaged the findings for the four tires.
Table III.A-23. NHTSA Onroad Tire Pressure Survey
Difference between Onroad and Manufacturer's Recommend Tire
Pressure (psi) (Average of four tires)
-12
-11
-10
-9
-8
-7
-6
-5
-4
O
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
Cumulative Frequency
LDVs
3.0%
5.0%
7.5%
9.5%
12.0%
16.0%
20.0%
26.0%
31.5%
40.0%
46.5%
54.0%
61.0%
68.0%
74.5%
79.0%
85.5%
88.0%
90.0%
94.0%
95.0%
96.0%
97.0%
98.0%
99.0%
LDTs
3.0%
6.0%
8.5%
11.0%
15.0%
20.0%
25.5%
34.5%
40.0%
47.0%
54.5%
62.0%
69.0%
75.0%
80.0%
84.0%
88.0%
91.0%
92.0%
93.0%
95.0%
97.0%
98.0%
99.0%
100.0%
83
-------
As can be seen from the table, 54-62% of cars and light trucks have under-inflated tires, while
31-39% have over-inflated tires. Using these estimates, we found that the tires of the average car
were under-inflated by 1.1 psi, while those on light trucks were under-inflated by 1.9 psi.
NHTSA presented two estimates of the effect of tire pressure on fuel economy. A 1978
study by Aerospace Corp. found that fuel economy decreased by 1% for every 3.3 psi decrease in
tire pressure, while more recent test data submitted by Goodyear showed a 1% decrease in fuel
economy for every 2.96 psi decrease in tire pressure. Using these two factors, the 1.1 psi under-
inflation of car tire pressure causes a 0.3-0.4% decrease in onroad fuel economy. The 1.9 psi
under-inflation of light truck tire pressure causes a 0.6% decrease in onroad fuel economy.
Assuming that new vehicles average close to their CAFE fuel economy standards (27.5 mpg for
cars and 20.6 mpg for light trucks) and a 50/50 mix of the two types of vehicles, the fleet-wide
effect of under-inflation is 0.5% using either factor.
NHTSA recently promulgated a regulation requiring manufacturers to monitor tire
pressure.34 This rule requires vehicles to be equipped with sensors to detect a tire which is
under-inflated by 25% or more. With a few exceptions, the regulation begins phasing in with the
2006 model year and all new cars and light trucks must have the monitoring systems by the 2008
model year. Assuming a tire's recommended pressure is about 32 psi, this implies catching tires
under-inflated by 8 psi or more. If we assume that the regulation is 100% effective and eliminate
all tires under-inflated by 8 psi or more in Table III. A-23 above, passenger car tires are no longer
under-inflated on average in-use; they exceed their specifications by roughly 0.1 psi. Light
trucks are still under-inflated by about 0.4 psi. Across the light duty fleet, the net effect on fuel
economy decreases to 0.1%, or roughly one-fifth the level prior to the rule. Of course, the
effectiveness of the rule could be less than 100%. At the same time, vehicles with a single tire
under-inflated by 8 psi could have other tires with a lower degree of under-inflation. Thus, the
rule could have some effect on tires with smaller levels of under-inflation than assumed above.
In any event, the effect should be less than 0.5% and could be close to zero.
A third factor which can be quantitatively estimated is the effect of wind. Wind affects
fuel economy by changing the road load of the vehicle. Wind can affect both rolling resistance
and aerodynamic drag. Rolling resistance is primarily affected by a side wind, which pushes the
vehicle sideways. The driver must compensate by turning the steering wheel into the wind. This
increases the drag caused by the tires on the roadway surface. However, the effect of wind on
aerodynamic drag is far the larger of the two effects.
Aerodynamic drag is generally assumed to be the product of three factors: 1) the frontal
area of the vehicle, 2) the air speed going by the vehicle squared, and 3) the "drag coefficient or
Cd." A headwind increases the speed of the air going by the vehicle directly (i.e., a 10 mph wind
increases air speed 10 mph). A tailwind decreases air speed by the vehicle. Even if the
frequency of a headwind and a tailwind is the same, total aerodynamic drag increases, due to the
fact that drag is proportional to air speed squared. For example, 40 mph squared is 1600 mph2.
Given a headwind of 10 mph, air speed increases to 50 mph and 50 mph squared is 2500 mph2.
Given a tailwind of 10 mph, air speed decreases to 30 mph and 30 mph squared is 900 mph2.
The average of 2500 and 900 mph2 is 1700 mph, which is more than 6% greater than 1600 mph2.
84
-------
Thus, even a randomly directional wind will increase total aerodynamic drag and decrease fuel
economy.
An even greater effect of wind, however, is that it changes the drag coefficient, Cd, and
increases the effective frontal area of the vehicle in the direction of the wind. For example, as
large as the frontal area is of a semi-tractor trailer combination, its area from a side view is 5-10
times as large. The greater the side wind relative to vehicle speed, the more the truck is actually
driving sideways down the road as far as aerodynamic drag is concerned. With respect to cars
and light trucks, their body shapes are designed to reduce aerodynamic drag when traveling into
the wind. Front and rear ends are sloped. Spoilers and other rear end shapes are designed to
minimize the creation of vortices behind the vehicle which "pull" the vehicle back as it is driving
forward. However, as soon as a significant side wind occurs, these benefits start to diminish.
For the 1984 label adjustment rule, EPA estimated that wind reduced onroad fuel
economy by 3% for a small car and 2% for a large car.29'35 These estimates were based on
several estimates made by the Department of Transportation: 1) the effect of 10, 15, and 20 mph
winds on aerodynamic drag at a constant speed of 55 mph as a function of wind angle, 2) the
effect of increased aerodynamic drag on 55 mph fuel economy, and 3) a distribution of onroad
VMT as a function of wind speed (with the national average wind speed being 9 mph). EPA
applied these estimates directly to highway fuel economy, but reduced the fuel economy effect
by 80% for city driving. This reduction was based on the fact that roughly 20% of the FTP test
is at speeds near 55 mph.
We reviewed this methodology in detail to determine if any improvement could be made.
Two areas were identified. The first area was the fact that the effect was estimated only for cars,
as that was the focus of the study. The second area was the assumption that wind had no effect
on fuel economy at vehicle speeds below roughly 55 mph. While aerodynamic drag is much
lower at city driving like speeds than highway speeds, wind speed is a higher fraction of vehicle
speed at low vehicle speeds. The effect of wind on a vehicle's effective drag coefficient
increases as the effective angle of the air speed increases. Thus, the effect of a side wind can be
significant, even at low vehicle speeds.
In order to expand the previous analysis, we developed a model of aerodynamic drag and
its impact on fuel economy as a function of wind speed and angle. We broke down the speed of
the vehicle through the air in terms of its x and y coordinates (i.e., parallel and perpendicular to
the direction of the vehicle). The parallel component is the speed of the vehicle plus the cosine
of the wind angle times wind speed. The perpendicular component is the sine of the wind angle
times wind speed. We then calculated the net angle of the air flowing past the vehicle and its
speed from these two x-y components. The net angle of the air flowing past the vehicle is the
arctangent of the ratio of perpendicular air speed to parallel air speed. Net air speed is the square
root of the sum of the square of the perpendicular air speed and the square of the parallel air
speed. Aerodynamic drag is the square of the net air speed times the vehicle drag coefficient.
DOT estimated that the vehicle drag coefficient increased 1.5% for every degree increase
in yaw angle, or angle of net air flow past the vehicle. Using this estimate, we were able to
85
-------
reproduce the estimates of the change in aerodynamic drag as a function of wind speed and
direction on a vehicle traveling at 55 mph, which were presented Figure 26 of the EPA report
33
In order to broaden the estimate to include light trucks, we obtained estimates of the
effect of wind angle on a vehicle's drag coefficient from Gillespie.36 Gillespie presents the
estimated absolute increase in drag coefficient as a function of wind speed for four vehicle
designs: pick-up trucks, station wagons, family sedans, and sports cars. The results are presented
in Table III. A-24 below in tabular form. Gillespie did not present estimates for sport utility
vehicles (SUV). We estimated the effect for SUVs by averaging the impacts for pick-up trucks
and station wagons.
Table III.A-24. Effect of Wind Angle on Vehicle Drag Coefficient
Wind Angle
(Deg)
0
5
10
15
20
Pick-Up
Truck
0
0.045
0.120
0.195
0.240
Station
Wagon
0
0.015
0.050
0.090
0.110
Family
Sedan
0
0.010
0.040
0.080
0.125
Sports
Car
0
0.010
0.025
0.050
0.070
SUV
0
0.030
0.085
0.143
0.175
Average
0
0.025
0.070
0.121
0.155
We estimated a fleet average change in the effective drag coefficient by averaging the
estimates for the five model types. We averaged the estimates for the three types of passenger
cars equally (33/33/33), the two estimates for light trucks equally (50/50) and then averaged the
averages for car and light trucks equally (50.50). For a wind angle of 20 degrees, the average
change in drag coefficient for cars is 0.102 and 0.208 for light trucks. Assuming average drag
coefficients in still air of 0.30 for passenger cars and 0.40 for light trucks,25 these changes
represent increases of 1.7% and 2.6% per degree of wind angle. The figure for cars matches the
DOE estimate from 1974 quite well, while that for light trucks is much larger. We performed a
regression of the change in drag coefficient versus wind angle in degrees and found the
following relationship:
Change in drag coefficient = -0.003 76 + 0.006815 x wind angle + 0.000065 x (wind angle}2
We also performed a similar regression used a linear model. The linear model yielded larger
average increase in vehicle drag coefficient. Therefore, we retained the non-linear model.
In order to expand the estimate to include city, as well as highway driving, we again used
PERE.25 Using PERE, we estimate that a 10% increase in aerodynamic drag or drag coefficient
decreases city fuel economy by 0.93%. Likewise, highway fuel economy decreases 3.11%.
Implied in the DOT estimate of the effect of wind speed on 55 mph fuel economy is a decrease
of roughly 4%. Thus, PERE estimates a somewhat lower effect of wind speed on fuel economy
even at highway speeds.
We then applied our model using the DOT estimates of the national average distribution
of wind speeds, which is shown in Table III.A-25 below. We assumed that the average wind
86
-------
speed within a range of wind speeds was the average of the lower and upper limit of the range.
We assumed that the average wind speed for winds above 25 mph was 27.5 mph.
Table III.A-25. Frequency of Wind Speeds in the U.S.
Wind Speed
(mph)
0-3
4-7
8-12
13-18
19-24
25
Assumed Average
Wind Speed
(mph)
1.5
5.5
10
15.5
21.5
27.5
% of National VMT
16%
28%
30%
18%
6%
2%
Using an average vehicle speed of 19.9 mph for city driving and 57.1 mph for highway
driving (from Draft MOVES2004), the vehicle drag coefficient increases by 73.5% and 15.9%,
respectively, on average. Assuming average city and highway fuel economy of 18.8 and 25.5
mpg (from our database of 615 recent model year vehicles without any factor for non-
dynamometer effects) and city and highway VMT weights of 43% and 57%, respectively,
composite fuel economy is 22.1 mpg in still air and 20.8 mpg with a typical distribution of wind.
Thus, taking wind into account reduces onroad fuel economy by 6%. This is more than twice
that estimated for the 1984 label adjustment rule.
Roughly 60% of this 6% increase is due to the increase in drag coefficient during city
driving. This portion of the estimate is likely the most uncertain, due to the large wind angles
which can occur at relatively low vehicle speeds (e.g., 45% or more). This means that the
figures taken from Gillespie are being extrapolated to a significant degree. We are not certain
that the drag coefficient would continue to increase beyond 20 degrees wind angle at the same
rate as below 20 degrees. However, the effective frontal area of the vehicle would continue to
increase. Rolling resistance is also likely to increase, as the vehicle must be driven increasingly
sideways to travel in the direction that the vehicle is pointing (i.e., down the road). It is unlikely
that either the DOT or Gillespie estimates consider an increase in rolling resistance, as they were
likely developed in wind tunnels where the vehicle is standing still. Thus, it is likely that the
estimate for the effect of wind on onroad fuel economy is more uncertain than those for fuel
quality or tire pressure. Still, the effect of wind appears to be very significant and likely larger
than either of the other two factors.
The final factor which we reevaluated was road roughness. Road roughness was
estimated to reduce onroad fuel economy by 4.2% relative to that measured on the
dynamometer.29 The model developed in 1980 included estimates of the percentage of VMT
driven on dry (69%), wet (25%) and snow-covered (6%) roads.33 It also included estimates of
roadway miles which were unsurfaced, gravel, low-load asphalt, and concrete and high load
asphalt. It also included estimates of the percentages of VMT on each roadway type, as well as
the effect of each roadway type on fuel economy relative to that measured on a dynamometer.
As the vehicle coast downs used to determine dynamometer road loads are conducted on dry
concrete or high load asphalt roads, this combination of roadway type and driving condition was
87
-------
assumed to have no effect on fuel economy relative to a dynamometer. All other roadway types
and driving conditions reduced onroad fuel economy. Table III.A-26 presents the inputs to the
model developed in 1980 (with highway and VMT estimates from FHWA's Highway Statistics
1977).
Table III.A-26. Effect of Road Roughness on Onroad Fuel Economy: 1977
Percent of Roadway
Miles
Percent of VMT
Unsurfaced
18.2%
1.8%
Gravel
31.1%
9.7%
Low-load
Asphalt
27.3%
30.2%
Concrete or High-
load Asphalt
23.4%
58.3%
Effect on Onroad Fuel Economy
Dry
Wet
Snowy
-20%
-30%
-35%
-15%
-18%
-20%
-4%
-5%
-10%
0%
-3%
-7%
The net effect of the inputs shown in Table III.A-26 and the percentages of VMT with various
road conditions cited above was an average fuel economy shortfall of 4.4%.
We updated the model using more recent estimates of VMT by roadway type contained
in FHWA's Highway Statistics 2003. We had to combine estimates from a set of tables in order
to estimate VMT by cars and light-duty trucks by roadway surface. We began with estimates of
national VMT by cars and 2-axle, 4 tire trucks by roadway type in 2003 from Table VM-1. We
then converted the VMT by 2-axle, 4 tire trucks to the VMT by EPA-defined light-duty trucks by
multiplying by 0.9234, which was derived from an Oak Ridge National Laboratory study.
According to this study, 7.66% of the VMT by 2-axle, 4 tire trucks is by trucks which have a
curb weight above 6000 pounds or a gross vehicle weight rating of above 8500 pounds, which
put these vehicles into EPA's heavy-duty vehicle class.
We then obtained estimates of the length of roadway by surface type for each roadway
class from Table HM-12. The surface types used in Table HM-12 differ somewhat from those
cited in the EPA fuel economy study and shown in Table III.A-26. Table HM-12 uses two major
classes of roadway surface: Unpaved and Paved. There are five sub-classes of paved roadway
surfaces: low, intermediate, high-flexible, high-composite, and high-rigid, where low,
intermediate and high refer to the weight carrying capacity of the roadway. According to
FHWA, Paved includes the following categories:
Low type— an earth, gravel, or stone roadway which has a bituminous surface course less
than 1" thick suitable for occasional heavy loads;
Intermediate Type-- a mixed bituminous or bituminous penetration roadway on a flexible
base having a combined surface and base thickness of less than 7";
High-Type Flexible-- a mixed bituminous or bituminous penetration roadway on a
flexible base having a combined surface and base thickness of 7" or more; also
includes brick, block, or combination roadways;
88
-------
High-Type Composite-- a mixed bituminous or bituminous penetration roadway of more
than 1" compacted material on a rigid base with a combined surface and base
thickness of 7" or more;
High-Type Rigid— a Portland Cement Concrete roadway with or without a bituminous
wearing surface of less than 1".
From these definitions, it seemed that the gravel referred to in the definition of the Low
Paved class was the support for the bituminous surface. This would imply that both unsurfaced
and gravel surfaced roadways of Table III. A-26 fell into the unpaved categories. However, it
also seemed possible that the gravel roadways referred to in Table III. A-26 were included in the
low paved category. It also seemed uncertain whether the high, flexible paved roadway fell into
the low-load asphalt or high-load asphalt. In order to reflect these uncertainties, we developed to
mappings of these six roadway surfaces onto the four types used in the EPA study in order to
bracket the potential fuel economy impact. Table III.A-27 shows these two mappings.
Table III.A-27. Mapping of Roadway Surfaces
Low Fuel Economy
High Fuel Economy
Unsurfaced
Unpaved
None
Gravel
Low Paved
Unpaved +
Low Paved
Low-load Asphalt
Intermediate + High-
Flexible Paved
Intermediate Paved
Concrete or High-
load Asphalt
High-Composite +
High-Rigid Paved
High-Flexible +
High-Composite +
High-Rigid Paved
Using these two sets of roadway mappings, we converted the total roadway lengths for each
roadway surface class within each highway class from Table HM-12 into lengths of roadway by
the surface classes shown in Table III. A-26 within each highway class.
We then estimated the effect of roadway condition on the fuel economy of vehicles
driving on each roadway surface type. We did this by multiplying the estimates of the
percentage of VMT driven on dry (69%), wet (25%) and snow-covered (6%) roads to the
changes in fuel economy for each roadway condition for each roadway surface (shown in Table
III. A-26). The result is that the average fuel economy on unpaved roads, gravel roads, low load
paved road and high load paved roads are 72.5%, 83.9%, 95.1%, and 98.8% of that on dry high
load paved roads like those simulated during dynamometer testing.
We then applied these effects of roadway conditions to the distribution of roadway
surfaces for each highway class in order to develop estimates of the average effect of roadway
conditions for each highway class. We then weighted these effects by the distribution of car and
light truck VMT by highway class in order to develop a national average effect of roadway
surface and condition on fuel economy. Using the two mappings, we estimate that the national
average impact of roadway surface and condition on fuel economy is 1.4-3.2%. Based on the
decisions underlying the two mappings, we believe that the low end of this range is more likely
than the high end. Thus, it appears that the effect of roadway surface and condition is lower
today than it was in 1977.
89
-------
The other non-dynamometer factors are more difficult to estimate. Fortunately, they
appear to be smaller in magnitude. The following table shows the breakdown of the impact of all
the non-dynamometer factors as estimated for the 1984 rule. It also updates the impacts for the
four factors discussed above, as well as eliminating those factors which are now addressed by the
US06, SC03 and cold FTP tests.
Table III.A-28. Effect of Non-Dynamometer Factors on Onroad Fuel Economy
Factor
Ambient temperature
Fuel Quality
Altitude
Wind
Road grade
Road surface
Road curvature
Trip length
Average vehicle speed
Cold starts
Acceleration intensity
Brake drag
Wheel alignment
Tire switching
Tire pressure
Vehicle load
Dynamometer loading
Tire effects
Weight classification
Manual transmissions
Power accessories, air
conditioning
Sum
Analysis for 1984 Rule
-5.3%
0%
-0.1%
-2.3%
-1.9%
-4.2%
-0.1%
0.8%
10.6%
-0.7%
-11.8%
-0.3%
-0.3%
-0.4%
-3.3%
-0.4%
-2.7%
-5.1%
-1.0%
-1.8%
-0%
-30%
Effect Applicable to 5-Cycle Fuel Economy
Included
-1.1 to -1.5%
-0.1%
-6%
-1.9%
-1.4% to -3. 2%
-0.1%
Included
Included
Included
Included
-0.3%
-0.3%
-0.4%
-0.5%
-0.4%
Revised test procedures may have removed
most of these effects
Air conditioning included
-12% to -15%
As can be seen, the net impact of non-dynamometer factors applicable to a 5-cycle fuel economy
estimate is 12-15%. The four factors evaluated in detail above comprise the majority of the
impact. Together, other factors like road grade, road curvature, altitude and vehicle condition
add only 3.5% to the overall estimate.
We received little comment on the estimates of the impacts of the individual untested
factors on onroad fuel economy. One commenter indicated that he thought that the effect of
90
-------
wind seemed high, but offered no specific information on how the methodology or estimate
should be changed.
Due to the fact that this type of analysis seemed to over-estimate the impact of these
factors compared to onroad fuel economy estimates from owner diaries, we performed a second
analysis starting from fleet-wide fuel economy estimates.
As described above in section II.C above, FHWA develops annual estimates of car and
light truck fuel economy based on estimates of total VMT and fuel consumption across the
nation. For the NPRM, we utilized FHWA fuel economy estimates for the 2002 and 2003
vehicle fleets. Since the time of the NPRM, FHWA has updated their estimate of onroad fuel
economy for the 2003 fleet and published an estimate for the 2004 fleet. These latest estimates
show lower onroad fuel economy for light trucks than those estimated recently. After adjusting
for the difference FHWA's and EPA's definition of light trucks, onroad fuel economy was 19.7
and 19.9 mpg in 2003 and 2004, respectively. Using MOBILE6.2, we estimate that fleetwide
label fuel economy for these calendar years were 21.1 and 21.2 mpg, respectively. This indicates
a current shortfall of roughly 6.5-7.0%.
Absent any non-dynamometer factor, the average combined (55/45) current label value
for the 601 conventional vehicles in our certification fuel economy database is 20.9 mpg, while
the average combined (43/57) 5-cycle fuel economy is 21.6 mpg, or 3.5% higher. Thus, the
shortfall between the combined 5-cycle fuel economy and the FHWA-based fleet estimates is
roughly 10-10.5%.
In the NPRM, we added one more factor to account for changes in FTP and HFET test
procedures when EPA implemented the Supplemental FTP standards. Specifically, we reduced
combined fuel economy by 3% to compensate for the removal of a 10% upward adjustment to
the vehicle's tractive road load horsepower setting on the dynamometer. However, we received
a comment that, at the time of the SFTP rule, EPA had found that the net fuel economy effect of
all the changes in test procedures was zero, not 3%. We agree with this comment. Therefore, we
have removed this 3% adjustment from this analysis. Thus, the shortfall between the combined
5-cycle fuel economy and the FHWA-based fleet estimates remains at roughly 10-10.5%.
This 10-10.5% difference is slightly lower than the 12-15% estimate for the
impact of non-dynamometer factors shown in Table III.A-28. In the NPRM, we decided to
average the two estimates, rounding down, and include a factor of 0.89 in the 5-cycle city and
highway formulae (i.e., a reduction of 11% in both city and highway fuel economy) to account
for the impact of these factors. As discussed above, however, the bottom-up approach over-
estimated the net effect of these factors back in 1984 when the current label adjustments were
developed. Thus, for the final 5-cycle formulae, we decided to place more emphasis on the top-
down comparison. Therefore, we set the value of the non-dynamometer factor so that the
fleetwide combined 5-cycle fuel economy matches onroad fuel economy as estimated by FHWA.
As indicated above, the average of the current combined fuel economy label values in our
certification fuel economy database (20.9 mpg) is slightly lower than that for the entire onroad
fleet (21.1-21.2 mpg per MOBILE6.2). Thus, the certification database appears to be biased low
91
-------
by 1.0-1.5% relative to the onroad fleet. Thus, the average of the combined 5.cycle fuel
economy label values in our certification fuel economy database should also be 1.0-1.5% lower
than the onroad fuel economy estimated by FHWA (19.7-19.9 mpg), or about 19.6 mpg.
Incorporating a non-dynamometer factor of 0.905 into the 5-cycle city and highway formulae
produces an average combined label value of 19.6 mpg. Thus, we have set the value of the non-
dynamometer factor in the final 5-cycle formulae to 0.905.
6. 5-Cycle Fuel Economy Formulae
The complete 5-cycle fuel economy formulae are developed by combining the results of
the sections on start fuel use, running fuel use, air conditioning, cold temperature, and non-
dynamometer effects. The resultant formulae are described below.
Under the final rule, a special situation could exist where the city fuel economy of a
model type could be developed using the mpg-based formula and its highway fuel economy
developed using an alternative 5-cycle formula based on testing over only 3 test cycles (FTP,
HFET, and US06). This alternative 5-cycle fuel economy formula is also described below.
a. 5-Cycle Fuel Economy Formulae
Vehicles Tested Over a Three-Bag FTP at 75 F
5-Cycle City Fuel Economy Formula
The final 5-cycle city fuel economy would be calculated as follows:
CityFE = 0.905 x where
Start FC + Running FC '
^ _/ „ .,\ »„ ((0.76xStartFuel75)+(0.24xStartFuel2Q)^
StartFC (gallons per mile) = 0.33x -^ ^—^ ^
v ' ^ 4.1 )'
where,
1 1
Start Fuel =3.6x
Bag\FEx Bag3FEx
where
Bag y FEX = the fuel economy in miles per gallon of fuel during Bag 1 or Bag 3 of the FTP test
conducted at an ambient temperature of 75 or 20°F.
92
-------
Running FC =
0.82 x
0.48
0.41
0.11
Bag275 FE Bag3 75 FE US06 City FE
0.18x
0.5
0.5
Bag220FE Bag320FE
•0.133 xl.083x[4/CFC]
where
A/CFC =
1
0.61
0.39
SC03FE I Bag3 75 FE Bagl 75 FE
where
US06 City FE = fuel economy in miles per gallon over the city portion of the US06 test,
US06 Highway FE = fuel economy in miles per gallon over the Highway portion of the
US06 test,
HFET FE = fuel economy in miles per gallon over the HFET test,
SC03 FE = fuel economy in miles per gallon over the SC03 test.
5-Cycle Highway Fuel Economy Formula
The final 5-cycle highway fuel economy would be calculated as follows:
1
Highway FE = 0.905 x
Start FC + Running FC '
where
c, _/ a ., x „„ ((0.16xStartFuel75)+(0.24xStartFuel20)~]
StartFC (gallons per mile) = 0.33 x - ——^ —
^ F J ( 60
where
Start Fuel =3.6:
1
1
BaglFEx Bag3FEx
, and
Running FC = 1.007 x
0.79
0.21
US06 Highway FE HFET FE
where the various symbols have the same definitions as just described above.
93
-------
Hybrid Vehicles Tested over a Four bag FTP at 75 F
5-Cycle City Fuel Economy Formula
The final 5-cycle city fuel economy would be calculated as follows:
CityFE = 0.905 x
Start FC = 0.33 x
Start FC + Running FC 'where
'(0.76 x Start FueI75 + 0.24 x Start Fuel
4.1
, where
Start Fuel15 = 3.6 x
and
1
1
BaglFE75 Bag3FE,
75
+ 3.9x
Start FueLn = 3.6x
1
1
Bag\FE20 Bag3FE2
1
1
Bag2FE75 Bag4FE
75
where
Running FC =
0.82x
0.48
0.41
0.11
Bag415 FE Bag315 FE US06 City FE
0.18x
0.5
0.5
Bag220FE Bag320FE
•0.133xl.083x[,4/CFC]
where
A/CFC =
1
0.61
0.39
SC03FE (Bag375FE Bag475FE
, where
US06 City FE = fuel economy in miles per gallon over the city portion of the US06 test,
US06 Highway FE = fuel economy in miles per gallon over the Highway portion of the
US06 test,
HFET FE = fuel economy in miles per gallon over the HFET test,
SC03 FE = fuel economy in miles per gallon over the SC03 test.
94
-------
5-Cycle Highway Fuel Economy Formula
The final 5-cycle highway fuel economy would be calculated as follows:
Highway FE = 0.905 x
1
Start FC + Running FC
, where
0 ^/ „ ,\ „„„ ((Q.76xStartFueLs)+(Q.24xStartFueLQ)
StartFC (gallons per mile) = 0.33 x ^ ^—^ 2°Z
where
Start Fuel75 = 3.6 x
and
Start Fuel20 =3.6x
3.9x
1
Bag2FE75
75
1
1
BaglFE20 Bag3FE
20
, and
Running FC = 1.007 x
0.79
0.21
US06 Highway FE HFET FE
+ 0.133 x 0.377
where the various symbols have the same definitions as just described above.
Hybrid Vehicles Tested over a Two-Bag FTP at 75°F
5-Cycle City Fuel Economy Formula for Vehicles Tested Over a 2-BagFTP at 75°F
The final 5-cycle city fuel economy for vehicles tested over a 2-Bag FTP at 75°F would
be calculated as follows:
CityFE = 0.905 x
1
Start FC + Running FC '
where
Start FC = 0.33 x
(0.76 x Start Fuel75 + 0.24 x Start Fuel20)
4.1
where
95
-------
Start Fuel'75 = 7.5x
1
1
Bag\/2FE75 Bag3/4FE75
, where
Start Fuel20 = 3.6x
1
1
BaglFE20 BaglFE
20
, where
Bag y FEX = the fuel economy in miles per gallon of fuel during Bag 1 or Bag 3 of the FTP test
conducted at an ambient temperature of 75 or 20°F.
Bag x/y FEX= fuel economy in miles per gallon of fuel during Bags 1 and 2 or Bags 3 and 4 of
the FTP test conducted at an ambient temperature of 75°F.
Running FC =
0.82x
0.90
0.10
Bag31475 FE US06 City FE
0.18x
0.5
0.5
Bag220FE Bag320FE
0.133xl.083x[^/CFC]
where
A/CFC =
1
1.0
SC03FE (Bag3/475FE
, where
US06 City FE = fuel economy in miles per gallon over the city portion of the US06 test,
US06 Highway FE = fuel economy in miles per gallon over the Highway portion of the
US06 test,
HFET FE = fuel economy in miles per gallon over the HFET test,
SC03 FE = fuel economy in miles per gallon over the SC03 test.
5-Cycle Highway Fuel Economy Formula for Vehicles Tested Over a 2-Bag FTP at 75°F
The final 5-cycle highway fuel economy for vehicles tested over a 2-Bag FTP at 75°F
would be calculated as follows:
Highway FE = 0.905:
1
Start FC + Running FC
, where
96
-------
(0.76 x Start FueL, + 0.24 x Start FueLn)
Start FC = 0.33 x ^ ^ 22Z ^ where
Start FueL, =7.5
75
60
1
1
Bagl/2FE75 Bag3/4FE,
75
, and
Start Fuel'„ =3.6x
1
1
Bag 1FE20 Bag 3 FE20
, and
Running FC = 1.007:
0.79
0.21
OT06 Highway FE HFET FE
0.133x0.377:
where the various symbols have the same definitions as just described above.
b. Alternative 5-cycle Highway Fuel Economy Formula
Beginning with the 2011 model year, manufacturers would be allowed to continue to use
the mpg-based formulae if the available 5-cycle fuel economy estimates indicated close
alignment with the mpg-based formulae. Fuel economy values over all five cycles will be
available for one or more vehicle configurations within each durability data group or basic
engine group. If the 5-cycle fuel economy values for a specific emission data vehicle are no
more than 4% below the mpg-based estimate for city fuel economy and no more than 5% below
the mpg-based estimate for highway fuel economy, all the vehicle configurations represented by
that emission data vehicle would be allowed to use the mpg-based formulae in complying with
the fuel economy label requirements. If the 5-cycle fuel economy values for a specific emission
data vehicle are more than 4% below the mpg-based estimate for city fuel economy and more
than 5% below the mpg-based estimate for highway fuel economy, all the vehicle configurations
represented by that emission data vehicle would be required to use the 5-cycle formulae in
complying with the fuel economy label requirements.
It is possible for the 5-cycle fuel economy values to meet the above criteria for either city
or highway fuel economy, but not the other. If the 5-cycle fuel economy values for a specific
emission data vehicle are more than 4% below the mpg-based estimate for city fuel economy, but
no more than 5% below the mpg-based estimate for highway fuel economy, all the vehicle
configurations represented by that emission data vehicle would be required to use the 5-cycle
formulae in complying with the fuel economy label requirements for both city and highway fuel
economy. All five cycles play a significant role in the 5-cycle city fuel economy formula. Once
the five tests have been performed for the city estimate, there is little reason not to use the same
information to derive the highway fuel economy estimate.
We proposed a different approach for the opposite situation. If the 5-cycle fuel economy
values for a specific emission data vehicle are no more than 4% below the mpg-based estimate
for city fuel economy, but more than 5% below the mpg-based estimate for highway fuel
economy, all the vehicle configurations represented by that emission data vehicle would be
97
-------
allowed to use the mpg-based formulae in deriving the city fuel economy label value. The
highway fuel economy value, however, would be based on an alternative, simplified 5-cycle
formula as opposed to the full 5-cycle highway fuel economy formula. This alternative 5-cycle
highway formula would be based on fuel economy values over the FTP, HFET and US06 tests.
The impact of the SC03 test is relatively small due to the speed adjustment factor of 0.377 and
air conditioning usage factor of 0.133 in the 5-cycle highway fuel economy formula. The impact
of the cold FTP test is small due to the 60 mile trip length assumed for highway driving and the
fact that we do not use the actual cold FTP test results to adjust running fuel consumption for
colder temperatures during highway driving.
This approach requires that we develop a simplified 5-cycle highway fuel economy
formula which is consistent with the full 5-cycle formula. We developed this simplified formula
using estimates of the average impact of the SC03 and cold FTP test results on 5-cycle highway
fuel economy. In both cases, we estimated this average impact by regressing the impact of these
test cycles on the 5-cycle highway fuel economy for the 615 vehicles in our certification
database against fuel economy values which would be available from FTP, HFET and US06
testing.
Regarding the impact of the cold FTP on highway fuel economy, we regressed start fuel
use in highway driving under a mix of ambient temperature against start fuel use in highway
driving at 75°F. As described above, start fuel use in highway driving under a mix of ambient
temperature is as follows:
„_ ^ „ „ (0.76 x Start FueL, + 0.24 x Start FueLn)
Start FC = 0.33 x-^ 2 2°Z where
60
Start Fuel = 3.6x
1 1
and x can be either 20°F or 75°F.
Bag\FEx Bag3FEx
The result of the regression was:
Start FC at ambient = 0.005515 + 1.13637* Start FC at 75°F.
The adjusted r-squared of the regression was very good, over 0.92.
Regarding the impact of SC03 on highway fuel economy, we regressed fuel use due to air
conditioning use in highway driving against several estimates of running fuel use, namely Bags 2
and 3 of the FTP, FIFET and US06. As described above, fuel use due to air conditioning use is
as follows:
AICFC = .
SC03FE \Bag3FE75 Bag2FE
75
98
-------
In the analysis performed for the NPRM, fuel use over US06 showed the highest level of
correlation with air conditioning use. The same was true with the expanded certification fuel
economy database. The result of the regression is:
AICFuelUse = 0.540 + -
0.1357
US'06 Fuel Economy
The adjusted r-squared of this regression is much lower than that for cold start fuel use (0.15).
However, the p-values of both coefficients were less than 0.0000001, and thus, are quite
statistically significant.
These two relationships can be inserted directly into the 5-cycle fuel economy formula.
The result is:
Alternative Highway FE = 0.905 x
1
Start FC + Running FC
, where
(0.005515 + 1.13637 x StartFueL,}
Start FC = 0.33 x ^ £Z where
60.0
Start Fuel75 = 3.6 x
1
1
BaglFE75 Bag3FE
and
75
Running FC =
[l.0 + (0.04x0.18)] >
0.79
0.21
US06 Highway FE HFETFE
0.377x0.133x 0.00540 +
0.1357 ]
US06FEJ
Hybrid gasoline-electric vehicles using this modified 5-cycle highway calculation use one
of the following equations for start fuel, depending upon whether the vehicle is tested on a 4-bag
FTP or a 2-bag FTP.
For a 4-bag FTP:
Start Fuel75 =3.6x
1 1
Bag\FE75 Bag3FE75
+ 3.9x
1 1
Bag2FE75 Bag4FE75
For a 2-bag FTP:
Start Fuel'75 =7.5:
1
1
Bagl/2FE75 Bag3/4FE,
75
99
-------
where the various symbols have the same definitions as just described above.
B. Derivation of the MPG-Based Approach
The 5-cycle fuel economy formulae derived above assume that fuel economy estimates
are available for specific vehicles for all five dynamometer cycles and their respective bags of
emission measurements. As discussed in the preamble to the final rule, these estimates may be
based on fuel economy measurements, or on estimates based on test results from a similar
vehicle. A simplified approach to implementing the 5-cycle formulae is to apply these formulae
to test results on recent model vehicles and develop correlations between the 5-cycle city and
highway fuel economy estimates for these vehicles and their fuel economy over the FTP and
HFET, respectively. This simplified approach is referred to as the mpg-based approach, since
the resultant label adjustment will vary depending on the measured fuel economy (i.e., mpg) of a
vehicle over the FTP and FIFET tests.
The database from which the mpg-based correlations were derived consisted of 615
2003-2006 model year vehicles, including 14 hybrids and one diesel vehicle. All vehicles had
been tested over all five certification test cycles. In most cases, bag-specific fuel economy
measurements were also available, but in some cases they were not. In the latter cases, we
estimated FTP bag-specific fuel economy using relationships between bag and whole cycle fuel
economy which were developed from those vehicles with bag fuel economy data. The following
table shows the relationships between bag and cycle fuel economy from our 5-cycle fuel
economy database for the standard and cold FTP.
Table III.B-1. Ratio of FTP Bag to Cycle Fuel Consumption
No. of Vehicles
Bagl
Bag 2
Bag 3
Standard FTP
Mean
Standard Deviation
Coefficient of Variation
389
1.047
0.040
3.8%
1.036
0.029
2.8%
0.897
0.036
4.0%
Cold FTP
Mean
Standard Deviation
Coefficient of Variation
330
1.171
0.055
4.7%
1.013
0.028
2.8%
0.855
0.036
4.2%
The 5-cycle formulae also require separate fuel economy estimates for the city and
highway portions of US06. These measurements have not been taken on a regular basis. In the
Draft Technical Support Document to the NPRM, we analyzed US06 city and US06 highway
fuel economy data for 85 vehicles which was available. There we found that the fuel economy
of the US06 city bag averaged 68% of that over the entire US06 cycle for conventional vehicles,
and 77% for two hybrid vehicles. We also found that the fuel economy of the US06 highway
bag averaged 116% of that over the entire US06 cycle for conventional vehicles, and 109% for
two hybrid vehicles.
100
-------
For the NPRM, we projected the impact of the 5-cycle approach by applying the
relationships for conventional vehicles to all vehicles, including hybrids. For the FRM, we
believe that it would be more accurate to apply the relationships for the two hybrids that were
tested to all the hybrid vehicles in the certification database. While only two hybrids were tested
over a two-bag US06 test, the fact that fuel economy over the US06 city bag was closer to that
over the entire US06 cycle than with conventional vehicles is very consistent with the effect of
hybrid technology on fuel economy. That is, hybrid technology is generally more effective
during lower speed, stop and go driving than at consistently high vehicle speeds. Which
relationship is used to project US06 city and highway fuel economy values has no effect in the
future for vehicles whose label values are set using the 5-cycle formulae. In this case, US06 city
and highway fuel economy values will be measured, not estimated. However, the projections
made here can affect the mpg-based equations, as these equations are based on projected 5-cycle
fuel economy values. These projections for hybrids with the highest fuel economy values are
particularly important, as these vehicles can affect the shape of the mpg-based equations at high
fuel economy values. Fortunately, the two hybrids for which we have US06 city and highway
fuel economy testing are hybrids with very high fuel economy values (a Prius and a Civic
hybrid). Thus, using the relationships between US06 city, US06 highway and US06 fuel
economy values based on the testing of these two vehicles to all hybrids in the certification
database should be most accurate in the range of fuel economy where the mpg-based equations
are most affected by hybrids.
One additional adjustment was made to Cold FTP fuel economy values of all vehicles.
This adjustment is related to the new requirement that the heater or defroster be turned on during
the Cold FTP test. In order to estimate the impact of this change on fuel economy, EPA tested
two conventional vehicles and two hybrid vehicles at 20°F with the heater turned on and off.37
The results are shown in Table III.B-2.
101
-------
Table III.B-2. Effect of Heater/Defroster Use on Cold FTP Fuel Use
Bagl
Bag 2
Bag 3
FTP
Fuel Economy: Heater off (mpg)
Odyssey
Trailblazer
Prius
Civic Hybrid
13.5
11.4
32.4
30.2
16.3
13.8
50.4
38.2
18.5
15.6
39.0
40.8
16.2
13.6
44.1
38.2
Fuel Economy: Heater on (mpg)
Odyssey
Trailblazer
Prius
Civic Hybrid
13.0
11.2
28.5
26
15.2
13.2
34.1
29.1
17.4
15.2
37.9
34.6
15.3
13.2
36.2
31.7
Fuel Consumption: Heater off (gallon per 100 miles)
Odyssey
Trailblazer
Prius
Civic Hybrid
7.42
8.79
3.086
3.311
6.12
7.25
1.984
2.618
5.42
6.40
2.564
2.451
6.16
7.33
2.268
2.618
Fuel Consumption: Heater on (gallon per 100 miles)
Odyssey
Trailblazer
Prius
Civic Hybrid
7.67
8.89
3.509
3.846
6.58
7.56
2.933
3.436
5.76
6.59
2.639
2.890
Increase in Fuel Consumption (%)
Odyssey
Trailblazer
Average
Prius
Civic Hybrid
Average
3.5%
1.2%
2.3%
13.7%
16.2%
14.9%
7.5%
4.4%
6.0%
47.8%
31.3%
39.5%
6.3%
2.9%
4.6%
2.9%
17.9%
10.4%
6.55
7.57
2.762
3.155
6.4%
3.2%
4.8%
21.8%
20.5%
Increase in Fuel Consumption after adjusting for test procedure differences (%)
Odyssey
Trailblazer
Average
Prius
Civic Hybrid
Average
2.3%
1.2%
1.7%
13.7%
16.2%
14.9%
3.8%
2.2%
3.0%
47.8%
31.3%
39.5%
3.2%
1.4%
2.3%
2.9%
17.9%
10.4%
N/A
N/A
N/A
21.8%
20.5%
The test procedure used in this testing differs from that being promulgated in the final
rule. In this testing, the defroster was turned on to the maximum position immediately at the
start of the test and held there throughout the test. This is a more severe setting than we are
promulgating, where the start of defrosting is being delayed two minutes and then reduced to a
more moderate setting during Bags 2 and 3. To account for these differences, we reduced the
impact of defrosting on fuel consumption for conventional vehicles. These reduced impacts are
shown in the final section in Table III.B-2. Specifically, we reduced the adjusted the impacts
during Bags 2 and 3 by a factor of two. We also reduced the impact on Bag 1 fuel consumption
102
-------
for the Odyssey to reflect the relative impact for the Odyssey on Bags 2 and 3 compared to those
for the Trailblazer. The Bag 1 impact for the Trailblazer was not adjusted, because it was tested
with the delay in defroster start-up being promulgated. Thus, the fuel consumption of
conventional vehicles over Bags 1, 2, and 3 of the Cold FTP test in our 5-cycle certification
database were increased by 1.7%, 2.9%, and 2.2%, respectively, to account for defroster use in
future testing.
As can be seen from Table III.B-2, the effect of defroster and heater use was much larger
for the two hybrids than for the two conventional vehicles. (The temperature control was also
turned to hot when the defroster was turned on during the hybrid testing.) This greater impact is
likely due to the fact that operating the heater prevents the engine from shutting off during
certain driving modes, like idling and decelerations. The same effect likely occurs when the
heater is turned on without the defroster. Since drivers regularly use their heater under colder
ambient conditions, this effect is occurring currently in-use. While the final Cold FTP test
procedure is more moderate than that used in the above testing of hybrids, we believe that the
great majority of the impact on hybrid fuel consumption was due to the elimination of the engine
shut-off feature, as opposed to the specific defroster/heater setting. Thus, we did not believe that
the retesting of these vehicles with the final test procedure would produce significantly lower
fuel consumption impacts. We expect that auto manufacturers will modify their hybrid designs
in the future to reduce this impact. However, as the mpg-based equations will be applied to
vehicles as early as the 2008 model year, we believe that they should reflect current technology
as much as possible. Also, hybrids which reflect improved technology in this regard can utilize
the 5-cycle formulae, especially since hybrid models are always tested over all five dynamometer
cycles during certification due to their unique features. Thus, we are applying the average
impacts on Bag 1, 2, and 3 fuel consumption, as measured in the above test program, to hybrid
fuel consumption in our 5-cycle certification database.
Using the fuel economy values which are now available for all bags and cycles for all 65
vehicles, we calculated 5-cycle city and highway label values. We then developed relationships
between the 5-cycle city and highway label values and FTP and HFET fuel economy values,
respectively, using the least squares regression function in Excel. As we did for the NPRM, we
performed these regressions in terms of fuel consumption (i.e., gallons per mile or the inverse of
fuel economy). For 5-cycle city fuel economy, the best fit relationship was:
5-cycle city FE = 17(0.003259+ 1.18053 /FTP FE )
The adjusted r-squared for this regression was 0.990. For 5-cycle highway fuel economy, the
best fit relationship was:
5-cycle highway FE = 17(0.001376+ 1.3466 7 HFET FE )
The adjusted r-squared for this regression was again slightly worse, 0.952. Figures III-8 and 9
show the relationship between the inverse of 5-cycle city and highway fuel economy (i.e., fuel
consumption) versus the inverse of FTP or FIFET fuel economy. The first graph shows city fuel
consumption, while the second shows highway fuel consumption.
103
-------
Figure III-8. 5-Cycle City Versus FTP Fuel Consumption
LU
LL
SI
.' ro
"O O)
•5~
ro
O)
0.15
0.10
0.05
5-Cycle Highway Vs. HFET Fuel Consumption
0.01 0.02 0.03 0.04 0.05 0.06
Inverse of HFET Fuel Economy (gal/mi)
0.07
0.08
Figures III-10 and 11 show the relationship between 5-cycle city and highway fuel
economy (i.e., fuel consumption) versus of FTP or FIFET fuel economy. The first graph shows
city fuel consumption, while the second shows highway fuel consumption. As can be seen by
comparing the two sets of graphs, the relationships are linear in terms of fuel consumption, but
become curved in terms of fuel economy.
104
-------
Figure 111-10. MPG-Based City Fuel Economy
MPG-Based City FE
el Economy
5-Cycle City
10 20 30 40 50
FTP Fuel Economy (mpg)
60
70
Figure III-ll. MPG-Based Highway Fuel Economy
MPG Based Highway FE
10 20 30 40 50 60
Current HFET Fuel Economy (mpg)
70
The standard error of the difference between the mpg-based equations and the 5-cycle
fuel economies are 0.5 mpg and 1.15 mpg for city and highway fuel economy, respectively.
These differences represent 3% of the average 5-cycle city fuel economy and 5% of the average
5-cycle highway fuel economy, respectively. Thus, while the mpg-based equations are able to
reflect much of the difference in fuel economy represented by the 5-cycle formulae, differences
105
-------
between the fuel economy of individual vehicles on the order of 0.5-1.1 mpg are muted by the
mpg-based approach.
C. Variability in Onroad Fuel Economy
As described in the preamble to the final rule, EPA is proposing to continue to set the city
and highway mpg estimates at a level that reflects average fuel economy. However, we desire
the fuel economy label to indicate the range of onroad fuel economy that drivers might
experience. Therefore, it is important to understand the variability of onroad fuel economy.
We begin with a review of the work done in this area for the 1984 fuel economy
adjustment rule. At that time (circa 1982), EPA conducted a systematic review of the onroad
fuel economy experienced by over 40,000 drivers compared to the EPA fuel economy labels for
each vehicle.38 Vehicles were separated into 3 categories: 1) primarily city driven, 2) primarily
highway driven, and 3) mix of driving. The percentage differences between onroad and either
the EPA city or highway fuel economy label was determined for the first two groups. The results
were generally normally distributed. The results for city driven vehicles are depicted in Figure
111-12.
Figure 111-12. Onroad FE Versus Pre-1984 EPA City Label for City Driven Cars
In-Use
9n°/
£.\J /O
y> 1 co/
»- I O /o
0
.>
J^ 1 n%
fi 1 U /u
M—
5 5%
Oo/
FE Before Adjustment: 1984 Label Rule
61% below 90% of
label FE
4 S~
/
/
^^/
70 1
-80%
\
\ 4% above 1 1 0% of
\ label FE
\
+ 10% V
I I I I I
-60% -40% -20% 0% 20% 40%
Difference from EPA City Label FE (FTP)
The analysis of the onroad fuel economy by primarily city drivers found that average onroad fuel
economy was 90% of the then current EPA city label (i.e., a 10% shortfall on average). At that
time, the city label value was simply the fuel economy over the FTP. As shown in Figure 111-12,
only 4% of all drivers achieved an onroad fuel economy more than 10% of their vehicle's city
fuel economy label. And 61% achieved an onroad fuel economy less than 90% of their vehicle's
city label.
106
-------
Assuming a normal distribution of the onroad fuel economy of any specific vehicle, as
was done in the 1984 label adjustment rule, it is possible to use this information to calculate a
coefficient of variation for the difference between onroad fuel economy and EPA's city label.
As shown in Figure III-12, 4% of all predominantly city drivers achieved more than 110% of the
EPA city fuel economy label. For a normal distribution, 4% of the population exceeds the mean
of the population plus 1.75 times the standard deviation. In this case, since the population is in
percentage terms (onroad fuel economy divided by label fuel minus 1.0), the standard deviation
is equal to the coefficient of variation in the ratio of onroad fuel economy to EPA label value.
Likewise, 61% of all predominantly city drivers achieved less than 90% of the EPA city fuel
economy label. Another way to put this is that 39% of all drivers achieved more than 90% of the
EPA city fuel economy label. For a normal distribution, 39% of the population exceeds the
mean of the population plus 0.279 times the coefficient of variation. Thus, 20% of the EPA city
fuel economy label (110% minus 90%) represents 1.471 (1.75 minus 0.279) times the coefficient
of variation. The coefficient of variation for this distribution is therefore 13.6% (20% divided by
1.471).
The comparison of onroad fuel economy and EPA highway label for highway driven
vehicles yielded similar results. In the case of onroad fuel economy during highway driving,
only 8% of all predominantly highway drivers achieved an onroad fuel economy more than
110% of their vehicle's highway fuel economy label. Of all predominantly highway drivers,
34% achieved an onroad fuel economy within 10% of their vehicle's highway label and 58%
achieved an onroad fuel economy below 90% of their vehicle's highway label. These
percentages apply prior to the 22% downward adjustment to the highway fuel economy label
implemented in that rulemaking.
Again for a normal distribution, 8% of the population exceeds the mean of the population
plus 1.405 times the coefficient of variation. Likewise, 42% (8% plus 38%) of all drivers
achieved more than 90% of the EPA highway fuel economy label. For a normal distribution,
42% of the population exceeds the mean of the population plus 0.202 times the coefficient of
variation. Thus, 20% of the EPA highway fuel economy label (110% minus 90%) represents
1.203 (1.405 minus 0.202) times the coefficient of variation. The coefficient of variation for this
distribution is therefore 16.6% (20% divided by 1.203).
The goal of the final 10% and 22% adjustments was to move the average onroad fuel
economy closer to either the city or highway label value, as applicable. These adjustments,
however, do not affect the underlying variability of the data. They increase the coefficients of
variation slightly, because they reduce the denominator (the EPA label value) by 10% or 22%,
respectively. They primarily shift the distribution over by 10% or 22%. However, EPA did
analyze the effect of more complex adjustments, which depended on the several vehicle factors,
such as front versus rear wheel drive, manual or automatic transmission, gasoline or diesel
engine, etc. Applying this more complex system of adjustments reduced the variability in onroad
versus EPA label fuel economy somewhat. For city driven vehicles, the coefficient of variation
decreased to 13.1%, while that for highway driven vehicles decreased to 14.1%. Since the 5-
cycle formulae basically adjust fuel economy in a similar fashion (i.e., some vehicles receive
more of an adjustment than others), we believe that it these somewhat smaller coefficients of
variation are more indicative of what drivers would experience with vehicles labeled using the 5-
107
-------
cycle formulae. Thus, we will use a coefficient of variation of 13-14% to determine an
adjustment that would convert a mean fuel economy into a 25th percentile fuel economy. As
discussed above, 75% of the population of a normal distribution exceeds the mean minus 0.675
times the standard deviation (or coefficient of variation in this case). Applying this factor to our
estimate of the coefficient of variation of 13-14% yields an offset of 9-10%.
Oak Ridge National Laboratory (ORNL), sponsored by the Department of Energy, has
recently begun a program where drivers can submit their own fuel economy measurements via
the Internet.39 The program is commonly referred to as "Your MPG." The Your MPG data are
similar in nature to the much larger databases analyzed for the 1984 label adjustment rule.
Drivers measure their own fuel economy and provide a perceived split of their driving into city
and highway categories. The strength of this type of data is the fact that the vehicle is being
operated by the owner or regular driver in typical use. The weaknesses are the unknown
representativeness of the sample, the unknown nature of the technique used by the owner/driver
to measure fuel economy and the short time period over which fuel economy is generally
assessed (e.g., as short as a tank full of fuel or two). In the particular case of the ORNL
database, its current size is still small (8180 estimates of fuel economy for 4092 vehicles)
compared to those available in 1984, though it is growing daily.
We compared the fuel economy estimates submitted to the ORNL website with each
vehicle's fuel economy label. We combined the city and highway labels using each driver's
estimate of the percentage of their driving that was city-like and highway-like. If a driver did not
provide an estimate of the breakdown of their driving pattern, we assumed that their driving was
55% city and 45% highway. We calculated the percentage difference between the onroad fuel
economy and the current composite EPA label value. (A more detailed discussion of these
estimates is presented in Chapter II of this Final Technical Support Document.) The average
difference for more than 7300 individual fuel economy estimates was -1.4%, meaning that
onroad fuel economy was just slightly lower than the composite EPA label value using the split
of city and highway driving estimated by the driver. This metric is analogous to those presented
above from the early 1980's, as vehicles were segregated then into those with predominantly city
or highway driving. The standard deviation in this percentage difference was 13%, very
consistent with the estimates derived above.
Another source of onroad fuel economy data is the recent testing of over 100 vehicles in
the Kansas City area. Valid onroad fuel economy measurements were obtained for roughly one
day of driving from roughly 100 vehicles. The average onroad fuel economy was 30.4 mpg,
while the average composite EPA label value was 31.4 mpg. The standard deviation of the
percentage difference between onroad fuel economy and EPA composite fuel economy was
14%. This is only slightly higher than the estimates from the early 1980's and the Your MPG
website. We would have expected a larger variability than these other sources for two reasons.
One, we did not segregate vehicles into primarily city and highway driving categories or account
for the predominance of one or the other type of driving. Two, driving can vary significantly
from day to day. With only one day's worth of driving measured, the variability in fuel economy
would be expected to be much higher than if a week or two of driving were assessed. At the
same time, all of these vehicles were located within a single metropolitan area, so their driving
did not reflect much urban/rural diversity, nor the diversity likely present between urban areas.
108
-------
And only 100 vehicles were assessed, including a unrepresentatively high number of hybrids.
(The standard deviation only increased to 15% when the hybrids were excluded.) Thus, the
small sample may be a factor.
Finally, we evaluated the variability in the Consumer Report fuel economy measurements
compared to the current EPA city and highway label values. (See section II.B. 1 for a more
complete discussion of the Consumer Report fuel economy estimates.) The standard deviations
of the percentage difference between the Consumer Report and current EPA fuel economy were
8% for city and 7% for highway. These figures are lower than the 13-16% value found during
the 1985 label adjustment rule and in the ORNL Your MPG database. However, Consumer
Report adjusts their fuel economy measurements to represent a single ambient temperature and
all of their testing follows the same road routes. Thus, both the ambient conditions and the
driving patterns are much more consistent than those experienced by the population of drivers in
the U.S. The fact that variability is still as high as 7-8% tends to confirm that the 13%
assumption described above is reasonable.
All of the above estimates of the standard deviation in the percentage difference between
onroad and EPA label fuel economy fall in the range of 13-16%. The more recent estimates fall
towards the lower end of this range. Thus, we will select 13% as the best point estimate of
variability. Multiplying the standard deviation by 67.5% produces an offset from mean fuel
economy which should encompass an additional 25% of drivers. The 5-cycle formulae derived
in section III. A and the mpg-based formulae derived in section III.B, including the 11%
downward adjustment for non-dynamometer effects, represent estimates of mean onroad fuel
economy. All of the inputs to the 5-cycle formulae are based on national averages of the relevant
parameter. Thus, reducing these estimates by 9% (i.e., multiplying them by 0.91) would convert
these figures from the mean fuel economy achieved on the road to the 25th percentile of the range
of onroad fuel economies achieved. This would produce a label value which would be achieved
or exceeded by 75% of all drivers. Figure 111-20 depicts this graphically.
109
-------
Figure 111-13. Variability in Onroad FE
Variability in Onroad FE
9(W
£.\) /O
5" 1 ^%
I_ I O /O
0)
_>
i 1 n%
M-
o
^s ^%
o^ O /O
0%
-4(
Median: 0%
25thpercentile:-9%
/
25% of /
drivers /
, /
' X
50% of N
drivers
th percentile:-18%
\
\ 75% of
\. drivers
\ fc
)% -20% -9% 0% 9% 20% 40%
Onroad FE minus Composite EPA Label FE
The frequency distribution of onroad fuel economy shown in Figure 111-13 assumes that the 5-
cycle formulae developed above match onroad fuel economy on average In this case, 25% of
drivers achieve an onroad fuel economy below 90% of their label value and 25% achieve an
onroad fuel economy above 110% of their label value. Half of all drivers achieve an onroad fuel
economy within plus or minus 9% of their label value.
The adjustment appropriate to convert mean fuel economy estimates to those
representative of other percentiles is straightforward. For example, the 10th and 90th percentile
fuel economy values would be 1.28 times the coefficient of variation off of the mean value, or
17% downward and upward adjustments from the mean. The 5th and 95th percentile fuel
economy values would be 1.645 times the coefficient of variation off of the mean value, or 21%
downward and upward adjustments from the mean.
D. Impact of the 5-Cycle and MPG-Based Formulae on Fuel Economy
Labels
The impact of the final rule on city and highway fuel economy label values was assessed
using the same database of 615 late model year vehicles used to develop the mpg-based
adjustments above. It should be noted that these data are not sales weighted. In fact, most
specific vehicle models included in the database are "worst case" for emission performance
purposes within their model group, as this is currently one of the criteria used by EPA to
determine which vehicles should be tested over the US06, SC03 and cold FTP tests for emissions
compliance. Table III.D-1 presents the results of this comparison for all 615 vehicles, as well as
various sub-sets of vehicles.
110
-------
Table III.D-1. Current and 5-Cycle Label Fuel Economies by Model Type1
Current City
5-Cycle City
Current Highway
5-Cycle Highway
Conventional Vehicles
Large car
Midsize car
Minivan
Pickup
Small car
Station wagon
SUV
Van
All conventional
All hybrids
Diesel (one midsize car)
All vehicles
15.7
20.5
17.4
15.1
20.7
20.3
16.8
12.5
18.6
41.6
26.2
19.1
13.8
17.8
15.2
13.2
18.1
17.6
14.6
10.9
16.2
32.0
22.7
16.4
21.9
27.9
23.6
18.9
27.3
26.6
21.6
16.0
24.6
40.6
35.3
24.9
19.7
25.6
20.9
17.2
25.3
23.5
19.5
14.3
22.4
36.8
31.4
22.7
The next table shows the effect of the 5-cycle formulae on conventional gasoline fueled
vehicles with particularly high or low fuel economy.
Table III.D-2. Current and 5-cycle Label Fuel Economy by Propulsion System
Hybrids
Diesel
City
Current
(mpg)
42.7
26.2
5-Cycle
(mpg)
33.0
23.4
Percent
Change
-22.3%
-10.7%
Highway
Current
(mpg)
42.8
35.3
5-Cycle
(mpg)
36.9
32.0
Percent
Change
-12.9%
-9.3%
Conventional Gasoline-Fueled Vehicles
12 Highest FE
12 Lowest FE
Average
30.9
10.2
18.6
26.9
9.5
16.5
-12.9%
-6.9%
-10.8%
36.6
14.8
24.6
34.0
14.8
22.8
-6.9%
-0.2%
-7.4%
As can be seen from Tables III.D-1 and III.D-2, use of the 5-cycle formulae would reduce both
current city and highway fuel economy label values. For conventional vehicles, city and
highway fuel economy values would be reduced an average of 13% and 9%, respectively. For
higher than average fuel economy vehicles, the reduction in city fuel economy would be slightly
higher, while for lower than average fuel economy vehicles, the reduction in city fuel economy
would be slightly lower. The change in highway fuel economy is essentially independent of
current highway fuel economy.
1 These figures may differ from those that might appear on the EPA fuel economy window sticker, as they
have not undergone any sales weighting. They have been derived by applying fuel economy label formulae (e.g.,
0.9 times FTP fuel economy) to fuel economy test results for individual vehicles.
Ill
-------
The impact on hybrid vehicles would be significantly greater for city fuel economy,
averaging a 23% reduction. However, the reduction in highway fuel economy would be the
same as for conventional gasoline-fueled vehicles. This greater impact occurs primarily because
a number of the fuel efficient aspects of hybrid vehicles produce their maximum benefit under
conditions akin to the FTP tests, and are somewhat less beneficial during aggressive driving,
colder ambient temperatures and when the air conditioner is turned on. The impacts of the 5-
cycle formulae on the single diesel vehicle in the database are very similar to those for
conventional gasoline fueled vehicles.
The impact of the mpg-based formulae would be very similar on average to those shown
in Tables III.D-1 and III.D-2 above for conventional vehicles, gasoline-fueled and diesel. This is
not surprising for conventional gasoline fueled vehicles, since the mpg-based formulae are based
essentially on the average results of the 5-cycle formulae and the vast majority of the vehicles in
the database are conventional gasoline-fueled vehicles. The 5-cycle fuel economy values for the
one diesel in the database also fall very near the mpg-based curves. However, the impact of the
mpg-based formulae on the current city fuel economy label values for hybrids would vary
significantly from the 5-cycle values. Basically, the impact on hybrids would reflect that of
conventional vehicles with the same current fuel economy levels. The mpg-based regressions
therefore, represent essentially the impact of the 5-cycle formulae on conventional vehicles,
which is less than that for hybrids. The impact on the mpg-based formulae for hybrids is shown
in Table III.D-3 below. The impact on the city fuel economy label is still somewhat higher (-
18%) than the average for conventional gasoline vehicles, because the average FTP city fuel
economy for hybrids is higher than that for even the top 12 conventional gasoline fueled vehicles
(-15%). The impact of the mpg-based formula on the highway label value for hybrids is 10%, or
just slightly higher than that for conventional gasoline-fueled (9%). With only one diesel vehicle
in the database, no general observations about this engine type can be made.
Table III.D-3. Effect of MPG-Based Formulae on City and Highway Fuel Economy
Hybrids
Conventional
City
Current
(mpg)
41.6
18.6
MPG-
Based
34.1
16.2
Percent
Change
-18%
-13%
Highway
Current
(mpg)
40.6
24.6
5-Cycle
(mpg)
36.8
22.4
Percent
Change
-10%
-9%
In addition to looking at the overall change in fuel economy estimates for all vehicles in
the database, we also focused on those manufacturers responsible for the majority of sales in the
U.S. This approach may better reflect the changes likely to be seen by the majority of
consumers. In effect, Table II-3 above includes vehicles by Aston Martin and Rolls-Royce in the
percent change, and these vehicles are weighted equally with cars made by GM, Ford,
DaimlerChrysler, and other top-selling manufacturers. According to Autodata Corporation, the
seven manufacturers with the greatest U.S. market share account for more than 90 percent of
U.S. sales. Table II.D-4 shows these manufacturers, their 2005 U.S. market share, and the
average percent change in city and highway fuel economy estimates for each of these
manufacturers as represented in our database. It is important to note, however, that these
112
-------
estimates are not intended to represent or include the entirety of a manufacturer's product line,
and should not be interpreted as such. These estimates are derived from our database of 615 test
vehicles for which data on all five emission and fuel economy test procedures is available, and
because of differing ways in which manufacturers test their vehicles and submit data to EPA, the
database may not reflect the range of makes and models similarly across manufacturers.
Table II.D-4.
Effect of New Methods on Fuel Economy Estimates for Major
Manufacturers
Manufacturer
General Motors
Ford Motor Co.
DaimlerChrysler
Toyota
Honda
Nissan
Hyundai
Average
2005 U.S. Market
Share (%)*
25.9
17.9
14.9
13.7
8.9
6.1
2.9
Average Change in
City Fuel Economy
Estimate
-10%
-12%
-10%
-11%
-13%
-11%
-13%
-12%
Average Change in
Highway Fuel
Economy Estimate
-11%
-10%
-11%
-7%
-7%
-7%
-8%
-8%
* Source: Autodata Corp., Woodcliff Lake, New Jersey.
The following table shows the effect of the various aspects of the 5-cycle formulae on 5-
cycle city and highway fuel economy relative to fuel economy over the FTP and HFET,
respectively.
Table III.D-5. Effect of Various Factors on 5-cycle Fuel
Economy
Start Fuel
Cold Temp
Start
Cold Temp
Running
A/C
Running
City FE
Conventional
1.2%
1.4%
2.1%
3.2%
3.8%
Hybrid
1.0%
4.5%
10.1%
5.5%
4.0%
Highway FE
Conventional
0.6%
0.7%
2.8%
16.8%
Hybrid
0.7%
0.7%
3.7%
20.9%
E. Sensitivities and Uncertainties in the 5-Cycle Fuel Economy
Formulae
In this section, we evaluate the impact of a series of alternative assumptions and
approaches to developing the 5-cycle fuel economy formulae. The organization of this section
basically follows that of section II. A above. Alternatives regarding start fuel use are addressed,
followed by those affecting running fuel use at 75°F, fuel associated with air conditioning use
and the impact of colder temperatures.
113
-------
1. Start Fuel Use
There are five key factors which affect start fuel use on a gallon per mile basis. These
are: 1) the distribution of starts as a function of soak time and time of day, 2) the effect of soak
time and ambient temperature on start fuel use, 3) fuel consumption associated with a cold start,
4) average trip length, and 5) heater/defroster use. Each of these factors will be addressed below
in turn.
Regarding the distribution of starts as a function of soak time and time of day, we know
of no other existing estimate which is as representative and extensive as the Baltimore-Spokane
data used to develop the MOBILE6.2 and Draft MOVES2004 distributions. The recent testing in
Kansas City covers nearly as many vehicles. However, the number of days of driving assessed
for each vehicle is well below that achieved in Baltimore and Spokane. Georgia Tech has been
studying driving patterns in Atlanta via vehicle instrumentation for some time. The amount of
data which they have collected to date exceeds that obtained in Baltimore and Spokane. EPA
has begun to obtain the start and trip related information from this study. However, the work
involved is considerable and the results are not yet available.
Regarding the effect of soak time and ambient temperature on start fuel use, the
correlations used in Draft MOVES2004 and EMFAC2000 are both recent and addressed
essentially all data available at that time. We know of no other data addressing the effect of soak
time on start fuel use. Some information regarding the impact of ambient temperature on start
fuel use at 50°F is available from California certification testing. California requires a small
number of vehicles to have their emissions tested at 50°F each year. We obtained this data for
nine conventional Honda vehicles, two Honda hybrids and two Toyota hybrids.
The nine conventional Honda vehicles showed slightly lower sensitivity of start fuel use
to temperature than that estimated in Draft MOVES2004. At 50°F and 20°F, the nine vehicles
showed 1.51 and 2.30 times the start fuel use as at 75°F, respectively. (Start fuel use being
defined as 3.59 miles times the difference in fuel consumption in Bags 1 and 3 of the FTP.)
Draft MOVES2004 estimates ratios of 1.63 and 2.75 at 50°F and 20°F, respectively. The 90%
confidence intervals around the means of the Honda data were roughly half of the difference
between the means for the Honda vehicles and the Draft MOVES2004 estimates (i.e., 0.07 and
0.20, respectively). We do not have estimates for the confidence limits around the Draft
MOVES2004 estimates. However, given the greater number of vehicles tested, the confidence
intervals around the Draft MOVE2004 projection are probably smaller. Thus, while 1-4
individual Honda vehicles showed a greater sensitivity to temperature than projected by Draft
MOVES2004, on average the Honda vehicles are less sensitive at a 90% confidence level.
We re-estimated the weighting for start fuel use at 20°F and 75°F using the temperature
sensitivity of the average Honda conventional vehicle. We modified the coefficients of the Draft
MOVES2004 equation for the ratio of start fuel use as a function of temperature to match the
ratios of 1.51 and 2.30 at 50°F and 20°F, respectively. The resulting equation was:
Start fuel at T / Start fuel use at 75°F = 1 - 0.0170 * ( T - 75 )+ 0.00013 * ( T - 75 )2
114
-------
Using this equation, we found that the weighting for start fuel use at 20°F and 75°F changed in
the third decimal place, but remained the same when rounded to two decimal places (0.24 and
0.76, respectively). This is likely the result of the fact that the Honda vehicles were less sensitive
at both 50°F and 20°F. Thus, the effect of lower temperature sensitivity at 50°F is appropriately
reflected in the 5-cycle formulae by the lower start fuel use measured during the cold FTP at
20°F.
All four hybrid vehicles showed greater sensitivity to temperature at 20°F than projected
by Draft MOVES2004, but only three showed greater sensitivity at 50°F. Table III.E-1 shows
the temperature sensitivities of these four vehicles.
Table III.E-1. Sensitivity of Hybrid Start Fuel Use to Ambient Temperature
Vehicle
Honda Insight
Honda Accord
Toyota 2004 Prius
Toyota RX 400H
Draft MOVES2004
Ratio of Start Fuel Use at Temperature X to 75°F
X = 50°F
1.99
2.02
1.45
2.55
1.63
X = 20°F
3.68
2.93
4.40
4.68
2.74
Weight of Start Fuel
Use at 20°F
0.24
0.30
0.14
0.26
0.24
As can be seen, the calculated 20°F cold start weights range from 0.14-0.30 for the four hybrids,
compared to 0.24 based on Draft MOVES2004. On average, the results for the four hybrids
essentially match that based on Draft MOVES2004. Individually, the Prius and the RX 400H
data produce cold temperature weights very similar to that based conventional vehicles. The
Honda Accord data produces a greater weight for cold start fuel use at 20°F, due to the fact that
its cold start fuel use at 50°F is high relative to that at 20°F. The opposite is true for the 2004
Prius. Using a weight of 0.30 for cold start fuel use at 20°F for the Accord hybrid would reduce
its 5-cycle city fuel economy by 0.1 mpg from 21.6 to 21.5 mpg. Using a weight of 0.14 for cold
start fuel use at 20°F for the 2004 Prius would increase its 5-cycle city fuel economy by 0.5 mpg
from 43.6 to 44.1 mpg. In both cases, 5-cycle highway fuel economy would be unaffected.
Based on this limited data, it appears unlikely that uncertainty in the effect of ambient
temperature on start fuel use would significantly affect city 5-cycle fuel economy. It is certain to
have no effect on 5-cycle highway fuel economy, due to the extremely low contribution of start
fuel use in highway driving. Hybrids would likely show the greatest variability in this area, due
to the greater number of technological factors that could be affected. Even for these vehicles, the
5-cycle fuel economy for the vehicle reflecting the greatest difference in temperature sensitivity
was only changed 1%.
Regarding fuel consumption associated with a cold start, the primary issue is the
assumption that the vehicle is fully warmed up by the end of Bag 1 of the FTP. We are not
aware of any evidence that this is not the case at 75°F. As discussed in section III. A.4 above,
there is some evidence that the vehicle is still warming up after Bag 1 at 20°F. However, if some
of the difference between Bag 2 fuel consumption at 20°F and 75°F was due to continued vehicle
warm up, then the difference after the vehicle was fully warmed up would be commensurately
115
-------
smaller. In general, the second change would tend to mitigate the first. We assessed the
sensitivity of 5-cycle city fuel economy to just the first change by assuming that the difference in
fuel consumption (in gallons) between Bag 2 at 20°F and 75°F was associated with the cold start.
Doing so decreased 5-cycle fuel economy on average for non-hybrid vehicles by 0.16 mpg, or
1%. City fuel economy was reduced by 1.9 mpg for hybrids, or 5.7%. Reducing the impact of
cold temperature on running fuel use would reduce this impact. As discussed in section III.E.5
below, reducing the incremental running fuel use at 20°F by 60% would increase city fuel
economy for non-hybrid vehicles by 0.7% and by 4.5% for hybrids. Thus, the net effect of this
shift of fuel use from running fuel use to start fuel use is to decrease city fuel economy by 0.3%
for non-hybrids and 1.2% for hybrids.
One additional uncertainty regarding cold start fuel consumption involves the testing of
hybrids. Most current hybrid designs include a sizeable battery with which to store energy from
braking, provide launch power after extended idles, etc. Current EPA test procedures require
that hybrids undergo a four-bag FTP, the fourth bag being a repeat of Bag 2. The state of battery
charge is required to be the same at the beginning of the FTP and the end of the four bags.
However, the state of battery charge need not be the same at the beginning and end of Bag 1, nor
the beginning and end of Bag 3. Thus, the possibility exists that a portion of the difference in
fuel consumption between Bags 1 and 3 is related to a change in battery charge, which may not
occur on the road. One contributing factor towards this possibility is the fact that the driving
pattern of Bags 1 and 3 is not representative of driving immediately following an engine start.40
The speeds of the second hill in Bags 1 and 3 contain too much high speed driving. This affects
the rate of engine warm-up for all vehicles. But it could also affect the net change in battery
charge of hybrids relative to that occurring on the road.
As was the case with potential vehicle warm up during Bag 2, any difference in battery
use between Bags 1 and 3 should reverse in Bags 2 and 4. Thus, if the indicated cold start fuel
use is unrepresentatively high or low, the change in running fuel use should be in the opposite
direction. Given the difference in trip length and bag weights between the FTP and the 5-cycle
city formula, the opposing differences do not necessarily balance exactly. However, the net
effect is likely to be much smaller than the effect of a change in battery charge on cold start fuel
use.
We examined the potential impact of a change in battery capacity in Bag 1 relative to Bag
3 using two hybrids in our 5-cycle fuel economy database: a Honda Civic hybrid and a Toyota
Prius. In both cases, we subtracted 0.005 gallons per mile from the fuel consumption in Bag 1
added the same fuel consumption to Bag 2. This was an increase in Bag 1 fuel consumption of
23-25% increase for the two vehicles. Thus, these are significant shifts in battery storage and
probably exceed any change actually occurring during the FTP. This shift in fuel consumption
increased the 5-cycle city fuel economy of the Civic by 0.5 mpg, or 1.5%. It increased the 5-
cycle city fuel economy of the Prius by 1.0 mpg, or 2%. These changes in city fuel economy are
likely worse case, since the degree of shift in fuel consumption are large percentages of Bag 1
fuel consumption. Also, such a shift in Bag 1 fuel consumption would likely produce some
degree of shift in Bag 3 fuel consumption, as well. The above analysis assumed that Bag 3 fuel
consumption remained unchanged. Still, the potential for change in battery charge status during
Bags 1 and 3 could have a significant effect on 5-cycle city fuel economy values that may not
116
-------
reflect onroad operation. Without actual measurement of the state of battery charge after each
bag of the FTP and on-road following vehicle start-up, it is not possible to quantify this
uncertainty any further.
Regarding the effect of average trip length on start fuel consumption, the three studies
with the most extensive collections of data were addressed in section LA above: the instrumented
vehicle studies conducted in support of the Supplemental FTP rule, the National Household
Travel Study (NHTS), and the data currently being collected by Georgia Tech in Atlanta. One
obvious uncertainty in the current estimates of average trip length is the 11% downward
adjustment to the average trip length found in the NHTS. This reduction in average trip length
from 9.8 to 8.7 miles was applied in order to obtain consistency with the results of the various
instrumented vehicle studies performed in support of the Supplemental FTP rule and Atlanta. It
is also based on the belief that instruments more accurately measure short trips and short respites
between trips, such as moving a car out of the garage or stopping to refuel. Even if it is almost
certain that a diary-based measure of trip length should be adjusted downward, the degree of this
adjustment is uncertain. There is also uncertainty in the average trip length for urban dwellers,
since only three cities have been studied with vehicle instrumentation.
In order to assess the potential consequence of this uncertainty, we removed the 11%
adjustment from the national average trip length, leaving it at 9.8 miles. Retaining a highway
trip length of 60 miles, this increased the city trip length to 4.6 miles. Using 4.6 miles for the
average trip length for city driving increased the average fuel economy of the 615 vehicles in our
5-cycle fuel economy database from 16.9 mpg to 17.0 mpg, or by less than 1%. The 5-cycle city
fuel economy of non-hybrids increased from 16.5 mpg to 16.6 mpg, while that for hybrids
increased from 33.0 mpg to 33.2 mpg. These increases are quite small, particularly given the
fact that this represents removal of the entire adjustment. Since the uncertainty in the 11%
adjustment is likely much less than +11%, the uncertainty is average trip length is not a major
factor causing uncertainty in the 5-cycle fuel economy formulae.
Finally, we are promulgating a change in the test procedure of the Cold FTP which
involves activation of the heater or defroster during the test. The vast majority of drivers
obviously utilize their heaters at 20°F, but the number that do so between 20°F and 75°F has not
been studied. The weighting factor for cold start fuel use was developed from test data which
did not involve heater or defroster activation. Thus, consideration of the effect of
heater/defroster use at various temperatures could affect this factor. Also, drivers can differ in
the way they activate their heater and defroster at colder temperatures. As described in Section
III.B. above, EPA has tested several vehicles at 20°F with and without heater/defroster
activation. The impact on conventional vehicle start fuel use is minimal, decreasing 5-cycle city
fuel economy by only 0.1%. Thus, the uncertainty in this effect as estimated in the 5-cycle
formulae should be even smaller, given some heater and defroster use obviously occurs.
The effect is larger for hybrid vehicles, though it should be noted that our estimate of the
effect is only based on the test results from two hybrid vehicles. Actual label values developed
using the 5-cycle formulae will be based on actual Cold FTP fuel economy values measured with
the heater/defroster activated. Removing the estimated effect of heater and defroster use from
Bags 1 and 3 increases 5-cycle city fuel economy by 1%. The uncertainty in this effect as
117
-------
estimated in the 5-cycle formulae should be even smaller, given some heater and defroster use
obviously occurs. Thus, even uncertainty in the effect of heater/defroster use on start fuel use
does not appear to be a major source of uncertainty in 5-cycle city fuel economy.
2. Running Fuel Use At 75°F
In this section we evaluate several alternative approaches to determining the weighting of
the various test cycles to estimate running fuel use at 75°F. One alternative evaluates a more
ideal split of the US06 cycle into city and highway driving. In this case, both the second and
third hills described in Table III.A-16 are designated as highway driving. The average speed of
the US06 city bag decreases and the average speed of the US06 highway bag increases.
A second alternative eliminates the three highest speed freeway cycles which were not
derived from the 3-city studies performed in support of the Supplemental FTP rule. These three
cycles had to be developed subsequent to these instrumented vehicle studies due to the increase
in maximum speed limit from 55 mph nationwide to 70 mph and even higher today. The basis
for these three higher speed cycles is not as robust as that for the other 13 facility cycles. Thus,
there is more uncertainty in the VSP distributions of these three highest speed cycles than the
others.
A third alternative, actually a set of alternatives, evaluates the use of alternative fuel rates
by VSP bin in the regression of dynamometer cycles versus onroad operation. Fuel rates from
the EPA 15 car study, fuel rates from Draft MOVES2004 extrapolated to 23 VSP bins are
substituted for those found in the EPA Kansas City testing. The impact of using just the 17 VSP
bins current in Draft MOVES2004 is also evaluated.
A fourth alternative develops test cycle combinations which represent onroad VSP
distributions and fuel rates from EPA's recent test program in Kansas City. Test cycle
combinations are developed for non-hybrid and hybrid vehicles separately, given that significant
numbers of both vehicle types were tested.
A final alternative develops a set of cycle weighting factors using only entire test cycles
(FTP, HFET and US06) instead of allowing separate bag weights within the FTP and US06
cycles.
These alternatives and their effect on the cycle weighting factors are described below.
a. Alternative Definition of US06 City and Highway Bags
In section III.A.2, we defined the city bag of US06 to include hills number 1, 2, 4, and 5.
However, hill 2 was placed in the city bag for practical, testing related reasons. Here, we
redefine the city bag in a more ideal way to only include hills 1, 4, and 5. The highway bag
includes hills 2 and 3. The description of the various hills in US06 and their assignment to the
city and highway bags of US06 are shown in Table III.E-2.
118
-------
Table III.E-2. Split of US06 Cycle into City and Highway Portions
Hill
1
2
3
4
5
Portion of Driving Cycle
(cumulative seconds)
0-43
44-131
132-495
496-563
564-600
Maximum Speed (mph)
44.2
70.7
80.3
29.8
51.6
Proposed
Designation
City
City
Highway
City
City
Ideal
Designation
City
Highway
Highway
City
City
With this revised split, the average speed of the city bag decreases to 18.1 mph and the
average speed of the highway bag increases to 80.3 mph. Table III.E-3 shows the VSP
distributions for both the proposed and more ideal city and highway bags of US06.
119
-------
Table III.E-3. VSP Distributions for US06 City and Highway Bags (% of time)
VSP
Bin
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
Proposed Definition
US06 City
32.6%
14.3%
1.3%
1.7%
0.9%
1.3%
0.4%
2.2%
2.2%
1.3%
3.5%
3.3%
1.5%
1.7%
0.2%
1.1%
0.4%
1.5%
2.4%
9.1%
4.6%
3.7%
0.4%
0.9%
0.9%
6.5%
US06 Hwy
7.1%
2.5%
0.3%
0.3%
0.0%
0.0%
0.0%
0.3%
0.0%
0.0%
0.5%
0.3%
0.1%
0.1%
0.0%
0.0%
0.0%
0.1%
0.4%
1.6%
9.9%
20.7%
13.8%
16.2%
10.2%
15.5%
More Ideal Definition
US06 City
31.8%
19.6%
1.4%
2.7%
1.4%
2.0%
0.7%
2.7%
2.0%
1.4%
5.4%
5.2%
2.3%
2.7%
0.4%
1.6%
0.7%
2.5%
2.3%
9.5%
1.4%
0.0%
0.0%
0.0%
0.0%
0.7%
US06 Hwy
11.9%
3.5%
0.4%
0.2%
0.0%
0.0%
0.0%
0.4%
0.4%
0.2%
0.4%
0.2%
0.1%
0.1%
0.0%
0.0%
0.0%
0.1%
0.7%
2.9%
10.5%
19.2%
11.4%
13.4%
8.5%
15.1%
As can be seen from Table III.E-3, the changes in the VSP distribution of the highway bag
changes are slight. This is due to the fact that hill 2 is much shorter than hill 3, so its addition
has a smaller impact. Also, the driving in hills 2 and 3 are similar. However, the VSP
distribution of the city bag changes dramatically. This occurs because hill 2 varies dramatically
from hills 1, 4, and 5. In particular, the amount of time spent in VSP bins 33-39 decreases from
17% to 2%.
We repeated the regressions of the VSP distributions of the dynamometer bags and cycles
(Tables III. A-15 and III.E-3) against the VSP distributions of city and highway driving (Table
IE. A-14) weighted by the square root of the fuel rate in each VSP bin from the Kansas City test
120
-------
program (Table III.A-16). The results using both sets of definitions of US06 city and highway
bags are described in Table III.E-4 below.
Table III.E-4. Bag/Cycle Combinations for City and Highway Driving:
Alternative US06 Splits
Proposed Split
Ideal Split
City Driving
Bag 2 FTP
Bag 3 FTP
US06 City
48%
41%
11%
49%
43%
8%
Highway Driving
HFET
US06 Highway
21%
79%
17%
83%
As seen in Table III.E-5, the more ideal breakdown of the US06 cycle results in the same
bags and cycles being selected to represent city and highway driving as use of the proposed split
of US06. The contribution of the US06 city bag to city driving decreases slightly, while
contribution of the US06 highway bag to highway driving increases slightly.
We reanalyzed the results of the 80 vehicles tested over US06 on a second by second
basis in order to re-estimate the relative fuel economy over the US06 city and US06 highway
bags using the more ideal split. We found that the fuel economy over the more ideal US06 city
bag was only 50% of that over the entire US06 cycle, compared to 68% for the proposed US06
city bag. The fuel economy over the more ideal US06 highway bag was only 111% of that over
the entire US06 cycle, compared to 116% for the proposed US06 highway bag. Thus, hill 2
appears to have a somewhat mid-range fuel economy compared to the lower speed hills 1, 4, and
5 and to the higher speed hill 3.
Using these revised cycle combinations and the revised estimates of US06 city and
highway fuel economy relative to US06 fuel economy, we recalculated 5-cycle fuel economy
estimates for the 615 vehicles in our 5-cycle fuel economy database. The results are summarized
in Table III.E-5.
Table III.E-5. Average 5-Cycle Fuel Economy: Alternative US06 Splits
City Fuel Economy (mpg
Highway Fuel Economy (mpg)
Proposed US06 Split
16.9
23.1
Ideal US06 Split
16.7
22.1
As can be seen, 5-cycle city fuel economy decreases slightly, by roughly 1%. However,
highway fuel economy decreases by 4.5%. This decrease in both city and highway fuel economy
would increase the non-dynamometer factor by roughly 3%. Thus, the net change in city fuel
economy would be an increase of roughly 1% and highway fuel economy would decrease by
1.5%.
121
-------
b. Elimination of Three Highest Speed Freeway Cycles in Draft
MOVES2004
As described in section III.A-2, Draft MOVES2004 uses 16 facility driving cycles to
describe onroad driving. Most of these cycles are based on the results of the instrumented
vehicle studies conducted in the early 1990's in support of the Supplemental FTP rule.
However, since the maximum speed limit at this time was 55 mph and is now much higher,
additional high speed cycles had to be developed to describe this now common type of driving.
During the development of MOBILE6.2, EPA developed three higher speed freeway
cycles (High Speed Freeway 1, High Speed Freeway 2, and High Speed Freeway 3 in Table
III.A-12). High Speed Freeway 3, the fastest of the three cycles, was based on a segment of
onroad driving of one of the vehicles tested in the EPA 15-car study.17 While this driving
segment was actually driven on the highway by a vehicle, the EPA 15-car study was a pilot study
and neither the vehicle nor driver selection was designed to be random. Also, since the drivers
were either EPA employees or contractors and were aware of the purpose of the study, the
driving was not designed to be representative of typical vehicle use. The portion of onroad
driving represented by High Speed Freeway 3 is based on its average speed and onroad speeds
predicted by travel demand models and California rural chase car data, the acceleration rates are
not. Thus, the power demands of this cycle could differ from those on the road. The
representativeness of the specific speed-time trace of High Speed Freeway 3 and its effect on
vehicle power is being evaluated as part of the further development of MOVES.
The High Speed Freeway 2 cycle is a portion of the US06 cycle. This portion of the
cycle met the desired average speed, which was slightly below that of High Speed Freeway 3.
As discussed in section III.A.2, the US06 cycle is a concentrated version of the REP05 cycle.
REP05 is a driving cycle which is representative of higher speed and/or higher power driving
found onroad during the vehicle studies performed in support of the EPA Supplemental FTP
rulemaking.12 In the early 1990's, it represented roughly 28% of U.S. driving. US06, consisting
of the most aggressive portions of REP05, represented a smaller percentage of driving at that
time. However, both speed limits and the power to weight ratio of vehicles have both increased
since that time. US06 driving represents more driving today than it did in the early 1990's.
However, the exact percentage is not known. As is the case with High Speed Freeway 3, the
portion of onroad driving represented by High Speed Freeway 2 is based on its average speed
and onroad speeds predicted by travel demand models and California rural chase car data.
However, the power demands of the cycle derive from the specific portion of US06 which was
selected to comprise High Speed Freeway 2. Thus, the power demands of this cycle could differ
from those of in-use vehicles which are driving at these speeds. The representativeness of the
specific speed-time trace of High Speed Freeway 2 and its effect on vehicle power is being
evaluated as part of the further development of MOVES.
High Speed Freeway 1 is the most representative of the three highest speed cycles.41 It
was developed for use in MOBILE6.2 from data obtained during the 3-city studies. Basically,
vehicle operation which occurred on relatively uncongested freeways (LOS A-C) which lasted at
least 30 seconds with a minimum speed of 50 mph were binned and used to create the High
Speed Freeway 1 cycle. Thus, the underlying driving characteristics reflect real world operation
122
-------
in the early 1990's. However, vehicles operating at these speeds today might be driven
differently, due to a change in the types of roadways which carry vehicles at these speeds and to
the fact that many vehicles on the same roads are driving faster and not slower. As is the case
with High Speed Freeway 2 and 3, the representativeness of the specific speed-time trace of High
Speed Freeway 1 and its effect on vehicle power is being evaluated as part of the further
development of MOVES.
We evaluated the impact of these three cycles on the proposed cycle combinations in the
5-cycle formulae by shifting the onroad driving assigned to these cycles to the fourth highest
speed cycle, LOS AC Freeway. This is an extreme change and goes far beyond any possible
uncertainty related to these three high speed freeway cycles. Thus, it bounds the potential
uncertainty in this area and likely over-estimates it. Only the highway fuel economy formula is
affected, since we assumed in section III.A.2 that 100% of the driving over these cycles was
highway driving. As shown in Table III.A-13, this shift represents 20% of onroad highway like
driving.
We repeated the regressions of the VSP distributions of the dynamometer bags and cycles
(Tables III.A-15) against the VSP distribution of highway driving (Table III.A-14 with shift to
LOS A-C Freeway) weighted by the square root of the fuel rate in each VSP bin from the Kansas
City test program (Table III. A-16). The results with and without the three highest speed cycles
are described in Table III.E-6 below.
Table III.E-6. Bag/Cycle Combinations for Highway Driving: High
Speed Freeway Cycles
HFET
US06 Highway
With 3 High Speed
Cycles
21%
79%
Without 3 High Speed Cycles
25%
75%
Using this revised cycle combination for highway driving, we recalculated 5-cycle
highway fuel economy estimates for the 615 vehicles in our 5-cycle fuel economy database.
Eliminating the three highest speed freeway cycles increased average 5-cycle highway fuel
economy from 23.1 mpg to 23.3 mpg, or by less than 1%.
c. Alternative Fuel Rates and Number of VSP Bins
In this section, we evaluate the use of three sets of alternative fuel rates by VSP bin in the
regression of dynamometer cycles versus onroad operation. The three sets of fuel rates are based
on: 1) the EPA 15 car study, 2) Draft MOVES2004 extrapolated to 26 VSP bins using fuel rates
from the EPA 15 car study, and 3) Draft MOVES2004 extrapolated to 26 VSP bins using fuel
rates from the EPA Kansas City testing. We also evaluate the impact of using just the 17 VSP
bins currently used in Draft MOVES2004. In this case, the fuel rates are those currently in Draft
MOVES2004.
These three sets of 26 VSP bin fuel rates were presented in Table III. A-16 above. Those
currently in Draft MOVES2004 are simply those rates shown in Table III. A-16 under either of
123
-------
the "Extrapolated MOVES" columns for bins 0-16, 21-26 and 33-36. The 17 bin VSP
distributions for both the dynamometer cycles and bags and the MOVES facility cycles are the
same as those with 26 bins, except that the x6 bin contains all of the driving in bins x6, x7, x8
and x9.
We repeated the regressions of the VSP distributions of the dynamometer bags and cycles
(Tables III.A-15) against the VSP distributions of city and highway driving (Table III.A-14)
weighted by the square root of the various fuel rates. The results under the base case and the
alternatives are shown in Table III.E-7 below.
Table III.E-7. Bag/Cycle Combinations for City and Highway Driving: Alternative Fuel
Rates
Base Case: Kansas
City Fuel Rates
EPA 15 Car
Fuel Rates
Extrapolated MOVES
EPA 15 Car
Kansas City
MOVES
17 bin
City Driving
Bag 2 FTP
Bag 3 FTP
US06 City
48%
41%
11%
50%
50%
0%
50%
50%
0%
50%
50%
0%
50%
39%
11%
Highway Driving
HFET
US06 Highway
21%
79%
21%
79%
21%
79%
21%
79%
20%
80%
As seen in Table III.E-7, the three alternative sets of fuel rates yield cycle combinations
of city driving which are identical and do not include the US06 city bag. The MOVES 17 bin
approach yields a cycle combination of city driving which is much more similar to that of the
base case. Regarding the cycle combinations of highway driving, all of the various alternatives
produce essentially identical results.
The results for city driving indicate that the contribution of the US06 city bag is the most
uncertain of the 3 bags. This is confirmed by the p-values for the various bags in the final
regressions. While the inclusion of the US06 city bag in the regression improves the adjusted r-
squared value, the p-values for the US06 city bag are the highest of the three remaining bags,
falling in the range of 0.2-0.3.
Using the 50/50 combination of bag 2 and 3 (zero weight for US06 city), we recalculated
5-cycle fuel economy estimates for the 615 vehicles in our 5-cycle fuel economy database. The
50/50 combination of Bags 2 and 3 increases 5-cycle city fuel economy for conventional vehicles
from 16.5 mpg to 17.2 mpg, or by 4%, compared to the proposed cycle combination. The 50/50
combination of Bags 2 and 3 increases 5-cycle city fuel economy for hybrid vehicles from 33.0
mpg to 34.8 mpg, or by 5.5%, compared to the proposed cycle combination. The increase in city
fuel economy also increases combined fuel economy by 2% for conventional vehicles and 3%
for hybrids. This increase in combined 5-cycle fuel economy would lead to a lower factor for
non-dynamometer effects, roughly decreasing from 0.905 to 0.885. This change would decrease
both 5-cycle city and highway fuel economy by 2%. The uncertainty related to the contribution
of the US06 city bag to city driving is the sum of the change in city fuel economy plus the
change in both city and highway fuel economy due to a lower non-dynamometer factor. Thus,
124
-------
this uncertainty is roughly 2% in both city and highway fuel economy for conventional vehicles
and 4% for city and 2% for highway fuel economy for hybrid vehicles.
d. Kansas City VSP Distributions and Fuel Rates
EPA's recent testing of roughly 100 recent model year vehicles in Kansas City is
described in detail in Appendix A. We aggregated the driving activity and fuel rates measured
during this testing (basically one day's driving for each vehicle) and developed VSP distributions
and average fuel rates by VSP bin for two sets of vehicles, conventional vehicles and hybrids. In
order to develop VSP distributions for city and highway driving, we segregated driving activity
on a second by second basis into two groups: those at 45 mph or lower and those above 45 mph.
This segregation is not exactly consistent with our definition of city and highway driving, since
these are based on longer time frames than one second. However, this was the only method
applicable in the near term. With additional time, each vehicle's driving trace could be reviewed
visually and segregated into city and highway driving. We may attempt to do this in the future.
The city and highway VSP distributions and fuel rates are shown in Table III.E-8.
Table III.E-8. Kansas City VSP Distributions and Fuel Rates
VSP
Bin
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
VSP Distribution: Non-Hybrids
City
8.4%
29.1%
5.5%
9.7%
3.5%
2.1%
1 .6%
0.7%
0.4%
0.2%
0.2%
7.3%
8.8%
9.7%
2.8%
4.0%
2.5%
1 .0%
0.8%
1 .6%
Highway
0.8%
1 .8%
0.1%
3.6%
1 .9%
0.0%
1 .0%
1 .0%
0.0%
0.7%
14.5%
36.8%
15.7%
6.5%
4.5%
1 1 .2%
VSP Distribution: Hybrids
City
9.7%
10.6%
7.5%
13.7%
4.9%
2.5%
2.0%
0.8%
0.4%
0.1%
0.0%
9.5%
16.8%
6.9%
4.8%
4.9%
2.4%
1.1%
0.7%
0.6%
Highway
0.7%
1 .8%
5.4%
1 .8%
0.2%
0.5%
0.5%
19.2%
49.8%
6.9%
3.9%
3.6%
6.1%
Fuel Rate (g/sec)
Non-Hybrids
0.412
0.384
0.454
0.641
1.122
1.406
1.715
2.006
2.172
2.358
2.296
0.593
0.833
0.986
1.180
1.352
1.631
1.917
2.272
2.424
1.069
1.486
1.753
1.937
1.948
2.309
Hybrids
0.097
0.248
0.122
0.188
0.466
0.690
0.857
0.976
1.126
1.182
1.319
0.225
0.359
0.578
0.569
0.755
0.993
1.107
1.327
1.588
0.645
0.920
1.175
1.160
1.210
1.441
125
-------
We repeated the regressions of the VSP distributions of the dynamometer bags and cycles
(Tables III. A-15) against the VSP distributions of city and highway driving weighted by the
square root of the various fuel rates shown in Table III.E-8. The results for the base case and the
two vehicle types in Kansas City are shown in Table III.E-9 below.
Table III.E-9. Bag/Cycle Combinations for City and Highway Driving: Kansas City
Base Case
Kansas City Non-Hybrids
Kansas City Hybrids
City Driving
Bag 2 FTP
Bag 3 FTP
US06 City
48%
41%
11%
39%
61%
0%
70%
30%
0%
Highway Driving
HFET
US06 Highway
21%
79%
38%
62%
56%
44%
The cycle combinations for non-hybrid and hybrid vehicles tested in Kansas City vary
dramatically from those of the base case and from each other. The US06 city bag does not
appear in either of the final regressions for city driving in Kansas City. However, Bag 3
dominates city driving by non-hybrids, while Bag 2 dominates city driving by hybrids. The
potential causes for this difference in the driving of hybrids and conventional vehicles are
discussed in some detail in Appendix A. First, this difference is based on limited driving;
roughly one hour of driving per vehicle. Second, while it is possible that this difference, if real,
will persist in the future, it seems more likely that any difference will disappear as more high
power hybrids enter the fleet and more drivers purchase hybrids.
The contribution of the US06 highway bag to highway driving in Kansas City is also
lower than that predicted by MOVES for both vehicle types. The contribution of the US06
highway bag to highway driving by hybrids in Kansas City is also lower than that by
conventional vehicles. A more detailed review of the regression results for highway driving
indicates that the model is having difficulty in matching the onroad VSP distributions for
highway driving. In all the other regressions of highway driving, the coefficients for HFET and
US06 highway from the raw regression results sum very close to 1.0 (e.g., 0.97). In Kansas City,
the coefficients for HFET and US06 highway from the raw regression results sum to roughly
1.25. Mathematically, this means that in order to minimize error, the model wants to use 1.25
seconds of cycle driving to match the fuel consumption of one second of onroad driving. This
means that the normalized coefficients presented in Table III.E-8 will under-estimate onroad fuel
consumption significantly. The prediction of relative fuel consumption across models might be
reasonable for the highway driving performed in Kansas City, but the absolute prediction will be
quite low. This would necessitate use of a non-dynamometer factor well below 1.0, as opposed
to the 0.98 factor proposed in section III.A-5.
In general, a sum of coefficients well above or below 1.0 indicates that the model cannot
match the onroad fuel consumption per second using the cycles made available to it. In the case
of highway driving, the model would normally increase the contribution of US06 highway, as
this cycle has higher average fuel consumption than HFET. However, this is creating too great
an error in the individual VSP bins and the sum of squared errors increases. This means that the
126
-------
highway driving observed in Kansas City differs dramatically (in terms of VSP) from both the
HFET and US06 highway cycles and that another cycle is needed.
As will be seen in the next section, driving observed in California yield very different
cycle combinations compared to those based on the Kansas City data. This indicates the need for
driving to be characterized in a number of urban areas in order to be representative of the country
as a whole, and not based on just one or two areas.
e. California Chase Car Studies
As described in Appendix A, Sierra Research has performed several chase car studies of
both urban and rural driving in California since 1998. We have not yet been able to perform a
detailed analysis of the second by second data from this testing. However, we have been able to
develop approximate VSP distributions for urban and rural driving from the speed-acceleration
frequency distributions. As in Kansas City, we assumed that driving at or below 45 mph was
city driving and driving above 45 mph was highway driving. However, one of the SAFD bins
was centered at 45 mph (i.e., speeds of 42.5-47.5 mph). This bin was assigned to city driving.
Also, all driving in bins 17-19 were assigned to bin 16. These VSP distributions are shown in
Table III.E-10.
Table IILE-10. California Urban and Rural VSP Distributions
VSP
Bin
0
1
11
12
13
14
15
16
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
Urban
City
16.3%
14.9%
4.9%
14.6%
5.3%
4.6%
3.9%
2.8%
5.2%
4.0%
12.2%
2.1%
2.1%
4.4%
0.8%
0.9%
0.9%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Highway
2.9%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
2.6%
0.0%
0.0%
8.6%
0.0%
0.0%
3.2%
0.0%
0.7%
12.1%
19.8%
14.7%
13.4%
8.7%
13.2%
Rural
City
19.8%
6.8%
2.8%
9.3%
3.5%
3.0%
3.1%
4.4%
6.4%
2.4%
18.1%
1 .2%
1 .9%
8.9%
1 .8%
2.9%
3.8%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Highway
2.2%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
1 .9%
0.0%
0.0%
7.4%
0.0%
0.0%
2.6%
0.0%
0.6%
9.6%
27.0%
1 1 .3%
1 1 .8%
1 1 .5%
14.1%
Fuel Rate (g/sec)
Non-Hybrids
0.412
0.384
0.454
0.641
1.122
1.406
1.715
2.006
0.593
0.833
0.986
1.180
1.352
1.631
1.917
2.272
2.424
1.069
1.486
1.753
1.937
1.948
2.309
Hybrids
0.097
0.248
0.122
0.188
0.466
0.690
0.857
0.976
0.225
0.359
0.578
0.569
0.755
0.993
1.107
1.327
1.588
0.645
0.920
1.175
1.160
1.210
1.441
127
-------
We repeated the regressions of the VSP distributions of the dynamometer bags and cycles
(Tables III. A-15) against the VSP distributions of California city and highway driving in urban
and rural (Table III.E-10) weighted by the square root of the Kansas City fuel rates (Table III.A-
16). We performed the same analysis for combined urban/rural VSP distributions of city and
highway driving, using an urban/rural weighting of 60/40 (consistent with FHWA estimates of
urban and rural VMT for the U.S.). The chase car studies showed that 75.5% of all urban
operation was city driving and 24.5% was highway. In contrast, 30.4% of all rural operation was
city driving and 69.6% was highway. Thus, the VSP distribution for combined urban/rural city
driving was developed by weighting the urban city driving VSP distribution by 78.8% and the
rural city driving VSP distribution by 21.2%. Likewise, the VSP distribution for combined
urban/rural highway driving was developed by weighting the urban highway driving VSP
distribution by 34.5% and the rural city driving VSP distribution by 65.5%. The results for the
base case and the California cases are shown in Table III.E-11 below.
Table III.E-1 1. Bag/Cycle Combinations for City and Highway Driving: California
City Driving
Bag 2 FTP
Bag 3 FTP
US06 City
Highway Driving
HFET
US06 Highway
Base Case
48%
41%
11%
21%
79%
California Urban
42%
42%
16%
13%
87%
California Rural
0%
75%
25%
21%
79%
All California
35%
46%
19%
18%
82%
Overall, the dynamometer cycle combinations for California driving tend to be more
aggressive than those based on Draft MOVES2004 and are decidedly more aggressive than those
found in Kansas City. City driving in California shows a 16-19% contribution for the US06 city
bag, versus 11% based on Draft MOVES2004 and zero in Kansas City. Highway driving in
California shows a 79-87% contribution for the US06 highway bag, versus 79% based on Draft
MOVES2004 and 44% for hybrids and 63% for non-hybrids in Kansas City. Chase car studies
can tend to under-represent driving on neighborhood and local roads. Thus, some increased
percentage of US06 city driving can be expected in the cycle combinations for city driving.
However, one would not expect over-estimation by a factor of two. The characterization of
highway driving should not be affected by this factor, since vehicle speeds on local and
neighborhood roads tend to be less than 45 mph.
Thus, the results from Kansas City and California indicate the need for driving to be
characterized in a number of urban areas in order to be representative of the country as a whole.
The cycle combinations based on Draft MOVES2004 fall in between those found in Kansas City
and California and thus, are not inconsistent with the driving in the two specific geographical
areas.
f Alternative Splits of City/Highway Driving
In Sections III.A.I and III.A.2, we define city driving as that below 45 mph and highway
driving as that above 45 mph. According to this definition, we assigned all of the driving over
128
-------
the LOS D Freeway cycle to highway driving. This definition of city and highway driving
produces a city/highway VMT split of 43/57, which differs dramatically from the current 55/45
split. We believed it appropriate to investigate some options which yielded city/highway VMT
splits closer to the current split of 55/45, specifically 50/50 and 55/45. This was accomplished
by adjusting the split of driving over the LOS D freeway cycle to city and highway categories.
The LOS D freeway cycle has an average speed of 53 mph and over three-quarters of the
driving time of this cycle is above 45 mph. However, the cycle does include some driving below
30 mph (about 3% in terms of driving time). Thus, there is some rationale to assigning at least a
portion of this cycle to city driving. Also, the LA4 road route which is the basis for the FTP test
included some freeway operation. Bag 3 of the FTP has a maximum speed of 55 mph and about
20% of the driving time during this bag is above 45 mph. All the cycle combinations
representing city driving include Bag 3, so all these representations of city driving include some
driving over 45 mph regardless of how city driving is defined with respect to the Draft
MOVES2004 facility cycles. Thus, assigning some driving over 45 mph to city driving could
improve the representation of city driving (i.e., increase the adjusted r-squared values in the
regression of the VSP distribution of city driving versus those of the dynamometer cycles).
To assess this possibility, we developed two alternative assignments of driving of the
LOS D Freeway cycle to city and highway driving. The first produced an overall city/highway
VMT split of 50/50, while the second produced an overall city/highway VMT split of 55/45.
The first split required assigning 26% of LOS D Freeway driving to city driving, while the
second the required assigning 43% of LOS D Freeway driving to city driving. We recalculated
the VSP distributions of city and highway driving with the two alternative assignments of LOS D
Freeway driving. The base case and alternative VSP distributions are shown in Table III.E.12.
129
-------
Table III.E-12. VSP Distributions for U.S. Driving with Alternative Definition of City
Driving (% of time)
VSP
Bin
0
1
11
12
13
14
15
16
17
18
19
21
22
23
24
25
26
27
28
29
33
35
36
37
38
39
Base: 43/57 City/Highway Split
City VSP
1 1 .8%
20.4%
8.3%
13.0%
6.0%
3.5%
2.2%
0.6%
0.1%
0.1%
0.0%
5.5%
6.2%
4.7%
4.2%
3.0%
2.2%
0.8%
0.4%
0.3%
1 .9%
2.3%
1.1%
0.9%
0.5%
0.2%
Highway VSP
3.9%
0.2%
0.1%
0.1%
0.1%
0.1%
0.0%
0.0%
0.0%
0.0%
0.0%
2.8%
2.9%
3.2%
2.9%
2.6%
2.4%
0.9%
0.5%
0.9%
16.7%
21 .0%
10.1%
9.8%
8.0%
10.7%
50.50 City/Highway Split
City VSP
1 1 .4%
19.1%
7.8%
12.1%
5.6%
3.3%
2.0%
0.5%
0.1%
0.1%
0.0%
5.4%
6.1%
4.7%
4.3%
3.1%
2.3%
0.9%
0.4%
0.4%
2.8%
3.3%
1 .6%
1 .3%
0.8%
0.6%
Highway VSP
3.6%
0.2%
0.2%
0.1%
0.2%
0.1%
0.0%
0.0%
0.0%
0.0%
0.0%
2.5%
2.5%
2.8%
2.6%
2.4%
2.2%
0.8%
0.4%
0.9%
16.7%
21 .5%
10.2%
10.2%
8.4%
1 1 .4%
55/45 City/Highway Split
City VSP
11.1%
18.3%
7.5%
1 1 .7%
5.4%
3.1%
2.0%
0.5%
0.1%
0.1%
0.0%
5.4%
6.1%
4.7%
4.3%
3.1%
2.3%
0.9%
0.4%
0.4%
3.4%
3.9%
1 .9%
1 .5%
1 .0%
0.8%
Highway VSP
3.5%
0.3%
0.2%
0.2%
0.2%
0.1%
0.0%
0.0%
0.0%
0.0%
0.0%
2.3%
2.3%
2.5%
2.3%
2.1%
1 .9%
0.8%
0.4%
0.8%
16.7%
21 .9%
10.3%
10.5%
8.7%
1 1 .9%
Shifting some of the driving over the LOS D Freeway cycle to city driving increases
operation in Bins 33-39 significantly in the VSP distribution for city driving. It also increases
operation in Bins 35-39 and reduces operation in Bins 21-29 in the description of highway
driving.
We repeated the regressions of the VSP distributions of the dynamometer bags and cycles
(Tables III. A-15) against the alternative VSP distributions of city and highway driving shown in
Table III.E-12 weighted by the square root of the Kansas City fuel rates (Table III.A-16). As has
been done in all such regressions, we first dropped those cycles or bags with negative
coefficients and then dropped the least significant cycles or bags until the adjusted R-squared
value decreased. Then, the results of the previous regression were selected as the final
combinations of cycles. The results for the base case and the revised split of city and highway
driving are shown in Table III.E-13 below.
130
-------
Table IILE-13. Cycle Combinations for City and Highway Driving: Revised City/Highway
Split
Base Case
Cycle %
p-value
50/50 City/Highway Split
Cycle %
p-value
55/45 City/Highway Split
Cycle %
p-value
City Driving
Bag 2 FTP
Bag 3 FTP
US06 City
US06 Highway
Adjusted r-squared
48%
41%
11%
0%
0.7262
<0.001
0.01
0.26
—
40%
43%
0%
17%
0.7048
<0.001
0.003
—
0.11
35%
43%
0%
22%
0.6958
<0.001
0.0015
—
0.034
Highway Driving
HFET
US06 Highway
Adjusted r-squared
21%
79%
0.8682
<0.001
<0.001
20%
80%
0.8784
<0.001
<0.001
18%
82%
0.8855
<0.001
<0.001
Including some operation over 45 mph in the definition of city driving has a significant
effect on the cycles which represent city driving. For both alternative definitions of city driving,
US06 city drops out of the cycle combination and US06 highway is added at roughly twice the
previous US06 city weight. The contribution of Bag 2 drops, as well. These changes are
progressive. The greater the percentage of city driving overall (i.e., the greater the percentage of
LOS D Freeway driving assigned to city driving), the greater degree that the above changes
occur. The effect on highway driving is much smaller: the weight of US06 highway increases 1-
4%, while that of HFET decreases commensurately.
Some of the regression statistics appear to be better for the base definition of city driving,
while others appear better for the higher city driving fraction. Again, the changes are
progressive with respect to the two alternatives. The adjusted r-squared value for the base case
combination of city driving is higher than that for either alternative definition of city driving.
However, the p-values for the US06 bag with the alternative definitions of city driving are better
than that for the base definition. The more high speed driving is included in city driving, the
greater the significance of the US06 highway coefficient. With the base definition of city
driving, including US06 city improves the adjusted r-squared value, but the p-value of the US06
coefficient is relatively high, 0.26.
For highway driving, the p-values for both HFET and US06 are very low in both cases.
However, the adjusted r-squared value increases as the overall city driving fraction increases. In
both alternative cases, the adjusted r-squared value is higher than that with the base definition of
city driving.
The impact of increasing the VMT fraction of city driving on the 5-cycle fuel economy of
the 615 vehicles in our certification database is shown in Table III.E-14. In addition to the
revised cycle combinations shown in Table III.E-13, increasing the city fraction of VMT also
increases the average trip length of city driving. For a city/highway VMT split of 50/50, the
average trip length for city driving increases from 4.1 miles to 4.7 miles, while that for a split of
55/45 is 5.1 miles. The effect of this increased trip length is included in the 5-cycle fuel
economy estimates with the 505/50 and 55/45 city/ highway VMT splits.
131
-------
Table IILE-14. 5-Cycle Fuel Economy Values: Effect of the Definition of City Driving
Current EPA Label
City
Non-Hybrid
18.6
Hybrid
42.7
Highway
Non-Hybrid
24.6
Hybrid
42.8
Composite
Non-Hybrid
20.9
Hybrid
42.6
5-Cycle
43/57 City/Hwy Split
50/50 City/Hwy Split
55/45 City/Hwy Split
16.5
17.7
17.9
33.0
34.2
34.0
22.8
22.7
22.6
36.9
36.8
36.6
19.6
20.2
20.3
35.0
35.5
35.4
Moving some of the operation over LOS D Freeway to city driving increases 5-cycle city
fuel economy for both non-hybrids and hybrids. This is not surprising, given that fuel economy
over the US06 city bag is the lowest of any of the dynamometer cycles or bags. The effect is
again progressive with the degree of the shift of driving from highway to city. Five-cycle
highway fuel economy decreases very slightly due to the increased fraction of US06 highway in
the 5-cycle formula. Composite fuel economy increases slightly. This implies that the effect of
higher city fuel economy is slightly greater than the effect of the increased city VMT fraction.
The ORNL Your MPG website, discussed in detail in section II.A, contains consumers
estimates of their onroad fuel economy, as well as their estimate of their city and highway
driving fractions. Across the 8180 estimates of fuel economy, the average percentage of city
driving is 43%. This is closer to the 43% estimate resulting from placing all the operation over
LOS D Freeway into the highway driving category than splitting this operation between city and
highway driving. At the present time, it is the primary source of information about how typical
drivers label their own driving.
g. Complete Cycles
A final set of alternatives develops cycle combinations using entire test cycles (FTP,
HFET and US06) instead of allowing separate bag weights within the FTP and US06 cycles.
One alternative investigates the impact of not splitting US06 into city and highway bags, but
retaining separate bags for the FTP. The other alternative investigates the impact of not splitting
US06 into city and highway bags, as well as using a combined Bag 2 plus Bag 3 fuel economy.
The latter might be applicable to current hybrid testing. A hybrid FTP test consists of two
complete LA4 driving cycles, one with a cold start and one with a hot start. This is often
accomplished using four emission bags, with Bag 4 being similar to Bag 2. However, EPA
regulations allow emissions to be measured in only two bags, Bag 1 consisting of the normal
Bags 1 and 2 and Bag 2 consisting of the normal Bag 3 plus the extra Bag 4. In these cases, fuel
economy measurements for each of the four bags would not be available.
We repeated the regressions of the VSP distributions of the dynamometer bags and cycles
(Tables III. A-15) against the Draft MOVES2004 VSP distributions of city and highway driving
(Table III.A-14) weighted by the square root of the Kansas City fuel rates (Table III.A-16). The
results for the base case and the alternative complete cycle cases are shown in Table III.E-15
below.
132
-------
Table III.E-15. Bag/Cycle Combinations for Complete Cycle Alternatives
Base Case
Complete US06
Complete US06 and
LA4
City Driving
Bag 2 FTP
Bag 3 FTP
LA-4
US06 City
48%
41%
—
11%
50%
50%
—
—
—
—
100%
—
Highway Driving
HFET
US06 Highway
US06
21%
79%
—
25%
—
75%
25%
—
75%
Not surprisingly, maintaining US06 as a single cycle eliminates any contribution of US06
to city driving. The weights of Bag 2 and Bag 3 become 50% each, as was the case for the
alternative fuel rates in Table III.E-7. This is also consistent with the modeling of city driving
under cold temperature conditions where the US06 city bag was not allowed in the regression.
Maintaining US06 as a single cycle reduces the contribution of US06 to highway driving
modestly.
The results are quite similar when the LA-4 cycle is substituted for Bags 2 and 3. The
weighting of Bags 2 and 3 in the LA-4 is 52/48, which is very similar to the 50/50 weighting
found when separate bag estimates are allowed into the model. The impact of using whole
cycles on 5-cycle city and highway fuel economy values is shown in Table III.E-16.
Table III.E-16. Effect of Using Whole Cycles on 5-Cycle Fuel Economy Values (mpg)
Current EPA label
Base Case 5-cycle
Complete US06 5-cycle
Complete US06 and LA4 5-cycle
Conventional Vehicles
City
18.6
16.5
17.2
17.2
Highway
24.6
22.8
20.5
20.5
Hybrids
City
41.6
33.0
34.8
35.0
Highway
40.6
36.9
34.7
34.7
As expected, eliminating the contribution of US06 city from the 5-cycle city formula
increases city fuel economy (by about 4% in both alternative cases for conventional vehicles).
While increasing the contribution of HFET increases highway fuel economy, shifting from US06
highway to the whole US06 cycle decreases fuel economy. The latter impact predominates and
highway fuel economy decreases substantially (-10% for conventional vehicles). City fuel
economy for hybrids is slightly more sensitive to these changes than conventional vehicles, while
highway fuel economy is less sensitive. Hybrid city fuel economy increases 6%, while hybrid
highway fuel economy decreases 6%.
133
-------
3. Air Conditioning Effects
The primary factors affecting the estimation of fuel use related to air conditioning are: 1)
the degree of compressor engagement during the SC03 test, 2) the degree of compressor
engagement on the road, 3) the effect of vehicle speed on fuel use related to running the
compressor, and 4) the effect of ambient conditions on the load of the compressor on the engine.
The potential uncertainty in each of these factors and its affect on 5-cycle fuel economy will be
addressed below.
We assume that every vehicle's air conditioning compressor is engaged 100% of the time
during the SC03 test. This is based on general knowledge of the SC03 test itself. For example,
the test only lasts 10 minutes and the vehicle is subjected to strong sunlight at 95°F during
vehicle preconditioning operation (e.g., running a Bag 1 or 2 of the FTP, both Bags 1 and 2 of
the FTP, a SC03 test) and during the ten minute soak prior to the test. The air conditioning
system and fan are turned to its maximum setting and the recirculation option is chosen, if
available.
EPA also made the same assumption when estimating the emission benefits associated
with the SC03 emission standards.42 It is clear from the description of various test programs
performed in support of this rule that the engagement of the compressor was measured over the
SC03 and other test cycles and that the Agency had this information when it assumed that the
compressor was engaged 100% of the time over the SC03 test. However, this data is no longer
easily accessible and the level of compressor use which was actually measured over these tests
was not presented in any of the official rulemaking documents. Also, it is possible that vehicles'
air conditioning system designs have changed since the early 1990's. The degree of compressor
engagement over SC03 today might differ from that measured at that time. Therefore, in order to
assess the possible uncertainty in this assumption, we assume here that the compressor is
engaged only 80% of the time over SC03.
Reducing the amount of time that the compressor is engaged over SC03 from 100% to
80% can be modeled by simply dividing the incremental fuel use over SC03 versus the 69/31
mix of fuel use over Bags 3 and 2 of the FTP by a factor of 0.8. This change decreases the 5-
cycle city fuel economy of non-hybrid vehicles from 16.5 mpg to 16.4 mpg and that for hybrids
decreases from 33.0 mpg to 32.6 mpg. Thus, the change in city fuel economy is roughly 1%.
The change in compressor engagement over SC03 affects highway fuel economy even less.
Five-cycle highway fuel economy of non-hybrid vehicles decreases from 22.8 mpg to 22.7 mpg
and that for hybrids decreases from 36.9 mpg to 36.7 mpg. Thus, the change in highway fuel
economy is roughly half of one percent. Overall, the changes in both city and highway fuel
economy are small.
Regarding the degree of compressor engagement on the road, the Phoenix study used to
develop the 0.133 factor in the 5-cycle fuel economy formulae is based on the only instrumented
vehicle study performed to date. As discussed above, the NREL-OAP model, based on a
person's comfort at a given ambient temperature and humidity, yields an estimate of drivers
turning on the air conditioning of 29%, versus that based on the Phoenix work of 24%.
Considering the fact that the compressor is not always engaged when the air conditioning system
134
-------
is turned on, and the fact that the ambient temperature is usually less than 95 F, we estimate that
the compressor is on 13.3% of the time at a load equivalent to that occurring at 95 F. The
comparable percentage based on the NREL-OAP system-on estimate would be 16.1%. Defroster
use, based on the NREL-OAP work, could add another 0.4% to overall compressor use (in terms
of compressor load at 95°F). Thus, if both of the NREL-OAP estimates were correct, the 0.133
factor for compressor use in the 5-cycle formulae could be as high as 0.165. This increase is
roughly equivalent to the increase associated with assuming that the compressor is only engaged
80% of the time over the SC03 test. Thus, the fuel economy effects of increasing the air
conditioning usage factor to 0.152 would essentially the same as those just presented above (i.e.,
1.0% for city fuel economy and 0.5% for highway fuel economy). These are very small changes.
Regarding the effect of vehicle speed on fuel use related to running the compressor, this
uncertainty primarily applies to highway fuel economy. The speed of SC03 is within 10% of
that for average city driving. So there is little uncertainty in applying the incremental fuel use of
SC03 to city driving. The extrapolation to highway driving is much larger, the ratio of highway
speeds to the speed over SC03 being a factor of 2.67. Our testing of six vehicles with the air
conditioning on and off over the FTP, SC03 and HFET showed that the fuel use per unit of time
tended to be within 10% over the various cycles. However, these tests were not conducted in an
environmentally controlled test cell. Thus, further differences could occur under more realistic
ambient conditions. Also, the FIFET was the only highway cycle tested. The differences could
be larger if, for example, vehicles would have been tested over the US06. Therefore, to assess
the uncertainty in this factor, we assume here that the difference in fuel use per unit of time over
the various cycles could vary by as much as 20%. This can be represented by applying an
exponent of 0.8 to the ratio of speeds over the SC03 and highway driving. Thus, instead of a
factor 0.366 (21.5 mph divided by 57.1 mph), the factor would be 0.458. With this change,
highway fuel economy changed less than 0.5%. Therefore, this factor is not likely a significant
source of uncertainty in the 5-cycle highway fuel economy formulae.
Finally, regarding the effect of ambient conditions on the load of the compressor on the
engine, consideration of this factor reduces the effective compressor on time from 15.2% to
13.3%, or by a factor of 0.875. A reasonable estimate of the uncertainty in this factor would be
to double its effect (i.e., double its difference from 1.0). This would mean reducing the factor to
0.75. This change increases city fuel economy from non-hybrid vehicles by 0.1 mpg and by 0.3
mph for hybrids, at most 1% in the latter case. The effect on highway fuel economy is even
smaller, less than 0.5% in either case. Therefore, uncertainty in this area is unlikely to
significantly affect 5-cycle fuel economy.
In all, uncertainty in the four factors affecting the impact of air conditioning on onroad
fuel economy appears to be quite small. No single factor appears to affect fuel economy by more
than 1%, with most being significantly less than this.
4. Cold Temperature Running Fuel Use
The primary factors affecting the impact of colder temperatures on fuel economy are: 1)
the assumption that vehicles are fully warmed up by the end of Bag 1 of the FTP, 2) the effect of
heater/defroster use on warmed up fuel use, 3) the use of fuel consumption over just Bags 2 and
135
-------
3 to represent city driving and 4) the assumption that fuel consumption increases 4% during
highway driving at 20°F.
The potential impact of the assumption that vehicles are fully warmed up by the end of
Bag 1 of the FTP was addressed above in section III.E.l. As described there, if the vehicle
continued to warm up during Bag 2 at 20°F relative to that at 75°F, cold start fuel use would
increase and warmed up fuel use decrease. The net effect was an increase in the 5-cycle city fuel
economy of conventional vehicles of 0.3% and 1.2% for hybrids.
As described in Section III.E. 1, the effect of heater/defroster use is estimated to be very
small for conventional vehicles, but more significant for hybrids. Some effect of heater/defroster
use is certain, given the fact that most drivers use the heater and/or the defroster under certain
conditions. However, given the absence of precise data on their use under specific ambient
conditions, it is difficult to establish a precise uncertainty in their effect on onroad fuel economy
as indicated by the 5-cycle formulae. An upper bound on the uncertainty is the overall impact of
heater/defroster use on 5-cycle fuel economy. For conventional vehicles, removing the impact of
the heater/defroster increases 5-cycle city fuel economy by 0.4%. There is no effect on 5-cycle
highway fuel economy, as the effect of colder temperatures on running fuel use is modeled and
not based on the Cold FTP test. For hybrids, the effect is roughly 4%. Since heater/defroster use
clearly affects warmed up fuel use with current hybrid systems, some loss in onroad fuel
economy is certain. Thus, the uncertainty in the 5-cycle city fuel economy due to this factor is
likely well below 4%.
Regarding the absence of the US06 city cycle in the estimate of running fuel use at cold
temperatures, this means that the driving cycles underlying our estimated fuel use at 75°F and
20°F are inconsistent to some degree. We developed three approaches to estimate the potential
uncertainty associated with this difference. Each approach estimates the effect of cold
temperature on running fuel use during city driving slightly differently. For review, the approach
included in the final 5-cycle fuel economy formula bases running fuel use at 20°F only on Bags 2
and 3 of the FTP (Bags 3 and 4 for hybrids and other vehicles tested over a 4-bag FTP).
Therefore, the effect of the US06 city cycle is included in the estimate of running fuel use at
75°F (using a weighting factor of 16%), but not at 20°F. The first alternative approach calculates
the effect of including the US06 city cycle at 75°F and then applies this effect to the running fuel
use at 20°F, which is based only on Bags 2 and 3 of the FTP. To do this, we first remove the
effect of cold temperature completely. We then determined the impact of replacing the 11%
weight of the US06 city cycle by increasing the weights of Bags 2 and 3 each by 5.5%. This
replacement increased the city fuel economy from non-hybrid vehicles from 16.7 mpg to 17.5
mpg, or by 5%. Hybrid city fuel economy increased from 35.9 mpg to 38.6 mpg, or by 8%.
With an 18% weighting factor for running fuel use at 20°F, this means that the current approach
in this area could be under-estimating onroad fuel economy by 1% for non-hybrids and 2% for
hybrids.
The second alternative approach assumes that the effect of cold temperature on running
fuel use over Bags 2 and 3 of the FTP applies to all city driving, including the US06 city cycle.
Therefore, under this approach, running fuel use at 75°F is estimated using the 43/41/16%
weights for Bag 2, Bag 3 and US06 city cycles. Running fuel use at 20°F is equal to running
136
-------
fuel use at 75°F plus the difference in running fuel use over a 50/50% weight of Bags 2 and 3 at
20°F and 75°F. This approach yields the same effect as those of the first alternative approach.
The city fuel economy of non-hybrid vehicles decreases by 1% and that for hybrids decreases by
2%.
The third and final alternative approach is similar to the second approach in that it
assumes that the effect of cold temperature on running fuel use over Bags 2 and 3 of the FTP
applies to all city driving, including the US06 city cycle. However, rather than calculating an
effect of cold temperature in terms of an absolute difference in fuel consumption, it calculates a
relative difference in percentage terms and applies this percentage to relative fuel consumption at
75°F. Therefore, under this approach, running fuel use at 75°F is estimated using the 43/41/16%
weights for Bag 2, Bag 3 and US06 city cycles. Running fuel use at 20°F is equal to running
fuel use at 75°F times the percentage increase in running fuel use over a 50/50% weight of Bags
2 and 3 at 20°F and 75°F. This approach yields roughly the same effect as those of the first and
second alternative approaches. The city fuel economy of non-hybrid vehicles decreases by 1%
and city fuel economy for hybrids decreases by 2%.
137
-------
Chapter III References
1. Glover, E. L. and D. J. Brzezinski. Exhaust Emission Temperature Correction Factors for
MOBILE6: Adjustments for Engine Start and Running LA4 Emissions for Gasoline Vehicles
(M6.STE.004). U.S. Environmental Protection Agency, No. EPA420-R-01-029, April 2001.
Website: http://www.epa.gov/otaq/models/mobile6/r01029.pdf
2. Koupal, J., and L. Landman, E. Nam, J. Warila, C. Scarbro, E. Glover, R. Giannelli.
MOVES2004 Energy and Emissions Report -Draft Report. U.S. Environmental Protection
Agency, No. EPA420-P-05-003, March 2005, pp 57-63.
Website: http://www.epa.gov/otaq/models/ngm/420p05003.pdf
3. California Air Resources Board. Public Meeting to Consider Approval of Revisions to the
State's On-Road Motor Vehicle Emissions Inventory - Technical Support Document.
California Environmental Protection Agency, March 2000. See Section 6.7 (Start Correction
Factors). Website: http://www.arb.ca.gov/msei/on-road/doctable_test.htm
4. U.S. Environmental Protection Agency. MOVES 2004 - Software Design and Reference
Manual. U.S. EPA Office of Transportation and Air Quality, September 8, 2004. See data
table ZoneMonthHour. ZoneMonthHour is an information table which contains temperature
and humidity information. Table can be accessed by downloading and installing the MOVES
2004 model, which can be found at http://www.epa.gov/otaq/ngm.htm#moves2004.
5. U.S. Environmental Protection Agency and E.H. Pechan & Associates, Inc. Documentation
for the Draft 2002 Mobile National Emissions Inventory. Prepared for U.S. Environmental
Protection Agency/OAQPS, Research Triangle Park, NC, March 2005.
Web site: http ://www. epa.gov/ttn/chief/net/2002inventory. html
6. Glover, E. L. and D. J. Brzezinski. Soak Length Activity Factors for Start Emissions
(M6.FLT.003). U.S. Environmental Protection Agency, No. EPA420-R-01-011, April 2001.
See Tables 2a, 2b, 3a, 3b. Website: http://www.epa.gov/otaq/models/mobile6/r01011.pdf
7. U.S. Environmental Protection Agency. MOVES 2004 - Software Design and Reference
Manual. U.S. EPA Office of Transportation and Air Quality, September 8, 2004. See data
table MonthVMTFraction. MonthVMTFraction is a distribution information type table that
contains a distribution of each sourceTypelD's VMT across the MonthsOfAny Year. Table
can be accessed by downloading and installing the MOVES 2004 model, which can be found
at http://www.epa.gov/otaq/ngm.htm#moves2004.
8. Beardsley, M., and D. Brzezinski, B Giannelli, J. Koupal, S. Srivastava. MOVES2004
Highway Vehicle Population and Activity Data - Draft. U.S. Environmental Protection
Agency, No. EPA420-P-04-020, December 2004. See Table 15-1.
Website: http://www.epa.gov/otaq/models/ngm/420p04020.pdf
9. Ibid.
138
-------
10. U.S. Department of Transportation, Federal Highway Administration. Highway Statistics
2003. See Table VM-1. Website: http://www.fhwa.dot.gov/policy/ohim/hs03/htm/vml.htm
11. Davis, S. C. and S. W. Diegel. Transportation Energy Data Book: Edition 24. Oak Ridge
National Laboratory, prepared for the U.S. Department of Energy, Report No. ORNL-6973,
December 2004, Table 4.3. Website: http://cta.ornl.gov/data/download24.shtml
12. U.S. Environmental Protection Agency. Federal Test Procedure Review Project: Preliminary
Technical Report. U.S. Environmental Protection Agency, No. EPA420-R-93-007, May
1993. Website: http://www.epa.gov/otaq/regs/ld-hwy/ftp-rev/ftp-tech.pdf
13. U.S. Department of Transportation, Bureau of Transportation Statistics. Highlights of the
2001 National Household Travel Survey. Report No. BTS-03-05, 2003. See Tables 2 and
A-8. Website:
http://www.bts.gov/publications/highlights_of_the_2001_national_household_travel_survey/
14. Hu, P. S. and T. R. Reuscher. Summary of Travel Trends - 2001 National Household Travel
Survey. Prepared for the U.S. Department of Transportation, Federal Highway
Administration, December 2004. See Table 3.
Website: http://nhts.ornl.gov/2001/pub/STT.pdf
15 Guensler, Randall, Seungju Yoon, Hainan Li, and Vetri Elango, "Atlanta Commute Vehicle
Soak and Start Distributions and Engine Starts per Day: Impact on Mobile Source Emission
Rates - DRAFT FOR REVIEW," School of Civil and Environmental Engineering, Georgia
Institute of Technology, Atlanta, GA, for the U.S. Environmental Protection Agency,
February 19, 2005.
16 Rykowski, Richard, U.S. EPA, Memorandum to the Record, "Average Trip Length in
Atlanta," October 11, 2006.
17. Hu, P. S. and T. R. Reuscher. Summary of Travel Trends - 2001 National Household Travel
Survey. Prepared for the U.S. Department of Transportation, Federal Highway
Administration, December 2004. See Table 3.
Website: http://nhts.ornl.gov/2001/pub/STT.pdf
18. Ibid., see Table 13-1.
19. U.S. Environmental Protection Agency. Development of Methodology for Estimating VMT
Weighting by Facility Type (M6.SPD.003). U.S. Environmental Protection Agency, No.
EPA420-R-01-009, April 2001.
Website: http://www.epa.gov/otaq/models/mobile6/r01009.pdf
20. Rykowski, Richard A., Edward K. Nam, and George Hoffman, "On-road Testing and
Characterization of Fuel Economy of Light-Duty Vehicles," SAE 2005-05FL-174.
139
-------
21. U.S. Environmental Protection Agency. Response to Comments for the Final Regulations for
Revisions to the Federal Test Procedure for Emissions from Motor Vehicles. U.S.
Environmental Protection Agency, August 15, 1996.
Web site: http ://www. epa. gov/otaq/regs/ld-hwy/ftp-rev/sftp-rtc. pdf
22. U.S. Environmental Protection Agency. Support Document to the Proposed Regulations for
Revisions to the Federal Test Procedure: Detailed Discussion and Analysis. U.S.
Environmental Protection Agency, January 31, 1995.
Web site: http ://www. epa. gov/otaq/regs/ld-hwy/ftp-rev/ftp-supp .pdf
23. Meisner, B.N. and Graves, L.F. Apparent Temperature. Weatherwise, August 1985, 211-
213.
24. Koupal, J. W. Air Conditioning Activity Effects in MOBILE6 (M6.ACE.001). U.S.
Environmental Protection Agency, No. EPA420-R-01-054, November 2001.
Website: http://www.epa.gov/otaq/models/mobile6/r01054.pdf
25. U.S. Environmental Protection Agency. Air Conditioning Survey Data, Phoenix AZ, 1994.
Website: http://www.epa.gov/otaq/sftp.htm#phoenix. Excel or CSV data format.
26. Nam, Edward K., "Understanding and Modeling NOx Emissions From Air Conditioned
Automobiles," 2000, SAE #2000-01-0858.
27. Nam, Edward K., "Understanding and Modeling NOx Emissions From Air Conditioned
Automobiles," 2000, SAE #2000-01-0858.
28. Mitcham, A., Fernandez, A., & Bochenek, D. "Impacts of Ambient Temperature and Air
Conditioning Usage on Fuel Economy" U.S. EPA, Office of Transportation & Air Quality,
2005.
29. Johnson, Valerie H., "Fuel Used for Vehicle Air Conditioning: A State-by-State Thermal
Comfort-Based Approach," SAE 2002-01-19574, 2002.
30. Rugh, John P, Valerie Hovland, and Stephen O. Andersen, "Significant Fuel Savings and
Emission Reductions by Improving Vehicle Air Conditioning," SAE ,
31. Hellman, Karl H. and J. Dillard Murrell, U.S. EPA, SAE 820791, 1982, Tables 5 and 6.
32. U.S. Department of Transportation. Final Regulatory Impact Analysis: Tire Pressure
Monitoring System FMVSS No. 138. National Highway Traffic Safely Administration,
National Center for Statistics and Analysis, March 2005. Website:
http://www.nhtsa.gov/staticfiles/DOT/NHTSA/Rulemaking/Rules/Associated%20Files/TPM
S-2005-FMVSS-Nol38.pdf
33. U.S. Department of Transportation. National Automotive Sampling System -Tire Pressure
140
-------
Special Study. National Highway Traffic Safety Administration, National Center for
Statistics and Analysis , July 26, 2001. Website: http://www-nrd.nhtsa.dot.gov/pdf/nrd-
0 l/NRDmtgs/2001/0701 TirePressure.pdf
34. U.S. Department of Transportation, National Highway Traffic Safety Adminstration. Federal
Motor Vehicle Safety Standards; Tire Pressure Monitoring Systems; Controls and Displays -
Final Rule. U.S. Government Printing Office, Federal Register Volume 70, No. 67, p.
18136, April 8, 2005. Website:
http://www.nhtsa.dot.gOv/cars/rules/rulings/TPMSfmalrule.6/TPMSfmalrule.6.html
35. "Passenger Car Fuel Economy," US EPA, EPA 460/3-80-010, September 1980.
36. Gillespie, Thomas D., Fundamentals of Vehicle Dynamics, Society of Automotive
Engineers, 1992 pp. 98-102.
37. Stanard, Alan P., Southwest Research Institute, "VOC/PM Cold Temperature
Characterization and Interior Climate Control Emissions/Fuel Economy Impact, Final Report
Volume II, for U.S. EPA, October 2005.
38. "Analysis of In-Use Fuel Economy Data: Stage I," Technical Report, U.S. EPA, August,
1982.
39. U.S. Department of Energy, U.S. Environmental Protection Agency. "YourMPG" is a feature
of the http://www.fueleconomy.gov website. Select "MPG Estimates from Users." Direct
internet address: http://www.fueleconomy.gov/mpg/MPG.do?action=browseList. YourMPG
data as of September 12, 2005 is available as a Microsoft Excel spreadsheet in the public
docket for this rulemaking.
40. U.S. Environmental Protection Agency. Federal Test Procedure Review Project: Preliminary
Technical Report. U.S. Environmental Protection Agency, No. EPA420-R-93-007, May
1993.
41. Sierra Research, Inc. Development of Speed Correction Cycles - Draft (M6.SPD.001).
Prepared for the U.S. Environmental Protection Agency, National Vehicle and Fuel
Emissions Laboratory, June 26, 1997. Sierra Research No. Report No. SR97-04-01.
Website: http://www.epa.gov/otaq/models/mobile6/m6spd001 .pdf
42. U.S. Environmental Protection Agency. Regulatory Impact Analysis (Final Rule) - Federal
Test Procedure Revisions. U.S. EPA, August 15, 1996. See page 14.
Website: http://www.epa.gov/otaq/regs/ld-hwy/ftp-rev/sftp-ria.pdf
141
-------
Chapter IV: Economic Impacts
A. Testing and Facilities Burden
This section gives additional details for the Cost Analysis under Part VI. of the Preamble,
"Projected Cost Impacts ."
The final rule requires calculation of fuel economy values based on the five-cycle
formulae beginning with model year 2011 for some engine families. As discussed in detail
elsewhere, for model years 2008 through 2010, manufacturers may use the mpg-based
calculation for the five-cycle fuel economy values or they may conduct voluntary testing. For
model year 2011 and after, if the five-cycle city and highway fuel economy values for an
emission data vehicle group are within 4% and 5% of the mpg-based regression line,
respectively, then all the vehicle configurations represented by the emission data vehicle (e.g., all
vehicles within the vehicle test group) would use the mpg-based approach. Vehicles within a test
group falling below the 5% tolerance band for highway fuel economy values would be required
to conduct US06 tests; those falling below the city fuel economy band would be required to
conduct SC03, US06, and Cold FTP tests. In addition, we expect that some of these vehicles
falling below the tolerance levels may be eligible to estimate fuel economy for a given test
through the application of analytically derived fuel economy (ADFE) values. Some data is
currently available for vehicles that have conducted all five tests; based on this data, EPA has
estimated the number of vehicles for which additional testing would be required because they fall
below the 4 and 5 percent bands, as discussed below.
We prepared a range of burden estimates for this analysis and the discussion will mention
minimum and maximum burden scenarios. These low and high estimates are intended to provide
EPA's estimate of the outer boundaries of the likely testing and information costs, and EPA
solicited comments on the basis of these estimates, including the number of additional tests, and
the costs of performing such tests and of the additional tests that will be likely under the new
regulations. EPA received no comments on the basis of these estimates, the number of additional
tests, or the basic cost estimates for performing tests as presented in the proposal. Some
comments were received on more specific costs issues, and these have resulted in some
modifications to the cost estimate that will be noted below.
1. Test Volume
a. Testing Burden for MY 2008 through 2010
EPA estimates no additional tests during MY 2008 through MY 2010 based on the fact
that the mpg-based fuel economy estimates will be available for all manufacturers. No
additional testing would be required because manufacturers simply apply the mpg-based scale of
adjustments to the same FTP and HFET test results that they otherwise would conduct for the
fuel economy labeling program. While manufacturers have the option of conducting and
reporting full five-cycle test results, such tests are not required. This cost analysis is limited to
burdens that are mandated by the final rule.
142
-------
b. Testing Burden for MY 2011 and After
The term "Test Volume" includes the labor and operations and maintenance (O&M) costs
for running additional tests under the proposed requirements. Because of EPA's facilities cost
methodology, it also sets the capital cost estimates for ongoing facilities costs, as discussed
below.
Based on MY 2004 data, 1250 fuel economy vehicles were tested with the FTP and
highway fuel economy tests. (The figure is approximate because the city FTP test may be used
and recorded primarily as a fuel economy test, an emissions test, or both.) Data show that 330
Supplemental FTP (US06 and SC03) tests were conducted and 220 Cold FTP tests.
Consequently, if all fuel economy vehicles were required to conduct full five-cycle tests,
approximately 920 additional Supplemental FTP tests and 1,030 Cold FTP tests would be
required. Based on an analysis of our 615 vehicle dataset, we estimate that 8% of the test groups
will fall outside a band of (= ~ 4% of the regression for the city test and 23% outside a band of
(=~ 5% of the highway regression. Taking the 2004 numbers above as a baseline, 92% of the
additional SC03 and Cold FTP tests otherwise required would therefore be avoided for city fuel
economy; 77% of the additional US06 tests would be avoided. Thus, for example, the initial
estimate of increased testing burden for SC03 would be 8% of the difference between 1250 and
330.
This approach is retained in the final cost analysis, with one adjustment. The percent of
vehicles triggering additional testing requirements because they fall outside the tolerance bands
for the city and highway tests should only count those that are below the band in both cases; that
is, only those with fuel economy lower than 4 and 5 percent below the regressions, respectively.
With this correction applied to our updated 615-vehicle dataset, 4 percent of the test groups
would trigger additional testing because they fall more than 4 percent below the city regression
line, and 13 percent because they would fall more than 5 percent below the highway line. Thus,
for example, the initial estimate of the increased testing burden for SC03 would be 4% of the
difference between 1250 and 330, rather than 8%. The effect of this correction is that the
baseline estimated ranges of additional tests in the proposal of 169-212 US06 tests, 59-74 SC03
tests, and 66-82 cold FTP tests, become 96-120, 29-37, and 33-41, respectively, for the final rule.
Finally, the high and low estimates under these assumptions are generated by differing
estimates of the effect of another feature that will be available for MY 2011 and after: an
expanded use of analytically derived fuel economy (ADFE) as an alternative to conducting
vehicle tests. Current guidance (CCD-04-06) limits ADFE to 20% of the values that would
otherwise be derived from tests; the 1250 test baseline already excludes such analytically derived
results. Expanded ADFE guidance will be prepared in time for MY 2011 to allow for derivation
of fuel economy values for some of the additional test cycles that otherwise would be required as
described above. The low and high burden estimates assumes that 20% and 0% of the additional
tests would thereby be avoided, respectively.
143
-------
1) Fuel Economy Labeling for Medium-Duty Passenger Vehicles (MDPVs)
As discussed earlier, MDPVs will be included in the labeling program beginning with
model year 2011. This change is congruent with NHTSA's expansion of the CAFE program to
include MDPVs beginning the same model year. As discussed in Section II.D, we are finalizing
fuel economy test methods for MDPVs that will not require additional testing beyond that which
will already be required by the CAFE program beginning in model year 2011 (i.e., the FTP and
HFET tests). Therefore, we are projecting no additional costs in this final rule to extend labeling
to MDPVs.
2) Cold FTP Diesel Testing
The estimated cost impact of requiring cold FTP testing for light-duty diesel vehicles is
small. As an example, in model year 2006, only five light-duty diesel vehicles were certified for
sale in the U.S. A total of eight city/highway tests were performed on those vehicles to
determine fuel economy estimates. Applied to the 2006 model year, our proposal would have
required that an additional maximum of eight cold FTP tests be performed in addition to the
city/highway tests. Our proposed cost analysis accounted for additional cold FTP testing across
the entire automotive industry, including diesel vehicles.
While anticipating the makeup of the MY2011 diesel fleet is uncertain at this point, it
seems prudent to anticipate some addition in the number of diesel vehicles certified and, as a
result, we have assumed a doubling in the number of light duty diesels certified by MY 2011,
and thus increased the number of light-duty diesel test groups to 10. This has increased the
estimated Cold FTP test volume from 66 to 82 tests (proposed) and 33-41 tests as corrected
above, to 41 to 49 tests (final). The consequent adjustment in testing costs is approximately
$20,000 per year. This adjustment also has an effect on the estimated capital costs for testing
facilities (see below for the methodology for estimating facility capital costs).
A separate capital cost addition for Cold FTP diesel testing facility upgrades is discussed
below under Facility Burden.
The labor and O&M costs of conducting these estimated tests are derived from prior
Information Collection Requests submitted for EPA's light duty certification program. Those
estimates are based on the number of tests and the hours of labor used at EPA's testing facility
combined with industry data supplied in response to questionnaires; these have been somewhat
adjusted to reflect current information. These costs are estimated to range from $1,860 to $2,441
per test. These costs per test are applied to the numbers of tests estimated under the minimum
and maximum scenarios above, and now amount to $343,000 to $424,000 and 5,000 to 6,200
hours per year for MY 2011 and after, compared to $$606,000 to $757,000 and 8,800 to 11,000
hours in the proposal.
This analysis is summarized in the tables below:
144
-------
Table IV.A-1.
Estimated Cost per Test: Proposed
and Final
Test Cycle
FTP/HWY
US06
SC03
Cold FTP
Cost/Test
$1,860
$1,860
$2,206
$2,441
Hours/Test
30
30
30
30
Table IV.A-2. Estimated Increase in Number of Tests for Model Years 2008-2010:
Proposed and Final
Test Cycle
FTP/HWY
US06
SC03
Cold FTP
2004
Model
Year
Number of
Tests
1250
330
330
220
Increase In Number of Tests: MY 2008 - 2010
Min Tests
0
0
0
Min
Increase =
Min$
Increase
$0
$0
$0
$0
Max Tests
0
0
0
Max
Increase =
Max$
Increase
$0
$0
$0
$0
Increase in Hours
Min
0
0
0
0
Max
0
0
0
0
Table IV.A-3. Estimated Increase in Number of Tests for Model Years 2011 and Later:
Proposed
Test Cycle
FTP/HWY
US06
SC03
Cold FTP
2004
Model
Year
Number of
Tests
1250
330
330
220
Increase In Number of Tests: MY 2010 And
After
Min Tests
169
59
66
Min
Increase =
Min$
Increase
$314,861
$129,907
$160,904
$605,672
Max Tests
212
74
82
Max
Increase =
Max$
Increase
$393,576
$162,384
$201,130
$757,090
Increase in Hours
Min
5,078
1,766
1,978
8,822
Max
6,348
2,208
2,472
11,028
145
-------
Table IV.A-4. Estimated Increase in Number of Tests for Model Years 2011 and Later:
Final
Test Cycle
FTP/HWY
US06
SC03
Cold FTP
2004
Model
Year
Number of
Tests
1250
330
330
220
Increase In Number of Tests: MY 2010 And
After
Min Tests
96
29
41
Min
Increase =
Min$
Increase
$177,965
$ 64,954
$ 99,979
$342,898
Max Tests
120
37
49
Max
Increase =
Max$
Increase
$222,456
$ 81,192
$120,092
$423,740
Increase in Hours
Min
2,870
883
1,229
4,982
Max
3,588
1,104
1,476
6,168
2. Facilities Burden
"Facilities" refers to the capital costs for constructing facilities to accommodate the
increases in test volume estimated in the table above. For these capital costs we used an FTP
facility cost of $4 million per facility able to perform 750 US06 tests per year, a cost of $9
million for an environmental test facility able to conduct 300 to 428 SC03 tests per year, and $10
million for an environmental facility able to conduct 300 to 428 Cold FTP tests per year. The
new tests were deemed to require these facilities in proportion to the number of tests needed, and
the costs were then annualized over ten years with a 7% discount rate. This is likely a very
conservative assumption since it does not attempt to account for the excess capacity that exists in
manufacturers' current test facilities. We assume that there is no excess capacity in our analysis.
Furthermore, consistent with other information burden analyses for the emissions and fuel
economy programs, we have considered these as ongoing capital rather than startup capital costs
(i.e., as the facilities depreciate they are continually being replaced). Annualized and depreciated
over ten years at 7%, these capital costs per year under the above analysis under the proposal
were $0 for each of model years 2008, 2009 and 2010, and range from $524,000 to $866,000 per
year for model years 2011 and after. Under the corrected and adjusted number of tests described
the preceding section, the final annualized and discounted capital costs for model year 2011 and
after range from $375,000 to $560,000.
In addition, commenters raised a number of technical issues regarding laboratory
configurations and the difficulty of establishing cold test facility retrofits to accommodate diesel
testing without a transition period. The 2011 effective date of the requirement is intended to
address some of these concerns, particularly the lead time needed to implement laboratory
adjustments, but we recognize that some facility updates will still be necessary. An additional
capital cost of $55,000 for each often manufacturers has been added the proposed facilities costs
to account for these adjustments in the final cost estimate. This figure includes the cost of flame
ionization detectors (FIDs) as well as heated sample probes, sample lines and sample filters.
This cold facility upgrade for diesel testing, along with the corrected and adjusted
projected number of tests, accounts for the changes in facility capital costs from the proposal.
146
-------
This analysis is summarized in the tables below:
Table IV.A-5. Estimated Facility Costs: Proposed
Un-depreciated capital costs
FTP/HW
US06
SC03
Cold FTP
Total
Depreciated, lOyrs @ 7 %
MY 2008 -20 10
Minimum
$0
$0
$0
$0
$0
$0
Maximum
$0
$0
$0
$0
$0
$0
MY 20 11 And After
Minimum
$0
$902,827
$1,238,131
$1,540,187
$3,681,144
$524,112
Maximum
$0
$1,128,533
$2,208,000
$2,746,667
$6,083,200
$866,111
Table IV.A-6. Estimated Facility Costs: Final
Un-depreciated capital costs
FTP/HW
US06
SC03
Cold FTP
Cold FTP facility upgrades
Total
Depreciated, lOyrs @ 7 %
MY 2008 -20 10
Minimum
$0
$0
$0
$0
$0
$0
$0
Maximum
$0
$0
$0
$0
$0
$0
$0
MY 20 11 And After
Minimum
$0
$510,293
$619,065
$957,009
$550,000
$2,636,368
$375,359
Maximum
$0
$637,867
$1,104,000
$1,640,000
$550,000
$3,931,867
$559,809
3. Startup Burden
"Startup" refers to one-time costs beginning with model year 2008 to implement the new
requirements in the final rule. These startup burdens are primarily information technology costs
involving familiarization with the new data reporting requirements and reformatting management
information systems to carry out and report the necessary data and calculations. With the
exception noted below, all these burdens are add-ons to well established reporting requirements:
manufacturers already submit data to EPA on all five test cycles, have the option of applying
analytically derived fuel economy numbers, and report vehicle class determinations and
supporting information. This part of the proposed estimate assumed four weeks for four
information technology specialists for analysis and coding, and four weeks for two IT specialists
for testing, at $100 per hour, for 35 manufacturers, based on a count of manufacturers in 2004
with an allowance for inter- and intra-corporate relationships. The estimate also includes 1,120
hours industry wide for label redesign. Startup information technology costs finally also include
one-time costs and hours for implementing US06 split phase sampling, assuming one to seven
days of programming. The remaining startup cost is the one-time costs associated with validation
tests for the split phase sampling, and assumes one to seven tests at the costs per test given
above. These one-time tests are not considered to entail ongoing capital costs but are treated as
startup capital costs. EPA's proposal estimated all startup costs, discounted at 7% and annualized
over ten years, as $526,000 to $615,000 and 3,815 to 4,718 hours.
147
-------
a. Dual Information Systems Needed for CAFE/Gas Guzzler and Labeling
Currently, EPA's data system has a single "path" through which fuel economy data is
passed. Up to the point of determining the so-called "model type" fuel economy, the calculations
performed by the database are identical for both CAFE and labeling purposes. Then, the single
path diverges: The label value is determined by applying the 10% and 22% downward
adjustment factors to the city and highway model-type values, whereas, for CAFE, the model-
type values are adjusted by a minor, but statutorily required, adjustment to account for test
procedure differences since 1975.
The new rule would now require two separate data paths, since the new derived 5-cycle
label calculations for labeling purposes would be carried through "from the bottom up" resulting
in a model type fuel economy that needs no adjustment for "real world" conditions (i.e., model-
type fuel economy for labeling purposes would be different from model-type fuel economy for
CAFE purposes.). Consequently, it would no longer be possible to use the single existing
calculation systems to determine the CAFE values.
EPA agrees that there may be some convenience in applying the derived 5-cycle equation
at the model-type level for manufacturers who wish to use the same data management system for
reporting CAFE and label values. This may be particularly true for the early part of the
transition period. However, this approach is not available for the vehicle-specific 5-cycle label
calculations, and any manufacturers who use it during the phase-in period during the 2008-2010
model years will encounter an information cost not contemplated in the proposal; a dual
calculation procedure will be needed. Similarly, a dual calculating system will be needed for
model years 2011 and after.
The cost analysis has been updated to account for this increased information system
startup burden. Based on a projection of EPA's information development contract costs, and an
estimate of the portion of those costs attributable to the dual information system possibility, we
have increased the industry information startup costs (unamortized) by $933,450. This increases
the annualized and discounted startup costs to $659,000 to $748,000 for the industry as a whole,
from the proposed level of $526,000 to $615,000.
The Final Technical Support Document has also been updated to delete labor hours
attributed to startup costs. Startup costs are properly treated as capital costs, not labor.
The startup burdens are summarized in the following tables:
148
-------
Table IV.A-7. Estimated Startup Costs: Proposed
Item
Information Tech/Paperwork
Adjustment to new FE and label value
computations and reporting of FE data
for SFTP and Cold FTP and new ADFE
calculations— analysis, code
development, and testing; label redesign
Sample system changes for US06 split
phase
O&M
Validation testing form US06 split phase
sampling
TOTAL
Depreciate 10 years at 7%
Cost
Minimum
Maximum
Hours
Minimum
Maximum
$3,472,000
$28,000
$3,472,000
$196,000
34,720
280
34,720
1,960
$195,300
$3,695,300
$526,128
$651,000
$4,319,000
$614,928
3,150
38,150
3,815
10,500
47,180
4,718
Table IV.A-8. Estimated Startup Costs: Final
Item
Information Tech/Paperwork
Adjustment to new FE and label value
computations and reporting of FE data
for SFTP and Cold FTP and new ADFE
calculations— analysis, code
development, and testing; label redesign
Sample system changes for US06 split
phase
O&M
Validation testing form US06 split phase
sampling
TOTAL
Depreciate 10 years at 7%
Cost
Minimum
Maximum
Hours
Minimum
Maximum
$4,405,450
$28,000
$4,405,450
$196,000
0
0
0
0
$195,300
$4,628,750
$659,002
$651,000
$5,252,450
$747,830
0
0
0
0
0
0
4. Summary
The combined results of the above tables can be summarized as follows: the final total
estimated costs for each of Model Years 2008, 2009, and 2010 range from $659,000 to $748,000,
and for Model Years 2011 and after, from $1,377,000 to $1,731,000. This is shown the tables
below:
149
-------
Table IV.A-9. Estimated Total Costs: Proposed
MY 2008-2010
MY2011 and After
Min
Max
Min
Max
Test Volume
Facilities (annualized
10yrs/7%)
Startup: one-time
IT/Paperwork and O&M
(annualized 10yrs/7%)
Total
$0
$0
$526,128
$526,128
$0
$0
$614,928
$614,928
$605,672
$524,112
$526,128
$1,655,912
$757,090
$866,111
$614,928
$2,238,129
Table TV.A-10. Estimated Total Costs: Final
MY 2008-2010
MY2011 and After
Min
Max
Min
Max
Test Volume
Facilities (annualized
10yrs/7%)
Startup: one-time IT and
validation (annualized
10yrs/7%)
Total
$0
$0
$659,002
$659,002
$0
$0
$747,830
$747,830
$342,898
$375,359
$659,002
$1,377,259
$423,740
$559,809
$747,830
$1,731,380
The final combined burden hours shown in these tables is 4,982 to 6,168 hours for each model
years beginning 2011 and continuing thereafter, as summarized below:
Table IV.A-11. Estimated Total Hours: Proposed
MY 2008-2010
Min
Max
MY2011 and After
Min
Max
Test Volume (Labor,
O&M)
Facilities (Capital,
annualized 10yrs/7%)
Startup: one-time
IT/Paperwork and
O&M (Capital,
annualized 10yrs/7%)
Total
0
0
3,815
3,815
0
0
4,718
4,718
8,822
0
3,815
12,637
11,028
0
4,718
15,746
150
-------
Table IV.A-12. Estimated Total Hours: Final
MY 2008-2010
Min
Max
MY2011 and After
Min
Max
Test Volume (Labor,
O&M)
Facilities (Capital,
annualized 10yrs/7%)
Startup: one-time
IT/Paperwork and
O&M (Capital,
annualized 10yrs/7%)
Total
0
0
0
0
0
0
0
0
4,982
0
0
4,982
6,168
0
0
6,168
B. Impact on Confirmatory Testing of Vehicles
EPA can conduct confirmatory testing of any vehicle tested for emissions or fuel
economy compliance purposes for any reason it deems necessary. EPA's fuel economy
regulations currently specify certain conditions under which test results from manufacturers are
to be confirmed, either by EPA at its own laboratory, or by the manufacturer. See 40 CFR
600.008-01. These conditions are tied to the FTP and highway test procedures, but do not
address the US06, SC03 and Cold temperature FTP tests that will now included in fuel economy
measurement. There are separate confirmatory test provisions in EPA's emission compliance
regulations at 40 CFR 86.1835-01, which would continue to apply to all five test cycles.
Confirmatory testing is generally indicated for fuel economy purposes when the fuel
economy falls near the cut point for Gas Guzzler Tax assessment, when the fuel economy (either
city or highway) is unexpectedly high for that vehicle or falls near the leader within the
comparable class. EPA provides guidance to manufacturers defining the actual criteria and
cutpoints. Confirmatory testing is also required when the test results are close to or exceed the
federal emission standard associated with the test. EPA also selects a random number of
vehicles for confirmatory testing, and will request confirmatory testing on vehicles employing
new or unusual technology.
In considering the impact of confirmatory testing of US06, SC03 and Cold temperature
tests, some of the above criteria may or may not apply. For instance, the Gas Guzzler tax is
based only on the FTP and Highway test procedures, so there would be no impact on
confirmatory testing for the other procedures. However, if, in the course of performing a US06,
SC03 or Cold temperature fuel economy test, an applicable emission standard should be
exceeded, we believe that a confirmatory test would be necessary, per the current regulations.
Another reason to conduct additional tests is to resolve a disparity between confirmatory
fuel economy test results. If a manufacturer's initial fuel economy test result does not compare
closely to that of the confirmatory test, EPA requires a retest. (This requirement helps to
establish correlation between manufacturer and EPA testing, providing assurance that
151
-------
manufacturer testing is properly conducted.) This currently applies only to FTP and Highway
testing. If this requirement was applied equally to the US06, SC03 and Cold temperature FTP
fuel economy tests, this could result in the need for additional tests. Obviously, EPA does not
have established fuel economy correlation data for the US06, SC03 or Cold temperature FTP
tests, because fuel economy has never been required to be measured. There is not enough fuel
economy data to determine what constitutes reasonable correlation from one test to the next.
Therefore, EPA does not plan at this time to conduct retests of US06, SC03 or cold temperature
FTP tests on the basis of test-to-test fuel economy disparities. Once the final regulations for 5-
cycle fuel economy are implemented, and more correlation data becomes available, we will
assess whether or not we should establish retest criteria for US06, SC03 and cold temperature
FTP tests on the basis of fuel economy test-to-test disparities.
C. Changes to Label Format and Content
Manufacturers are currently required to print and post fuel economy window stickers on
all new cars and light trucks available for sale in the U.S. Our final rule does not change this
requirement, thus no new costs for these activities will be incurred. The final changes to the
format and content may require a one-time design change, but any additional costs are
anticipated to be very slight, and mitigated by the design templates for the new labels which EPA
is providing in the regulations. The reporting and recordkeeping requirements associated with
the fuel economy label are set forth in 40 CFR sections 600.312 to 600.314. These sections
require that manufacturers supply EPA with the label values and the data used to derive them,
and provide schedules for the updating of this information. Under the final rule, these values will
be recalculated and new data will be submitted. The costs for these efforts are very minimal and
are addressed in the startup cost figures above. There will be a similar one-time set-up charge
associated with the new label format based on the effort required for each manufacturer to apply
the new EPA templates to the labels they must print. This cost item also has been included in the
paperwork startup costs portion of the cost analysis.
D. Certification Fees
EPA collects fees from manufacturers to recover its costs associated with its emission
and fuel economy compliance efforts. The fee amount is adjusted from time to time according to
any changes to EPA's costs. The impact of the final rule changes is not anticipated to
significantly increase EPA's efforts with the fuel economy labeling program; therefore, we are
not proposing to adjust the fee amount at this time. However, we reserve the right to evaluate
our actual incurred costs once the final rule is implemented and to propose fee changes if deemed
appropriate.
152
-------
Appendix A: EPA Kansas City Test Program
During 2004-2005, EPA in association with the Coordinating Research Council,
STAPPA/ALAPCO, DOT and DOE/NREL,, recruited and tested over 600 privately owned
passenger vehicles in the Kansas City area. The vehicles included an assortment of compact
cars, mid-size cars, pick-ups and SUVs from a variety of manufacturers. The program was split
into 3 rounds (1, 1.5 and 2), each consisting of 120-300 vehicles. In all three rounds, vehicles
were recruited randomly from lists of vehicle registrations in the Kansas City area. Care was
taken to ensure that the sample were random with respect to the geographic location of the owner
and socio-economic status. In rounds 1 and 2, the desired sample of vehicles was stratified into
1 0
four groups of model years, with emphasis on older vehicles ' . The primary purpose of Rounds
1 and 2 was the quantification of particulate emissions, particularly those from high emitters. In
Round 1.5, only 2001 and later model year vehicles were sampled. (Details about the design and
performance of Round 1.5 are described the study's final report.3) The primary purpose of
Round 1.5 was the measurement of onroad fuel economy from vehicles for which we could
estimate 5-cycle fuel economy. This meant that we had to have fuel economy estimates over all
five cycles for these vehicles (i.e., that the vehicle had to be certified to the Supplemental FTP
standards). These standards began phasing in with the 2001 model year.
All of the vehicles In Rounds 1 and 2 were tested on a dynamometer, but only a subset of
the vehicles were instrumented with a Portable Emissions Measurement System (PEMS) and
tested in the hands of their owners. As these vehicles ranged in model year from 1968 to 2005,
very few of the vehicles tested in Rounds 1 and 2 had been certified to the Supplemental FTP
standards.
In Round 1.5, none of the vehicles were tested on a dynamometer. However, all of these
vehicles were instrumented with PEMS and had their fuel economy measured while being driven
in normal use by their owners. The round 1.5 vehicle fleet consisted of approximately 120
vehicles, including over 30 hybrid electric vehicles. The PEMS measures driving activity, as well
as second-by-second mass emissions of CO2, CO, HC, and NOx for roughly 24 hours while the
owners of the vehicles are utilizing their vehicles on the road under normal, real-world
conditions. Two aspects of the PEMS limit the duration of its operation. One is the capacity of
gas which is needed to operate the flame ionization detector used to measure HC emissions. The
other is battery capacity needed to operate both the emissions measurement equipment and the
onboard computer which scans vehicle activity and stores information. Using the carbon balance
method, fuel economy can be accurately calculated on a second by second basis.
Since the focus of this proposed rule is fuel economy labeling, we are solely interested in
the onroad fuel economy data collected during this program. We will not present, or discuss the
results of the dynamometer testing, nor the emission test results. Because of this, we will focus
primarily on Round 1.5, but will also include relevant data from Rounds 1 and 2, as appropriate.
The reader is referred to this report for further information related to vehicle recruitment,
selection, instrumentation, data processing and delivery.
The onroad fuel economy data will be used for several purposes:
1) To compare the onroad fuel economy for each vehicle to its fuel economy label values,
153
-------
2) To compare the driving activity of vehicles in Kansas City to the findings of past studies, and
3) To develop 5-cycle formulae for each vehicle and compare the 5-cycle fuel economy to
similar vehicles in our 5-cycle certification fuel economy database
Before moving to these three tasks, the following section will describe the methods we
used to process and quality assure the data obtained from the PEMS.
A. Quality Assurance
Upon initial delivery of the data, we performed a number of tests on the data to identify
equipment malfunctions. The removal of inaccurate data is referred herein as "flagging" (i.e.,
data are "flagged" when it meets a criterion designed to identify data which are not of acceptable
quality). Flagging was performed using the SAS statistical package.
The initial step in quality assurance involved segregating vehicles into those with
acceptable data and those with bad data. An entire set of bad data for a given vehicle was
removed for a variety of reasons. Sometimes, the data collected for a given vehicle was
completely missing vehicle speed or exhaust flow data, which are critical to proper processing of
the data. In some cases, the vehicle was only driven for a few miles before the equipment
stopped functioning. These data were insufficient to develop robust estimates of fuel use in the
various VSP bins and to have confidence that the measured onroad fuel economy was indicative
of that vehicle's typical use. In one case, the CO2 emissions were outside a reasonable range.
After this initial filtering of the data, there remained 9 vehicles from round 1, 97 vehicles from
round 1.5 (including 33 hybrids), and 42 from round 2. These all constitute "good tests". Data
from 18 vehicles were removed in this step.
The next step in data processing involved removing or modifying individual seconds of
data. We often found instances where the exhaust flow meter signal was zero, or erratic. There
were several instances where the flow meter stopped functioning properly permanently (i.e., for
the rest of the testing of that vehicle). This could have been due to freezing in the cold
temperatures or simply from an electronic failure. For non-hybrids, if the vehicle was moving,
but exhaust flow dropped to a value equivalent to engine off conditions, then the data was
flagged and omitted from the final dataset. Unfortunately, this condition was impossible to flag
systematically with hybrids, since the engine can shut off while the vehicle is on. Obvious cases
of lengthy equipment failures during hybrid testing were flagged by hand. However, exhaust
flowmeters were assumed to have been functioning properly during all hybrid runs.
Another frequent problem was that the signal from the vehicle's onboard diagnostic
(OBD) system, which provides measurements of vehicle speed, engine rpm, engine coolant
temperature and other vehicle information, disappeared permanently or acted erratically. This is
easy to spot since all of the OBD signals disappear at the same time. This can be caused by the
driver accidentally knocking the connector, or it can be a result of manufacturer software
settings. Fortunately, the PEMS units were equipped with a geographical positioning system
(GPS) and vehicle speed could be calculated from the output from the GPS unit. These GPS
estimates of vehicle speed were substituted for OBD speeds when the OBD speeds were absent
or obviously erratic.
154
-------
Sometimes the GPS speeds appeared to reflect anomalies (i.e., were clearly
discontinuous, or drifted unrealistically over time). This was more of a concern regarding the
calculation of vehicle acceleration than vehicle speed, due to the fact that required engine power
is very dependent on small changes in acceleration. To identify unrealistically high
accelerations, a cut-off was chosen at a specific power (the product of acceleration and speed, or
v*a) of 300 mph2/sec. Beyond this, the data were flagged (though these events were rare).
Choosing a lower (and probably still realistic) maximum value would have resulted in excessive
removal of data. . In addition, there were instances where the GPS speeds drifted slowly over
time, usually when the car was not operating. These time segments were removed.
Table A-l summarizes the processing of the data obtained with the PEMS.
Table A-l. Processing of Raw Data Obtained In Kansas City
Round 1 Round 1.5 Round 2
Vehicles tested onroad with PEMS 18 117 52
Vehicles with some PEMS data 13 109 47
Vehicles with acceptable PEMS data 9 97 42
Total seconds of acceptable PEMS data while vehicle is 53,319 501,405 181,915
in operation
Seconds of PEMS data removed 3,808 44,996 2,773
As can be seen, a total of 187 vehicles were equipped with PEMS units in Kansas City and the
equipment worked to some degree on 169 of these vehicles (90%). Data from an additional 22
vehicles were rejected due to obvious equipment malfunctions, leaving 148 vehicles (78% of the
original sample) with usable data. Of these 148 vehicles, 125 were conventional vehicles and 33
were hybrids. This usable activity and fuel economy data were obtained for a total of 736,000
seconds (over 200 hours) of driving from 148 vehicles, or roughly 80 minutes of driving per
vehicle.
B. Onroad Fuel Economy
Total fuel consumption for each vehicle was determined from the carbon balance of the
CO2, HC, and CO emissions. The total distance of driving was determined by summing vehicle
speed and multiplying by total time of operation. This total distance traveled was then divided
by total fuel consumption to determine onroad fuel economy.
EPA city and highway label fuel economy values were obtained from EPA mileage
guides. The test vehicles were matched to those tested in Kansas City to the closest degree
possible. The following figure compares the measured fuel economy to the 55/45 composite
label fuel economy from round 1.5 (newer vehicles). We segregated the vehicles into two
groups: conventional gasoline-fueled vehicles and hybrids. A linear regression with no constant
of the conventional vehicles showed nearly one-to-one correlation, with a slope of 1.006. The
correlation was also quite good (r-squared value of 0.77). The largest difference was only 6
mpg, or about 30%. Thus, the onroad fuel economy data indicate no offset from the current EPA
label values on average.
155
-------
The correlation of hybrid data shows much more scatter. This is partially explained by
the fact that only three hybrid models were tested, the Toyota Prius and the Honda Insight and
Civic. On average, hybrid fuel economy was 11% less than the composite EPA label values.
The average onroad fuel economy of the Toyota Prius vehicles was closer to their composite
label values than those for the two Honda models. On average, the onroad fuel economy of the
hybrids tested varied more than the conventional vehicles. This could be due to hybrids' greater
sensitivity to operating conditions which can either take full advantage of the hybrid technology
or essentially nullify it. The fact that most vehicles started out testing with a hot start likely
biased onroad fuel economy upwards to some degree. Thus, the actual shortfalls found would
have been greater to some degree if testing had begun with a cold start, which is more
representative of a typical day of driving.
Figure A-l.
60 T
Comparison Onroad to Current Label Economy: Kansas City
~50
D)
Q.
>.
E40
o
o
o
-------
There are two basic ways to display a SAFD. The more complex way is to show the
percentage of driving for each combination (or bin) of speed and acceleration. The simpler way
is to distinguish between those speed-acceleration bins for which onroad driving was found and
those which did not. For ease of comparison with the SAFDs of dynamometer cycles, we use the
simpler form of display here. We developed three SAFDs for the driving monitored in Kansas
City: one for all vehicles, one for hybrids and one for conventional, or non-hybrid vehicles.
The SAFD for all of the vehicles tested with PEMS units in Kansas City is shown in the
following figure. The breadth of onroad driving in Kansas City generally exceeds that of the
FTP and FIFET. Thus, for comparison, we show the SAFD for the FTP and FIFET, as it
represents an inner envelope of driving, per se. In some cases, the breadth of onroad driving in
Kansas City generally exceeds that of the US06 cycle, as well. However, in some cases, it does
not. Thus, in the figure, we indicate the driving in Kansas City which exceeds that of the FTP
and FIFET cycles. Then we show driving included in the US06 cycle which exceeds that found
in Kansas City. In most cases, where the Kansas City driving forms the edge of the envelope,
the US06 cycle also includes that type of driving. However, this is not true in all cases. Since
the onroad data and the dynamometer cycles provide speed and acceleration rates for one second
intervals and the HFET and US06 cycles are about 600 seconds long, the smallest amount of
driving which can fall into a single bin in these two cases is 0.16%. Of course, smaller
percentages of onroad driving can fall into a single bin due to the large number of observations
available. To be comparable with the driving cycles, we only considered onroad driving to
adequately populate a specific speed-acceleration combination if at least 0.1% of all onroad
driving fell into that bin.
Figure A-2. Speed-Acceleration Frequency Distribution: Kansas City Vs. Test Cycles
SPEED BIN
Imphl
15 | 20 | 25 | 30 | 35 | 40 | 45 | 50 | 55 | 60 | 65 | 70 | 75 | 80 | 85>\
KEY
Driving frequency covered by FTP/HFET style driving
Driving frequency covered by Kansas City Real-World Driving
Driving frequency covered by US06 style driving
KEY
"i
KC activity cut off < 0.1 %
18% of KC driving outside FTP/HFET driving envelope
0.6% of KC driving is outside US06 driving envelope
90% of US06 driving is within the 0.1% KC boundary
157
-------
Overall, 18% of the onroad driving activity (time based) in Kansas City fell outside of the
FTP/HFET envelope. This corresponds to 33% in VMT terms. However, only 0.6% fell outside
the US06 (0.4% of the VMT). These are data in the white cells, which are populated but fall
below 0.1% of the total. Overall, 25% of the driving activity in Kansas City fell outside the FTP
envelope alone (40% of the VMT). Comparing to most speed limits, 28% of the VMT was
compiled greater than 65 mph. The SAFD envelopes for hybrid and conventional vehicles did
not differ significantly from each other. However, the percentages of driving in the various bins
did vary, as will become more evident below when we evaluate the VSP frequency distributions
for the two types of vehicles.
We also show the SAFD driving activity found from a chase car study conducted in Los
Angeles in 2000 in this figure4. This study observed roughly 390 hours of vehicle operation.
According to the data, 20% of California urban driving lies outside the FTP/HFET envelope
(34% of the VMT). 1.2% fell outside the US06 envelope (1.3% of the VMT). 25% of the driving
occurred outside of the FTP alone (42% of the VMT). However note that much more of the rural
driving was outside the US06 envelope (3%).
Figure A-3. Speed-Acceleration Frequency Distribution: Urban California Vs. Test
Cycles
SPEED BIN
(mphl
10 | 15 | 20 | 25 | 30 | 35 | 40 | 45
50
55 | 60 | 65 | 70 | 75 | 80 | 85 | 90]
KEY
Driving frequency covered by FTP/HFET style driving
Driving frequency covered by real-world California urban style drivini
Driving frequency covered by US06
Driving frequency covered by real-world CA but NOT US06
CA activity cut off < 0.1 %
20% of CA URBAN driving is outside FTP/HFET driving envelope
1.2% of CA URBAN driving is outside US06 driving envelope
41% of CA RURAL driving is outside FTP/HFET driving envelope
3% of CA RURAL driving is outside US06 driving envelope
Moving to VSP, we calculated the level of VSP for each second of each vehicle's driving
from the vehicle's speed and acceleration, plus its road load characteristics and vehicle mass.
The equation used is shown below:
VSP = (Av + Bv2 + Cv3 + Mva)/M,
158
-------
where v is vehicle speed, a is acceleration, A, B, and C are the road load (coast-down)
coefficients of the vehicles and M is the mass of the vehicle.
This is the same method used in section IA.2 to convert the facility driving cycles in MOVES
into VSP frequency distributions. The only difference is that here, the constants A, B, C and M
are specific to the vehicle being monitored on the road. For the MOVES cycles, we used
coefficients indicative of the average car and light truck on the road.
Road grade can also be included in the above equation. We did not include this term,
because the altitude data from the GPS unit appeared quite unreliable. Therefore, the effect of
road grade is not reflected in our analysis and remains a source of uncertainty in the results. We
obtained the specific vehicle parameters from EPA's Inspection and Maintenance database based
on the model year, make and model of the vehicle. The vehicles were not weighed on a scale, so
the actual weights may vary somewhat from those in the database.
The Kansas City driving activity was binned by VSP and vehicle speed as described in
section IA.2. We used 26 VSP bins, which includes the addition of 9 high power bins beyond
those used currently in Draft MOVES2004. Bin 0 contains significant decelerations, while Bin 1
contains idle operation. Bins 11-19 include other operation below 25 mph. Bins 21-29 include
operation between 25 and 50 mph and Bins 33-39 include operation above 50 mph. Bins xl
contain operation with very low or negative engine power, while Bins x9 contain very high
power operation.
The following figure compares the Kansas City activity in terms of VSP bin to that used
in Draft MOVES2004 and described in section IA.2 (Table I A-12). The match in trends is
reasonable, though Draft MOVES2004 projects roughly 4% more activity in the high speed, high
power Bins 36-39 and 4% less activity in the lower power Bin 35.
Figure A-4. Kansas City and California VSP Frequency Distributions vs. MOVES
019 -
0-1
n nft -
n OR
n ro
n -
-
1
Tn
1
If
rrflrJl _n
1
D MOVES
Q California Urban
r
p
M
flflmifHl
1
-
1 n i
_r T_
f [1
0 1 11 12 13 14 15 16 17 18 19 21 22 23 24 25 26 27 28 29 33 35 36 37 38 39
VSP Bin
159
-------
We also show the driving activity found from a chase car study conducted in Los Angeles in
2000 in this figure5. This study observed roughly 390 hours of vehicle operation. This VSP
frequency distribution was estimated from the speed acceleration frequency distributions
developed from the chase car data6. The portion of driving found the Bins 36-39 in Los Angeles
matches that in MOVES2004, though there is about 1% less operation in Bins 26-29. One must
be cautious when comparing activity data from instrumented vehicles and chase cars. Chase cars
do not follow drivers throughout their entire trip, so operation on local neighbor roads is often
missed. Also, the VSP distribution from Draft MOVES2004 includes both urban and rural
operation, while those of Kansas City and Los Angeles are primarily urban. This could
introduce some differences, as well.
The following figure shows the VSP frequency distributions for the three groups of
Kansas City vehicles mentioned above: all vehicles, hybrids and non-hybrids. We removed idle
(bin 0) prior to calculating the VSP distributions for two reasons. One, idle percentages were
extremely high and if depicted in the figure, would have made the rest of the bars difficult to
read. Second, given that little fuel is consumed during idle, including the percentage of time at
idle distorts the comparison of those VSP bins where fuel consumption is more significant.
Figure A-5. VSP Frequency Distributions in Kansas City: Hybrids vs. Non-Hybrids
n 9
0-1 Q
01R
014.
0 12
D"
-------
either a Prius or a Civic, which were clearly designed for fuel efficiency with some sacrifice in
power. Another cause could be driver behavior. People purchasing hybrids may have an
unusually high interest in achieving high fuel economy on the road and drive their vehicles less
aggressively, knowing that aggressive driving reduces fuel economy. This could either represent
a change in driving behavior towards less aggressiveness with the purchase of a hybrid, or these
drivers may have driven their previous vehicle the same way.
Of particular interest is whether the difference in the operation of conventional vehicles
and hybrids will continue in the future or disappear. If the difference is due to a low power
vehicle design, then the difference depends on the design of future hybrids. Most of the recently
introduced hybrids match (or even exceed) the power output of their non-hybrid counterparts.
Thus, this difference in vehicle operation found in Kansas City may soon disappear.
If the drivers of hybrids in Kansas City always drove in the manner captured in this
study, then this implies that people conscious of fuel efficiency tend to be the first people to buy
hybrids. In this case, the difference in operation would continue until hybrids become the
dominant drivetrain. As more and more people purchase hybrids, the driving of hybrids would
likely become more aggressive, since the new hybrid purchasers are coming from the remaining,
more aggressive pool of drivers. The driving of conventional vehicles would also become more
aggressive as the less aggressive drivers of conventional vehicles buy hybrids. Overall driving
operation would likely remain the same, but the driving of the two pools would shift over time,
tending to remain distinct from each other.
Finally, the drivers of hybrids in Kansas City may have changed their driving behavior
when they purchased their hybrids. Some manufacturers of hybrids offer training classes or
videos to help hybrid owners get the most from their hybrid technology. If this trend continued,
hybrid driving activity would tend to always be less aggressive than that of conventional
vehicles. As more hybrids enter the fleet, the overall driving of the fleet would become less
aggressive. Even if the drivers of current hybrids have changed their driving behavior, the
question exists whether this trend would continue as hybrids become more popular.
This is the first systematic study of the onroad operation of hybrids in the hands of typical
owners. Only about 45 hours of operation were studied. Thus, there is significant uncertainty in
the difference found in the operation of conventional and hybrid vehicles. As discussed in
section LA, we are in the process of obtaining a large volume of operational (activity) data
obtained in Atlanta. Some hybrid vehicles may be included in this database. We plan to
compare the operation of conventional and hybrid vehicles in that study as soon as we receive
the data. Still, the potential impact of the difference in driving behavior for the two types of
vehicles is examined in section III.E.4.
D. Evaluation of 5-Cycle Approach to Fuel Economy Estimation
In section I.A.2.b, we developed combinations of dynamometer cycles which best
represented the driving activity represented in Draft MOVES2004. In an earlier publication, we
also developed a methodology by which real-world driving could be fit by 3 cycles (FTP, HFET,
and US06).7 We repeated this analysis using the VSP frequency distributions shown in Figure
161
-------
II-5 above and additional cycle combinations. This analysis and its results are described in detail
in section III.E.4. Because the Kansas City study measured the driving activity of each vehicle
separately, as well as its second by second fuel consumption, we can develop dynamometer cycle
combinations which best represent the driving of each vehicle. These cycle combinations can
then be used in the 5-cycle formula to predict onroad fuel economy. These vehicle specific 5-
cycle fuel economies can then be compared to the measured onroad fuel economy to assess the
accuracy of the 5-cycle methodology.
One caveat regarding this comparison is that we need estimates of fuel economy over the
five dynamometer cycles to use in the 5-cycle formula. These generally only exist for vehicles
certified to the Supplemental FTP standards, which began phasing in with the 2001 model year.
Very few of the vehicles tested with a PEMS unit on the road in Rounds 1 and 2 of the Kansas
City test program were from the 2001 and later model years. Therefore, we only performed this
analysis for vehicles from Round 1.5. Even with this restriction, we still do not have fuel
economy measurements over the 5 dynamometer cycles for many of the vehicles tested in Round
1.5.
The first step in the analysis was to develop a VSP frequency distribution for each
vehicle's onroad driving. The same methods were used to do this as were used above in
developing the VSP frequency distributions for all the Kansas City vehicles. The only difference
is that it was done for each vehicle individually.
The second step was to develop VSP frequency distributions for the five dynamometer
bags and cycles for each vehicle. These distributions are very similar to those shown in Table
IA-15. The only difference is that we used vehicle-specific values for the constants A, B, C and
M instead of the fleet-average estimates contained in Draft MOVES2004.
The third step was to calculate an average rate of fuel consumption for each vehicle's
operation in each VSP bin. This was done by calculating the rate of fuel consumption for each
second of vehicle operation using the carbon balance method. For each vehicle, the seconds of
operation were grouped by VSP bin and the fuel rates for these times of operation averaged. If
no onroad operation occurred in a particular bin for a particular vehicle, the fuel consumption in
that bin was set to the rate of fuel consumption in the nearest bin having the same level of power
(e.g., bin 38 average fuel consumption is set to that found for bin 28). If no operation occurred
in any of the other bins of the same power level, the rate of fuel consumption was set to the
average rate of fuel consumption for other similar vehicles in that bin. Here, similar means
either conventional or hybrid. In other words, if no operation occurred in bin 18, 28, or 28, then
the rate of fuel consumption in each of these bins for a conventional vehicle was set to the rate of
fuel consumption in each of these bins for all the conventional vehicles which operated in each
bin. The same approach was taken for hybrid vehicles.
We restricted the calculation of fuel rates to warmed up driving (i.e., after the effects of
the cold start had ceased). We first had to determine when an engine start occurred. Along with
this we determined how long the engine had been turned off (i.e., the soak time).
162
-------
We used two criteria for determining the point in each trip when the engine was fully
warmed up. The first applied to those vehicles without coolant temperature data. The second
applied to those vehicles with coolant temperature data. Engine coolant temperature data was
only available for some of the vehicles, so the latter criterion could be not extended to all of the
Kansas City vehicles
For vehicles without coolant temperature data, we assumed that the engine was fully
warmed up after the engine had been running for 200 seconds. For those with such data, we first
smoothed out fluctuations in the coolant temperature data by calculating a five-second average
temperature for a time t which is the average of the coolant temperature at that time, the two
seconds prior to time t and the two seconds subsequent to time t. For all operation beyond the
200th second after a start, we calculated the mean and standard deviation of what we considered
to be the warmed-up coolant temperature. The first second when the five-second average
temperature fell within 1 standard deviation of the mean coolant temperature was estimated to be
the time at which the engine was first fully warmed up.
As described in section I.A.2, we regressed the VSP distributions for the 5 dynamometer
cycles and bags against that of the onroad driving for each vehicle. We weighted the square of
the error by the rate of fuel consumption in each bin, then minimized the weighted error. Unlike
the analyses described in section I.A.2, here we used SAS to perform the analyses due to the
large number of vehicles involved. In all cases, the sum of the dynamometer cycle coefficients
was set to equal 1.0 and the intercept was set to zero.
We developed two sets of dynamometer cycle combinations for each vehicle. One used
the five dynamometer cycles and bags used in section I.A.2, namely: Bags 2 and 3 of the FTP,
HFET, US06 city and US06 highway. A second set of regressions was performed using only
three of these cycles and bags, namely Bags 2 and 3 of the FTP and HFET. This second set of
regressions represents what is called here the "3-cycle" methodology. It represents a way of
estimating onroad fuel economy using only the current two fuel economy tests, the FTP and
HFET. It does not include the US06 cycle, either in whole or in part. No attempt was made to
segregate a vehicle's driving into city or highway modes. Thus, each regression represents all of
the driving by an individual vehicle and the predicted fuel economy is then comparable to the
overall fuel economy of that vehicle on the road during the duration of the PEMS testing.
The initial weighted regression for each vehicle sometimes produced cycle coefficients
which were negative. Those cycles with negative coefficients were removed, one-by-one,
starting with the cycle with the coefficient of the largest magnitude. Once all regression
coefficients were non-negative, the regression procedure was stopped and the results accepted.
No attempt was made to further remove cycles with positive coefficients which might not pass a
statistical significance test of some sort.
For example, vehicle number 521, a 2003 Mitsubishi Montero Sport, was driven 46 miles
while its fuel economy and activity were being monitored. The cycle representation of its
driving is shown in Table A-2.
163
-------
Table A-2. Cycle Combinations for the Mitsubishi Montero Sport
Bag 3 Bag 2 HFET US06 Highway US06 City
Time Basis
3-cycle combination 68.6% 0% 31.4%
5-cycle combination 66.0% 0.2% 0% 33.7% 0%
Mileage Basis
3-cycle combination 53.7% 0% 46.3%
5-cycle combination 45.0% 0.1% 0% 54.9% 0%
Fuel Economy 16.7 19.9 23.4 19.4 11.4
The regression of VSP, weighted by fuel consumption, provides cycle combinations in
terms of the percentage of time spent driving in each cycle. As was described in section I.A.2,
these percentages of time are converted to percentages of miles traveled using the average speed
of each dynamometer cycle and bag.
As can be seen in Table A-2, without consideration of US06, about two-thirds of the
Montero's driving time is represented by Bag 3 and one-third by HFET. When we include US06
in the calculation, the Bag 3 contribution changes very little. The driving previously represented
by HFET shifts entirely to US06 highway.
In order to use the cycle combinations to predict onroad fuel economy, we need estimates
of each vehicle's fuel economy over the five cycles. The vehicles tested in Kansas City were not
tested over the dynamometer cycles. We estimated the fuel economy for each bag or cycle for
each Kansas City vehicle from the test results of similar vehicles in our 5-cycle fuel economy
database of 615 vehicles. The fuel economy estimates for the Montero are shown in the last row
of Table A-2. Warmed up onroad fuel economy can be estimated by simply summing the
product of the fuel economy of each bag and cycle and that bag's or cycle's contribution to 3-
cycle or 5-cycle driving.
We were able to match up 71 of the vehicles tested in Round 1.5 to those in our 5-cycle
fuel economy database. Of these vehicles, 53 were conventional vehicles and 18 were hybrids.-"
Generally, differences exist between the Kansas City and certification vehicles, as the latter are
selected based on their worse case nature regarding emissions and the former are likely to be
high sales volume vehicles. This could cause the 3-cycle and 5-cycle estimates of fuel economy
to somewhat low. We present an analysis at the end of this section which sheds some light on
this issue.
In order to further refine the 3-cycle and 5-cycle estimates, we then included fuel use
related to engine starts. This basically involved determining how many times the engine was
started while the vehicle's operation was being monitored, the soak time prior to each engine
start and the ambient temperature at the time of each start. If a trip lasted less than 10 seconds,
we assumed that it did not occur. The time associated with this trip was made part of the soak
J Nearly half of the hybrids tested were pre-2004 model year Prius vehicles. We do not have 5-cycle fuel economy
values for this model.
164
-------
time prior to the next trip. We also deleted trips when more than 5% of the trip's data were
discarded for reasons described above.
The ambient temperature at the time of the engine start was estimated from the time and
day of the engine start and local meteorology data from the National Weather Service. We
estimated the total fuel related to engine starts using the same methodology applied to the Draft
MOVES2004 estimates of annual engine starts throughout the nation and described in section
LA. 1. Each engine start had a previous soak time and ambient temperature associated with it.
The start fuel use related to this engine start was estimated in terms of the start fuel use related to
a cold start at 75°F using the equations for the effect of start fuel use as a function of soak time
and ambient temperature presented in section I.A.I. The 75°F cold start equivalents for each
start were summed across all of the starts for each vehicle to estimate the total number of 75°F
cold start equivalents for each vehicle's monitored driving. We did not have a measured soak
time prior to the first start of monitored operation, as the equipment was by necessity installed
while the engine was turned off. We assumed that amount of time that the equipment was
operating prior to the first engine start was the soak time prior to that start. This tended to be
only a matter of minutes. Practically, this was equivalent to assuming that the first engine start
was a hot start, though in some rare cases the vehicle had sat overnight since its last operation.
Thus, this is one certain source of uncertainty in the estimation procedure. We estimated the
sensitivity of the estimated onroad fuel economy due to this uncertainty by evaluating the impact
of adding one cold start to the number of equivalent cold starts.
For example, the Montero took three trips while being monitored. The first was preceded
by a soak of 12 minutes, the second by 99 minutes and the third by 3 minutes. The ambient
temperature for all the starts ranged only from 24-25 F. Using the equations presented in section
LA. 1, the total number of cold starts ignoring the effect of ambient temperature was 0.47 for the
three starts. The total number of 75°F cold start equivalents including the effect of ambient
temperature was 0.96 for the three starts. The latter estimate assumes that the cold start fuel use
(the difference in fuel consumption in Bags 1 and 3 multiplied by 3.59 miles) at 20°F is 2.75
times that at 75°F. We determined this ratio for each vehicle using the actual or estimated fuel
economy values for Bags 1 and 3 at 20°F and 75°F from our 5-cycle fuel economy database.
The ratio for the Montero was 2.16, meaning that the cold start fuel use only increased by a
factor of 2.16 between 75°F and 20F, or 22% less than that assumed in Draft MOVES2004 for
the typical vehicle. Thus, we multiplied the equivalent number of cold starts at 75°F (0.96) by
the ratio of 2.16 to 2.75 to make the estimate of the equivalent number of cold starts at 75°F
specific to the cold start fuel use of the Montero. This reduced the estimate of the equivalent
number of cold starts at 75°F for the Montero to 0.75.
We then multiplied the cold start fuel use at 75°F for the Montero by the number of cold
start equivalents. For the 3-cycle estimate of onroad fuel economy, we used a value of 0.47 for
the number of cold start equivalents, since the FTP and LtFET only provide fuel economy
information at 75°F. For the 5-cycle estimate of onroad fuel economy, we used a value of 0.75
for the number of cold start equivalents, since the availability of the cold FTP provides the
additional fuel economy information at 20°F.
165
-------
Finally, for the 5-cycle estimate of onroad fuel economy, we included an estimate of the
effect of temperature on running fuel use. We estimated the ambient temperature at the time of
the start of each trip and then weighted these temperatures by the length of each trip in order to
estimate an average temperature while the vehicle was operating. For the Montero, this was 24.7
F. We then calculated a percentage indicating the degree to which this temperature represented
the temperature drop from 75°F to 20°F; in this case, 91%. Consistent with the 5-cycle
formulae, for the percentage of driven mileage represented by FIFET and US06 highway, we
increased running fuel use by 4% multiplied by the percentage of the temperature drop from
75°F to 20°F. For the percentage of driven mileage represented by Bags 2 and 3 and US06 city,
we calculated running fuel use as the sum of:
1) 100% minus the temperature percentage (9%) times the running fuel use at 75°F (a
function of fuel economy over Bags 2 and 3 and US06 City), plus
2) the temperature percentage (91%) times the running fuel use at 20°F (a function of fuel
economy over Bags 2 and 3, with half of the mileage weight of US06 City added to each Bag).
We were able to match 71 vehicles from those successfully tested in Kansas City to
vehicles in our 5-cycle fuel economy database. Using the procedures just described, we
estimated 3-cycle and 5-cycle fuel economies for each vehicle and averaged the results across all
91 vehicles. The results are shown in Table A-3.
Table A-3. Onroad and Predicted Fuel Economy: Kansas City Test Program
Onroad fuel economy
Predicted Fuel Economy
Warmed up fuel economy
With cold starts at 75°F
With cold starts at ambient
temp.
With running fuel use at
ambient temp
3-Cycle Predicted
Fuel Coefficient
Economy of Vari ati on
(mpg) of Error
28.3
5-Cycle Predicted
Fuel Coefficient of
Economy Variation of
(mpg) Error
28.3
32.2
32.0
18%
18%
29.9
29.7
29.2
27.7
17%
17%
17%
17%
As indicated in Table A-3, the average onroad fuel economy of the 71 vehicles (each vehicle
weighted equally, not by mileage of travel) was 28.3 mpg. Using only the cycle combinations to
predict onroad fuel economy over-predicted onroad fuel economy using both three cycles and
five cycles, as would be expected. The overprediction is smaller for five cycles than three cycles
(1.6 mpg versus 3.9 mpg), indicating the benefit of including the US06 city and highway bags in
predicting onroad fuel economy. Also shown in Table A-5 are the coefficients of variation of the
percentage differences in the predicted versus onroad fuel economy for each of the 71 vehicles.
This metric provides an indication of the consistency of the prediction. A low coefficient of
variation, even if there is a large, but consistent offset, would indicate that a significant factor
was missing from the prediction, like cold start fuel use, but to a consistent degree across all of
the vehicles.
166
-------
Adding cold start fuel use as if all starts were at 75°F (as assumed in the FTP) reduces the
difference between predicted and onroad fuel economy in both cases by the same 0.2 mpg. The
coefficients of variation remain unchanged.
Adding cold start fuel use at the estimated ambient temperature is only applicable to the
5-cycle prediction. This factor reduces the difference between predicted and onroad fuel
economy by 0.5 mpg. The coefficient of variation, however, does not change. One would have
expected it to decrease at least slightly, since the ambient temperatures at which the vehicles
were started varied. Three possible reasons for this outcome exist. One, the effect of these
temperature differences was so small that total fuel use varied very little even for specific
vehicles. Two, the data for soak times were inaccurate, particularly due to the unknown soak
prior to the first start. Or, three, the difference in Bag 1 and Bag 3 fuel economy at 20°F is not a
good predictor of excess fuel use at temperatures which tended to fall in the range of 20 and
75°F. In terms of predicting onroad fuel economy, though, considering the ambient temperature
of engine starts reduces the difference between the 5-cycle fuel economy estimate and onroad
fuel economy by 2%.
Adding the effect of ambient temperature on running fuel use had a larger effect on
predicted fuel economy than the effect of cold starts at 75°F or ambient temperature. This
reduced the predicted 5-cycle from 0.9 mpg above onroad fuel economy to 0.5 mpg below it, for
a final difference of less than 2%.
The difference between the 3-cycle and 5-cycle formulae is even more dramatic for non-
hybrid vehicles. The difference between the best 3-cycle prediction and the onroad measurement
for hybrids averages 24%, while that for the best 5-cycle prediction is only 3%.
The VSP based approach to predicting onroad fuel economy was equally successful when
applied in a 15 car study conducted by EPA in 2001. Table A-4 presents the measured onroad
fuel economy of the 15 cars and fuel economy predictions using two different approaches. One
approach used the VSP methodology developed for Draft MOVES2004 (see section I. A.2) to
determine the mix of FTP, HFET and US06 driving which best matched each vehicle's onroad
driving pattern. This approach under-estimated onroad fuel economy by 4%. The other
approach used average onroad speed to determine the mix of FTP and HFET driving which best
matched each vehicle's onroad driving pattern. This approach over-estimated onroad fuel
economy by 24%. One advantage of the 15-car test program was that most of the 15 cars were
tested over the FTP and US06 as part of the study. Therefore, the dynamometer fuel economy
values used in the predictions were those for the specific vehicles tested onroad. In the Kansas
City program, we had to match the vehicles tested onroad to those in our 5-cycle certification
database. These matches may have been closer in some cases than others, causing the 5-cycle
fuel economy to slightly under-predict onroad fuel economy on average in Kansas City.
167
-------
Table A-4. On-road and Modeled Fuel Economies Using Vehicle-Specific Cycle
Weights (mpg)
Veh.
No.
1
2
3
4
5
6
7
8
11
12
13
14
15
17
18
Fuel Economy (mpg)
On-road
24.8
26.2
27.2
19.5
32.0
24.7
40.9
29.5
26.0
27.9
22.5
15.6
26.8
26.5
17.8
Modeled
VSP (FTP,
HFET, US06)
26.3
23.2
28.7
N/A
N/A
25.1
N/A
28.5
23.8
25.4
22.8
14.5
N/A
N/A
15.5
Speed (FTP,
HFET)
27.4
34.0
31.5
30.5
32.5
28.5
40.0
38.0
34.6
40.7
30.7
19.9
30.9
26.7
18.3
Average (10 vehicles w/ VSP based estimates)
Standard Deviation (10 vehicles)
% Difference
VSP (FTP,
HFET, US06)
6.3%
-11.4%
5.5%
N/A
N/A
1.6%
N/A
-3.3%
-8.6%
-8.9%
1.0%
-7.1%
N/A
N/A
-12.8%
-3.8%
7.0%
Speed (FTP,
HFET)
10.4%
30.0%
15.8%
56.6%
1.7%
15.3%
-2.4%
29.0%
33.1%
46.0%
36.1%
27.4%
15.2%
0.5%
2.5%
24.6%
13.2%
Finally, in an effort to better understand the cause of the difference between predicted
and onroad fuel economy reflected in Table A-4, we reversed this analysis. Instead of using fuel
economy measured over dynamometer cycles to predict onroad fuel economy, we used the
onroad fuel measurements to predict cycle fuel economy. We did this using the VSP
methodology. Each combination of vehicle and dynamometer cycle has its own VSP frequency
distribution. We simply weighted the measured onroad fuel consumption in each VSP bin for
each vehicle by this cycle-specific VSP frequency distribution to estimate the average rate of fuel
use (in gallons per second) over that cycle. We then converted this fuel rate to fuel economy
using the average speed of each dynamometer cycle. Table A-5 shows the measured and
predicted fuel economy values for four dynamometer bags or cycles. When comparing the
predicted fuel economy over the dynamometer cycles to measured values, we found significant
differences between hybrids and conventional vehicles. Thus, Table A-5 presents the results of
this analysis separately for conventional vehicles and hybrids.
168
-------
Table A-5. Comparison of Cycle Fuel Economy
Bag 2 Bag 3 HFET US06
Conventional Vehicles
Predicted from onroad fuel rates (mpg) 16.5 20.7 28.6 24.3
Measured in lab @ 75°F (mpg) 21.5 25.0 34.0 22.5
% Difference 30% 21% 19% -7%
Measured in lab: adjusted for temperature (mpg) 20.5 24.0
% Difference 26% 16%
Hybrids
Predicted from onroad fuel rates (mpg) 32.4 38.4 51.5 43.3
Measured in lab @ 75°F (mpg) 61.8 53.9 61.8 41.6
% Difference 91% 40% 20% -4%
Measured in lab: adjusted for temperature (mpg) 47.6 46.9
% Difference 46% 22%
For conventional vehicles, cycle/bag fuel economy from the 5-cycle certification
database are higher than those predicted from second-by-second onroad fuel rates for Bags 2 and
3 of the FTP and FIFET. The differences are fairly significant, ranging from 19-30%. However,
the situation reverses for US06, with the measured cycle fuel economy being 7% lower than that
predicted from the onroad fuel measurements. These differences can be due to differences in
vehicle operation on the road and on the dynamometer and to differences between the physical
vehicles tested in both cases. As mentioned above, the vehicles which are generally included in
our 5-cycle certification database represent the worse case vehicle configuration within their
broader vehicle groupings. Worse case might include four wheel drive, higher inertia weight
setting, higher TRLHP, etc. The vehicles tested in Kansas City would tend to be high sales
volume models. This might explain the 7% difference seen for the US06 cycle, but the
differences observed with the other cycles go in the wrong direction. Differences in ambient
temperature could explain this difference in fuel economy, as the Kansas City testing was
conducted during the late fall and early winter.
We attempted to correct for the difference in temperature using the average temperature
for each vehicle's operation, as described above. We then used this average temperature to
interpolate between the Bag 2 and Bag 3 fuel economy values measured and 20°F and 75°F,
respectively, to estimate a dynamometer-measured fuel economy at the ambient temperature of
each vehicle's testing in Kansas City. We cannot perform a similar adjustment to the HFET and
US06 fuel economy values as these tests are not run under cold temperature conditions. The
results of this adjustment for temperature are shown just below the unadjusted results in Table A-
5. As can be seen, for conventional vehicles, this adjustment reduces the difference between the
predicted and measured fuel economy values over Bags 2 and 3 from 21-30% to 16-26%, a
modest decrease. As described in section II.A.4, we expect the effect of ambient temperature on
HFET and US06 fuel economy to be relatively small. Thus, there appears to be factors which
are causing onroad fuel consumption to be higher than dynamometer measurements during low
speed and mild driving which is not affecting higher speed, more aggressive driving. Generally,
this argues for including US06 fuel economy in the fuel economy label calculation. It also
supports the inclusion of the 9.5% downward adjustment for non-dynamometer factors.
169
-------
The comparison for hybrids is similar, but the differences are much more dramatic,
particularly at low speed, stop and go driving where hybrid technology functions. Bag 2 and 3
fuel economy measured on the dynamometer exceeds those estimated from onroad
measurements by 40-91%. In contrast, at higher speeds, but with either mild or aggressive
driving styles, the differences are both smaller and very similar to those found with conventional
vehicles. Of course, at higher speeds without much stopping, hybrids operate like conventional
vehicles.
As for conventional vehicles, we adjusted the Bag 2 and 3 dynamometer fuel economy
values for temperature. The effect is much more dramatic for hybrids, given their greater
sensitivity to ambient temperature. The 40-91% difference decreases to 22-46%. These
differences are still greater than those found for conventional vehicles. We considered the
possibility that our analysis of the data was somehow ignoring the engine shut-off feature of the
hybrid vehicles (i.e., the zero fuel consumption occurring during these times were being
excluded from the average fuel rates being calculated for lower speed, low power VSP bins).
However, we confirmed that our measurements included significant amounts of time when the
engine was off. Overall, we measured zero carbon monoxide emissions 12% of the time from
the hybrid vehicles. The Prius models had the highest percentage of engine off operation (19%),
followed by the Civic (3%), following by the one Insight in the test fleet (1%). Clearly, the
Honda hybrids turned off their engines less frequently than the Toyota Prius. This may be due to
differences in hybrid technology utilized by the two manufacturers.
We evaluated whether this difference in engine off time affected the comparisons shown
in Table A-5. For both the Prius and Civic models, the dynamometer measured fuel economy
over Bag 3 is 21% higher than those values estimated from onroad fuel rate measurements, after
adjusting for ambient temperature. For Bag 2, the Prius models show a 63% difference, while
that for the Civic hybrid models is 38%. Thus, the lower percentage of indicated engine off
operation for the Civics is not likely the cause of the greater difference in hybrid fuel economy
over Bags 2 and 3 of the FTP shown in Table A-5 for hybrids compared to conventional
vehicles. The difference could be due to the efficiency of regenerative braking onroad versus on
the dynamometer, as the severity of deceleration is not considered in the VSP methodology.
Further study of the data obtained in Kansas City and additional data collected elsewhere in the
future will be needed to better identify the cause of the difference. Overall, however, for
hybrids, as well as conventional vehicles, the dynamometer measured fuel economy over US06
appears to be much more directly representative of onroad fuel consumption than those measured
over the current fuel economy cycles.
170
-------
Appendix A References
1. Eastern Research Group. Kansas City PM Characterization Study: Round 1 Testing Report.
ERG No. 0133.18.001.001, March 7, 2005.
2. Eastern Research Group. Kansas City Field Testing: Round 2 Testing Final Progress Report.
ERG No. 0133.18.005.003, April 15, 2005.
3. Eastern Research Group. Late Model Vehicle Emissions and Fuel Economy Characterization
Study: Addendum to the Kansas City Exhaust Characterization Study-Draft Report. ERG No.
0133.18.004.001, September 26, 2005.
4. Sierra Research, Inc., "Task Order No. 2 SCF Improvement - Field Data Collection," Sierra
Report No. SR02-07-04, July, 2002.
5. Sierra Research, Inc., "Task Order No. 2 SCF Improvement - Field Data Collection," Sierra
Report No. SR02-07-04, July, 2002.
6. Brzezinski, D., E. Nam, J. Koupal, G. Hoffman. Changes in Real World Driving Behavior:
Analysis of Recent Driving Activity Data. Proceedings of the 15th Coordinating Research
Council On Road Vehicle Emissions Workshop, 2005.
7. R.A. Rykowski, Nam, E.K., Hoffman, G., "On-Road testing and Characterization of Fuel
Economy of Light-duty Vehicles," SAE 2005-01-0677.
171
------- |