Population and Activity of Onroad
Vehicles in MOVES5
£% United States
Environmental Protect
Agency
-------
Population and Activity of Onroad
This technical report does not necessarily represent final EPA decisions
or positions. It is intended to present technical analysis of issues using
data that are currently available. The purpose in the release of such
reports is to facilitate the exchange of technical information and to
inform the public of technical developments.
Vehicles in MOVES5
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
NOTICE
4>EPA
United States
Environmental Protection
Agency
EPA-420-R-24-019
November 2024
-------
Table of Contents
1. Introduction 7
2. MOVES Vehicle and Activity Classifications 11
2.1. HPMS Class 11
2.2. Source Use Types 11
2.3. Regulatory Classes 15
2.4. Fuel and Technology Types 17
2.5. Road Types 18
2.6. Source Classification Codes (SCC) 19
2.7. Model Year Groups 19
2.8. Source Bins 20
2.9. Allowable Vehicle Modeling Combinations 21
2.10. Default Inputs and Fleet and Activity Generators 23
3. Vehicle Miles Traveled by Calendar Year 25
3.1. Historic Vehicle Miles Traveled (1990 and 1999-2022) 25
3.2. Projected Vehicle Miles Traveled (2023-2060) 28
4. Vehicle Populations by Calendar Year 31
4.1. Historic Source Type Populations (1990, 1999-2022) 31
4.2. Projected Vehicle Populations (2023-2060) 35
5. Fuel Type, Regulatory Class, and Engine Technology Distributions by Calendar Year 38
5.1. Sample Vehicle Population 38
5.2. Historic Distributions (MY1950-2022) 40
5.3. Projected Distributions (MY2023-2060) 46
5.4. Summary Results 52
6. Vehicle Age-Related Characteristics 57
6.1. Age Distributions 57
6.2. Relative Mileage Accumulation Rate 65
7. VMT Distribution of Source Type by Road Type 75
8. Average Speed Distributions 77
8.1. Description of Telematics Dataset 77
8.2. Derivation of Default National Average Speed Distributions 79
8.3. Updated Average Speed Distributions 80
9. Driving Schedules and Ramps 82
9.1. Driving Schedules 82
9.2. Modeling of Ramps in MOVES 88
10. Off-Network Idle Activity 89
10.1. Off-Network Idle Calculation Methodology and Definitions 89
10.2. Light-Duty Off-Network Idle 91
10.3. Heavy-Duty Off-Network Idle 98
10.4. Off-network Idling Summary 104
11. Hotelling Activity 106
11.1. Hotelling Activity Distribution 106
11.2. National Default Hotelling Rate 108
12. Engine Start Activity Ill
1
-------
12.1. Light-Duty Start Activity 112
12.2. Heavy-Duty Start Activity 118
12.3. Motorcycle and Motorhome Starts 137
13. Temporal Distributions 140
13.1. VMT Distribution by Month of the Year 141
13.2. VMT Distribution by Type of Day 142
13.3. VMT Distribution by Hour of the Day 143
13.4. Parking Activity 145
13.5. Hourly Hotelling Activity 149
14. Geographical Allocation of Activity 154
14.1. Source Hours Operating Allocation to Zones 154
14.2. Parking Hours Allocation to Zones 155
15. Vehicle Mass and Road Load Coefficients 156
15.1. Source Mass and Fixed Mass Factor 157
15.2. Road Load Coefficients 161
16. Air Conditioning Activity Inputs 166
16.1. ACPenetrationFraction 166
16.2. FunctioningACFraction 167
16.3. Air Conditioning Activity Demand 169
17. Conclusion and Areas for Future Research 171
Appendix A Fuel Type and Regulatory Class Fractions from Previous Versions of MOVES ... 173
Al. Distributions for Model Years 1960-1981 173
A2. Distributions for Model Years 1982-1999 175
Appendix B 1990 Age Distributions 185
Bl. Motorcycles 185
B2. Passenger Cars 185
B3. Trucks 185
B4. Other Buses 186
B5. School Buses and Motor Homes 186
B6. Transit Buses 186
Appendix C Detailed Derivation of Forecast and Backcast Age Distributions 188
CI. Vehicle Survival by Source Type 188
C2. Vehicle Sales by Source Type 191
C3. Historic Age Distributions 193
C4. Projected Age Distributions 195
Appendix D Driving Schedules 198
Appendix E Total Idle Fraction Regression Coefficients 202
Appendix F Source Masses for Light-Duty Vehicles 205
Fl. Motorcycles 206
F2. Passenger Cars 207
F3. Light-Duty Trucks 207
Appendix G NREL Fleet DNA Preprocessing Steps 209
Appendix H Averaging Methods for Heavy-duty Telematics Activity Data 213
HI. Evaluated Methods 213
2
-------
H2. Comparison of Evaluated Methods 216
H3. Future Work 217
Appendix I Road Load Coeffiecient for Combination Trucks in HD GHG Rule 219
Appendix J MOVES5 SourceUseTypePhysics Table 229
Appendix K AVFTTool 233
Kl. Proportional 233
K2. National Average 234
K3. Known Fractions 234
K4. Constant 235
18. References 236
3
-------
List of Acronyms
AEO Annual Energy Outlook
AMPO Association of Metropolitan Planning Organizations
APU auxiliary power unit
ARCADIS design & consultancy firm for natural and built assets
ASD average speed distribution
AVFT Alternate Vehicle Fuel and Technology
AVGCK VIUS broad average weight category
CAN Controller Area Network
CARB California Air Resources Board
CBI confidential business information
CE-CERT College of Engineering - Center for Environmental Research and Technology
CFR Code of Federal Regulations
CNG compressed natural gas
CO carbon monoxide
CO2 carbon dioxide
CRC Coordinating Research Council
DOT U.S. Department of Transportation
EIA U.S. Energy Information Administration
EPA U.S. Environmental Protection Agency
ERG Eastern Research Group, Inc.
E85 gasoline containing 70-85 percent ethanol by volume
FFV flexible fuel vehicle
FHWA Federal Highway Administration
FMCSA Federal Motor Carrier Safety Administration
FTA Federal Transit Administration
GEM Greenhouse Gas Emissions Model
GHG greenhouse gases
g/hr grams per hour
GPO U.S. Government Publishing Office
GPS Global Positioning System
GVWR Gross Vehicle Weight Rating
HC hydrocarbons
HD heavy-duty
HDDBT heavy-duty diesel transit buses
HDV heavy-duty vehicle
HHD heavy-heavy-duty
HHDDT heavy-heavy-duty diesel truck
HPMS Highway Performance Monitoring System
ID identification
IHS Information Handling Services (research consulting firm)
l/M Inspection and Maintenance program
kg/m kilogram per meter
LD light-duty
4
-------
LDT
light-duty truck
LDV
light-duty vehicle
LHD
light-heavy-duty
MAR
mileage accumulation rate
MC
motorcycle
MD
medium-duty
MHD
med i u m-heavy-d uty
M0BILE6
(predecessor to MOVES)
MOVES
MOtor Vehicle Emission Simulator
mph
miles per hour
MPO
Metropolitan Planning Organization
MSA
Metropolitan Statistical Area (U.S. Census)
MSOD
Mobile Source Observation Database
NACFE
North American Council for Freight Efficiency
NCHRP
National Cooperative Highway Research Program
NCTCOG
North Central Texas Council of Government
NEI
National Emission Inventory
NHTS
National Household Travel Survey
NHTSA
National Highway Traffic Safety Administration
NOx
nitrogen oxide
NPMRDS
National Performance Management Research Dataset
NREL
National Renewable Energy Laboratory
NTD
National Transit Database
NVPP
National Vehicle Population Profile
OHIM
Office of Highway Information Management
ONI
off-network idle
OPCLASS
operator classification
ORNL
Oak Ridge National Laboratory
PAMS
portable activity measurement systems
PM
particulate matter
PM2.5
fine particles of particulate matter
PM10
particles of particulate matter 10 micrometers and smaller
RMAR
relative mileage accumulation rate
RPM
revolutions per minute
RT
road type
SBDG
Source Bin Distribution Generator
see
source classification codes
SCR
selective catalytic reduction
SHI
source hours idle
SHO
source hours operating
SIP
state implementation plan
SMOKE
Sparse Matrix Operator Kernel Emissions
SPG
S&P Global
ST
source type
5
-------
STP
scaled-tractive power
SUV
sport utility vehicle
SVP
sample vehicle population
TDM
travel demand model
TEDB
Transportation Energy Data Book
TIF
total idle fraction
TIUS
Truck Inventory and Use Survey
TRB
Transportation Research Board
TRLHP
tractive road load horsepower
TxDOT
Texas Department of Transportation
VIUS
Vehicle Inventory and Use Survey
VMT
vehicle miles traveled
VSP
vehicle specific power
VTRIS
Vehicle Travel Information System
WMATA
Washington Metropolitan Area Transit Authority
6
-------
1. Introduction
The United States Environmental Protection Agency's Motor Vehicle Emission Simulator—
commonly referred to as MOVES—is a set of modeling tools for estimating air pollution
emissions produced by onroad (highway) and nonroad mobile sources. MOVES estimates the
emissions of greenhouse gases (GHGs), criteria pollutants and selected air toxics. The MOVES
model is currently the official model for use for state implementation plan (SIP) submissions to
EPA and for transportation conformity analyses outside of California. The model is also the
primary modeling tool for estimating the impact of mobile source regulations on emission
inventories.
MOVES calculates emission inventories by multiplying emission rates by the appropriate
emission-related activity, applying correction and adjustment factors as needed to simulate
specific situations and then adding up the emissions from all sources and regions.
Vehicle population and activity data are critical inputs for calculating emission inventories from
emissions processes such as running exhaust, start exhaust and evaporative emissions. In
MOVES, most running emissions are distinguished by operating modes, depending on road type
and vehicle speed. Start emissions are determined based on the time a vehicle has been parked
prior to the engine starting, known as a "soak." Evaporative emission modes are affected by
vehicle operation and the time that vehicles are parked. Emission rates are further categorized
by grouping vehicles with similar fuel type, regulatory classification, and other vehicle
characteristics into "source bins."
This report describes the sources and derivation for onroad vehicle population and activity
information and associated adjustments as stored in the MOVES default database. In particular,
this report describes the data used to fill the default database tables listed below in Table 1-1.
Note that technical details on the default database values for emission rates, correction factors
and other inputs, including information on nonroad equipment, are described in other MOVES
technical reports.1
These data have been updated for MOVES5 from previous versions of MOVES. In MOVES5, we
have updated vehicle activity based on newer data from Highway Statistics, Transportation
Energy Data Book, and School Bus Fleet Fact Book. We also updated vehicle distributions based
on newer vehicle registration data from S&P Global (SPG, formerly IHS Markit).
Properly characterizing emissions from onroad vehicles requires a detailed understanding of
the vehicles that comprise the national fleet and their patterns of operation. The MOVES
default database has a domain that encompasses the entire United States, Puerto Rico and the
Virgin Islands. In MOVES, users may analyze emission inventories in 1990 and every year from
1999 to 2060. The national default activity information in MOVES provides a reasonable basis
for estimating national emissions. As described in this report, the most important of these
inputs, such as vehicle miles travelled (VMT) and population estimates, come from long-term
systematic national measurements.
7
-------
Due to the availability of these national measurements, the most recent year of measured data
in the model and the base year for projected emissions is 2022.
It is important to note that uncertainties and variability in the default data contribute to the
uncertainty in the resulting emission estimates. Therefore, MOVES has been specifically
designed to accommodate the input of alternate, user-supplied activity data. In particular,
when modelers estimate emissions for specific geographic locations, EPA guidance
recommends replacing many of the MOVES fleet and activity defaults with local data. This is
especially true for inputs where local data is more detailed or up to date than those provided in
the MOVES defaults. EPA's Technical Guidance2 provides more information on customizing
MOVES with local inputs.
Population and activity data are ever changing as new historical data becomes available and
new projections are generated. As part of the MOVES development process, the model
undergoes major updates and review every few years. The significant updates made to MOVES
since the MOVES2014 release were peer-reviewed under EPA's peer review guidance3 in
separate reviews conducted in 2017, 2019, and 2020. Materials from each peer review,
including peer-review comments and EPA responses are located on the EPA's science inventory
webpage.4 5
The development of fleet and activity inputs will continue to be an important area of focus and
improvement for MOVES.
8
-------
Table 1-1 MOVES database elements covered in this report
Database Table Name
Content Summary
Sections
AvgSpeedDistribution
Distribution of time among average speed bins
Section 8
DayVMTFraction
Distribution of VMT between weekdays and weekend
days
Section 13
DriveSchedule
Average speed of each drive schedule
Section 9
DriveScheduleAssoc
Mapping of which drive schedules are used for each
combination of source type and road type
Section 9
DriveScheduleSecond
Speed for each second of each drive schedule
Section 9
FuelType
Broad fuel categories that indicate the fuel vehicles are
capable of using
Section 2
HotellingActivityDistribution
Distribution of hotelling activity to the various
operating modes
Section 11
HotellingCalendarYear
Rate of hotelling hours per total restricted access VMT
Section 11
HourVMTFraction
Distribution of VMT among hours of the day
Section 13
HPMSVtypeYear
Annual VMT by HPMS vehicle types
Section 3
IdleRegion
Map of idle regions to idle region IDs.
Section 10
ModelYearGroup
A list of years and groups of years corresponding to
vehicles with similar emissions performance
Section 2
MonthGroupHour
Coefficients to calculate air conditioning demand as a
function of heat index
Section 16
MonthVMTFraction
Distribution of annual VMT among months
Section 13
OpModeDistribution
The distribution of engine start soak times for each
source type, day type, hour of the day and pollutant.
Section 12
PollutantProcessModelYear
Assigns model years to appropriate groupings, which
vary by pollutant and process
Section 2
RegulatoryClass
Categorizes vehicles into weight-rating based groups
used to assign emission rates.
Section 2
RoadType
Distinguishes roadways as urban or rural and by type of
access, particularly the use of ramps for entrance and
exit
Section 2
RoadTypeDistribution
Distribution of VMT among road types
Section 7
SampleVehicleDay
Identifies vehicles in the SampleVehicleTrip table
Section 13
SampleVehiclePopulation
Fuel type and regulatory class distributions by source
type and model year.
Section 5
SampleVehicleTrip
Trip start and end times used to determine parking
times for evaporative emission calculations.
Section 13
see
Source Classification Codes that identify the vehicle
type, fuel type, road type and emission process in
MOVES output
Section 2
StartsHourFraction
The fraction of total starts that occur in each hour of
the day. This allocationFraction varies by county
(zonelD) and day type.
Sectionl2
Sta rts M onth Adj u st
The monthAdjustFactor adjusts the starts per day to
reflect monthly variation in the number of starts.
Section 12
9
-------
Table 1-1 MOVES database elements covered in this report
Database Table Name
Content Summary
Sections
StartsPerDay
StartsPerDay value is the number of starts per average
vehicle (of all source types). This value varies by county
(zonelD) and day type.
Section 12
StartsSourceTypeFraction
The allocation of total starts per day for all vehicles to
each of the MOVES source types.
Section 12
SourceBinDistribution
Distribution of population among different vehicle sub-
types (source bins)
Section 2
SourceTypeAge
Rate of survival to subsequent age, relative mileage
accumulation rates and fraction of functional air
conditioning equipment
Appendix C
Section 6
Section 16
SourceTypeAgeDistribution
Distribution of vehicle population among ages
Section 6
SourceTypeHour
The distribution of total daily hotelling among hours of
the day
Section 13
SourceTypeModelYear
Prevalence of air conditioning equipment
Section 16
SourceTypePol Process
Indicates which source bin discriminators are relevant
for each source type and pollutant/process
Section 2
SourceTypeYear
Source type vehicle counts by year
Section 4
SourceUseType
Mapping from HPMS class to source type, including
source type names
Section 2
SourceUseTypePhysics
Road load coefficients and vehicle masses for each
source type used to calculate vehicle specific power
(VSP) and scaled tractive power (STP)
Section 15
TotalldleFraction
Fraction of vehicle operating time when speed is zero.
Section 10
Zone
Allocation of activity to zone (county)
Section 14
ZoneRoadType
Allocation of driving time to zone (county) and road
type
Section 14
10
-------
2. MOVES Vehicle and Activity Classifications
Fundamentally, onroad mobile source emission inventories are estimated by applying vehicle
populations and activity to appropriate emission rates. We wanted to enter vehicle population
and activity data in a form as close as possible to how this data is collected by highway
departments and vehicle registrars, but we had to map these to existing emission standards
and in-use emission rates. Thus, EPA developed MOVES-specific terminology classifying
vehicles according to how they are operated, such as "source types," and to emission-related
characteristics, such as "regulatory classes" and "fuel types." At the most detailed level,
vehicles are classified into "source bins" which have a direct mapping to emission rates by
vehicle operating mode in the MOVES emission rate tables.
This section provides definitions of the various vehicle classifications used in MOVES. The
MOVES terms introduced in this section will be used throughout the report. Later sections
explain how default vehicle populations and activity are assigned and allocated to these
classifications.
2.1. HPMS Class
In this report, MOVES HPMS class refers to one of five categories derived from the US
Department of Transportation (DOT) Highway Performance Monitoring System (HPMS) based
vehicle classes used by the Federal Highway Administration (FHWA) in the Table VM-1 of their
annual Highway Statistics report.6 The five HPMS classes used in MOVES are as follows:
motorcycles (HPMSVTypelD 10), light-duty vehicles (25), buses (40), single unit trucks (50) and
combination trucks (60). Please note that the light-duty vehicles class (25) here represents the
combination of the VM-1 categories for long wheelbase and short wheelbase light-duty cars
and trucks. More details on how HPMS classes are used in MOVES may be found in Section 3.
2.2. Source Use Types
The primary vehicle classification in MOVES is source use type, or, more simply, source type.
Source types are groups of vehicles with similar activity and usage patterns and are more
specific than the HPMS vehicle classes described above. In addition, the road load coefficients
(rolling load, mechanical rotating friction, aerodynamic drag) are defined by source type as
discussed in Section 15.
Vehicles are classified into source types based on body type as well as other characteristics,
such as whether they are registered to an individual, a commercial business, or a transit agency;
whether they have specific travel routines such as a refuse truck; and whether they typically
travel short- or long-haul routes (greater than 200 miles per day). The MOVES source types are
listed in Table 2-1 along with the associated HPMS classes. More detailed source type
definitions are provided below, organized by HPMS class. These source type definitions include
a discussion on relevant regulatory classes; see Section 2.3 for more information regarding
regulatory classes.
11
-------
Table 2-1 Onroad Source Types in MOVES
sourceTypelD
Source Type Name
HPMSVTypelD
HPMS Description
11
Motorcycles
10
Motorcycles
21
Passenger Cars
25
Light-Duty Vehicles
31
Passenger Trucks
(primarily personal use)
25
Light-Duty Vehicles
32
Light Commercial Trucks
(primarily non-personal use)
25
Light-Duty Vehicles
41
Other Buses
(non-school, non-transit)
40
Buses
42
Transit Buses
40
Buses
43
School Buses
40
Buses
51
Refuse Trucks
50
Single Unit Trucks
52
Single Unit Short-Haul Trucks
50
Single Unit Trucks
53
Single Unit Long-Haul Trucks
50
Single Unit Trucks
54
Motor Homes
50
Single Unit Trucks
61
Combination Short-Haul Trucks
60
Combination Trucks
62
Combination Long-Haul Trucks
60
Combination Trucks
2.2.1. Motorcycles
According to the HPMS vehicle description, motorcycles (sourceTypelD 11) are, "all two- or
three-wheeled motorized vehicles, typically with saddle seats and steered by handlebars rather
than a wheel."7 This category usually includes any registered motorcycles, motor scooters,
mopeds, and motor-powered bicycles. Please note that off-road motorcycles are regulated as
nonroad equipment and are not covered in this report.
All motorcycles are categorized in their own regulatory class (regClassID 10).
2.2.2. Passenger Cars
Passenger cars are defined as any coupes, compacts, sedans, or station wagons with the
primary purpose of carrying passengers.7 For consistency with vehicle emission standards, the
category also includes some small crossover vehicles.8
All passenger cars (sourceTypelD 21) are categorized in the light-duty vehicle regulatory class
(regClassID 20).
2.2.3. Light-Duty Trucks
Light-duty trucks include pickups, most sport utility vehicles (SUVs) and vans.7 In MOVES,
passenger trucks (sourceTypelD 31) and light commercial trucks (sourceTypelD 32) are defined
using FHWA's vehicle classification, which specifies that light-duty vehicles are those weighing
less than 10,000 pounds, specifically vehicles with a gross vehicle weight rating (GVWR) in Class
1 and 2 (including 2a and 2b). An exception to this is that Class 2b trucks (8,500 to 10,000 lbs)
12
-------
with two axles or more and at least six tires, colloquially known as "duallies", are classified into
the single unit truck category (see Section 2.2.5).
Light-duty trucks are categorized in the light-duty truck regulatory class (regClassID 30) if they
have a GVWR in Class 1 or 2a. They are categorized in the light heavy-duty (Class 2b and 3)
regulatory class (regClassID 41) if they are Class 2b.
2.2.3.1. Passenger Trucks
A light-duty truck as defined above is considered a passenger truck (sourceTypelD 31) if it is
registered to an individual.
2.2.3.1. Light Commercial Trucks
A light-duty truck as defined above is considered a light commercial truck (sourceTypelD 32) if it
is registered to an organization or business.
2.2.4. Buses
MOVES has three bus source types: transit (sourceTypelD 42), school (sourceTypelD 43), and
other (sourceTypelD 41).
2.2.4.1. Transit Buses
Transit buses in MOVES are defined as any active vehicle with a bus body type ("bus",
"articulated bus", "over-the-road bus", "double decked bus" and "cutaway") that must be
reported to Federal Transit Administration's (FTA) National Transit Database (NTD). According
to the FTA, these are buses owned by a public transit organization for the primary purpose of
transporting passengers on fixed routes and schedules.9
Transit buses are categorized in the light heavy-duty (GVWR Class 4 and 5) regulatory class
(regClassID 42) or the medium heavy-duty (Class 6 and 7) regulatory class (regClassID 46) if their
GVWR falls within Class 4 to 7. Heavy heavy-duty buses in Class 8 are assigned to a regulatory
class based on fuel type:
• Gasoline Class 8 buses are categorized in the heavy heavy-duty (Class 8) regulatory class
(regClassID 47).
• Diesel, CNG, and electric Class 8 buses are categorized in the urban bus regulatory class
(regClassID 48).
2.2.4.2. School Buses
School buses in MOVES are defined as according to FHWA: vehicles designed to carry more than
ten passengers and are used to transport K-12 students between their home and school.10
School buses can be categorized in any of the following regulatory classes: regClassID 41, 42,
46, or 47.
13
-------
2.2.4.3. Other Buses
Any other buses that do not fit into the transit or school bus categories are modeled in MOVES
as "other" buses.3 For example, these may include intercity buses not owned by transit
agencies. Please note that these definitions allow similar vehicle types to be modeled in both
the transit and other bus source types. For example, a shuttle bus operated by a transit agency
would be modeled as a transit bus, but an airport shuttle bus operated by a private company
would be modeled as an "other" bus. Due to the similarities between these source types, they
have identical fuel type and regulatory class distributions and mileage accumulation rates.
However, they have different age distributions and driving schedules as described in
subsequent sections.
Other buses can be categorized in any of the following regulatory classes: regClassID 42, 46, or
47.
2.2.5. Single Unit Trucks
Single unit source types in MOVES include refuse trucks (sourceTypelD 51), single unit short-
haul trucks (sourceTypelD 52), single unit long-haul trucks (sourceTypelD 53) and motor homes
(sourceTypelD 54). FHWA's vehicle classification specifies that single unit trucks are single-
frame trucks with a gross vehicle weight rating of greater than 10,000 pounds or with two axles
and at least six tires—colloquially known as "dualies."
All single unit trucks can be categorized in any of the following regulatory classes: regClassID 41,
42,46, or 47.
2.2.5.1. Refuse Trucks
Refuse trucks are single unit trucks that are specifically designed for garbage, refuse, or solid
waste disposal operations. They are typically equipped with a self-compactor or an industrial
roll-off hoist and roll-off container.
2.2.5.2. Single Unit Short-haul Trucks
The single unit short-haul truck vehicle category is quite diverse, including (but not limited to)
delivery, box, flatbed, cement mixers, and tow trucks, among other body styles and vocations.
This source type includes all single unit trucks that are not refuse trucks or motor homes, which
travel less than or equal to 200 miles a day.
a Note, in MOVES2014, "other" buses were called "intercity" buses and defined differently.
14
-------
2.2.5.3. Single Unit Long-haul Trucks
Single unit long-haul trucks are defined similarly as single unit long-haul trucks, except they
travel more than 200 miles a day.
2.2.5.4. Motor Homes
Motor homes are self-propelled recreational vehicles. In MOVES, sourceTypelD 54 includes
Class A, B, and C motor homes.
2.2.6. Combination Trucks
Combination trucks in MOVES include two source types: short-haul (sourceTypelD 61) and long-
haul combination trucks (sourceTypelD 62). These are heavy-duty trucks that are not single-
frame. Instead, they consist of a tractor and one or more trailers.
Combination trucks are categorized in the medium heavy-duty (Class 6 and 7) regulatory class
(regClassID 46) or the heavy heavy-duty (Class 8) regulatory class (regClassID 47). An exception
to this categorization is for glider vehicles, which are assigned to regClassID 49. See Section 2.3
for more information about glider vehicles.
2.2.6.1. Combination Short-haul Trucks
Short-haul combination trucks travel less than or equal to 200 miles a day. Frequently, these
trucks are older than long-haul combination trucks and they are often purchased in secondary
markets, such as for drayage applications, after being used primarily for long-haul trips.11
2.2.6.1. Combination Long-haul Trucks
Long-haul combination trucks travel more than 200 miles a day. These trucks frequently have a
sleeper cab, allowing the driver to "hotel" in their truck during mandated rest times. See
Section 11 for more information regarding hotelling activity.
2.3. Regulatory Classes
In contrast to source types, regulatory classes are used to group vehicles subject to similar
emission standards. The EPA regulates vehicle emissions based on groupings of technologies
and classifications that do not necessarily correspond to DOT activity and usage patterns. To
properly estimate emissions, it is critical for MOVES to account for these emission standards.
The regulatory classes used in MOVES are summarized in Table 2-2 below. The "doesn't matter"
regulatory class is used internally in the model if the emission rates for a given pollutant and
process are independent of regulatory class. The motorcycle (MC) and light-duty vehicle (LDV)
regulatory classes have a one-to-one correspondence with source type. Other source types are
allocated between regulatory classes based primarily on gross vehicle weight rating (GVWR)
classification, which is a set of eight classes defined by FHWA based on the manufacturer-
defined maximum combined weight of the vehicle and its load. "Urban buses" are distinguished
from other heavy heavy-duty vehicles (GVWR >33,000 lbs.) because they have tighter PM
emission standards for the 1994 through 2006 model years.12 Urban bus is a regulatory class
15
-------
that is defined by its intended use as well as the GVWR. EPA regulations define urban buses as
"heavy heavy-duty diesel-powered passenger-carrying vehicles with a load capacity of fifteen or
more passengers and intended primarily for intra-city operation."13 In MOVES, we also assign
all CNG transit buses and some electric transit buses to the Urban Bus regulatory class.
Table 2-2 Regulatory Classes in MOVES
regClassID
Regulatory
Class Name
Description
0
Doesn't Matter
Doesn't Matter
10
MC
Motorcycles
20
LDV
Light-Duty Vehicles
30
LDT
Light-Duty Trucks
41
LHD2b3
Class 2b and 3 Trucks (8,500 lbs < GVWR <= 14,000 lbs)*
42
LHD45
Class 4 and 5 Trucks (14,000 lbs. < GVWR <= 19,500 lbs.)*
46
MHD
Class 6 and 7 Trucks (19,500 lbs. < GVWR < =33,000 lbs.)
47
HHD
Class 8a and 8b Trucks (GVWR > 33,000 lbs.)
48
Urban Bus
Urban Buses
49
Gliders
Gliders14
• Class 2b trucks (GVWR 8,501-10,000 lbs) are only modeled in passenger trucks and light
commercial trucks (source types 31 and 32, respectively). Class 3 trucks (GVWR 10,001-14,000
lbs) are only modeled in heavy-duty source types (buses and single unit trucks). Model year
2017-and-later engine-certified Class 3 trucks (only present within source types 52, 53, and 54)
are classified as LHD45 (regclassID 42) for modeling purposes.
MOVES distinguishes between light-duty and heavy-duty vehicles by GVWR.
• Vehicles with a GVWR at or below 10,000 lbs are considered light-duty and are modeled
as one of the light-duty source types: passenger cars, passenger trucks, or light
commercial trucks.
o Passenger cars are assigned regulatory class LDV (regClassID 20)
o Passenger trucks and light commercial trucks have two relevant regulatory
classes, depending on the truck's GVWR. Trucks with a GVWR at or below 8,500
lbs (Class 1 and 2a) are assigned regulatory class LDT (regClassID 30) Trucks with
a GVWR between 8,500-10,000 lbs (Class 2b) are assigned regulatory class
LHD2b3 (regClassID 41).
• Vehicles with a GVWR greater than 10,000 lbs are considered heavy-duty and are
modeled as any of the heavy-duty source types falling in the bus, single unit truck, or
combination truck categories. The relevant regulatory classes for these vehicles are
LHD2b3 (regClassID 41), LHD45 (42), MHD (46), HHD (47), Urban Bus (48), and Gliders
(47).
Note that vehicles commonly referred to as "medium-duty" (ranging from 8,500-14,000 lbs,
covering the LHD2b3 regulatory class) are modeled as either light-duty or heavy-duty vehicles in
MOVES, depending on if their GVWR is above or below 10,000 lbs.
16
-------
There is an exception to the assignment of the heavy-duty regulatory classes strictly based on
GVWR. Some Class 3 trucks may be engine-certified or chasis-certified. Because Class 3 engine-
certified vehicles are subject the same emission standards as Class 4 and 5 engine-certified
vehicles, these vehicles are classified as LHD45 vehicles instead of LHD2b3. This exception is
applied to model year 2017 and later because these are the only model years where the
emission rates are different between LHD2b3 and LHD45.18
In the MOVES model, "Gliders" refers to post-2007 heavy-duty diesel vehicles with new chassis
but with older engines that do not meet 2007 or 2010 emissions standards and thus are treated
as a separate regulatory class.
Section 5.1 provides more information on the distribution of vehicles among regulatory classes.
Vehicle weights in MOVES are defined by both regulatory class and source type as discussed in
Section 15.
2.4. Fuel and Technology Types
MOVES models vehicles powered by the following fuel types: gasoline, diesel, E-85 (fuels
containing 70 percent to 85 percent ethanol by volume), compressed natural gas (CNG) and
electricity. Note that in some cases, a single vehicle can use more than one fuel. For example,
flexible fuel vehicles (FFV) are capable of running on either gasoline or E-85. In MOVES, fuel
type refers to the capability of the vehicle rather than the fuel in the tank. The fuel use actually
modeled depends on a number of factors including the location, year and month in which the
fuel was purchased, as explained in the MOVES technical report on fuel supply.15 MOVES also
allows the modeling of engine technology types, although these are not distinguished in MOVES
output. Specifically, engine technology is used to distinguish battery and fuel-cell electric
vehicles. Table 2-3 below summarizes the fuel types and technology types populated in MOVES.
These are recorded in the default database FuelType, EngineTech and FuelEngTechAssoc tables.
Table 2-3 A List of Allowable Fuel Types to Power Vehicles in MOVES
fuelTypelD
Description
Default Fuel
FormulationID15
EngTechID
Technology Description
1
Gasoline
10
1
Conventional Internal
Combustion
2
Diesel Fuel
20
1
Conventional Internal
Combustion
3
Compressed
Natural Gas
(CNG)
30
1
Conventional Internal
Combustion
5
Ethanol (E-85)
50
1
Conventional Internal
Combustion
9
Electricity
90
30
Electric
40
Fuel Cell
17
-------
It is important to note that not all fuel type/source type combinations can be modeled in
MOVES. Specifically:
• For motorcycles, MOVES can model only the gasoline fuel type.
• MOVES cannot model gasoline-fueled long-haul combination trucks.
• Flexible fueled vehicles (i.e., E85-compatible) are only modeled for passenger cars,
passenger trucks, and light commercial trucks.
• CNG and fuel cell electric vehicles are only modeled in heavy-duty source types.
In addition, MOVES does not explicitly model hybrid powertrains, but accounts for these
vehicles in calculating fleet-average energy consumption and CO2 rates.b For more information
on how MOVES models the impact of fuels on emissions, please see the MOVES documentation
on fuel effects.16
2.5. Road Types
MOVES calculates onroad emissions separately for each of four road types and for "off-
network" activity when the vehicle is not moving. The road types used in MOVES are listed in
Table 2-4. The four MOVES road types (2-5) are aggregations of FHWA functional facility types.
Table 2-4 Road Types in MOVES
roadTypelD
Description
FHWA Functional Types
1
Off Network
Off Network
2
Rural Restricted Access
Rural Interstate & Rural Freeway/Expressway
3
Rural Unrestricted Access
Rural Other Principal Arterial, Minor Arterial,
Major Collector, Minor Collector & Local
4
Urban Restricted Access
Urban Interstate & Urban Freeway/Expressway
5
Urban Unrestricted Access
Urban Other Principal Arterial, Minor Arterial,
Major Collector, Minor Collector & Local
The MOVES road types are based on two important distinctions in how FWHA classifies roads:
1) urban versus rural roadways are distinguished based on surrounding land use and human
population density and 2) unrestricted versus restricted are distinguished based on roadway
access—restricted roads require the use of ramps. The urban/rural distinction is used primarily
for national level calculations. It allows different default speed distributions in urban and rural
settings. Of course, finer distinctions are possible. Users with more detailed information on
b While we have considered creating a separate category for hybrid vehicles, modeling their emissions separately is
not required for regulatory purposes and presents a number of challenges, including obtaining representative
detailed data on hybrid vehicle emissions and usage and accounting for offsetting emissions allowed under the
fleet-averaging provisions of the relevant emissions standards.
18
-------
speeds and acceleration patterns may run MOVES at project level where emissions can be
calculated for individual links.
2.6. Source Classification Codes fSCC)
Source Classification Codes (SCC) are used to group and identify emission sources in large-scale
emission inventories. They are often used when post-processing MOVES output to further
allocate emissions temporally and spatially when preparing inputs for air quality modeling. In
MOVES, SCCs are numerical codes that identify the vehicle type, fuel type, road type and
emission process using MOVES identification (ID) values in the following form:
AAAFVVRRPP
Where:
• AAA indicates mobile source (this has a value of 220 for both onroad and nonroad),
• F indicates the MOVES fuelTypelD value,
• VV indicates the MOVES sourceTypelD value,
• RR indicates the MOVES roadTypelD value, and
• PP indicates the MOVES emission processID value.
Building the SCC values in this way allows additional source types, fuel types, road types and
emission processes to be easily added to the list of SCCs as changes are made to future versions
of MOVES. The explicit coding of fuel type, source type, road type and emission process also
allows the new SCCs to indicate aggregations. For example, a zero code (00) for any of the
sourceTypelD, fuelTypelD, roadTypelD and processID strings that make up the SCC indicates
that the reported emissions are an aggregation of all categories of that type. Using the mapping
described above, modelers can also easily identify the sourceTypelD, fuelTypelD, roadTypelD
and processID of emissions reported by SCC. Refer to earlier sections in this document for the
descriptions of the sourceTypelD, fuelTypelD and roadTypelD values currently used by MOVES.
Emission processes are discussed in other MOVES reports on emission rate development1718
and are not described here. All feasible SCC values are listed in the SCC table within the default
database.
2.7. Model Year Groups
MOVES uses model year groups to avoid unnecessary duplication of emission rates for vehicles
with similar technology and similar expected emission performance. For example, there is a
model year group for "1980 and earlier." In MOVES, model year refers to the year in which the
vehicle was produced, built and certified as compliant with emission standards.
The default ModelYearGroup table provides information on the model year group names,
beginning and ending years and a two-digit shorthand identifier (shortModelYrGroupID).
However, the model year groups that are relevant for a given calculation can vary depending on
pollutant and emission process as defined in the PollutantProcessModelYear table. Because
these groupings are determined based on analysis of the actual or expected emissions
19
-------
performance, the rationale for each model year grouping is provided in the MOVES emission
rate reports.1718
2.8. Source Bins
The MOVES default database identifies emission rates by emission-related characteristics such
as the type of fuel that a vehicle uses and the emission standards it is subject to. These
classifications are called "source bins." They are named with a sourceBinID that is a unique 19-
digit identifier in the following form:
1FFEERRMM0000000000
Where:
• 1 is a placeholder,
• FF is a MOVES fuelTypelD,
• EE is a MOVES engTechID,
• RR is a MOVES regClassID,
• MM is a MOVES shortModYrGroupID, and
• 10 trailing zeros for future characteristics.
The model allocates vehicle activity and population to these source bins as described below.
A mapping of model year to model year groups is stored in the PollutantProcessModelYear
table. Distributions of fuel type and regulatory class by source type are stored by model year in
the SampleVehiclePopulation table. MOVES combines information from these two tables (see
Table 2-5) to create a detailed SourceBinDistribution. In general, fuel type is relevant for all
emission calculations, but the relevance of regulatory class and model year group depend on
the pollutant and process being modeled. See Section 2.10 for more information on how
MOVES uses generators to calculate detailed activity information.
20
-------
Table 2-5 Data Tables Used to Allocate Source Type to Source Bin
Table Name
Key Fields*
Additional Fields
Notes
SourceTypePol Process
sourceTypelD
polProcessID
isRegClassReqd
isMYGroupReqd
Indicates which pollutant-processes
the source bin distributions may be
applied to and indicates which
discriminators are relevant for each
sourceTypelD and polProcessID
(pollutant/process combination)
PollutantProcessModelYear
polProcessID
modelYearlD
modelYearGroupID
Assigns model years to appropriate
model year groups for each
polProcessID.
SampleVehiclePopulation
sourceTypelD
modelYearlD
fuelTypelD
engTechID
regClassID
stmyFuelEngFraction
stmyFraction
Includes fuel type and regulatory class
fractions for each source type and
model year, even for some source
type/fuel type combinations that do
not currently have any appreciable
market share (i.e. CNG motor homes).
This table provides default fractions for
the Alternative Vehicle Fuel &
Technology (AVFT) importer.
* In these tables, the sourceTypelD and modelYearlD are combined into a single sourceTypeModelYearlD.
While the relevance of model year groups and regulatory classes to specific pollutant processes
is detailed in other MOVES tehcnical reports, the SampleVehiclePopulation (SVP) table is a topic
for this report and is discussed in Section 5.1
2.3. Allowable Vehicle Modeling Combinations
In theory, the MOVES source bins would allow users to model any combination of source type,
model year, regulatory class and fuel type. However, each combination must have
accompanying emission rates; combinations that lack data from emissions testing or have
negligible market share cannot be directly modeled in MOVES.
Table 2-6 summarizes the allowable source type-fuel type combinations. Most of the gasoline
and diesel combinations exist with a few exceptions, but options for alternative fuels are
limited, as discussed earlier in Section 2.4. MOVES also stores regulatory class distributions by
source type in the SampleVehiclePopulation table. Table 2-7 summarizes the allowable source
type-regulatory class combinations. Table 2-8 shows the full set of allowable source type, fuel
type and regulatory class combinations. Also see the mapping of fuel types and technology
types in Table 2-3. Additional discussion about decisions to include and exclude certain types of
vehicles can be found in Section 5.
21
-------
Table 2-6 Matrix of the Allowable Source Type-Fuel Type Combinations in MOVES5
(Allowable combinations are marked with an X)
Source Use Type
Motorcycles
Passenger Cars
Passenger Trucks
Light Commercial Trucks
Other Buses
Transit Buses
School Buses
Refuse Trucks
Short-Haul Single Unit
Trucks
Long-Haul Single Unit
Trucks
Motor Homes
Short-Haul Combination
Trucks
Long-Haul Combination
Trucks
Fuel Type
11
21
31
32
41
42
43
51
52
53
54
61
62
Gasoline
1
X
X
X
X
X
X
X
X
X
X
X
X
Diesel
2
X
X
X
X
X
X
X
X
X
X
X
X
CNG
3
X
X
X
X
X
X
X
X
X
E85-Capable
5
X
X
X
Battery
Electric
9
X
X
X
X
X
X
X
X
X
X
X
X
Electric Fuel
Cell
9
X
X
X
X
X
X
X
X
X
Table 2-7 Matrix of the Allowable Source Type-Regulatory Class Combinations in MOVES5
(Allowable combinations are marked with an X)
Source Use Type
Motorcycles
Passenger Cars
Passenger Trucks
Light Commercial
Trucks
Other Buses
Transit Buses
School Buses
Refuse Trucks
Short-Haul Single
Unit Trucks
Long-Haul Single Unit
Trucks
Motor Homes
Short-Haul
Combination Trucks
Long-Haul
Combination Trucks
Regulatory Class
11
21
31
32
41
42
43
51
52
53
54
61
62
MC
10
X
LDV
20
X
LDT
30
X
X
LHD2b3
41
X
X
X
X
X
X
X
LHD45
42
X
X
X
X
X
X
X
MHD67
46
X
X
X
X
X
X
X
X
X
HHD8
47
X
X
X
X
X
X
X
X
X
Urban Bus
48
X
Gliders
49
X
X
22
-------
Table 2-8 Summary of source type, fuel type, technology, and regulatory class combinations
in MOVES5
sourceTypelD
fuelTypelD
engTechID
regClassID
11
1
1
10
21
1, 2,5
1
20
9
30
20
31, 32
1,2
1
30,41
5
1
30
9
30
30,41
9
40
41
41, 42
1, 2,
1
42, 46, 47
3
1
47, 48
9
30, 40
42, 46, 47
43
1,2
1
41, 42, 46, 47
3
1
47
9
30, 40
41, 42, 46, 47
51, 52, 53, 54
1,2
1
41, 42, 46, 47
3
1
47
9
30, 40
41, 42, 26, 47
61, 62
1,9
1
46, 47
2
1
46, 47, 49
3
1
47
9
30, 40
46,47
2.10. Default Inputs and Fleet and Activity Generators
As explained in the introduction, vehicle population and activity data are critical inputs for
calculating emission inventories and MOVES calculators require information on vehicle
population and activity at a very fine scale. In project-level modeling, this detailed information
may be available and manageable. However, in other cases, the fleet and activity data used in
the MOVES calculators must be generated from inputs in a condensed or more readily available
format. MOVES uses "generators" to create fine-scale information from user inputs and MOVES
defaults.
The MOVES Total Activity Generator estimates hours of vehicle activity using vehicle miles
traveled (VMT) and speed information to transform VMT into source hours operating (SHO).
Other types of vehicle activity are generated by applying appropriate factors to vehicle
populations. Vehicle starts, extended idle hours and source hours (including hours operating
and not-operating) are also generated. The default database for MOVES contains national
estimates for VMT and vehicle population for every possible analysis year (1990 and 1999-
2060). For national inventory runs, annual national activity is distributed temporally and
spatially using allocation factors and age distributions (see Section 6.1) are applied to allocate
activity by model year.
The Source Bin Distribution Generator (SBDG) uses information on model year groupings and
fuel type and regulatory class distributions to estimate activity fractions of each source bin as a
23
-------
function of source type, model year, pollutant and process. MOVES maps the activity data (by
source types) to source bins which map directly to the MOVES emission rates.
There are a number of MOVES modules that generate operating mode distributions based on
vehicle activity inputs. For running, brake wear, and tirewear emissions, and to calculate off-
network idling activity, MOVES uses information on speed distributions and driving patterns
(driving schedules) to develop operating mode fractions for each source type, road type, and
time of day. Similarly, other generators use other MOVES inputs to develop operating mode
distributions for hotelling activity, starts and vapor venting.
24
-------
3. Vehicle Miles Traveled by Calendar Year
At the national level, MOVES calculates source operating hours from national vehicle miles
traveled (VMT) by vehicle type. The default database contains national VMT estimates for all
analysis years, which include 1990 and 1999-2060. Years 1991-1998 are excluded because there
is no regulatory requirement to analyze them and including them would increase model
complexity. Calendar year 1990 is available to be modeled in MOVES because of the Clean Air
Act Amendments of 1990.
The national VMT estimates are stored in the HPMSVTypeYear table,c which includes three data
fields: HPMSBaseYearVMT (discussed below), baseYearOffNetVMT and VMTGrowthFactor. Off
network VMT refers to the portion of activity that is not included in travel demand model
networks or any VMT that is not otherwise reflected in the other four road types. The field
baseYearOffNetVMT is provided in case it is useful for modeling local areas. However, the
reported HPMS VMT values, used to calculate the national averages discussed here, are
intended to include all VMT. Thus, for MOVES national defaults, the baseYearOffNetVMT is zero
for all vehicle types. Additionally, the VMTGrowthFactor field is not used in MOVES and is set to
zero for all vehicle types.
3.1. Historic Vehicle Miles Traveled (1990 and 1999-2022)
In MOVES5, VMT estimates for the historic years 1990 and 1999-2022 come from the VM-1
table of US DOT Federal Highway Administration's (FHWA) Highway Statistics series.6 In
reporting years 2007 and later, the VM-1 data are calculated with an updated methodology,19
which implements state-reported data directly rather than a modeled approach and which has
different vehicle categories. The current HPMS-based VM-1 categories are 1) light-duty short
wheelbase, 2) light-duty long wheelbase, 3) motorcycles, 4) buses, 5) single unit trucks and 6)
combination trucks. Because MOVES categorizes light-duty source types based on vehicle type
and not wheelbase length, the short and long wheelbase categories are combined into a single
category of light-duty vehicles (HPMSVTypelD 25). Internally, the MOVES Total Activity
Generator allocates this VMT to MOVES source types and ages using vehicle populations, age
distributions, and relative mileage accumulation rates.
For years prior to 2007, the VM-1 data with historical vehicle type groupings are inconsistent
with the current VM-1 vehicle categories used in MOVES and cannot be used as they are
currently reported. However, in early 2011, FHWA released revised VMT data for years 2000-
2006 to match the new category definitions. Shortly afterward, the agency replaced these
revised numbers with the previously published VMT data stating, "[FHWA] determined that it is
c In MOVES, users can enter VMT estimates using four different input methods: annual miles by HPMS class, annual
miles by source type, annual average daily miles by HPMS class and annual average daily miles by source type.
However, the national defaults are stored as annual miles by HPMS class and any discussion in this report on
annual VMT estimates will be in this context.
25
-------
more reliable to retain the original 2000-2006 estimates because the information available for
those years does not fully meet the requirements of the new methodology."d However,
needing continuity of the VM-1 vehicle categories, we used these FHWA-revised values by the
new categories as the VMT for 2000-2006.
This left two years, 1990 and 1999, that needed to be adjusted to be consistent with the new
HPMS vehicle categories. Since the methodology that FHWA used to revise the 2000-2006 data
is undocumented, we adjusted 1990 and 1999 using the average ratio of the change for each
vehicle category. This was found by dividing the FHWA-adjusted VMT for each vehicle category
by the original VMT for each year 2000-2006 and then calculating the average ratio for each
category. This ratio was then applied to the corresponding VMT values reported in VM-1 for
1990 and 1999. Since FHWA's adjustments conserved the original total VMT estimates, we
normalized our adjusted values such that the original total VMT for the years were unchanged.
The resulting values for historic years by HPMS vehicle class are listed in Table 3-1. The VMT for
1990 and 1999 were EPA-adjusted from VM-1, 2000-2006 were FHWA-revised and 2007 and
later were unadjusted, other than the simple combination of the short and long wheelbase
classes into light-duty vehicles. In addition to these adjustments, for some years, the VMT
values were revised by FHWA in subsequent publications. Table 3-2 summarizes the data
source and revision date we used for each historical year.
d This text appears in a footnote to FHWA's Highway Statistics Table VM-1 for publication years 2000-2009.
26
-------
Table 3-1 Historic VMT by HPMS vehicle class for calendar years 1990 and 1999-2022 (millions
o1
: miles)
Year
Motorcycles
Light-Duty
Vehicles
Buses
Single Unit
Trucks
Combination
Trucks
1990
11,404
1,943,194
10,279
70,861
108,624
1999
13,619
2,401,408
14,853
100,534
160,921
2000
12,175
2,458,221
14,805
100,486
161,238
2001
11,120
2,499,069
12,982
103,470
168,969
2002
11,171
2,555,468
13,336
107,317
168,217
2003
11,384
2,579,195
13,381
112,723
173,539
2004
14,975
2,652,092
13,523
111,238
172,960
2005
13,773
2,677,641
13,153
109,735
175,128
2006
19,157
2,680,537
14,038
123,318
177,321
2007
21,396
2,691,034
14,516
119,979
184,199
2008
20,811
2,630,213
14,823
126,855
183,826
2009
20,822
2,633,248
14,387
120,207
168,100
2010
18,513
2,648,456
13,770
110,738
175,789
2011
18,542
2,650,458
13,807
103,803
163,791
2012
21,385
2,664,060
14,781
105,605
163,602
2013
20,366
2,677,730
15,167
106,582
168,436
2014
19,970
2,710,556
15,999
109,301
169,830
2015
19,606
2,779,693
16,230
109,597
170,246
2016
20,445
2,849,718
16,350
113,338
174,557
2017
20,149
2,877,378
17,227
116,102
181,490
2018
20,076
2,897,083
18,303
120,699
184,165
2019
19,688
2,924,053
17,980
124,746
175,305
2020
17,947
2,572,988
15,037
117,832
179,817
2021
19,642
2,768,999
16,744
131,637
195,389
2022
23,765
2,822,664
18,490
136,224
195,049
27
-------
Table 3-2 Highway Statistics publications used for historical years
Year
FHWA Publication Source (Publication/Revision Date)
1990
Highway Statistics 1991 (October 1992)
1999
Highway Statistics 1999 (October 2000)
2000
Highway Statistics 2000 (Apr
1 2011)
2001
Highway Statistics 2001 (Apr
1 2011)
2002
Highway Statistics 2002 (Apr
1 2011)
2003
Highway Statistics 2003 (Apr
1 2011)
2004
Highway Statistics 2004 (Apr
1 2011)
2005
Highway Statistics 2005 (Apr
1 2011)
2006
Highway Statistics 2006 (Apr
1 2011)
2007
Highway Statistics 2007 (Apr
1 2011)
2008
Highway Statistics 2008 (Apr
1 2011)
2009
Highway Statistics 2010 (December 2012)
2010
Highway Statistics 2010 (December 2012)
2011
Highway Statistics 2012 (January 2014)
2012
Highway Statistics 2013 (January 2015)
2013
Highway Statistics 2014 (December 2015)
2014
Highway Statistics 2014 (December 2015)
2015
Highway Statistics 2015 (January 2017)
2016
Highway Statistics 2016 (May 2018)
2017
Highway Statistics 2017 (March 2019)
2018
Highway Statistics 2018 (March 2020)
2019
Highway Statistics 2019 (November 2020)
2020
Highway Statistics 2021 (February 2023)
2021
Highway Statistics 2022 (November 2023)
2022
Highway Statistics 2022 (November 2023)
3.2. Projected Vehicle Miles Traveled (2023-2060)
The Annual Energy Outlook (AEO)20 describes the future energy consumption forecasted by
Department of Energy. Vehicle sales and miles traveled are included in the projections because
they strongly influence fuel consumption. In MOVES5, VMT for years beyond 2022 are based on
the reference case VMT projections from AEO2023. Because AEO vehicle categories are
different from HPMS classes, the AEO projections were not used directly. Instead, year-to-year
percent changes in the projected values were calculated and applied to the 2022 base year
HPMS data. Since AEO2023 only projects out to 2050, VMT for years 2051-2060 were assumed
to continue to grow at the same growth rate as between 2049 and 2050.
Table 3-3 shows the mappings between AEO VMT categories and HPMS categories. Where
multiple AEO categories are listed, their VMT were summed before calculating the year-over-
year growth rates. AEO's light-duty category was mapped to both the combined HPMS light-
duty and the motorcycle categories. Motorcycles were included here because they were not
28
-------
explicitly accounted for elsewhere in AEO. Since buses span a large range of heavy-duty vehicles
and activity, the total heavy-duty category was mapped to the HPMS bus category. AEO's light-
medium- and medium-heavy-duty categories were combined for mapping to the HPMS single
unit truck category and AEO's heavy-heavy-duty category was mapped to the HPMS
combination truck category. We acknowledge that using VMT growth estimates from different
vehicle types as surrogates for motorcycles and buses, in particular, will introduce additional
uncertainty into these projections.
Table 3-3 Mapping AEO categories to HPMS classes for projecting VMT
AEO VMT Category Groupings
HPMS Class
Total Light-Duty VMT1
+
Total Commercial Light Truck VMT"
10 - Motorcycles
25 - Light Duty Vehicles
Total Heavy-Duty VMTNi
40 - Buses
Light-Medium Subtotal VMT1"
+
Medium Subtotal VMT1"
50 - Single Unit Trucks
Heavy Subtotal VMT1"
60 - Combination Trucks
1 From AEO2023 Table 41: Light-Duty VMT by Technology Type
" From AEO2023 Table 46: Transportation Fleet Car and Truck VMT by Type and
Technology
111 From AEO2023 Table 49: Freight Transportation Energy Use
The percent growth over time was calculated for each of the groups described above and
applied by HPMS category to the 2022 base year VMT from Highway Statistics Table VM-1. The
resulting values are presented in Table 3-4 below.
29
-------
Table 3-4 VMT projections for 2023-2060 by HPMS class (millions of miles)
Year
Motorcycles
Light-Duty
Vehicles
Buses
Single Unit
Trucks
Combination
Trucks
2023
24,242
2,879,385
18,419
135,421
194,589
2024
24,415
2,899,852
18,444
135,630
194,829
2025
24,509
2,911,035
18,610
137,076
196,342
2026
24,688
2,932,275
18,863
139,280
198,667
2027
24,922
2,960,106
19,085
141,298
200,617
2028
25,138
2,985,812
19,290
143,320
202,242
2029
25,309
3,006,071
19,447
145,233
203,128
2030
25,439
3,021,553
19,593
147,357
203,577
2031
25,551
3,034,758
19,766
149,897
204,094
2032
25,647
3,046,248
19,995
153,001
205,067
2033
25,783
3,062,346
20,189
155,825
205,673
2034
25,953
3,082,617
20,378
158,630
206,209
2035
26,111
3,101,301
20,580
161,538
206,874
2036
26,240
3,116,669
20,755
164,237
207,268
2037
26,390
3,134,434
20,958
167,169
207,928
2038
26,558
3,154,411
21,156
170,025
208,583
2039
26,727
3,174,458
21,353
172,972
209,121
2040
26,920
3,197,371
21,570
176,128
209,803
2041
27,118
3,220,941
21,800
179,426
210,572
2042
27,324
3,245,393
22,026
182,660
211,330
2043
27,524
3,269,108
22,238
185,818
211,924
2044
27,729
3,293,529
22,434
188,907
212,308
2045
27,947
3,319,384
22,618
192,009
212,450
2046
28,200
3,349,397
22,831
195,478
212,735
2047
28,471
3,381,615
23,050
199,076
212,998
2048
28,749
3,414,667
23,240
202,441
212,975
2049
29,031
3,448,123
23,445
205,988
213,033
2050
29,341
3,484,935
23,724
210,228
213,725
2051
29,654
3,522,140
24,006
214,555
214,419
2052
29,971
3,559,742
24,292
218,971
215,116
2053
30,291
3,597,746
24,581
223,479
215,814
2054
30,614
3,636,155
24,873
228,079
216,515
2055
30,941
3,674,974
25,169
232,773
217,218
2056
31,271
3,714,208
25,469
237,565
217,924
2057
31,605
3,753,860
25,772
242,455
218,632
2058
31,942
3,793,936
26,078
247,445
219,342
2059
32,283
3,834,440
26,389
252,538
220,054
2060
32,628
3,875,376
26,703
257,737
220,769
30
-------
4. Vehicle Populations by Calendar Year
MOVES uses vehicle populations to characterize emissions activity that is not directly
dependent on VMT. These population data are also used to allocate VMT from HPMS class to
source type and age (for more details, see Section 6). The default database stores historic
estimates and future projections of total US vehicle populations in 1990 and 1999-2060 by
source type. The MOVES database stores this information in the SourceTypeYear table, which
has three data fields: sourceTypePopulation, salesGrowthFactor and migrationRate. However,
the salesGrowthFactor and migrationRate fields are not used in MOVES.
4.1. Historic Source Type Populations (1990,1999-2022)
MOVES populations for calendar years 1990 and 1999-2022 are derived primarily from
registration data summarized in the Federal Highway Administration's annual Highway
Statistics report. Motorcycle populations are from vehicle registrations reported in Table VM-
l,6 and passenger car populations are from registrations reported in Table MV-1.21 The general
categories for truck and bus registrations presented in Highway Statistics were allocated to
specific MOVES source types as described below.
The numbers of single unit and combination trucks were determined for each calendar year
using registration data in the Highway Statistics Table VM-1. The remaining MV-1 truck
registrations were allocated to the light-duty trucks. The populations were further allocated
from the light-duty, single unit and combination truck categories to individual source types
using the source type distribution fractions shown below in Table 4-1.
The source type distribution fractions were calculated from national vehicle registration data
purchased from Polk, IHS, and SPG22 for calendar years 1999, 2011, 2014, 2020, and 2023.
These fractions were calculated as the ratio of the total registrations by source type to their
corresponding HPMS class totals (see Table 2-1 for this mapping). These fractions were then
linearly interpolated to estimate the source type distribution fractions for all years between
1999 and 2022. However, there are a few nuances to this analysis:
• Starting with MOVES3, the distinction between passenger light-duty trucks (31) and
commercial light-duty trucks (32) was revised. Now in MOVES, a light-duty truck is
considered a passenger truck if it is registered to an individual and a commercial light-
duty truck if it is registered to an organization or business. Since this is inconsistent with
the source type definitions used by the 1999 data, the ratio of passenger to commercial
light-duty trucks from the 2014 data were used for all calendar years prior to 2014.
• Vehicle registration data data are unable to distinguish between short-haul (52) and
long-haul (53) single unit trucks, and consequentially these vehicles are frequently
grouped together. These vehicles are differentiated in MOVES5 using the results of an
analysis performed for MOVES2014. This analysis relied on 2011 vehicle registration
data combined with data from VIUS200223 to differentiate between these vehicles,
which showed that of short-haul and long-haul single unit trucks, 95.8 percent are short-
haul. This percentage fraction was applied for all historic years to differentiate between
these two source types.
31
-------
• The vehicle registration data were not able to consistently differentiate between short-
haul (61) and long-haul (62) combination trucks in the same way for all years. Therefore,
we calculated the proportion of short-haul vs. long-haul combination trucks for calendar
years 1999 and 2021 and linearly interpolated between the resulting values.
o For 1999, we used the ratio of combination trucks that had a primary trip length
of 200 miles or less to those with a primary trip length greater than 200 miles in
the VIUS2002 data.
o For 2021, we used the ratio of tractor trailer vehicles with a day cab to those
with a sleeper cab in the VIUS2021 data.24
Table 4-1 Source type distributions used to a locate truck populations in MOVES5
Year
31/30
32/30
51/50
52/50
53/50
54/50
61/60
62/60
1990**
0.895179
0.104821
0.013311
0.772548
0.029035
0.185107
0.625648
0.374352
1999
0.895179
0.104821
0.015472
0.780681
0.046175
0.157671
0.574437
0.425563
2000
0.895179
0.104821
0.015838
0.783677
0.045329
0.155156
0.572101
0.427899
2001
0.895179
0.104821
0.016204
0.786673
0.044482
0.152641
0.569765
0.430235
2002
0.895179
0.104821
0.016570
0.789669
0.043636
0.150126
0.567429
0.432571
2003
0.895179
0.104821
0.016936
0.792665
0.042789
0.147610
0.565093
0.434907
2004
0.895179
0.104821
0.017302
0.795660
0.041943
0.145095
0.562757
0.437243
2005
0.895179
0.104821
0.017668
0.798656
0.041096
0.142580
0.560421
0.439579
2006
0.895179
0.104821
0.018034
0.801652
0.040249
0.140065
0.558085
0.441915
2007
0.895179
0.104821
0.018400
0.804648
0.039403
0.137550
0.555750
0.444250
2008
0.895179
0.104821
0.018766
0.807643
0.038556
0.135034
0.553413
0.446587
2009
0.895179
0.104821
0.019131
0.810639
0.037710
0.132519
0.551078
0.448922
2010
0.895179
0.104821
0.019497
0.813635
0.036863
0.130004
0.548742
0.451258
2011
0.895179
0.104821
0.019863
0.816631
0.036017
0.127489
0.546406
0.453594
2012
0.895179
0.104821
0.015282
0.834433
0.036802
0.113483
0.544070
0.455930
2013
0.895179
0.104821
0.010700
0.852236
0.037587
0.099477
0.541734
0.458266
2014
0.895179
0.104821
0.006118
0.870039
0.038372
0.085471
0.539398
0.460602
2015
0.898254
0.101746
0.005968
0.867090
0.038242
0.088700
0.537062
0.462938
2016
0.901330
0.098670
0.005817
0.864142
0.038112
0.091928
0.534726
0.465274
2017
0.904405
0.095595
0.005667
0.861193
0.037982
0.095157
0.532390
0.467610
2018
0.907481
0.092519
0.005517
0.858245
0.037852
0.098386
0.530054
0.469946
2019
0.910556
0.089444
0.005366
0.855297
0.037722
0.101615
0.527718
0.472282
2020
0.913632
0.086368
0.005216
0.852348
0.037592
0.104844
0.525382
0.474618
2021
0.913406
0.086594
0.005046
0.854251
0.037676
0.103027
0.523047
0.476953
2022
0.913180
0.086820
0.004876
0.856154
0.037760
0.101210
0.523046
0.476954
Fractions may not
Fractions for 1990
sum to one due to rounding.
were retained from MOVES201425 with the exceptions noted in the text.
Buses were allocated using different data sources:
• School bus (43) populations for 2002-2023 come from the School Bus Fleet Fact Book26
publication series' School Transportation Statistics tables. Since these values are
presented as totals corresponding to academic years (e.g., 2016-2017) and MOVES
requires national values to be entered for calendar years, the data were taken to
32
-------
correspond to the year in which the school year ends (2017, in the example). For 1990
and 1999-2001, school buses were assumed to be a constant proportion of the total bus
population in each year based on the 2002 counts. Note that the School Bus Fleet Fact
Book series did not publish the School Transportation Statistics tables for the academic
years 2019-2020 or 2020-2021. Therefore, the school bus populations for calendar years
2020 and 2021 were linearly interpolated from the 2019 and 2022 school bus
population values.
• Transit bus (42) populations were calculated from the Federal Transit Administration's
National Transit Database (NTD)27 data series on Revenue Vehicle Inventory and Rural
Revenue Vehicle Inventory. See Section 2.2.4 for more information on the definition of
transit buses in MOVES. For 1990 and 1999-2001, transit buses were assumed to be a
constant proportion of the total bus population in each year based on the 2002 counts.
• Other bus (41) populations were calculated as the remainder of the MV-1 bus
registrations less the school bus and transit bus populations. Note that the Highway
Statistics series on bus populations show large changes in bus registrations for 2011,
2012, and 2016-2020, inconsistent with intermediate years as well as historic
populations. Lacking evidence that these specific data reflect actual changes in the
number of buses operating in the US, the bus registration values for those years were
dropped and estimated instead with linear interpolation/extrapolation. Specifically,
2011 and 2012 values were linearly interpolated from 2010 and 2013 registrations, and
2016-2020 values were linearly interpolated from 2015 and 2021 registrations.
For all source type populations derived from Table VM-1, note that this registration data has
the same vehicle categorization issue as the VMT data for reporting years prior to 2007 as
described in Section 3.1. Similar to the VMT analysis, we used the FHWA-revised values for
2000-2006 and adjusted the registration data ourselves for 1990 and 1999 as described in
Section 3.1.
33
-------
ycle
,632
,581
,068
,056
,156
,035
,870
,146
,958
,476
,926
,724
,503
,502
,939
,687
,718
,936
,380
,204
,185
,314
,435
,491
,664
Table 4-2 Historic source type populations for calendar years 1990 and 1999-2022
Passenger
Truck
Light Com.
Truck
Other
Bus
Transit
Bus
School
Bus
Refuse
Truck
Single
Unit
Short-haul
Single
Unit
Long-
haul
Motorhome
Combination
Short-haul
Combination
Long-haul
35,233,652
3,330,743
172,025
48,151
318,050
57,229
3,300,770
145,578
795,855
1,010,435
66,461,938
7,782,396
226,133
63,296
418,087
102,180
5,155,620
304,940
1,041,260
1,320,899
69,931,300
8,188,643
238,473
66,750
440,902
104,180
5,154,843
298,161
1,020,579
1,378,719
73,943,775
8,658,486
239,567
67,056
442,925
112,984
5,485,096
310,152
1,064,291
1,407,620
74,801,221
8,758,889
243,137
68,055
449,525
115,407
5,499,871
303,912
1,045,592
1,369,599
76,531,916
8,961,546
237,582
68,604
470,364
119,510
5,593,475
301,943
1,041,619
1,352,574
80,962,696
9,480,370
256,107
68,796
470,371
124,111
5,707,483
300,865
1,040,806
1,350,820
84,088,058
9,846,336
264,495
69,514
473,044
131,119
5,927,101
304,988
1,058,135
1,380,374
87,378,257
10,231,604
274,929
70,232
476,798
139,868
6,217,559
312,172
1,086,333
1,438,738
89,289,785
10,455,435
266,607
82,378
485,451
149,344
6,531,061
319,821
1,116,445
1,464,593
8,952,385
10,415,927
270,183
84,744
488,381
155,530
6,693,787
319,558
1,119,172
1,430,701
89,149,115
10,438,963
293,180
86,757
462,056
159,864
6,773,781
315,108
1,107,344
1,442,235
89,116,997
10,435,202
284,728
89,197
472,126
160,213
6,685,794
302,914
1,068,268
1,400,863
96,844,800
11,340,094
291,434
8,122
472,661
155,312
6,385,282
281,618
996,843
1,339,589
109,633,104
12,837,547
298,491
91,912
467,980
125,160
6,834,249
301,420
929,457
1,343,360
109,510,674
12,823,211
297,426
94,222
472,901
86,947
6,925,277
305,434
808,350
1,338,813
113,022,019
13,234,374
289,926
98,060
484,041
50,956
7,246,342
319,595
711,866
1,390,135
116,820,608
13,232,356
300,197
103,669
485,041
50,465
7,332,377
323,389
750,071
1,475,246
121,394,421
13,289,294
316,211
106,871
474,194
50,882
7,558,232
333,350
804,054
1,471,590
126,052,567
13,323,652
326,070
108,115
471,461
52,913
8,040,961
354,641
888,483
1,539,789
128,989,018
13,150,673
331,220
106,645
476,150
56,975
8,863,868
390,935
1,016,121
1,540,344
132,273,282
12,993,192
334,857
107,660
479,867
54,524
8,690,184
383,274
1,032,451
1,543,687
135,714,463
12,829,494
340,394
107,199
483,161
51,681
8,445,415
372,479
1,038,834
1,571,399
141,307,570
13,396,505
346,059
106,610
486,454
54,059
9,152,062
403,645
1,103,783
1,643,859
144,828,919
13,769,593
360,240
104,131
489,748
54,044
9,489,609
418,532
1,121,812
1,699,809
34
-------
4.2. Projected Vehicle Populations (2023-2060)
Vehicle stock estimates from the reference case of AEO2023 were used to project future
populations, using a methodology similar to the VMT projections as described in Section 3.2.
Because AEO vehicle categories differ from MOVES source types, the AEO projected vehicle
stocks were not used directly. Instead, year-to-year percent changes in the projected values
were calculated and applied to the base year populations. Since AEO2023 only projects out to
2050, populations for years 2051-2060 were assumed to continue to grow at the same growth
rate as between 2049 and 2050.
Table 4-3 shows the mappings between AEO stock categories and MOVES source types. Where
multiple AEO categories are listed, their stocks were summed before calculating the year-over-
year growth rates. AEO's car category was mapped to both motorcycle and passenger car
categories. Motorcycles were included here because they were not explicitly accounted for
elsewhere in AEO. Since buses span a large range of heavy-duty vehicles and activity, the total
heavy-duty category was mapped to each source type in the HPMS bus category. AEO's light-
medium- and medium-heavy-duty categories were combined for mapping to each source type
in the HPMS single unit truck category and AEO's heavy-heavy-duty category was mapped to
each source type in the HPMS combination truck category. We acknowledge that using stock
growth estimates from different vehicle types as surrogates for motorcycles and buses, in
particular, will introduce additional uncertainty into these projections.
Table 4-3 Mapping AEO categories to source types for projecting vehicle populations
AEO Stock Category Groupings
MOVES Source Type
Total Car Stock1
11 - Motorcycle
21 - Passenger Car
Total Light Truck Stock1
+
Total Commercial Light Truck Stock"
31 - Passenger Truck
32 - Light Commercial Truck
Total Heavy-Duty Stock1"
41-Other Bus
42 -Transit Bus
43 - School Bus
Light-Medium Subtotal Stock1"
+
Medium Subtotal Stock1"
51 - Refuse Truck
52 - Single Unit Short-haul Truck
53 - Single Unit Long-haul Truck
54 - Motor Home
Heavy Subtotal Stock1"
61 - Combination Short-haul Truck
62 - Combination Long-haul Truck
' From AEO2023 Table 45: Light-Duty Vehicle Stock by Technology Type
" From AEO2023 Table 46: Transportation Fleet Car and Truck Stock by Type and Technology
'" From AEO2023 Table 49: Freight Transportation Energy Use
The percent growth over time was calculated for each of the groups described above and
applied to the 2022 base year source type populations. The resulting populations are presented
in Table 4-4.
35
-------
Table 4-4 Projected source type populations for ca
yearlD
Motorcycle
Passenger
Car
Passenger
Truck
Light Com.
Truck
Other
Bus
Transit
Bus
School
Bus
Refuse
Truck
Single Unit
Short-haul
Single
Unit
Long-
haul
Motorhome
Combination
Short-haul
Combination
Long-haul
2023
9,588,387
97,950,475
147,741,538
14,046,510
368,351
106,476
500,775
55,505
9,746,209
429,849
1,152,146
1,725,984
1,573,883
2024
9,627,377
96,079,762
151,037,970
14,359,918
376,892
108,944
512,386
57,070
10,020,895
441,964
1,184,618
1,752,277
1,597,859
2025
9,672,530
94,141,077
154,616,987
14,700,192
385,772
111,511
524,459
58,709
10,308,783
454,661
1,218,650
1,778,974
1,622,203
2026
9,719,362
92,183,949
158,279,863
15,048,440
394,925
114,157
536,903
60,389
10,603,725
467,669
1,253,517
1,806,997
1,647,757
2027
9,761,851
90,181,690
161,889,657
15,391,641
404,286
116,863
549,629
62,092
10,902,722
480,856
1,288,863
1,836,398
1,674,567
2028
9,797,208
88,107,368
165,397,453
15,725,144
413,475
119,519
562,121
63,773
11,197,930
493,876
1,323,761
1,864,778
1,700,446
2029
9,832,622
86,046,008
168,885,842
16,056,802
422,047
121,997
573,775
65,389
11,481,713
506,392
1,357,308
1,888,893
1,722,436
2030
9,862,757
83,984,330
172,234,045
16,375,132
430,086
124,321
584,704
66,952
11,756,082
518,493
1,389,743
1,909,190
1,740,944
2031
9,887,812
81,938,567
175,419,885
16,678,026
437,540
126,475
594,838
68,428
12,015,244
529,923
1,420,379
1,926,662
1,756,877
2032
9,909,241
79,955,493
178,443,879
16,965,531
444,815
128,578
604,728
69,901
12,273,979
541,335
1,450,966
1,942,081
1,770,937
2033
9,927,373
78,041,522
181,283,816
17,235,538
451,433
130,491
613,725
71,264
12,513,206
551,885
1,479,246
1,955,025
1,782,740
2034
9,943,225
76,232,705
183,930,306
17,487,153
457,560
132,262
622,055
72,576
12,743,676
562,050
1,506,491
1,964,476
1,791,358
2035
9,962,447
74,574,885
186,476,063
17,729,190
463,733
134,047
630,448
73,889
12,974,134
572,214
1,533,734
1,974,487
1,800,487
2036
9,987,663
73,123,865
188,930,883
17,962,582
470,207
135,918
639,249
75,238
13,211,067
582,664
1,561,743
1,986,323
1,811,280
2037
10,018,606
71,878,523
191,284,014
18,186,306
476,797
137,823
648,208
76,612
13,452,299
593,303
1,590,261
1,998,366
1,822,262
2038
10,056,737
70,846,556
193,567,740
18,403,431
483,360
139,720
657,131
77,988
13,694,028
603,965
1,618,836
2,009,936
1,832,812
2039
10,101,118
70,001,880
195,797,147
18,615,391
489,601
141,524
665,615
79,345
13,932,305
614,474
1,647,004
2,018,568
1,840,684
2040
10,152,827
69,333,536
198,008,279
18,825,615
495,518
143,234
673,659
80,665
14,163,964
624,691
1,674,390
2,025,125
1,846,663
2041
10,207,183
68,786,720
200,155,731
19,029,783
501,234
144,887
681,430
82,009
14,399,970
635,100
1,702,289
2,028,020
1,849,303
2042
10,264,167
68,342,485
202,256,398
19,229,504
507,216
146,616
689,562
83,350
14,635,442
645,485
1,730,126
2,034,297
1,855,027
2043
10,322,095
67,953,475
204,335,311
19,427,156
513,860
148,536
698,595
84,813
14,892,408
656,818
1,760,503
2,042,564
1,862,565
2044
10,381,193
67,622,150
206,389,451
19,622,454
521,006
150,602
708,310
86,365
15,164,793
668,832
1,792,703
2,052,572
1,871,691
2045
10,442,100
67,338,410
208,434,012
19,816,840
527,969
152,615
717,776
87,882
15,431,262
680,584
1,824,203
2,062,027
1,880,313
2046
10,504,248
67,138,902
210,409,243
20,004,635
534,455
154,490
726,594
89,342
15,687,625
691,891
1,854,509
2,068,533
1,886,246
2047
10,566,202
66,947,924
212,369,280
20,190,985
540,818
156,329
735,245
90,815
15,946,302
703,300
1,885,089
2,072,906
1,890,233
2048
10,627,410
66,782,027
214,277,619
20,372,420
547,041
158,128
743,704
92,284
16,204,221
714,675
1,915,578
2,075,778
1,892,853
2049
10,689,995
66,633,868
216,186,374
20,553,895
552,693
159,762
751,389
93,673
16,448,018
725,427
1,944,399
2,075,711
1,892,791
2050
10,756,125
66,529,452
218,130,466
20,738,729
557,938
161,278
758,520
95,019
16,684,339
735,850
1,972,335
2,072,804
1,890,140
endar years 2023-2060
36
-------
yearlD
Motorcycle
Passenger
Car
Passenger
Truck
Light Com.
Truck
Other
Bus
Transit
Bus
School
Bus
Refuse
Truck
Single Unit
Short-haul
Single
Unit
Long-
haul
746,423
Motorhome
Combination
Short-haul
Combination
Long-haul
2051
10,822,663
66,425,199
220,092,042
20,925,226
563,233
162,808
765,718
96,384
16,924,055
2,000,673
2,069,901
1,887,493
2052
10,889,614
66,321,110
222,071,257
21,113,399
568,578
164,353
772,985
97,769
17,167,215
757,147
2,029,418
2,067,002
1,884,849
2053
10,956,979
66,217,184
224,068,270
21,303,265
573,974
165,913
780,321
99,173
17,413,869
768,025
2,058,577
2,064,107
1,882,210
2054
11,024,760
66,113,421
226,083,242
21,494,839
579,422
167,488
787,726
100,598
17,664,066
779,060
2,088,154
2,061,216
1,879,573
2055
11,092,961
66,009,820
228,116,334
21,688,135
584,920
169,077
795,202
102,043
17,917,859
790,254
2,118,156
2,058,329
1,876,941
2056
11,161,583
65,906,382
230,167,709
21,883,169
590,471
170,682
802,749
103,510
18,175,297
801,608
2,148,589
2,055,447
1,874,312
2057
11,230,630
65,803,106
232,237,531
22,079,957
596,075
172,301
810,367
104,997
18,436,435
813,125
2,179,459
2,052,568
1,871,687
2058
11,300,104
65,699,991
234,325,967
22,278,515
601,732
173,937
818,057
106,505
18,701,324
824,808
2,210,773
2,049,693
1,869,066
2059
11,370,008
65,597,039
236,433,183
22,478,858
607,442
175,587
825,821
108,036
18,970,020
836,658
2,242,537
2,046,823
1,866,448
2060
11,440,345
65,494,247
238,559,349
22,681,003
613,207
177,254
833,658
109,588
19,242,576
848,679
2,274,757
2,043,956
1,863,834
37
-------
5. Fuel Type, Regulatory Class, and Engine Technology Distributions by Calendar
Year
Despite the availability of vehicle registration databases, comprehensive surveys for
characterizing travel patterns, and sophisticated sensors and cameras for measuring vehicle
activity, it is still difficult to estimate vehicle populations in the categories needed for emissions
inventory modeling. Differentiating, for example, between passenger car and trucks, or
between light-duty and heavy-duty trucks presents substantial modeling challenges since the
characteristics that are important for emissions are not always readily observable.28 29 To
develop MOVES defaults, we have used a variety of registration and survey data to identify key
vehicle parameters.
This section explains how the national default fuel type, regulatory class, and engine technology
distributions were developed by source type. These national defaults are stored in the
SampleVehiclePopulation table.
Note that MOVES may be run at the county or project scale with local information to accurately
capture fuel type variation by geographic region. As explained in the MOVES Technical
Guidance,2 this can be done through the AVFT importer in the MOVES interface.
5.1. Sample Vehicle Population
To match source types to emission rates, MOVES must associate each source type with specific
fuel types, technologies (EngTech), and regulatory classes. As vehicle markets shift, these
distributions change with model year. This information is stored in the
SampleVehiclePopulation (SVP) table, which contains two fractions: stmyFraction and
stmyFuelEngFraction.
The stmyFraction represents the default national fuel type, EngTech, and regulatory class
allocation for each source type and model year. We define the stmyFraction as shown in
Equation 5-1.
j., Nst,my,ft,et,rc
f yStmy)stmy jtetrc —
\\ a] Equation 5-1
7 ftEFT st,my,ft,et,rc
^—'etEET
rcERC
where the number of vehicles N in a given source type st, model year my, fuel type ft,
EngTech et, and regulatory class rc is divided by the sum of vehicles across the set of all fuel
types FT, EngTechs ET, and regulatory classes RC. That is, the denominator is the total number
of vehicles in a given source type and model year and so the stmyFraction must sum to one for
each source type and model year. For example, model year 2010 passenger trucks have
stmyFractions that indicate the distribution of these vehicles between gasoline, diesel, E85, and
electricity fuel types with their associated EngTechs, and regulatory classes 30 and 41. A value
of zero indicates that the MOVES default population of vehicles of that source type, model year,
38
-------
fuel type, EngTech, and regulatory class is negligible in the national population or does not
exist.
Because a modeler may modify fuel type and EngTech distributions by source type and model
year to simulate local conditions through the Alternative Vehicle Fuel and Technology (AVFT)
table—but is not expected to modify regulatory class distributions—the
SampleVehiclePopulation table also contains the stmyFuelEngFraction. When a modeler
supplies an AVFT table, MOVES will use the stmyFuelEngFraction to apply a default regulatory
class distribution to the user-supplied fuel type and EngTech distributions, regardless of
whether these vehicles exist in the default. Similar to the stmyFraction above, we define
stmyFuelEngFraction as shown in Equation 5-2.
for number of vehicles N in a given source type st, model year my, fuel type ft, EngTech et,
regulatory class rc, and the set of all regulatory classes RC. In this case, the denominator is the
total for a given source type, model year, fuel type, and EngTech, and so the
stmyFuelEngFraction must sum to one for each combination of source type, model year, fuel
type,and EngTech.
For a concrete example of how stmyFraction and stmyFuelEngFraction are used in MOVES, take
the example of MY2030 combination long-haul trucks. The stmyFraction assigns the default
fuel, EngTech, and regulatory classes to the population of MY2030 combination long-haul
trucks, which are mostly HHD class 8 diesel. However, a modeler could create a future scenario
in which there is a high penetration of fuel cell electric trucks. The stmyFuelEngFraction allows
MOVES to assign vehicles to regulatory classes without also requiring the modeler to supply
future regulatory class distributions.
As noted in Section 2.4, these fuel type fractions indicate the fuel capability of the vehicle,
which is not necessarily the fuel being used by the vehicle. MOVES allocates fuel to specific
vehicles in a two-step process: 1) vehicles are classified by the type of fuel they can use in the
fuel type fraction and then 2) fuels are distributed according to how much of each fuel is used
relative to the vehicles' total fuel consumption in the fuel usage fraction. For example, Figure
5-2 shows the national default fuel type fractions for all light-duty vehicles among the different
MOVES fuel types. In this report's nomenclature, E85-capable and flexible fuel vehicles are
synomous—they describe vehicles that can accept either gasoline or E-85 fuel. The amount of
E-85 versus the amount of gasoline used out of all the fuel consumed by the vehicle is stored in
the fuelUsageFraction table. Discussion on fuel usage can be found in the MOVES Fuel Supply
Report.15
f (stTnyfuclcTig^ Sf- my ^ rc
ZNst,my,ft,
rceRC
Equation 5-2
st,my,ft,et,rc
39
-------
5.2. Historic Distributions (MY1950-2022)
Historic fuel type, regulatory class, and engine technology distributions were derived from
vehicle registration data22 pulled in 1999, 2011, 2014, 2020, and 2023, as well as other data
sources as detailed in the sections below. This analysis relied on the following fields in the
vehicle registration data:
• Source type (assigned by the data provider following MOVES source type definitions
using vehicle/truck type, owner type, body style, and cab configuration data, among
others)
• Fuel type
• Gross vehicle weight rating (GVWR)
• Model year
• Make / model
• Body style / cab configuration
Before calculating fuel type, regulatory class, and engine technology distributions, the vehicle
registration data as delivered to EPA were cleaned using the following steps:
1. Some source type assignments were adjusted:
a. Some compact SUVs were classified as light trucks where EPA emission
certification data showed that those makes and models were regulated as cars;8
we reassigned these vehicles as passenger cars (sourceTypelD 21).
b. The vehicle registration data were unable to properly differentiate transit buses
and other buses, so all vehicles classified as transit bues (sourceTypelD 42) were
reassigned as other buses (sourceTypelD 41).
c. Class 3 trucks classified as light commercial trucks were reassigned as single unit
short-haul trucks (sourceTypelD 52).
d. Trucks classified as combination trucks with a body style of "straight truck" were
reassigned as single unit short-haul trucks (sourceTypelD 52).
e. Vehicles classified as single unit trucks with a body style of "bus non school"
were reassigned as other buses (sourceTypelD 41).
f. Trucks classified as single unit trucks with a body style of "tractor truck" were
reassigned as combination trucks. If the vehicle had a "long conventional" cab, it
was assigned to the long-haul category, otherwise it was assigned to as a short-
haul.
2. Fuel type cleaning:
a. Electric hybrids with gasoline or diesel engines were grouped with fully gasoline
or diesel vehicles since MOVES does not model hybrids separately.
b. Light-duty fuel cell vehicles were reassigned as battery electric vehicles because
MOVES does not model light-duty fuel cell vehicles.
c. Vehicles categorized as "ethanol" or "flexible" were grouped with E-85 vehicles.
d. If a vehicle's fuel type was unknown, it was set to be the most common fuel type
for the vehicle's source type and model year.
40
-------
e. Any remaining vehicles with alternative fuels (such as methanol) orvehicles with
source type/fuel type combinations that MOVES cannot model (such as CNG
light commercial trucks) were dropped from the data.
f. MOVES fuelTypelDs and engTechlDs were then assigned based on the cleaned
fuel type data using the definitions in Section 2.4.
3. MOVES regClasslDs were assigned based on the GVWR data using the definitions in
Section 2.3 with the following exceptions:
a. All E-85 light-duty trucks were assigned to regClassID 30 because MOVES does
not have emission rates for E-85 vehicles in regClassID 41.
b. Light-duty trucks (sourceTypelDs 31 and 32) with an unknown GVWR were
assigned to regClassID 30.
c. Diesel Class 3 engine-certified vehicles (identified by make and model) in model
year 2017 and beyond were reassigned to regClassID 42.
d. Class 1 and Class 2 motor homes (sourceTypelD 54) were assigned to regClassID
41.
After cleaning the vehicle registration data, historic fuel type, regulatory class, and engine
technology distributions could be calculated. The following outlines which data sources were
used for which model years:
• For pre-2000 model years, the 1999 vehicle registration data were combined with data
from the 1997 and 2002 Vehicle Inventory and Use Surveys (VIUS). Documentation for
the pre-2000 model years may be found in Appendix A. Note that there are two
exceptions to our reliance on the VlUS-based analysis for model years before 2000:
o For passenger trucks and light commercial trucks, we used the 2014 vehicle
registration data for model years 1981-2000 because the MOVES definition of
these vehicle types is no longer consistent with the VIUS definition.
Unfortunately, the data are too scarce in the 2014 vehicle registration data for
pre-1981 model years, so we continue to rely on the previous analysis as
described in Appendix A analysis for those model years.
o We also relied exclusively on the 2014 vehicle registration data for all pre-2000
model years of transit buses, other buses, and motor homes.
• Model years 2000-2013 rely on the 2014 vehicle registration data.
• Model years 2014-2019 rely on the 2020 vehicle registration data.
• Model years 2020-2022 rely on the 2023 vehicle registration data.
The subsections below contain detail on additional data sources which were used in developing
historic fuel type, regulatory class, and engine technology distributions.
5.2.1. Light-duty Vehicles
The fraction of electric light-duty vehicles for model years 2017-2022 come from the 2022 EPA
Automotive Trends report.30 This data source includes the proportion of passenger cars and
passenger trucks that are electric. The proportion of electric passenger trucks was also applied
to light commercial trucks.
41
-------
5.2.2. Buses
Transit buses and "other buses" are not well distinguished from each other in the vehicle
registration data. The National Transit Database is a potential alternate data source for transit
buses, but since it lacks weight class information, it could not be used to calculate regulatory
class distributions. Instead, considering that the vehicles in the transit and "other bus"
categories may overlap, we grouped these categories together when determining fuel type,
regulatory class, and engine technology distributions. The only difference between the transit
and other bus distributions is in the categorization of Class 8 buses, since urban transit buses
are regulated separately from other heavy-duty vehicles under 40 CFR 86.091-2.13 For this
reason, Class 8 CNG and diesel transit buses were classified in regulatory class 48, whereas
Class 8 gasoline transit buses and all Class 8 other buses were classified as regulatory class 47.
Additionally, MOVES can only model CNG other buses in regulatory class 47.
5.2.3. Single Unit Trucks
Lacking a way to differentiate single unit short-haul and long-haul trucks in the registration
data, we used the fuel type, regulatory class, and engine technology distributions for "non-
refuse, non-RV single unit trucks" identically for both short-haul and long-haul single unit
trucks.
For refuse trucks, we found that electric refuse trucks were not well represented in the vehicle
registration data, so we used electric refuse truck counts reported to EPA from the 2019 Annual
Production Volume Reports into Engine and Vehicle Compliance Information System31 for
applicable model years instead of the electric refuse truck counts in the registration data.
5.2.4. Combination Trucks
We found that battery electric combination trucks were not well represented in the vehicle
registration data, so we used electric combination truck counts reported to EPA from the 2019
Annual Production Volume Reports into Engine and Vehicle Compliance Information System31
instead of the battery electric combination truck counts in the registration data.
5.2.4.1. Glider Truck Populations
"Glider trucks" in MOVES refers to vehicles with new chassis but with older engines that do not
meet MY 2007 or 2010 emissions standards (Section 2.3). Most glider trucks are Class 8 vehicles
that use diesel heavy heavy-duty engines. For simplicity, in MOVES, we assume that all glider
vehicles are HHD but modeled as a separate regulatory class (regClassID 49) and are only
populated within the combination short- and long-haul truck source types (sourceTypelD 61
and 62, respectively).
We used sales data from both glider kit manufacturers and glider assembler manufacturers to
estimate glider truck populations in MOVES. The glider kits contain the vehicle chassis and cab,
but lack the engine and transmission. Glider assembler manufacturers (referred heareafter as
"glider assemblers") assemble the glider vehicle by installing the engine and transmission into
42
-------
the glider kit produced by the glider kit manufacturer. Most glider assemblers are small
businesses that sell less than 10 glider vehicles per year. However, most of the glider vehicles
made from 2016 to 2020 were produced by a handful of large glider assemblers.
We estimated the glider population based on annual glider production volume (sales) data for
model years 2010 to 2016 shared as claimed confidential business information (CBI) from the
two major glider kit manufacturers.32 For use in MOVES, we assumed annual sales of 500 for
glider vehicles for years prior to 2010 and rounded the reported production volumes in the
years 2010 to 2016 to the nearest thousand, as shown in Table 5-1. The rounded values reflect
the uncertainty regarding the number of gliders in the fleet, including the contribution of small
volume glider manufacturers and the number used in single unit vehicles.6
Table 5-1: Annual Glider Vehicle Sales Estimates Applied in MOVES Based on Claimed CBI
Data Shared by Manufacturers
MY
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
Glider
Population
500
500
1000
3000
4000
5000
8000
12000
7000
7500
3500
1500
0
0
For estimating the glider sales for 2017 and 2018, we did not have data from the two glider kit
manufacturers, but we have data from the glider assemblers. As part of EPA's Greenhouse Gas
Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty Engines and Vehicles -
Phase 2 rulemaking (Phase 2)102, the agency adopted new rules for glider kits, glider vehicles
and glider engines. Starting in model year 2018, a glider assembler could continue to sell glider
vehicles, without limit, if the glider engine was from a 2010 or later model year. If a glider
assembler wishes to sell glider vehicles with earlier model year engines, they are limited to the
lesser of 300 per year or the number of glider vehicles they sold in calendar year 2014. The
regulation requires glider assemblers to report their sales data to EPA, including their 2014
sales to identify their individual sales allowances. The number of manufacturers who reported
glider sales data to EPA for 2014, 2017, 2018, and 2019 are shown in Table 5-2, and the
reported sales are displayed in Figure 5-1. Prior to the 2018 model year, more than 260 glider
assemblers reported their sales data, including five manufacturers which produced more than
300 gliders and whose sales were capped starting in 2018. The reported sales from the glider
assemblers in 2014 was close to 9,000 vehicles, which compares well to the 8,000 glider kit
sales reported for 2014 shown in Table 5-1.
e In 2017, glider manufacturers are limited to producing their maximum production between MYs 2010 and 2014.
See 81 FR 73478 for more information.
43
-------
Table 5-2. Number of Glider Assemblers that reported to EPA, grouped by glider production in
2014
Glider
Manufacturers
Manufacturers
Manufacturers
Manufacturers
Production in
reporting in
reporting in
reporting in
reporting in
2014
2014
2017
2018
2019
<=10
208
70
73
21
10-50
53
23
23
6
51-300
8
5
5
0
300 +
5
2
2
0
Total
274
100
103
27
7000
6000
5000
4000
3000
12014 Manufacturer Reported Sales
12017 Manufacturer Reported Sales
I 2018 Manufacturer Reported Sales
2019 Manufacturer Reported Sales
2000
1000
I..
<=10
10-50
51-300
I.
300 +
Figure 5-1. Reported Glider Sales by Glider Assemblers for Calendar Year 2014, 2017, 2018,
and 2019.
The total number of glider sales by glider assemblers is significantly reduced in 2017, 2018, and
2019 as shown in Figure 5-1. One reason for the decrease is that we only received reported
sales from a subset of the assemblers who reported their sales in 2014. For the assemblers who
reported sales in 2017, we calculated the ratio of sales to the number of sales these assemblers
reported in 2014. For the assemblers who reported sales in 2018 and 2019, we calculated the
ratio of the sales to maximum allowable (the smaller of their sales in 2014 or 300 glider
vehicles). As shown in Table 5-3, the larger glider assemblers tended to produce more gliders in
comparison to the 2014 sales and their maximum allowable sales.
44
-------
Table 5-3. Ratio of 2017, 2018, and 2019 Glider Sales to 2014 Sales or the Maximum
Allowable Sales (2017 and 2018) from Reporting Glider Assemblers
A
B
C
Glider
Ratio of 2017 sales to
Ratio of 2018 sales to
Ratio of 2019 to 2014 sales
Production
2014 sales for
maximum allowable for
to maximum allowable for
in 2014
assemblers that
assemblers that
assemblers that reported
reported in 2017
reported in 2018
in 2019
<=10
0.30
0.46
0.62
10-50
0.57
0.63
0.70
51-300
0.68
0.71
300 +
1.02
1.00
To estimate the number of glider sales in 2017, we multiplied the ratios in Column (A) of Table
5-3 by total number of gliders sales by glider assembler size reported in 2014. This is under the
assumption that the sale growth/reduction rate from 2014 to 2017 is the same between
reported glider assemblers and the overall glider assemblers. And similarly, to estimate the
number of glider sales in 2018, we multiplied the ratios in Column (B) by the maximum
allowable of all glider assembler manufacturers who reported in 2014. When rounded to the
nearest 500, this yielded total glider sales estimates of 7,500 in 2017 and 3,500 in 2018 as
shown in Table 5-1.
To estimate 2019 sales, we had limited data from the glider assembler manufacturers (only 27
assembler manufacturers reported 2019 sales at the time of the analysis, all of which sold less
than 50 gliders per manufacturer). The two major glider kit manufacturers informed EPA that
they had stopped production of glider kits in 2018. As such, we assumed that the 2019 glider
vehicles would be 1500 vehicle sales, which is calculated from 50 gliders per manufacturer and
27 reported assembler manufacturers, rounded to the nearest 500. We assumed zero sales for
2020 and later model years. Assuming an insignificant number of gliders in future years is
appropriate due to decreasing availability of pre-2010 engines and requirements for 2021 and
later model year glider vehicles to meet the Medium and Heavy-duty Greenhouse Gas Phase 2
emissions and fuel economy standards.94
We calculated the fraction of gliders (stmyFraction) by dividing the estimated glider production
by the number of age-0 combination trucks for each model year using Equation 5-3. We applied
this fraction to both short-haul combination (sourceTypelD 61) and long-haul combination
trucks (sourceTypelD 62). Note all gliders are regClassID 49 diesel trucks.
/ (,Stmy)regciass 49 model year i,
_ Glidersi Equation 5-3
Combination TvuckSgQ^g^ypg si+62,i
45
-------
5.3. Projected Distributions (MY2023-2060)
Vehicle sales estimates from the reference case of AEO2023 were used to project future fuel
type, regulatory class, and engine technology distributions using a methodology similar to the
VMT projections as described in Section 3.2. However, because AEO2023 does not account for
EPA's Light- and Medium-Duty Multi-Pollutant Rule (LMDV)33 and Greenhouse Gas Emissions
Standards for Heavy-Duty Vehicles—Phase 3 (HDP3)34 or CARB's Advanced Clean Trucks rule,35
we replaced the EV projections in AEO2023 with EPA analyses as described in the sections
below.
5.3.1. Light-duty and Medium-duty Vehicles
For model years 2023 and later, electric light-duty vehicle (regClassID 20), light-duty truck
(regClassID 30), and medium-duty (regClassID 41) sales fractions are based on electric vehicle
cost and consumer preferences as modeled in EPA's OMEGA (Optimization Model for reducing
Emissions of Greenhouse Gases from Automobiles) model under its 2027 LMDV FRM
scenario.36
The remaining sales fractions for each vehicle class were allocated to the relevant internal
combustion engine (ICE) fuels using their proportion of sales fractions in AEO2023.20
Specifically:
• Light-duty vehicle (regClassID 20) ICE proportions by model year were calculated using
the new car sales projections presented in the "Light-Duty Vehicle Sales by Technology
Type" table.
o Note: The new car sales projections in this table include a very small fraction of
diesel passenger car sales in the future. Since there are no longer any diesel
passenger car models for sale in the U.S. (as confirmed using our vehicle
registration data for model years 2020 and later), we set the diesel passenger car
fractions to 0 for all projected model years.
• Light-duty truck (regClassID 30) ICE proportions by model year were calculated using the
new light truck sales projections presented in the "Light-Duty Vehicle Sales by
Technology Type" table.
• Medium-duty (regClassID 41) ICE proportions by model year were calculated using the
commercial light truck sales projections presented in the "Transportation Fleet Car and
Truck Sales by Type and Technology" table.
Since AEO2023 only projects out to 2050, the ICE proportions for model years 2051-2060 used
the same values as model year 2050.
5.3.2. Heavy-duty Vehicles
As discussed above, we projected national default heavy-duty fuel type distributions based on
AEO as well as finalized regulations or legislative actions which may impact fuel type
distributions. First, we projected fuel type distributions using AEO2023,20 which does not
include the impact of CARB's Advanced Clean Trucks (ACT) rule35 or EPA's HD GHG Phase 3
rule.34 Second, we projected HD ZEV adoption that would result from CARB's ACT rule in the
46
-------
states that have adopted it. Third, we updated fuel type distributions based on the
electrification projected in the HD GHG Phase 3. The Phase 3 rule is a performance-based
tailpipe standard and does not mandate a shift in fuel distributions across the HD fleet, but we
expect compliance with Phase 3 to be predominantly achieved via electrification.
We calculated HD electrification by calculating the proportion of HD sales which would be zero-
emission vehicles (ZEV). In MOVES, HD ZEVs can be either battery electric vehicles (BEV) or fuel
cell electric vehicles (FCEV). After we calculated ZEV adoption, we apportioned ZEV sales to
BEVs and FCEVs and reduced the ICE sales percentages to preserve the total number of HD
sales modeled by MOVES.
5.3.2.1. Annual Energy Outlook Projections
Because AEO vehicle categories differ from MOVES source types, the AEO projected vehicle
sales were not used directly. Instead, we calculated percent changes in the AEO projections and
applied them to base year data used in MOVES.
To start, we first calculated model year 2022 vehicle populations by source type, regulatory
class and fuel type from the vehicle registration data pulled in calendar year 2023.22 Next, we
calculated year-over-year percent changes in the projected sales by AEO vehicle categories and
fuel type as presented in the "Freight Transportation Energy Use" table.20 To use these year-
over-year changes, we mapped the AEO vehicle categories to MOVES regulatory class as
presented in Table 5-4.
Table 5-4 Mapping Heavy-duty AEO Vehicle Categories to MOVES Regulatory Classes
AEO Freight Sales
Categories
MOVES regClassID
Light Medium
42
Medium
46
Heavy
47
We grew the model year 2022 vehicle populations using the year-over-year changes by fuel
type and regulatory class, applying the growth uniformly across all relevant source types. Since
AEO2023 only projects out to 2050, populations for years 2051-2060 were assumed to continue
to grow at the same growth rate as between 2049 and 2050.
We calculated fuel type, regulatory class, and engine technology distributions from the grown
populations. This formed the starting point for the electric vehicle adjustments described in the
next section.
5.3.2.2. Advanced Clean Trucks and HD GHG Phase 3
ACT imposes minimum HD ZEV sales requirements beginning in MY 2024 in California and the
states that have adopted the program under CAA section 177. EPA granted the waiver of
preemption for California's ACT rule under CAA section 209(b) on March 30, 2023. At the time
47
-------
of our analysis for M0VES5, ACT has been adopted by 10 other states, beginning in various
model years. The states in which we model ACT adoption, referred to collectively as ACT states,
are shown in Table 5-5 with the model year in which their adoption of ACT begins.
Table 5-5 States that have adopted ACT
State
Beginning MY
California37
2024
Massachusetts38
2025
New Jersey39
2025
New York40
2025
Oregon41
2025
Washington42
2025
Vermont43
2026
Colorado44
2027
Maryland45
2027
New Mexico46
2027
Rhode Island47
2027
The ACT rule enforces minimum ZEV sales separately for HD vocational vehicles and tractorsf
and compliance is determined by the balance of ACT deficits and credits within each vehicle
group. An ACT deficit is incurred with the sale of every HD vehicle while ACT credits are
generated for the sale of HD ZEVs.g Each vehicle incurs a given number of deficits (and
generates credits, if the vehicle is a ZEV) based on its weight class modifier as outlined in the
ACT rule. Table 5-6 maps each MOVES source type and regulatory class to the appropriate ACT
vehicle group and weight class modifier.
Table 5-(
5 Mapping MOVES vehicle types to ACT vehic
e groups and weight classes
ACT Vehicle
Group
Weight Class
MOVES
regClasslD(s)
MOVES
sourceTypelD(s)
ACT Weight
Class
Modifier
Vocational
Light Heavy-Duty
42
41, 42, 43
51, 52, 53, 54
1.0
Medium Heavy-Duty
46
41, 42, 43
51, 52, 53, 54
1.5
Heavy Heavy-Duty
47, 48
41, 42, 43
2.0
f ACT also has a ZEV sales mandate for medium-duty (Class 2b and 3) vehicles. We do not present or discuss them
in this section because the impact of ACT on medium-duty ZEV adoption is captured by OMEGA (see Section 5.3.1).
g Manufacturers can also incur credits for near-zero emissions vehicles (i.e., plug-in hybrid electric vehicles). NZEVs
incur fewer credits than ZEVs do, and we do not model NZEV adoption as part of our modeled compliance pathway
for ACT in MOVES.
48
-------
ACT Vehicle
Group
Weight Class
MOVES
regClasslD(s)
MOVES
sourceTypelD(s)
ACT Weight
Class
Modifier
51, 52, 53, 54
Tractor
Medium Heavy-Duty
46
61, 62
2.5
Heavy Heavy-Duty
47
61, 62
2.5
Table 5-7 shows the ZEV sales percentages mandated by model year in ACT or, to be more
precise, the percentage of ACT credits, relative to ACT deficits, that must be incurred to comply
with ACT, which does not necessitate the exact sales percentages shown.
Table 5-7 Mandated ACT ZEV credit percentages by model year
Model Year
ACT Vehicle Group Mandated ZEV Sales Percentage
Vocational
Tractor
2024
9%
5%
2025
11%
7%
2026
13%
10%
2027
20%
15%
2028
30%
20%
2029
40%
25%
2030
50%
30%
2031
55%
35%
2032
60%
40%
2033
65%
40%
2034
70%
40%
2035 and beyond
75%
40%
To project ZEV adoption by source type, regulatory class, and model year, we relied on heavy-
duty vehicle modeling in EPA's Heavy-Duty Technology Resource Use Case Scenario Tool (HD
TRUCS). HD TRUCS was EPA's technology assessment tool for the final HD Phase 3 standards. It
was peer reviewed48 in 2023 and is more fully discussed in the HD GHG Phase 3 Regulatory
Impact Analysis Chapter 2.49 For our analysis, we used the FRM version of HD TRUCS available
in the HD Phase 3 docket.50
HD TRUCS evaluates the energy and power demands of 101 representative HD vehicles. The
representative vehicles cover many aspects of work performed by vehicles in the heavy-duty
sector. This work was translated into total energy and power demands by vehicle type based on
everyday use of HD vehicles. HD TRUCS then identifies the technical properties and costs
required for electric vehicles (BEV, FCEV, or plug-in hybrids) to meet the operational needs of a
comparable diesel vehicle. Based on these assessments, HD TRUCS projects the extent to which
specific HD vehicle types may electrify in the future.
49
-------
Using the HD TRUCS Output Calculator,50 we calculated sales-weighted average ZEV adoption
rates by source type and regulatory class from the projected rates for the 101 vehicle types.
This was done in the same way as was done to calculate the HD Phase 3 standards, but for
different vehicle groupings. We calculated the standards by averaging across regulatory
subcategories, but for this update to MOVES, we averaged across combinations of source type
and regulatory class.
We refer to the resulting adoption rates, shown in Table 5-8, as the baseline adoption rates.
These baseline adoption rates reflect the relative suitability of ZEV technology for different
vehicle types. For example, ZEV technologies are generally more suitable to school buses than
combination long-haul trucks and are generally more suitable to LHD vehicles than MHD
vehicles. While the baseline adoption rates represent the ZEV adoption we expect to be driven
by the HD GHG Phase 3 rule, they do not match the ZEV sales percentages mandated by ACT.
The ACT sales mandates are defined for two broad HD vehicle groups, Class 4-8 vocational
vehicles and Class 7-8 tractors, as shown in Table 5-7. We used the baseline adoption rates as
the starting point for modeling how ACT would impact ZEV sales by MOVES source type and
regulatory class.
Table 5-8 ZEV baseline adoption rates by MOVES vehicle type and model year
MOVES Vehicle Type
Model Year Adoption Rates
Source Type
Reg Class
2027 and earlier*
2028
2029
2030
2031
2032 and beyond
Other Bus (41)
LHD (42)
20%
25%
30%
35%
52%
68%
MHD (46)
14%
13%
13%
12%
13%
14%
HHD (47)
0%
0%
0%
6%
10%
13%
Transit Bus
(42)
LHD (42)
20%
25%
30%
35%
52%
68%
MHD (46)
10%
10%
10%
11%
13%
16%
HHD (47)
0%
0%
16%
18%
28%
37%
School Bus
(43)
LHD (42)
20%
25%
30%
35%
52%
68%
MHD (46)
19%
24%
28%
33%
51%
69%
HHD (47)
0%
0%
13%
14%
20%
25%
Refuse Truck
(51)
MHD (46)
14%
15%
16%
18%
22%
26%
HHD (47)
0%
0%
16%
18%
16%
13%
Single Unit
Short-Haul
Truck (52)
LHD (42)
19%
24%
30%
35%
51%
67%
MHD (46)
12%
15%
18%
21%
28%
36%
HHD (47)
0%
0%
14%
16%
25%
32%
Single Unit
Long-Haul
Truck (53)
LHD (42)
10%
15%
20%
24%
33%
41%
MHD (46)
14%
17%
20%
23%
31%
39%
HHD (47)
0%
0%
16%
18%
28%
37%
Combination
Short-Haul
Truck (61)
MHD (46)
0%
13%
17%
12%
21%
29%
HHD (47)
0%
6%
10%
18%
31%
45%
Combination
Long-Haul
Truck (62)
MHD (46)
0%
0%
0%
6%
12%
25%
HHD (47)
0%
0%
0%
6%
12%
25%
* ACT's phase-in begins in MY 2024, but HD TRUCS modeling begins in MY 2027, corresponding with the
phase-in for Phase 3. For calculating ACT-driven adoption in model years 2024 through 2026, we copied
the baseline adoption rates in MY 2027.
50
-------
To calculate the ZEV adoption needed to comply with ACT, we first calculated the number of
ACT deficits incurred across all ACT states by model year. The deficit calculation appears in
Equation 5-4:
DeficitSAcj
-------
Once we calculate national ZEV adoption, we then divide that adoption rate into BEV and FCEV
adoption rates consistent with the HD Phase 3 technology assessment and HD TRUCS, which
evaluates FCEV adoption for certain coach buses and long-range day cab and sleeper cab
tractors, but otherwise evaluates BEV adoption.
In accounting for CARB and EPA rules that we expect to drive HD electrification, we model an
increased proportion of HD BEV and FCEV sales while keeping the total number of HD sales
constant, which in turn, requires a reduction in ICE vehicle sales. In reducing ICE vehicle sales,
we assume that no ICE fuel type is preferentially displaced by ZEVs. In other words, we
decrease the sales percentage of all ICE vehicle types by the same proportion to account for
ZEV adoption, rather than decreasing a particular fuel type more than any other.
5.4. Summary Results
This section presents summary result graphs of the analysis described above. Figure 5-2, Figure
5-3, Figure 5-4, and Figure 5-5 show fuel type and engine technology distributions by source
type. Figure 5-6, Figure 5-7, Figure 5-8, and Figure 5-9 show regulatory class distributions by
source type.
52
-------
Figure 5-2 Summary fuel type distributions for light-duty Figure 5-3 Summary fuel type distributions for buses
source types
Fuel Type
— Gas
Diesel
CNG
BEV
FCEV
100%-
100%-
0% -
100%-
c
o
'¦g 75% -
2
it
® 50% -
Q.
>.
t-
® 25% -
0% -
100%-
Fuel Type
Gas
— Diesel
¦— E-85
BEV
100%-
c
o
75% -
CO
LL
0
50% -
£1,
1-
0
25% -
LL
100%-
75% -
50% -
25% -
75% -
50% -
25% -
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060
Model Year
Model Year
75% -
50% -
25% -
o% -
75% -
50% -
25% -
53
-------
100%"
75% -
50% -
25% -
0% -
100%-
75% -
50% -
c
o
25% -
CO
LL
0
100%-
1-
"q5
75% -
_3
LL
50% -
25% -
0% -
100%-
75% -
25% -
n
Fuel
Type
Gas
Diesel
CNG
BEV
FCEV
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060
Model Year
Figure 5-4 Summary fuel type distributions for single unit
trucks
100%-
25% -
u_
CD
.>• 100%"
1950 1960 1970 19
I 2000 2010 2020 2030 2040 2050 2060
Model Year
Fuel Type
— Gas
— Diesel
— CNG
BEV
FCEV
Figure 5-5 Summary fuel type distributions for
combination trucks
54
-------
25% -
~0
0)
c/>
a>
a>
H
c
o
Regulatory Class
— LDT
LHD2b3
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060
Model Year
Figure 5-6 Summary regulatory class distributions for
light-duty trucks
55
25% -
Regulatory Class
— LHD2b3
— LHD45
— MHD67
HHD8
Urban Bus
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060
Model Year
Figure 5-7 Summary regulatory class distributions for
buses
-------
100%-
75% -
50% -
25% -
0% -
100%"
75% -
o
=3
50% -
-Q
)
25% -
Q
CO
...
cn
TO
O
100%"
>>
o
75% -
CO
50% -
(I)
rr
25% -
0% -
100%-
75% -
25% -
Regulatory Class
— LHD2b3
"Nv"Va
I
CO
CO
MHD67
flA l K A /"
C
HHD8
o
23
=r
O)
cz
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060
Model Year
Figure 5-8 Summary regulatory class distributions for
single unit trucks
56
100%"
25% -
co 0% -
o 100%"
o
JS
3
a? 75%-
Regulatory Class
HHD8
Gliders
1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060
Model Year
Figure 5-9 Summary regulatory class distributions for
combination trucks
-------
6. Vehicle Age-Related Characteristics
Age is an important factor in calculating vehicle emission inventories. MOVES employs a
number of different age dependent factors, including deterioration of engine and emission
after-treatment technology due to tampering and mal-maintenance, vehicle scrappage and
fleet turnover and mileage accumulation over the lifetime of the vehicle. Deterioration effects
are detailed in the MOVES reports on the development of light-duty and heavy-duty emission
rates.1718 This section describes vehicle age distributions and relative mileage accumulation
rates by source type.
6.1. Age Distributions
Vehicle age is defined in MOVES as the difference between a vehicle's model year and the year
of analysis. Age distributions in MOVES vary by source type and range from 0 to 40+ years, so
that all vehicles 40 years and older are modeled together. Therefore, an age distribution is
comprised of 41 fractions, where each fraction represents the number of vehicles present at a
certain age divided by the vehicle population for all ages. Since sales and scrappage rates are
not constant, these distributions vary by calendar year. Ideally, all historic age distributions
could be derived from registration data sources. However, acquiring such data is prohibitively
costly, so MOVES only contains registration-based age distributions for two analysis years: 1990
and 2023. The age distributions for all other analysis years in MOVES were projected forwards
or backwards from the 2023 base age distribution. All default age distributions are available in
the SourceTypeAgeDistribution table in MOVES database.
The derivation of the base 2023 age distribution and the forwards and backwards projections
for all years other than 1990 are detailed in the following subsections. The last subsection
describes the algorithm used in the MOVES5 input database conversion tools. The 1990 age
distributions are discussed in Appendix B.
6.1.1. Base Age Distributions
The 2023 base age distributions for cars and trucks were primarily derived from vehicle
registration data22 pulled in 2023 and the National Transit Database (NTD).27The registration
data had vehicle counts by age for motorcycles (11), passenger cars (21), passenger trucks (31),
light commercial trucks (32), school buses (43), refuse trucks (51), motor homes (54),
combination short-haul trucks (61) and combination long-haul trucks (62), as well as other
single unit trucks and non-school buses. The age distribution for the other single unit trucks was
applied to both short-haul (52) and long-haul (53) single unit trucks and the age distribution for
non-school buses was applied to the other bus source type (41). Transit bus (42) age
distributions were calculated from the NTD active fleet vehicles using the definition of a transit
bus in Section 2.2.4.
Since the age distributions in MOVES represent the full calendar year, additional calculations
were necessary for determining the fraction of age 0 vehicles in the fleet because the 2023
vehicle registration data did not capture all vehicles sold in 2023. Vehicle sales by source type in
2023 were calculated from a variety of sources as described in Appendix C.2. The source type
57
-------
sales were divided by the 2023 source type populations (see Section 4.1) to determine the age
0 fractions. The other fractions for ages 1-40 were renormalized so that each source type's age
distribution summed to 1. This was done instead of directly using the sales numbers to
calculate the age distributions (i.e., using the sales values as age 0 counts) because the vehicle
registration data were used in MOVES only to determine vehicle distributions, not for vehicle
populations.
Figure 6-1 shows the fraction of vehicles by age and source type for calendar year 2023, which
formed the basis for forecasting and back-casting age distributions as described in the following
sections. Please note that since all vehicles age 40 and older are grouped together, there is an
uptick in this age bin for some source types.
58
-------
10
10
10
10
20
30
20
30
20
30
20
Vehicle Age
30
40
40
40
40
Source Type
Motorcycle
Passenger Car
— Passenger Truck
— Light Com Truck
Source Type
Other Buses
— Transit Bus
School Bus
Source Type
Refuse Truck
— Short- and Long-haul
— Motor Home
Source Type
Combination Short-haul
— Combination Long-haul
Figure 6-1 Calendar year 2023 age distributions by source type
59
-------
6.1.2. Historic Age Distributions
The 2000-2022 age distributions were backcast from the 2023 base age distribution using
historic population and sales estimates. Age distributions are calculated from population counts
if the populations are known by age:
In Equation 6-1, fay is the age fraction, pa is the population of vehicles at age a and Py is the
total population in calendar year y. In this section, arrow notation will be used if the operations
are to be performed for all ages. For example, fy is used to represent all age fractions in
calendar year y. Another example is Py; it represents an array of pa values at each permissible
age in calendar year y. In contrast, Py represents the total population in year y.
Intuitively, backcasting an age distribution one year involves removing the new vehicles sold in
the base year and adding the vehicles scrapped in the previous year, as shown in Equation 6-2:
where Py-i is the population (known at each age) of the previous year, Py is the population in
the base year, Ny is new vehicles sold in the base year and Ry_ 1 is the population of vehicles
removed in the previous year. Please note that the sales term only includes new vehicles at age
0. This can be represented algorithmically as follows:
1. Calculate the base population distribution (Py) by multiplying the base age distribution
(fy) and base population (Py).
2. Remove the age 0 vehicles (Ny).
3. Decrease the population age index by one (for example, 3-year-old vehicles are
reclassified as 2-year-old vehicles).
4. Add the vehicles that were removed in the previous year (Ry_1).
5. Convert the resulting population distribution into an age distribution using Equation 6-1.
6. Replace the new age 39 and 40+ fractions with the base year age 39 and 40+ fractions
and renormalize the new age distribution to sum to 1 while retaining the original age 39
and 40+ fractions.
7. This results in the previous year age distribution (fy_x). If this algorithm is to be
repeated, fy_1 becomes fy for the next iteration.
The fraction of age 40+ vehicles is kept constant because some source types have a fraction in
this age bin in the base age distributions. If left unconstrained, the algorithm can either grow
this age bin unreasonably large or shrink it unreasonably small, depending on the source type.
This indicates that the base survival rates for the oldest age bins may be inappropriate.
Equation 6-1
Py-1 — Py ~ Ny + Ry-!
Equation 6-2
60
-------
However, lacking better data, we decided to keep the age 40+ bin at a constant fraction for all
historic age distributions.
Age 39 is additionally retained because when the number of scrapped vehicles is calculated, a
large proportion of them come from the age 40 bin. In reality, these scrapped vehicles have a
distribution well beyond age 40, but they are all grouped together in this analysis. When the
scrapped vehicles are added to the index-shifted population distribution, this results in a large
addition to the age 39 bin. To prevent this from happening, the base year age 39 fractions are
also retained in each backcasted year.
Please see Appendix C, Detailed Derivation of Forecast and Backcast Age Distributions, for
more information on how this algorithm was applied to derive the historic national default age
distributions in MOVES.
6.1.3. Projected Age Distributions
The method used to forecast the 2024-2060 age distributions from the 2023 distribution is
similar to the backcasting method described above. To forecast an age distribution one year,
Equation 6-2 of the previous section can be rewritten as Equation 6-3:
Py+l — Py Ry + Ny+1 Equation 6-3
Essentially, this is done by taking the base year's population distribution, removing the vehicles
scrapped in the base year and adding the new vehicles sold in the next year. This can be
represented algorithmically as follows:
1. Calculate the base population distribution (Py) by multiplying the base age distribution
(fy) and base population (Py).
2. Remove the vehicles that did not survive (Ry) at each age level.
3. Increase the population age index by one (for example, 3-year-old vehicles are
reclassified as 4-year-old vehicles).
4. Add new vehicle sales (Ny+1) as the age 0 cohort.
5. Convert the resulting population distribution into an age distribution using Equation 6-1.
6. Replace the new age 40+ fraction with the base year age 40+ fraction and renormalize
the new age distribution to sum to 1 while retaining the original age 0 and age 40+
fractions.
7. This results in the next year age distribution (fy+1). If this algorithm is to be repeated,
fy+1 becomes fy for the next iteration.
The fraction of age 40+ vehicles is kept constant in the projection algorithm for the same
reasons given for the backcasting algorithm. However, there is no issue with an artificially
growing population of age 39 vehicles when projecting forward. Therefore, the age 39 bin is
calculated as the others are instead of being retained from the base age distribution.
61
-------
Please see Appendix C, Detailed Derivation of Forecast and Backcast Age Distributions, for more
information on how this algorithm was applied to derive the historic national default age
distributions in MOVES.
In addition to producing the default projected age distributions, this algorithm was
implemented in the spreadsheet-based Age Distribution Projection Tool.h This tool can be used
to project future local age distributions from user-supplied baseline distributions, provided that
the baseline year is 2011 or later. This requirement ensures that the 2008-2009 recession is
fully accounted for in the baseline. The sales rates and scrappage assumptions are the same in
the tool as they are in the national default. This is because local projections of sales and
scrappage are generally unavailable and the national trends are the best available data. Thus,
projections made with the tool tend to converge with the national age distributions for far
future years.
6.1.4. Converting 30-Year Age Distributions to 40-Year Distributions
The MOVES4 to MOVES5 database conversion tool (as well as the MOVES3 to MOVES5 version)
allows users to convert an input database created with an older version of MOVES to be
compatible with the latest version. Since older versions of MOVES used 31-year age
distributions, the conversion tool needs to be able to transform an age distribution that covers
ages 0-30+ into a reasonable age distribution covering ages 0-40+.
In earlier versions of MOVES, the age 30 category included all vehicles ages 30 and older. This
resulted in most age distributions having a "tail", or a larger proportion of vehicles in age 30
than age 29. Since the definition of ages 0-29 did not change, the fractions associated with
those ages are not changed by the converter tool. However, since MOVES5 requires individual
age fractions for ages 30-39, as well as a grouped age 40+ category, the converter tool needs to
allocate the input age 30+ fraction across ages 30-39 and to put the remainder in 40+.
To estimate fractions for individual ages from the grouped 30+ age fraction, we first calculated
the national average maximum age by source type across all counties in the US using vehicle
registration data22 pulled in 2023, as shown in Figure 6-2. The error bars represent two times
the standard deviation of the average, which indicate that heavy-duty vehicles in particular
have significant variability in maximum vehicle age across US counties.
h The Age Distribution Projection Tool for MOVES is a macro-enabled spreadsheet and is available for download at:
https://www.epa.gov/moves/tools-develop-or-convert-moves-inputs
62
-------
Figure 6-2 National average maximum vehicle age by source type
When the converter tool is run, it operates on a single county input database. For simplicity, we
assume the maximum age for that county is equal to the national average maximum vehicle
age by source type.1 To estimate fractions for individual ages from the grouped 30+ age
fraction, we calculated a linear trend for each source type subject to the following constraints:
Zu —
n =
a=MaxAge
fa ~ /30+
a=30
Equation 6-4
ffrlaxAge+l ~ 0
Equation 6-5
' From Figure 6-2, note that refuse trucks have an average vehicle age of approximately 26 years. This means that
on average, refuse trucks do not have age fractions in the 30+ category. However, for the converter tool to work
for counties that do have refuse trucks in this category, we need to make an assumption as to what the actual
maximum age is in that county. For this purpose, in the rest of this analysis, we treat the average maximum age of
refuse trucks to be the average age plus one standard deviation, or 35 years old.
63
-------
In the equations above, fa represents the new age fraction at age a, and /30+ represents the
original age fraction in the age 30+ category (i.e., the "tail" from the input age distribution).
Essentially, the new age fractions for ages 30 through the maximum age must sum to the
original age 30+ fraction, and the age fraction beyond the maximum age is 0.
Since we are assuming a linear trend, we can solve the summation presented in Equation 6-4
using the constraint in Equation 6-5 and the formula for the geometric area of a triangle to
calculate the new age 30 fraction, /30, as shown in Equation 6-6 below. Linearly interpolating
between /30 and fMaxAge+i allows us to calculate fa for the intermining ages, as shown in
Equation 6-7.
foa = ^30+ X ^ Equation 6-6
J30 MaxAge + 2 — 30
fa = ho X 777T777TTTTT) Equation 6-7
a — 30 >
^30 — (MaxAge + 1)>
Finally, since MOVES groups together all ages over 40 into the 40+ category, we use Equation
6-8:
Za=MaxAge
fa Equation 6-8
a=40
The following figure shows an example input age distribution developed for MOVES4 (using
ages 0-30+) and the resulting output from the database conversion tool (using ages 0-40+):
64
-------
6%-
c
o
4%-
-------
commercial trucks are defined to have a RMAR of one (1.0)j and new passenger cars have a
RMAR of 0.95. This means that when MOVES allocates the VMT assigned to the light-duty
vehicle HPMS class to passenger cars, passenger trucks and light commercial trucks, a
passenger car of age 0 will be assigned only 95 percent of the annual VMT assigned to a
passenger truck or light commercial truck of age 0. The RMAR values are shown in Figure 6-4.
J Within each HPMS vehicle class, an RMAR value of one is assigned to the source type and age with the highest
annual VMT accumulation. Because we use the same mileage accumulation data for passenger trucks and light
commercial trucks, they both have a value of one.
66
-------
10
20
30
40
10
20
30
40
Source Type
— Motorcycle
Source Type
— Passenger Car
— Pass, and Light Com Trucks
Source Type
— Transit and Other Buses
— School Bus
Source Type
Refuse Truck
— Single Unit Short-haul
Single Unit Long-haul
-- Motor Home
Source Type
Combination Short-haul
— Combination Long-haul
1.00-
0.75-
0.50-
0.25-
0.00-,
Figure
0 10 20 30 40
Vehicle Age
6-4. Relative Mileage Accumlation Rates (RMAR)
by HPMS Class and SourceTypelD
The derivation of the RMAR values for each sourcetype and HPMS class are discussed in the
following subsections.
67
-------
6.2.1. Motorcycles
The RMAR values were calculated from mileage accumulations for motorcycles (sourceTypelD
11) based on the model years and odometer readings listed in motorcycle advertisements. A
stratified sample of about 1,500 ads were examined. A modified Weibull curve was fit to the
data to develop the relative mileage accumulation rates used in MOVES.51
6.2.2. Passenger Cars, Passenger Trucks and Light-Commercial Trucks
The RMAR values for passenger cars, passenger trucks and light commercial trucks
(sourcetypelD 21, 31, and 32) are based on an analysis by NHTSA of vehicle miles traveled,
performed for their CAFE standards for MY 2024-2026 and presented in their Technical Support
Documentation.52 NHTSA used a random national sample of one million vehicles based on
vehicle registration data pulled in 2016, which provides a longitudinal dataset with information
on individual vehicles over time, providing data on vehicle mileage as they age. The smoothed
annual mileage schedules used by NHTSA are presented in Table 6-1 below. Since MOVES
models vehicles up to age 40, and the NHSTA data included through only age 39, we
extrapolated vehicle mileage to age 40 based on the selected fits described by NHTSA in their
documentation.
68
-------
Table 6-1VMT Annual Mileage Schedules Derived by NHTSA and Weights Used to Generate LDT
Mileage for MOVES
Vehicle Age
Cars
Vans/SUVs
Pickups
Weight 31
(Vans/SUVs)
Weight 32
(Pickups)
0
15,922
16,234
18,964
0.831
0.169
1
15,379
15,805
17,986
0.831
0.169
2
14,864
15,383
17,076
0.884
0.116
3
14,378
14,966
16,231
0.896
0.104
4
13,917
14,557
15,449
0.900
0.100
5
13,481
14,153
14,726
0.905
0.095
6
13,068
13,756
14,060
0.917
0.083
7
12,677
13,366
13,448
0.924
0.076
8
12,305
12,982
12,886
0.918
0.082
9
11,952
12,605
12,372
0.915
0.085
10
11,615
12,234
11,903
0.930
0.070
11
11,294
11,870
11,476
0.928
0.072
12
10,986
11,512
11,088
0.940
0.060
13
10,690
11,161
10,737
0.945
0.055
14
10,405
10,816
10,418
0.945
0.055
15
10,129
10,477
10,131
0.947
0.053
16
9,860
10,146
9,871
0.944
0.056
17
9,597
9,820
9,635
0.942
0.058
18
9,338
9,501
9,421
0.943
0.057
19
9,081
9,189
9,226
0.946
0.054
20
8,826
8,883
9,047
0.944
0.056
21
8,570
8,583
8,882
0.946
0.054
22
8,313
8,290
8,726
0.946
0.054
23
8,051
8,004
8,577
0.950
0.050
24
7,785
7,724
8,433
0.957
0.043
25
7,511
7,450
8,290
0.962
0.038
26
7,229
7,183
8,146
0.964
0.036
27
6,938
6,923
7,998
0.965
0.035
28
6,635
6,669
7,842
0.968
0.032
29
6,319
6,421
7,676
0.971
0.029
30
5,988
6,180
7,497
0.984
0.016
31
5,641
5,946
7,302
0.977
0.023
32
5,277
5,718
7,089
0.980
0.020
33
4,893
5,496
6,853
0.982
0.018
34
4,488
5,281
6,593
0.998
0.002
35
4,061
5,072
6,305
0.990
0.010
36
3,610
4,870
5,987
0.994
0.006
37
3,133
4,674
5,635
0.999
0.001
38
2,629
4,485
5,248
0.999
0.001
39
2,096
4,303
4,821
0.987
0.013
40
1,535
4,126
4,080
0.999
0.001
69
-------
Passenger cars, passenger trucks and light commercial trucks were considered together as
light-duty vehicles (HPMSVTypelD 25).
We chose to use the same RMAR for all light trucks rather than deriving individual mileage
accumulation rates for sourceTypelD 31 and 32 because the NHTSA results did not distinguish
passenger and commercial trucks. However, rather than computing a simple arithmetic
average of the annual mileage for the SUV/Van and Pickup categories, we weighted these
values using factors from a separate vehicle registration data sample obtained by EPA for 2017
in five states: Colorado, Georgia, New Jersey, California, and Illinois. We originally purchased
this sample to update the RMAR for light-duty vehicles but given the availability of a national
sample with similar longitudinal characteristics, we opted to use the NHTSA mileage schedules
as our main data source. EPA's five-state sample did, however, classify vehicles to MOVES
sourcetypes based on body shape, GVWR and personal vs commercial registration type. We
calculated weights for SUV/Vans and Pickups based on the five-state samples for sourceTypelDs
31 and 32, respectively, and applied them to the NHTSA schedules (see Table 6-1), assuming
that the SUV/Van category loosely mapped to passenger trucks while the Pickup category
mapped to light commercial trucks and assuming no significant difference between the
populations in 2016 and 2017.
Since the trucks had a higher MAR than passenger cars, each source type's mileage by age was
divided by truck mileage at age 0 to determine a relative MAR. Analysis of the NHTSA car data
determined that new passenger cars (age 0) accumulate 95 percent of the annual miles
accumulated by new light-duty trucks.
In summary, MOVES passenger car mileage came from NHTSA car mileage, and the light-duty
truck mileage was estimated by multiplying the SUV/Vans mileage by the 31 weighting value
and adding the Pickup mileage multiplied by the 32 weighting value. Then each source type's
mileage by age was divided by truck mileage at age 0 to determine a relative MAR.
6.2.3. Buses
The transit bus (sourceTypelD 42) annual mileage accumulation rates are taken from the
MOBILE6 values for diesel transit buses (HDDBT). This mileage data was obtained from the 1994
Federal Transportation Administration survey of transit agencies as shown in Table 6-3 and a
smoothing function applied to remove the variability in the data.53 The MOBILE6 results were
linearly extrapolated to calculate values for ages 26 through 40.
Since other buses (sourceTypelD 41) are similar to transit buses, we assigned them the same
RMAR as transit buses (sourceTypelD 42).
The school bus (sourceTypelD 43) annual mileage accumulation rate (9,939 miles per year) is
derived from the 1997 School Bus Fleet Fact Book.26 Specifically, we set the RMAR for school
buses to have the same shape as the transit bus RMAR, but adjusted down such that year 0 is
based on the 9,939 miles per year from the School Bus Fleet Fact Book. The same relative shape
is evident in of the Bus RMAR in Figure 6-4.
70
-------
Table 6-2 Annual mileage accumulation of transit buses from 1994 Federal Transit
Administration data
Age
Miles
Age
Miles
Age
Miles
1
*
11
32,540
21
19,588
2
*
12
32,605
22
22,939
3
46,791
13
27,722
23
26,413
4
41,262
14
28,429
24
23,366
5
42,206
15
32,140
25
11,259
6
39,160
16
28,100
26
23,228
7
38,266
17
24,626
27
21,515
8
36,358
18
23,428
28
25,939
9
34,935
19
22,575
29
20,117
10
33,021
20
23,220
30
17,515
* Insufficient data
6.2.4. Other Heavy-Duty Vehicles
The RMAR values for source types 51 (refuse trucks), 52 (short-haul single unit trucks), 53 (long-
haul single unit trucks), 61 (short-haul combination trucks) and 62 (long-haul combination
trucks) use the data from the 2002 Vehicle Inventory and Use Survey (VIUS).23 The total
reported annual miles traveled by truck in each source type by age, as shown in Table 6-3, was
divided by the vehicle population by age to determine the average annual miles traveled per
truck by source type.
71
-------
Table 6-3 VIUS2002 annual mileage by vehicle age
Age
Model
Year
Single Unit Trucks
Combination Trucks
Refuse (51)
Short-Haul
(52)
Long-Haul
(53)
Short-Haul
(61)
Long-Haul
(62)
0
2002
26,703
21,926
40,538
60,654
109,418
1
2001
32,391
22,755
28,168
59,790
128,287
2
2000
31,210
24,446
30,139
61,651
117,945
3
1999
31,444
23,874
49,428
62,865
110,713
4
1998
31,815
21,074
33,266
55,113
99,925
5
1997
28,450
21,444
23,784
54,263
94,326
6
1996
25,462
16,901
21,238
40,678
85,225
7
1995
30,182
15,453
27,562
38,797
85,406
8
1994
20,722
13,930
21,052
33,485
71,834
9
1993
25,199
13,303
11,273
30,072
71,160
10
1992
23,366
11,749
18,599
27,496
67,760
11
1991
18,818
13,675
15,140
24,175
80,207
12
1990
12,533
11,332
13,311
22,126
48,562
13
1989
15,891
9,795
9,796
21,225
64,473
14
1988
19,618
9,309
12,067
21,163
48,242
15
1987
12,480
9,379
16,606
20,772
58,951
16
1986
12,577
4,830
8,941
11,814
35,897
0-3
1999-2002
Average
30,437
23,250
37,069
61,240
116,591
For each source type, in the first few years, the data showed only small differences in the
annual miles pervehicle and no trend. Afterthat, the average annual miles pervehicle declined
in a fairly linear manner, at least until the vehicles reach age 16 (the limit of the data). MOVES,
however, requires mileage accumulation rates for all ages up to age 40.
Mileage accumulation rates for these vehicles were determined for each age from 0 to 40 using
the following method:
1) Ages 0 through 3 use the same average annual mileage accumulation rate for age 0-
3 vehicles of that source type.
2) Ages 4 through 16 use mileage accumulation rates calculated using a linear
regression of the VIUS data. The average mileage accumulation rate of ages 0 to 3
were used for age 3 in the regression. The resulting coefficients are summarized in
Table 6-4.
3) Age 30 uses the relative mileage accumulation rate derived from the 1992 Truck
Inventory and Use Survey (TIUS) as documented in the ARCADIS report.53 These
relative mileage accumulation rates were allocated to the MOVES source types from
the MOBILE6 mileage accumulation rates, they were converted to mileage based on
72
-------
the mileage data used in MOVES, then converted back to an RMAR consistent with
the other ages.
4) Ages 17 through 29 use values from interpolation between the values in age 16 and
age 30.
5) Ages 31 through 40 use a linear extrapolation beyond age 30. However, when this
would result in RMAR values less than 0.01, we set the RMAR value equal to 0.01.
This is because we assume the oldest heavy-duty vehicles are still driven, just
proportionally much less than newer vehicles.
Table 6-4 Regression coefficients for heavy-duty truck average annual mileage accumulation
rates (ages 4-16)
Measurement
Refuse
Truck(51)
Single Unit
Short-Haul (52)
Single Unit
Long-Haul (53)
Combination
Short-Haul (61)
Combination
Long-Haul (62)
Average 0-3*
30,437
23,250
37,069
61,240
116,591
Intercept**
30,437
23,250
37,069
61,240
116,591
^ ¦ **
Slope
-1,361
-1,368
-2,476
-4,092
-6,418
Age 30 RMAR
0.027
0.0115
0.086
0.015
0.052
Average sample annual miles traveled for ages 0 through 3.
" Intercept at age 3; slope from ages 4 through 16.
The resulting relative mileage accumulation rates are shown in Table 6-5 below and Figure 6-4
above.
6.2.5. Motor Homes
Data from the 2017 National Household Travel Survey54 was used for the motor home RMAR
calculation. The calculation methodology is different from the other heavy-duty trucks. The
same average annual mileage accumulation rate was used for age 0-3 motor homes. Age 4
through 30 used mileage accumulation rates that were calculated using a linear regression of
the National Household Travel Survey data.
Based on this data, the average annual vehicle miles of travel per vehicle for age 0 to 3 is 6003.
In the regression analysis, this value was used as intercept at age 3. The slope from age 4
through 30 was calculated at -83 miles/year. Since the survey data did not include results for
motor homes ages 31-40, we linearly extrapolated beyond age 30. However, like we did for
other heavy-duty vehicles, when this would result in RMAR values less than 0.01, we set the
RMAR value equal to 0.01. This is because we assume the oldest motor homes are still driven,
just proportionally much less than newer ones.
The motor home mileage accumulation values were then converted to RMARs by dividing by
the average mileage for age 0-3 long-haul single unit trucks (37,069). The resulting relative
mileage accumulation rates of motor homes are shown in Table 6-5 below and Figure 6-4
above.
73
-------
Table 6-5 Relative mileage accumulation rates for heavy-duty trucks in
Age
Refuse
(51)
Short-Haul
Single Unit (52)
Long-Haul
Single Unit (53)
Motor Home
(54)
Short-Haul
Combination
(61)
Long-Haul
Combination
(62)
0
0.8211
0.6272
1.0000
0.1620
0.5253
1.0000
1
0.8211
0.6272
1.0000
0.1620
0.5253
1.0000
2
0.8211
0.6272
1.0000
0.1620
0.5253
1.0000
3
0.8211
0.6272
1.0000
0.1620
0.5253
1.0000
4
0.7844
0.5903
0.9332
0.1597
0.4902
0.9473
5
0.7477
0.5534
0.8664
0.1575
0.4551
0.8945
6
0.7110
0.5165
0.7996
0.1552
0.4200
0.8418
7
0.6743
0.4796
0.7328
0.1529
0.3849
0.7891
8
0.6376
0.4427
0.6660
0.1507
0.3498
0.7363
9
0.6009
0.4058
0.5992
0.1484
0.3147
0.6836
10
0.5642
0.3689
0.5323
0.1462
0.2796
0.6309
11
0.5275
0.3320
0.4655
0.1439
0.2445
0.5781
12
0.4908
0.2950
0.3987
0.1417
0.2094
0.5254
13
0.4541
0.2581
0.3319
0.1394
0.1743
0.4727
14
0.4174
0.2212
0.2651
0.1372
0.1392
0.4199
15
0.3807
0.1843
0.1983
0.1349
0.1041
0.3672
16
0.3440
0.1474
0.1315
0.1327
0.0690
0.3145
17
0.3214
0.1380
0.1282
0.1304
0.0652
0.2957
18
0.2987
0.1285
0.1249
0.1282
0.0613
0.2769
19
0.2761
0.1191
0.1216
0.1259
0.0575
0.2581
20
0.2535
0.1097
0.1184
0.1236
0.0536
0.2394
21
0.2309
0.1002
0.1151
0.1214
0.0498
0.2206
22
0.2083
0.0908
0.1118
0.1191
0.0460
0.2018
23
0.1857
0.0814
0.1085
0.1169
0.0421
0.1830
24
0.1631
0.0719
0.1052
0.1146
0.0383
0.1642
25
0.1405
0.0625
0.1019
0.1124
0.0344
0.1454
26
0.1179
0.0530
0.0986
0.1101
0.0306
0.1267
27
0.0953
0.0436
0.0954
0.1079
0.0267
0.1079
28
0.0727
0.0342
0.0921
0.1056
0.0229
0.0891
29
0.0500
0.0247
0.0888
0.1034
0.0191
0.0703
30
0.0274
0.0153
0.0855
0.1011
0.0152
0.0515
31
0.0100
0.0100
0.0822
0.0989
0.0114
0.0328
32
0.0100
0.0100
0.0789
0.0966
0.0100
0.0140
33
0.0100
0.0100
0.0757
0.0944
0.0100
0.0100
34
0.0100
0.0100
0.0724
0.0921
0.0100
0.0100
35
0.0100
0.0100
0.0691
0.0898
0.0100
0.0100
36
0.0100
0.0100
0.0658
0.0876
0.0100
0.0100
37
0.0100
0.0100
0.0625
0.0853
0.0100
0.0100
38
0.0100
0.0100
0.0592
0.0831
0.0100
0.0100
39
0.0100
0.0100
0.0559
0.0808
0.0100
0.0100
40
0.0100
0.0100
0.0527
0.0786
0.0100
0.0100
MOVES
74
-------
7. VMT Distribution of Source Type by Road Type
For each source type, the RoadTypeVMTFraction field in the RoadTypeDistribution table stores
the fraction of total VMT for each source type that is traveled on each of the MOVES five road
types nationally. Users may supply the VMT distribution by vehicle class for each road type for
individual counties when using County Scale. For National Scale, the default distribution is
allocated to individual counties using the SHOAIIocFactor found in the ZoneRoadType table.
The national default distribution of VMT to source type for each road type were derived to
reflect the VMT data included in the 2017 National Emission Inventory (NEI) Version 2.55 This
data is provided by states every three years as part of the NEI project and is supplemented by
EPA estimates based on data provided by FHWA Highway Statistics56 when state supplied
estimates are not available. The FHWA road types mapped to the MOVES roadTypelD values
(the eighth and ninth digits of the 10-digit onroad SCC) are shown below in Table 7-1.
Table 7-1 Mapping of FHWA road types to MOVES road types
FHWA Road Type
MOVES
roadTypelD
MOVES Road Type
Rural Interstate
2
Rural Restricted Access
Rural Other Freeways and Expressways
2
Rural Restricted Access
Rural Other Principal Arterial
3
Rural Unrestricted Access
Rural Minor Arterial
3
Rural Unrestricted Access
Rural Major Collector
3
Rural Unrestricted Access
Rural Minor Collector
3
Rural Unrestricted Access
Rural Local
3
Rural Unrestricted Access
Urban Interstate
4
Urban Restricted Access
Urban Other Freeways & Expressways
4
Urban Restricted Access
Urban Other Principal Arterial
5
Urban Unrestricted Access
Urban Minor Arterial
5
Urban Unrestricted Access
Urban Major Collector
5
Urban Unrestricted Access
Urban Minor Collector
5
Urban Unrestricted Access
Urban Local
5
Urban Unrestricted Access
The national distribution of road type VMT by source type is calculated from the NEI VMT
estimates and is summarized in Table 7-2. The off-network road type (roadTypelD 1) is
allocated no VMT.
Note that because it is difficult to distinguish single unit short-haul and long-haul trucks in
roadway VMT measurements, the distributions for single unit short-haul trucks are virtually the
same as those for single unit long-haul trucks.
75
-------
Table 7-2 M0VES5 road type distribution by source type
Road Type*
Source
Type
Description
Rural
Restricted
Rural
Unrestricted
Urban
Restricted
Urban
Unrestricted
2
3
4
5
All
11
Motorcycle
0.0825631
0.267313
0.198403
0.451721
1.000
21
Passenger Car
0.08177
0.204595
0.259544
0.454091
1.000
31
Passenger Truck
0.0958223
0.265213
0.222866
0.416098
1.000
32
Light Commercial Truck
0.0839972
0.217512
0.262385
0.436105
1.000
41
Other Bus
0.131819
0.246451
0.222309
0.399421
1.000
42
Transit Bus
0.122177
0.232623
0.259237
0.385963
1.000
43
School Bus
0.133622
0.290446
0.202762
0.37317
1.000
51
Refuse Truck
0.133744
0.281628
0.244409
0.340218
1.000
52
Single Unit Short-Haul Truck
0.133827
0.290565
0.233264
0.342345
1.000
53
Single Unit Long-Haul Truck
0.124627
0.288468
0.224945
0.36196
1.000
54
Motor Home
0.146173
0.297276
0.211836
0.344715
1.000
61
Combination Short-Haul Truck
0.172224
0.327849
0.244772
0.255155
1.000
62
Combination Long-Haul Truck
0.338174
0.240709
0.256685
0.164432
1.000
* RoadTypelD = 1 (Off Network) is assigned no VMT.
76
-------
8. Average Speed Distributions
Average speed is used in MOVES to convert VMT inputs into the source hours operating (SHO)
units that MOVES uses for internal calculations. It is also used to select appropriate driving
cycles, which are then used to calculate exhaust running operating mode distributions at the
national, county and sometimes project level. Instead of using a single average speed in these
tasks, MOVES uses a distribution of average speeds by bin. The AvgSpeedDistribution table lists
the default fraction of driving time for each source type, road type, day and hour in each
average speed bin. The fractions sum to one for each combination of source type, road type,
day and hour. The MOVES average speed bins are defined in Table 8-1.
Table 8-1 MOVES speed bin categories
Bin
Average Speed (mph)
Average Speed Range (mph)
1
2.5
speed < 2.5 mph
2
5
2.5 mph <= speed < 7.5 mph
3
10
7.5 mph <= speed < 12.5 mph
4
15
12.5 mph <= speed < 17.5 mph
5
20
17.5 mph <= speed < 22.5 mph
6
25
22.5 mph <= speed < 27.5 mph
7
30
27.5 mph <= speed < 32.5 mph
8
35
32.5 mph <= speed < 37.5 mph
9
40
37.5 mph <= speed < 42.5 mph
10
45
42.5 mph <= speed < 47.5 mph
11
50
47.5 mph <= speed < 52.5 mph
12
55
52.5 mph <= speed < 57.5 mph
13
60
57.5 mph <= speed < 62.5 mph
14
65
62.5 mph <= speed < 67.5 mph
15
70
67.5 mph <= speed < 72.5 mph
16
75
72.5 mph <= speed
As described below, the default average speed distributions for all sourcetypes were last
updated for MOVES3 using the telematics data.
8.1. Description of Telematics Dataset
In a study done by the Coordinating Research Council (CRC A-100),57 the GPS data collected by
StreetLight Data were used to develop inputs for the 2014 National Emissions Inventory. The
dataset consists of data from billions of trips derived from smart phone applications, in-
dashboard car navigation systems and commercial fleet management systems on vehicles
operating over a period of 12 consecutive months between September 2015 and August 2016
at a high temporal and spatial resolution.
The data included latitude, longitude and timestamps corresponding to the instantaneous
position that each vehicle sends to a central server. StreetLight overlays the coordinates on
their roadway network to determine distance traveled between consecutive points. From the
77
-------
distance and time between points, average speeds were calculated and further classified by
month, day of the week and hour. The dataset also was able to discriminate between personal
vehicles, medium-duty commercial trucks (Class 6 and lower) and heavy-duty commercial
trucks (Class 7 and 8). The personal dataset was available at high resolution (1 Hz) and low
resolution (one point every 10 or 30 seconds) while the commercial dataset was available at a
lower resolution with one point every 60 or 180 seconds. The data included a GIS shapefile
containing road information classified into the four MOVES road types and a second shapefile
containing county boundaries to generate data with the appropriate mapping.
Note that since the CRC A-100 project was developed to improve inputs used in the 2014 NEI,
the definitions of urban and rural applied to the CRC study were consistent with the
requirements of EPA's platform modeling for the NEI and regulatory impact analyses,58 which
follow the definitions established by the U.S. Census Bureau. This is inconsistent with the
urban-rural roadtype definitions used in MOVES, which follow those established by FHWA. The
main difference in the definitions established by the U.S. Census Bureau and FHWA is the
population threshold used to distinguish between urban and rural. The U.S. Census Bureau
defines an urban area as areas with a population of 2500 or more, whereas the FHWA defines
an urban area as areas with a population of 5000 or more. Therefore, telematics speed data
gathered by StreetLight Data in some areas that are considered rural by FHWA and MOVES may
have been assigned to "urban" roadtypes. For MOVES modeling purposes, this discrepancy
implies that the average speed distributions derived from this dataset could be biased high by
some degree, since vehicles on rural roads generally spend more time traveling at faster speeds
than those on urban roads.
Due to restrictions in time and resources, the final dataset consisted of only l/16th of the
information available to StreetLight Data. This aggregated subset totaled 250 million records
classified into 3 vehicle categories:
Personal Passenger vehicles
Medium-Duty commercial trucks (under 26,000 lbs of GVWR)
Heavy-Duty commercial trucks (over 26,000 lbs of GVWR)
The final dataset contains information for the three vehicle categories mentioned above across
3,109 counties in the mainland US. The dataset was classified into MOVES roadtypes and
MOVES speedbins, for 12 months of the year, seven days of the week and 24 hours of the day.
For further details, see the CRC A-100 report.57
A single set of default national average speed distributions for the MOVES default database
were developed using the national database which contains average speed distributions for
each county and hour of the day, for weekday/weekend, varying by road type and source type.
Additionally, we used activity (VMT and average speed by county, fuel, source type and road
type) from the beta version of the NEI collaborative 2016 modeling platform.59 The following
section describes the procedure to generate the average speed distributions included in
MOVES.
78
-------
8.2. Derivation of Default National Average Speed Distributions
The general steps for the derivation of default average speed distributions were:
1. Calculation of source hours operating (SHO) for each source type on each road type
aggregated over all counties to represent the entire U.S.
2. Calculation of average speed distributions for each hour of the day, day of the week,
road type and source type, weighted by the fraction of SHO in each county in reference
to the national SHO for a given source type and road type combination.
For the first step, we used county-specific annual VMT classified by fuel, source type and road
type as well as county-specific annual average speed values classified by source type and road
type. Both data files were used in the development of activity for the NEI collaborative 2016
beta modeling platform and are based on FHWA and CRC A-100 information (where available),
respectively. We calculated a county-specific annual value of source-hours operating (SHO) for
each source type - road type combination, as shown in Equation 8-1, by adding all the VMT
assigned to different fuels (/') for each source type (ST) - road type (RT) combination in each
county (Co) and dividing by the corresponding annual average speed:
. , „,m Hi=fuei Annual VMTSTiRTi Co miles ]
Annual SH0STRT Co = — - —- Equation 8-1
AyiyiuclI Avevcige SpccdsT rt co .ytiiIgs/fioiiv.
Then, we aggregate over all counties /' to obtain a national annual SHO for each source type (ST)
- road type (RT) combination following Equation 8-2:
National Annual SHOst RT = / Annual SHO(st RTv [hours] Equation 8-2
£—>i=Co
In the second step, we used a data file from the CRC A-100 project containing average speed
distributions by hour of the day and day typefor each source type - road type combination for
each county. These values were weighted togheter using the SHO for each county developed in
Equation 8-1 divided by the national annual SHO determined in Equation 8-2. This results in
average speed distributions (ASD) weighted by the national activity for a given source type -
road type combination for each hour (h) of each weekday/weekend (d). This is summarized in
Equation 8-3:
ASDhdSTRT =
ZAverageSpeedFractioni h d STRT Co x Annual SHOST RT Co Equation 8-3
i=16 National Annual SHOst rt
Note that the sum over all 16 speed bins should be equal to 1 for each hour and type of day for
a given source type and road type combination.
For the default national average speed distributions, we used the same mapping of telematics
data to MOVES source type used in the NEI to maintain consistency. For buses, refuse trucks,
79
-------
and motor homes for which no direct mapping was provided, we assigned the medium-duty
commercial profile. The final mapping is detailed in Table 8-2:
Table 8-2 Map of MOVES Source Types to telematics data vehicle type
SourceTypelD
Source Type Name
Telematics Vehicle Type
11
Motorcycle
Personal
21
Passenger Car
Personal
31
Passenger Truck
Personal
32
Light Commercial Truck
Medium-Duty Commercial
41
Other Bus
Medium-Duty Commercial
42
Transit Bus
Medium-Duty Commercial
43
School Bus
Medium-Duty Commercial
51
Refuse Truck
Medium-Duty Commercial
52
Single Unit Short-haul Truck
Medium-Duty Commercial
53
Single Unit Long-haul Truck
Heavy-Duty Commercial
54
Motor home
Medium-Duty Commercial
61
Combination Unit Short-haul Truck
Heavy-Duty Commercial
62
Combination Unit Long-haul Truck
Heavy-Duty Commercial
8.3. Updated Average Speed Distributions
As an example, the resulting default average speed distributions for different vehicle types are
shown in Figure 8-1 for all road types and day types at 5 pm.
• Differences between Personal, Medium-Duty and Heavy-Duty commercial are most
noticeable on rural restricted roads, where the Personal category (mapped to Passenger
Cars, Passenger Trucks and Motorcycles) shows notably more time traveling at speeds
above 75 mph.
• For all vehicle types, weekday-weekend differences between average speed profiles are
generally small; the exception is for urban restricted access roads, reflecting the
expected difference between weekend and weekday traffic volumes at 5 pm on urban
freeways.
80
-------
HD commercial
MD commercial
Personal
c
o
o
Cfl
u_
"D
CD
CD
Q.
CO
O)
>
cu
Weekday
Weekend
0 20 40 60 80 0 20 40 60 80 0 20 40 60 80
Average Speed (mph)
Figure 8-1 Average speed distributions for 5pm (hourlD 17) on the different MOVES road
types. For mapping between MOVES source types and telematics vehicle types, see Table 8-2.
The StreetLight data improves on previous estimates, with more data and enough detail to
differentiate between Personal, Medium-Duty and Heavy-Duty vehicle types. However, it does
not provide information to differentiate between vocation-specific trucks or buses. As new
datasets beome available, we will continue to update and improve these inputs.
81
-------
9. Driving Schedules and Ramps
Drive schedule refers to a second-by-second vehicle speed trajectory. The drive schedules in
MOVES are intended to include all vehicle operation from the time the engine starts until the
engine is keyed off, both driving (travel) and idling time.k Drive schedules are used in MOVES to
determine the operating mode distribution for MOVES running processes for calculation of
emissions and energy consumption. The drive schedules in MOVES5 are unchanged from
MOVES3 and are mostly the same as those in MOVES2014, with the exception of drive
schedules for transit and school buses, as described below, and the handling of ramps as
described in Section 9.2.
More specifically, each second of vehicle operation is assigned to an operating mode as a
function of vehicle velocity in each second and the specific power (VSP) for light-duty vehicles
or scaled tractive power (STP) for heavy-duty vehicles. The distinction between VSP and STP is
discussed in Section 15. Each operating mode is associated with an emission rate (in grams per
hour of vehicle operation). The average speed distribution is used to weight the operating
mode distributions determined from driving schedules with different average speeds into a
composite operating mode distribution that represents overall travel by vehicles. The
distribution of operating modes is used by MOVES to weight the emission rates to account for
the vehicle operation.
9.1. Driving Schedules
A key feature of MOVES is the capability to accommodate many schedules to represent driving
patterns across source type, road type and average speed. For the national default case,
MOVES uses 49 drive schedules with various average speeds, mapped to specific source types
and road types.
MOVES stores all drive schedule information in three database tables. The DriveSchedule table
provides the drive schedule name, identification number and the average speed of the drive
schedule. The DriveScheduleSecond table contains the second-by-second vehicle trajectories
for each schedule. In some cases, the vehicle trajectories are not contiguous; as detailed below,
they may be formed from several unconnected microtrips that overall represent driving
behavior. The DriveScheduleAssoc table defines the set of schedules which are available for
each combination of source use type and road type.
Table 9-1 through Table 9-6 below list the driving schedules used in MOVES. Some driving
schedules are used for both restricted access (freeway) and unrestricted access (non-freeway)
driving. In these cases, for example, at extreme congestion or unimpeded high speeds, we
k However, as described in Section 10, recent data suggests that drive schedules miss a substantial fraction of real-
world idling. MOVES has been updated to better account for the idling that was not captured in previous versions
of the model.
82
-------
assume that the road type itself has little impact on the expected driving behavior (driving
schedule). Similarly, some driving schedules are used for multiple source types where vehicle
specific information was not available.
Table 9-1 MOVES driving cycles for motorcycles, passenger cars, passenger trucks and light
commercial trucks (11, 21, 31, 32)
ID
Cycle Name
Average
Speed
Unrestricted Access
Restricted access
Rural
Urban
Rural
Urban
101
LD Low Speed 1
2.5
X
X
X
X
1033
Final FC14LOSF
8.7
X
X
1043
Final FC19LOSAC
15.7
X
X
1041
Final FC17LOSD
18.6
X
X
1021
Final FC11LOSF
20.6
X
X
1030
Final FC14LOSC
25.4
X
X
153
LD LOS E Freeway
30.5
X
X
1029
Final FC14LOSB
31.0
X
X
1026
Final FC12LOSE
43.3
X
1020
Final FC11LOSE
46.1
X
X
1011
Final FC02LOSDF
49.1
X
1025
Final FC12LOSD
52.8
X
1019
Final FC11LOSD
58.8
X
X
1024
Final FC12LOSC
63.7
X
X
1018
Final FC11LOSC
64.4
X
X
1017
Final FC11LOSB
66.4
X
X
1009
Final FC01LOSAF
73.8
X
X
X
X
158
LD High Speed Freeway 3
76.0
X
X
X
X
83
-------
Table 9-2 MOVES d
riving eye
es for other buses (4]
L)
ID
Cycle Name
Average
Unrestricted access
Restricted access
Speed
Rural
Urban
Rural
Urban
398
CRC E55 HHDDT Creep
1.8
X
X
X
X
404
New York City Bus
3.7
X
X
201
MD 5mph Non-Freeway
4.6
X
X
X
X
405
WMATA Transit Bus
8.3
X
X
202
MD lOmph Non-Freeway
10.7
X
X
X
X
203
MD 15mph Non-Freeway
15.6
X
X
X
X
204
MD 20mph Non-Freeway
20.8
X
X
X
X
205
MD 25mph Non-Freeway
24.5
X
X
X
X
206
MD 30mph Non-Freeway
31.5
X
X
X
X
251
MD 30mph Freeway
34.4
X
X
X
X
252
MD40mph Freeway
44.5
X
X
X
X
253
MD 50mph Freeway
55.4
X
X
X
X
254
MD 60mph Freeway
60.4
X
X
X
X
255
MD High Speed Freeway
72.8
X
X
X
X
397
MD High Speed Freeway Plus
5mph
77.8
X
X
X
X
Table 9-3 MOVES driving cycles for transit and school buses
ID
Cycle Name
Average
Speed
Unrestricted access
Restricted access
Rural
Urban
Rural
Urban
398
CRC E55 HHDDT Creep
1.8
X
X
X
X
401
Bus Low Speed Urban
3.1
X
X
404
New York City Bus
3.7
X
X
201
MD 5mph Non-Freeway
4.6
X
X
405
WMATA Transit Bus
8.3
X
X
202
MD lOmph Non-Freeway
10.7
X
X
402
Bus 12mph Non-Freeway
11.5
X
X
203
MD 15mph Non-Freeway
15.6
X
X
204
MD 20mph Non-Freeway
20.8
X
X
403
Bus 30mph Non-Freeway *
21.9
X
X
205
MD 25mph Non-Freeway
24.5
X
X
206
MD 30mph Non-Freeway
31.5
X
X
251
MD 30mph Freeway
34.4
X
X
252
MD40mph Freeway
44.5
X
X
253
MD 50mph Freeway
55.4
X
X
X
X
254
MD 60mph Freeway
60.4
X
X
X
X
255
MD High Speed Freeway
72.8
X
X
X
X
397
MD High Speed Freeway Plus
5mph
77.8
X
X
X
X
42, 43)
84
-------
Table 9-4 MOVES driving cycles for refuse trucks (51)
ID
Cycle Name
Average
Unrestricted access
Restricted access
Speed
Rural
Urban
Rural
Urban
398
CRC E55 HHDDT Creep
1.8
X
X
501
Refuse Truck Urban
2.2
X
X
301
HD 5mph Non-Freeway
5.8
X
X
302
HD lOmph Non-Freeway
11.2
X
X
X
X
303
HD 15mph Non-Freeway
15.6
X
X
X
X
304
HD 20mph Non-Freeway
19.4
X
X
X
X
305
HD 25mph Non-Freeway
25.6
X
X
X
X
306
HD 30mph Non-Freeway
32.5
X
X
X
X
351
HD 30mph Freeway
34.3
X
X
X
X
352
HD 40mph Freeway
47.1
X
X
X
X
353
HD 50mph Freeway
54.2
X
X
X
X
354
HD 60mph Freeway
59.4
X
X
X
X
355
HD High Speed Freeway
71.7
X
X
X
X
396
HD High Speed Freeway Plus
5mph
77.8
X
X
X
X
Table 9-5 MOVES driving cycles for single unit trucks and motor homes (52, 53, 54)
ID
Cycle Name
Average
Unrestricted access
Restricted access
Speed
Rural
Urban
Rural
Urban
398
CRC E55 HHDDT Creep
1.8
X
X
X
X
201
MD 5mph Non-Freeway
4.6
X
X
X
X
202
MD lOmph Non-Freeway
10.7
X
X
X
X
203
MD 15mph Non-Freeway
15.6
X
X
X
X
204
MD 20mph Non-Freeway
20.8
X
X
X
X
205
MD 25mph Non-Freeway
24.5
X
X
X
X
206
MD 30mph Non-Freeway
31.5
X
X
X
X
251
MD 30mph Freeway
34.4
X
X
X
X
252
MD40mph Freeway
44.5
X
X
X
X
253
MD 50mph Freeway
55.4
X
X
X
X
254
MD 60mph Freeway
60.4
X
X
X
X
255
MD High Speed Freeway
72.8
X
X
X
X
397
MD High Speed Freeway Plus
5mph
77.8
X
X
X
X
85
-------
Table 9-6 MOVES driving cycles for combination trucks (61, 62)
ID
Cycle Name
Average
Speed
Unrestricted access
Restricted access
Rural
Urban
Rural
Urban
398
CRC E55 HHDDT Creep
1.8
X
X
X
X
301
HD 5mph Non-Freeway
5.8
X
X
X
X
302
HD lOmph Non-Freeway
11.2
X
X
X
X
303
HD 15mph Non-Freeway
15.6
X
X
X
X
304
HD 20mph Non-Freeway
19.4
X
X
X
X
305
HD 25mph Non-Freeway
25.6
X
X
X
X
306
HD 30mph Non-Freeway
32.5
X
X
X
X
351
HD 30mph Freeway
34.3
X
X
X
X
352
HD 40mph Freeway
47.1
X
X
X
X
353
HD 50mph Freeway
54.2
X
X
X
X
354
HD 60mph Freeway
59.4
X
X
X
X
355
HD High Speed Freeway
71.7
X
X
X
X
396
HD High Speed Freeway Plus
5mph
77.8
X
X
X
X
The default drive schedules for light-duty vehicles listed in the tables above were developed
from several sources. "LD LOS E Freeway" and "HD High Speed Freeway" were retained from
MOBILE6 and are documented in report M6.SPD.001.60 "LD Low Speed 1" is a historic cycle
used in the development of speed corrections for MOBILE5 and is meant to represent extreme
stop-and-go "creep" driving. "LD High Speed Freeway 3" was developed for MOVES to
represent very high-speed restricted access driving. It is a 580-second segment of restricted
access driving from an in-use vehicle instrumented as part of EPA's On-Board Emission
Measurement Shootout program,61 with an average speed of 76 mph and a maximum speed of
90 mph. Fifteen additional light-duty "final" cycles were developed for MOVES based on urban
and rural data collected in California in 2000 and 2004.51 These cycles were selected to best
cover the range of road types and average speeds modeled in MOVES.
The driving schedules (ID 201-206, 251-255, 397, and 398) used for all buses (41,42,43) are
borrowed directly from driving schedules used for single unit trucks. The "New York City Bus"62
and "WMATA Transit Bus"63 drive schedules are included for urban driving that includes transit-
type bus driving behavior. The "CRC E55 HHDDT Creep"64 cycle was included to cover
extremely low speeds for heavy-duty trucks. The "Bus 12 mph Non-Freeway" (ID 402) and the
"Bus 30 mph Non-Freeway" (ID 403) cycles used for transit and school buses were based on
Ann ArborTransit Authority buses instrumented in Ann Arbor, Michigan.65 The bus "flow"
cycles were developed using selected non-contiguous snippets of driving from one stop to the
next stop, including bus-stop idling, to create cycles with the desired average driving speeds.
The "Bus Low Speed Urban" bus cycle (ID 401) is the last 450 seconds of the standard New York
City Bus cycle.
For MOVES3 and later versions of MOVES, we revised the handling of bus speeds; we changed
the driving cycle mapping in the DriveSchedule table to be the actual speed for all bus drive
86
-------
cycles. Consistent with our changes, users should input the actual average speed distribution
for transit buses.
The "Refuse Truck Urban" cycle represents refuse truck driving with many stops and a
maximum speed of 20 mph but an average speed of 2.2 mph. This cycle was developed by West
Virginia University for the State of New York. For restricted access driving of refuse trucks at
extremely low speeds, the CRC E55 HHDDT Creep cycle is used instead. All of the other driving
cycles used for refuse trucks are the same as the driving cycles developed for heavy-duty
combination trucks, described below.
Single unit and combination trucks use driving cycles developed specifically for MOVES, based
on data from 150 medium- and heavy-duty vehicles instrumented to gather instantaneous
speed and GPS measurements.66 The drive cycle data were segregated into restricted access
and unrestricted access driving for medium- and heavy-duty vehicles and then further stratified
by vehicles trips according the pre-defined ranges of average speed covering the range of
vehicle operation. The medium-duty cycles are used with single unit trucks and heavy-duty
cycles are used with combination trucks.
The developed schedules are not contiguous schedules which could be run on a chassis
dynamometer but are made up of non-contiguous "snippets" of driving (microtrips) meant to
represent target distributions. For use with MOVES, we modified the schedules' time field in
order to signify when one microtrip ended and one began. The time field of the driving
schedule table increments two seconds (instead of one) when each new microtrip begins. This
two-second increment signifies that MOVES should not regard the microtrips as contiguous
operation when calculating accelerations.
Both single unit and combination trucks use the CRC E55 HHDDT Creep cycle for all driving at
extremely low speeds. At the other end of the distribution, none of the existing driving cycles
for heavy-duty trucks included average speeds sufficiently high to cover the highest speed bin
used by MOVES. To construct such cycles, EPA started with the highest speed driving cycle and
added 5 mph to each point, effectively increasing the average speed of the driving cycle
without increasing the acceleration rate at any point. We have checked the feasibility of these
new driving cycles (396 and 397) using simulations with the EPA's Greenhouse Gas Emissions
Model (GEM)67 for medium- and heavy-duty vehicle compliance. GEM is a forward-looking full
vehicle simulation tool that calculates fuel economy and GHG emissions from an input drive
trace and series of vehicle parameters. One of the aspects of forward-looking models is that
the driver model is designed to demand torque until the vehicle drive trace is met. Our results
indicate that the simulated vehicles could follow the speed demands of the proposed driving
cycles without exceeding maximum torque or power.
We compared the operating mode distrition estimated for a national scale run in MOVES to the
operating mode distribution measured from the Heavy-Duty In-Use Testing (HDIUT) program in
the Appendix G of the heavy-duty exhaust report. Overall, the operating mode distributions
compare well. One notable differene is, for a national scale run, MOVES estimates a higher
87
-------
percentage of activity in the highest power, high speed operating mode bins.18 This may be
reasonable because the manufactur-run testing for the HDIUT data are expected to under-
represent high power operation due to steep grades, high speeds, and heavy-pay loads (e.g.,
multiple trailers, over-weight trailers) compared to the in-use fleet. Or perhaps, the discrepancy
could be due in part to the high-speed driving cycle being overly aggressive compared to in-use
driving. As mentioned in the Conclusions section, we suggest that a further evaluation of the in-
use operating mode distributions and heavy-duty driving cycles be considered for future work
for MOVES.
3.2. Modeling of Ramps in MOVES
For MOVES3 and later versions of MOVES, we simplified the modeling of emissions on
restricted access roadways by modeling ramps as part of highway driving. The MOVES3
Population and Activity Report has a detailed discussion of this change.68 For future versions of
MOVES, we hope to investigate whether drive cycles can be further improved by incorporating
a representative mix of ramp and highway driving.
At the project-scale, it is important to model ramps separately to identify localized areas where
high acceleration and deceleration events cause increases in exhaust emissions69 and brake
emissions. Users can continue to estimate ramps as individual links in project-scale. Preferably,
project-level users can characterize the specific operating mode or driving cycle of the ramps
they are evaluating.
88
-------
10. Off-Network Idle Activity
To better account for observed levels of idling, we added a new emission calculation to
MOVES3 for County and National Scale runs' allowing the model to estimate idle emissions that
occur off the road network (i.e., on roadTypelD=l) for all soucetypes. This section summarizes
the calculation methodology employed by MOVES and then provides information on the idling
data available for both light-duty and heavy-duty vehicles.
10.1. Off-Network Idle Calculation Methodology and Definitions
We are defining the total idle fraction (TIF) as the ratio of the total source hours idling and total
source hours operating. This value can be derived from instrumented vehicles as explained
below. MOVES defines "idle" as any seconds in the driving schedules where the speed is less
than one mile per hour (opModelD=l) during engine operation. Using the fraction of vehicle
operation hours that are opModelD=l, the source hours idle (SHI2-5) during normal daily vehicle
operation for each of the four onroad road types (roadTypelDs 2, 3, 4, & 5) can be determined
from the driving schedules used for vehicle operation on roadways. We exclude any extended
engine idle that occurs during the mandated rest period for combination long-haul truck
(sourceTypelD 62), which we call hotelling (see Section 11). Total idle fractions are stored in
the new TotalldleFraction table in the MOVES default database.
Since the estimates of TIF are greater than the idle time accounted for in the MOVES driving
schedules (SHI2-5), we also need to increase MOVES' estimate of total source hours operating
(SHO). In particular, the off-network idle (ONI) time is defined as the additional idle hours that
need to be added to the on-network source hours operating (SHO2-5) in order to account for the
additional idle time. The on-network SHO2-5 is derived from the VMT and speed distribution.
Starting with MOVES3, the additional ONI hours are assigned to the running exhaust process
(processlD=l) for the off-network road type (roadTypelD=l).
In MOVES2014, total SHO was calculated from vehicle miles traveled (VMT) and average speed
for all onroad roadTypelDs 2, 3, 4 and 5. Starting with MOVES3, we renamed this value as on-
network SHO2-5 to indicate that additional time needs to be added to account for off-network
idle time. The SHO for all road types will now include the "extra" operating time (ONI) implied
by the larger total idle fraction value:
Equation 10-1
Where /' = roadTypelD.
1 In Project Scale, MOVES does not adjust activity to account for off-network idling. Instead, the user can provide
location-specific idling activity as appropriate.
89
-------
Source hours idle (SHI) then is the total hours of idle, excluding diesel long-haul combination
truck hotelling idle:
SHI = (Y SHIi) + ONI Equation 10-2
*—>i=2
Where /' = roadTypelD.
All running exhaust activity for roadTypelD=l is idle, so SHOi=SHIi and represent ONI. Since the
TIF values are the measured fraction of idle time during vehicle operation, the SHI is also the
result of applying the TIF to the SHO:
SHI = TIF x SHO Equation 10-3
Thus, from Equation 10-1, Equation 10-2, and Equation 10-3:
(EU SHIi) + ONI Equation 10-4
T'F flf=2SHOiJ+ONI
And, by by re-arranging Equation 10-4 and using the TIF, on-network source hours operating
(SHO2-5) and on-network source hours idling (SHI2-5) from the four network road types, MOVES
can calculate the hours for off-network idle (ONI):
(I?=2 SHOj ) X TIF - Ei=2 SHIi Equation 10-5
(1 - TIF)
Where /' = roadTypelD.
As an example, the default values of TIF for light-duty vehicles in idleRegionlD=101 (New Jersey)
are presented in Table E-2 in Appendix E.
In cases where the ONI is calculated to be less than zero, the ONI will be set to zero. This is
currently true for motorcycles and motorhomes.
Off-network idle emissions are calculated for each hour by using the corresponding emission
rate (grams per hour) for opModelD=l for that hour. All of the adjustments (e.g., fuel effects,
air condition effects) made to the emission rates for opModelD=l for other road types apply to
off-network idle emissions as well. MOVES3 separately reports the emissions from the off-
network idle hours in the movesOutput table as exhaust running process (processlD=l) for road
type "off-network" (roadTypelD=l).
90
-------
10.2. Light-Duty Off-Network Idle
10.2.1. Verizon Telematics Data
MOVES defaults for light-duty off network idling continues to rely on our MOVES3 analysis. In
developing MOVES3, Verizon Telematics data for light-duty vehicles was purchased only for the
following five states due to costs - California, New Jersey, Illinois, Georgia and Colorado. These
states were selected for a variety of reasons, including geographic coverage, urban and rural
mix, use of inspection and maintenance programs, and number of vehicles participating in the
program. The data were collected August 2015 through August 2016 using on-board diagnostic
data loggers under contracts with State Farm insurance, Mercedes-Benz and Volkswagen. The
data includes vehicles from model year 2017 back to model year 1996, which is also the first
year manufacturers were required to equip all vehicles with on-board diagnostic (OBD)
systems.70 Vehicle owners allowed their vehicles to be measured for a variety of reasons and
the data cannot be considered a random sample. The Verizon Telematics data were used as a
primary data source for the light-duty off-network idle defaults described in this section and
also for the soak and start defaults described in Section 12.1 The data characteristics and pre-
processing steps for both analyses are described here.
The Verizon data includes activity information gathered on vehicles for all or some subset of the
entire year. The information collected was summarized and processed into individual trips for
analysis. The analysis summary database includes trip start time and date, trip end time and
date, total trip time, total idle time, trip average speed, trip maximum speed and trip distance.
Trips were defined as the time period from key-on to key-off. Engine idle was defined as any
time during the trip where the recorded engine RPM was greater than zero and the vehicle
speed was less than one mile per hour. Total idle time is a fraction defined as the ratio of the
sum of the idle time periods in a trip and the total time of the trip from key-on to key-off. In
addition to the trip data, each trip was associated with a vehicle ID. For each vehicle ID, the
model year and vehicle registration postal ZIP code was provided. All vehicles were light-duty,
either passenger car or light-duty truck. No information about where the trips occurred was
provided in the samples.
Using the provided data, all of the activity by vehicles was assumed to occur within the county
in which they were registered. The counties were categorized as urban or rural based on the
U.S. Census Metropolitan Statistical Area (MSA) classifications. Counties were also grouped as
either having a State Inspection and Maintenance (l/M) program or not.
91
-------
10.2.2. QA/QC of the Verizon Telematics Data
Table 10-1 shows a high-level summary of Verizon Telematics data. The original dataset
provided by Verizon included around 41 million trip summary records from the five states. Such
large datasets pose several challenges related to data quality and sampling. For example, for
some trips, data were found to be missing or incomplete. Such trips were removed from the
original dataset and the remainder were used to analyze the idle fraction as summarized in the
"Total Trips (Idle)" column of Table 10-1.
Table 10-1 Verizon Telematics data sample summary
State
Total Trips
(Original)
Total Trips
(Idle)*
Total Trips (Soak
Time & Starts)"
%Trips"*
California
1,958,858
1,886,947
1,761,184
90%
Colorado
5,644,374
5,390,417
4,977,334
88%
Georgia
15,457,392
14,654,336
13,465,865
87%
Illinois
12,955,252
12,318,387
11,448,257
88%
New Jersey
5,139,506
4,947,792
4,615,346
90%
Only valid trips included in idle analysis.
Only valid trips with previous recorded valid trips included in start and soak analysis.
*** Percent of total trips remaining after all screening (starts divided by original total).
In addition, not all vehicles in the sample had 12 complete months of data, due to termination
of subscriptions, instrumentation failures, etc. during the sampling period. To distinguish
infrequently used vehicles from those that had left the program, we developed an algorithm to
extract only those vehicles and their associated monthly data for which there was at least one
trip in the current month, the preceding and succeeding months. In addition, for a given
vehicle, the first and last month of the data for each vehicle was kept in the sampling frame
only if there was at least one trip in the first week and the last week for the month,
respectively. Figure 10-1 shows the Verizon Telematics sample vehicle population by state and
month derived using this sampling approach. Appropriate weighting was then applied to the
monthly results to generate annual averages.
92
-------
12015-8 >2015-9 B2015 - 10 12015 - 11 B2015 - 12 B2016 - 1
12016-3 ¦ 2016 - 4 B2016-5 B2016-6 B2016-7 B2016-8
12016-2
7000
6000
£ 5000
¦a
> 4000
t 3000
J
a
2000
1000
21 31
California
21 31 21 31 21 31
Colorado Georgia Illinois
sourceTypelD/State
21 31
New Jersey
Figure 10-1 Sample vehicle population in the Verizon Telematics data by month, state and
sourceType. Note: the legend indicates the "year-month" of the data collection.
There were a few instances where the trip time was less than 1 second, or the soak time was
less than two seconds, for example, when a vehicle crossed into a different time zone or when
the data logger recorded erroneous trip starts at midnight for trips that included midnight
driving. Such trips represented less than 1 percent of the total trips for any given state and
were removed from the idle and starts/soak analysis. The remaining trips were used to analyze
engine starts and soaks (see the "Total Trips (Soak Time & Starts)" column in Table 10-1 for the
total trip counts). The erroneous trip starts removed from the start/soak analysis do not affect
the results for the analysis of total idle time.
10.2.3. Estimating MOVES National Defaults from Verizon Telematics
Data
The Verizon Telematics data covered only five states, but MOVES must model the entire U.S.
Thus, we associated each state with nearby states to create vehicle-population weighted
national averages for starts and soaks and regional-specific values for idle time. Table 10-2 lists
the vehicle populations used for computing national averages. Figure 10-2 shows how we
mapped individual states to the Verizon data. We grouped the states qualitatively, considering
proximity and climate. Climate was considered because the monthly patterns varied between
areas with large temperature shifts between seasons (Colorado, Illinois) and states with
moderate seasonal changes (California and Georgia). The weighted average results for the light-
duty passenger trucks (indicated in the data as sourceTypelD 31) were used for light-duty
93
-------
commercial trucks (sourceTypelD 32) as well. Due to lack of data, motorcycle idle fractions
were set to zero. This results in the same roadway (drivecycle-based) idling as before and no
off-network idle.
Table 10-2 2014 Vehicle populations of the idle regions
Verizon data source state
sourceTypelD
Vehicle Population
idleRegionID
California
21
23,114,006
105
California
31
19,917,792
105
Colorado
21
6,902,041
104
Colorado
31
8,823,105
104
Georgia
21
38,269,101
102
Georgia
31
39,358,137
102
Illinois
21
26,768,198
103
Illinois
31
25,510,186
103
New Jersey
21
27,625,575
101
New Jersey
31
23,077,050
101
Calgary
- Vancouver
California
Colorado
S
Illinois
Georgia
M EXICO
Guadalajara
Ottj.va MontrfflH
I. o -O
roronl • ¦
o Mr
New Jersey
ERSPnilaclelphia
'in (it on
C UHA
Esn. HEKE. Ga... -
Figure 10-2 Default Regions for Weighting Light-Duty Activity"1
m Note, Alaska is associated with Colorado. Hawaii, Puerto Rico arid the Virgin Islands are associated with
California.
94
-------
In addtion to region, the Verizon Telematics data analysis suggested that the following factors
are important when estimating total idling fraction:
• Month of the year (which depends on the region)
• County type, i.e., whether registered in an urban (MSA) or rural county
• Passenger car or light truck
• Day type, i.e., weekend vs. weekday variation
The analysis showed no significant variation with age or hour of the day. A simplified linear
regression model was built to capture the variability of the total idle fraction (TIF) across
different variables (daylD, sourceTypelD, countyTypelD, idleRegionID and monthID). MOVES3
default values for TIF were calculated based on the equation below:
TIF = daylDi + sourceTypelDj + countyTypeIDk + idleRegionIDl ^ uation 10 6
+ monthIDm + idleRegionIDl x monthIDm + n
where, i, j, k, l,m are coefficient values for the combinations of daylD (2=Weekend,5=Weekday),
sourceTypelD, countyTypelD, idleRegionID and monthID and n is the intercept (a constant) for
Equation 10-6 above. The regression model handled ordinal categorical variables as
independent variables. The full set of coefficients are available in Appendix E.
As one might expect, idling activity is more common in winter months in colder states and
urban areas have more idling activity than rural areas. There is less idling activity on weekends
versus weekdays. Idling activity is similar for passenger cars and light trucks, but separate idle
fractions were developed for each of the source types.
In MOVES, we use the model fit TIF values from the multi-variable linear model (Equation 10-6,
rather than using the averages from the Verizon Telematics data, mainly to smooth the
variation in the Verizon Telematics data. Figure 10-3 below illustrates the model fit against
actual values. TIF model results are represented by solid lines versus average values from the
Verizon Telematics data, shown as dashed lines. As expected, Region 105 (California) which has
the smallest sample size also shows the most variation and deviation from the regression
results. For example, for Region 105 (California), passenger trucks (sourceTypelD 31), weekdays
(daylD 5), the model fit smooths out the abnormally high idle fraction measured for July
(monthID 7).
We also use the model estimated TIF values to estimate values that were not measured by
Verizon. Note that there was no data available for New Jersey from Verizon Telematics for rural
counties (i.e., countyTypelD=0) as shown in Figure 10-3. However, the regression model applies
the rural/urban effect without regard to region. Appendix E shows a sample calculation using
MOVES3 default values for passenger cars in rural counties in idleRegionlD=101 (New Jersey).
95
-------
The model fit TIF values apply to all calendar years in MOVES. Note that idleRegionID and
countyTypelD vary depending on the county location. Each state is assigned an idleRegionID in
the MOVES State table as shown in Figure 10-2. Each county is assigned an "urban" or "rural"
countyTypelD in the MOVES County table based on the MSA designation. As discussed earlier,
the results for the light-duty passenger trucks (indicated in the data as sourceTypelD 31) are
also used for light-duty commercial trucks (sourceTypelD 32).
96
-------
—•—TIF (Modeled) - 0 —•—TIF (Modeled) - 1 0 represents courilyTypcJD-0
-.-TIF (Actual)-0 - — TIF(Actual)- 1 1 comtvT»pelI)-l
sourceTypelD=21, idleRegionID=101
source'fypelD=21, id IeRegionlD=l 03
sourccIypell)=31, id I c Region I1)-101
0.3
E 0.28
'I 0.26
2
't 0.24
-5 0.22
5
0.2
0.18
0.16
monthlD/day'ID
soureeTypeTD=31, idIcRegionTD=l 03
0.34
0.32
^ 0.3
E 0.28
5
0.26
jt 0.24
I0'22
0.2
0.18
0.16
source'i'ypelD=21, idleRegionID=l 02
1 2 3 4 5 6 7 8 9 101112| 1 2 3 4 5 6 7 8 9 1011 12
montblDMayll)
sourceTypeID=21, idleRegionID=l04
0.34
0.32
0.3
E- 0.28
1
'§ 0.26
2
U 0.24
^ 0.22
|
0.2
0.18
0.16
12345678? 1011
monthlD/daylD
sourceTypeTD=31, idleRegionTD=l 02
sourecTypeID=31, idleRegionID=l04
monlhTD'ilayTT)
sourceTypeID=21, idleRegionlD=105 source l'ypelD=31, idleRegionlD=105
Figure 10-3: TIF model results compared to the values from the Verizon Telematics data
97
-------
10.3. Heavy-Duty Off-Network Idle
The Verizon Telematics data exclusively covered light-duty vehicles. Heavy-duty vehicles are
spread across a wide range of vocations and have activity patterns that are distinctly different
from light-duty. Currently, the idling captured in the MOVES driving cycles represents the idling
at intersections and on congested highways, but do not include a full estimate of "workday
idle" that many commercial heavy-duty trucks experience in their daily operation, such as
queuing at distribution centers, or loading and unloading payload. Off-network idle is also
intended to address these gaps in idle activity modeling.
The heavy-duty off-network idle defaults were derived from the National Renewable Energy
Laboratory (NREL) Fleet DNA clearinghouse of commercial fleet vehicle operating data. The
data processing applied to the Fleet DNA dataset is described in this section. This analysis has
not been updated since MOVES3. However, the University of California Riverside's Bourns
College of Engineering Center for Environmental Research and Technology (CE-CERT) has
concluded their data collection for a study to evaluate the selective catalytic reduction (SCR)
behavior of heavy-duty vehicles. We hope to apply the same processing steps to the latest CE-
CERT dataset and expect to combine the results the with Fleet DNA data in a future MOVES
update.
The same Fleet DNA dataset and pre-processing steps described in this section were used for
the soak and start defaults described in Section 12.2
10.3.1. NREL Fleet DNA Database
We partnered with NREL to make use of their expansive Fleet DNA database71 of heavy-duty
vehicles to develop idle activity estimates for heavy-duty vehicles. NREL's Fleet DNA database is
developed from vehicles operating in the field with devices to record 1-Hz telematics and CAN
(controller area network72) data.
While the Fleet DNA database includes a wide range of fuels, vehicle drivetrains and propulsion
mechanisms, only diesel-powered conventional vehicles were included in the analysis to ensure
the selected drive cycles are representative of traditional operation and not modified to
accommodate the vehicle architecture. This analysis used data from 415 conventional heavy-
duty vehicles with over 120,000 hours of operation, providing a diverse data encompassing 23
vehicle vocations in 36 states. The number of conventional vehicles in the Fleet DNA database
by MOVES source type are shown in Table 10-3. The table also includes the number of states
with activity in each Fleet DNA sample. The geographic distribution could influence average idle
98
-------
emission rates, due to differences in congestion, topography and regional policies." 73 However,
as presented in the NREL project report,74 truck idling and start activity was observed to be
largely a function of the truck vocation, rather than the US state of operation. Likely a larger
sample size of vehicles across vocations and states would be needed to elucidate geographic
differences in truck activity.
Table 10-3. Sample size of conventional vehicles in the Fleet DNA database by MOVES source
type
Number of
Number of States
sourceTypelD
Source Type Name
Vehicles in
Fleet DNA
with Recorded
Activity
41
Other Buses (non-school, non-
n
n
transit)
U
U
42
Transit Buses
16
3
43
School Buses
7
1
51
Refuse Trucks
37
4
52
Single Unit Short-Haul Trucks
119
8
53
Single Unit Long-Haul Trucks
0
0
54
Motor Homes
0
0
61
Combination Short-Haul Trucks
105
8
62
Combination Long-Haul Trucks
131
32
Total
415
Note: The number of trucks operating in each US state is listed in the NREL project report.74
Table 10-4 shows the vocational distribution of the short-haul source types (single unit and
combination short-haul trucks) and the sample size of each vocation category. A complete
description of the Fleet DNA dataset, additional pre-processing performed, and analyses not
discussed in this report can be found in the NREL report.74
n For example, California has a regulation prohibiting idling for more than five minutes for vehicles that
are not California clean idle certified." However, other states, counties and cities also have idling
regulations. In addition, most recent heavy-duty vehicles are California clean-idle certified. For example,
all fourteen of the MY 2008 and later heavy heavy-duty tractors tested for extended idling emission
rates (produced from four major engine manufacturers) were all clean idle certified.5
99
-------
Table 10-4. Vocation types of the Combination Short-Haul and Single Unit Short-Haul vehicles within the Fleet
DNA database
Combination Short-Haul Vehicle
Vocation
Number of
Vehicles in Fleet
DNA
Single Unit Short-Haul
Vehicle Vocation
Number of
Vehicles in Fleet
DNA
Beverage Delivery
10
Warehouse Delivery
9
Food Delivery
13
Parcel Delivery
39
Drayage
28
Linen Delivery
17
Transfer Truck
28
Food Delivery
30
Local Delivery
7
Snow Plow
11
Regional Haul
7
Towing
4
Dump Truck
4
Concrete
3
Parcel Delivery
5
Delivery
1
Dry Van
3
Shredder
1
Propane Tank
1
Dump Truck
3
100
-------
10.3.2. CE-CERT Study
The California Air Resources Board (CARB) contracted with CE-CERT to conduct a large-scale
study in which vehicle and engine activity data were collected from 90 heavy-duty vehicles that
are mapped to 19 different groups defined by a combination of vocational use, gross vehicle
weight rating and geographic region within California. EPA supported the test program by
providing data loggers and data quality analysis through a Cooperative Research and
Development Agreement with CE-CERT. Most of these vehicles were registered in California
and traveled a majority of their miles in-state. The study did include some out-of-state vehicles
in the line-haul and pick-up/delivery categories. Almost all the vehicles were of model year
2010 or newer and most were equipped with SCR technology. One drayage truck was model
year 2008 (with no SCR) and all the buses were CNG fueled. In addition, some of the vehicles in
the study were hybrids. We hope to incorporate data from the CE-CERT study in future versions
of MOVES.
10.3.3. Heavy-Duty Off-network Idle Data Processing
The NREL Fleet DNA data were preprocessed to identify starts and idle periods for the analysis.
Two key parameters are engine speed (revolutions per minute [rpm]) and wheel speed (miles
per hour [mph]). An engine speed greater than zero indicates that the vehicle engine is running
and a wheel speed greater than zero mph signifies that the vehicle is in motion. In this analysis,
vehicle starts are calculated by identifying the transition from an engine speed of zero to
greater than zero. Vehicle soak is defined as the length of time between engine off (engine
speed of zero) and the next time it is started (engine speed greater than zero). A vehicle is
considered to be idling when its wheel speed is less than one mph and the engine speed is
greater than zero. The total operating time (engine RPM > 0) occurring within each daylD is also
calculated.
Periods of contiguous idle are identified by length and the daylD corresponding to the start of
the idle. If an idle period started during one daylD and ended on another, the idle time was only
counted for the daylD in which the trip started. Idle periods longer than an hour were
categorized separately as "extended idle" for long-haul combination trucks (sourceTypelD 62)
and not included in the average idle time of the off-network idle fraction calculation below.
Vehicle activity values in MOVES represent average activity at a national scale. MOVES uses
"total idle fraction" to quantify off-network idle. In this analysis, total idle fraction was
calculated by first summing the daily average idle time for each individual vehicle across all
vehicles within the same vehicle (sourceType) and day type (daylD) classification. Those
summed idle times were then divided by the sum of the daily average operating time for each
individual vehicle across all vehicles within the same vehicle and day type. This sum-over-sum
approach normalizes the recorded activity by the amount of time each vehicle was
instrumented and weights the average idle fraction towards the vehicles with the most daily-
101
-------
average activity.0 Equation 10-7 shows the calculation of the total idle fraction for each source
type and specific day type (weekday or weekend).
Where:
i = individual vehicle ID
s = source type ID
d = day type ID
10.3.4. Heavy-duty Off-network Idle Results
As seen in Table 10-3, several heavy-duty source types were not available in the Fleet DNA
database at the time of this report. Additionally, none of the school buses instrumented for this
dataset operated on the weekend, so there is no data for daylD 2. We hope to have more of the
source types and daylD's covered when we process the CE-CERT dataset and combine it with
the Fleet DNA dataset in a future version of MOVES. In the interim, we assumed the idle
behavior of the missing vehicles closely matched others. We chose to use the transit bus
(sourceTypelD 42) to represent other buses (sourceTypelD 41), applied the weekday data from
the school bus (sourceTypelD 43) for the missing weekend data, used the single unit short-haul
data (sourceTypelD 52) to represent the single unit long-haul trucks (sourceTypelD 53). Lacking
data for motorhomes (sourceTypelD 54), we set their total idle fraction to zero. This will result
in the same roadway (drivecycle-based) idling as in MOVES2014 and no off-network idle. While
this is an area that would benefit from more research, we think it is unlikely for motorhomes to
idle significantly when they are not on roadways since they are equipped with APUs and often
park where auxiliary power is available.
Figure 10-4 and Figure 10-5 show the idle fraction values for weekends and weekdays,
respectively. In both figures, the solid blue bars represent the off-network idle for each heavy-
duty vehicle sourceType. The hashed bars represent the extended idle portion, which is only
available to the long-haul combination trucks (sourceTypelD 62). The specific values added to
the MOVES TotalldleFraction database table for this update are shown in Table 10-5.
° We evaluated several approaches for calculating average idle fraction using the Fleet DNA data. The approach
presented here (Equation 10-7) is equivalent to Equation 17-23 (Method 3 "normalized sum over sum") in
Appendix I. Appendix I includes an overview of each approach and a comparison between calculation approaches.
Idle fractions d =
Equation 10-7
102
-------
Idle Fraction, Weekends
¦ Off-Network
a Extended Idle
Vehicle Description / MOVES SourceType
Figure 10-4 Weekend idle fractions for heavy-duty vehicle sourceTypes based on data from
NREL's Fleet DNA database
Idle Fraction, Weekdays
0.6
0.5 -I
Off-Network
B Extended Idle
lllllll.
Other Bus Transit Bus School Bus Refuse Single Unit, Single Unit, Combo, Combo,
Truck Short Long Short Long
41 42 43 51 52 53 61 62
Vehicle Description / MOVES SourceType
Figure 10-5 Weekday idle fractions for heavy-duty vehicle sourceTypes based on data from
NREL's Fleet DMA database
103
-------
Table 10-5 Idle fraction values for heavy-duty sourceTypes based on data from NREL's Fleet
DNA database
SourceType
Vehicle Description
Weekend Idle Fractions
Weekday Idle Fractions
Off-
Network
Extended
Off-
Network
Extended
41
Other Bus
0.388
0.000
0.390
0.000
42
Transit Bus
0.388
0.000
0.390
0.000
43
School Bus
0.314
0.000
0.314
0.000
51
Refuse Truck
0.503
0.000
0.469
0.000
52
Single Unit, Short
0.420
0.000
0.348
0.000
53
Single Unit, Long
0.420
0.000
0.348
0.000
61
Combo, Short
0.312
0.000
0.332
0.000
62
Combo, Long
0.130
0.127
0.145
0.138
10.4. Off-network Idling Summary
Figure 10-6 displays the off-network idling fraction and the on-network idling fraction for an
urban county in the midwestern idle region. The off-network idling accounts for most of the
idling for most source types. Note that the idle fraction, and subsequently, the off-network
idling fraction changes significantly between January and July for the light-duty vehicles.
However, it is unchanged for the heavy-duty vehicles. Also, note that the idling fraction for
long-haul combination trucks is lower than for other vehicles because long-duration idling (> 1
hour) for long-haul combination trucks is modeled as hotelling activity discussed in the Section
11.
104
-------
daylD: 2
0.5
0.4
0.3
0.2
C0.1
o
!o.ol
0.5
0.41
0.3
0.2
0.1
0.0H
daylD: 5
u
idle_fraction
| off_network
I on network
21 31 32 41 42 43 51 52 53 61 62 21 31 32 41 42 43 51 52 53 61 62
sourceTypelD
Figure 10-6. On-network idle and Off-network idle fractions for an Urban County in the
Midwestern Region
105
-------
11. Hotelling Activity
MOVES defines "hotelling" as any long period of time (e.g., > 1 hour) that drivers spend in their
vehicles during mandated rest times during long distance deliveries by tractor/trailer
combination heavy-duty trucks. During the mandatory rest time, drivers can stay in motels or
other accommodations, but most of these trucks have sleeping berths built into the cab of the
truck and drivers stay in their vehicles.
Hotelling hours are included in MOVES to account for the energy used and pollutants generated
to power air conditioning, heat and other amenities. These amenities require power for
operation, which can be obtained by running the main truck engine (extended idle) or by use of
smaller on-board power generators (auxiliary power units, APU). Some truck stop locations
include power hookups (truck stop electrification or "shore power") to allow use of amenities
without running either the truck engines or APUs. Some of the rest time may occur without the
use of amenities at all.
In MOVES, only the long-haul combination truck source use type (sourceTypelD 62) is assumed
to have any hotelling activity. All source use types other than long-haul combination trucks
have hotelling activity fractions set to zero.
11.1. Hotelling Activity Distribution
In MOVES, hotelling hours are divided into operating modes which define the emissions
associated with the type of hotelling activity. As explained above, long-haul trucks are often
equipped with sleeping berths and other amenities to make the drive rest periods more
comfortable. Table 11-1 shows the hotelling operating modes available in MOVES.
Table V
L-l Hotelling activity operating modes in MOVES
OpModelD
Description
200
Extended Idling of Main Engine
201
Hotelling Diesel Auxiliary Power Unit (APU)
203
Hotelling Shore Power (plug in)
204
Hotelling Battery or All Engines and Accessories Off
The hotelling activity distributions in MOVES are consistent with the hotelling assumptions used
in EPA's Heavy-Duty Greenhouse Gas Phase 2 rulemaking, which included increasing adoption
of battery or electric supplemental power.75 Additionally, we updated the model to include a
fraction of hotelling time when the driver did not require any supplemental power. Starting in
2011, the hours-of-service regulations from the Federal Motor Carrier Safety Administration
(FMCSA) were updated to encourage longer periods of rest.76 Drivers could split their 10 hours
of mandated off-duty time between the sleeper berth for at least 8 hours and another location
for the remaining 2 hours. We assume the drivers do not require power when not in the sleeper
berth and applied a constant 20 percent of hotelling time to represent the 2 hours off-duty time
not in the sleeper berth for all years.
106
-------
The HotellingActivityDistribution table, shown in Table 11-2, contains the MOVES default values
for the distribution of hotelling activity to the operating modes. For model years before 2010,
we assume 80 percent of time is extended idling and 20 percent does not require supplemental
power, as mentioned previously. Starting with the 2010 model year, an increased number of
trucks equipped with APUs are expected as a result of the Phase 1 Heavy Duty Greenhouse Gas
Standards77 and a fraction of the time that previously was assigned to extended idle is now
assigned to opModelD 201 (the use of APUs). In later model years, we continue to assume a
constant fraction of time with no supplemental power and distribute the remaining time among
extended idle, APU use and increasing battery use based on EPA's assessment of technologies
expected to be used by tractor manufacturers to comply with the Heavy-Duty Greenhouse Gas
standards Phase 2, with stepwise increases in model years 2021, 2024 and 2027.78 Similar to
pre-2010 model years, we assumed drivers would not require supplemental power 20 percent
of the time for model years 2010 and later.
Alternative fueled long-haul trucks are assumed to have the same hotelling activity distribution
as diesel trucks as described above with the following adjustments:
• CNG trucks are assumed to not have diesel APUs and instead rely on the main engine
idling.
• Fuel Cell EVs are assumed to use shore power for the 80 percent time in sleeper berth.
Table 11-2 Default hotelling activity distributions
Fuel Type
Begin Model
Year ID
End Model
Year ID
opModeFraction for given opModelD
200
201
203
204
Idle
APU
Shore Power
Battery/Off
Diesel
1950
2009
0.80
0.00
0.00
0.20
2010
2020
0.73
0.07
0.00
0.20
2021
2023
0.48
0.24
0.00
0.28
2024
2026
0.40
0.32
0.00
0.28
2027
2060
0.36
0.32
0.00
0.32
CNG
1950
2020
0.80
0.00
0.00
0.20
2021
2026
0.72
0.00
0.00
0.28
2027
2060
0.68
0.00
0.00
0.32
Fuel Cell EV
1950
2060
0.00
0.00
0.80
0.20
Based on peer-review comments on the above analysis for diesel trucks in 2017, we
reevaluated our assumptions about APU and hotelling battery penetration rates. The diesel
APU usage assumptions for model year 2010 through 2020 in Table 11-2 are qualitatively
107
-------
consistent with two fleet surveys: NACFE 2018 Annual Fleet Fuel Study79 p and Shoettle et al.
(2016).80 q On the other hand, both surveys suggested a higher (non-zero) penetration of
hotelling battery units in 2010-2020, as well as projecting a higher penetration in future years.
However, given concerns about the representativeness of the surveys, we have decided to
retain the current assumption regarding fleet-average APU and battery usage in MOVES5 and
recognize that the current hotelling battery usage may be a low estimate. Future MOVES
updates could utilize instrumented truck and APU measurements to replace these projections.
11.2. National Default Hotelling Rate
To estimate hotelling activity, MOVES uses a hotelling rate. As shown in Equation 11-1, the
default hotelling rate is the national total hours of hotelling divided by the national total miles
driven by long-haul combination trucks on all restricted access roads (both urban and rural).
Hotelling Hours „ . _ _ _
Hotelling Rate = — — ; ——— ;—r Equation 11-1
Total Restricted Miles Traveled
The hotelling rate is used to estimate hotelling in different calendar years and to spatially
allocate hotelling to counties across the US. The hotelling rate is based on travel on rural and
urban restricted access roads (freeways), because this is where long-haul trucks are most
frequently operated and most hotelling occurs at locations near those roadways (i.e., rest stops
or truck stops).
The national default hotelling rate is based on data collected and analyzed by the National
Renewable Energy Laboratory (NREL) Fleet DNA74 as discussed in Section 10.3.1. For the
hotelling analysis, NREL analyzed data collected from 131 long-haul combination diesel trucks
operating in the United States. The 131 trucks had broad coverage across the United States,
with home bases in 32 states.
Because the NREL data did not include information on all operating modes of hotelling activity,
we back-calculated the hours of hotelling from the data on extended idling using Equation 11-2.
First, we estimated the extended idle hours per mile from the NREL data. Vehicles were
assumed to be extended idling (hotelling with the main engine running in idle), if the vehicle
p NACFE (2018) reported increasing diesel APU and and battery penetration rates for model year 2010-2016
vehicles. The diesel APU values span the MOVES values for model year 2010-2020. The NACFE 2013 Annual Fleet
Fuel Study reports survey values from 20 participating fleets, which are likely earlier adopters and may not be
representative of the entire fleet.
q Shoettle et al. reports that 38.7 percent of fleets use auxiliary power sets and 30.1 percent use battery packs
based on a survey of 96 heavy-duty fleet managers. However, information regarding the percentage of vehicles
within a fleet is not provided.
108
-------
speed = 0 and the duration of the idling was > 1 hour. For the 131 long-haul trucks, the trucks
averaged 3.45 extended idle hours for every 1,000 miles driven. Then, we calculated a ratio of
total miles traveled to restricted access miles using the MOVES national default values
presented in Table 7-2 (the rural restricted VMT fraction = 0.34 and urban restricted VMT
fraction = 0.26). This allows better spatial allocation of hoteling activity to counties with
freeways. Finally, we multiply the extended idle hours by the ratio of hotelling hours to the
extended idle hours. We did not have information from NREL about use of auxiliary power units
from any of the trucks in the Fleet DNA data, so we used the 80 percent extended idling value
for pre-2010 model year trucks which assumes no APU usage as presented in Table 11-2.
Hotelling Rate =
(Extended Idle Hours\( Total Miles Traveled \ / Hotelling Hours \
\Total Miles Traveled) \Restricted Access Miles Traveled) \Extended Idle Hours)
Figure 11-1 compares the hotelling rate in MOVES3 derived from NREL Fleet DNA, with the
default value used in MOVES2014 for the 2014 NEI version 281 and two other studies. Lutsey et
al.82 presented data from a nationwide truck surveyr and NCHRP 08-10183 conducted an
analysis of an instrumented truck dataset with 300 trucks.s The hotelling rate was unchanged
between MOVES3 and MOVES5.
r Lutsey reported average idling hours and driving hours per day. Using default national hours driving and
restricted access miles driven reported in Table 11-1 of the MOVES2014 Population and Activity Report, we
derived an estimate of extended idle hours per restricted access miles. We also used the ratio of hotelling hours to
extended idle hours as was done in Equation 11-2.
s Equation 11-2 was also used to calculate the hotelling rates from the data reported from NCHRP 08-101. The
definition of hotelling for the NCHRP 08-101 data was idling between 8 and 16 hours of duration, which is different
than used by the NREL analysis.
Equation
11-2
7.2 Hotelling Hours
1000 Restriced Access Miles Traveled
109
-------
30.00 :
25.00 :
20.00 :
15.00 :
(0
i_
Z5
i 10.00 :
5.00 :
0.00 :
Figure 11-1. Hotelling hours per 1000 miles driven on freeways compared across different
datasets.
In MOVES, the national rate of hotelling hours per mile of restricted access roadway VMT is
stored in the HotellingCalendarYear table for each calendar year. When the hotelling rate is
applied, it is multiplied by the rural and urban restricted access VMT by long-haul combination
trucks to estimate the default hotelling hours for any location, month or day. In MOVES, the
national rate of hotelling hours per mile of restricted access roadway VMT is stored in the
HotellingCalendarYear table for each calendar year. When the hotelling rate is applied, it is
multiplied by the rural and urban restricted access VMT by long-haul combination trucks to
estimate the default hotelling hours for any location, month or day.
The County Data Manager includes the HotellingHours table which provides the opportunity for
states and other users to provide their own estimates of hotelling hours specific to their
location and time. Whenever possible, states and local areas should obtain and use more
accurate local estimates of hotelling hours when modeling local areas.
The allocation of hotelling to specific hours of the day is described below in Section 13.5.
27.3
11.0
I
1
.2
6.3
1
MOVES2014
(2014 NEI v2)
MOVES3
(NREL Fleet DNA)
Lutsey (UC Davis) et
al. (2004)
NCHRP 08-101
110
-------
12.Engine Start Activity
Immediately following the start of an internal combustion engine, fuel is inefficiently burned
due to the relatively cool temperature of the engine and the need to provide excess fuel to
promote combustion. During this time, the quantity and profile of the pollutants generated by
the engine are significantly different than when the running engine is fully warm. Additionally,
the after-treatment technology employed on modern vehicles often requires time to become
fully functional. For these reasons, MOVES accounts for the effects of engine starts separately
from the estimates for hot running emissions.
The temperature of the engine and after-treatment systems depend not only on ambient
temperature, but the time since the last engine operation (soak time) as discussed in the light-
duty17 and heavy-duty18 emission rate reports. MOVES accounts for the soak time using "soak
time operating modes." The distribution of the soak times for engine starts can have a
significant effect on the emissions estimated for trips.
MOVES uses the following set of tables in the default database to determine the default
number of starts, soak times and their temporal distributions:
• StartsPerDayPerVehicle
• StartsAge Adjustment
• StartsHourFraction
• StartsMonthAdjust
• StartsOpModeDistribution
The StartsPerDayPerVehicle table contains a factor (StartsPerDayPerVehicle) which, when
multiplied by the total number of vehicles of a given source type calculates the number of starts
in a day. The StartsPerDayPerVehicle factor represents the average starts per day for each
sourcetype and day type (weekday/weekend)
Starts vary by source type, day type and vehicle age to account for the lower average start
activity that is expected to occur as vehicles age (see Section 12.1.1 for light-duty and Section
12.2.3 for heavy-duty). Figure 12-1 shows the calculation of starts per day per vehicle by vehicle
age for light-duty vehicles. Note that the age 0 starts per day are greater than the fleet-average
starts per day, and the starts at age 40 are lower than the fleet-average starts per day.
MOVES accounts for the effect of age using the ageAjustment factors stored in the
StartsAgeAdjustment table. This table stores the number of starts by vehicle age within each
sourcetype, relative to the number of starts at age 0. All of the ageAdjustment factors in
MOVES are based on the mileage accumulation rates (discussed in Section 6.2). By using the
mileage accumulation rates to derive the start ageAdjustement factors, we are assuming that
the starts per mile is constant over the life of the vehicle. In other words, as vehicles travel
fewer miles per day as they age, they similarly conduct fewer starts. The ageAdjustment factors
for each source type are set equal to one at age zero, and decrease from one as the age
increases, reflecting relatively lower starts as the vehicles age. MOVES does not use the
111
-------
absolute values in this table, but scales the ageAdjustment factors in conjunction with the
source type age distributions of the MOVES run (Section 6.1) such that the average starts
reported in in the StartsPerDayPerVehicle table is conserved. Using this method, MOVES
estimates starts by vehicle age without having the default or input age distribution impact the
estimated number of starts. However, the StartsPerDayPerVehicle factor value stored in the
StartsPerDayPerVehicle is intended to be representative of the fleet-average starts, and we
consider the age distributions when estimating these fleet-average starts as discussed in the
following subsections.
The StartsMonthAdjust table contains the monthAdjust factor which adjusts the starts per day
to reflect monthly variation in the number of engine starts (see Section 12.1.2.2 for light-duty
and Section 12.2.3.2 for heavy-duty). The monthAdjustment is used as a raw multiplicative
factor, with values greater and less than one. Unlike the startsageadjustment table, MOVES
does not scale the monthAdjustment factors to conserve starts for each model year. The
average monthAdjust values across all 12 months is one, so the annual number of starts
estimated by MOVES is consistent with the values in the StartsPerDayPerVehicle table.
However, the numbers of starts for a given month vary from the values in the
StartsPerDayPerVehicle table according to the monthAdjustment factors.
The StartsHourFraction distributes the starts in a day to the hours of the day. The
allocationFraction value varies by source type, day type and hour of the day (see Section
12.1.2.1 for light-duty and Section 12.2.3.1 for heavy-duty).
The StartsOpModeDistribution table contains the distribution of engine start soak times for
each source type, age, day type and hour of the day (see Section 12.1.3 for light-duty and
Section 12.2.4 for heavy-duty).
MOVES allows users to update the starts table if they have more representative data for their
purposes. MOVES provides additional start input tables and flexibilities for entering starts as
described in the Technical Guidance2.
The data inputs for motorcycles and motorhomes for the four start tables are discussed in
Section 12.3.
As discussed in Section 13.4, the MOVES2014 SampleVehicleTrip table is still used in MOVES5
for estimating evaporative emission activity. Thus, the number and time of starts used to
estimate start emissions is inconsistent with the trips and parking time used for evaporative
emissions in MOVES3 and later versions. While we think the impact of these inconsistencies is
small, we may consider addressing this conflict in future versions of MOVES.
12.1. Light-Duty Start Activity
Light-duty start activity are calculated from the same sample of vehicles from the Verizon
Telematics data discussed in Section 10.2.1.
112
-------
The vehicle starts input format was substantially updated for M0VES3 to better allow inputs
based on the summary activity from large telematic datasets. In addition, the start inputs have
been updated to account for differences in start activity by month, day type, hour of day, and
vehicle age. To calculate the national average light-duty starts per day for MOVES, we
calculated the average starts from a set of telematics data obtained from Verizon (discussed in
Section 10.2.1) and adjusted this average to account for vehicle age. There were no additional
updates for MOVES5.
Table 12-1 below shows the starts per day per vehicle derived from the Verizon telematics
dataset for passenger cars (sourceTypelD 21) and passenger trucks (sourceTypelDs 31) and by
weekend days (daylD 2) and weekdays (daylD 5). We calculated a weighted-average starts per
day per vehicle from the Verizon dataset using the regional populations from each state
sampled (California New Jersey, Illinois, Georgia and Colorado) as documented in Table 10-2.
The values shown for passenger trucks (sourceTypelD 31) are also being used for light
commercial trucks (sourceTypelD 32).
Next, we calculated the average age of the vehicles in the Verizon dataset, using the model year
for each vehicle stored in the the vehicle metadata file from all the vehicles in the Verizon
dataset. We assumed the base year = 2015.6 (5 months of the Verizon dataset were in 2015
and 6 months occurred in 2016). We calculated the average vehicle age for each vehicle using
Equation 12-1.
Age = Base Year (2015.6) — Average Model Year
Equation 12-1
We then calculated the average age for each state included in the dataset and then calculated a
Verizon weighted-average shown in Table 12-1 using the regional populations used previously
(Table 10-2)
Table 12-1 National Average Starts per Day per Vehicle for Light-duty Vehicles based on
Verizon Telematics data per Vehicle
Source Type
Source
Type ID
Verizon
weighted
average age
(years)
MOVES3 CY
2016
average age
(years)
Day of the
Week
Verizon
weighted
average starts
per vehicle
per day
Calculated national
average starts per
day per vehicle
Passenger Cars
21
7.3
9.55
Weekend
3.36
3.13
Weekday
3.96
3.68
Passenger Trucks
31
8.54
10.1
Weekend
3.49
3.32
Weekday
4.09
3.89
Light-Commercial
Trucks
32
8.54
8.47
Weekend
3.49
3.52
Weekday
4.09
4.13
113
-------
Next, we adjusted the starts for each vehicle age. We could not use the Verizon data directly
because it did not include a full range of vehicle ages. Instead, we used factors derived from
the mileage accumulation rates as discussed in the beginning of Section 12. We scaled the age
adjustment factors, such that at the average age (e.g., 7.3 years for passenger cars), the starts
per day is equal to the average estimated from Verizon (e.g., 3.96 per day for weekdays for
passenger cars). The resulting starts per day by age for light-duty vehicles are presented in
Figure 12-1. The starts per day for age 0 are higher than the Verizon weighted average starts
per day, while the starts at age 30 are substantially lower.
a> 5
~o
0 4
>
>
CL
(n
t1
TO
Passenger Car
Passenger Truck
Light Commercial Truck
\
n\
%
*«r
%
\
s X.
%
%
*
*
%
%
%
% \
%
*
*
%
* - - - ¦
C
10 20 30 0 10
20 30 0 10 20 30
vehicle age
dayName — Weekdays ¦ ' Weekend
Figure 12-1. Starts per day per vehicle by vehicle age calculated from the Verizon dataset and
MOVES ageAdjustment factors
We then used Equation 12-2 to calculate the MOVES age-weighted average starts per vehicle
per day using the starts per day per vehicle by age calculated in Figure 12-1 and the 2016
default age distributions in MOVES. The purpose of this calculation is to adjust the average
starts per day from the Verizon sample to represent the nation, given that the national age
distribution is different than the age distribution of vehicles sampled in the Verizon datasets. As
shown in Table 12-1, the age in MOVES3 for CY 2016 passenger cars and passenger trucks is
older than in the Verizon dataset, while the average age of light commercial trucks is slightly
older in MOVES3 than in the Verizon dataset. We adjusted the average starts using the 2016
default age distribution because the Verizon dataset was conducted in 2015-2016. Note that
since this analysis was performed for MOVES3, it used age distributions covering ages 0-30+.
National Average Starts per Vehicle per Day
30
= ^ (Starts per Day Per Vehicle)age x ageFractionage
Equation 12-2
age=0
114
-------
Table 12-2 demonstrates the calculation of Equation 12-2 for passenger cars on weekdays.
Table 12-1 shows the calculated national average starts per vehicle per day which are used in
MOVES. The national average starts per vehicle day in 12-1 are used to estimate the average
starts for these source types and day types for all calendar years in MOVES.
Table 12-2. Calculation of the National Average Starts per Vehicle per Day for Passenger Cars
(SourceType21
on Weekdays (DaylD 5)
Vehicle age (agelD)
Starts per Day Per
Vehicle by Age
CY 2016 Age Distribution
(ageFraction)
Starts per Day per
Vehicle x ageFraction
0
4.76
0.061
0.29
1
4.67
0.066
0.31
2
4.57
0.067
0.31
3
4.47
0.062
0.28
4
4.36
0.056
0.24
5
4.24
0.043
0.18
6
4.12
0.044
0.18
7
4.00
0.040
0.16
8
3.87
0.050
0.19
9
3.74
0.055
0.21
10
3.61
0.051
0.19
11
3.48
0.050
0.17
12
3.35
0.046
0.15
13
3.22
0.045
0.15
14
3.09
0.040
0.12
15
2.97
0.034
0.10
16
2.84
0.033
0.09
17
2.72
0.025
0.07
18
2.61
0.021
0.05
19
2.50
0.017
0.04
20
2.39
0.013
0.03
21
2.30
0.012
0.03
22
2.21
0.009
0.02
23
2.12
0.007
0.01
24
2.05
0.006
0.01
25
1.99
0.005
0.01
26
1.93
0.004
0.01
27
1.89
0.003
0.01
28
1.86
0.003
0.00
29
1.84
0.002
0.00
30
1.84
0.030
0.05
National Age-Weighted Average Starts per Vehicle per Day =
3.68
115
-------
12.1.2. Temporal Distributions
There were no updates to default temporal distibutions in MOVES5.
12.1.2.1. Hourly Distribution
The number of starts varies by hour of day. National values for the distribution of starts per day
by hour for passenger cars and light-duty trucks were calculated from the five-state Verizon
sample data described above in Section 10.2.1. The resulting national defaults for start
distribution in MOVES3 and later are illustrated in Figure 12-2. The start fraction values for
hourlDs 1 through 24 sum to 1.0 for a given sourceTypelD and daylD combination. The start
distribution curve in MOVES3 is much smoother than the start distribution based on the
SampleVehicleTrip table in MOVES2014 owing to the larger sample size of the Verizon data.
However, the overall trends are similar.
MOVES3 MOVES3 - - MOVES2014 MOVES2014
Weekday Weekend Weekday Weekend
0.12
0.1
0.08
g
Jj 0.06
S
00
0.04
0.02
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Houi" ED
Figure 12-2 Start distribution for source type 21: MOVES3 derived from Verizon data vs.
MOVES2014
12.1.2.2. Monthly Distribution
For MOVES, we assume that the starts/mile is the same across months. We use the same
monthly distribution for starts in the MonthAdjust table as for VMT in the MonthVMTFraction
table discussed in Section 13.1. Light-duty vehicles and all other source types (except
116
-------
motorcycles) follow the same monthly variation, with slightly elevated starts during the
summer months, and corresponding decrease in starts in the winter.
12.1.3. Start Soak Distributions
As discussed in the beginning of Section 12, soak times are binned into different operating
modes, shown in Table 12-3. The fraction of starts assigned to each soak bin is the "soak
distribution." The light-duty soak distributions derived from Verizon differ by source type, day
type and hour of the day.
The engine soak time distributions for all source types are available in the OpModeDistribution
table of the default database (see Section 10.2) for the national default value calculation
method from the Verizon sample data and Table 10-1 for the number of sample trips used.
Figure 12-3 illustrates the MOVES national default soak distribution for a weekday for
passenger cars (sourceTypelD 21).
Table 12-3 MOVES engine soak operating modes
opModelD
Description
101
Soak Time < 6 minutes
102
6 minutes <= Soak Time < 30 minutes
103
30 minutes <= Soak Time < 60 minutes
104
60 minutes <= Soak Time < 90 minutes
105
90 minutes <= Soak Time < 120 minutes
106
120 minutes <= Soak Time < 360 minutes
107
360 minutes <= Soak Time < 720 minutes
108
720 minutes <= Soak Time
117
-------
¦ 101 ¦ 102 ¦ 103 ¦ 104 ¦ 105 ¦ 106 ¦ 107 ¦ 108
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
HourlD
Figure 12-3 MOVES national average engine soak distribution for source type 21 and weekday
MOVES has the capability to model different soak distributions by vehicle age, but we are
currently using the same soak distribution across all vehicle ages. In general, as vehicles age, we
would expect less vehicle starts on average and a soak distribution to shift towards longer soak
times. Access to a large data covering a wider range of ages would help us better quantify this.
12.2. Heavy-Duty Start Activity
Like light-duty vehicles, starts from heavy-duty vehicles can be an important contributor to
emission inventories (e.g., THC and NOx). Additionally, heavy-duty diesel aftertreatment
technologies such as selective catalytic reduction (SCR) systems are also not fully active at
controlling NOx emissions below the catalyst light-off temperature.
Compared to light-duty vehicles, less data are available on heavy-duty vehicle start activity and
there are more subgroups of vehicles with potentially unique activity patterns. For example,
delivery vehicles have different start and soak patterns than long-haul trucks. For MOVES3 and
later, data that covers a wider range of heavy-duty vocations was available. The engine start
analysis below was applied to the same NREL Fleet DNA dataset used in the off-network idle
analysis discussed in Section 10. We are aware of the additional heavy-duty activity data
118
-------
collected by CE-CERT (also described in Section 10) and we expect to incorporate the data in a
future version of MOVES.
12.2.1. Heavy-Duty Engine Start Activity Data Processing
Starts were identified in the data using the data channel for engine speed, measured in
revolutions per minute (RPM). All the instances when the engine speed transitioned from zero
to greater than zero were considered new starts. If the data logger was installed but did not
record any activity, the start fraction is zero; however, if the data logger was not installed on a
specific day type, those values were denoted as "nan" (not-a-number) and were removed from
the analysis. The sum of the hourly fractions across all hours of the day is one.
The number of starts per day was calculated on a per vehicle basis and averaged equally across
all vehicles as shown in Equation 12-3. Thus, vehicles that start frequently or infrequently are
equally weighted in the average starts per day.
Similar to the off-network idle discussion in Section 10, we applied a sum-over-sum approach to
our hourly start fractions. Using Equation 12-4, a start fraction for each hour was calculated by
dividing the daily average starts-per-hour by the average starts-per-day for each combination of
sourceType and daylD. This sum-over-sum approach normalizes the recorded start activity by
the amount of time each vehicle was instrumented and weights the average start fraction
towards the vehicles with the most daily-average starts.*
Equation 12-3
/' = Vehicle ID within a given sourceType, s
days/ = days within a given daylD, d, when vehicle, is instrumented
n = number of VehiclelDs withing a given sourceType, s
rstartshi . \
^ V ' day Si J
startshi
Start fractionhsd =
Equation 12-4
' We evaluated several approaches for calculating average start and soak fractions using the Fleet DNA data. The
equations presented in this section are equivalent to the equations labeled "Method 3 'normalized sum over sum'"
in Appendix I. Appendix I includes an overview and a comparison of the calculation approaches.
119
-------
h = hour of the day
i = Vehicle ID within a given sourceType, s
days, = days within a given daylD, d, when vehicle, is instrumented
Vehicle soak is defined as the time difference between when an engine stops and the next time
the engine starts, as shown in Equation 12-5. The engine stop is defined as the time when
engine speed transitions from greater than zero to zero and engine start is defined as the time
when engine speed transitions from zero to greater than zero.
Every start was assigned a soak opModelD based on the definitions in Table 12-3.u We then
calculated the average soak fraction, using a normalized sum-over-sum approach like we did for
the start fraction.* For each vehicle, hour and daytype, an average number of starts by soak
length was calculated by summing the number of starts matching each soak opModelD for each
hourlD and daylD and dividing by the number of unique days of measurement for that vehicle.
The hourly soak fraction distribution for each opModelD, sourceType and daylD was then
calculated using Equation 12-6. The sum of the eight opModelD soak fractions will equal 1.0 for
each combination of daylD, hourlD and sourceTypelD.
soak time = engine stop time — engine start time
Equation 12-5
Soak
h = hour of the day
i = Vehicle ID within a given sourceType, s
o = operating mode/soak length
days, = days within a given day ID, d, vehicle, is instrumented
Equation 12-6
12.2.2. Starts Per Vehicle Per Day
u The first start identified for each vehicle was not considered when calculating soak time due to lack of a previous
recorded stop time.
120
-------
As seen in Table 10-3, several heavy-duty source types were not available in the Fleet DNA
database at the time of this analysis. In the future, we hope to update the analysis using both
Fleet DNA and CE-CERT datasets. In the interim, we assumed the start behavior of the missing
vehicles closely matched others. We chose to use the transit bus (sourceTypelD 42) to
represent other bus (sourceTypelD 41), used the single unit short-haul data from the weekend
(sourceTypelD 52) to represent both the weekday and weekend data of the single unit long-
haul trucks (sourceTypelD 53) and continued to use the same starts per day for motorhomes
(sourceTypelD 54) as in MOVES2014 (See Table 13-8).
None of the school buses (sourceTypelD 43) instrumented in the Fleet DNA dataset operated
on the weekend, so there is no data for that daylD. In Section 10, we applied the weekday
school bus off-network idle data for the weekend data, assuming the idle behavior of buses was
similar regardless of day type. We opted to retain the zero starts-per-day value for weekends
(daylD 2), assuming the frequency of school bus starts differed between weekends and
weekdays.
Figure 12-4 and Figure 12-5 show the starts-per-day values for weekends and weekdays,
respectively.
Daily Starts, Weekends
Vehicle Description / MOVES SourceType
Figure 12-4 Weekend starts per day for heavy-duty source types based on data from NREL's
Fleet DNA database
121
-------
Daily Starts, Weekdays
40 -1
O
35 -
1c
30 -
.
03
20 -
Q
15 -
CD
10 -
Q_
CO
I—
5 -
TO
CO
0 -
Other Bus
41
Transit Bus
42
School Bus
43
Refuse
Truck
51
Single Unit,
Short
52
Single Unit,
Long
53
Combo,
Short
61
Combo,
Long
62
Vehicle Description / MOVES SourceType
Figure 12-5 Weekday starts per day for heavy-duty source typesbased on data from NREL's
Fleet DNA database
As shown in Figure 12-5, the single unit short-haul trucks (sourceTypelD 52) have significantly
more starts on weekdays. To understand this, we evaluated the impact of the vehicles'
vocations on their start behavior. Figure 12-6 shows that the parcel delivery vocation
contributes many more starts than the other vocations. While we did see differences in starts
activity due to vehicle vocation, we did not account for vocation differences when calculating
starts for MOVES because we could not identify a means to map the vocations represented in
this dataset to a nationally-representative population of vocations. We plan to revisit these
estimates in future versions of MOVES.
122
-------
Daily Starts by Vocation, Weekdays
Single-Unit Short-Haul Trucks
$
0
0
0
O
>
>
>
Q.
15
"0
"0
>
o
Q
O
>
o
~0
o
c
0
"O
o
c
(/)
C
o
03
Q_
_l
U_
CD
C
0
0
o
c
o
o
0
>
0
Q
CD
"O
"O
0
_c
CO
c
03
I-
0
C
03
Q_
O
o
Q_
E
Q
Vehicle Vocation
Figure 12-6 Vocation impacts on weekday starts-per-day for heavy-duty, single unit short-
haul vehicles (sourceType 52) based on data from NREL's Fleet DNA database
As discussed in the beginning of Section 12, the startsPerDayPerVehicle factor stored in the
MOVES startsPerDayPerVehicle table represents the national average starts per day by
sourcetype and day of the week. We developed adjusted the starts measured from Fleet DNA
to be consistent with the national average age distribution in MOVES.
First, we adjusted the heavy-duty start values obtained from Fleet DNA using the
ageAdjustment factors (derived from the mileage accumulation rates in Section 6.2), by
assuming that the Fleet DNA starts are representative of vehicles at age 0. We assumed the
Fleet DNA starts are representative of activity at age 0 because:
• Of the vehicles with a recorded age (112 out of 415 vehicles in the Fleet DNA database),
most are younger than 3 years of age.
• NREL has informed the US EPA that vehicles chosen to be instrumented in the Fleet DNA
database tend to be active vehicles.
Figure 12-7 displays the resulting starts per day per vehicle across all ages for heavy-duty
vehicles calculated using these assumptions. Note that the starts per day per vehicle at age 0
are equivalent to the average values reported from the Fleet DNA database, while the starts per
day for age 30 source types are significantly lower.
123
-------
Buses
41
Buses
42
Buses
43
7.5
5.0-
2.5
0.0-
4i
2
0-L
Single-Unit Trucks
Single-Unit Trucks
Single-Unit Trucks
— Weekdays
- 1 Weekend
10 20
vehicle age
Figure 12-7, Starts per day per vehicle by vehicle age calculated from the Fleet DNA dataset
and MOVES ageAdjustment factors
Next, we calculated the national average starts per day for heavy-duty vehicles using Equation
12-2 with the starts per day per vehicle by age in Figure 12-7 and the 2014 heavy-duty default
age distributions in MOVES3. We used the 2014 age distributions because it was the calendar
year with the most vehicle measurements in the FleetDNA dataset; the average age from the
MOVES 2014 age distributions are shown in Table 12-4. The resulting national average starts
per day per vehicle are also displayed in Table 12-4, which are significantly lower than the
average starts per day as measured from the FleetDNA database. Note that since this analysis
was performed for MOVES3, it used age distributions covering ages 0-30+.
124
-------
Table 12-4 National Average Starts Per Day Per Vehicle for Heavy-duty Vehicles based on data
from NREL's Fleet DNA database
Source Type
SourceTypelD
MOVES3 CY
2014
Average Age
(years)
Day of the
Week
FleetDNA
Starts per
day per
vehicle
Calculated
national
average
starts per
day per
vehicle
Other Bus
41
10.4
Weekend
2.64
1.93
Weekday
8.70
6.38
Transit Bus
42
6.5
Weekend
2.64
2.16
Weekday
8.70
7.13
School Bus
43
10.3
Weekend
0.00
0.00
Weekday
5.41
3.98
Refuse Truck
51
11.7
Weekend
0.16
0.10
Weekday
2.74
1.71
Single Unit Short-haul
52
11.8
Weekend
2.58
1.41
Weekday
36.26
19.86
Single Unit Long-haul
53
11.8
Weekend
2.58
1.33
Weekday
2.58
1.33
Combination Short-
haul
61
12.0
Weekend
2.82
1.35
Weekday
12.25
5.87
Combination Long-
haul
62
10.5
Weekend
0.60
0.37
Weekday
0.81
0.51
12.2.3. Temporal Distribution
12.2.3.1. Hourly Distribution
This section describes the temporal distribution of starts (also referred to as the start fractions)
for heavy-duty vehicles in MOVES based on data from NREL's Fleet DNA. As seen in Table 10-3,
several heavy-duty source types were not available in Fleet DNA at the time of the analysis. We
expect data collected by CE-CERT to cover more of the source types and daylDs and we hope to
update the analysis using both Fleet DNA and CE-CERT datasets. In the interim, we assumed
the start behavior of the missing vehicles closely matched others, as described when the figures
are presented below.
The Fleet DNA dataset did not conain any information from buses meeting the MOVES
definition of "other buses." We assumed the start distributions from transit bus (sourceTypelD
42) represented other buses (sourceTypelDs 41) for both weekends and weekdays. Figure 12-8
shows the resulting starts distribution for these two bus types.
125
-------
Transit Buses (sourceType 42) & Other Buses (sourceType 41)
0.12
0.10
c 0.08
o
£ 0.06
t
TO
« 0.04
0.02
0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-8 Start fraction temporal distribution for transit buses (sourceType 42) and other
buses (sourceType 41) based on data from NREL's Fleet DNA database
The school buses and refuse vehicles in the Fleet DNA dataset did not operate during certain
hours of the day. To avoid tables with zero values for those hours, we averaged adjacent blocks
of time, so those zeros were replaced with very small, nonzero values. The school buses in this
dataset did not operate from the hours of 8:00 PM to 6:00 AM on weekdays. We replaced the
zeros in those hours with 0.0008, which was the average start fraction from 6:00 PM through
6:00 AM, as depicted by the red boxes in Figure 12-9. The refuse trucks did not operate on
weekends from 7:00 PM to 4:00 AM and no data were collected in the hour between 6:00 AM
and 7:00 AM. The missing night hours' data were replaced with 0.01 (the average of the data
for 7:00 PM to 4:00 AM) and the missing 6:00 AM data point was replaced with 0.07 (the
average of the 4:00 AM to 7:00 PM data). For each case, we renormalized the results once the
zeros were replaced, so the start fractions across all 24 hours of the day continued to sum to
1.0.
Mji
~ Weekends
¦ Weekdays
i in 11
126
-------
School Bus, Weekdays
0.25
0.20
C
1 0.15
2
% 0.10
CO
0.05
0.00
¦
1
I
l
¦
1
1
¦
1
¦
1 2
3
4
5
6
7
8
9
10 11
12 13 14 15 16 17
18 19 20 21 22 23 24
Refuse Trucks, Weekends
0.30
0.25
o 0.20
o
ro
•c
CO
0.15
0.10
0.05
0.00
1
¦
1
1
1
1
1
1
1
¦
1
1
1
1
1
1
1
1
2
3
4
5
6
7
8
9
10 11
12 13 14 15 16
17 18 19 20 21 22 23 24
Figure 12-9 Approach for renormalizing the start fraction results to avoid zeros in hours when
no data were collected. The zero-value start fraction in the red boxes were replaced with the
average start fractions from the range of hours in the red boxes. For refuse trucks, the 6:00
AM missing datapoint was replaced with an average of the hours not outlined in red boxes.
Figure 12-10 and Figure 12-11 show the resulting start fractions by hour for school buses and
refuse trucks, respectively. Note that, in MOVES, school buses (sourceTypelD 43) have zero
starts per day on weekends, and none of the school buses instrumented in the Fleet DNA
dataset operated on the weekend, so we applied the start fractions from the weekday school
bus data to weekends.
127
-------
School Buses (sourceType 43)
0.25 i
0.20 -
2 0.15 -
o
TO
TO
-t—'
CO
0.10 -
0.05 -
0.00
II rr
~ Weekends
¦ Weekdays
J1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-10 Start fractions by hour for school buses (sourceType 43) based on data from
NREL's Fleet DNA database
Refuse Trucks (sourceType 51)
0.25 n
0.20 -
5 0.15 -
O
CO
I 0.10 h
CO
0.05 -
0.00
EL
~ Weekends
¦ Weekdays
lljjjwllrlrlr
I ml nl rl nl _Lz
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-11 Start fractions by hour for refuse trucks (sourceType 51) based on data from
NREL's Fleet DNA database
The Fleet DNA dataset did not contain any single unit long-haul trucks (sourceType 53) so we
assumed their start distribution was similar to the single short-haul trucks (sourceType 52).
128
-------
Figure 12-12 shows the start distribution applied to both single unit truck types for weekends
and weekdays.
0.12
0.10
c 0.08
o
H—»
£ 0.06
¦e
TO
w 0.04
0.02
0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-12 Start fraction by hour for single unit short-haul trucks (sourceType 52) and single
unit long-haul trucks (sourceType 53) based on data from NREL's Fleet DNA database
Additional consideration was given before using the FleetDNA data to populate the hourly
fraction tables. The Fleet DNA dataset contained many combination long-haul trucks
(sourceType 62) and NREL staff are confident that the average idle data described in Section 10
and average starts-per-day data described earlier in this section represent the activity of
combination long-haul trucks. However, NREL believes there was a time zone-related logging
error when the data were reported to NREL. Most of the data from the 131 combination long-
haul trucks in the Fleet DNA dataset were collected by an industry partner and NREL was unable
to accurately confirm which time zone the activity was recorded in. The data consistently
showed that the trucks operated more at night with hotelling during the day, which conflicted
with other data sources as discussed in Section 13.5. Figure 12-13 shows the original Fleet DNA
data for long-haul combination trucks that was not applied in MOVES due to the possible time
misalignment. Instead, we assumed the start distribution from short-haul combination trucks
was a better representation. Figure 12-14 shows the starts distribution that was applied in
MOVES for both short- (sourceType 61) and long-haul combination trucks (sourceType 62).
Single-Unit Trucks (sourceTypes 52 & 53)
JL
~ Weekends
¦ Weekdays
i
129
-------
0.14
0.12
2 0.08
LL
ra 0.06
CO
0.04
0.02
0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-13 Start fractions from on NREL's Fleet DNA database that were not applied for
combination long-haul trucks (sourceType 62); we suspect a time misalignment in the the
data
Combination Trucks (sourceTypes 61 & 62)
0.12
0.10
c 0.08
o
£ 0.06
t
ra
w 0.04
0.02
0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-14 Start fraction temporal distribution for combination short-haul trucks
(sourceType 61) and combination long-haul trucks (sourceType 62) based on data from NREL's
Fleet DNA database
Suspected Time Misalignment for
Combination Long-Haul Trucks (sourceType 62)
~ Weekends
r.
¦ Weekdays
-
-1
p.
ill
in
i
i
¦
pi
i-i
p.
n
ill
nl
nl
ni
na
nl
wii
1
1
1
1
~ Weekends
¦ Weekdays
130
-------
12.2.3.2. Monthly Distribution
In MOVES, we assume that the starts/mile is the same across all months. We use the same
monthly distribution for starts in the MonthAdjust table as for the VMT in the
MonthVMTFraction table discussed in Section 13.1. Heavy-duty vehicles follow the same
monthly variation as light-duty vehicles, with slightly elevated starts during the summer
months, and corresponding decrease in starts in the winter.
12.2.4. Start Soak Distributions
This section describes the heavy-duty vehicles' soak distributions in MOVES. As seen in Table
10-3, several heavy-duty source types were not available in the Fleet DNA database at the time
of this analysis. We hope to update the analysis using both Fleet DNA and CE-CERT datasets. In
the interim, we applied several assumptions for the soak behavior as described below.
The Fleet DNA dataset did not contain any information from buses meeting the MOVES
definition of "other buses". We assumed the start distributions from transit bus (sourceTypelD
42) represented other buses (sourceTypelDs 41) for both weekends and weekdays. Figure 12-15
and Figure 12-16 show the resulting starts distributions for these two bus types on weekends
and weekdays, respectively.
Transit & Other Buses (sourceTypes 42 & 41), Weekends
100%
90%
80%
£=
70%
o
o
60%
TO
¦.
LL
50%
(0
o
40%
CO
30%
20%
10%
0%
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-15 Weekend soak distributions transit buses (sourceType 42) and other buses
(sourceType 41) based on data from NREL's Fleet DNA database
131
-------
Transit & Other Buses (sourceTypes 42 & 41), Weekdays
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-16 Weekday soak distributions transit buses (sourceType 42) and other buses
(sourceType 41) based on data from NREL's Fleet DMA database
As mentioned previously in the start distribution discussion, school buses (sourceTypelD 43) in
this dataset did not operate from 8:00 PM to 6:00 AM on weekdays. For soak distribution, we
replaced these hours with the average hourly soaks over the period from 6:00 PM through 6:00
AM. The school buses in the Fleet DNA dataset did not operate on the weekends, so we applied
the weekday school bus soak distribution to weekends. Figure 12-17 shows the soak
distribution applied to school buses for both weekends and weekdays.
School Buses (sourceType 43), Weekdays & Weekends
100%
90%
l!l
OpMode
80%
¦ 108
c 70%
¦ 107
| 60%
¦ 106
OJ
¦ 105
it 50%
| 40%
¦ 104
W 30%
¦ 103
20%
III llll
¦ 102
10%
¦ 101
0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-17 Weekend and weekday soak distributions for school buses (sourceType 43)
based on data from NREL's Fleet DMA database
132
-------
Refuse trucks (sourceType 51) in the Fleet DNA dataset did not operate on weekends from 7:00
PM to 4:00 AM and no data were collected in the hour between 6:00 AM and 7:00 AM. The
missing night hours' data were replaced with the average hourly soaks over the period from
7:00 PM to 4:00 AM and the missing 6:00 AM data point was replaced with the average hourly
soaks from 4:00 AM to 7:00 PM. Figure 12-18 and Figure 12-19 show the soak distributions for
refuse trucks on weekends and weekdays, respectively.
Refuse Trucks (sourceType 51), Weekends
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-18 Weekend soak distributions for refuse trucks (sourceType 51) based on data
from NREL's Fleet DNA database
133
-------
Refuse Trucks (sourceType 51), Weekdays
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-19 Weekday soak distributions for refuse trucks (sourceType 51) based on data
from NREL's Fleet DNA database
The Fleet DNA dataset did not contain any single unit long-haul trucks (sourceType 53) and we
assumed their soak distribution was similar to the single unit short-haul trucks (sourceType 52).
Figure 12-20 and Figure 12-21 show the soak distributions applied to both single unit truck
types for weekends and weekdays, respectively.
Single-Unit Trucks (sourceTypes 52 & 53), Weekends
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-20 Weekend soak distributions for single unit short-haul trucks (sourceType 52) and
single unit long-haul trucks (sourceType 53) based on data from NREL's Fleet DNA database
134
-------
Single-Unit Trucks (sourceTypes 52 & 53), Weekdays
100%
90%
80%
c
70%
o
o
60%
CO
1—
LL
50%
TO
o
40%
CO
30%
20%
10%
0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-21 Weekday soak distributions for single unit short-haul trucks (sourceType 52) and
single unit long-haul trucks (sourceType 53) based on data from NREL's Fleet DMA database
Figure 12-22 shows the weekend soak distribution that was applied in MOVES for combination
short-haul trucks (sourceType 61). Figure 12-23 shows the weekday soak distribution for the
same vehicles.
Combination Short-Haul Trucks (sourceType 61), Weekends
100%
90%
80%
c
70%
o
o
60%
ra
LL
50%
nj
o
40%
CO
30%
20%
10%
0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-22 Weekend soak distributions for combination short-haul trucks (sourceType 61)
based on data from NREL's Fleet DNA database
135
-------
Combination Short-Haul Trucks (sourceType 61), Weekdays
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-23 Weekday soak distributions for combination short-haul trucks (sourceType 61)
based on data from NREL's Fleet DNA database
As mentioned in the start distribution section, we believe there was a time zone-related logging
error for many of the combination long-haul trucks (sourceType 62) in the Fleet DNA dataset.
Consequently, we opted to apply the same average hourly soak distribution from each day type
across all hours of the day. The soak distributions applied for combination long-haul trucks on
weekends are shown in Figure 12-24, The weekday soak distrbituions are in Figure 12-25.
Combination Long-Haul Trucks (sourceType 62), Weekends
100%
90%
80%
c
70%
o
o
60%
ra
LL
50%
oi
o
40%
CO
30%
20%
10%
0%
123456789
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-24 Weekend soak distributions for combination long-haul trucks (sourceType 62)
applying the average hourly soak distribution from NREL's Fleet DNA database across all
hours of the day
136
-------
Combination Long-Haul Trucks (sourceType 62), Weekdays
100%
90%
80%
c
70%
o
o
60%
W
¦.—
LL
50%
CO
o
40%
CO
30%
20%
10%
0%
123456789
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-25 Weekday soak distributions for combination long-haul trucks (sourceType 62)
applying the average hourly soak distribution from NREL's Fleet DNA database across all
hours of the day
As mentioned for light-duty vehicles, MOVES has the capability to model different soak
distributions by vehicle age. We are currently using the heavy-duty soak distribution estimated
from Fleet DNA across all vehicle ages. In general, as vehicles age, we would expect fewer
vehicle starts and a soak distribution shift towards longer soak times. However, the available
data on heavy-duty vehicles at older ages is much more limited. Future work could evaluate the
dependency of soak distributions on vehicle age.
12.3. Motorcycle and Motorhome Starts
Motorcycle and motorhome data are not captured in the Verizon and Fleet DNA datasets used
to update the other source types. The data we used to model starts from motorcycles and
motorhomes is outlined in Table 12-5.
Table 12-5. Motorcycle and Motorhome Start Data
MOVES Table
Motorcycles
Motorhomes
(SourceTypelD 11)
(SourceTypelD 54)
startsPerDayPerVehicle
Starts from Table 13-8
Table 13-8 adjusted to
adjusted to represent CY
represent CY 2014 age
2014 age distribution
distribution
startsHourFraction
Passenger Cars(21)
Passenger Trucks(31)
startsOpmodeDistribution
Passenger Cars(21)
Passenger Trucks(31)
(soaks)
startsMonthAdjust
Table 13-2
Table 13-1
137
-------
For national average starts per day per vehicle, we used the starts per day estimated in
MOVES2014 as presented in Table 13-8. Because these start rates were calculated from
instrumented vehicle data, we assume these start rates are respresentative of active, age 0
vehicles. We thus followed similar steps to calculate national average starts per day per vehicle
as was conducted for heavy-duty vehicles above which used the same assumptions. We
calculated starts per day by vehicle age by applying the ageAdjustment factors to the start data
as shown in Figure 12-26.
a>
O 1.5
>
.
OJ
Q
£0.5
CL
(/)
r
ra ^ „
0 .0
Motorcycle
0.6
0.4-
0.2-
0.0-
Motor Home
—
dayName
Weekdays
Weekend
10
20
30
10
20
30
vehicle age
Figure 12-26. Starts per day per vehicle by vehicle age for motorcycles and motorhomes
calculating using MOVES ageAdjustment factors
We then calculated the national average starts per day for motorcycles and motorhomes using
Equation 12-2 with the starts per day per vehicle by age in Figure 12-26, Figure 12-7 and the
2014 heavy-duty default age distributions in MOVES3. We used the 2014 age distributions
because it is the year from which the source type age distributions are based in MOVES3
(Section 6.1). The resulting national average starts per day per vehicle for motorcycles and
motorhomes are displayed in Table 12-4, which are significantly lower than the age zero start
rate.
Table 12-6. National Average Starts Per Day Per Vehicle for Motorcycles and Motorhomes
Source Type
SourceTypelD
MOVES3 CY
2014 Average
Age (years)
Day of the
Week
Starts per
day per
vehicle at
age 0
Calculated
national average
starts per day
per vehicle
Motorcycle
11
10.5
Weekend
1.52
0.37
Weekday
0.45
0.11
Motorhome
54
15.0
Weekend
0.57
0.48
Weekday
0.57
0.47
The hourly distribution of starts (stored in the startsHourFraction table) for motorcycles is
assumed to be the same as for passenger cars. For motorhomes, the hourly distribution of
138
-------
starts is assumed to be the same as for passenger trucks, both of which are estimated from the
Verizon data base.Motorcyles soak distributions are the same as passenger cars and
motorhomes are the same as passenger trucks. We assume that the montly pattern of starts
(stored in the startsMonthAdjust table) follows the same pattern as VMT as described in in
Section 13.1. Thus, motorcycle have a pronounced increase in starts during summer months.
Motorhomes starts follow the monthly variation of all other source types, which are only
slightly elevated during the summer months.
139
-------
IS.Iemporal Distributions
MOVES is designed to estimate emissions for every hour of every day type in every month of
the year. This section describes how VMT is allocated to months of the year, the two day types
and to hours of the day. This section also addresses how sample vehicle trip data is used to
determine and allocate evaporative soak periods to hours of the day. Finally, this section
discusses the derivation of the allocation of hotelling activity for long-haul combination trucks.
See also the discussion of temporal allocations for off-network idle in Section 10 and for engine
starts in Section 12.
In MOVES, VMT are provided in terms of annual miles. These miles are allocated to months,
days and hours using allocation factors, either using default values or values provided by users.
Default values for most temporal VMT allocations are derived from a 1996 report from the
Office of Highway Information Management (OHIM).84 The report describes analysis of a
sample of 5,000 continuous traffic counters distributed throughout the United States. EPA
obtained the data from the report and used it to generate the VMT temporal distribution inputs
in the form needed for MOVES. This information has not been updated since MOVES2010b.
The OHIM report does not specify VMT by vehicle type, so MOVES uses the same values for all
source types, except motorcycles, as described below.
In MOVES, daily truck hotelling hours are calculated as proportional to VMT on restricted access
road types for long-haul combination trucks. However, the hours of hotelling activity in each
hour of the day are not proportional to VMT, as described in Section 13.5.
The temporal distributions for engine start are described in Section 12.1.2. These values are
stored in the StartsMonthAdjust and StartsHourFraction Tables. However, we have not yet
updated the data used to estimate vehicle parking time and associated evaporative emissions.
As in MOVES2014, the engine soak (parked) distributions for evaporative emissions are
calculated from vehicle activity data stored in the SampleVehicleDay and SampleVehicleTrip
tables of the MOVES database. The inconsistency between the updated activity defaults now
being used to calculate engine starts and soaks and the older defaults that MOVES will continue
using for evaporative emissions is not ideal. We plan to resolve this inconsistency in future
versions of MOVES when the code used for the calculation of evaporative emissions is updated.
The temporal allocation of vehicle activity will vary from location to location and EPA guidance
encourages states and local areas to determine their own local vehicle activity parameters for
use with MOVES. When EPA runs MOVES for air quality modeling purposes, we use local
activity data including temporal allocations as explained in EPA technical support documents.58
EPA plans to update the temporal allocations currently in MOVES using more recent data
sources, such as telematics data, as they become available.
140
-------
13.1. VMT Distribution by Month of the Year
In MOVES, when VMT is entered as an annual value, it is allocated to months of the year using
the factors in the MonthVMTFraction table. For MOVES, we modified the data from the OHIM
report to fit MOVES specifications. Table 13-1 shows VMT/day taken from the OHIM report
(Figure 2.2.1 "Travel by Month, 1970-1995"), normalized to one for January. The VMT per day
in Table 13-1 were used to calculate the fraction of total annual VMT in each month using the
number of days in each month, assuming a non-leap year (365 days). These monthly VMT
allocations are used for all source types, except motorcycles, as described below.
Table 13-1 MonthVMTFraction
Month
Normalized
MOVES
VMT/day
Distribution
January
1.0000
0.0731
February
1.0560
0.0697
March
1.1183
0.0817
April
1.1636
0.0823
May
1.1973
0.0875
June
1.2480
0.0883
July
1.2632
0.0923
August
1.2784
0.0934
September
1.1973
0.0847
October
1.1838
0.0865
November
1.1343
0.0802
December
1.0975
0.0802
Sum
1.0000
FHWA does not report monthly VMT information by vehicle classification. However, it is clear
that in many regions of the United States, motorcycles are driven much less frequently in the
winter months. For MOVES, an allocation for motorcycles was derived using monthly national
counts of fatal motorcycle crashes from the National Highway Traffic Safety Administration
Fatality Analysis System for 2010.85 This allocation increases motorcycle activity (and emissions)
in the summer months and decreases them in the winter compared to the other source types.
These default values in Table 13-2 for motorcycles are only a national average and do not
reflect the strong regional differences that would be expected due to climate.
141
-------
Table 13-2 Mont
lVMTFraction for motorcycles
Month
Month ID
Distribution
January
1
0.0262
February
2
0.0237
March
3
0.0583
April
4
0.1007
May
5
0.1194
June
6
0.1269
July
7
0.1333
August
8
0.1349
September
9
0.1132
October
10
0.0950
November
11
0.0442
December
12
0.0242
Sum
1.0000
The monthly allocation of VMT will vary from location to location and EPA guidance encourages
states and local areas to determine their own monthly VMT allocation factors for use with
MOVES.
13.2. VMT Distribution by If pe of Day
The distributions in the DayVMTFraction table divide the weekly VMT estimates into the two
MOVES day types. The OHIM report provides VMT percentage values for each day and hour of
a typical week for urban and rural roadway types for various regions of the United States. Since
the day-of-the-week data obtained from the OHIM report is not disaggregated by month or
source type, the same values were used for every month and for every source type. MOVES
uses the 1995 data displayed in Figure 2.3.2 of the OHIM report.84
The DayVMTFraction needed for MOVES has only two categories: weekdays (Monday, Tuesday,
Wednesday, Thursday and Friday) and weekend (Saturday and Sunday) days. The OHIM
reported percentages for each day of the week were summed in their respective categories and
converted to fractions, as shown in Table 13-3. The OHIM report explains that data for "3am"
refers to data collected from 3am to 4am. Thus, the data labeled "midnight" was summed with
the upcoming day.
Table 13-3 DayVMTFractions
Fraction
Rural
Urban
Weekday
0.72118
0.762365
Weekend
0.27882
0.237635
Sum
1.00000
1.000000
142
-------
We assigned the "rural" fractions to the rural road types (roadTypelDs 2 and 3) and the "urban"
fractions to the urban road types (roadTypelDs 4 and 5). The fraction of weekly VMT reported
for a single weekday in MOVES will be one-fifth of the weekday fraction and the fraction of
weekly VMT for a single weekend day will be one-half the weekend fraction.
The day type allocation of VMT will vary from location to location and EPA guidance encourages
states and local areas to determine their own VMT allocation factors for use with MOVES.
13.3. VMT Distribution by Hour of the Day
HourVMTFraction uses the same data as for DayVMTFraction. We converted the OHIM report's
VMT data by hour of the day in each day type to percent of day by dividing by the total VMT for
each day type, as described for the DayVMTFraction. There are separate sets of
HourVMTFractions for "urban" and "rural" road types, but unrestricted and unrestricted roads
use the same HourVMTFraction distributions. All source types use the same HourVMTFraction
distributions and Table 13-4 and Figure 13-1 summarize these default values.
143
-------
Table 13-4 MOVES distribution of VMT by hour of the day
hourlD
Description
Urban
Rural
Weekday
Weekend
Weekday
Weekend
1
Hour beginning at 12:00
midnight
0.00986
0.02147
0.01077
0.01642
2
Hour beginning at 1:00 AM
0.00627
0.01444
0.00764
0.01119
3
Hour beginning at 2:00 AM
0.00506
0.01097
0.00655
0.00854
4
Hour beginning at 3:00 AM
0.00467
0.00749
0.00663
0.00679
5
Hour beginning at 4:00 AM
0.00699
0.00684
0.00954
0.00722
6
Hour beginning at 5:00 AM
0.01849
0.01036
0.02006
0.01076
7
Hour beginning at 6:00 AM
0.04596
0.01843
0.04103
0.01768
8
Hour beginning at 7:00 AM
0.06964
0.02681
0.05797
0.02688
9
Hour beginning at 8:00 AM
0.06083
0.03639
0.05347
0.03866
10
Hour beginning at 9:00 AM
0.05029
0.04754
0.05255
0.05224
11
Hour beginning at 10:00 AM
0.04994
0.05747
0.05506
0.06317
12
Hour beginning at 11:00 AM
0.05437
0.06508
0.05767
0.06994
13
Hour beginning at 12:00 Noon
0.05765
0.07132
0.05914
0.07293
14
Hour beginning at 1:00 PM
0.05803
0.07149
0.06080
0.07312
15
Hour beginning at 2:00 PM
0.06226
0.07172
0.06530
0.07362
16
Hour beginning at 3:00 PM
0.07100
0.07201
0.07261
0.07446
17
Hour beginning at 4:00 PM
0.07697
0.07115
0.07738
0.07422
18
Hour beginning at 5:00 PM
0.07743
0.06789
0.07548
0.07001
19
Hour beginning at 6:00 PM
0.05978
0.06177
0.05871
0.06140
20
Hour beginning at 7:00 PM
0.04439
0.05169
0.04399
0.05050
21
Hour beginning at 8:00 PM
0.03545
0.04287
0.03573
0.04121
22
Hour beginning at 9:00 PM
0.03182
0.03803
0.03074
0.03364
23
Hour beginning at 10:00 PM
0.02494
0.03221
0.02385
0.02622
24
Hour beginning at 11:00 PM
0.01791
0.02457
0.01732
0.01917
Sum of All Fractions
1.00000
1.00000
1.00000
1.00000
144
-------
Hour of the Day
Figure 13-1 Hourly VMT fractions by day type and road type
The allocation of VMT to the hours of the day will vary from location to location and EPA
guidance encourages states and local areas to determine their own VMT allocation factors for
use with MOVES. For example, an analysis by CRC has made county specific hourly VMT
distributions available for calendar year 2014.58
13.4. Parking Activity
To properly estimate evaporative fuel vapor losses, it is important to estimate the number of
starts by time of day and the duration of time between vehicle trips. The time between trips
with the engine off is referred to as "soak time". To determine typical patterns of trip starts and
ends, MOVES uses information from instrumented vehicles. This data is stored in two tables in
the MOVES default database, as discussed below. Unlike the information used to determine
exhaust start emissions (see Section 12.1.2), these tables are unchanged from MOVES2014,
since updating the activity data in these tables was beyond the scope of more recent MOVES
145
-------
updates/ Note that the activity described below is applied only to gasoline vehicles since diesel
evaporative emissions (other than refueling spillage) are expected to be negligible and are not
calculated by MOVES.
The first table, SampleVehicleDay, lists a sample population of vehicles, each with an identifier
(vehID), an indication of vehicle type (sourceTypelD) and an indication (daylD) of whether the
vehicle is part of the weekend or weekday vehicle population. Some vehicles were added to
this table to increase the number of vehicles in each day which do not take any trips to better
match a more representative study of vehicle activity in Georgia.86 This analysis is described in
greater detail in the report describing evaporative emissions in MOVES.87
The second table, SampleVehicleTrip, lists the trips in a day made by each of the vehicles in the
SampleVehicleDay table. It records the vehID, daylD, a trip number (tripID), the hour of the trip
(hourlD), the trip number of the prior trip (priorTriplD) and the times at which the engine was
turned on and off for the trip. The keyOnTime and keyOffTime are recorded in minutes since
midnight of the day of the trip. 439 trips (about 1.1 percent) were added to this table to assure
that at least one trip is done by a vehicle from each source type in each hour of the day to
assure that emission rates will be calculated in each hour. Table 13-5 shows the resulting
number of vehicles in the SampleVehicleDay table with trip information.
Table 13-5 SampleVehic
eDay table
sourceTypelD
# of Weekday (daylD 5) Records
# of Weekend (daylD 2) Records
11
2214
983
21
821
347
31
834
371
32
773
345
41
190
73
42
110
14
43
136
59
51
205
65
52
112
58
53
123
50
54
5431
2170
61
130
52
62
122
49
v Updating the SampleVehicleTrip table to use the data also used to update starts is not straightforward. For
example, the current SampleVehicleTrip table used for evaporative emissions currently contains 37,216 vehicle
trips, whereas the Verizon light-duty database used for starts contains millions of trips. Another approach would
be to change the MOVES algorithm to calculate evaporative emissions based on summarized trip information (as
was done for start emissions) but creating this new algorithm would be a significant programing effort.
146
-------
To account for overnight soaks, many first trips reference a prior trip with a null value for
keyOnTime and a negative value for keyOffTime. The SampleVehicleDay table also includes
some vehicles that have no trips in the SampleVehicleTrip table to account for vehicles that sit
for one or more days without any driving.
The data and processing algorithms used to populate these tables are detailed in two
contractor reports.88 89 The data come from a variety of instrumented vehicle studies,
summarized in Table 13-6. These data were cleaned, adjusted, sampled and weighted to
develop a distribution intended to represent average urban vehicle activity.
Table 13-6 Source data for sample vehicle trip information
Study
Study Area
Study
Years
Vehicle Types
Vehicle
Count
3-City FTP
Study
Atlanta, GA; Baltimore, MD;
Spokane, WA
1992
Passenger cars & trucks
321
Minneapolis
Minneapolis/St. Paul, MN
2004-
2005
Passenger cars & trucks
133
Knoxville
Knoxville, TN
2000-
2001
Passenger cars & trucks
377
Las Vegas
Las Vegas, NV
2004-
2005
Passenger cars & trucks
350
Battelle
California, statewide
1997-
1998
Heavy-duty trucks
120
TxDOT
Houston, TX
2002
Diesel dump trucks
4
For vehicle classes that were not represented in the available data, the contractor synthesized
trips using trip-per-operating hour information from the EPA MOBILE653 model and soak time
and time-of-day information from source types that did have data. The application of synthetic
trips is summarized in Table 13-7.
147
-------
Table 13-7 Synthesis of sample vehicles for source types lacking data
Source Type
Based on
Direct Data?
Synthesized From
Motorcycles
No
Passenger Cars
Passenger Cars
Yes
n/a
Passenger Trucks
Yes
n/a
Light Commercial Trucks
No
Passenger Trucks
Other Buses
No
Combination Long-Haul Trucks
Transit Buses
No
Single Unit Short-Haul Trucks
School Buses
No
Single Unit Short-Haul Trucks
Refuse Trucks
No
Combination Short-Haul Trucks
Single Unit Short-Haul Trucks
Yes
n/a
Single Unit Long-Haul Trucks
No
Combination Long-Haul Trucks
Motor Homes
No
Passenger Cars
Combination Short-Haul trucks
Yes
n/a
Combination Long-Haul trucks
Yes
n/a
The resulting trip-per-day estimates are summarized in Table 13-8. The same estimate for trips
per day is used for all ages of vehicles in any calendar year.
Table 13-8 Trip per day by source type used for evaporative emissions activity
Source Type
Weekday
Weekend
Motorcycles
0.45
1.52
Passenger Cars
5.38
4.99
Passenger Trucks
5.58
4.7
Light Commercial Trucks
6.02
5.06
Other Buses
2.88
1.19
Transit Buses
4.75
4.93
School Buses
5.88
1.64
Refuse Trucks
3.85
1.28
Single Unit Short-Haul Trucks
7.14
1.67
Single Unit Long-Haul Trucks
4.45
1.74
Motor Homes
0.57
0.57
Combination Short-Haul trucks
6.07
1.6
Combination Long-Haul trucks
4.29
1.29
The trip activity used for determination of emissions resulting from parked vehicles differs from
the activity used to determine engine start emissions, described in Section 12. Ideally, trips,
engine soak periods, and parking hours would be consistent. However, since both approaches,
although different, are describing the same vehicle activity, the differences are not expected to
have a negative impact on total emission estimates.
148
-------
Knowing the sequence of starts for each vehicle in the sampleVehicleTrip table allows MOVES
to calculate the length and time of day when each soak occurs. Using this information, the
distribution of soak times in each hour of the day can be calculated for use in the determination
of evaporative emissions from parked vehicles.
The evaporative vapor losses from gasoline vehicle fuel tanks are affected by many factors,
including the number of hours a vehicle is parked without an engine start, referred to as engine
soak time. Most modern gasoline vehicles are equipped with emission control systems
designed to capture most evaporative vapor losses and store them. These stored vapors are
then burned in the engine once the vehicle is operated. However, the vehicle storage capacity
for evaporative vapors is limited and multiple days of parking (diurnals) will overload the
storage capacity of these systems, resulting in larger losses of evaporative vapors in subsequent
days.
The detailed description of the calculation for the number of vehicles that have been soaking
for more than a day and the amount of time that the vehicles have been soaking can be found
in the MOVES technical report on evaporative emissions.89
Note, the MOVES County Data Manager allows users to specify the number of engine starts in
each month, day type and hour of the day, as well as by source type and vehicle age. These
user inputs override the default start activity values provided by MOVES (see Section 12).
However, these user inputs will not update the soak times used in the calculations for
evaporative emissions, which rely solely on the sample trip data.
13.5. Hourly Hotelling Activity
In Section 11, we updated the hotelling activity rate based on instrumented truck data from the
NREL Fleet DNA database. However, this dataset was not deemed appropriate for updating the
hourly hotelling activity.™ As discussed below, we found that the hotelling hourly distribution
assumed in MOVES2014 compared well to other datasets. Thus, the hourly hotelling activity is
unchanged from MOVES2014.
To derive the hotelling hour distribution in MOVES, we used the assumption that
the hotelling hours in each day should not directly correlate with the miles traveled in each
hour, since hotelling occurs only when drivers are not driving. Instead, the fraction of hours
spent hotelling by time of day can be derived from other sources. In particular, the report,
"The NREL long-haul dataset yielded an hourly hotelling distribution with most of the activity occuring during the
daytime hours, while the MOVES2014, NCHRP 08-101 and truck survey data suggests that most occurs during the
nightime. As discussed in Section 12, NREL could not confirm the time stamp of the data for the long-haul trucks in
the FleetDNA was the local time, or a reference time because the long-haul truck was provided by an industry
partner, not collected by NREL. For these reasons, we decided not to use the FleetDNA data to update the
hotelling hourly distributions.
149
-------
Roadway-Specific Driving Schedules for Heavy-Duty Vehicles66 combines data from several
instrumented truck studies and contains detailed information about truck driver behavior.
While none of the trucks in that study were involved in long-haul interstate activity, for lack of
better data, we have assumed that long-haul truck trips have the same hourly truck trip
distribution as the heavy heavy-duty trucks that were studied.
For each hour of the day, we estimated the number of trips that would end in that hour, based
on the number of trips that started 10 hours earlier. The hours of hotelling in that hour is the
number that begin in that hour, plus the number that began in the previous hour, plus the
number that began in the hour before that and so on, up to the required eight hours of rest
time. Table 13-9 shows the number of trip starts and inferred trip ends over the hours of the
day in the sample of trucks assuming all trips are 10 hours long. For example, the number of
trip ends in hour 1 is the same as the number of trip starts 10 hours earlier in hour 15 of the
previous day.
Table 13-9
Hourly distribution of truck trips used to calculate hotelling hours
hourlD
Hour of the Day
Trip Starts
Trip Ends
1
Hour beginning at 12:00 midnight
78
171
2
Hour beginning at 1:00 AM
76
167
3
Hour beginning at 2:00 AM
65
144
4
Hour beginning at 3:00 AM
94
98
5
Hour beginning at 4:00 AM
107
71
6
Hour beginning at 5:00 AM
131
73
7
Hour beginning at 6:00 AM
194
71
8
Hour beginning at 7:00 AM
230
52
9
Hour beginning at 8:00 AM
279
85
10
Hour beginning at 9:00 AM
267
48
11
Hour beginning at 10:00 AM
275
78
12
Hour beginning at 11:00 AM
240
76
13
Hour beginning at 12:00 Noon
201
65
14
Hour beginning at 1:00 PM
211
94
15
Hour beginning at 2:00 PM
171
107
16
Hour beginning at 3:00 PM
167
131
17
Hour beginning at 4:00 PM
144
194
18
Hour beginning at 5:00 PM
98
230
19
Hour beginning at 6:00 PM
71
279
20
Hour beginning at 7:00 PM
73
267
21
Hour beginning at 8:00 PM
71
275
22
Hour beginning at 9:00 PM
52
240
23
Hour beginning at 10:00 PM
85
201
24
Hour beginning at 11:00 PM
48
211
150
-------
An estimate of the distribution of truck hotelling duration times is derived from a 2004 TRB
paper82 based on a survey of 365 truck drivers at six different locations. Table 13-10 lists the
fraction of trucks in each duration bin. Some trucks are hotelling for more than the required
eight hours, but some are hotelling for less than eight hours.
Table 13-10 Distribution of truck hotelling activity duration
Hotelling Duration (hours)
Fraction of Trucks
2
0.227
4
0.135
6
0.199
8
0.191
10
0.156
12
0.057
14
0.014
16
0.021
Total
1.000
We assume that all hotelling activity begins at the trip ends shown in Table 13-9. However, not
all trip ends have the same number of hotelling hours. The distribution of hotelling durations
from Table 13-10 is applied to the hotelling that occurs at each of these trip ends.
Table 13-11 illustrates the hotelling activity calculations based on the number of trip starts and
trip ends. The hours of hotelling in any hour of the day is the number of trip ends in the current
hour plus the trip ends from the previous hours that are still hotelling. However, since not all
trips begin and end precisely on the hour, we have discounted the oldest hour included in the
calculation by 60 percent to account for those unsynchronized trips. For example, there are 171
trip ends in hourlD 1. If all trip ends idle for two hours, the number of hours is 171 (for hourlD
1) and 40 percent of 211 (for hourlD 24) and thus 171 + (0.4*211) = 255.4 hours of hotelling.
Similarly, the number of hours can be calculated for other hotelling time periods. For four-hour
hotelling periods, the hotelling hours would be 171 + 211 + 201 + (0.4*240) = 679. Only the
oldest hour of the hotelling time period is discounted.
This calculation accounts for the time in the current hour of the day which is a result of
hotelling from trips that ended in the current hour and trips that ended in previous hours. This
approach assumes that all hotelling begins at the trip end. For example, in the hour of the day
1 for the four hours hotelling bin, the trip ends in hourlD 22 contribute to the hours of hotelling
in hourlD 1, since these trip ends are still hotelling (four hours) after the trip end. The trip ends
in hourlD 21 do not contribute to the four hours hotelling bin, since it has been more than four
hours since the trip ends occurred.
The initial calculated hours assume that all trucks idle the same amount of time, indicated by
the hotelling hours bin. The distribution (weight) from Table 13-10 is applied to the hour
estimate in each hotelling hours bin to calculate the weighted total idle hours for each hour of
the day.
151
-------
fable 13-11 Calcu
ation of
lourly distributions of hotelling activity
hourlD
Trip
Starts
Trip
Ends*
2
hours
4
hours
6 hours
8 hours
10
hours
12
hours
14 hours
16 hours
Weighted Total
Idle Hours
Distribution
1
78
171
255.4
679
1204.8
1736
2120.4
2343.6
2495.4
2638.2
1276
0.0628
2
76
167
235.4
629.4
1100
1643.6
2118.6
2408.8
2593
2739.2
1234
0.0611
3
65
144
210.8
566.4
990
1515.8
2047
2431.4
2654.6
2806.4
1166
0.0577
4
94
98
155.6
477.4
871.4
1342
1885.6
2360.6
2650.8
2835
1056
0.0526
5
107
71
110.2
379.8
735.4
1159
1684.8
2216
2600.4
2823.6
930
0.0458
6
131
73
101.4
299.6
621.4
1015.4
1486
2029.6
2504.6
2794.8
823
0.0407
7
194
71
100.2
254.2
523.8
879.4
1303
1828.8
2360
2744.4
728
0.0357
8
230
52
80.4
224.4
422.6
744.4
1138.4
1609
2152.6
2627.6
630
0.0306
9
279
85
105.8
237.2
391.2
660.8
1016.4
1440
1965.8
2497
581
0.0289
10
267
48
82
213.4
357.4
555.6
877.4
1271.4
1742
2285.6
507
0.0255
11
275
78
97.2
231.8
363.2
517.2
786.8
1142.4
1566
2091.8
479
0.0238
12
240
76
107.2
236
367.4
511.4
709.6
1031.4
1425.4
1896
457
0.0221
13
201
65
95.4
238.2
372.8
504.2
658.2
927.8
1283.4
1707
434
0.0221
14
211
94
120
266.2
395
526.4
670.4
868.6
1190.4
1584.4
447
0.0221
15
171
107
144.6
296.4
439.2
573.8
705.2
859.2
1128.8
1484.4
476
0.0238
16
167
131
173.8
358
504.2
633
764.4
908.4
1106.6
1428.4
526
0.0255
17
144
194
246.4
469.6
621.4
764.2
898.8
1030.2
1184.2
1453.8
635
0.0323
18
98
230
307.6
597.8
782
928.2
1057
1188.4
1332.4
1530.6
767
0.0374
19
71
279
371
755.4
978.6
1130.4
1273.2
1407.8
1539.2
1693.2
933
0.0458
20
73
267
378.6
853.6
1143.8
1328
1474.2
1603
1734.4
1878.4
1068
0.0526
21
71
275
381.8
913
1297.4
1520.6
1672.4
1815.2
1949.8
2081.2
1194
0.0594
22
52
240
350
893.6
1368.6
1658.8
1843
1989.2
2118
2249.4
1268
0.0628
23
85
201
297
822.8
1354
1738.4
1961.6
2113.4
2256.2
2390.8
1289
0.0645
24
48
211
291.4
762
1305.6
1780.6
2070.8
2255
2401.2
2530
1308
0.0645
Totals
3428
3428
4799
11655
18511
25367
32223
39079
45935
52791
20213
1.0000
Weight
0.227
0.135
0.199
0.191
0.156
0.057
0.014
0.021
*Assumes every trip ends 10 hours after it starts, such that all trips are 10
column sum is reduced by 60 percent to account for trip ends in a column
hours long
that are not
For the first
a full hour.
iour of hotelling in each hour bin, the
152
-------
The distribution calculated using this method is similar to the behavior observed in a
dissertation90 at the University of Tennessee, Knoxville. This study observed the trucks parking
at the Petro truck travel center located at the 140/175 and Watt Road interchange between mid-
December 2003 and August 2004. Rather than using results from a single study at a specific
location, MOVES uses the more generic simulated values to determine the diurnal distribution
of hotelling behavior. The distribution of total hotelling hours to hours of the day is calculated
from the total hotelling hours and stored in the SourceTypeHour table in MOVES.
MOVES uses this same default hourly distribution from Table 13-11 for all days and locations, as
shown below in Figure 13-2. Note this distribution of hotelling by hour of the day is similar to
the inverse of the VMT distribution used for these trucks by hour of the day.
0.08
0.07
.2 0.06
+-<
-Q
t 0.05
h
0.04
3
o
x
m 0.03
"ai
o 0.02
x
0.01
0
0 2 4 6 8 10 12 14 16 18 20 22 24
Hour of the Day
Figure 13-2 Truck hotelling distribution by hour of the day in MOVES
In Figure 13-2, we also compare the hotelling distribution to hotelling activity derived from the
Vnomics data analyzed by the NCHRP 08-101 project.83 As shown, it provides a constent diurnal
pattern, with most of the hotelling activity occuring during the nighttime hours. This is a
consistent pattern displayed by truck parking results at unofficial locations and truck stops
reported by the Federal Highway Administration.91 The data from the NCHRP 08-101 project
was used only for comparison purposes and not used dirctly to update the hotelling hourly
distribution, because NCHRP 08-101 utilized different definitions of hotelling activity than
MOVES/
x The definition of hotelling used in the draft NCHRP 08-101 project estimates idling activity with duration > 8 hour,
whereas in Section 11 we used an idle duration of > 1 hour.
153
-------
14. Geographical Allocation of Activity
The vehicle miles traveled (VMT) and vehicle populations contained in the default database
(see Section 3 and Section 4, respectively) are for all 50 states and Washington D.C. However,
MOVES can estimate activity in individual states or counties (including Puerto Rico and the
Virgin Islands) when running at Default Scale using geographic allocation factors stored in two
tables: Zone and ZoneRoadType. The geographic allocations in MOVES are based on the 2020
NEI92 distribution of VMT by county.
In MOVES, hotelling hours (including extended idling and auxiliary power unit usage) are
calculated from combination long-haul combination truck VMT in each location and have their
own allocation factors (see Section 11). Similarly, engine starts are calculated based on county
vehicle populations and are not allocated from national estimates, although county vehicle
populations themselves are calculated using allocation factors (SHO and SHP as explained in the
sections below).
Note that the MOVES design only allows for one set of geographic allocations to be stored in
the default database. While real-world geographic allocations change over time, the MOVES
defaults are used for all calendar years. Thus, it is often more accurate to use information other
than the default values. National-level emissions can be generated with calendar year specific
geographical information by running each year separately, with different user-input allocations
for each run. County- and Project-level calculations do not use the default geographical
allocation factors at all. Instead, County and Project scales require that the user input local total
activity for each individual year being modeled.
For MOVES4, we updated the default list of counties (and therefore zones) to incorporate
changes in the state of Alaska that went into effect in January 2019 when the Valdez-Cordova
census area (FIP 02261) was divided into two new census areas: Chugach Census Area (FIP
02263) and Copper Census Area (FIP 02066). For the default database, we split the activity
existing in the zone and zoneroadtype tables for the original Valdez-Cordova Census Area and
allocated it to the new census areas using the population estimate reported in the 2020 US
Census93 for these new areas.
14.1. Source Hours Operating Allocation to Zones
The national total source hours of operation (SHO) are calculated from the estimates of VMT as
described in sections above. This total VMT for each road type is allocated to county using the
SHOAIIocFactor field in the ZoneRoadType table. Although the field is named "source hours
operating", it is used only for allocating VMT and not hours of operation.
The 2020 NEI VMT was aggregated into the annual sum for the four MOVES road types in each
county and nationally and used to calculate the SHOAIIocFactor using Equation 14-1.
SHOAllocF actorRoadTypeID = ^°^^^oadrype/D Equation 14-1
NationalVMTRoadTypelD
154
-------
The county allocation values for each roadway type sum to one (1.0) over all 50 states and
Washington D.C. The same SHOAIIocFactor set is the default for all calendar years at the
National scale. County- and Project-level calculations do not use the default SHOAIIocFactor
allocations at all. Instead, County and Project scales require that the user input all local activity.
14.2. Parking Hours Allocation to Zones
The allocation of the domain-wide hours of parking (time when vehicles are not operating but
continue to have evaporative emissions) to zones is stored in the SHPAIIocFactor in the Zone
table. In the default database for MOVES, the domain is the nation and the zones are the
counties. There is no national source for hours of parking by county, so we have used a VMT-
based allocation.
The allocation is determined using the VMT estimates for each county in each state as
calculated using Equation 14-2, where i represents each individual county and / is the set of all
US counties.
SHPAllocFactori = CountyVMTj CountyVMTj Equation 14-2
iel
The county allocation values for parking hours sum to one (1.0) over all 50 states and
Washington D.C. The same SHPAIIocFactor set is the default for all calendar years at the
National scale. County- and Project-level calculations do not use the default SHPAIIocFactor
allocations at all. Instead, County and Project scales require that the user input all local activity.
Note that the same allocation values are used for the StartAllocFactor column, also saved in the
Zone table.
155
-------
15. Vehicle Mass and Road Load Coefficients
The MOVES model calculates emissions using a weighted average of emisson rates by operating
mode. For running exhaust emissions, the operating modes are defined by either vehicle
specific power (VSP) or scaled tractive power (STP). Both VSP and STP estimate the tractive
power exerted by a vehicle and are calculated based on a vehicle's speed and acceleration, but
differ in how they are scaled (or normalized). VSP is used for the motorcycle, light-duty vehicles
and light-duty truck regulatory classes 10, 20, and 30 and STP is used for heavy-duty regulatory
classes.
The SourceUseTypePhysics table describes the vehicle characteristics needed for the VSP and
STP calculations, including average vehicle mass, a fixed mass factor and three road load
coefficients for each combination of source type and regulatory class averaged over all ages. In
MOVES2014, the SourceUseTypePhysics table varied only by source type. However, regulatory
class and model year were added in MOVES3 as one of the key changes to model the Heavy-
Duty Greenhouse Gas Phase 2 rule94 which anticipates improvements to vehicle and trailer
design. MOVES uses values in the SourceUseTypePhysics table to calculate VSP and STP for each
source type/regulatory class combinations according to Equation 15-1 and Equation 15-2:
Av + Bv2 + Cv3 + M ¦ (a + a ¦ sinO) ¦ v r „
VSP = — Equation 15-1
M
Av + Bv2 + Cv3 + M ¦ (a + q ¦ sinO) ¦ v
STP = Equation 15-2
f scale
where A, B and C are the road load coefficients in units of kW-s/m, kW-s2/m2 and kW-s3/m3
respectively. A is associated with tire rolling resistence, B with mechanical rotating friction as
well as higher order rolling resistance losses and C with aerodynamic drag. M is the source
mass for the source type in metric tons, g is the acceleration due to gravity (9.8 m/s2), v is the
instantaneous vehicle speed in m/s, a is the instantaneous vehicle acceleration in m/s , sin 0
is the (fractional) road gradey and fscaie is a scaling factor. Note that the only difference
between the VSP and STP equations is the term in the denominator. For light-duty vehicles
using VSP, the power is normalized by the mass of the vehicle (fscaie = M). For heavy-duty
vehicles, the fscaie is similar, but not equal to the average source mass of the vehicle source
type Cfscaie * M).
When conducting light-duty emissions analysis, emissions data from individual vehicles are
assigned to VSP operating mode bins using Equation 15-1, with the individual vehicle's
y MOVES does not model grade at the national and county scale. Road grade may be entered at the project scale.
156
-------
measured weight as the source mass (hence the term "vehicle-specific"). When developing
emissions rates for MOVES, the emissions from individual vehicles are averaged across
operating mode bins to calculate average emission rates for each regulatory class. Because
individual vehicle weights within the same regulatory class vary, the absolute tractive power
produced by individual vehicle activity assigned to the same VSP-defined operating mode also
varies. In contrast, when MOVES calculates VSP from driving cycles and assigns operating
modes for an entire source type, the average source type mass is used instead.
For heavy-duty vehicles, STP is calculated with Equation 15-2, which is very similar to the VSP
equation except the tractive power is normalized by a fixed fscaie values for all vehicles within
the same regulatory class and model year group. The fscaie is used to bring the numerical range
of tractive power from heavy-duty vehicles into the same numerical range as the VSP values
when assigning operating modes. When developing emission rates for MOVES, operating
modes are assigned to individual vehicles using both the individual truck weight, and the
common fscaie value used for all heavy-duty vehicles from the same regulatory class, source
type and model year group. Because a common fscaievalue is used, individual vehicles assigned
to the same STP-defined operating mode bin are producing the same absolute tractive power,
regardless of differences in their individual source masses. When MOVES estimates STP and
assigns operating mode distributions for the heavy-duty fleet, it uses the average source type
mass (M) for each regulatory class, source type, and model year group in the numerator and
uses the common fscaie value which was used in the emission rate analysis.
Additional discussion regarding VSP and STP (including the selection of fscaie values) are
provided in the MOVES light-duty17 and heavy-duty18 exhaust emission rate reports,
respectively.
In both cases, MOVES derives operating mode distributions by combining second-by-second
speed and acceleration data from a specific drive schedule with the proper coefficients for a
specific source type. More information about drive schedules can be found in Section 9.1 The
following sections detail the derivation of values used in Equation 15-1 and Equation 15-2.
15.1. Source Mass and Fixed Mass Factor
The two mass factors stored in the SourceUseTypePhysics table are the source mass and fixed
mass factor. The source mass represents the average weight of vehicles of a given regulatory
class within a source type, which includes the weight of the vehicle, occupants, fuel and
payload (M in Equation 15-1 and Equation 15-2) and the fixed mass factor represents the STP
scaling factor (fscaie 'n Equation 15-2). The mass factors in the SourceUseTypePhysics table are
in units of metric tons (1000 kilograms). The source masses are reported in this section both in
units of weight in lbs (used in the regulatory class defintions), and mass in kilograms (used in
MOVES calculations).
In MOVES5, the source masses of light-duty vehicles were unchanged from previous versions of
MOVES, as presented in Table 15-1 and documented in Appendix F.
157
-------
Table 15-1. Average Vehicle Weight and Mass for Motorcycles, Light-duty Vehicles, and Light-
duty Trucks Regulatory C
asses
Source Type (sourceTypelD)
Regulatory Class (regClassID)
Average Vehicle
Weight (lbs)
Average Vehicle
Mass (kg)
Motorcycle (11)
Motorcycle (10)
628
285
Passenger Car(31)
Light-duty Vehicle, LDV (20)
3,260
1,479
Passenger Truck(31)
Light-duty Truck, LDT (30)
4,116
1,867
Light Commercial Truck (32)
Light-duty Truck, LDT (30)
4,541
2,060
The source masses for light heavy-duty trucks are based on a report from the National Research
Council.95 This report included data on empty vehicle weight ranges, typical payload capacity
and annual fleet VMT by truck class. For light heavy-duty trucks, the average source mass was
assumed to be the midpoint of the empty vehicle weight range plus 50 percent of the typical
payload capacity. The source mass for passenger trucks and light commercial trucks in
regulatory class 41 was calculated using the data presented for class 2b trucks only. A VMT-
weighted average mass was calculated for single unit trucks in regulatory class 41 using data for
class 2b and 3 trucks, and single unit trucks in regulatory class 42 were assigned a VMT-
weighted average for class 4 and 5 trucks.
Table 15-2. Average Vehicle Weight Mass for LHD2b3 and LHD45 Regulatory Classes by Source
Type
Source Type (sourceTypelD)
Regulatory Class
(regClassID)
VMT-weighted
Average Vehicle
Weight (lbs)
VMT-weighted
Average Vehicle
Mass (kg)
Passenger Truck(31)
Light Commercial Truck (32)
LHD2b3 (41)
7,500
3,402
Refuse Truck (51)
Single Unit Short-haul Truck (52)
Single Unit Long-haul Truck (53)
Motor Home (54)
LHD2b3 (41)
7,879
3,574
LHD45 (42)
12,716
5,768
The source masses for medium and heavy heavy-duty single unit trucks and combination trucks
were estimated based on weigh-in-motion data made available through FHWA's Vehicle Travel
Information System (VTRIS).96 This data source presents average gross vehicle weights by truck
type (single unit, single trailer and multi-trailer), axle count and state. An approximate mapping
between MOVES source types/regulatory classes and the VTRIS truck types/axle counts is
presented in Table 15-3. National average masses were calculated for regulatory classes 46, 47
and 49 in the single unit truck and combination truck source types using 2013 VTRIS data,
weighted by VMT data for the same year by state, as presented in Highway Statistics Table VM-
2.
158
-------
Table 15-3 Average Vehicle Weight and Mass for MHD, HHD and Glider Regulatory Classes
and Source Type and VTRIS Vehicle Classes and Axle Count
Source Type
(sourceTypelD)
Regulatory
Class
(regClassID)
VTRIS Vehicle Class
and Axle Count
VMT-weighted
Average Vehicle
Weight (lbs)
VMT-weighted
Average Vehicle
Mass (kg)
Refuse Truck (51)
MHD (46)
Single unit Trucks:
3-axle
30,424
13,800
HHD (47)
-
45,645
20,704
Single Unit Short-haul
Truck(52)
Single Unit Long-haul
Truck(53)
Motor Home (54)
MHD (46)
Single unit Trucks:
3-axle
30,424
13,800
HHD (47)
Single unit Trucks:
4-axle
55,221
25,048
Combination Short-
haul Truck (61)
Combination Long-
haul Truck(62)
MHD (46)
Single Trailer Trucks:
4-axles or less
30,891
14,012
HHD (47)
Single Trailer Trucks:
5-axle, 6-axle, or more
54,741
24,830
All Multi-trailer Trucks
Glider (49)
Single Trailer Trucks:
5-axle, 6-axle, or more
54,741
24,830
All Multi-trailer Trucks
The exception to the single unit truck analysis described above is the average source mass for
class 8 (HHD) refuse trucks because these trucks are subject to a lower Federal weight limit due
to their typical vehicle length and axle configuration.97 These vehicles are assumed to have an
average source mass of 45,645 lbs, based on several studies of in-use refuse truck
activity.98 99 100 101
The medium heavy- and heavy heavy-duty truck source masses were adjusted from the
baseline masses as calculated above to account for expected changes by model year due to
both the Heavy-Duty GHG Phase 1 and Phase 2 rules. With the Phase 1 rule, decreases were
expected for combination trucks. With the Phase 2 rule, weight reductions were also expected
for single unit trucks. The changes in source masses from the baseline masses reflecting the
Phase 1 and 2 rules are shown in Table 15-4. The details of the analyses used to estimate the
changes in source masses can be found in the Phase 1 Regulatory Impact Analysis102 and in the
docket for the Phase 2 rule.103104
159
-------
Table 15-4 MHD and HHD Changes in Vehicle Wei
Source Type*
Model Years
Change in Vehicle
Weight from Baseline
(lbs)
Single Unit Short-haul Truck
2021-2023
-4.4
2024-2026
-10.4
2027+
-16.5
Single Unit Long-haul Truck
2021-2023
-7.9
2024-2026
-23.6
2027+
-39.4
Combination Short-haul Truck
2014+
-321
Combination Long-haul Truck
2014+
-400
ght by Model Year
* No change in vehicle weights is modeled for other source types.
The source masses for all medium heavy-duty and heavy heavy-duty buses are based on a
report from the American Public Transit Association (APTA).105 This report included data on the
ranges of seating capacity, curb weight and fully-loaded weight for different types and lengths
of buses. Lacking specific data on in-use bus masses, we assume that the average source mass is
the mid-point between the curb weight and fully-loaded weight. We also assume that seating
capacity is the driving variable for the curb weight and fully-loaded weight of a bus. Under this
simplifying assumption, linear functions of seating capacity for average and fully-loaded masses
were determined by bus type and length using the ranges presented in the APTA report.
To calculate national average source masses for medium heavy-duty and heavy heavy-duty
buses, the mass functions derived from the APTA report were weighted by bus activity data
from FTA's 2017 National Transit Database (NTD).27 The NTD contains estimates of transit bus
populations and VMT and also includes data on seating capacity, bus type and vehicle length.
For each entry in the NTD that described a type of bus, double decked bus or articulated bus,
we calculated an average mass and a fully-loaded mass based on seating capacity, bus type and
vehicle length. We assigned vehicles with fully-loaded masses between 19,500 and 33,000 lbs
to the medium heavy-duty regulatory class (regClassID 46) and vehicles above 33,000 lbs were
assigned to the heavy heavy-duty regulatory class (regClassID 48 for diesel and CNG transit
buses and regClassID 47 for all the remaining bus source type and fuel type combinations).
Using these regulatory class assignments, we calculated VMT-weighted average masses for
regulatory classes 46, 47, and 48 and applied them to all bus source types (i.e., sourceTypelDs
41, 42, and 43). In the future, we will consider updating the bus mass estimates to incoporate
data on mile-weighted average passenger load.
Because the APTA report did not include data for light heavy-duty buses, we calculated source
mass based on a number of assumptions regarding vehicle parameters from manufacturer
specifications for the most popular vehicle models in the NTD with a bus type of cutaway.
Specifically, we assumed the following:
160
-------
• The fully-loaded vehicle mass is at the upper bounds of allowable GVWR for each class
(i.e., 14,000 lbs for regClassID 41 and 19,500 lbs for regClassID 42).
• Passenger capacity is 15 for regClassID 41 and 25 for regClassID 42 and the average
passenger weighs 175 lbs.
• Fuel tank capacity is 50 gallons (gasoline).
If the average operating conditions for these vehicles are at 50 percent passenger and 50
percent fuel capacity, then the average vehicle mass for LHD2b3 (regClassID 41) buses can be
calculated as 12,531 lbs and LHD45 (regClassID 42) buses as 17,156 lbs. The average weights
and source masses are shown in Table 15-5.
Table 15-5. Bus Weights and Mass by Regulatory Classes by Source Type
Source Type
(sourceTypelD)
Regulatory Class
(regClassID)
Vehicle Weight
(lbs)
Vehicle Mass
(kg)
Other Bus (41)
School Bus (43)
LHD2b3 (41)
12,531
5,684
Other Bus (41)
Transit Bus (42)
School Bus (43)
LHD45 (42)
17,156
7,782
Other Bus (41)
Transit Bus (42)
School Bus (43)
MHD (46)
25,060
11,367
Other Bus (41)
Transit Bus (42)
School Bus (43)
HHD (47)
34,399
15,603
Transit Bus (42)
Urban Bus(48)
34,399
15,603
The complete list of sourceMass and fixedMassFactor in MOVES is presented in Appendix J.
15.2. Road Load Coefficients
As indicated above, in MOVES, road load coefficients are used in the calculation of both VSP
and STP. A, B and C are the road load coefficients in units of kW-s/m, kW-s2/m2, and
kW-s3/m3, respectively. A is associated with rolling resistance, B with mechanical rotating
friction as well as higher order rolling resistance losses and C with aerodynamic drag. The
information available on road load coefficients varied by regulatory class.
15.2.1. Light-Duty and Motorcycles
Motorcycle road load coefficients, given in Equation 15-3 through Equation 15-5, were
empiricially derived in accordance with standard practice:106 107
A = 0.088 ¦ M Equation 15-3
161
-------
5 = 0
Equation 15-4
C = 0.00026 + 0.000194 ¦ M Equation 15-5
For light-duty vehicles, the road load coefficients were calculated according to Equation 15-6
through Equation 15-8:108
0-7457 „
A = 50-0 447 ' 035 ' TRLHP@somph Equation 15-6
0.7457
B = (50 ¦ 0 447)2 " °-10 " TRLHPmomph Equation 15-7
0.7457
C = (50-0 447)3 ' 0-55 ' ™HP@50mPh Equation 15-8
In the three equations above, the first factor is the appropriate unit conversion to allow A, B
and C to be used in Equation 15-1 and Equation 15-2, the second factor is the power
distribution into each of the three load categories and the third is the tractive road load
horsepower rating (TRLHP). Average values for A, B and C for source types 21, 31 and 32 were
derived from applying TRLHP values recorded in the Mobile Source Observation Database
(MSOD)109 to Equation 15-6 through Equation 15-8. While we expect light-duty road load
coefficients to improve over time due to the 2017 and Later Model Year Light-Duty Vehicle
Greenhouse Gas Emissions Rule, the impact of these changes have been directly incorporated
into the emission and energy rates.110 Therefore, these coefficients remain constant over time
in the MOVES (if not in the real-world) to avoid double counting the impacts of actual road load
improvements in the fleet.
15.2.2. Heavy-Duty Vehicles
For heavy-duty source types, no road load parameters were available in the MSOD. Therefore,
for the heavy-duty source types other than combination trucks, relationships of historical road
load coefficent to vehicle mass came from a study done by V.A. Petrushov111 and then were
revised to reflect EPA rulemakings as described later in this section. The resulting road load
coefficents are shown in Table 15-6. These relationships are grouped by regulatory class; source
type values were determined by weighting the combination of weight categories that comprise
162
-------
the individual source types.2 As noted in the table below, the B term is set to zero to reflect that
the frictional forces that are linearly related to vehicle speed in heavy-duty vehicles are very
low when compared to the rolling resistance and aerodynamic forces.
Table 15-6 Road Load Coefficients for MY 1950-2013 Buses, Motor Homes and
Single Unit Heavy-duty Trucks
Coefficient
8500 to 14000
lbs.
(3.855 to 6.350
metric ton)
14000 to 33000
lbs.
(6.350 to 14.968
metric ton)
>33000 lbs.
(>14.968 metric
ton)
Buses and Motor
Homes
/kW-s\
n m )
0.0996-M
0.0875-M
0.0661-M
0.0643 ¦ M
B(kwf)
\ m/ J
0
0
0
0
fkW-s3\
\ m3 J
0.00147 +
5.22 X id"5 ¦ M
0.00193 +
5.90 X 10-5 ¦ M
0.00289 +
4.21 X 10 ~5-M
0.0032 +
5.06 X 10-5 ¦ M
EPA set greenhouse gas (GHG) emission standards for heavy-duty vehicles in three separate
rulemakings. We revised the road load coefficients to model the impacts of the first two,
refered to in this report as the Phase l102 and Phase 294 HD GHG rules. The Phase 1 rulemaking
became effective for the 2014 model year. The Phase 2 rulemaking became effective in 2021
model year and is fully phased-in by the 2027 model year.
z The A and C coefficients were derived in MOVES2010 based on the equations in Table 15-3 and the population
fraction of regulatory classes within the sourcetypes in MOVES2010. For MOVES2014 and MOVES3, we updated
the vehicle source masses, and we scaled the A coefficients from MOVES2010 according to the changes in vehicle
mass, because there is a direct relationship between rolling resistance and vehicle weight. In contrast, for all but
the combination trucks, the aerodynamic drag coefficients, C, are unchanged from MOVES2010 because we lacked
new data and there is not a direct relationship between arodynamic drag and vehicle weight.
163
-------
The updated road load coefficients in MOVES reflect the projected improvements to the
vehicles in different model year groups. The first model year group includes model years 1950-
2013 to reflect the time period prior to the first heavy-duty truck GHG emission standards. Due
to improvements in trailers over this time period, the first model year group is split into pre-
2008 and 2008-2013 for combination tractor-trailers. The Phase 1 standards are applied to
model years 2014-2017 (or through 2020 depending on category). The Phase 2 standards are
phased-in using model year groups 2021-2023, 2024-2026 and 2027-and-later. To account for
the improvements due to the HD GHG rules, road load forces were separated into individual
road load coefficients because significant improvements are expected in aerodynamic drag and
rolling resistance, particularly for tractor-trailers. The aerodynamic and rolling resistance
components of the overall road load are determined separately and updated in MOVES4 as a
result of information on Phase 2 HD GHG rule implemention.aa
For reference, the aerodynamic drag force, Faem as a function of speed is represented as:
_ 1 2
Faero — ^P^dAfVair Equation 15-9
where p is the density of air, Cd is the aerodynamic drag coefficient, At is the frontal area of the
vehicle and i/0/r is the air speed relative to the vehicle as it is traveling. In zero wind conditions,
the relative air speed is equal to vehicle speed. Consequently, the aerodynamic drag
component of STP can be represented as:
ctd ( 1 \ 1 r A 3 Equation
STPaero = I I 1 - oC^Af V
I? j'l pCdAfvC
vJ scale' **
15-10
Thus, the C road load coefficient can be represented as:
C = \pCdAf Equation
2 15-11
The quantity CaAf, shortened to CdA, is called the drag area and is used to characterize the
overall aerodynamic drag forces for a vehicle.
aa Due to a 2021 appeals court ruling vacating the portions of the Heavy-duty Greenhouse Gas Phase 2 standards
(HDGHG2) that apply to trailers, we revised MOVES inputs that describe weight, aerodynamics, rolling resistance
and "other efficiency" improvements for combination trucks of model year 2018 and later to better represent the
implemented program.
164
-------
The tire rolling resistance force is represented using the A coefficient in the
SourceUseTypePhysics table. It is related to the coefficient of rolling resistance, Crr and source
mass M, using the following equation:
A = CRRMg Equation
15-12
where g is the gravitational acceleration.
MOVES3 included updates to both the aerodynamic and rolling resistance components of the
overall road load reflecting the greenhouse gas emissions standards for combination trucks. An
aerodynamic assessment of all model years of combination trucks was conducted to utilize a
consistent method in MOVES, and the aerodynamic values were updated for all model years to
reflect the aerodynamic technology analysis and projections in HD GHG Phase 2 rulemakings.
These values were further updated in MOVES4 to reflect the HD GHG Phase 2 rulemaking as
implemented. The average road load coefficients are updated by source type and regulatory
class through the beginModelYearlD and endModelYearlD fields in the SourceUseTypePhysics
table. Appendix I describes how the aerodynamic improvements were developed as part of the
rulemaking and how they were used to update MOVES.
For buses, refuse trucks, motor homes, and long-haul and short-haul single unit trucks
(sourceTypelDs 41 through 54), the A coefficient values determined through tire rolling
resistance reductions projected in the HD GHG Phase 1 and Phase 2 rulemakings were used
directly. The aerodynamic drag coefficient (C coefficient) was not updated for these heavy-duty
vehicles because no significant improvements in C coefficients is expected from the Phase 2
standards.102
While we expect road load coefficients for Heavy-Duty Pickups and Vans (regclassID 41) to
improve over time due to the Phase 1 and Phase 2 HD GHG rules, the impact of these changes
have been directly incorporated into the emission and energy rates.18 Since nearly all HD pickup
trucks and vans are certified on a chassis dynamometer, the improvements in road loads
expected from the greenhouse gas standards are modeled as total vehicle improvements
without separating out the engine and road load components. Therefore, these coefficients
remain constant over time in MOVES (if not in the real-world) to avoid double counting the
impacts of actual road load improvements in the fleet.
The final road load coefficients for all regulatory classes and sourcetypes in MOVES are shown
in Appendix J.
165
-------
16. Air Conditioning Activity Inputs
This section describes three inputs used in determining the impact of air conditioning on
emissions. The ACPenetrationFraction is the fraction of vehicles equipped with air conditioning.
FunctioningACFraction describes the fraction of these vehicles in which the air conditioning
system is working correctly. The ACActivityTerms relate air conditioning use to local heat and
humidity. These factors remain the same as in earlier versions of MOVES. More information on
air conditioning effects and how air conditioning affects Electric Vehicle energy consumption is
provided in the MOVES technical report on adjustment factors.112
16.1. ACPenetrationFraction
The ACPenetrationFraction is a field in the SourceTypeModelYear table that describes the
fraction of vehicles equipped with air conditioning. Default values, by source type and model
year, were taken from MOBILE6.113 Market penetration data by model year were gathered from
Ward's Automotive Handbook for light-duty vehicles and light-duty trucks for model years 1972
through 1995 for cars and 1975-1995 for light trucks. Rates in the first few years of available
data were quite variable, so values for early model years were estimated by applying the 1972
and 1975 rates for cars and trucks, respectively. Projections beyond 1995 were developed by
calculating the average yearly rate of increase in the last five years of data and applying this
rate until a predetermined cap was reached. A cap of 98 percent was placed on cars and 95
percent on trucks under the assumption that there will always be vehicles sold without air
conditioning, more likely trucks than cars. No data were available on heavy-duty trucks. While
VIUS asks if trucks are equipped with A/C, "no response" was coded the same as "no," making
the data unusable for this purpose. For MOVES, the light-duty vehicle rates were applied to
passenger cars and the light-duty truck rates were applied to all other source types (except
motorcycles, for which A/C penetration is assumed to be zero), as summarized in Table 16-1.
166
-------
Table 16-1 AC penetration fractions in MOVES
Motorcycles
Passenger Cars
All Trucks and Buses
1972-and-earlier
0
0.592
0.287
1973
0
0.726
0.287
1974
0
0.616
0.287
1975
0
0.631
0.287
1976
0
0.671
0.311
1977
0
0.720
0.351
1978
0
0.719
0.385
1979
0
0.694
0.366
1980
0
0.624
0.348
1981
0
0.667
0.390
1982
0
0.699
0.449
1983
0
0.737
0.464
1984
0
0.776
0.521
1985
0
0.796
0.532
1986
0
0.800
0.544
1987
0
0.755
0.588
1988
0
0.793
0.640
1989
0
0.762
0.719
1990
0
0.862
0.764
1991
0
0.869
0.771
1992
0
0.882
0.811
1993
0
0.897
0.837
1994
0
0.922
0.848
1995
0
0.934
0.882
1996
0
0.948
0.906
1997
0
0.963
0.929
1998
0
0.977
0.950
1999+
0
0.980
0.950
16.2. FunctioningACFraction
The FunctioningACFraction field in the SourceTypeAge table (see Table 16-2) indicates the
fraction of the air-conditioning-equipped fleet with fully functional A/C systems, by source type
and vehicle age. A value of one means all systems are functional. This is used in the calculation
of total energy to account for vehicles without functioning A/C systems. Default estimates were
developed for all source types using the "unrepaired malfunction" rates used for 1992-and-later
model years in MOBILE6. The MOBILE6 rates were based on the average rate of A/C system
failure by age reported in the 1997 Consumer Reports Magazine Automobile Purchase Issue
and assumptions about repair frequency during and after the warranty period. The MOBILE6
rates were applied to all source types except motorcycles, which were assigned a value of zero
for all years.
167
-------
Table 16-2 FunctioningACFraction by age (for all source types except motorcycles)
agelD
functioningACFraction
0
1
1
1
2
1
3
1
4
0.99
5
0.99
6
0.99
7
0.99
8
0.98
9
0.98
10
0.98
11
0.98
12
0.98
13
0.96
14
0.96
15
0.96
16
0.96
17
0.96
18
0.95
19
0.95
20
0.95
21
0.95
22
0.95
23
0.95
24
0.95
25
0.95
26
0.95
27
0.95
28
0.95
29
0.95
30
0.95
31
0.95
32
0.95
33
0.95
34
0.95
35
0.95
36
0.95
37
0.95
38
0.95
39
0.95
40
0.95
168
-------
16.3. Air Conditioning Activity Demand
In the MonthGroupHour table, ACActivityTerms A, B and C are coefficients for a quadratic
equation that calculates air conditioning activity demand as a function of the heat index. These
terms are applied in the calculation of the A/C adjustment in the energy consumption
calculator. The methodology and the terms themselves were originally derived for MOBILE6
and are documented in the report, Air Conditioning Activity Effects in MOBILE6.113 They are
based on analysis of air conditioning usage data collected in Phoenix, Arizona, in 1994.
In MOVES, ACActivityTerms are allowed to vary by monthGroup and Hour, in order to provide
the possibility of different A/C activity demand functions at a given heat index by season and
time of day (this accounts for differences in solar loading observed in the original data).
However, the default data uses one set of coefficients for all MonthGroups and Hours. These
default coefficients represent an average A/C activity demand function over the course of a full
day. The coefficients are listed in Table 16-3.
Table 16-3 Air conditioning activity coefficients
A
B
C
-3.63154
0.072465
-0.000276
The A/C activity demand function that results from these coefficients is shown in Figure 16-1. A
value of 1 means the A/C compressor is engaged 100 percent of the time; a value of 0 means no
A/C compressor engagement.
169
-------
Heat Index (F)
Figure 16-1 Air conditioning activity demand as a function of heat index
170
-------
17.Conclusion and Areas for Future Research
Properly characterizing emissions from vehicles requires a detailed understanding of the cars
and trucks that make up the vehicle fleet and their patterns of operation. The national default
information in MOVES5 provide a reliable basis for estimating national emissions. The most
important of these inputs are well-established: base year VMT and population estimates come
from long-term, systematic national measurements by US Department of Transportation. The
relevant characteristics for prevalent vehicle classes are well-known; base year age distributions
are well-measured and driving activity has been the subject of much study in recent years.
Still, the fleet and activity inputs do have significant limitations. In particular, local variations
from the national defaults can contribute to discrepancies in resulting emission estimates. Thus,
it is recommended to replace many of the MOVES fleet and activity defaults with local data
when available as explained in EPA's Technical Guidance.2
The fleet and activity defaults also are limited by the necessity of forecasting future emissions.
EPA utilizes annual US Department of Energy forecasts of vehicle sales and activity. The inputs
for MOVES5 were developed for a 2023 base year and much of the source data is from 2023
and earlier. This information needs to be updated periodically to assure that the model defaults
reflect the latest available data and projections on the US fleet.
Moreover, for data that is specific to MOVES, we are also limited by available staff and funding.
Collecting data on vehicle fleet and activity is expensive, especially when the data is intended to
accurately represent the entire United States. Even when EPA does not generate data directly
(for example, compilations of state vehicle registration data), obtaining the information needed
for MOVES can be costly and, thus, dependent on budget choices.
Future updates to vehicle population and activity defaults will need to continue to focus on the
vehicles that contribute the most air pollution nationally, namely gasoline light-duty cars and
trucks and diesel heavy-duty trucks. Information collection on motorcycles, refuse trucks,
motor homes, diesel light-duty vehicles,gasoline heavy-duty vehicles, and vehicles fueled by
CNG and E-85 will be a lower priority. Similarly, in addition to updating the model defaults, we
will need to consider whether the current MOVES design continues to meet our modeling
needs. Simplifications to the model to remove categories, such as source types or road types,
might simplify data collection and make noticeable improvements in run time without affecting
the validity of fleet-wide emission estimates.
In addition to these general limitations, there are also specific MOVES data elements that could
be improved with additional research, including:
• Updates to the trip information used to generate evaporative activity to be
consistent with the new engine start and soak distributions based on the telematics
data; this will likely require modification to the MOVES code as well as updates to
the default database;
171
-------
• Updated real-world highway driving cycles and operating mode distributions,
including incorporating ramp activity into the default highway driving cycles and
accounting for grade;
• Additional instrumented vehicle data from a wider sample of heavy-duty vehicles to
better characterize off-network behavior including vehicle starts and soaks;
• Improved information on truck hotelling durations, locations and temporal
distributions, and variation in operating modes for different vehicle fuel types and
technologies;
• VSP/STP adjustments for road grade and vehicle load;
• Better data on activity changes with age, such as mileage accumulation rates, start
activity and soak distributions. Telematics will provide important insights here, but
gathering representative data for the oldest vehicles in the fleet will continue to be a
challenge;
• Updated estimates of vehicle scrappage rates used to project vehicle age
distributions;
• Updated air conditioning system usage, penetration and failure rates;
• Finer vehicle type distinctions in temporal activity and road type distributions;
• Updated information on light-duty vehicle mass;
• Information on shifts in vehicle activity patterns with population shifts to electric,
shared, connected and automated vehicles.
At the same time, the fundamental MOVES assumption that vehicle activity varies by source
type and not by fuel type or other source bin characteristic may be challenged by the growing
market share of alternative vehicles such as autonomous, shared and electric vehicles which
may have distinct activity patterns. As we progress with MOVES, the development of vehicle
population and activity inputs will continue to be an essential area of research.
172
-------
Appendix A Fuel If pe and Regulatory Class Fractions from Previous
Versions of MOVES
Fuel type and regulatory class distributions for most source types are described in Section 5.1.
In the current version of MOVES, the fuel type and regulatory class distributions were
unchanged from previous versions of the model for the following source type and model year
combinations:
• Passenger cars, school buses, refuse trucks, short-haul and long-haul single unit trucks
and all combination trucks prior to model year 2000
• Passenger trucks and light commercial trucks prior to model year 1981
This appendix describes the derivation of these fuel type and regulatory class distributions.
Note that the analysis described here uses 1960 as the earliest model year that MOVES can
model. In MOVES5, we expanded this back to 1950 and used the fuel type and regulatory class
distributions for model year 1960 for all model years 1950-1959.
Al. Distribution. i Ni Model ) ' i • 1960-1981
The fuel type distributions between 1960 and 1981 for each source type have been summarized
in Table A-l and Table A-2. Truck diesel fractions in Table A-l were derived using the 1999 IHS
vehicle registrations and the 1997 VIUS,114 except for refuse trucks and motor homes. We
assumed 96 percent of refuse trucks were manufactured to run on diesel fuel in 1980 and
earlier according to the average diesel fraction from VIUS across all model years.
Table A-l Diesel fractions for truck source types
Source Type*
Light
Refuse
Single Unit
Short-Haul
Long-Haul
Model Year
Passenger
Commercial
Trucks
Trucks
Combination
Combination
Trucks(31)
Trucks(32)
(51)
(52 & 53)
Trucks(61)
Trucks(62)
1960-1979
0.0139
0.0419
0.96
0.2655
0.9146
1.0000
1980
0.0124
0.1069
0.96
0.2950
0.9146
1.0000
1981
0.0178
0.0706
0.96
0.3245
0.9146
1.0000
*AII other true
-------
Table A-2 Diesel fractions for non-truck source types
Source Type*
Model Year
Motorcycles (11)
Passenger Cars (21)
School Buses (43)
1960-1974
0
0.0069
0.0087
1975
0
0.0180
0.0087
1976
0
0.0165
0.0086
1977
0
0.0129
0.0240
1978
0
0.0151
0.0291
1979
0
0.0312
0.0460
1980
0
0.0467
0.0594
1981
0
0.0764
0.2639
*AII other vehicles are assumed to be gasoline-powerec
. Values forTransit
Buses and Other Buses were estimated as described in Section 5.1.
The 1960-1981 regulatory class distributions were derived from the 1999 IHS data and VIUS.
Motorcycles (sourceTypelD 11 and regClassID 10) and passenger cars (sourceTypelD 21 and
regClassID 20) have one-to-one relationships between source types and regulatory classes for
all model years. Passenger trucks (sourceTypelD 31) and light commercial trucks (sourceTypelD
32) are split between fuel type and regulatory class (regClassID 30 and 40) as shown in Table A-
3.
Table A-3 Percentage by regulatory class and fuel type for passenger trucks (sourceTypelD 31)
and light commercial truck (sourceTypelD 32)
Passenger Trucks(31)
Light Commercial Trucks (32)
Gasoline
Diesel
Gasoline
Diesel
LDT
LHD
LDT
LHD
LDT
LHD
LDT
LHD
Model Year
(30)
(40)
(30)
(40)
(30)
(40)
(30)
(40)
1960-1966
81%
19%
38%
62%
24%
76%
7%
93%
1967
90%
10%
38%
62%
72%
28%
7%
93%
1968
88%
12%
38%
62%
67%
33%
7%
93%
1969
100%
0%
38%
62%
91%
9%
7%
93%
1970
99%
1%
38%
62%
80%
20%
7%
93%
1971
96%
3%
38%
62%
94%
6%
7%
93%
1972
96%
4%
38%
62%
75%
25%
7%
93%
1973
95%
5%
38%
62%
59%
41%
7%
93%
1974
95%
5%
38%
62%
65%
35%
7%
93%
1975
97%
3%
38%
62%
72%
28%
7%
93%
1976
95%
5%
38%
62%
88%
12%
7%
93%
1977
89%
11%
38%
62%
79%
21%
7%
93%
1978
85%
15%
38%
62%
81%
19%
7%
93%
1979
87%
13%
38%
62%
78%
22%
7%
93%
1980
90%
10%
38%
62%
74%
26%
40%
60%
1981
96%
4%
38%
62%
89%
11%
12%
88%
174
-------
The school bus regulatory class fractions were reused from MOBILE6, originally based on 1996
and 1997 IHS data. The 1960-1981 regulatory class distributions for diesel-fueled single unit and
combination trucks have been summarized in Table A-4 below. All 1960-1981 gasoline-fueled
single unit and combination trucks fall into the medium heavy-duty (MHD) regulatory class
(regClassID 46).
Table A-4 Percentange of MHD trucks (regClass 46) among diesel-fueled single unit and
combination trucks*
Model
Year
Refuse Trucks
(51)
Single Unit
Trucks (52 & 53)
Short-haul
Combination Trucks
(61)
Long-haul
Combination Trucks
(62)
1960-
1972
100%
0%
0%
0%
1973
100%
3%
8%
0%
1974
0%
6%
30%
0%
1975
0%
14%
3%
0%
1976
0%
44%
13%
0%
1977
0%
43%
31%
0%
1978
0%
36%
18%
0%
1979
0%
34%
16%
0%
1980
0%
58%
29%
5%
1981
0%
47%
31%
6%
* For these source types, all remaining trucks are in the HHD regulatory class (regClassID 47)
Distribution. i Ni Model Years 1982-1999
VIUS was our main source of information for determining fuel and regulatory class fractions for
these model years. Table A-5 summarizes how the VIUS200223 parameters were used to classify
the VIUS data to calculate fuel and regulatory class fractions for the light-duty, single unit and
combination truck source types.
Axle arrangement (AXLE_CONFIG) was used to define four categories: straight trucks with two
axles and four tires (codes 1, 6, 7, 8), straight trucks with two axles and six tires (codes 2, 9, 10,
11), all straight trucks (codes 1-21) and all tractor-trailer combinations (codes 21+). Primary
distance of operation (PRIMARY_TRIP) was used to define short-haul (codes 1-4) for vehicles
with primary operation distances less than 200 miles and long-haul (codes 5-6) for 200 miles
and greater. The VIN-decoded gross vehicle weight (ADM_GVW) and survey weight
(VIUS_GVW) were used to distinguish vehicles less than 10,000 lbs. as light-duty and vehicles
greater than or equal to 10,000 lbs. as heavy-duty. Any vehicle with two axles and at least six
tires was considered a single unit truck regardless of weight. We also note that refuse trucks
have their own VIUS vocational category (BODYTYPE 21) and that MOVES distinguishes
between personal (OPCLASS 5) and non-personal use.
175
-------
Table A-5 VIUS2002 parameters used to distinguish trucks in previous versions of MOVES
Source Type
Axle
Arrangement
Primary Distance
of Operation
Weight
Body Type
Operator
Class
Passenger
Trucks
AXLE_CONFIG in
(1,6,7,8)*
Any
ADM_GVW in (1,2) &
VIUS_GVW in (1,2,3)
Any
OPCLASS=5
Light
Commercial
Trucks
AXLE_CONFIG in
(1,6,7,8)*
Any
ADM_GVW in (1,2) &
VIUS_GVW in (1,2,3)
Any
OPCLASS*5
Refuse
Trucks"
AXLE_CONFIG in
(2,9,10,11)
TRIP_PRIMARY in
(1,2,3,4)
Any
BODYTYPE
=21
Any
AXLE_CONFIG
<=21
TRIP_PRIMARY in
(1,2,3,4)
ADM_GVW > 2 &
VIUS_GVW > 3
BODYTYPE
=21
Any
Single Unit
Short-Haul
Trucks"
AXLE_CONFIG in
(2,9,10,11)
TRIP_PRIMARY in
(1,2,3,4)
Any
BODYTYPE
*21
Any
AXLE_CONFIG
<=21
TRIP_PRIMARY in
(1,2,3,4)
ADM_GVW > 2 &
VIUS_GVW > 3
BODYTYPE
*21
Any
Single Unit
Long-Haul
Trucks"
AXLE_CONFIG in
(2,9,10,11)
TRIP_PRIMARY in
(5,6)
Any
Any
Any
AXLE_CONFIG
<=21
TRIP_PRIMARY in
(5,6)
ADM_GVW > 2 &
VIUS_GVW > 3
Any
Any
Combination
Short-Haul
Trucks
AXLE_CONFIG
>=21
TRIP_PRIMARY in
(1,2,3,4)
Any
Any
Any
Combination
Long-Haul
Trucks
AXLE_CONFIG
>=21
TRIP_PRIMARY in
(5,6)
Any
Any
Any
In the MOVES2014 analysis, we did not constrain axle configuration of light-duty trucks, so there are some,
albeit very few, light-duty trucks that have three axles or more and/or six tires or more. These vehicles are
classified as light-duty trucks based primarily on their weight. Only 0.27 percent of light-duty trucks have
such tire and/or axle parameters and they have a negligible impact on vehicle populations and emissions.
For a source type with multiple rows, the source type is applied to any vehicle with either set of parameters.
hi - >• 1 I-' I finitions
Motorcycles and passenger cars in MOVES borrow vehicle definitions from the FHWA Highway
Performance Monitoring System (HPMS) classifications from the Highway Statistics Table MV-1.
Source type definitions for school buses are taken from various US Department of
Transportation sources. While refuse trucks were identified and separated from other single
unit trucks in VIUS, motor homes were not.
Light-C Iks
Light-duty trucks include pickups, sport utility vehicles (SUVs) and vans.7 Depending on use and
GVWR, we categorize them into two different MOVES source types: 1) passenger trucks
(sourceTypelD 31) and 2) light commercial trucks (sourceTypelD 32). FHWA's vehicle
classification specifies that light-duty vehicles are those weighing less than 10,000 pounds,
specifically vehicles with a GVWR in Class 1 and 2, except Class 2b trucks with two axles or more
and at least six tires are assigned to the single unit truck category.
176
-------
VIUS contains many survey questions on weight; we chose to use both a VIN-decoded gross
vehicle weight rating (ADM_GVW) and a respondent self-reported GVWR (VIUS_GVW) to
differentiate between light-duty and single unit trucks. For the passenger trucks, there is a final
VIUS constraint that the most frequent operator classification (OPCLASS) must be personal
transportation. Inversely, light commercial trucks (sourceTypelD 32) have a VIUS constraint that
their most frequent operator classification must not be personal transportation.
Buses
Previous versions of MOVES had three bus source types: intercity (sourceTypelD 41), transit
(sourceTypelD 42) and school buses (sourceTypelD 43). Since the definition of sourceTypelDs 41
and 42 changed in MOVES3, only school bus distributions for model years prior to 2000 were
retained in the current version of MOVES. According to FHWA, school buses are defined as
vehicles designed to carry more than ten passengers, used to transport K-12 students between
their home and school.
Siiri|li ' 111iii 111:ks
The single unit HPMS class in MOVES consists of refuse trucks (sourceTypelD 51), short-haul
single unit trucks (sourceTypelD 52), long-haul single unit trucks (sourceTypelD 53) and motor
homes (sourceTypelD 54). FHWA's vehicle classification specifies that a single unit truck as a
single-frame truck with a gross vehicle weight rating of greater than 10,000 pounds or with two
axles and at least six tires—colloquially known as a "dualie." As with light-duty truck source
types, single unit trucks are sorted using VIUS parameters, in this case that includes axle
configuration (AXLE_CONFIG) for straight trucks (codes 1-21), vehicle weight (both ADM_GVW
and VIUS_GVW), most common trip distance (TRIP_PRIMARY) and body type (BODYTYPE). All
short-haul single unit trucks must have a primary trip distance of 200 miles or less and must not
be refuse trucks and all long-haul trucks must have a primary trip distance of greater than 200
miles. Refuse trucks are short-haul single unit trucks with a body type (code 21) for trash,
garbage, or recyclable material hauling. Motor home distributions from previous versions were
not retained in the current version of MOVES, and therefore these vehicles are not discussed
further in this section.
Combination Trucks
A combination truck is any truck-tractor towing at least one trailer according to VIUS. MOVES
divides these tractor-trailers into two MOVES source types: short-haul (sourceTypelD 61) and
long-haul combination trucks (sourceTypelD 62). Like single unit trucks, short-haul and long-
haul combination trucks are distinguished by their primary trip length (TRIP_PRIMARY) in VIUS.
If the tractor-trailer's primary trip length is equal to or less than 200 miles, then it is considered
short-haul. If the tractor-trailer's primary trip length is greater than 200 miles, then it is
considered long-haul. Short-haul combination trucks are older than long-haul combination
trucks and these short-haul trucks often purchased in secondary markets, such as for drayage
applications, after being used primarily for long-haul trips.115
177
-------
Fuel Type and Regulatory Class Distributions
The SampleVehiclePopulation table fractions were developed by EPA using the sample vehicle
counts data, which primarily joins calendar year 2011 registration data from IHS and the 2002
Vehicle Inventory and Use Survey (VIUS) results. The sample vehicle counts data were
generated by multiplying the 2011 IHS vehicle populations by the source type allocations from
VIUS.
While VIUS provide source type classifications, we relied primarily on the 2011 IHS vehicle
registration data to form the basis of the fuel type and regulatory class distributions in the
SampleVehiclePopulation table. The IHS data were provided with the following fields: vehicle
type (cars or trucks), fuel type, gross vehicle weight rating (GVWR) for trucks, household vehicle
counts and work vehicle counts. We combined the household and work vehicle counts. The
MOVES distinction between personal and commercial travel for light-duty trucks comes from
VIUS.
The IHS records by FHWA truck weight class were grouped into MOVES GVWR-based regulatory
classes, as shown in Table A-6 below. As stated above, all passenger cars were assigned to
regClassID 20. The mapping of weight class to regulatory class is straightforward with one
notable exception: delineating trucks weighing more or less than 8,500 lbs.
Tab
e A-6 Initial mapping from FHWA truck classes to MOVES regulatory classes
Vehicle
Category
FHWA Truck Weight Class
Weight Range (lbs.)
regClassID
Trucks
1
< 6,000
30
Trucks
2a
6,001-8,500
30*
Trucks
2b
8,501-10,000
41*
Trucks
3
10,001-14,000
41
Trucks
4
14,001-16,000
42
Trucks
5
16,001-19,500
42*
Trucks
6
19,501-26,000
46
Trucks
7
26,001-33,000
46
Trucks
8a
33,001-60,000
47
Trucks
8b
> 60,001
47
Cars
20
*Afterthe IHS data had been sorted into source types (described later in this section), some
regulatory classes were merged or divided. Any regulatory class 41 vehicles in light-duty truck
source types were reclassified into the new regulatory class 40 (see explanation in Section
2.3), any regulatory class 30 vehicles in single unit truck source types were reclassified into
regulatory class 41 and any regulatory class 42 vehicles in combination truck source types
were reclassified into regulatory class 46.
Since the IHS dataset did not distinguish between Class 2a (6,001-8,500 lbs.) and Class 2b
(8,501-10,000 lbs.) trucks, but MOVES regulatory classes 30, 40 and 41 all fall within Class 2, we
178
-------
needed a secondary data source to allocate the IHS gasoline and diesel trucks between Class 2a
and 2b. We derived information from an Oak Ridge National Laboratory (ORNL) paper,116
summarized in Table A-7, to allocate the IHS Class 2 gasoline and diesel trucks into the
regulatory classes. Class 2a trucks fall in regulatory class 30 and Class 2b trucks fall in regulatory
class 41.
Table A-7 Fractions used to distribute Class 2a and 2b trucks
Truck Class
Fuel Type
Gasoline
Diesel
2a
0.808
0.255
2b
0.192
0.745
Additionally, the IHS dataset includes a variety of fuels, some that are included in MOVES and
others that are not. Only the IHS diesel, gasoline, or gasoline and another fuel were included in
our analysis; all other alternative fuel vehicles were omitted. While MOVES2014 did model
light-duty E-85 and electric vehicles, these relative penetrations of alternative fuel vehicles have
been developed from secondary data sources rather than IHS because IHS excludes some
government fleets and retrofit vehicles that could potentially be large contributors to these
alternative fuel vehicle populations. Instead, we used flexible fuel vehicle sales data reported
for EPA certification. Table A-8 illustrates how IHS fuels were mapped to MOVES fuel types and
which IHS fuels were not used in MOVES.
The "N/A" mapping shown in Table A-8 led us to discard 0.22 percent, roughly 530,000 vehicles
(mostly dedicated or aftermarket alternative fuel vehicles), of IHS's 2011 national fleet in
developing the default fuel type fractions. However, because the MOVES national population is
derived top-down from FHWA registration data, as outlined in Section 4.1, the total population
is not affected. We considered the IHS vehicle estimates to be a sufficient sample for the fuel
type and regulatory class distributions in the SampleVehiclePopulation table.
179
-------
Table A-8 List of fuels from the IHS dataset used to develop MOVES fuel type distributions
IHS Fuel Type
MOVES fuelTypelD
MOVES Fuel Type
Unknown
N/A
Undefined
N/A
Both Gas and Electric
1
Gasoline
Gas
1
Gasoline
Gas/Elec
1
Gasoline
Gasoline
1
Gasoline
Diesel
2
Diesel
Natural Gas
N/A
Compressed Natural Gas
N/A
Natr.Gas
N/A
Propane
N/A
Flexible (Gasoline/Ethanol)
1
Gasoline
Flexible
1
Gasoline
Electric
N/A
Cnvrtble
N/A
Conversion
N/A
Methanol
N/A
Ethanol
1
Gasoline
Convertible
N/A
Next, we transformed the VIUS dataset into MOVES format. The VIUS vehicle data were first
assigned to MOVES source types using the constraints in Table A-5, and then to MOVES
regulatory classes using the mapping described in Table A-6, including the allocation between
Class 2a and 2b trucks from the ORNL study in Table A-7. Similar to our fuel type mapping of the
IHS dataset, we chose to omit alternative fuel vehicles, as summarized below in Table A-9.
180
-------
Table A-9 Mapping of VIUS2002 fuel types to MOVES fuel types
VIUS Fuel Type
VIUS Fuel Code
MOVES fuelTypelD
MOVES Fuel
Type
Gasoline
1
1
Gasoline
Diesel
2
2
Diesel
Natural gas
3
N/A
Propane
4
N/A
Alcohol fuels
5
N/A
Electricity
6
N/A
Gasoline and natural gas
7
1
Gasoline
Gasoline and propane
8
1
Gasoline
Gasoline and alcohol fuels
9
1
Gasoline
Gasoline and electricity
10
1
Gasoline
Diesel and natural gas
11
2
Diesel
Diesel and propane
12
2
Diesel
Diesel and alchol fuels
13
2
Diesel
Diesel and electricity
14
2
Diesel
Not reported
15
N/A
Not applicable
16
N/A
This process yielded VIUS data by MOVES source type, model year, regulatory class and fuel
type. The VIUS source type distributions were calculated in a similar fashion to the
SampleVehiclePopulation fractions discussed above for each regulatory class-fuel type-model
year combination. Stated formally, for any given model year i, regulatory class j, and fuel type
k, the source type population fraction / for a specified source type I will be the number of VIUS
trucks N in that source type divided by the sum of VIUS trucks across the set of all source types
L. The source type population fraction is summarized in Equation A-l:
Ni i k i
f(VIUS)i jkl= ^—
n Jl,J,u — Equation A-l
*—HeL
The VIUS data in our analysis spanned model year 1986 to 2002. The 1986 distribution was used
for all prior to MY 1986.
From there the source type distributions from VIUS were multiplied by the IHS vehicle
populations to generate the sample vehicle counts by source type. Expressed in Equation A-2,
the sample vehicle counts are:
N(SVP)ijjjkjl = P(Polk)i j k l ¦ f (VIUS)Equation A-2
181
-------
where N is the number of vehicles used to generate the SampleVehiclePopulation table, P is
the 2011IHS vehicle populations and / is the source type distributions from VIUS.
VIUS 2002
Polk 2011
Interim Polk
Vehicle Category
Model Year
Fuel Type
GVWR
Household Units
Work Units
modelYearlD
fuelTypelD
regClassID
totalCounts
V
Interim VIUS
sourceTypelD
modelYearlD
fuelTypelD
regClassID
s ourceTyp eF ractions
V
4
INTERCITY BUSES
~
SAMPLE ID
AXLE_CONFI G
TRIPPRIMARY
OPCLASS
FUEL
VIUSOVW
ADM_MODELYEAR
ADMGVW
TAB TRUCKS
TRANSIT BUSES
sourceTypelD
modelYearlD
fuelTypelD
regClassID
SCHOOL BUSES
MOTORCYCLES
MOTOR HOMES
Sample Vehicle Counts
Figure A-l Flowchart of data sources of fuel and regulatory class distributions for model years
1982-1999
These sample vehicle counts by source type were then utilized to calculate the sample vehicle
population fractions, stmyFraction and stmyFuelEngFraction, as defined above. For simplicity,
we also moved the small number of LHD45 (regClassID 42) vehicles in combination truck source
types to MHD (regClassID 46). The source mass and road-load coefficients for combination
trucks are only developed for MHD, HHD and Glider vehicles.
As noted above, the initial sample vehicle counts dataset did not contain buses, so information
on these source types was appended. In the subsections below, we have provided more
detailed descriptions by source type.
Appendix A.2.2.1 Motorcycles
The representation of motorcycles in the SampleVehiclePopulation table is straightforward. All
motorcycles fall into the motorcycle regulatory class (regClassID 10) and must be fueled by
gasoline.
Appendix A.2.2.2 Passenger Cars
Any passenger car is considered to be in the light-duty vehicle regulatory class (regClassID 20).
Cars were included in the IHS dataset purchased in 2012 and EPA's subsequent sample vehicle
counts dataset, which provided the split between gasoline and diesel cars in the
182
-------
SampleVehiclePopulation table. Flexible fuel (E85-capable) cars were also included in the SVP
fuel type distributions but added after the sample vehicle counts analysis. We assume that a
flexible fuel vehicle would directly displace its gasoline counterpart. For model years 2011 and
earlier, we used manufacturer reported sales to EPA in order to calculate the fraction of sales of
flexible fuel cars among sales of all gasoline and flexible fuel cars and added those penetrations
as the fraction of E85 (fuelTypelD 5) vehicles and deducted them from the gasoline cars in the
IHS dataset.
Appendix A.2.2.3 Light-Du cks
Since passenger and light commercial trucks are defined as light-duty vehicles, they are
constrained to regulatory class 30 and 40. Within the sample vehicle counts, GVWR Class 1 and
2a trucks were classified as regulatory class 30 and Class 2b trucks with two axles and four tires
were classified as regulatory class 40. Both light-duty truck source types are divided between
gasoline and diesel using the underlying splits in the sample vehicle counts data. Passenger
trucks and light commercial trucks have similar but distinct distributions. Similar to cars, a
penetration of flexible fuel (E-85-capable) light-duty trucks was calculated using EPA
certification sales for MY 2011 and earlier.
Appendix A.2.2.4 Buses
Only school bus distributions from MOVES2014 for model years prior to 2000 were retained in
the current version of MOVES. The MOVES2014 school bus fuel type distributions were based
on MOBILE6 estimates, originally calculated from 1996 and 1997 IHS bus registration data, for
model years 1982-1996 and are summarized in Table A-10. The Union of Concerned Scientists
estimates that roughly one percent of school buses run on non-diesel fuels, so we have
assumed that one percent of school buses were gasoline fueled for MY 1997 and later.117 The
school bus regulatory class distributions were also derived from 2011 FHWA data118 as listed in
Table A-ll, which were applied to model years prior to 2000 for both gasoline and diesel.
183
-------
Table A-10 Fuel type market s
lares by model year for school buses
Model Year
MOVES Fuel Type
Gasoline
Diesel
1982
67.40%
32.60%
1983
67.62%
32.38%
1984
61.55%
38.45%
1985
48.45%
51.55%
1986
32.67%
67.33%
1987
26.55%
73.45%
1988
24.98%
75.02%
1989
22.90%
77.10%
1990
12.40%
87.60%
1991
8.95%
91.05%
1992
1.00%
99.00%
1993
12.05%
87.95%
1994
14.75%
85.25%
1995
11.43%
88.57%
1996
4.15%
95.85%
1997-1999
1.00%
99.00%
Table A-ll Regulatory class fractions of school buses using 2011 FHWA data
Vehicle Type
MOVES regClassID
41
42
46
47
Total
School Buses
0.0106
0.0070
0.9371
0.0453
1
Appendix A.2.2.5 Single i11111 11.11 ¦ 111hii tati 11 111 > li s
The fuel type and regulatory class distributions for the single unit and combination trucks were
calculated directly from the EPA's sample vehicle counts datasets. The single unit and short-
haul combination truck source types were split between gasoline and diesel only and long-haul
combination trucks only contained diesel vehicles. Single unit vehicles were distributed among
all the heavy-duty regulatory classes (regClasslDs 41, 42, 46 and 47) and combination trucks
were distributed among the MHD and HHD regulatory classes (46 and 47) based on the
underlying sample vehicle data.
184
-------
Appendix B 1990 Age Distributions
For MOVES5, we updated the 1990 age distributions by extending their "tails" in age 30+ to
cover ages 30-40+. This appendix describes the original 1990 age distribution derivation. The
extension from 31 age categories to 41 age categories used the same algorithm as described in
Section 6.1.4 Converting 30-Year Age Distributions to 40-Year Distributions.
Details on the derivations of the other age distributions in MOVES may be found in Appendix C.
Motorcycles
The motorcycle age distributions are based on Motorcycle Industry Council estimates of the
number of motorcycles in use, by model year, in 1990. However, data for individual model
years starting from 1978 and earlier were not available. A logarithmic regression curve (R2 value
= 0.82) was fitted to available data, which was then used to extrapolate age fractions for earlier
years beginning in 1978.
Passenger Cars
To determine the 1990 age fractions for passenger cars, we began with IHS NVPP® 1990 data on
car registration by model year. However, this data presents a snapshot of registrations on July
1, 1990 and we needed age fractions as of December 31, 1990. To adjust the values, we used
monthly data from the IHS new car database to estimate the number of new cars registered in
the months July through December 1990. Model Year 1989 cars were added to the previous
estimate of "age 1" cars and Model Year 1990 and 1991 cars were added to the "age 0" cars.
Also the 1990 data did not detail model year for ages 15+. Hence, regression estimates were
used to extrapolate the age fractions for individual ages 15+ based on an exponential curve (R2
value =0.67) fitted to available data.
11 Trucks
For the 1990 age fractions for passenger trucks, light commercial trucks, refuse trucks, short-
haul and long-haul single unit trucks and short-haul and long-haul combination trucks, we used
data from the TIUS92 (1992 Truck Inventory and Use Survey) database. Vehicles in the TIUS92
database were assigned to MOVES source types as summarized in Table B-l. TIUS92 does not
include a model year field and records ages as 0 through 10 and 11-and-greater. Because we
needed greater detail on the older vehicles, we determined the model year for some of the
older vehicles by using the responses to the questions "How was the vehicle obtained?" (TIUS
field "OBTAIN") and "When did you obtain this vehicle?" (TIUS field "ACQYR") and we adjusted
the age-ll-and-older vehicle counts by dividing the original count by model year by the fraction
of the older vehicles that were coded as "obtained new."
185
-------
Ta
ble B-l VIUS1997 codes used for distinguishing truck source types
Source Type
Axle Arrangement
Primary Area of
Operation
Body Type
Major Use
Passenger
2 axle/4 tire
Any
Any
personal
Trucks
(AXLRE= 1,5,6,7)
transportation
(MAJUSE=20)
Light
2 axle/4 tire
Any
Any
any but personal
Commercial
(AXLRE= 1,5,6,7)
transportation
Trucks
Refuse Trucks
Single Unit
(AXLRE=2-4, 8-16)
Off-road, local
or short-range
(AREAOP <=4)
Garbage hauler
(BODTYPE=30)
Any
Single Unit
Single Unit
Off-road, local
Any except
Any
Short-Haul
(AXLRE=2-4, 8-16)
or short-range
garbage hauler
Trucks
(AREAOP<=4)
Single Unit
Single Unit
Long-range
Any
Any
Long-Haul
(AXLRE=2-4, 8-16)
(AREAOP>=5)
Trucks
Combination
Combination
Off-road, local
Any
Any
Short-Haul
(AXLRE>=17)
or medium
Trucks
(AREAOP<=4)
Combination
Combination
Long-range
Any
Any
Long-Haul
(AXLRE>=17)
(AREAOP>=5)
Trucks
Other Buses
For 1990, we were not able to identify a data source for estimating age distributions of other
buses. Because the purchase and retirement of these buses is likely to be driven by general
economic forces rather than trends in government spending, we will use the 1990 age
distributions that were derived for short-haul combination trucks, as described above.
I" School Bus ' 111 Motor Homes
To determine the age fractions of school buses and motor homes, we used information from
the IHS TIP® 1999 database. School bus and motor home counts were available by model year.
Unlike the IHS data for passenger cars, these counts reflect registration at the end of the
calendar year and, thus, did not require adjustment. We converted model year to age and
calculated age fractions. Because we did not have access to 1990 data, these fractions were
used for 1990.
B6. sit Buses
For 1990 Transit Bus age distributions, we used the MOBILE6 age fractions since 1990 data on
transit buses was not available from the Federal Transit Administration database. MOBILE6 age
fractions were based on fitting curves through a snapshot of vehicle registration data as of July
1, 1996, which was purchased from IHS (then known as R.L. Polk Company). To develop a
general curve, the 1996 model year vehicle populations were removed from the sample
because it did not represent a full year and a best-fit analysis was performed on the remaining
population data. The best-fit analyses resulted in age distribution estimates for vehicles ages 1
186
-------
through 25+. However, since the vehicle sales year begins in October, the estimated age 1
population was multiplied by 0.75 to account for the fact that approximately 75 percent of the
year's sales will have occurred by July 1st of a given calendar year.
Both Weibull curve fitting and exponential curve fitting were used to create the age
distributions. The nature of the Weibull curve fitting formula is to produce an "S" shaped curve,
which is relatively flat for the first third of the data, decreases rapidly for the next third and
flattens again for the final third. While using this formula resulted in a better overall fit for
transit buses, the flatness of the final third for each curve resulted in unrealistically low vehicle
populations for the older vehicle ages. For this reason, the original Weibull curve was used
where it fit best and exponential curves were fit through the data at the age where the Weibull
curves began to flatten. Table B-2 presents the equations used to create the age distribution
and the years in which the equations were used.
B-2 Curve fit equations for registration distribution data
Vehicle
Age
Equation
1-17
(( aSe \12.53214119\
y = 3462 * e (A17.16909475J J
18-25+
24987.0776 * e-0,2000*age
187
-------
Appendix C Detailed Derivation of Forecast and Backcast Age Distributions
Since purchasing vehicle registration data for all calendar years is prohibitively costly for historic
years, the base age distribution described in Section 6.1 is forecast and backcast for all other
calendar years in the model. While sales data for historic years are well known and projections
for future years are common in economic modeling, national trends in vehicle survival for every
MOVES source type at all ages are not well studied. For the analysis presented in this appendix,
a generic survival rate was scaled up or down for each calendar year based on our assumptions
of sales and changes in total populations. The following sections summarize the derivation of
the generic survival rate, the estimation of vehicle sales by source type and the algorithms used
to forecast and backcast age distributions for each year.
CI. Vehi' rail by Source Type
The survival rate describes the fraction of vehicles of a given source type and age that remain
on the road from one year to the next. Although this rate changes from year to year, a single
generic rate was calculated from available data.
Survival rates for motorcycles were calculated based on a smoothed curve of retail sales and
2008 national registration data as described in a study conducted for the EPA.51 Survival rates
for passenger cars, passenger trucks and light commercial trucks came from NHTSA's
survivability Table 3 and Table 4.119 These survival rates are based on a detailed analysis of
vehicle registration data from 1977 to 2002. We modified these rates to be consistent with the
MOVES format using the following guidelines:
• NHTSA rates for light trucks were used for both the MOVES passenger truck and light
commercial truck source types.
• MOVES calculates emissions forvehicles up to age 40 (with all older vehicles lumped
into the age 40 category), but NHSTA car survival rates were available only to age 25.
Therefore, we extrapolated car rates to age 40 using the estimated survival rate
equation in Section 3.1 of the NHTSA report. When converted to MOVES format, this
caused a striking discontinuity at age 26 which we removed by interpolating between
ages 25 and 27.
• According to the NHTSA methodology, NHTSA age 1 corresponds to MOVES agelD 2, so
the survival fractions were shifted accordingly.
• Because MOVES requires survival rates for agelDs < 2, these values were linearly
interpolated with the assumption that the survival rate prior to agelD 0 is 1.
• NHTSA defines survival rate as the ratio of the number of vehicles remaining in the fleet
at a given year as compared to a base year. However, MOVES defines the survival rate
as the ratio of vehicles remaining from one year to the next, so we transformed the
NHTSA rates accordingly.
Quantitatively, the following piecewise formulas were used to derive the MOVES survival rates.
In them, sa represents the MOVES survival rate at age a and oa represents the NHTSA survival
rate at age a. When this generic survival rate is discussed below, the shorthand notation S0 will
188
-------
represent a one-dimensional array of sa values at each permissible age a as described in
Equation C-l through Equation C-3 below:
Age 0:
Equation C-l
Age 1:
s1 = l-
2(1 - ff2)
Equation C-2
3
Ages 2-40:
°a-1
sa ~ s2...40 ~ ~
Equation C-3
With limited data available on heavy-duty vehicle scrappage, survivability for all other source
types came from the Transportation Energy Data Book.120 We used the heavy-duty vehicle
survival rates for model year 1980 (TEDB40, Table 3.16). The 1990 model year rates were not
used because they were significantly higher than rates for the other model years in the analysis
(i.e. 45 percent survival rate for 30 year-old trucks) and seemed unrealistically high. While
limited data exists to confirm this judgment, a snapshot of 5-year survival rates was derived
from VIUS 1992 and 1997 results for comparison. According to VIUS, the average survival rate
for model years 1988-1991 between the 1992 and 1997 surveys was 88 percent. The
comparable survival rate for 1990 model year heavy-duty vehicles from TEDB was 96 percent,
while the rate for 1980 model year trucks was 91 percent. This comparison lends credence to
the decision that the 1980 model year survival rates are more in line with available data. TEDB
does not have separate survival rates for medium-duty vehicles; the heavy-duty rates were
applied uniformly across the bus, single unit truck and combination truck categories. The TEDB
survival rates were transformed into MOVES format in the same way as the NHTSA rates.
Additionally, since those survival rates did not extend beyond age 30, we extended the trend in
the data to age 40.
The resulting survival rates are listed in the default database's SourceTypeAge table, shown
below in Table C-l. Please note that since MOVES does not calculate age distributions during a
run, these survival rates are not actively used by MOVES. However, they were used in the
development of the national age distributions stored in the SourceTypeAgeDistribution table
and remain in the default database for reference. In addition, the survival rates in the
SourceTypeAge table are listed by source type, but the values are identical for the grouping of
vehicles listed in the table header.
189
-------
Table C-l Vehicle survival rate by age
Age
Motorcycle
s
Passenger
Cars
Light-duty Trucks
(Passenger and
Light Commercial)
Heavy-duty Vehicles
(Buses, Single Unit Trucks
and Combination Trucks)
0
1.000
0.997
0.991
1.000
1
0.979
0.997
0.991
1.000
2
0.940
0.997
0.991
1.000
3
0.940
0.993
0.986
1.000
4
0.940
0.990
0.981
0.990
5
0.940
0.986
0.976
0.980
6
0.940
0.981
0.970
0.980
7
0.940
0.976
0.964
0.970
8
0.940
0.971
0.958
0.970
9
0.940
0.965
0.952
0.970
10
0.940
0.959
0.946
0.960
11
0.940
0.953
0.940
0.960
12
0.940
0.912
0.935
0.950
13
0.940
0.854
0.929
0.950
14
0.940
0.832
0.913
0.950
15
0.940
0.813
0.908
0.940
16
0.940
0.799
0.903
0.940
17
0.940
0.787
0.898
0.930
18
0.940
0.779
0.894
0.930
19
0.940
0.772
0.891
0.920
20
0.940
0.767
0.888
0.920
21
0.940
0.763
0.885
0.920
22
0.940
0.760
0.883
0.910
23
0.940
0.757
0.880
0.910
24
0.940
0.757
0.879
0.910
25
0.940
0.754
0.877
0.900
26
0.940
0.754
0.875
0.900
27
0.940
0.567
0.875
0.900
28
0.940
0.752
0.873
0.890
29
0.940
0.752
0.872
0.890
30
0.940
0.751
0.871
0.890
31
0.940
0.750
0.870
0.880
32
0.940
0.750
0.869
0.880
33
0.940
0.749
0.868
0.880
34
0.940
0.748
0.867
0.870
35
0.940
0.747
0.866
0.870
36
0.940
0.746
0.865
0.870
37
0.940
0.746
0.864
0.860
38
0.940
0.745
0.863
0.860
39
0.940
0.744
0.862
0.860
40
0.940
0.743
0.861
0.850
190
-------
C2. Vehicle Sales by Source Type
Knowing vehicle sales by source type for every calendar year is essential for estimating age
distributions in both historic and projected years. Since MOVES doesn't calculate age
distributions at run time, this information isn't stored in the default database.bb However, sales
data are used in the age distribution backcasting and projection algorithms, which are
described in subsequent sections. They are also used in calculating the age 0 fractions of
vehicles in the base age distribution, which is described in Section 6.1.1.
Historic motorcycles sales came from the Motorcycle Industry Council's 2015 Motorcycle
Statistical Annual,121 which contains estimates of annual on-highway motorcycle sales going
back to 1989. Sales for calendar years 2015-2022 were estimated as a constant proportion of
the total motorcycle stock, using the ratio of 2014 sales to population.
Historic passenger car sales came from the TEDB40 Table 4.6 estimate for total new retail car
sales.
Historic light truck sales came from the TEDB40 Table 4.7 estimate for total light truck sales.
These were then split into passenger truck and light commercial truck sales using the source
type distribution fractions described in Section 4.1.
Historic school bus sales came from the 2001, 2010, 2020, and 2024 publications of School Bus
Fleet Fact Book.26 Each publication contains estimates for 10 years of historic annual national
sales. Sales for before 1990 were estimated as a constant proportion of the total school bus
stock, using the ratio of 1999 sales to population.
Historic transit bus sales were calculated from the Federal Transit Administration's National
Transit Database (NTD)27 data series on Revenue Vehicle Inventory and Rural Revenue Vehicle
Inventory. Since the annual publication does not necessarily contain all model year vehicles sold
in the year of publication, transit bus sales are instead estimated from 1-year-old buses. This
assumes 0 scrappage of new transit buses, which is consistent with the heavy-duty survival rate
presented in Table C-l. The 1-year-old transit bus populations were estimated from the NTD
active fleet vehicles using the definition of a transit bus as given in Section 2.2.4. Since the
Revenue Vehicle Inventory tables are not available for years before 2002, sales for 1990 and
1999-2001 were estimated as a constant proportion of the total transit bus stock, using the
ratio of 2002 sales to population.
bb Early versions of MOVES calculated age distributions at runtime and therefore required sales data to be stored in
the default database. Consequently, the SourceTypeYear table has a salesGrowthFactor column. Since MOVES no
longer needs this information, this column contains 0s in the current MOVES default database.
191
-------
Lacking a direct source of historic other bus sales, these were derived from the average sales
rate for school buses and transit buses. The ratio of total school and transit bus sales to school
and transit bus populations was applied to the other bus population, as shown in Equation C-4
below. The historic populations for each of the bus source types were determined as described
in Section 4.1.
„ , ScLleSSChool S^teSf-ransit n _ .
Salesother = — — Popother Equation C-4
PoVschool tOPtransit
Historic sales for heavy-duty trucks were derived from the TEDB40 Table 5.3 estimate for truck
sales by gross vehicle weight. These were translated to source type sales by calculating the
source type distribution for each weight class 3-8 from vehicle registration data22 pulled in
2023. Since the registration data grouped short-haul (52) and long-haul (53) single unit trucks,
sales were further allocated to the individual source types 52 and 53 using the source type
distribution fractions described in Section 4.1.
Projected sales for all source types were derived from AEO2023. Because AEO vehicle
categories differ from MOVES source types, the AEO projected vehicle sales were not used
directly. Instead, ratios of vehicle sales to stock were calculated and applied to the projected
populations (see Section 4.2 for the derivation of projected populations). Since AEO2023 only
projects out to 2050, sales for years 2051-2060 were assumed to continue to grow at the same
growth rate as between 2049 and 2050.
Table C-2 shows the mappings between AEO sales categories and MOVES source types. Where
multiple AEO categories are listed, their values were summed before calculating the sales to
stock ratios. These are the same groupings as presented for the stock categories in Table 4-3
and more details on the selection of the groupings may be found in Section 4.2. We
acknowledge that using sales projections from different vehicle types as surrogates for
motorcycles and buses in particular will introduce additional uncertainty into these projections.
The sales to stock ratios for each year and group were calculated and applied to the projected
source type populations using the mappings given in to derive projected sales for each source
type.
192
-------
Table C-2 Mapping AEO categories to source types for projecting vehicle populations
AEO Sales Category Groupings
MOVES Source Type
Total Car Sales'
11 - Motorcycle
21 - Passenger Car
Total Light Truck Sales'
+
Total Commercial Light Truck Sales"
31 - PassengerTruck
32 - Light Commercial Truck
Total Sales'"
41-Other Bus
42-Transit Bus
43 - School Bus
Light Medium Subtotal Sales'"
+
Medium Subtotal Sales'"
51 - Refuse Truck
52 - Single Unit Short-haul Truck
53 - Single Unit Long-haul Truck
54 - Motor Home
Heavy Subtotal Sales'"
61 - Combination Short-haul Truck
62 - Combination Long-haul Truck
1 From AEO2023 Table 38: Light-Duty Vehicle Sales by Technology Type
" From AEO2023 Table 44: Transportation Fleet Car and Truck Sales by Type and
Technology
111 From AEO2023 Table 49: Freight Transportation Energy Use
C3. Historic Age Distributions
The algorithm for backcasting age distributions is as follows:
1. Calculate the base population distribution (Py) by multiplying the base age distribution
(fy) and base population (Py).
2. Remove the age 0 vehicles {Ny).
3. Decrease the population age index by one (for example, 3-year-old vehicles are
reclassified as 2-year-old vehicles).
4. Add the vehicles that were removed in the previous year (Ry_1).
5. Convert the resulting population distribution into an age distribution using Equation 6-1.
6. Replace the new age 39 and 40+ fractions with the base year age 39 and 40+ fractions
and renormalize the new age distribution to sum to 1 without changing the age 39 and
40+ fractions.
7. This results in the previous year age distribution (fy_i). If this algorithm is to be
repeated, fy_1 becomes fy for the next iteration.
This is mathematically described with the following equation (reprinted from Section 6.1.2 for
reference):
193
-------
Py-1 — Py ~ Ny + Ry-l
Equation 6-2
Unfortunately, as described in Section CI, the only survival information we have is a single
snapshot. Because vehicle populations and new sales change differentially (for example, the
total vehicle populations as shown in Section 4.1 leveled off early on in the pandemic; at the
same time, sales of most vehicle types plummeted), it is important to adjust the survival curve
in response to changes in population and sales. We did so by defining a scalar adjustment factor
ky that can be algebraically calculated from population and sales estimates. Its use in
calculating the scrapped vehicles with generic survival rate S0 is given by Equation C-5. Note
that the open circle operator (°) represents entrywise product; that is, each element in an array
is multiplied by the corresponding element in the other one and it results in an array with the
same number of elements. In this case, the scalar adjustment factor is applied to the scrappage
rate (1 minus the survival rate) at each age, which is then applied to the population of vehicles
at each corresponding age; this results in the number of removed vehicles by age.
Ry_1 = ky_t ¦ (l - 5^) o Py_1 Equation C-5
Substituting Equation C-5 into Equation 6-2 yields Equation C-6:
P^ = Ty - JVy + ky_x ¦ (l - T0) o P^[ Equation C-6
To solve for ky_lt Equation C-6 can be transformed into Equation C-7 using known total
populations and sales:
Py-1 = Py- Ny + ky_t ¦ ^ ((l - S0) ° Py—i ^ Equation C-7
However, this still leaves a Py-i term, which is unavoidable because the total number of
vehicles removed is dependent on the age distribution of those vehicles. To solve Equation C-7,
an iterative approach was used. The first time the algorithm described above is run, Py-i is
approximated by applying the base age distribution fy to the population of the previous year
Py-i¦ The scaling factor ky_t is calculated using this approximation in Equation C-7 and then a
guess for Py-\ is calculated from Equation C-6. The guess for the resulting age distribution fy_1
is then calculated using the known Py_The algorithm is repeated for the same year using the
updated guess for the resulting age distribution. This is repeated until the resulting age
distribution matches the guessed age distribution at each age fraction within lxlO6, which
occurred within 10 iterations for most source types and calendar years.
This algorithm was then repeated for each historic year from 2022 to 1999 and for each source
type using the following data sources:
• Total populations Py and Py-\ as described in Section 4.
194
-------
• Generic survival rates S0 as described in Section CI.
• Vehicle sales Ny as described in Section C2.
• Base age distributions /2023 as described in Section 6.1.1. All other fy come from the
/y_! of the previous iteration.
With all of this information, the age distributions were algorithmically determined for years
1999-2022 and are stored in the SourceTypeAgeDistribution table of the default database.
Projected Age Distributions
The base algorithm for forecasting age distributions is as follows:
1. Calculate the base population distribution (Py) by multiplying the base age distribution
(fy) and base population (Py).
2. Remove the vehicles that did not survive (Ry) at each age level.
3. Increase the population age index by one (for example, 3-year-old vehicles are
reclassified as 4-year-old vehicles).
4. Add new vehicle sales (Ny+1) as the age 0 cohort.
5. Convert the resulting population distribution into an age distribution using Equation 6-1.
6. Replace the new age 40+ fraction with the base year age 40+ fraction and renormalize
the new age distribution to sum to 1 without changing the age 0 and the age 40+
7. This results in the next year age distribution (fy+1). If this algorithm is to be repeated,
fy+1 becomes fy for the next iteration.
This is mathematically described with the following equation (reprinted from Section 6.1.3 for
reference):
As with the backcasting algorithm, the scrapped vehicles need to be estimated by scaling the
generic survival rate. The equation governing vehicle removal discussed the previous section is
also applicable here. Taking careful note of the subscripts, Equation 6-3 and Equation C-5 can
be combined into Equation C-8:
To solve for ky, Equation C-8 can be transformed into Equation C-9 using the population and
sales totals:
fractions.
Equation 6-3
Py.(-I — Py ky " (1 Sq) ° Py + Ny+^
Equation C-8
Equation C-9
195
-------
This can be algebraically solved for ky and evaluated for each source type as all of the other
values are known. Please note that the iterative approach to solving this equation as described
in the back-casting section is not necessary here, as the number of scrapped vehicles depends
on the base age distribution, which is known. After ky is calculated, Equation C-8 is used to
determine Py+l- The resulting age distribution fy+1 is then calculated using the known Py+l-
This algorithm was then repeated for each projected year from 2024 to 2060 and for each
source type using the following data sources:
• Total populations Py and Py+i as described in Section 4.
• Generic survival rates S0 as described in Section CI.
• Vehicle sales Ny+1 as described in Section C2.
• Base age distributions /2023 as described in Section 6.1.1. All other fy come from the
fy+1 of the previous iteration.
With all of this information, the age distributions were algorithmically determined for years
2024-2060 and are stored in the SourceTypeAgeDistribution table of the default database. To
illustrate the backcast and forecast algorithm, Figure C-l presents four selected age
distributions: the base year 2023 distribution, one backcast distribution (2010), and two
forecast distributions (2030 and 2040). For ease of comparing multiple age distributions for
different calendar years, the x-axis is model year (calculated as the calendar year minus age),
not age.
196
-------
1970 1980 1990 2000 2010 2020 2030 2040
Model Year
Figure C-l Selected age distributions for passenger cars in MOVES5
197
-------
Appendix D Driving Schedules
A key feature of MOVES is the capability to accommodate a number of drive schedules to
represent driving patterns across source type, roadway type and average speed. For the
national default case, MOVES5 employs 49 drive schedules with various average speeds,
mapped to specific source types and roadway types. These are unchanged from MOVES2014.
Table D-l below lists the driving schedules used in MOVES. Some driving schedules are used for
both restricted access (freeway) and unrestricted access (non-freeway) driving. Some driving
schedules are used for multiple source types or multiple road types where vehicle specific
information was not available.
198
-------
Drive
tied i
ID
101
153
158
201
202
203
204
205
206
251
252
253
254
255
301
302
303
304
305
306
351
352
353
354
355
396
Table D-l MOVES default driving sc
hedule statistics
Drive Schedule Name
Avg
Speed
Max
Speed
Idle
Time
(sec)
Percent of
Time Idling
Miles
Time (sec)
Minutes
LD Low Speed 1
2.5
10.00
280
46.5%
0.419
602.00
10.03
LD LOS E Freeway
30.5
63.00
1.1%
3.863
456.00
7.60
LD High Speed Freeway 3
76.0
90.00
0.0%
12.264
581.00
9.68
MD 5mph Non-Freeway
4.6
24.10
85
29.0%
0.373
293.00
4.88
MD lOmph Non-Freeway
10.7
34.10
61
19.6%
0.928
311.00
5.18
MD 15mph Non-Freeway
15.6
36.60
57
12.6%
1.973
454.00
7.57
MD 20mph Non-Freeway
20.8
44.50
95
9.1%
6.054
1046.00
17.43
MD 25mph Non-Freeway
24.5
47.50
63
11.1%
3.846
566.00
9.43
MD 30mph Non-Freeway
31.5
55.90
54
5.5%
8.644
988.00
16.47
MD 30mph Freeway
34.4
62.60
0.0%
15.633
1637.00
27.28
MD40mph Freeway
44.5
70.40
0.0%
43.329
3504.00
58.40
MD 50mph Freeway
55.4
72.20
0.0%
41.848
2718.00
45.30
MD 60mph Freeway
60.1
68.40
0.0%
81.299
4866.00
81.10
MD High Speed Freeway
72.8
80.40
0.0%
96.721
4782.00
79.70
HD 5mph Non-Freeway
5.8
19.90
37
14.2%
0.419
260.00
4.33
HD lOmph Non-Freeway
11.2
29.20
70
11.5%
1.892
608.00
10.13
HD 15mph Non-Freeway
15.6
38.30
73
12.9%
2.463
567.00
9.45
HD 20mph Non-Freeway
19.4
44.20
84
15.1%
3.012
558.00
9.30
HD 25mph Non-Freeway
25.6
50.70
57
5.8%
6.996
983.00
16.38
HD 30mph Non-Freeway
32.5
58.00
43
5.3%
7.296
809.00
13.48
HD 30mph Freeway
34.3
62.70
0.0%
21.659
2276.00
37.93
HD 40mph Freeway
47.1
65.00
0.0%
41.845
3197.00
53.28
HD 50mph Freeway
54.2
68.00
0.0%
80.268
5333.00
88.88
HD 60mph Freeway
59.7
69.00
0.0%
29.708
1792.00
29.87
HD High Speed Freeway
71.7
81.00
0.0%
35.681
1792.00
29.87
HD High Speed Freeway Plus 5mph
76.7
86.00
0.0%
38.170
1792.00
29.87
199
-------
Drive
Schedule
ID
Drive Schedule Name
Avg
Speed
Max
Speed
Idle
Time
(sec)
Percent of
Time Idling
Miles
Time (sec)
Minutes
Hours
397
MD High Speed Freeway Plus 5mph
77.8
85.40
0
0.0%
103.363
4782.00
79.70
1.328
398
CRC E55 HHDDT Creep
1.8
8.24
107
42.3%
0.124
253.00
4.22
0.070
401
Bus Low Speed Urban
3.1
19.80
288
63.9%
0.393
451.00
7.52
0.125
402
Bus 12mph Non-Freeway
11.5
33.80
109
37.5%
0.932
291.00
4.85
0.081
403
Bus 30mph Non-Freeway
21.9
47.00
116
28.3%
2.492
410.00
6.83
0.114
404
New York City Bus
3.7
30.80
403
67.2%
0.615
600.00
10.00
0.167
405
WMATA Transit Bus
8.3
47.50
706
38.4%
4.261
1840.00
30.67
0.511
501
Refuse Truck Urban
2.2
20.00
416
66.9%
0.374
622.00
10.37
0.173
1009
Final FCOILOSAF Cycle (C10R04-
00854)
73.8
84.43
0
0.0%
11.664
569.00
9.48
0.158
1011
Final FC02LOSDF Cycle (C10R05-
00513)
49.1
73.06
34
5.0%
9.283
681.00
11.35
0.189
1017
Final FC11LOSB Cycle (C10R02-
00546)
66.4
81.84
0
0.0%
9.567
519.00
8.65
0.144
1018
Final FC11LOSC Cycle (C15R09-
00849)
64.4
78.19
0
0.0%
16.189
905.00
15.08
0.251
1019
Final FC11LOSD Cycle (C15R10-
00068)
58.8
76.78
0
0.0%
11.922
730.00
12.17
0.203
1020
Final FC11LOSE Cycle (C15R11-00851)
46.1
71.50
1
0.1%
12.468
973.00
16.22
0.270
1021
Final FC11LOSF Cycle (C15R01-00876)
20.6
55.48
23
2.5%
5.179
905.00
15.08
0.251
1024
Final FC12LOSC Cycle (C15R04-
00582)
63.7
79.39
0
0.0%
15.685
887.00
14.78
0.246
1025
Final FC12LOSD Cycle (C15R09-
00037)
52.8
73.15
12
1.5%
11.754
801.00
13.35
0.223
1026
Final FC12LOSE Cycle (C15R10-00782)
43.3
70.87
0
0.0%
10.973
913.00
15.22
0.254
1029
Final FC14LOSB Cycle (C15R07-
00177)
31.0
63.81
27
3.6%
6.498
754.00
12.57
0.209
1030
Final FC14LOSC Cycle (C10R04-
00104)
25.4
53.09
41
8.0%
3.617
513.00
8.55
0.143
1033
Final FC14LOSF Cycle (C15R05-00424)
8.7
44.16
326
38.2%
2.066
853.00
14.22
0.237
200
-------
Drive
Schedule
ID
Drive Schedule Name
Avg
Speed
Max
Speed
Idle
Time
(sec)
Percent of
Time Idling
Miles
Time (sec)
Minutes
Hours
1041
Final FC17LOSD Cycle (C15R05-
00480)
18.6
50.33
114
16.1%
3.659
709.00
11.82
0.197
1043
Final FC19LOSAC Cycle (C15R08-
00267)
15.7
37.95
67
7.7%
3.802
870.00
14.50
0.242
201
-------
Appendix E Total Idle Fraction Regression Coefficients
Table E-l displays the regression coefficients for the linear model used to estimate variation in
total idle fraction for light-duty vehicles presented in Equation 10-6 discussed in Section 10.2.3.
Table E-l Total idle fraction regression coefficients for light-duty vehicles trucks in urban
counties for weekdays
Variable
Coefficients
Comments
(Intercept)
0.20977
daylD5
0.011262
Applicable when daylD=5
sourceTypelD31
0.001329
Applicable when sourceTypelD=31
countyTypelDl
0.03058
Applicable when equation is used for an urban county
(countyTypelD=l)
idleRegionlD104
0.021342
Appl
cable when idleRegionlD=104
idleRegionlD102
0.026097
Appl
cable when idleRegionlD=102
idleRegionlD103
0.05461
Appl
cable when idleRegionlD=103
idleRegionlDlOl
0.057216
Appl
cable when idleRegionlD=101
monthlD2
0.002789
Appl
cable when monthlD=2
monthlD3
-0.00429
Appl
cable when monthlD=3
monthlD4
-0.00609
Appl
cable when monthlD=4
monthlD5
-0.00412
Appl
cable when monthlD=5
monthlD6
-0.00264
Appl
cable when monthlD=6
monthlD7
0.002914
Appl
cable when monthlD=7
monthlD8
-0.00066
Appl
cable when monthlD=8
monthlD9
-0.00296
Appl
cable when monthlD=9
monthlDIO
0.007288
Appl
cable when monthlD=10
monthlDll
0.00585
Appl
cable when monthlD=ll
monthlD12
0.007586
Appl
cable when monthlD=12
dleRegionlD104:monthlD2
-0.01478
Appl
cable when monthlD=2 and
dleRegionlD=104
dleRegionlD102:monthlD2
-0.00664
Appl
cable when monthlD=2 and
dleRegionlD=102
dleRegionlD103:monthlD2
-0.0173
Appl
cable when monthlD=2 and
dleRegionlD=103
dleRegionlD101:monthlD2
-0.01595
Appl
cable when monthlD=2 and
dleRegionlD=101
dleRegionlD104:monthlD3
-0.02666
Appl
cable when monthlD=3 and
dleRegionlD=104
dleRegionlD102:monthlD3
-0.01167
Appl
cable when monthlD=3 and
dleRegionlD=102
dleRegionlD103:monthlD3
-0.04358
Appl
cable when monthlD=3 and
dleRegionlD=103
dleRegionlD101:monthlD3
-0.0334
Appl
cable when monthlD=3 and
dleRegionlD=101
dleRegionlD104:monthlD4
-0.02855
Appl
cable when monthlD=4 and
dleRegionlD=104
dleRegionlD102:monthlD4
-0.01194
Appl
cable when monthlD=4 and
dleRegionlD=102
dleRegionlD103:monthlD4
-0.04759
Appl
cable when monthlD=4 and
dleRegionlD=103
dleRegionlD101:monthlD4
-0.03841
Appl
cable when monthlD=4 and
dleRegionlD=101
dleRegionlD104:monthlD5
-0.04011
Appl
cable when monthlD=5 and
dleRegionlD=104
dleRegionlD102:monthlD5
-0.01453
Appl
cable when monthlD=5 and
dleRegionlD=102
dleRegionlD103:monthlD5
-0.05713
Appl
cable when monthlD=5 and
dleRegionlD=103
dleRegionlD101:monthlD5
-0.0465
Appl
cable when monthlD=5 and
dleRegionlD=101
dleRegionlD104:monthlD6
-0.04388
Appl
cable when monthlD=6 and
dleRegionlD=104
dleRegionlD102:monthlD6
-0.01298
Appl
cable when monthlD=6 and
dleRegionlD=102
202
-------
Variable
Coefficients
Comments
idleRegionlD103:monthlD6
-0.05729
Appl
cable when monthlD=6 and
dleRegionlD=103
idleRegionlD101:monthlD6
-0.05025
Appl
cable when monthlD=6 and
dleRegionlD=101
idleRegionlD104:monthlD7
-0.04935
Appl
cable when monthlD=7 and
dleRegionlD=104
idleRegionlD102:monthlD7
-0.0138
Appl
cable when monthlD=7 and
dleRegionlD=102
idleRegionlD103:monthlD7
-0.06494
Appl
cable when monthlD=7 and
dleRegionlD=103
idleRegionlD101:monthlD7
-0.05502
Appl
cable when monthlD=7 and
dleRegionlD=101
idleRegionlD104:monthlD8
-0.04589
Appl
cable when monthlD=8 and
dleRegionlD=104
idleRegionlD102:monthlD8
-0.01495
Appl
cable when monthlD=8 and
dleRegionlD=102
idleRegionlD103:monthlD8
-0.06051
Appl
cable when monthlD=8 and
dleRegionlD=103
idleRegionlD101:monthlD8
-0.05
Appl
cable when monthlD=8 and
dleRegionlD=101
idleRegionlD104:monthlD9
-0.04807
Appl
cable when monthlD=9 and
dleRegionlD=104
idleRegionlD102:monthlD9
-0.02195
Appl
cable when monthlD=9 and
dleRegionlD=102
idleRegionlD103:monthlD9
-0.06001
Appl
cable when monthlD=9 and
dleRegionlD=103
idleRegionlD101:monthlD9
-0.04851
Appl
cable when monthlD=9 and
dleRegionlD=101
idleRegionlD104:monthlDl
0
-0.05049
Applicable when monthlD=10 and idleRegionlD=104
idleRegionlD102:monthlDl
0
-0.03221
Applicable when monthlD=10 and idleRegionlD=102
idleRegionlD103:monthlDl
0
-0.06831
Applicable when monthlD=10 and idleRegionlD=103
idleRegionlD101:monthlDl
0
-0.05287
Applicable when monthlD=10 and idleRegionlD=101
idleRegionlD104:monthlDl
1
-0.02092
Applicable when monthlD=ll and idleRegionlD=104
idleRegionlD102:monthlDl
1
-0.0262
Applicable when monthlD=ll and idleRegionlD=102
idleRegionlD103:monthlDl
1
-0.04514
Applicable when monthlD=ll and idleRegionlD=103
idleRegionlD101:monthlDl
1
-0.04651
Applicable when monthlD=ll and idleRegionlD=101
idleRegionlD104:monthlDl
2
-0.0075
Applicable when monthlD=12 and idleRegionlD=104
idleRegionlD102:monthlDl
2
-0.02558
Applicable when monthlD=12 and idleRegionlD=102
idleRegionlD103:monthlDl
2
-0.04263
Applicable when monthlD=12 and idleRegionlD=103
idleRegionlD101:monthlDl
2
-0.04724
Applicable when monthlD=12 and idleRegionlD=101
Table E-2 shows a sample calculation of MOVES default total idle fractions using the coefficients
for passenger cars (sourceTypelD=21) in rural counties (countyTypelD=0) in idleRegionlD=101
(represented by New Jersey). The total idle fractions for all the sourceTypelD 21 and 32 derived
from the TIF regression equation is available in the MOVES totalldleFraction table.
203
-------
Table E-2 Example total idle fractions for rural New Jersey passenger cars
sourceTypelD
monthID
daylD
idleRegionID
countyTypelD
TIF
21
1
2
101
0
0.2670
21
2
2
101
0
0.2538
21
3
2
101
0
0.2293
21
4
2
101
0
0.2225
21
5
2
101
0
0.2164
21
6
2
101
0
0.2141
21
7
2
101
0
0.2149
21
8
2
101
0
0.2163
21
9
2
101
0
0.2155
21
10
2
101
0
0.2214
21
11
2
101
0
0.2263
21
12
2
101
0
0.2273
21
1
5
101
0
0.2782
21
2
5
101
0
0.2651
21
3
5
101
0
0.2406
21
4
5
101
0
0.2337
21
5
5
101
0
0.2276
21
6
5
101
0
0.2254
21
7
5
101
0
0.2261
21
8
5
101
0
0.2276
21
9
5
101
0
0.2268
21
10
5
101
0
0.2327
21
11
5
101
0
0.2376
21
12
5
101
0
0.2386
204
-------
Appendix F Source Masses for Light-Duty Vehicles
In M0VES5, the source masses of light-duty vehicles were unchanged from MOVES2010b. This
appendix describes the derivation of these source masses. Information on the updated source
masses for heavy-duty vehicles is provided in Section 15.
In MOVES2010b, weight data (among other kinds of information) were used to allocate source
types to source bins using a field called weightClasslD. While that information is no longer used
in MOVES and has not been updated, it provides a reasonable basis for estimating source mass
for the MOVES source types. As described in Equation F-l, each source type's source mass was
calculated using an activity-weighted average of their associated source bins' midpoint weights:
where M is the source mass factor for the source type, fa is the age fraction at age a, ab is the
source bin activity fraction for source bin b and m is the vehicle midpoint mass. Table F-l lists
the vehicle midpoint mass for each weightClasslD. The source bin activity fraction in
MOVES2010b is a calculated value of activity based on fuel type, engine technology, regulatory
class, model year, engine size and weight class.
Equation F-l
205
-------
Table F-l MQVES2010b weight classes
WeightClassID
Weight Class Name
Midpoint
Weight
0
Doesn't Matter
[NULL]
20
weight < 2000 pounds
1000
25
2000 pounds <= weight < 2500 pounds
2250
30
2500 pounds <= weight < 3000 pounds
2750
35
3000 pounds <= weight < 3500 pounds
3250
40
3500 pounds <= weight < 4000 pounds
3750
45
4000 pounds <= weight < 4500 pounds
4250
50
4500 pounds <= weight < 5000 pounds
4750
60
5000 pounds <= weight < 6000 pounds
5500
70
6000 pounds <= weight < 7000 pounds
6500
80
7000 pounds <= weight < 8000 pounds
7500
90
8000 pounds <= weight < 9000 pounds
8500
100
9000 pounds <= weight < 10000 pounds
9500
140
10000 pounds <= weight < 14000 pounds
12000
160
14000 pounds <= weight < 16000 pounds
15000
195
16000 pounds <= weight < 19500 pounds
17750
260
19500 pounds <= weight < 26000 pounds
22750
330
26000 pounds <= weight < 33000 pounds
29500
400
33000 pounds <= weight < 40000 pounds
36500
500
40000 pounds <= weight < 50000 pounds
45000
600
50000 pounds <= weight < 60000 pounds
55000
800
60000 pounds <= weight < 80000 pounds
70000
1000
80000 pounds <= weight < 100000 pounds
90000
1300
100000 pounds <= weight < 130000 pounds
115000
9999
130000 pounds <= weight
130000
5
weight < 500 pounds (for MCs)
350
7
500 pounds <= weight < 700 pounds (for MCs)
600
9
700 pounds <= weight (for MCs)
700
The following sections detail how weight classes were assigned to light-duty vehicles in MOVES.
Fl. Motorcycles
The Motorcycle Industry Council Motorcycle Statistical Annual provides information on
displacement distributions for highway motorcycles for model years 1990 and 1998. These
were mapped to MOVES engine displacement categories. Additional EPA certification data
were used to establish displacement distributions for model year 2000. We assumed that
displacement distributions were the same in 1969 as in 1990 and interpolated between the
established values to determine displacement distributions for all model years from 1990 to
206
-------
1997 and for 1999. Values for 2000-and-later model years are based on model year 2000
certification data.
We then applied weight distributions for each displacement category as suggested by EPA
motorcycle experts. The average weight estimate includes fuel and rider. The weight
distributions depended on engine displacement but were otherwise independent of model
year. This information is summarized in Table F-2.
Table F-2 Motorcycle engine size and average weight distributions for selected model years
Displacement
Category
1969 MY
distribution
(assumed)
1990 MY
distribution
(MIC)
1998 MY
distribution
(MIC)
2000 MY
distribution
(certification
data)
Weight distribution
(EPA staff)
0-169 cc (1)
0.118
0.118
0.042
0.029
100%: <=500 lbs.
170-279 cc (2)
0.09
0.09
0.05
0.043
50%: <=500 lbs.
50%: 500lbs. -700lbs.
280+ cc (9)
0.792
0.792
0.908
0.928
30%: 500 lbs.-700 lbs.
70%: > 700lbs.
senger Cars
Passenger car weights come from the 1999 IHS dataset. The weightClassID was assigned by
adding 300 lbs. to the IHS curb weight and grouping into MOVES weight bins. For each fuel
type, model year, engine size and weight bin, the number of cars was summed and fractions
were computed. In general, entries for which data were missing were omitted from the
calculations. Also, analysis indicated a likely error in the IHS data (an entry for 1997 gasoline-
powered Bentleys with engine size 5099 and weight class 20). This fraction was removed and
the 1997 values were renormalized. 1999 model year values were used for all 2000-and-later
model years.
I l ight-Di i liin-Iks
Light truck weights came from VIUS1997 data, which combines information from two different
survey forms. The first form was administered for VIUS "Strata" 1 and 2 trucks: pickup trucks,
panel trucks, vans (including mini-vans), utility type vehicles (including jeeps) and station
wagons on truck chassis. The second form was administered for all other trucks. While both
surveys requested information on engine size, only the second form requested detailed
information on vehicle weight. Thus, for Strata 1 and 2 trucks, VIUS classifies the trucks only by
broad average weight category (AVGCK): 6,000 lbs. or less, 6,001-10,000 lbs., 10,001-14,000
lbs., etc. To determine a more detailed average engine size and weight distribution for these
vehicles, we used an Oak Ridge National Laboratory (ORNL) light-duty vehicle database,
compiled from EPA test vehicle data and Ward's Automotive Inc.122 data, to correlate engine
size with vehicle weight distributions by model year.
For source types 31 and 32 (Passenger Trucks and Light Commercial Trucks) in regClassID 30
(Light-Duty Trucks):
207
-------
• VIUS1997 trucks of the source type in Strata 3, 4 and 5 were assigned to the appropriate
MOVES weight class based on VIUS detailed average weight information.
• VIUS1997 trucks of the source type in Strata 1 and 2 were identified by engine size and
broad average weight category.
• Strata 1 and 2 trucks in the heavier (10,001-14,000 lbs., etc.) VIUS1997 broad categories
were matched one-to-one with the MOVES weight classes.
• For trucks in the lower broad categories (6,000 lbs. or less and 6001-10,000 lbs.), we
used VIUS1997 to determine the fraction of trucks by model year and fuel type that fell
into each engine size/broad weight class combination (the "VIUS fraction").
• We assigned trucks in the ORNL light-duty vehicle database to a weightClassID by adding
300 lbs. to the recorded curb weight and determining the appropriate MOVES weight
class.
• For the trucks with a VIUS1997 average weight of 6,000 lbs. or less, we multiplied the
VIUS1997 fraction by the fraction of trucks with a given weightClassID among the trucks
in the ORNL database that had the given engine size and an average weight of 6,000 lbs.
or less. Note, the ORNL database did not provide information on fuel type, so the same
distributions were used for all fuels.
Because the ORNL database included only vehicles with a GVW up to 8500 lbs., we did
not use it to distribute the trucks with a VIUS1997 average weight of 6,001-10,000 lbs.
Instead these were distributed equally among the MOVES weightClassID 70, 80, 90 and
100.
Note that the source mass for source types 31 and 32 in regClassID 41 (class 2b trucks) was
calculated as described in Section 15.
208
-------
Appendix G NREL Fleet DNA Preprocessing Steps
This appendix discusses the preprocessing steps undertaken on the NREL's Fleet DNA database,
which is used to derive default activity for heavy-duty vehicles, including idle fractions (see
Section 10) and starts activity (see Section 12).
Prior to calculation and preprocessing, the data is collected from the database which involves
loading and combining all of the 1 Hz data from Fleet DNA into a single 1-dimensional data
array for each parameter. Each data file is arranged in the database by vehicle, day and
parameter as shown in Figure G-l. To create one contiguous array per parameter, the
processing script loads each parameter and appends it to the parameter from the previous day
resulting in five 1-D arrays of equal length which can be joined on index.
Fleet
DNA
~T_
Raw
Data
j ~ Vehicle
T
I .
Day
Time.json
Latitude.json
Longitude.json
Engine Speed.json
"X
Wheel Speed.json
Figure G-l Diagram of Fleet DNA database file structure
After collecting the data, a processing step is performed to ensure the data is an accurate
representation of a vehicle's activity. Two of the key activity analyses from this report are
vehicle soak lengths and starts which are defined by the engine speed parameter that indicates
if the vehicle is running or not. A start is calculated by identifying a transition of the engine
speed from 0 to greater than zero and a soak is the length of time the engine was off before it is
started. Both parameter calculations depend on the engine being off; however, in some
instances the data logger will shut off before recording a zero for engine speed raising the
concern that starts and soak times may be missed or not accurately categorized.
To account for these instances in data preprocessing an algorithm was developed to look at the
time stamp and identify large leaps or gaps from one data point to the next. If the algorithm
finds a gap, the engine speed is replaced with a zero at that point to indicate the vehicle's
engine has shut off.
One of the major questions with this time gap method is what time length would constitute an
engine-off event. If the selected time length is too short, then instances such as the logger
updating its timestamp from the GPS may be characterized as a start. Conversely, if the time
length is too long, starts and vehicle soaks may be missed. A possible scenario resulting in a
209
-------
mischaracterization of starts could be when the GPS updates the data logger's clock while
crossing a time zone or the logger pausing its recording for a few seconds when creating a new
log file. Depending on the type of data logger used, some will create a new file at a specified
time interval or when a file size limit is reached requiring the logger to shift computing power
to saving the file to memory. If the gap length is set to an hour or less, the algorithm may count
these normal logger operations as vehicle starts. Similarly, if the logger was taken off of a
vehicle on the west coast and placed on a vehicle on the east coast, the timestamp may jump 3
hours should the GPS update the internal clock to local time.
To avoid these types of timestamp jumps which may show for soak operation modes 101
through 106, the gap length was set to 6 hours for this analysis. Plots of vehicle soak
distribution weighted by start fraction for various gap lengths are provided in Figure G-2 and
Figure G-3 to demonstrate what effect changes in gap length might have. Finally, after running
the gap filling routine, the first and last days of data are eliminated to avoid counting
incomplete or unrepresentative operation when the data logger is being installed or removed.
Plots of vehicle soak distribution weighted by start fraction for gap lengths varying between 1
second and 30 hours are provided in Figure G-2 and Figure G-3 to demonstrate what effect
changes in gap length might have. Figure G-2 provides the distributions for source type 62
which consists of combination long-haul trucks that have very few starts per day and Figure G-3
provides the distributions for source type 52 which consists of single unit short-haul trucks that
have a large number of starts per day. Intuitively the gap length algorithm had the most
noticeable effect on source type 62 due to the high weighting placed on each start as a result of
having very few starts per day.
210
-------
ID 62 Day 5 Gap: 0.02min
ID 62 Day 5 Gap: 6.0min
ID 62 Day 5 Gap: 30.0min
9 12 15 18 21 24
Hour ID
211
-------
Figure G-2 Start fraction weights soak distribution weighted by gap length: source type 62
ID 52 Day 5 Gap: 0.02min
3 6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 60.0min
0.00
ID 52 Day 5 Gap: 6.0min
D 52 Day 5 Gap: 30.0min
3 6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 90.0min
ID 52 Day 5 Gap: 120.0min
!"i
¦ =
¦¦I
4
'Li
I
Op Modes
101
102
103
104
105
106
107
108
Hi
0.08
0.06
i'p =
:-di
Op Modes
101
102
103
104
105
106
107
108
*
hi
3 6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 360.0miri
2 0.06
0.00
|H =
Op Modes
101
102
103
104
105
106
107
108
3 6 9 12 15 18
Hour ID
t; 0.06
,1 =
Op Modes
101
102
11
h
¦
104
105
. .1
III*
3 6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 720.0min
9 12 15 18
Hour ID
ID 52 Day 5 Gap: 18QQ.0min
21 24
9 12 15 18
Hour ID
Figure G-3 Start fraction weights soak distribution weighted by gap length: source type 52
212
-------
Appendix H Averaging Methods for Heavy-duty Telematics Activity Data
Telematics data provides great detail on vehicle activity, but to calculate MOVES inputs for
heavy duty starts, soaks and idle fractions, we need to compute averages across the available
data. Because different averaging methods lead to different results, we evaluated several
different approaches to calculating these averages. The following discussion uses the
calculation of the idle fraction by sourcetypelD and daylD (weekend and weekday) to illustrate
the strengths and weaknesses of each of the methods.
Evaluated Methods
Initally, we used Method 1 (Equation H-l) to average the idle fractions across all vehicles within
the same sourcetypelD and daylD (weekday vs weekend). Method 1 could also be referred to as
an average of ratios. We initially chose to use Method 1 because it is simple to implement and
it equally weights each vehicle in the sample.
£ Idle fractiorii
Idle fractions d =
n
Method 1 -
"Average /' = individual vehicle ID Equation H-l
of Ratios" n = vehicles sampled within each sourcetype
s = sourceTypelD
d = dayTypelD
However, to estimate a representative idle fraction, the vehicle should be weighted by its
contribution to real-world activity. By weighting each individual vehicle idle fraction equally,
Method 1 over-represents the vehicles with little real-world activity (with possibly unrealistic
idle fractions) and under-respresents the idle fractions from the vehicles with the most activity.
We then considered Method 2. shown in Equation H-2, which is referred to as the "Sum over
Sum Method." In Method 2, the average is weighted according to vehicle activity. Vehicles that
are operated for long work days will have more operating hours and idle hours than vehicles
that are only operated intermittently. Multiplying the Idle fraction estimated from Method 2 by
the the total operating hours, £ operating hoursi, will yield the total idle hours,
£ idle hoursi, measured in our sample. This property assures that the relationship between
idle hours and operating time is consistent between our model estimates and the source data.
£ idle hoursi
Idle fractions d = — : ;
2j operating hoursi
Method 2 -
"Sum over . . , , , . , Equation H-2
„ i = individual vehicle ID
Sum
s = sourceTypelD
d = dayTypelD
213
-------
One disadvantage of Method 2 is that the Idle fraction is dependent on the instrumentation
time. For example, a vehicle that is instrumented for two months will be weighted twice as
much in the idle fraction calculation as a vehicle with the same duty cycle that is only measured
for one month. In some cases, using more information from vehicles that are instrumented
longer would be a desirable property; however, if instrumentation times are not random, this
can skew the average to overrepresent certain groups of vehicles. For example, we hope in the
future to develop idle, start and soak inputs from multiple datasets with different
instrumentation times, such as the Fleet DNA dataset with an average instrumentation time of
35 days/vehicle and the HD SCR (CE-CERT) data with an 86-day average instrumentation time.
Using Method 2 to combine data from these two datasets, the HD SCR vehicle data would be
weighted twice as much as the vehicles in the Fleet DNA sample. This is undesirable, because
we have no reason to assume that the HD SCR sample vehicles are more representative of the
national fleet than the Fleet DNA sample vehicles.
In Method 3. we propose using a "Normalized Sum over Sum" approach as shown in Equation
H-3. Method 3 is similar to Method 2, except that the sum of idle hours and the operating hours
from each vehicle is divided (or normalized) by the number of days each sample vehicle was
instrumented. Method 3 controls for the different lengths of time each vehicle is instrumented
and the Idle fraction is weighted most heavily by the vehicles with the most daily average
activity, rather than the most measured activity. Method 3 (Equation H-3) is the current
approach we are using for developing MOVES inputs and is equivalent to
Equation 10-7 presented in Section 10.
Method 3 -
"Normalized
Sum over
Sum"
Idle fractions a =
Y /idle hourSi / \
^ V 'daysj
Y /operating hourSi / \
^ V ' days J
i = individual vehicle ID
days, = # of days vehicle, is instrumented
s = sourceTypelD
d = dayTypelD
Equation H-3
The methods we explained above are also applicable to estimating start fractions. The start
fraction determines at fraction of total daily starts occur at each hour of the day. The following
table contains the equations for the start fractions for each of the three methods.
Method 1— Start fractionhsd
"Ave rage of h= hour of the day
Ratios" / = individual vehicle ID
n = # of sampled vehicles
Y, Start fractionh £
n
Equation H-4
214
-------
s = sourceTypelD
d = dayTypelD
Method 2 -
"Sum over
Sum"
Method 3 -
"Normalized
Sum over
Sum"
Start fractionhsd =
£ starts
h,i
h= hour of the day
i = individual vehicle ID
s = sourceTypelD
d = dayTypelD
Start fractionhsd =
£ startsi
„ /startsKi . \
^ V ' day Si)
y /starts^ / \
^ V ' day Si)
h= hour of the day
i = individual vehicle ID
days, = days vehicle, is instrumented
s = sourceTypelD
d = dayTypelD
Equation H-5
Equation H-6
The three different averaging method was also applied for calculating the soak fractions in the
table below. The soak fraction determines the distribution of starts occuring for the 8 different
start operating modes in MOVES (or soak lengths) as defined in Table 12-3.
Method 1 -
"Average of
Ratios"
Method 2 -
"Sum over
Sum"
Method 3 -
"Normalized
Sum over
Sum"
Soak fractionho s d — —
h= hour of the day
i = Vehicle ID
n = # of sampled vehicles
s = sourceTypelD
d = dayTypelD
Soak fractionho sd =
h= hour of the day
i = Vehicle ID
o = operating mode (soak length)
s = sourceTypelD
d = dayTypelD
^ Soak fractionh i o
n
2startshio
Y, startsh i
Soak fractionhosd =
yi fstartsh i o i \
^ V ' dayst)
(startshij
/1
h= hour of the day
days
,)
Equation
H-7
Equation
H-8
Equation
H-9
215
-------
/ = Vehicle ID
o = operating mode (soak length)
days/ = days vehicle, is instrumented
s = sourceTypelD
d = dayTypelD
H2. Comparison of Evaluated Methods
Figure H-l graphically compares the idle fractions calculated using Method 1 and Method 3,
using the data for single unit short-haul trucks. In the graph, the darker colors represent idling
that is over 1 hour in duration, classified as extended idle. The height of the bars represents the
total idle fraction. For refuse trucks and single unit short haul trucks there is a significantly
higher idle fraction on the weekends, when using method 3. This implies that the refuse and
single unit truck vehicles that operate most on the weekend, also have higher idle fractions.
Method 3 appropriately weights the idle fractions from each vehicle according to its average
daily activity.
Method 1 "Average of Ratios" Method 3 "Normalized Sum over Sum"
Figure H-l. Idle fraction calculated using Method 1 and Method 3.
Figure H-2 graphically compares the start fractions and soak fractions calculated using Method
1 and Method 3, using the data for single unit short-haul trucks on weekdays. The start
distribution calculated with Method 1 weights all vehicles the same and thus overrepresents
the start times and soak times of vehicles which have few starts (and long soak periods). With
Method 3 the start and soak distribution more accurately characterize all the starts. The start
distribution with Method 3 is dominated by vehicles that have many starts per day. The starts
occur more evenly across the work-day and have shorter soak periods. Because emission rates
216
-------
increase with longer soak periods, the differences in averaging methods can have significant
impacts on the total emissions, as well as the temporal allocation of the emissions.
Single-Unit Short-Haul | Weekday
Single-Unit Short-Haul | Weekday
6 9 12 15 18 21 24
Hour of the Day
6 9 12 15 18 21 24
Hour of the day
Method 1 "Average of Ratios" Method 3 "Normalized Sum over Sum"
Figure H-2 Start fraction and Soak fraction calculated using Method 1 and Method 3.
H3. Future Work
The previous methods are all based on the assumption that the sample of vehicles are
representative of the entire vehicle population. However, as presented in Section 12,12.2.2,
there is significant variation in idle fractions and starts per day by truck vocation within the
MOVES sourcetypes. For example, parcel delivery trucks and concrete mixers are are both
single unit short haul trucks, but parcel delivery trucks have many more starts per day in the
Fleet DNA database.
The truck samples we are currently using (FleetDNA) and which we intend to use in the future
(CE-CERT), made efforts to collect data from a variety of important vocational classes. However,
the truck samples in these programs were not systematically chosen to be representative of
U.S. truck vocations. To address this deficiency, we would like to use a method that weights
each vehicle according to its average activity as well as the population of each vocation. The
proposed Method 4 "Vocation and Activity Weighted fraction" would use a weighting factor to
weight the vehicles within each vocation according to how many vehicles were sampled,
compared to how many exist in the national population.
217
-------
Method 4 -
"Vocation
and
Activity
Weighted
fraction"
/' = individual vehicle ID
days/ = # of days vehicle/is instrumented
v = vehicle vocation
wv= (population/sample size of each vocation, v)
s = sourceTypelD
d = dayTypelD
Equation H-10
In practice, we are not yet able to implement Method 4, because we are unable to accurately
map instrumented truck vocations to national truck populations because we lack information
on both parts of the equation.
1) We lack information on the total number of vehicles in each vocation. The IHS vehicle
registration data provides sufficient information to classify trucks by the MOVES
sourcetype, but not by vocation or specific firm. Some are characterized by the
industry sector of the firm that owns the truck, but, with large populations of trucks
classified in sectors such as: individual, general freight, government/miscellaneous,
lease/rental, wholesale/retail, manufacturing and services, these sector distinctions are
insufficient to determine the vocation of the truck. For example, should a "service truck"
be classified as a utility truck or a single unit box delivery truck?
2) The trucks in the Fleet DNA database only represent a subset of truck vocations
classified by these industrial sectors. For example, we do not have instrumented truck
data from many of the industry sectors in the registration data including:
agriculture/farm, petroleum, landscaping, mining, logging and emergency vehicles,
Additional work is needed to have confidence that the additional data needs and complexity of
Method 4 would yield meaningful improvements in emissions accuracy.
218
-------
Appendix I Road Load Coeffiecient for Combination Trucks in HD GHG Rule
Certification test procedures were developed to evaluate the aerodynamic performance of
tractors and trailers in HD GHG Phase 1 and 2. The test procedures varied between Phase 1 and
Phase 2 of the standards. Trailers were not included in the Phase 1 program and tractor
aerodynamic performance was measured at no wind conditions. While trailers were ultimately
not regulated in Phase 2, new test procedures were developed for trailers that approximate a
wind-averaged drag performance. Wind-averaged drag reflects a vehicle's average
performance for a range of yaw angles (the angle of attack of the air during travel) at a given
vehicle speed and wind speed and is more representative of real-world performance. The wind-
averaged drag result modeled in the Phase 2 rule is determined by an average of drag values
two yaw positions which represents a vehicle speed of 65 mph and a wind speed of 7 mph. In
the tractor analysis, the drag value is represented by the aerodynamic drag area, CdA In the
trailer analysis, the drag value is represented as a reduction in drag area, hC^A, relative to a
commonly available baseline trailer that is not equipped with aerodynamic devices.
The GHG rules also create bins for aerodynamic certification, so that a precise drag value is not
needed to certify every tractor. A representative aerodynamic value from each bin is used,
along with other aspects of the powertrain and vehicle, as an input into the Greenhouse Gas
Emissions Model (GEM) to determine a vehicle configuration's CO2 emissions result. Tractors
are categorized in the rule by their roof height and cab type - sleeper cabs and day cabs - and
different aerodynamic bins exist for each category and a mid-point from each bin is used as the
GEM input. The trailer analysis used the bottom boundaries of the bins for GEM input values,
which represent a conservative estimate of aerodynamic improvements. For this analysis,
midpoints of the bins were used to reflect average performance within the trailer bins. Bin I
represents no improvement, so a hC^A value of 0 m2 was used in this analysis. Non-box trailers
including flatbed and tank trailers, have standards based on tire technologies in the HD Phase 2
GHG program and aerodynamic improvements for those trailer types are neither expected nor
included in this analysis. The QA bin structures for tractors and trailers are shown respectively
below in Table 1-1 and Table I-2123 124 The trailer bin structure is common to all box van trailer
types.
Table 1-1 Phase 2 GHG Aerodynamic Drag Area Bin Structure for Tractors [m2]
Low-roof Sleeper &
Mid-roof Sleeper &
High-roof Sleeper Cab
High-roof Day Cab
Day Cabs
Day Cabs
Tractor
CdA Bin
CdA range
CdA input
CdA range
CdA input
CdA range
CdA input
CdA range
CdA input
1
>6.9
7.15
>7.2
7.45
>5.4
6.00
>5.9
7.00
II
6.3-6.8
6.55
6.6.7.1
6.85
4.9-5.3
5.60
5.5-5.8
6.65
III
5.7-6.2
5.95
6.0-6.5
6.25
4.5-4.8
5.15
5.1-5.4
6.25
IV
5.2-5.6
5.40
5.5-5.9
5.70
4.1-4.4
4.75
4.7-5.0
5.85
V
4.7-5.1
4.90
5.0-5.4
5.20
3.8-4.0
4.40
4.4-4.6
5.50
VI
4.2-4.6
4.40
4.5-4.9
4.70
3.5-3.7
4.10
4.1-4.3
5.20
VII
<4.1
3.90
<4.4
4.20
<3.4
3.80
<4.0
4.90
219
-------
Tab
e 1-2 Phase 2 GHG Aerodynamic Drag Area Bin Structure for Box Van Trailers [m2]
Trailer ACdA Bin
ACdA range
ACdA input for GEM
Midpoint of ACdA range
1
<0.09
0.0
0
II
0.10-0.39
0.1
0.25
III
0.40-0.69
0.4
0.55
IV
0.70-0.99
0.7
0.85
V
1.00-1.39
1.0
1.2
VI
1.40-1.79
1.4
1.6
VII
>1.80
1.8
1.9
The tractor and trailer bin structures were used to estimate adoption rates of improved
aerodynamic technologies. For tractors, EPA conducted such analyses for Phase 1 GHG and
Phase 2 GHG rulemakings, for both their respective baselines and the rulemaking scenarios. For
tractor certification in the GHG rules, different tractor types are assumed to be matched with
specific trailer types. High-roof tractors are matched with 53-foot box van trailers. Mid-roof
tractors are matched with tank trailers and low-roof tractors are matched with flatbed trailers.
The Phase 1 GHG baseline analysis was used for model years prior to implementation of the
Phase 1 GHG rule (pre-2014 model years). The Phase 2 GHG baseline analysis was used for
model years 2014 through 2020, which are predominantly the Phase 1 GHG implementation
years. The Phase 2 GHG technology penetration analysis was the basis for the adoption rates
for model years 2021 and later, with different rates for different types of cabs and each of the
major steps established in the rulemaking - model years 2021-2023, 2024-2026 and 2027 and
beyond. The bin-weighted average CdA (i.e., the "CdA input" from Table J-l) was then calculated
by model year group. For the high-roof sleeper cab and high-roof day cab subcategories, the
effect of the trailer skirt was removed to calculate the CdA of a tractor-trailer combination with
a baseline trailer. Through extensive testing in the Phase 2 GHG rulemaking development, the
trailer skirt was estimated to have Trailer Bin III performance of 0.55 m2, as seen in Table 1-3.
Table 1-3 Tractor aerodynamic technology adoption rates by model year groups
Tractor
Bin
Tractor Bin CdA
input [m2]
1950-2013
Phase 1 GHG
2014-2020
Phase 2 GHG
2021-2023
Phase 2 GHG
2024-2026
Phase 2 GHG
2027+
1
7.15
25%
0%
0%
0%
0%
II
6.55
70%
10%
0%
0%
0%
-Q
(0
III
5.95
5%
70%
60%
40%
20%
s_
0)
IV
5.40
0%
20%
30%
40%
30%
Q.
VI
4.40
0%
0%
0%
0%
0%
O
O
VII
3.90
0%
0%
0%
0%
0%
¦
-C
M
X
Mean CdA (w/ skirt) [m2]
6.67
5.9
5.68
5.52
5.26
Skirt effect [m2]
0.55
0.55
0.55
0.55
0.55
Mean CdA (w/o skirt) [m2]
7.22
6.45
6.23
6.07
5.81
220
-------
Tractor
Bin
Tractor Bin CdA
input [m2]
1950-2013
Phase 1 GHG
2014-2020
Phase 2 GHG
2021-2023
Phase 2 GHG
2024-2026
Phase 2 GHG
2027+
1
7.45
25%
0%
0%
0%
0%
II
6.85
70%
30%
0%
0%
0%
V)
-Q
III
6.25
5%
60%
60%
40%
30%
(D
U
>
IV
5.70
0%
10%
35%
40%
30%
(0
TS
V
5.20
0%
0%
5%
20%
40%
O
O
VI
4.70
0%
0%
0%
0%
0%
i
-C
txo
VII
4.20
0%
0%
0%
0%
0%
X
Mean CdA (w/ skirt) [m2]
6.97
6.375
6.005
5.82
5.665
Skirt effect [m2]
0.55
0.55
0.55
0.55
0.55
Mean CdA (w/o skirt) [m2]
7.52
6.925
6.555
6.37
6.215
1
7.00
100%
15%
10%
0%
0%
-Q
(D
II
6.65
0%
15%
10%
20%
20%
s_
0)
III
6.25
0%
70%
70%
60%
50%
Q.
ai
IV
5.85
0%
0%
10%
20%
30%
to
V
5.50
0%
0%
0%
0%
0%
o
o
VI
5.20
0%
0%
0%
0%
0%
1
¦O
VII
4.90
0%
0%
0%
0%
0%
§
Mean CdA [m2]
7.00
6.4225
6.325
6.25
6.21
1
7.00
100%
20%
10%
0%
0%
(/)
_D
II
6.65
0%
20%
10%
20%
20%
(B
U
III
6.25
0%
60%
70%
60%
50%
>
OJ
~o
IV
5.85
0%
0%
10%
20%
30%
M-
o
V
5.50
0%
0%
0%
0%
0%
0
1_
1
VI
5.20
0%
0%
0%
0%
0%
VII
4.90
0%
0%
0%
0%
0%
Mean CdA [m2]
7.00
6.48
6.325
6.25
6.21
1
6.00
100%
15%
10%
0%
0%
-Q
OJ
II
5.60
0%
15%
10%
20%
20%
(D
III
5.15
0%
70%
70%
60%
50%
a.
ai
Q)
V)
IV
4.75
0%
0%
10%
20%
30%
V
4.40
0%
0%
0%
0%
0%
o
o
VI
4.10
0%
0%
0%
0%
0%
1
s
VII
3.80
0%
0%
0%
0%
0%
iJ
Mean CdA [m2]
6.00
5.345
5.24
5.16
5.12
1
6.00
100%
20%
10%
0%
0%
(/)
II
5.60
0%
20%
10%
20%
20%
aj
u
III
5.15
0%
60%
70%
60%
50%
>
IV
4.75
0%
0%
10%
20%
30%
M-
o
V
4.40
0%
0%
0%
0%
0%
0
1_
1
5
o
VI
4.10
0%
0%
0%
0%
0%
VII
3.80
0%
0%
0%
0%
0%
Mean CdA [m2]
6.00
5.41
5.24
5.16
5.12
221
-------
A survey conducted by the North American Council for Freight Efficiency (NACFE) was used to
estimate that trailer aerodynamic technologies were not in significant use prior to 2008.125
Therefore, the model years between 1950-2007 reflect the time period prior to the use of
trailer aerodynamic improvements. The model year groups of 2008-2014 and 2014-2017
reflect voluntary improvements to trailer aerodynamics. Since trailers were not regulated in the
Phase 2 HDGHG rulemaking and lacking data on further voluntary improvements, trailer
aerodynamics for model years 2018 and beyond are modeled as MY2017.
Table 1-4 shows the trailer technology adoption rates were used to determine the average ACd>4
by model year group for for several trailer categories. Long box vans represent 53-ft box van
trailers. Short box vans are 50 feet and shorter, and the shortest ones are often pulled in
tandem. However, for simplicity and consistency with the compliance framework of the HD
GHG Phase 2 rule, a single-trailer configuration is the basis for this analysis for both long and
short trailers.
222
-------
Table 1-4 Trailer aerodynamic technology adoption rates by model year groups
Trailer Bin
1950-
2008-
2014-
2018-2020
Phase 2
Phase 2 GHG
Phase 2
2007
2013
2017
GHG
2021-2023
2024-2026
GHG
2027+
1
100%
65%
55%
55%
55%
55%
55%
II
0%
0%
0%
0%
0%
0%
0%
l/>
£
III
0%
35%
40%
40%
40%
40%
40%
(0
>
X
o
-Q
IV
0%
0%
5%
5%
5%
5%
5%
V
0%
0%
0%
0%
0%
0%
0%
two
c
o
VI
0%
0%
0%
0%
0%
0%
0%
_J
VII
0%
0%
0%
0%
0%
0%
0%
Average ACdA
[m2]
0
0.1925
0.2625
0.2625
0.2625
0.2625
0.2625
1
100%
100%
100%
100%
100%
100%
100%
II
0%
0%
0%
0%
0%
0%
0%
(/)
c
ro
>
X
III
0%
0%
0%
0%
0%
0%
0%
IV
0%
0%
0%
0%
0%
0%
0%
o
.Q
E
o
_c
V
0%
0%
0%
0%
0%
0%
0%
VI
0%
0%
0%
0%
0%
0%
0%
to
VII
0%
0%
0%
0%
0%
0%
0%
Average ACdA
[m2]
0
0
0
0
0
0
0
V)
1
100%
100%
100%
100%
100%
100%
100%
C
OJ
>
II
0%
0%
0%
0%
0%
0%
0%
X
o
-Q
two
c
III
0%
0%
0%
0%
0%
0%
0%
IV
0%
0%
0%
0%
0%
0%
0%
o
o
V
0%
0%
0%
0%
0%
0%
0%
ai
aj
VI
0%
0%
0%
0%
0%
0%
0%
.5
VII
0%
0%
0%
0%
0%
0%
0%
ro
Q_
Average ACdA
[m2]
0
0
0
0
0
0
0
V)
1
100%
100%
100%
100%
100%
100%
100%
(B
>
II
0%
0%
0%
0%
0%
0%
0%
X
o
-Q
III
0%
0%
0%
0%
0%
0%
0%
O
IV
0%
0%
0%
0%
0%
0%
0%
-C
V)
E
ai
ro
¦
aj
V
0%
0%
0%
0%
0%
0%
0%
VI
0%
0%
0%
0%
0%
0%
0%
VII
0%
0%
0%
0%
0%
0%
0%
t
ro
Q.
Average ACdA
[m2]
0
0
0
0
0
0
0
223
-------
The average ACdA values by model year group for tractor-trailer combinations were determined
by estimating the distribution of each trailer category within each tractor subcategory.
Following the analysis performed for the HD GHG Phase 2 rulemaking, the distribution in Table
1-5 was used. Trailers in the non-aero category are incompatible with aerodynamic
improvements and are assumed to be matched entirely within the low-roof and mid-roof
tractor types and no aerodynamic improvements are applied to these trailers. Trailers with
work-performing equipment that impedes the use of some aerodynamic devices are considered
partial-aero trailers. These trailers are assumed to be used in short haul operations and
assigned to high roof day cab tractors. The remaining trailers are full-aero box vans capable of
adopting a range of aerodynamic devices and we assume these trailer types are used in long
haul with sleeper cab tractors. Using a combination of data from the 2002 VIUS database and
trailer production results from ACT Research, over 70 percent of the full-aero capable trailers
are assumed to be long box vans (longer than 50-feet). Partial-aero box vans used in short-haul
applications, however, are more than 60 percent short trailer (50 feet and shorter).
Table 1-5 Trailer category distribution by tractor category
Trailer Category
Sleeper Cabs
Day Cabs
Low-roof
Mid-roof
High-roof
Low-roof
High-roof
Full-aero long
0%
0%
73%
0%
0%
Full-aero short
0%
0%
27%
0%
0%
Partial-aero long
0%
0%
0%
0%
36%
Partial-aero short
0%
0%
0%
0%
64%
Non-aero
100%
100%
0%
100%
0%
We assume no aerodynamic improvements for trailers pulled by low- and mid-roof tractors, so
all aerodynamic improvements for these vehicles come from the tractors only. Aerodynamic
improvements for the high-roof tractors pulling box trailers are calculated by combining the
aerodynamic drag estimates from the tractor and trailer. The average trailer ACdA values by
model year group and tractor category are listed in Table 1-6. Trailer aerodynamic
improvements are calculated using the trailer distribution shown in Table 1-5 and the adoption
rates of Table 1-4. The average CdA for a tractor-trailer combination by model year can be
calculated by subtracting the average trailer ACdA values from the average tractor CdA values in
Table 1-3.
Table 1-6 Average trailer ACdA values by tractor cate
years
Category ¦—^
Pre-2008
2008-2013
2014-2017
2018-2020
2021-2023
2024-2026
2027+
High-roof sleeper cab
0
0.140
0.192
0.192
0.192
0.192
0.192
High-roof day cab
0
0
0
0
0
0
0
gory and model year group [m2]
The resulting drag values that include aerodynamic improvements from tractors and trailers are
shown below.
224
-------
Table 1-7 Drag area, CdA [m2], by tractor-trailer subcategory and model year group
^^~^-^^_IVIodel years
Pre-2008
2008-2013
2014-2017
2018-2020
2021-2023
2024-2026
2027+
Category
High-roof sleeper cab
7.2200
7.0798
6.2584
6.2584
6.0384
5.8784
5.6184
High-roof day cab
7.5200
7.5200
6.9250
6.9250
6.5550
6.3700
6.2150
Mid-roof sleeper cab
7.0000
7.0000
6.4225
6.4225
6.3250
6.2500
6.2100
Mid-roof day cab
7.0000
7.0000
6.4800
6.4800
6.3250
6.2500
6.2100
Low-roof sleeper cab
6.0000
6.0000
5.3450
5.3450
5.2400
5.1600
5.1200
Low-roof day cab
6.0000
6.0000
5.4100
5.4100
5.2400
5.1600
5.1200
Vocational tractor
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
In MOVES, the values for sleeper cab tractors (with trailers) used for long-haul combination
trucks (sourceTypelD 62) and the values for day cab tractors (with trailers) are used for short-
haul combination trucks (sourceTypelD 61). Both the sleeper cab and day cab categories
contain a mix of high-roof, mid-roof and low-roof types. Day cab tractors also contain a
vocational tractor subcategory, for which the aerodynamic requirements of the Phase 2 rule do
not apply. They are of a low-roof height configuration and assumed to have the aerodynamic
characteristics of pre-2008 MY low-roof tractors for all model years. The combined average CdA
for the MOVES combination trucks shown in Table 1-9 was calculated using the distribution from
Table 1-8 and the drag areas from Table 1-7.
Table 1-8 Roof height distribution wit
hin cab types
Roof height
Sleeper Cab
Day Cab
Low-roof
5%
47%
Mid-roof
15%
0%
High-roof
80%
45%
Vocational
0%
8%
Table 1-9 Average CdA for each source type by model year group weighted by roof height
Pre-2008
2008-2013
2014-2017
2018-2020
2021-2023
2024-2026
2027+
Sleeper cab (sourceType 62)
7.1260
7.0136
6.2373
6.2373
6.0415
5.8982
5.6822
Day cab (sourceType 61)
6.6840
6.6840
6.1390
6.1390
5.8926
5.7717
5.6832
To convert from CdA to the C coefficient, Equation 15-11 was used with an estimate for air
density. A national annual MOVES run produced an average temperature of 61°F. At standard
atmospheric air pressure, the air density is 1.22 kg/m3. The resulting C coefficient values are
shown in Table 1-10.
Table 1-10 C coefficients [kW-s3/m3] of source types 61 and 62
2014-2017
2018-2020
2021-2023
2024-
2026
2027+
Sleeper cab (sourceType 62)
0.00380
0.00380
0.00369
0.00360
0.00347
Day cab (sourceType 61)
0.00374
0.00374
0.00359
0.00352
0.00347
by model year group
The Phase 1 and Phase 2 GHG emission standards also project improvements to the tire rolling
resistance. MOVES3 reflected these improvements through revisions to the A coefficient in the
225
-------
SourceUseTypePhysics table. It is related to the coefficient of rolling resistance, Crr and source
mass M, using the following equation where g is the gravitational acceleration:
A = CRRMg Equation 1-1
For combination tractor-trailers, the tires typically differ by axle position (steer, drive and
trailer). The HD GHG Phase 1 and Phase 2 rulemakings developed adoption rates of lower
rolling resistance tires for the steer and drive tires for all model years.126 127 The overall rolling
resistance of the vehicle is a weighted average of rolling resistance over axle based on axle
loading.
„ _ „ Msteer „ ^drive „ Mtrailer Equation 1-2
^RR — ^RR,steer ¦" ^RR,drive ¦" ^RR,trailer
Tire rolling resistance for tractor-trailers was updated using the same tractor type distributions
described in Table 1-8. Rolling resistance distributions, based on tire rolling resistance levels
from the GHG rules are shown in Table 1-11.
Table 1-11 Crr by axle and tractor type
Tire Crr
Tire Crr value
Pre-2014
Phase 1
2018-2020
Phase 2 GHG
Phase 2 GHG
Phase 2 GHG
level
[kg/metric
ton]
GHG
2014-2017
2021-2023
2024-2026
2027+
Base
7.8
100%
10%
10%
5%
5%
5%
(D
&—
1
6.6
0%
70%
70%
35%
15%
10%
(D
>
2
6.0
0%
20%
20%
50%
60%
50%
O
o
i—
Q
3
5.0
0%
0%
0%
10%
20%
35%
1
-C
Avg Crr [kg/metric ton]
8.1
6.84
6.84
6.32
6.04
5.845
.bp
if
"i_ ¦
1
6.9
0%
60%
60%
35%
25%
10%
0
2
6.0
0%
10%
10%
50%
65%
85%
226
-------
3
5.0
0%
0%
0%
0%
0%
0%
Avg Crr [kg/metric ton]
8.1
7.17
7.17
6.63
6.435
6.195
-Q
OJ
O
Avg Crr [kg/metric ton]
7.8
6.87
6.87
6.04
5.78
5.615
Base
8.1
100%
30%
30%
5%
5%
5%
>
(D
(D
1
6.9
0%
60%
60%
35%
15%
10%
M-
(D
>
2
6.0
0%
10%
10%
50%
60%
50%
o
i—
Q
3
5.0
0%
0%
0%
10%
20%
35%
-C
txo
Avg Crr [kg/metric ton]
8.1
7.17
7.17
6.32
6.04
5.845
X
00
3
4.9
0%
0%
0%
10%
15%
25%
l/>
Avg Crr [kg/metric ton]
7.8
6.99
6.99
6.04
5.91
5.785
OJ
u
Base
8.1
100%
40%
40%
15%
10%
5%
>
(D
(D
i_
1
6.9
0%
50%
50%
35%
25%
10%
M-
(D
>
2
6.0
0%
10%
10%
50%
65%
85%
0
1
§
o
Q
3
5.0
0%
0%
0%
0%
0%
0%
Avg Crr [kg/metric ton]
8.1
7.29
7.29
6.63
6.435
6.195
—1
(D
1
6.5
100%
100%
100%
100%
100%
100%
2
6.0
0%
0%
0%
0%
0%
0%
_0J
3
5.1
0%
0%
0%
0%
0%
0%
ro
1—
4
4.7
0%
0%
0%
0%
0%
0%
Avg Crr [kg/metric ton]
6.5
6.5
6.5
6.5
6.5
6.5
Base
7.8
100%
40%
40%
15%
10%
5%
o
QJ
1
6.6
0%
50%
50%
35%
20%
10%
u
(D
on
3
4.9
0%
0%
0%
0%
20%
35%
C
o
+-»
ro
Avg Crr [kg/metric ton]
7.8
6.99
6.99
6.33
5.93
5.615
(D
Base
8.1
100%
40%
40%
15%
10%
5%
o
>
>
"i_ ¦
1
6.9
0%
50%
50%
35%
20%
10%
2
6.0
0%
10%
10%
50%
50%
55%
227
-------
3
5.0
0%
0%
0%
0%
20%
30%
Avg Crr [kg/metric ton]
8.1
7.29
7.29
6.63
6.19
5.895
(D
»
1
6.5
100%
100%
100%
100%
100%
100%
2
6.0
0%
0%
0%
0%
0%
0%
i—
-------
ss
:85~
'88
186
179
5
179
179
5
179
5
5
179
5
5
5
5
5
7.1
7
7
7
7
7
7.1
10
10
10
10
10
179
5
5
5
7.1
7
7
7
7.1
10
10
10
7.1
10
10
10
179
5
5
179
MOVES5 SourceUseTypePhysics Table
Table J-l MOVES5 SourceUseTypePhysics Table
Begin
Model
Year
End
Model
Year
Rolling
Term A
(kW-s/m)
Rotating
Term B
(kW-s2/m2)
Drag
Term C
(kW-s3/m3)
Source Mass
(metric tons)
1950
2060
0.0251
0
0.000315
1950
2060
0.156461
0.0020019
0.0004926
1950
2060
0.22112
0.0028376
0.0006983
1950
2009
0.22112
0.0028376
0.0006983
2010
2060
0.22112
0.0028376
0.0006983
1950
2060
0.235008
0.0030386
0.0007478
1950
2009
0.235008
0.0030386
0.0007478
2010
2060
0.235008
0.0030386
0.0007478
1950
2009
1.29515
0
0.0037149
2010
2013
1.29515
0.0037149
2014
2060
1.23039
0.0037149
1950
2009
1.29515
0.0037149
2010
2013
1.29515
0.0037149
2014
2020
1.23039
0.0037149
2021
2023
1.00646
0.0037149
2024
2026
0.974469
0.0037149
2027
2060
0.926484
0.0037149
1950
2009
1.29515
0.0037149
2010
2013
1.29515
0.0037149
2014
2020
1.23039
0.0037149
2021
2023
1.00646
0.0037149
2024
2026
0.974469
0.0037149
2027
2060
0.926484
0.0037149
1950
2009
1.29515
0.0037149
2010
2013
1.29515
0.0037149
2014
2020
1.23039
0.0037149
2021
2023
1.00646
0.0037149
2024
2026
0.974469
0.0037149
2027
2060
0.926484
0.0037149
1950
2009
1.0944
0.003587
2010
2013
1.0944
0.003587
2014
2026
1.03968
0.003587
2027
2060
0.913879
0.003587
1950
2009
1.0944
0.003587
2010
2013
1.0944
0.003587
2014
2026
1.03968
0.003587
2027
2060
0.913879
0.003587
1950
2009
1.0944
0.003587
2010
2013
1.0944
0.003587
2014
2026
1.03968
0.003587
2027
2060
0.913879
0.003587
1950
2009
1.0944
0.003587
2010
2013
1.0944
0.003587
2014
2026
1.03968
0.003587
2027
2060
0.913879
0.003587
1950
2009
0.746718
0.0021758
2010
2013
0.746718
0.0021758
2014
2060
0.709382
0.0021758
1950
2009
0.746718
0.0021758
229
-------
Source
Type ID
Reg
Class ID
Begin
Model
Year
End
Model
Year
Rolling
Term A
(kW-s/m)
Rotating
Term B
(kW-s2/m2)
Drag
Term C
(kW-s3/m3)
Source Mass
(metric tons)
Fixed Mass
Factor
(metric
tons)
43
42
2010
2013
0.746718
0
0.0021758
7.78184
5
43
42
2014
2020
0.709382
0
0.0021758
7.78184
5
43
42
2021
2023
0.637734
0
0.0021758
7.78184
5
43
42
2024
2026
0.603684
0
0.0021758
7.78184
5
43
42
2027
2060
0.569634
0
0.0021758
7.78184
5
43
46
1950
2009
0.746718
0
0.0021758
11.3666
17.1
43
46
2010
2013
0.746718
0
0.0021758
11.3666
7
43
46
2014
2020
0.709382
0
0.0021758
11.3666
7
43
46
2021
2023
0.637734
0
0.0021758
11.3666
7
43
46
2024
2026
0.603684
0
0.0021758
11.3666
7
43
46
2027
2060
0.569634
0
0.0021758
11.3666
7
43
47
1950
2009
0.746718
0
0.0021758
15.6028
17.1
43
47
2010
2013
0.746718
0
0.0021758
15.6028
10
43
47
2014
2020
0.709382
0
0.0021758
15.6028
10
43
47
2021
2023
0.637734
0
0.0021758
15.6028
10
43
47
2024
2026
0.603684
0
0.0021758
15.6028
10
43
47
2027
2060
0.569634
0
0.0021758
15.6028
10
51
41
1950
2009
1.58346
0
0.0035723
3.57431
2.05979
51
41
2010
2013
1.58346
0
0.0035723
3.57431
5
51
41
2014
2060
1.50429
0
0.0035723
3.57431
5
51
42
1950
2009
1.58346
0
0.0035723
5.76818
2.05979
51
42
2010
2013
1.58346
0
0.0035723
5.76818
5
51
42
2014
2026
1.50429
0
0.0035723
5.76818
5
51
42
2027
2060
1.32227
0
0.0035723
5.76818
5
51
46
1950
2009
1.58346
0
0.0035723
13.8001
17.1
51
46
2010
2013
1.58346
0
0.0035723
13.8001
7
51
46
2014
2026
1.50429
0
0.0035723
13.8001
7
51
46
2027
2060
1.32227
0
0.0035723
13.8001
7
51
47
1950
2009
1.58346
0
0.0035723
20.7044
17.1
51
47
2010
2013
1.58346
0
0.0035723
20.7044
10
51
47
2014
2026
1.50429
0
0.0035723
20.7044
10
51
47
2027
2060
1.32227
0
0.0035723
20.7044
10
52
41
1950
2009
0.627922
0
0.0016030
3.57431
2.05979
52
41
2010
2013
0.627922
0
0.0016030
3.57431
5
52
41
2014
2060
0.596526
0
0.0016030
3.57431
5
52
42
1950
2009
0.627922
0
0.0016030
5.76818
2.05979
52
42
2010
2013
0.627922
0
0.0016030
5.76818
5
52
42
2014
2020
0.596526
0
0.0016030
5.76818
5
52
42
2021
2023
0.558348
0
0.0016030
5.76619
5
52
42
2024
2026
0.558348
0
0.0016030
5.76344
5
52
42
2027
2060
0.53568
0
0.0016030
5.76069
5
52
46
1950
2009
0.627922
0
0.0016030
13.8001
17.1
52
46
2010
2013
0.627922
0
0.0016030
13.8001
7
52
46
2014
2020
0.596526
0
0.0016030
13.8001
7
52
46
2021
2023
0.558348
0
0.0016030
13.7981
7
52
46
2024
2026
0.558348
0
0.0016030
13.7953
7
52
46
2027
2060
0.53568
0
0.0016030
13.7926
7
52
47
1950
2009
0.627922
0
0.0016030
25.0484
17.1
52
47
2010
2013
0.627922
0
0.0016030
25.0484
10
52
47
2014
2020
0.596526
0
0.0016030
25.0484
10
52
47
2021
2023
0.558348
0
0.0016030
25.0464
10
52
47
2024
2026
0.558348
0
0.0016030
25.0437
10
230
-------
Source
Type ID
Reg
Class ID
Begin
Model
Year
End
Model
Year
Rolling
Term A
(kW-s/m)
Rotating
Term B
(kW-s2/m2)
Drag
Term C
(kW-s3/m3)
Source Mass
(metric tons)
Fixed Mass
Factor
(metric
tons)
52
47
2027
2060
0.53568
0
0.0016030
25.0409
10
53
41
1950
2009
0.557262
0
0.0014738
3.57431
2.05979
53
41
2010
2013
0.557262
0
0.0014738
3.57431
5
53
41
2014
2060
0.529399
0
0.0014738
3.57431
5
53
42
1950
2009
0.557262
0
0.0014738
5.76818
2.05979
53
42
2010
2013
0.557262
0
0.0014738
5.76818
5
53
42
2014
2020
0.529399
0
0.0014738
5.76818
5
53
42
2021
2023
0.484929
0
0.0014738
5.76461
5
53
42
2024
2026
0.458989
0
0.0014738
5.75747
5
53
42
2027
2060
0.458989
0
0.0014738
5.75033
5
53
46
1950
2009
0.557262
0
0.0014738
13.8001
17.1
53
46
2010
2013
0.557262
0
0.0014738
13.8001
7
53
46
2014
2020
0.529399
0
0.0014738
13.8001
7
53
46
2021
2023
0.484929
0
0.0014738
13.7965
7
53
46
2024
2026
0.458989
0
0.0014738
13.7894
7
53
46
2027
2060
0.458989
0
0.0014738
13.7822
7
53
47
1950
2009
0.557262
0
0.0014738
25.0484
17.1
53
47
2010
2013
0.557262
0
0.0014738
25.0484
10
53
47
2014
2020
0.529399
0
0.0014738
25.0484
10
53
47
2021
2023
0.484929
0
0.0014738
25.0449
10
53
47
2024
2026
0.458989
0
0.0014738
25.0377
10
53
47
2027
2060
0.458989
0
0.0014738
25.0306
10
54
41
1950
2009
0.68987
0
0.0021055
3.57431
2.05979
54
41
2010
2013
0.68987
0
0.0021055
3.57431
5
54
41
2014
2060
0.655376
0
0.0021055
3.57431
5
54
42
1950
2009
0.68987
0
0.0021055
5.76818
2.05979
54
42
2010
2013
0.68987
0
0.0021055
5.76818
5
54
42
2014
2020
0.655376
0
0.0021055
5.76818
5
54
42
2021
2026
0.519058
0
0.0021055
5.76818
5
54
42
2027
2060
0.493498
0
0.0021055
5.76818
5
54
46
1950
2009
0.68987
0
0.0021055
13.8001
17.1
54
46
2010
2013
0.68987
0
0.0021055
13.8001
7
54
46
2014
2020
0.655376
0
0.0021055
13.8001
7
54
46
2021
2026
0.519058
0
0.0021055
13.8001
7
54
46
2027
2060
0.493498
0
0.0021055
13.8001
7
54
47
1950
2009
0.68987
0
0.0021055
25.0484
17.1
54
47
2010
2013
0.68987
0
0.0021055
25.0484
10
54
47
2014
2020
0.655376
0
0.0021055
25.0484
10
54
47
2021
2026
0.519058
0
0.0021055
25.0484
10
54
47
2027
2060
0.493498
0
0.0021055
25.0484
10
61
46
1950
2009
1.64062
0
0.0021055
14.0122
17.1
61
46
2010
2013
1.64062
0
0.0021055
14.0122
7
61
46
2014
2020
1.509
0
0.0037448
13.8666
7
61
46
2021
2023
1.407
0
0.0035945
13.8666
7
61
46
2024
2026
1.379
0
0.0035207
13.8666
7
61
46
2027
2060
1.353
0
0.0034667
13.8666
7
61
47
1950
2009
1.64062
0
0.0040772
24.8298
17.1
61
47
2010
2013
1.64062
0
0.0040772
24.8298
10
61
47
2014
2020
1.509
0
0.0037448
24.6842
10
61
47
2021
2023
1.407
0
0.0035945
24.6842
10
61
47
2024
2026
1.379
0
0.0035208
24.6842
10
61
47
2027
2060
1.353
0
0.0034667
24.6842
10
231
-------
Source
Type ID
Reg
Class ID
Begin
Model
Year
End
Model
Year
Rolling
Term A
(kW-s/m)
Rotating
Term B
(kW-s2/m2)
Drag
Term C
(kW-s3/m3)
Source Mass
(metric tons)
Fixed Mass
Factor
(metric
tons)
61
49
1950
2013
1.64062
0
0.0040772
24.8298
17.1
61
49
2014
2020
1.509
0
0.0037448
24.6842
17.1
61
49
2021
2023
1.407
0
0.0035945
24.6842
17.1
61
49
2024
2026
1.379
0
0.0035208
24.6842
17.1
61
49
2027
2060
1.353
0
0.0034667
24.6842
17.1
62
46
1950
2007
1.73882
0
0.0043469
14.0122
17.1
62
46
2008
2009
1.73882
0
0.0042785
14.0122
17.1
62
46
2010
2013
1.73882
0
0.0042785
14.0122
7
62
46
2014
2020
1.576
0
0.0038048
13.8308
7
62
46
2021
2023
1.502
0
0.0036853
13.8308
7
62
46
2024
2026
1.466
0
0.0035979
13.8308
7
62
46
2027
2060
1.44
0
0.0034661
13.8308
7
62
47
1950
2007
1.73882
0
0.0043469
24.8298
17.1
62
47
2008
2009
1.73882
0
0.0042785
24.8298
17.1
62
47
2010
2013
1.73882
0
0.0042785
24.8298
10
62
47
2014
2020
1.576
0
0.0038048
24.6484
10
62
47
2021
2023
1.502
0
0.0036853
24.6484
10
62
47
2024
2026
1.466
0
0.0035979
24.6484
10
62
47
2027
2060
1.44
0
0.0034661
24.6484
10
62
49
1950
2007
1.73882
0
0.0043469
24.8298
17.1
62
49
2008
2013
1.73882
0
0.0042785
24.8298
17.1
62
49
2014
2020
1.576
0
0.0038048
24.6484
17.1
62
49
2021
2023
1.502
0
0.0036853
24.6484
17.1
62
49
2024
2026
1.466
0
0.0035979
24.6484
17.1
62
49
2027
2060
1.44
0
0.0034661
24.6484
17.1
232
-------
Appendix K AVFT Tool
The AVFT Tool is a user tool available in the MOVES GUI that can be used to develop the AVFT
(Alternate Vehicle Fuel and Technology) table, which allows users to modify the fraction of
vehicles capable of using different fuels and technologies. The purpose of this tool is to project
future fuel type distributions based on the combination of local historic data and projected
national trends. The projections are applied to model years beyond the user-specified last
complete model year in the input data to the user-specified analysis year. The last complete
model year is needed, as partial model years are common in vehicle registration data. For
simplicity, the last complete model year will be referred to as the "base model year"
throughout this appendix.
The tool contains the following methods to project future fuel type distributions: proportional,
national average, known fractions, and constant.
The proportional, national average, and known fraction projection methods are dependent on
knowing the national default fuel type distributions. These are calculated from the default
SampleVehiclePopulation table by summing the stmyFraction values associated with each
source type, model year, fuel type, and engine technology combination (see Section 5.1 for
information on how the default SampleVehiclePopulation table was calculated). In effect, this
calculation is aggregating over the regClassID column in the SampleVehiclePopulation table to
calculate a nationally representative AVFT table. In keeping with the naming convention of the
AVFT columns, the resulting sum of the stmyFraction column will be referred to as the
fuelEngFraction throughout this appendix.
The following subsections describe each projection method. Note that these projection
methods are selected by source type, so different source types can use different methods. As
such, the algorithm descriptions below assume the algorithm is operating on a single source
type for clarity.
IK1. Pro portion a II
This method projects future fuel type distributions based on proportional differences between
the local and the national distributions in the base model year. The intention with this method
is to preserve differences between local conditions and the national average, while still
accounting for expected changes in national fuel type distribution trends.
To implement this method, the ratio between the input AVFT fuelEngFraction and the national
default fuelEngFraction is calculated for the base model year for each fuel type and engine
technology. This ratio is subject to the boundary limits of 0.5 and 2.0, which effectively limits
the proportional projection to stay between 50% lower and 200% higher than the national
default fraction for any given fuel type and engine technology combination. These limits were
chosen because there is considerable uncertainty as to the future geographic distribution of
EVs, and we did not want extreme differences between the national averages and local
conditions in the base model year to inappropriately bias the projected data. For example, if an
233
-------
area has no EVs in the base model year, a proportional projection without a lower boundary
limit would result in no EVs in the future. Given the projected future EV sales fractions
incorporated into MOVES (see Section 5.1), this does not seem reasonable. Conversely, if an
area is an early adopter of EVs in the base model year, a proportional projection without an
upper boundary limit could result in an unrealistic projection of 100% EVs. We chose boundary
limits of 50% and 200% (instead of looser limits, like 33% and 300% or 25% and 400%) because
the ratio is based on a single model year and we wanted to prevent the single model year from
introducing too much bias in the results.
After calculating the ratio and applying the boundary limits, the projected fuelEngFraction is
calculated by multiplying the base model year ratio with the national default fuelEngFraction
for all model years between the base model year and the analysis year. Since the base model
year ratio is calculated independently for each fuel type, the resulting projected fuelEngFraction
values are normalized so that the sum of the fuelEngFraction values by model year sum to 1.0.
K2. National Average
This method applies the national default fuel type distributions for all model years beyond the
base model year in the input data. The intention of this method is to allow users to provide
AVFT inputs for source types where local data do not reflect local vehicle activity (for example,
long haul source types), or where local data are not available for a particular source type.
To implement this method, the national default fuelEngFraction for each model year, fuel type,
and engine technology is used as-is for the projected fuelEngFraction.
K3. Known Fractions
This method allows the user to provide known fuel fractions for specific fuel types. The
intention with this method is to allow users to input values mandated by local programs (such
as a ZEV program), and the tool will handle fuel types not explicitly covered by the local
program. Essentially, this method applies the proportional method for all model years, fuel
types, and engine technologies not included in the user provided fractions.
To implement this method, the known fractions are used directly for the projected
fuelEngFraction values. Then, like the proportional method, the ratio between the input AVFT's
fuelEngFraction and the national default fuelEngFraction is calculated for the base model year
for each fuel type and engine technology. This ratio is subject to the same boundary limits of
50% and 200%.
For all model years, fuel types, and engine technologies not included in the input known
fractions, the projected fuelEngFraction is calculated by multiplying the base model year ratio
with the national default fuelEngFraction. Finally, the projected fuelEngFraction values are
normalized so that the sum of the fuelEngFraction by model year sums to 1.0, without changing
the input known fractions.
234
-------
K4. Cons
This method uses the fuelEngFraction for the base model year for each fuel type and engine
technology as-is for the projected fuelEngFraction values for all projected model years.
235
-------
18. References
1 Motor Vehicle Emission Simulator (MOVES) technical reports are available on the US
Environmental Protection Agency website:
https://www.epa.gov/moves/moves-technical-reports
2 US EPA. (2024). M0VES5 Technical Guidance: Using MOVES to Prepare Emission Inventories
for State Implementation Plans and Transportation Conformity. EPA-420-B-24-043. Ann Arbor,
Ml.
https://www.epa.gov/state-ancl-local-transportation/policv-ancl-technical-guiclance-state-ancl-
local-transportation#emission
3 US EPA. (2015). U.S. Environmental Protection Agency Peer Review Handbook. EPA-100-B-15-
001. Washington, D.C. 20460.
epa.gov/sites/default/files/2020-08/documents/epa peer review handbook 4th edition.pdf
4 US EPA. (2017). Population and Activity of Onroad Vehicles in MOVES201X - Draft Report.
Draft report and peer-review documents. Record ID 328870. EPA Science Inventory.
https://cfpub.epa.gov/si/si public record report.cfm?dirEntryld=328870
5 US EPA. (2019). Exhaust Emission Rates of Heavy-Duty Onroad Vehicles in MOVES_CTI_NPRM -
Draft Report. Draft report and peer-review documents. Record ID 347135. EPA Science
Inventory.
https://cfpub.epa.gov/si/si public record report.cfm?dirEntryld=347135
6 US FHWA. (1990-2023) Table VM-1: Annual Vehicle Distance Traveled in Miles and Related
Data. Highway Statistics. Washington, DC.
https://www.fhwa.dot.gov/policyinformation/statistics/2022/vml.cfm
7 US FHWA. (2011). Vehicle Type Codes and Descriptions. Highway Performance Monitoring
System Field Manual. Washington, DC.
http://www.fhwa.dot.gov/ohim/hpmsmanl/chapt3.cfm
8 Weatherby, M., Jackson, D., Koupal, J., and DenBleyker, A., Eastern Research Group, Inc.
(2017). Analysis of IHS Registration Data and Preparation of WA 5-08 Task 1 Deliverables,
Memorandum to David Brzezinski, EPA. EPA EP-C-12-017, Work Assignment 5-08. Austin, TX.
236
-------
9 US FTA. (2024). NTD Glossary. National Transit Database.
https://www.transit.dot.gov/ntd/national-transit-database-ntd-glossarv
10 US FHWA, Office of Planning, Environment, and Realty. (2012). Planning Glossary.
https://www.fhwa.dot.gov/planning/glossary/
11 Stanard, A., Fincher, S., Kishan, S., and Sabisch, M., Eastern Research Group, Inc. (2012). Data
Analyses on Drayage Heavy-Duty Vehicles. EPA EP-C-12-017, Work Assignment 0-2. Austin, TX.
12 US EPA. Heavy-Duty Highway Compression-Ignition Engines and Urban Buses—Exhaust
Emission Standards. Accessed August 2023.
https://www.epa.gov/emission-standards-reference-guide/epa-emission-standards-heavy-
duty-highway-engines-and-vehicles
13 US GPO. (2012). Definitions. Code of Federal Regulations, Title 40 - Protection of
Environment, Vol. 19, CFR 86.091-2.
14 US EPA. (2015). Frequently Asked Questions about Heavy-Duty "Glider Vehicles" and "Glider
Kits". EPA-420-F-15-904. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZvPDF.cgi?Dockev=P100MUVI.PDF
15 US EPA. (2024). Fuel Supply Defaults: Regional Fuels and the Fuel Wizard in M0VES5. EPA-
420-R-24-017. Office of Transportation and Air Quality. Ann Arbor, Ml.
https://www.epa.gov/moves/moves-onroad-technical-reports
16 US EPA. (2020). Fuel Effects on Exhaust Emissions from Onroad Vehicles in M0VES3. EPA-420-
R-20-016. Office of Transportation and Air Quality. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P1010M 6C.pdf
17 US EPA. (2024). Exhaust Emission Rates for Light-Duty Onroad Vehicles in M0VES5. EPA-420-
R-24-016. Office of Transportation and Air Quality. Ann Arbor, Ml.
https://www.epa.gov/moves/moves-onroad-technical-reports
18 US EPA (2024). Exhaust Emission Rates of Heavy-Duty Onroad Vehicles in M0VES5. EPA-420-
R-24-015. Office of Transportation and Air Quality. Ann Arbor, Ml.
https://www.epa.gov/moves/moves-onroad-technical-reports
237
-------
19 US FHWA. (2011). Annual Vehicle Miles Travelled and Related Data: Procedures Used to
Derive Data Elements Contained in Highway Statistics Table VM-1 for Years 2009 and after and
2007 and 2008 Historical Data. FHWA-PL-11-031. Washington, DC.
http://www.fhwa.dot.gov/ohim/vml methodology 2007.pdf
20 US Energy Information Administration. (2023). Annual Energy Outlook 2023. Washington, DC.
https://www.eia.gov/outlooks/aeo
21 US FHWA. (1990-2022). Table MV-1: State Motor-Vehicle Registrations. Highway Statistics.
Office of Highway Policy Information. Washington, DC.
https://www.fhwa.dot.gov/policyinformation/statistics/2022/mvl.cfm
22 S&P Global, formerly R.L. Polk & Co and IHS Markit. (1999, 2011, 2014, 2020, and 2023).
Vehicles in Operation (VIO) & Vehicle Registration Data Analysis.
https://www.spglobal.com/mobilitv/en/products/automotive-market-data-analysis.html
23 US Census Bureau. (2004). 2002 Vehicle Inventory and Use Survey. EC02TV-US.
www.census.gov/librarv/publications/2002/econ/census/vehicle-inventory-and-use-
survey.html
24 Bureau of Transportation Statistics and US Census Bureau. (2023). 2021 Vehicle Inventory and
Use Survey Datasets: 2021 Public Use File (PUF).
https://www.census.gov/data/datasets/2021/econ/vius/2021-vius-puf.html
25 US EPA. (2010). MOVES2010 Highway Vehicle Population and Activity Data. EPA-420-R-10-
026. Washington, DC.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P100ABRO.pdf
26 Bobit Publications. (1992, 2002, 2004-2024). School Bus Fleet Fact Book. Torrance, CA.
http://www.schoolbusfleet.com
27 US Federal Transit Administration. (2002-2022). National Transit Database.
https://www.transit.dot.gov/ntd
238
-------
28 Koupal, J., DeFries, T., Palacios, C., Fincher, S., and Preusse, D. (2014). Motor Vehicle
Emissions Simulator Input Data. Transportation Research Record: Journal of the Transportation
Research Board, 2427, 63-72.
https://doi.org/10.3141/2427-07
29 Yoon, S., Georgia Institute of Technology. (2005). A New Heavy-Duty Vehicle Visual
Classification and Activity Estimation Method for Regional Mobile Source Emissions Modeling
(student thesis). Atlanta, GA.
https://smartech.gatech.edu/bitstream/handle/1853/7245/seungju yoon 200508 phd.pdf
30 US EPA. (2023). 2022 EPA Automotive Trends Report.
www.epa.gov/automotive-trends/explore-automotive-trends-data
31 US EPA. (2019). 2019 Annual Production Volume Reports into Engine and Vehicle Compliance
Information System.
32 Brakora, J. (2018). Estimated Glider Vehicle Production for Use in EPA MOVES Models.
Memorandum to Docket EPA-HQ-OAR-2019-0055.
https://www.regulations.gov/docket/EPA-HQ-QAR-2019-0055
33 US EPA. (2024). Multi-Pollutant Emissions Standards for Model Years 2027 and Later Light-
Duty and Medium-Duty Vehicles. 89 Fed. Reg. 27842.
34 US EPA. (2024). Greenhouse Gas Emissions Standards for Heavy-Duty Vehicles—Phase 3. 89
Fed. Reg. 29440.
35 California Air Resources Board. 2021 Advanced Clean Trucks Regulation.
https://ww2.arb.ca.gov/rulemaking/2019/advancedcleantrucks
36 Sherwood, T. (2024). LMDV FRM OMEGA Effects, Pre-processing, Vehicle Rates - Memo.
Memorandum to Docket EPA-HQ-OAR-2022-0829.
https://www.regulations.gov/document/EPA-HQ-QAR-2022-0829-5553
37 California Air Resources Board. (2021). 13 C.C.R. Section 1963.
https://ww2.arb.ca.gov/sites/default/files/barcu/regact/2019/act2019/fro2.pdf
239
-------
38 US EPA. (2021). Final Advanced Clean Trucks Amendments. 310 CMR 7.40
https://www.epa.gov/sites/default/files/2017-10/documents/ma-310-cmr-7-40.pdf
39 New Jersey State Department of Environmental Protection. (2023). N.J.A.C. 7:27-31.
https://dep.ni.gov/wp-content/uploads/aqm/sub31.pdf
40 New York State Department of Environmental Conservation. 6 NYCRR Part 218 Section 218-4.
https://dec.ny.gov/sites/default/files/2024-01/218act.pdf
41 Oregon Department of Environmental Quality. OAR 340-257-0050.
https://secure.sos.state.or.us/oard/viewSingleRule.action?ruleVrsnRsn=296998
42 Washington State Legislature. Chapter 173-423-075 WAC.
https://apps.leg.wa.gov/wac/default.aspx?cite=173-423
43 State of Vermont Agency of Natural Resources. (2022). Chaper 40: Vermont Low Emission
Vehicle and Zero Emission Vehicle Rules.
https://dec.vermont.gov/sites/dec/files/aqc/laws-
regs/documents/Chapter 40 LEV ZEV rule adopted.pdf
44 Colorado Air Quality Control Commission Regulation 20, 5 CCR 1001-24. Part F.
https://drive.google.eom/file/d/lKyOLttXi9-JTGzrFHf59 RZGWIuvvxJa/view
45 Maryland Register Volume 50, Issue 25. December 15, 2023.
46 New Mexico Environmental Improvement Board. (2023). 0.2.91 NMAC.
https://www.srca.nm.gov/parts/title20/20.002.0Q91.html
47 Rhode Island Department of Environmental Management. Rhode Island's Low-Emission and
Zero-Emission Vehicle Programs. 250-RICR-120-05-37.
https://risos-apa-production-public.s3.amazonaws.com/DEM/REG 13147 20240701120209569.pdf
48 US EPA. (2024). Peer Review of Heavy-Duty Technology Resource Use Case Scnario (HD
TRUCS) Tool.
https://cfpub.epa.gov/si/si public record report.cfm?Lab=OTAQ&dirEntryld=360857
240
-------
49 US EPA. (2024). Greenhouse Gas Emissions Standards for Heavy-Duty Vehicles: Phase 3
Regulatory Impact Analysis. EPA-420-R-24-006.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P101A93R.pdf
50 US EPA. (2024). Docket Memo Heavy-Duty Technology Resource Use Case Scenario Tool (HD
TRUCS). Final Rule. EPA-HQ-OAR-2022-0985-3877.
https://www.regulations.gov/document/EPA-HQ-QAR-2022-0985-3877
51 US EPA. (2012). Use of Data from "Development of Emission Rates for the MOVES Model,"
Sierra Research, March 3, 2010. EPA-420-R-12-022. Ann Arbor, Ml.
http://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P100FlA5.pdf
52 National Highway Traffic Safety Administration. (2022). Final Rulemaking for Model Years
2024-206 Light-Duty Vehicle Corporate Average Fuel Economy Standards. Technical Support
Document.
53 Browning, L., Chan, M., Coleman, D., and Pera, C., ARCADIS Geraghty & Miller, Inc. (1998).
Update of Fleet Characterization Data for Use in M0BILE6: Final Report. EPA420-P-98-016.
Mountain View, CA.
http://nepis.epa.gov/Exe/ZyPDF.cgi? Dockey=P1001ZUK. PDF
54 National Household Travel Survey. 2017 NHTS Average Annual Vehicle Miles of Travel Per
Vehicle (Best Estimate).
https://nhts.ornl.gov
55 US EPA. The National Emission Inventory (NEI).
https://www.epa.gov/air-emissions-inventories/national-emissions-inventory-nei
56 US FHWA. Highway Statistics. Office of Highway Policy Information. Washington, DC.
https://www.fhwa.dot.gov/policyinformation/statistics.cfm
57 CRC Project A-100. Improvement of Default Inputs for MOVES and SMOKE-MOVES.
58 US EPA. (2015). Technical Support Document, Preparation of Emissions Inventories for the
Version 6.2, 2011 Emissions Modeling Platform. Office of Air Quality Planning and Standards.
241
-------
59 National Emissions Inventory Collaborative. (2019). 2016beta Emissions Modeling Platform.
http://views.cira.colostate.edu/wiki/wiki/10197
60 Sierra Research, Inc. (1997). Development of Speed Correction Cycles. M6.SPD.001, EPA-68-
C4-0056. Sacramento, CA.
61 Hart, C., Koupal, J., and Giannelli, R. (2002). EPA's Onboard Analysis Shootout: Overview and
Results. EPA420-R-02-026. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P10005PG.txt
62 DieselNet, Emission Test Cycles: New York Bus. Last accessed 8 June 2015.
https://www.dieselnet.com/standards/cycles/nybus.php
63 Melendez, M., Taylor, J., and Zuboy, J. (2005). Emission Testing of Washington Metropolitan
Area Transit Authority (WMATA) Natural Gas and Diesel Transit Buses. NREL/TP-540-36355.
http://www.afdc.energy.gov/pdfs/36355.pdf
64 Clark, N. and Gautam, M., West Virginia Research Corporation. (2005). Heavy-Duty Vehicle
Chassis Dynamometer Testing for Emissions Inventory, Air Quality Modeling, Source
Apportionment and Air Toxics Emissions Inventory. CRC Project E55/59. Morgantown, WV.
https://crcao.org/wp-content/uploads/2019/05/E-55 59 Final Report 23AUG2007.pdf
65 Sensors, Inc. (2002). On-Road Emissions Testing of 18 Tier 1 Passenger Cars and 17 Diesel
Powered Public Transport Buses. EPA-420-R-02-030. Saline, Ml.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P10005S8.txt
66 Eastern Research Group, Inc. (2003). Roadway-Specific Driving Schedules for Heavy-Duty
Vehicles. EPA420-R-03-018. Austin, TX.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P100LWCT.txt
67 US EPA. (2013). Greenhouse Gas Emissions Model (GEM) for Medium- and Heavy-Duty Vehicle
Compliance, Simulation Model v2.0.1.
https://www.epa.gov/regulations-emissions-vehicles-and-engines/greenhouse-gas-emissions-
model-gem-medium-and-heavy-duty
242
-------
68 US EPA. (2020). Population and Activity of Onroad Vehicles in MOVES3. EPA-420-R-21-012.
Office of Transportation and Air Quality. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZvPDF.cgi?Dockey=P1011TF8.pdf
69 Liu, H., Sonntag, D., Brzezinski, D., Fulper, C. R., Hawkins, D., and Warila, J. E. (2016).
Operations and Emissions Characteristics of Light-Duty Vehicles on Ramps. Transportation
Research Record: Journal of the Transportation Research Board. 2570, 1-11.
https://doi.org/10.3141/257Q-01
70 US EPA. (2001). Amendments to Vehicle Inspection Maintenace Program Requirements
Incorporating the Onboard Diagnositc Check. 66 FR 18156.
71 Walkowicz, K., Kelly, K., Duran, A. and Burton, E. (2014). Fleet DNA Project Data. National
Renewable Energy Laboratory.
https://www.nrel.gov/transportation/fleettest-fleet-dna.html
72 National Instruments. (2014). Controller Area Network (CAN) Overview.
http://www.ni.com/white-paper/2732/en
73 ATRI. (2015). Compendium of Idling Regulations.
https://truckingresearch.org/2021/10/idling-regulations-compendium/
74 Kotz, A. and Kelly, K. (2019). MOVES Activity Updates Using Fleet DNA Data: Interim Report.
NREL/TP-5400-70671. National Renewable Energy Laboratory. Golden, CO.
https://www.nrel.gov/docs/fyl9osti/70671.pdf
75 US EPA. (2016). Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and
Heavy-Duty Engines and Vehicles— Phase 2. 81 FR 206. Section III.D.l.c.v "Idle Reduction", page
73593.
76 National Academies Press. (2016). Commercial Motor Vehicle Driver Fatigue, Long-Term
Health, and Highway Safety: Research Needs.
https://www.ncbi.nlm.nih.gov/books/NBK384967
243
-------
77 US EPA. (2016). Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and
Heavy-Duty Engines and Vehicles— Phase 2. 81 FR 206 (October 25, 2016) Section III.D.l.a
"Tractor Baselines for Costs and Effectiveness", page 73587.
78 Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty
Engines and Vehicles— Phase 2. 81 FR 206. Section III.D.l.c.v "Idle Reduction", page 73593.
79 North American Council for Freight Efficiency. (2018). 2018 Annual Fleet Fuel Study.
https://nacfe.org/annual-fleet-fuel-studies/
80 Schoettle, B., Sivak, M., and Tunnell, M. (2016). A Survey of Fuel Economy and Fuel Usage by
Heavy-Duty Truck Fleets. SWT-2016-12. The University of Michigan. Sustainable Worldwide
Transportation.
https://trid.trb.org/View/1427438
81 US EPA. (2018). 2014 National Emissions Inventory, version 2: Technical Support Document.
Section 6.8.4. Office of Air Quality Planning and Standards. Research Triangle Park, North
Carolina.
82 Lutsey, N., Brodrick, C.-J., Sperling, D., and Oglesby, C. (2004). Heavy-Duty Truck Idling
Characteristics: Results from a Nationwide Truck Survey. Transportation Research Record. 1880
(1), 29-38.
https://doi.org/10.3141/1880-04
83 Transportation Research Board. (2019). Guide to Truck Activity Data for Emissions Modeling.
National Academies of Sciences, Engineering, Medicine. The National Academies Press.
Washington, DC.
https://www.nap.edu/catalog/25484/guide-to-truck-activity-data-for-emissions-modeling
84 Festin, S., US FHWA. (1996). Summary of National and Regional Travel Trends: 1970-1995.
Washington, DC.
http://www.fhwa.dot.gov/ohim/bluebook.pdf
85 National Highway Traffic Safety Administration. (2012). Fatality Analysis Reporting System
(FARS). Raw 2010 data.
http://www.nhtsa.gov/FARS
244
-------
86 Guensler, R., Yoon, S., Li, H., and Elango, V., Georgia Institute of Technology. (2007). Atlanta
Commute Vehicle Soak and Start Distributions and Engine Starts per Day: Impact on Mobile
Source Emission Rates. EPA-600-R-07-075. Atlanta, GA.
http://nepis.epa.gov/Adobe/PDF/P100AE2E.pdf
87 US EPA. (2024). Evaporative Emissions from Onroad Vehicles in MOVES5. EPA-420-R-24-014.
Office of Transportation and Air Quality. Ann Arbor, Ml.
https://www.epa.gov/moves/moves-onroad-technical-reports
88 Sierra Research, Inc. (2006). Development of Trip and Soak Activity Defaults for Passenger
Cars and Trucks in MOVES2006. SR2006-03-04. EPA Contract EP-C-05-037, Work Assignment
No. 0-01. Sacramento, CA.
89 US EPA. (2012). Development of Evaporative Emissions Calculations for the Motor Vehicle
Emissions Simulator MOVES2010. EPA-420-R-12-027. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P100F3ZY.txt
90 Indale, G., University of Tennessee. (2005). Effects of Heavy-Duty Diesel Vehicle Idling
Emissions on Ambient Air Quality at a Truck Travel Center and Air Quality Benefits Associated
with Advanced Truck Stop Electrification Technology (PhD dissertation). Knoxville, TN.
http://trace.tennessee.edu/utk graddiss/2085
91 US FHWA. (2017). Jason's Law Truck Parking Survey Results and Comparative Analysis.
ops.fhwa.dot.gov/freight/infrastructure/truck parking/jasons Iaw/truckparkingsurvey/ch3.htm
92 US EPA. (2023). 2020 National Emissions Inventory. Office of Air Quality Planning and
Standars. Research Triangle Park, North Carolina.
https://www.epa.gov/air-emissions-inventories/2020-national-emissions-inventory-nei-data
93 US Census Bureau. 2020 Census.
https://www.census.gov/2020results
94 US EPA. (2016). Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and
Heavy-Duty Engines and Vehicles—Phase 2. 81 FR 73478.
95 National Research Council. (2010). Technologies and Approaches to Reducing the
245
-------
Fuel Consumption of Medium- and Heavy-Duty Vehicles. Table 2-1. The
National Academies Press. Washington, DC.
https://doi.org/10.17226/12845
96 US FHWA. (2013). Vehicle Travel Information System. Table W-3.
http://www.fhwa.dot.gov/ohim/ohimvtis.cfm
97 US FHWA. (2006). Bridge Formula Weights. FHWA-HOP-06-105.
98 Sandhu, G. S., Frey, H. C., Bartelt-Hunt, S., and Jones, E. (2014). In-Use Measurement of the
Activity, Fuel Use, and Emissions of Front-Loader Refuse Trucks. Atmospheric Environment. Vol.
92, 2014, pp. 557-565.
https://doi.Org/10.1016/i.atmosenv.2014.04.036
99 Sandhu, G. S., Frey, H. C., Bartelt-Hunt, S., and Jones, E. (2015). In-Use Activity, Fuel Use, and
Emissions of Heavy Duty Diesel Roll-off Refuse Trucks. Journal of the Air & Waste Management
Association. Vol. 65, No. 3, pp. 306-323.
https://doi.org/10.1080/10962247.2014.99Q587
100 Sandhu, G. S., Frey, H. C., Bartelt-Hunt, S., and Jones, E. (2016). Real-World Activity, Fuel Use,
and Emissions of Diesel Side-Loader Refuse Trucks. Atmospheric Environment. Vol. 129, pp. 98-
104.
https://doi.Org/10.1016/i.atmosenv.2016.01.014
101 Sandhu, G. S., North Carolina State University. (2015). Evaluation of Activity, Fuel Use, and
Emissions of Heavy-Duty Diesel and Compressed Natural Gas Vehicles. Ph.D. Dissertation.
Raleigh, NC.
http://www.lib.ncsu.edu/resolver/1840.16/10222
102 US EPA. (2011). Final Rulemaking to Establish Greenhouse Gas Emissions Standards and Fuel
Efficiency Standards for Medium- and Heavy-Duty Engines and Vehicles: Regulatory Impact
Analysis. EPA-420-R-11-901. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P100EG9C.pdf
103 Choi, D. (2016). Tractor-Trailer Inputs to MOVES-FRM. Memorandum to Docket EPA-HQ-
OAR-2022-0829.
246
-------
https://www.regulations.gov/document/EPA-HQ-QAR-2014-0827-2221
104 Choi, D. (2016). Vocational Vehicle Inputs to MOVES-FRM. Memorandum to Docket No. EPA-
HQ-OAR-2014-0827-2222.
https://www.regulations.gov/document/EPA-HQ-QAR-2014-0827-2222
105 American Public Transportation Association. (2014). An Analysis of Transit Bus Axle Weight
Issues, Table 7.
106 US GPO. (2004). Road had force and inertia weight determination. Code of Federal
Regulations, Title 40 - Protection of Environment, Vol. 17, CFR 86.529-78.
107 Steven, H., United Nations Economic Commisssion for Europe. (2002). Worldwide
Harmonised Motorcycle Emissions Certification Procedure. UN/ECE-WP 29 - GRPE (informal
document no. 9, 46th GRPE, 13-17 January 2003, agenda item 3). Geneva, Switzerland.
http://www.unece.org
108 Warila, J. (2004). Derivation of Mean Energy Consumption Rates within the MOVES Modal
Framework. 14th Coordinating Research Council On-Road Vehicle Emissions Workshop (poster).
San Diego, CA.
109 US EPA. (2000). IM240 and Evap Technical Guidance. EPA-420-R-00-007. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P1008F0l.txt
110 US EPA. (2012). Final Rule for Model Year 2017 and Later Light-Duty Vehicle Greenhouse Gas
Emissions and Corporate Average Fuel Economy Standards. 11 FR 62624.
111 Petrushov, V. (1997). Coast Down Method in Time-Distance Variables. SAE International, SAE
970408. Detroit, Ml.
http://www.sae.org
112 US EPA. (2024). Emission Adjustments for Onroad Vehicles in MOVES5. EPA-420-R-24-013.
Office of Transportation and Air Quality. Ann Arbor, Ml.
https://www.epa.gov/moves/moves-onroad-technical-reports
247
-------
113 Koupal, J. (2001). Air Conditioning Activity Effects in MOBILE6. M6.ACE.001. EPA-420-R-01-
054. Ann Arbor, Ml.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P100226H.txt
114 US Census Bureau. (1999). 1997 Vehicle Inventory and Use Survey. EC97TV-US. Washington,
DC.
www.census.gov/librarv/publications/1997/econ/census/vehicle-inventory-and-use-
survey.html
115 Stanard, A., Fincher, S., Kishan, S., and Sabisch, M., Eastern Research Group, Inc. (2012).
Data Analyses on Drayage Heavy-Duty Vehicles. EPA EP-C-12-017. Work Assignment 0-2. Austin,
TX.
116 Davis, S. and Truitt, L. (2002). Investigation of Class 2b Trucks (Vehicles of 8,500 to 10,000 lbs
GVWR). Oak Ridge National Laboratory. ORNL/TM-2002.49, Oak Ridge, TN.
https://info.ornl.gov/sites/publications/Files/Pub57099.pdf
117 Union of Concerned Scientists (personal communication).
http://www.ucsusa.org
118 Data from 2011 was provided by US FHWA to replace Table 11.2 and 11.3 in the 1997 Federal
Highway Cost Allocation Study, specifying vehicles by weight (email correspondence), January
2013.
119 National Highway Traffic and Safety Administration. (2006). Vehicle Survivability and
Travel Mileage Schedules. DOT HS 809 952. Springfield, VA.
http://www-nrd.nhtsa.dot.gov/Pubs/809952.pdf
120 Davis, S., Diegel, S., and Boundy, R. Transportation Energy Data Book (TEDB) Edition 40. Oak
Ridge National Laboratory. ORNL/TM-2022/2376. Oak Ridge, TN.
https://tedb.ornl.gov
121 Motorcycle Industry Council. (2015). Motorcycle Statistical Annual. Irvine, CA.
https://www.mic.org
122 Ward's Automotive Inc.
248
-------
http://www.wardsauto.com
123 US GPO. (2016). Modeling CO2 emissions to show compliance for vocational vehicles and
tractors. Code of Federal Regulations, Title 40 - Protection of Environment, Vol. 36, CFR
1037.520.
124 US GPO. (2016). Determining CO2 emissions to show compliance for trailers. Code of Federal
Regulations, Title 40 - Protection of Environment, Vol. 36, CFR 1037.515.
125 North American Council for Freight Efficiency. (2015). 2015 Annual Fleet Fuel Study.
https://nacfe.org/downloads/nacfe-2015-annual-fleet-fuel-study/
126 US EPA. (2011). Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and
Heavy-Duty Engines and Vehicles. Federal Register Volume 76, No. 179. Page 57211.
127 US EPA. (2016). Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and
Heavy-Duty Engines and Vehicles— Phase 2. Federal Register Volume 81, No. 206. Pages
73587, 73608-73611.
249
------- |