Population and Activity of Onroad
Vehicles in MOVES3
&EPA
United States
Environmental PrulutUon
Agency
-------
Population and Activity of Onroad
Vehicles in MOVES3
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
NOTICE
This technical report does not necessarily represent final EPA decisions
or positions. It is intended to present technical analysis of issues using
data 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.
O EDA United States EPA-420-R-20-023
Environrrntntal Prolotliun „„„„
\^L_I Agency November2020
-------
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 12
2.4. Fuel Types 13
2.5. Road Types 14
2.6. Source Classification Codes (SCC) 14
2.7. Model Year Groups 15
2.8. Source Bins 15
2.9. Allowable Vehicle Modeling Combinations 16
2.10. Default Inputs and Fleet and Activity Generators 18
3. VMT by Calendar Year and Vehicle Type 20
3.1. Historic Vehicle Miles Traveled (1990 and 1999-2017) 20
3.2. Projected Vehicle Miles Traveled (2018-2060) 23
4. Vehicle Populations by Calendar Year 25
4.1. Historic Source Type Populations (1990 and 1999-2017) 25
4.2. Projected Vehicle Populations (2018-2060) 29
5. Fleet Characteristics 32
5.1. Source Type Definitions 32
5.2. Sample Vehicle Population 34
6. Vehicle Age-Related Characteristics 41
6.1. Age Distributions 41
6.2. Relative Mileage Accumulation Rate 45
7. VMT Distribution of Source Type by Road Type 53
8. Average Speed Distributions 55
8.1. Description of Telematics Dataset 55
8.2. Derivation of Default National Average Speed Distributions 57
8.3. Updated average speed distributions and comparison with MOVES2014 58
9. Driving Schedules and Ramps 62
9.1. Driving Schedules 62
9.2. Modeling of Ramps in MOVES 67
10. Off-Network Idle Activity 69
10.1. Off-Network Idle Calculation Methodology and Definitions 69
10.2. Light-Duty Off-Network Idle 71
10.3. Heavy-Duty Off-Network Idle 77
10.4. Off-network Idling Summary 83
11. Hotelling Activity 84
11.1. Hotelling Activity Distribution 85
11.2. National Default Hotelling Rate 87
12. Engine Start Activity 90
12.1. Light-Duty Start Activity 91
12.2. Heavy-Duty Start Activity 99
12.3. Motorcycle and Motorhome Starts 119
1
-------
13. Temporal Distributions 121
13.1. VMT Distribution by Month of the Year 122
13.2. VMT Distribution by Type of Day 123
13.3. VMT Distribution by Hour of the Day 124
13.4. Parking Activity 126
13.5. Hourly Hotelling Activity 130
14. Geographical Allocation of Activity 135
14.1. Source Hours Operating Allocation to Zones 135
14.2. Parking Hours Allocation to Zones 136
15. Vehicle Mass and Road Load Coefficients 137
15.1. Source Mass and Fixed Mass Factor 138
15.2. Road Load Coefficients 142
16. Air Conditioning Activity Inputs 148
16.1. ACPenetrationFraction 148
16.2. FunctioningACFraction 149
16.3. ACActivityTerms 150
17. Conclusion and Areas for Future Research 152
Appendix A Fuel Type and Regulatory Class Fractions from Previous Versions of MOVES 154
Al. Distributions for Model Years 1960-1981 154
A2. Distributions for Model Years 1982-1999 156
Appendix B 1990 Age Distributions 166
Bl. Motorcycles 166
B2. Passenger Cars 166
B3. Trucks 166
B4. Other Buses 167
B5. School Buses and Motor Homes 167
B6. Transit Buses 167
Appendix C Detailed Derivation of Age Distributions 169
CI. Generic Survival Rates 169
C2. Vehicle Sales by Source Type 171
C3. Base Year Age Distributions 174
C4. Historic Age Distributions 175
C5. Projected Age Distributions 176
Appendix D Driving Schedules 179
Appendix E Total Idle Fraction Regression Coefficients 182
Appendix F Source Masses from Previous Versions of MOVES 185
Fl. Motorcycles 186
F2. Passenger Cars 187
F3. Light-Duty Trucks 187
Appendix G Freeway Ramp Contribution at the County-Scale 189
Appendix H NREL Fleet DNA Preprocessing Steps 192
Appendix I Averaging Methods for Heavy-duty Telematics Activity Data 196
11. Evaluated Methods 196
12. Comparison of Evaluated Methods 199
13. Future Work 200
Appendix J Road Load Coefficient for Combination Trucks in HD GHG Rule 202
-------
Appendix K M0VES3 SourceUseTypePhysics Table 212
18. References 219
3
-------
List of Acronyms
AEO
Annual Energy Outlook publication
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
C02
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)
I/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
Medium-Heavy-Duty
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
ST
Source Type
STP
scaled-tractive power
SVP
Sample Vehicle Population
TDM
Travel Demand Models
5
-------
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 MOVES3 default database. These data
have been updated from previous versions of MOVES. 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
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 MOVES3, 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 2017.
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
modellers 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 MOVES3 since the
MOVES2014 release were peer-reviewed under EPA's peer review guidance3 in two separate
reviews conducted in 2017 and 2019. 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
Report Sections
AvgSpeedDistribution
Distribution of time among average speed bins
Section 8
Day VMTF raction
Distribution of VMT between weekdays and
weekend days
Section 13
Drive Schedule
Average speed of each drive schedule
Section 9
Drive Schedule Assoc
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
HotellingCalendarY ear
Rate of hotelling hours per total restricted access
VMT
Section 11
HourVMTFraction
Distribution of VMT among hours of the day
Section 13
HPMS VtypeY ear
Annual VMT by HPMS vehicle types
Section 3
IdleRegion
Map of idle regions to idle region IDs.
Section 10
ModelY earGroup
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
Mo nth VMTFractio n
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
StartsMonthAdjust
The monthAdjustFactor adjusts the starts per day
to reflect monthly variation in the number of
starts.
Sectionl2
StartsPerDay
StartsPerDay value is the number of starts per
average vehicle (of all source types). This value
varies by county (zonelD) and day type.
Sectionl2
StartsSourceTypeFraction
The allocation of total starts per day for all
vehicles to each of the MOVES source types.
Sectionl2
9
-------
Table 1-1 MOVES database elements covered in this report
Database Table Name
Content Summary
Report Sections
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
SourccTvpcAgcDistribution
Distribution of vehicle population among ages
Section 6
SourceTypeHour
The distribution of total daily hotelling among
hours of the day
Section 13
SourceTypeModelY ear
Prevalence of air conditioning equipment
Section 16
SourceTypePolProcess
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. MOWS Vehicle ami 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, source types have common body
types, and the road load coefficients (rolling load, aerodynamic drag, a) 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 MOVES3 source types are
listed in Table 2-1 along with the associated HPMS classes. More detailed source type
definitions are provided in Section 5.1.
11
-------
Table 2-1 Onroad Source Types in MOVES3
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
2.3.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 have their own
regulatory definition and therefore are an independent regulatory class.
Table 2-2 Regulatory Classes in MOVES3
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 (14,00 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 Bus (see CFR Sec. 86.091 2)
49
Gliders
Glider Vehicles7
12
-------
The EPA regulatory distinction between light-duty (LD) and heavy-duty (HD) trucks falls in the
midst of FHWA GVWR Class 2. Trucks of 6,001-8,500 lbs. GVWR are Class 2a; in MOVES,
they are considered light-duty trucks in regulatory class 30. Vehicles of 8,500-14,000 lbs.
GVWR are Class 2b and Class 3 and considered light heavy-duty vehicles (LHD) in regulatory
class 41. 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 5.2 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.
MOVES models vehicles powered by 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.8 Table 2-3 below
summarizes the fuel types available in MOVES.
fuelTypelD
defaultFormulationlD
Description
1
10
Gasoline
2
20
Diesel Fuel
3
30
Compressed Natural Gas (CNG)
5
50
Ethanol (E-85)
9
90
Electricity
ES3
It is important to note that not all fuel type/source type combinations can be modeled in
MOVES. For example, MOVES will not model gasoline-fueled long-haul combination trucks or
diesel motorcycles. Similarly, flexible fuel (E85-compatible) and electric vehicles are only
modeled for passenger cars, passenger trucks and light commercial trucks. In addition, MOVES
does not explicitly model hybrid powertrains, but accounts for these vehicles in calculating fleet-
average energy consumption and CO2 rates.a For more information on how MOVES models the
impact of fuels on emissions, please see the MOVES documentation on fuel effects.9
a 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.
13
-------
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 MOVES3
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
speeds and acceleration patterns may run MOVES at project level where emissions can be
calculated for individual links. In MOVES3, we removed the ramp road type as discussed in
Section 9.
2.6. Source Classification Codes (SCC)
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
14
-------
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
development10'11 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. For example,
a 2011 vehicle belongs to the "2011" model year group for estimating hydrocarbon running
exhaust emissions but belongs to the "2011-2020" group for estimating nitrous oxide running
emissions. Because these groupings are determined based on analysis of the actual or expected
emissions performance, the rationale for each model year grouping is provided in the MOVES
emission rate reports.10'11
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,b
• RR is a MOVES regClassID,
b In MOVES3, engTechID 1 is used for all fuel types except electric vehicles, where engTechID 30 is used instead.
Thus, in the current version, engTechID is somewhat redundant with fuel type and adds no new information when
determining source bin distributions or calculating emissions.
15
-------
• 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.
Table 2-5 Data Tables Used to Allocate Source r
"ype to Source Bin
Table Name
Key Fields*
Additional Fields
Notes
SourceTypePolProcess
sourceTypelD
polProcessID
isRegClassReqd
isMY GroupReqd
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
modelY earGroupID
Assigns model years to appropriate
model year groups for each
polProcessID.
SampleVehiclePopulation
sourceTypelD
modelYearlD
fuelTypelD
engTechID
regClassID
stmyFuelEngF raction
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.
Note:
* In these tables, the sourceTypelD and modelYearlD are combined into a single sourceTypeModelYearlD.
While details of the SourceTypePolProcess and PollutantProcessModelYear tables are discussed
in the reports on the development of the light- and heavy-duty emission rates,10'11 the
SampleVehiclePopulation (SVP) table is a topic for this report and is discussed in Section 5.2
2.9.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
16
-------
type in the SampleVehiclePopulation table. Table 2-7 summarizes the allowable source type-
regulatory class combinations in MOVES3. Table 2-8 shows the full set of allowable source
type, fuel type and regulatory class combinations. Additional discussion about decisions to
include and exclude certain types of vehicles can be found in Section 5.
Table 2-6 Matrix of the Allowable Source Type-fuel Type Combinations in MOVES3
(Allowable combinations are marked with an X)
Source Use Types
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 Types
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
E85-Capable
5
X
X
X
Electricity
9
X
X
X
17
-------
Table 2-7 Matrix of the allowable source type-regulatory class combinations in MOVES3
(Allowable combinations are marked with an X)
Source Use Types
Motorcycles
Passenger Cars
Passenger Trucks
Light Commercial
Trucks
Other Buses
Transit Buses
School Buses
Refuse Trucks
Short-Haul Single Unit
Tmcks
Long-Haul Single Unit
Tmcks
Motor Homes
Short-Haul Combination
Tmcks
Long-Haul Combination
Trucks
Regulatory Classes
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
Table 2-8 A summary of source type, fuel type and regulatory class combinations in MOVES3
sourceTypelD
fuelTypelD
regClassID
11
1
10
21
1, 2, 5, 9
20
31
1,2
30,41
5, 9
30
32
1,2
30,41
5, 9
30
41
1,2,3
42, 46, 47
42
1
42, 46, 47
2, 3
42, 46, 48
43
1,2,3
41,42, 46, 47
51
1,2,3
41,42, 46, 47
52
1,2,3
41,42, 46, 47
53
1,2,3
41,42, 46, 47
54
1,2,3
41, 42, 46, 47
61
1,2,3
46, 47
62
2
46, 47, 49
2.10. Default Inputs and Fleet ami 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
18
-------
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 for future years are generated from the base year
distribution.
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
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 emissions, 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 and to calculate off-network idling activity. Similarly, other
generators use MOVES inputs to develop operating mode distributions for hotelling activity,
starts and vapor venting.
19
-------
3. VY1T by Calendar Year ami Vehicle Type
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,0 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 1999-2017)
In MOVES3, VMT estimates for the historic years 1990 and 1999-2017 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,12
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
Generator13 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 more
reliable to retain the original 2000-2006 estimates because the information available for those
0 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. As in previous versions of MOVES, 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.
20
-------
years does not fully meet the requirements of the new methodology.'"1 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-
2016 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.
e Note that when MOVES is run at default scale using the "Nation" region aggregation option, it reduces the VMT
shown in Table 3-1 by 0.496 percent, which is the amount of national activity allocated to Puerto Rico and the
Virgin Islands based on allocation factors used for the 2017 NEI. However, the national VMT presented in
Highway Statistics Table VM-1 do not include activity occuring in Puerto Rico or the Virgin Islands. This results in
MOVES slightly underestimating VMT when running at default scale for the Nation region. EPA intends to address
this in future updates to MOVES.
21
-------
Table 3-1 Historic year VMT by HPMS vehicle class (millions of 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
Table 3-2
Year
FHWA Publication Source (Publication/Revision Date)
1990
Highway Statistics 1991 (October 1992)
1999
Highway Statistics 1999 (October 2000)
2000
Highway Statistics 2000 (April 2011)
2001
Highway Statistics 2001 (April 2011)
2002
Highway Statistics 2002 (April 2011)
2003
Highway Statistics 2003 (April 2011)
2004
Highway Statistics 2004 (April 2011)
2005
Highway Statistics 2005 (April 2011)
2006
Highway Statistics 2006 (April 2011)
2007
Highway Statistics 2007 (April 2011)
2008
Highway Statistics 2008 (April 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)
22
-------
3.2. Projected Vehicle Miles Traveled (2018-2060)
The Annual Energy Outlook (AEO)14 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 MOVES3, VMT for years beyond 2017 are based
on the reference case VMT projections from AEO2019. 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 2017 base year HPMS
data. Since AEO2019 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
explicitly accounted for elsewhere in AEO. Since buses span a large range of heavy-duty
vehicles and activity, the combination of AEO's light-medium-, medium- and heavy-heavy-duty
categories 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
iPMS classes for projecting VMT
AEO VMT Category Groupings
HPMS Class
Total Light-Duty VMT1
10 - Motorcycles
Total Commercial Light Truck VMT11
25 - Light Duty Vehicles
Total Heavy-Duty VMT111
40 - Buses
Light-Medium Subtotal VMT111
+
Medium Subtotal VMT111
50 - Single Unit Trucks
Heavy Subtotal VMT111
60 - Combination Trucks
Notes:
I From AEO2019 Table 42: Light-Duty VMT by Technology Type
II From AEO2019 Table 47: Transportation Fleet Car and Truck VMT by Type and Technology
III From AEO2019 Table 50: 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 2017 base year VMT from Highway Statistics Table VM-1. The
resulting values are presented in Table 3-4 below.
23
-------
Table 3-4 VMT projections for 2018-2060 by
Year
Motorcycles
Light-Duty
Vehicles
Buses
Single Unit Trucks
Combination Trucks
2018
20,489
2,925,906
17,702
119,053
186,717
2019
20,773
2,966,395
18,236
122,415
192,558
2020
20,986
2,996,822
18,518
124,844
195,055
2021
21,144
3,019,366
18,727
127,268
196,310
2022
21,272
3,037,686
18,986
129,735
198,375
2023
21,358
3,049,959
19,257
132,424
200,426
2024
21,421
3,058,950
19,523
135,271
202,249
2025
21,483
3,067,890
19,806
138,244
204,240
2026
21,628
3,088,503
20,104
141,420
206,301
2027
21,778
3,110,004
20,363
144,496
207,806
2028
21,927
3,131,268
20,654
148,035
209,397
2029
22,040
3,147,373
20,913
151,618
210,426
2030
22,153
3,163,519
21,178
155,143
211,610
2031
22,261
3,178,911
21,479
159,319
212,801
2032
22,364
3,193,616
21,761
163,175
213,962
2033
22,465
3,208,069
22,050
167,225
215,047
2034
22,564
3,222,206
22,362
171,457
216,374
2035
22,653
3,234,892
22,704
175,847
218,032
2036
22,781
3,253,155
23,041
180,341
219,523
2037
22,922
3,273,391
23,395
184,882
221,264
2038
23,073
3,294,957
23,752
189,570
222,909
2039
23,227
3,316,903
24,081
193,827
224,490
2040
23,383
3,339,124
24,401
198,488
225,545
2041
23,540
3,361,517
24,724
202,695
227,058
2042
23,698
3,384,089
25,042
207,087
228,329
2043
23,860
3,407,242
25,359
211,500
229,563
2044
24,023
3,430,581
25,704
216,067
231,125
2045
24,184
3,453,578
26,073
220,890
232,840
2046
24,348
3,476,893
26,439
225,878
234,360
2047
24,505
3,499,399
26,822
231,079
235,975
2048
24,651
3,520,213
27,221
236,461
237,688
2049
24,794
3,540,636
27,604
241,922
239,045
2050
24,928
3,559,739
28,004
247,589
240,521
2051
25,062
3,578,946
28,411
253,389
242,006
2052
25,197
3,598,257
28,824
259,325
243,501
2053
25,333
3,617,671
29,242
265,399
245,005
2054
25,470
3,637,191
29,667
271,617
246,518
2055
25,607
3,656,815
30,098
277,979
248,040
2056
25,746
3,676,546
30,535
284,491
249,572
2057
25,885
3,696,383
30,978
291,156
251,113
2058
26,024
3,716,327
31,428
297,976
252,664
2059
26,165
3,736,379
31,884
304,956
254,224
2060
26,306
3,756,539
32,347
312,100
255,794
JPMS class (millions of miles)
24
-------
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 ami 1999-2017)
MOVES populations for calendar years 1990 and 1999-2017 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-1,6 and
passenger car populations are from registrations reported in Table MV-1.15 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 IHS16'17 for calendar years 1999 and 2014. These fractions were calculated as the
ratio of the individual source type registrations 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 2014. However, there are a few
caveats to this analysis:
• The distinction between passenger light-duty trucks (31) and commercial light-duty
trucks (32) has been updated from previous versions of MOVES. In MOVES3, 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 IHS data, the ratio of
passenger to commercial light-duty trucks from 2014 IHS data was used for all calendar
years.
• The 2014 IHS data was unable to distinguish between short-haul (52) and long-haul (53)
single-unit trucks and consequentially grouped them together. These vehicles are
differentiated in MOVES3 using an earlier IHS data set for 2011 which was able to
differentiate between these vehicles. From the earlier data set, it was determined 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.
25
-------
• Source type distributions were needed to allocate the historic 2015-2017 populations.
Rather than projecting the linear interpolations, the distributions for 2014 were held
constant for 2015-2017.
Table 4-1 Source type distributions used to al
ocate true
t populations in MOVES3*
Year
31/30
32/30
51/50
52/50
53/50
54/50
61/60
62/60
1990"
0.895947
0.104053
0.013311
0.767722
0.033860
0.185107
0.625648
0.374352
1999""
0.895947
0.104053
0.015472
0.791929
0.034927
0.157671
0.574437
0.425563
2000
0.895947
0.104053
0.014852
0.797084
0.035155
0.152909
0.561208
0.438792
2001
0.895947
0.104053
0.014232
0.802239
0.035382
0.148146
0.547979
0.452021
2002
0.895947
0.104053
0.013612
0.807394
0.035610
0.143384
0.534750
0.465250
2003
0.895947
0.104053
0.012992
0.812549
0.035837
0.138622
0.521521
0.478479
2004
0.895947
0.104053
0.012372
0.817704
0.036064
0.133859
0.508292
0.491708
2005
0.895947
0.104053
0.011752
0.822859
0.036292
0.129097
0.495063
0.504937
2006
0.895947
0.104053
0.011133
0.828014
0.036519
0.124334
0.481835
0.518166
2007
0.895947
0.104053
0.010513
0.833169
0.036746
0.119572
0.468606
0.531394
2008
0.895947
0.104053
0.009893
0.838324
0.036974
0.114810
0.455377
0.544623
2009
0.895947
0.104053
0.009273
0.843479
0.037201
0.110047
0.442148
0.557852
2010
0.895947
0.104053
0.008653
0.848634
0.037428
0.105285
0.428919
0.571081
2011
0.895947
0.104053
0.008033
0.853789
0.037656
0.100523
0.415690
0.584310
2012
0.895947
0.104053
0.007413
0.858944
0.037883
0.095760
0.402461
0.597539
2013
0.895947
0.104053
0.006793
0.864099
0.038110
0.090998
0.389232
0.610768
2014*"
0.895947
0.104053
0.006173
0.869254
0.038338
0.086235
0.376003
0.623997
2015
0.895947
0.104053
0.006173
0.869254
0.038338
0.086235
0.376003
0.623997
2016
0.895947
0.104053
0.006173
0.869254
0.038338
0.086235
0.376003
0.623997
2017
0.895947
0.104053
0.006173
0.869254
0.038338
0.086235
0.376003
0.623997
Note:
* Fractions may not sum to one due to rounding.
" Fractions from 1990 were retained from MOVES201418 with the exceptions noted in the text.
*" Fractions from 1999 and 2014 were calculated from IHS registration data with the exceptions noted
in the text; fractions for other years were estimated from these values.
Buses were allocated using different data sources:
• School bus (43) populations for 2002-2017 come from the School Bus Fleet Fact Book19
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
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.
• Transit bus (42) populations were calculated from the Federal Transit Administration's
National Transit Database (NTD)20 data series on Revenue Vehicle Inventory and Rural
Revenue Vehicle Inventory. See Section 5.1.4 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.
26
-------
• 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, 2016, and 2017, 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
and 2017 values were estimated by linearly extrapolating from 2013, 2014, and 2015.
For all source type populations derived from Table VM-1, note that this registration data has the
same vehicle category differences 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.
Note that the national vehicle populations do not include Puerto Rico or the Virgin Islands.
When MOVES is run at the national scale for the entire country, it assumes Puerto Rico and the
Virgin Islands are included in the vehicle populations and accordingly reduces the national
activity, so the results correspond to just the 50 states and Washington DC. Therefore, the
national vehicle populations were increased by the proportion of activity allocated to Puerto Rico
and the Virgin Islands, so that when MOVES is run at the national scale for the entire country,
the correct populations are used. In MOVES3, 0.496 percent of national activity is allocated to
Puerto Rico and the Virgin Islands, based on allocation factors used for the 2017 NEI. See
Section 14 for more information on geographical allocation.
27
-------
Table <¦
-2 Historic source type populations for calem
ar years 1990 and
999-2017 (in thousands)
Year
Motorcycle
Passenger
Car
Passenger
Truck
Light
Commercial
Truck
Other
Bus
Transit
Bus
School
Bus
Refuse
Truck
Single Unit
Short-Haul
Truck
Single Unit
Long-Haul
Truck
Motor
Home
Combination
Short-Haul
Truck
Combination
Long-Haul
Truck
1990
3,676
144,265
34,694
4,062
173
48
320
58
3,317
146
800
1,015
608
1999
4,053
133,092
66,794
7,821
227
64
420
103
5,256
232
1,046
1,327
983
2000
4,368
134,288
70,280
8,229
240
67
443
98
5,269
232
1,011
1,359
1,063
2001
4,928
138,320
74,313
8,701
241
67
445
100
5,622
248
1,038
1,361
1,122
2002
5,029
136,598
75,174
8,802
244
68
452
95
5,651
249
1,004
1,297
1,129
2003
5,397
136,346
76,914
9,006
239
69
473
92
5,762
254
983
1,255
1,151
2004
5,810
137,111
81,367
9,527
257
69
473
89
5,895
260
965
1,226
1,186
2005
6,258
137,249
84,508
9,895
266
70
475
88
6,137
271
963
1,225
1,250
2006
6,712
136,075
87,814
10,282
276
71
479
87
6,454
285
969
1,248
1,342
2007
7,174
136,611
89,735
10,507
268
83
488
86
6,796
300
975
1,241
1,407
2008
7,792
137,763
89,396
10,468
272
85
491
82
6,983
308
956
1,183
1,415
2009
7,969
135,552
89,594
10,491
295
87
464
78
7,083
312
924
1,163
1,467
2010
8,049
131,545
89,562
10,487
286
90
474
71
7,008
309
869
1,100
1,465
2011
8,480
126,283
97,328
11,396
293
89
475
63
6,709
296
790
1,024
1,440
2012
8,497
111,845
110,180
12,901
300
92
470
61
7,070
312
788
999
1,483
2013
8,447
114,243
110,057
12,887
299
95
475
55
7,057
311
743
967
1,517
2014
8,460
114,467
113,586
13,300
291
99
486
52
7,276
321
722
974
1,616
2015
8,644
113,427
117,002
13,700
302
104
487
52
7,387
326
733
1,038
1,723
2016
8,723
113,525
121,168
14,188
320
107
477
54
7,641
337
758
1,040
1,726
2017
8,759
111,731
125,389
14,682
334
109
474
58
8,157
360
809
1,093
1,814
Note that the decline in sales seen in the 2008 recession results in a flattening of total population growth rates and eventually a decline
in total population for passenger cars and long-haul combination trucks as shown in Table 4-2. This suggests that the decline in sales
was accompanied by a delay in the scrappage of older vehicles. The dynamic vehicle survival rates in MOVES and their impact on
age distributions are discussed in Section Appendix C.
28
-------
4.2. Projected Vehicle Populations (2018-2060)
Vehicle stock estimates from the reference case of AEO2019 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 AEO2019 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
combination of AEO's light-medium-, medium- and heavy-heavy-duty categories 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 Stock11
31 - Passenger Truck
32 - Light Commercial Truck
Total Heavy-Duty Stock111
41 - Other Bus
42 - Transit Bus
43 - School Bus
Light-Medium Subtotal Stock111
+
Medium Subtotal Stock111
51 - Refuse Truck
52 - Single Unit Short-haul Truck
53 - Single Unit Long-haul Truck
54 - Motor Home
Heavy Subtotal Stock111
61 - Combination Short-haul Truck
62 - Combination Long-haul Truck
Notes:
I From AEO2019 Table 40: Light-Duty Vehicle Stock by Technology Type
II From AEO2019 Table 46: Transportation Fleet Car and Truck Stock by Type and Technology
III From AEO2019 Table 50: Freight Transportation Energy Use
The percent growth over time was calculated for each of the groups described above and applied
to the 2017 base year source type populations. The resulting populations are presented in Table
4-4.
29
-------
Table 4-4
rojected source type populations for 2018-2060 (in thousands)
Year
Motorcycle
Passenger
Car
Passenger
Truck
Light
Commercial
Truck
Other
Bus
Transit
Bus
School
Bus
Refuse
Truck
Single
Unit
Short-
Haul
Truck
Single
Unit
Long-
Haul
Truck
Motor
Home
Combination
Short-Haul
Truck
Combination
Long-Haul
Truck
2018
8,842
112,796
125,865
14,738
337
110
478
59
8,282
365
822
1,093
1,814
2019
8,920
113,794
126,450
14,806
341
111
484
60
8,434
372
837
1,099
1,824
2020
8,996
114,753
127,126
14,885
345
112
489
61
8,595
379
853
1,099
1,824
2021
9,066
115,650
127,849
14,970
348
113
494
62
8,747
386
868
1,097
1,820
2022
9,124
116,397
128,428
15,038
352
115
499
63
8,906
393
883
1,099
1,824
2023
9,190
117,237
129,052
15,111
355
116
504
64
9,056
399
898
1,098
1,822
2024
9,263
118,167
129,670
15,183
359
117
509
65
9,220
407
915
1,097
1,821
2025
9,341
119,159
130,193
15,244
363
118
516
67
9,397
414
932
1099
1,824
2026
9,419
120,151
130,680
15,301
368
120
522
68
9,580
423
950
1102
1,828
2027
9,497
121,148
130,996
15,339
372
121
529
69
9,768
431
969
1105
1,834
2028
9,573
122,114
131,124
15,353
375
122
533
70
9,915
437
984
1101
1,827
2029
9,647
123,058
131,138
15,355
380
124
539
72
10,111
446
1,003
1100
1,825
2030
9,723
124,039
130,943
15,332
383
125
544
73
10,254
452
1,017
1100
1,826
2031
9,810
125,140
130,905
15,328
389
127
553
75
10,517
464
1,043
1102
1,829
2032
9,900
126,288
130,645
15,297
393
128
558
76
10,673
471
1,059
1104
1,831
2033
9,995
127,502
130,382
15,267
396
129
563
77
10,846
478
1,076
1098
1,822
2034
10,097
128,807
130,027
15,225
400
130
568
78
11,014
486
1,093
1095
1,818
2035
10,204
130,170
129,628
15,178
405
132
575
80
11,214
495
1,112
1,098
1,822
2036
10,315
131,585
129,131
15,120
410
133
582
81
11,426
504
1,134
1,100
1,825
2037
10,432
133,074
128,577
15,055
416
135
590
83
11,657
514
1,156
1,104
1,833
2038
10,555
134,649
128,077
14,997
421
137
598
84
11,893
525
1,180
1,105
1,834
2039
10,682
136,261
127,475
14,926
426
139
604
86
12,059
532
1,196
1,109
1,840
2040
10,811
137,910
126,979
14,868
430
140
610
87
12,289
542
1,219
1,097
1,821
30
-------
Table 4-4 Projected source type population for 2018-2060 (in thousands)
Year
Motorcycle
Passenger
Car
Passenger
Truck
Light
Commercial
Truck
Other
Bus
Transit
Bus
School
Bus
Refuse
Truck
Single
Unit
Short-
Haul
Truck
Single
Unit
Long-
Haul
Truck
Motor
Home
Combination
Short-Haul
Truck
Combination
Long-Haul
Truck
2041
10,941
139,569
126,439
14,805
433
141
615
88
12,433
548
1,233
1,099
1,824
2042
11,070
141,212
126,011
14,755
439
143
623
90
12,680
559
1,258
1,100
1,826
2043
11,196
142,818
125,669
14,715
447
146
635
92
13,018
574
1,291
1,106
1,835
2044
11,320
144,406
125,332
14,675
456
148
647
95
13,320
587
1,321
1,114
1,849
2045
11,441
145,953
125,020
14,639
462
150
656
96
13,569
598
1,346
1,119
1,857
2046
11,557
147,428
124,766
14,609
469
153
666
98
13,865
611
1,375
1,123
1,864
2047
11,664
148,792
124,606
14,590
478
156
679
101
14,215
627
1,410
1,130
1,875
2048
11,761
150,031
124,344
14,560
485
158
688
103
14,477
639
1,436
1,135
1,883
2049
11,850
151,170
124,180
14,540
491
160
697
105
14,753
651
1,464
1,134
1,882
2050
11,930
152,192
124,018
14,521
498
162
707
107
15,044
663
1,492
1,137
1,886
2051
12,011
153,221
123,857
14,503
505
165
717
109
15,340
677
1,522
1,139
1,890
2052
12,092
154,256
123,696
14,484
513
167
728
111
15,643
690
1,552
1,142
1,895
2053
12,174
155,299
123,535
14,465
520
169
738
113
15,951
704
1,582
1,144
1,899
2054
12,256
156,348
123,375
14,446
527
172
749
116
16,265
717
1,614
1,147
1,903
2055
12,339
157,405
123,215
14,427
535
174
759
118
16,586
732
1,645
1,149
1,907
2056
12,422
158,468
123,054
14,409
543
177
770
120
16,913
746
1,678
1,152
1,911
2057
12,506
159,539
122,895
14,390
550
179
781
122
17,246
761
1,711
1,154
1,916
2058
12,591
160,618
122,735
14,371
558
182
793
125
17,586
776
1,745
1,157
1,920
2059
12,676
161,703
122,575
14,353
566
184
804
127
17,932
791
1,779
1,159
1,924
2060
12,762
162,796
122,416
14,334
574
187
815
130
18,286
806
1,814
1,162
1,928
31
-------
5. Fleet Characteristics
Despite the availability of vehicle registration databases, comprehensive surveys for
characterizing travel pattern 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.21'22 To develop
MOVES defaults, we have merged registration and survey data with activity measurements in an
effort to identify key vehicle parameters such as weight, axle and tire configuration and typical
trip range.
MOVES categorizes vehicles into thirteen source types as described in Section 2.1, which are
defined using physical characteristics, such as number of axles and tires and travel behavior
characteristics, such as typical trip lengths. This section describes the defining characteristics of
the source types in greater detail, explains how source type is related to fuel type and regulatory
class through the SampleVehiclePopulation table and how MOVES3 estimates and projects the
number of vehicles in each category.
5.1.Source Type Definitions
MOVES source types are intended to further divide HPMS vehicle classifications into groups of
vehicles with similar activity patterns. For example, passenger trucks and light commercial
trucks are expected to have different daily trip patterns.
5.1.1. 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."23 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.
5.1.2. Passenger Cars
Passenger cars are defined as any coupes, compacts, sedans, or station wagons with the primary
purpose of carrying passengers.23 For consistency with vehicle emission standards, the category
also includes some small crossover vehicles.24 All passenger cars (sourceTypelD 21) are
categorized in the light-duty vehicle regulatory class (regClassID 20).
5.1.3. Light-Duty Trucks
Light-duty trucks include pickups, most sport utility vehicles (SUVs) and vans.23 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; with the exception of Class 2b
trucks (8,500 to 10,000 lbs) with two axles or more and at least six tires, colloquially known as
"duallies", which FHWA classifies into the single-unit truck category.
32
-------
In MOVES, a light-duty truck is considered a passenger truck (sourceTypelD 31) if it is
registered to an individual, or a light-duty commercial truck (sourceTypelD 32) if it is registered
to an organization or business.
Because the Class 2b trucks with only 2 axles and only 4 tires are classified in the light-duty
source types, sourceTypelDs 31 and 32 contain vehicles in both the light-duty truck regulatory
class (regClassID 30) and the Class 2b and 3 truck regulatory class (regClassID 41) as discussed
in Section 5.2.3.
5.1.4. Buses
MOVES has three bus source types: other (sourceTypelD 41), transit (sourceTypelD 42) and
school buses (sourceTypelD 43).
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.25
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.26
Any other buses that do not fit into the transit or school bus categories are modeled in MOVES
as "other" buses/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. However, they do have different age
distributions and driving schedules as described in subsequent sections.
5.1.5. Single-Unit Trucks
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 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." The difference between short-haul
and long-haul single-unit trucks is their primary trip length; short-haul trucks travel less than or
equal to 200 miles a day and long-haul trucks travel more than 200 miles a day.
f Note, in previous versions of MOVES, "other" buses were called "intercity" buses and defined slightly differently.
33
-------
5.1.6. Combination Trucks
The combination truck HPMS class in MOVES consists of two source types: short-haul
(sourceTypelD 61) and long-haul combination trucks (sourceTypelD 62). These are heavy-duty
trucks that are not single-frame. Like single-unit trucks, short-haul and long-haul combination
trucks are distinguished by their primary trip length; short-haul trucks travel less than or equal to
200 miles a day and long-haul trucks travel more than 200 miles a day. Generally, short-haul
combination trucks are older than long-haul combination trucks and these short-haul trucks are
often purchased in secondary markets, such as for drayage applications, after being used
primarily for long-haul trips.27
5.2. Sample Vehicle Population
To match source types to emission rates, MOVES must associate each source type with specific
fuel types 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 and regulatory class allocation for
each source type and model year. Written out mathematically in Equation 5-1, we define the
stmyFraction as
where the number of vehicles JV in a given model year i, regulatory class j, fuel type k and
source type I is divided by the sum of vehicles across the set of all regulatory classes / and all
fuel types K. That is, the denominator is the total for 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 and regulatory classes 30 and 41. A value of zero indicates
that the MOVES default population of vehicles of that source type, model year, fuel type and
regulatory class is negligible or does not exist.
However, these default distributions in the stmyFraction may be modified by the user to model
local conditions through the Alternative Vehicle Fuel and Technology (AVFT) table. To allow
these user inputs, the StmyFuelEngFraction indicates the expected regulatory class distribution
for each allowable combination of source type, model year and fuel type, whether or not these
vehicles exist in the default. Similar to the stmyFraction above, we define StmyFuelEngFraction
in Equation 5-2 as
/ (stmy)ijkl =
Equation 5-1
34
-------
Nj ,¦ u i
f (stmyfueleng)iJkl = — ,
' ' Equation 5-2
' Ni,j,k,i
*-ijeJ
for number of vehicles JV, model year i, regulatory class j, fuel type /t, source type I and the set
of all regulatory classes/. In this case, the denominator is the total for a given source type, model
year and fuel type and so the stmyFuelEngFraction must sum to one for each combination of
source type, model year and fuel type. For example, for model year 2010 gasoline passenger
trucks, the table will list a StmyFuelEngFraction for regulatory class 30 and another for
regulatory class 41. In this example, while the stmyFraction indicates that the MOVES defaults
assign zero fraction of model year 2010 passenger trucks to the electricity fuel type, the
StmyFuelEngFraction indicates a default (hypothetical) regulatory class distribution if these
vehicles existed. In this case, MOVES would model any electric passenger trucks as belonging to
regulatory class 30. The stmyFraction is particularly important because users can edit fuel type
distributions using the Alternative Vehicle Fuel and Technology (AVFT) importer. For instance,
a user can create a future scenario in which there is a high penetration of electric passenger
trucks. The StmyFuelEngFraction allows MOVES to assign vehicles to regulatory class without
requiring this input from the user.
As noted in Section 2.4, these fuel type fractions indicate the fuel capability of the vehicle and
not 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 5-1 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—meaning they
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.8
35
-------
75%
50%
25%
0%
Fuel Type
-- Gasoline
-- Diesel
-- Flexible Fuel
-- Electricity
100%-
75% -
50% -
25% -
0% -
I960 1970 1980 1990 2000 20'l0 2020 2030 2040 2050 2060
Model Year
Figure 5-1 Default fuel fractions for light-duty source types in MOVES3
Both the stmyFractions and the stmyFuelEngFractions were calculated primarily using the 2014
IHS data set. However, in MOVES3, the fuel type and regulatory class distributions were
unchanged from MOVES2014 for the following source type and model year combinations:
• Passenger cars, school buses, refuse trucks, short-haul and long-haul single-unit trucks
and short-haul and long-haul combination trucks prior to model year 2000
• Passenger trucks and light commercial trucks prior to model year 1981
The previous versions of MOVES relied on combining vehicle registration data sets from IHS
with the Vehicle Inventory and Use Survey (VIUS). Because the last time the VIUS was
performed was in 2002, we retained the previous analysis for model years before 2000 but used
the 2014 IHS data set without combining it with the VIUS data for model years 2000 and later.
However, passenger trucks and light commercial trucks used the 2014 IHS data for 1981 and
later because we changed the definition of these vehicle types, as described in Section 5.1.2.
Therefore, they are no longer consistent with the VIUS definition. Unfortunately, the data are too
scarce in 2014 for pre-1981 model years, so we continued to rely on the previous analysis for
those model years. The documentation of the previous analysis may be found in Appendix A.
0"
s
s/
'—<
p
C/3
CD
p
CJQ
cd
>-J
O
p
'S.
S3
_o
O
J-l
Ph
CD
H
13
3
P-H
100%-
75% -
50% -
25% -
0% -
p
CD
P
CJQ
H
r
oq'
tr
O
o
36
-------
The fuel type and regulatory class distributions for all other source type and model year
combinations are described below.
As the 2014 IHS data set does not contain information on model year 2015 and later vehicles, we
held all distributions for these vehicles constant at the model year 2014 values, except where
noted below.
Before the fuel type and regulatory class distributions could be calculated from the 2014 IHS
data, the data set needed to be cleaned. For the source type field, there were many class 3 trucks
that were classified as a light-duty source type; as MOVES requires class 3 trucks to be modeled
in a heavy-duty source type, these were all re-classified as "other single-unit trucks" (see Section
5.2.5 for an explanation of this source categorization). Additionally, some compact SUVs were
originally classified as light trucks where EPA emission certification data showed that those
particular makes and models were regulated as cars;24 we re-classified these vehicles as
passenger cars. For the fuel type field, electric hybrids with gasoline or diesel were grouped with
fully gasoline or diesel vehicles, since MOVES does not model hybrids separately. Vehicles
categorized as "ethanol" or "flexible" were considered to be in the MOVES E-85 fuel category.
If the fuel type was unknown for light-duty source types or "other single-unit trucks," it was
assumed to be gasoline. If it was unknown for buses, refuse trucks, or combination trucks, the
fuel type was assumed to be diesel. All electric vehicles were dropped from the data set for
reasons described in the light-duty sections below. Any remaining vehicles with unknown fuel,
other alternative fuels (including hydrogen fuel cell, methanol and "convertible"), or vehicles
with source type/fuel type combinations that MOVES cannot model (such as CNG light
commercial trucks) were also dropped from the data set.
5.2.1. Motorcycles
All motorcycles fall into the motorcycle regulatory class (regClassID 10) and must be fueled by
gasoline. Although some alternative fuel motorcycles may exist, they account for a negligible
fraction of total US motorcycle sales and cannot be modeled in MOVES.
5.2.2. Passenger Cars
Any passenger car is considered to be in the light-duty vehicle regulatory class (regClassID 20).
The 2014 IHS data set provided the split between gasoline, diesel and E-85 capable cars in the
SampleVehiclePopulation table. For model years 2015 and later, we used Department of Energy
car sales projections from AEO's table "Light-Duty Vehicle Sales by Technology Type"28 to
derive fuel distributions and applied them to the SVP fractions for regulatory class 20. The
distribution for model year 2015 was derived from AEO2017, 2016 was derived from AEO2018
and model years 2017 and later were derived from AEO2019.g
g Each AEO contains historic data for a couple of years prior to the projected years. AEO2019, which is used as the
basis for the VMT and vehicle population projections, contains historic data only back to 2017. Therefore,
AEO2018 was used to split light-duty fuel types for model year 2016, AEO2017 was used for model year 2015, and
the 2014 IHS data was used for model years 2014 and earlier. .
37
-------
In MOVES, all electric passenger cars are modeled in the national case to have zero penetration.
This is because electric vehicle market penetration varies widely by geographic region and
MOVES does not have the capabilities to model this variance accurately at the national scale.
However, MOVES may be run at the county or project scale with local information to accurately
capture this detail. MOVES cannot model CNG passenger cars.
5.2.3. Light-Duty Trucks
Since passenger and light commercial trucks are defined as light-duty vehicles, they are
constrained to regulatory class 30 and 41. Light-duty trucks in the 2014 IHS data set with a
GVWR class of 1, 2, or 2a were classified as regulatory class 30 and Class 2b trucks were
classified as regulatory class 41. The 2014 IHS data set also provided the split between gasoline,
diesel and E-85 capable trucks. Please note that all E-85 light-duty trucks are modeled as
regulatory class 30.
For model years 2015 and later, we used Department of Energy light truck and light commercial
truck sales projections from AEO's tables "Light-Duty Vehicle Sales by Technology Type"28
and "Transportation Fleet Car and Truck Sales by Type and Technology"29 to derive fuel
distributions and applied them to the SVP fractions for regulatory class 30. The distribution for
model year 2015 was derived from AEO2017, 2016 was derived from AEO2018 and model
years 2017 and later were derived from AEO2019.g
In MOVES, all electric light-duty trucks are modeled in the national case to have zero
penetration. This is because electric vehicle market penetration varies widely by geographic
region and MOVES does not have the capabilities to model this variance accurately at the
national scale. However, MOVES may be run at the county or project scale with local
information to accurately capture this detail. Please note that all electric light-duty trucks are
modeled as regulatory class 30. MOVES cannot model CNG light-duty trucks.
5.2.4. Buses
Since school buses have a distinguishing characteristic in their VIN, they are well represented in
the 2014 IHS data set and we were able to calculate their fuel type and regulatory class
distributions. However, the 2014 IHS data set was unable to distinguish between transit buses
and other buses and so these categories were grouped together. As the National Transit Database
does not contain weight class information, that source could not be used to calculate regulatory
class distributions for transit buses. Considering that the vehicle types in both the transit and
"other" bus categories may overlap, we decided to keep these categories grouped together when
determining fuel type and regulatory class 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 86.091-2.30 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 in regulatory class 47.
Additionally, MOVES3 can only model CNG school buses and other buses in regulatory class 47
and it cannot model electric or E-85 buses.
38
-------
5.2.5. Single-Unit Trucks
Single-unit vehicles are distributed among the heavy-duty regulatory classes (regClassIDs 41,
42, 46 and 47) and between diesel and gasoline fuels based on the 2014 IHS data set. The 2014
IHS data set categorized single-unit trucks into refuse trucks (based on ownership), motor homes
and "other single-unit trucks." Lacking a way to differentiate these trucks into short-haul and
long-haul without resorting back to the VIUS, we used the fuel type and regulatory class
distributions for "other single-unit trucks" identically for both short-haul and long-haul single-
unit trucks. As with the other heavy-duty vehicles, MOVES3 can only model CNG single-unit
trucks in regulatory class 47. MOVES cannot model electric or E-85 single-unit trucks.
5.2.6. Combination Trucks
Combination trucks consist mostly of Class 8 trucks in the MOVES HHD regulatory class
(regClassID 47) but also contain Class 7 trucks in the MHD regulatory class (regClassID 46) as
well as glider trucks11 (regClassID 49).
Almost all combination trucks are diesel-fueled. MOVES does not model gasoline or CNG long-
haul combination trucks, but it models those fuel types for short-haul combination trucks. The
regulatory class and fuel type distributions are primarily based on the 2014 IHS data set.
Combination trucks were split between long-haul and short-haul using VMT fractions based on
estimates from the FHWA's 2007 Freight Analysis Framework as analyzed in CRC A-88.31
We estimated the glider population based on annual glider production volume (sales) data for
model years 2010 to 2016 shared as confidential business information (CBI) from two glider kit
manufacturers.32 A majority of the glider population is in Class 8 vehicles and therefore, in
MOVES, the gliders regulatory class only applies to combination short- and long-haul trucks
(MOVES sourcetypes 61 and 62, respectively).
For use in MOVES, we assumed a sale of 500 for glider vehicles for years prior to 2010 and
rounded the reported production volumes in the remaining years 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. Since only partial data was available for 2017, the value from 2014 was
used for 2017.1 The value used for 2018 and later is discussed below.
Table 5-1: Annual Glider Vehicle Sales Estimates Applied in MOVES Based on CBI Data Shared
MY
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018+
Glider
Population
500
500
1000
3000
4000
5000
8000
12000
7000
8000
4000
h "Glider trucks" refers to vehicles with new chassis but older engines that do not meet MY 2007 or 2010 emissions
standards.
1 In 2017, glider manufacturers are limited to producing their maximum production between MYs 2010 and 2014.
See 81 FR 73478 for more information.
39
-------
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)89, the agency adopted new
rules for glider kits, glider vehicles and glider engines. Starting in model year 2018, a
manufacturer could continue to sell glider vehicles, without limit, if the glider engine was from a
2010 or later model year. If a manufacturer 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.
Since the Phase 2 regulation set the sales cap at the lesser of manufacturers' 2014 sales volumes
or 300 vehicles per year, it is not straightforward to estimate the glider population in 2018
without individual manufacturer information from previous years. The regulation required
manufacturers to report their 2014 sales data to EPA to identify their individual sales allowances.
Prior to the 2018 model year, more than 260 glider manufacturers reported their sales data,
including 5 manufacturers that produced more than 300 gliders and who's sales would be capped
starting in 2018. Assuming glider production stabilizes starting with model year 2018 at the
limits set by the regulation, we apply a value of 4000 for model years 2018 and later in
MOVES3, consistent with the updated manufacturer-submitted data summarized in Figure 5-2.
Summary of Glider Manufacturer Sales Volumes for
« Model Year 2013 and Later
(L
7000
iS < = 10 10-50 51-300 > 300
-O
Sales Volumes Per Manufacturer Reported to EPA {Units Sold)
Figure 5-2 Summary of manufacturer-submitted glider sales volumes
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 (sourcetype 61) and long-haul combination trucks
(sourcetype 62).
/(Stmy)regciass 49 model year i,
_ Glidersi Equation 5-3
Combination Tvuckssoucetype 61+62,i
MOVES3 only models CNG short-haul combination trucks in regulatory class 47. As with the
other heavy-duty vehicles, MOVES does not model electric or E-85 combination trucks.
40
-------
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.10'11 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 30+ years, so
that all vehicles 30 years and older are modeled together. Therefore, an age distribution is
comprised of 31 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
MOVES3 only contains registration-based age distributions for two analysis years: 1990 and
2014. The age distributions for all other analysis years in MOVES3 were projected forwards or
backwards from the 2014 base age distribution. All default age distributions are available in the
SourceTypeAgeDistribution table in MOVES database.
Please note that the 1990 age distributions in MOVES3 have not been updated in this model
release. Please refer to Appendix B for more information.
6.1.1. Base Age Distributions
The 2014 base age distributions for cars and trucks were primarily derived from the 2014 IHS
data set and the 2014 National Transit Database (NTD). The 2014 IHS data set 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 5.1.4.
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 2014 IHS
data set and 2014 NTD did not capture all vehicles sold in 2014. Vehicle sales by source type in
2014 were calculated from a variety of sources as described in Appendix C.2. The source type
sales were divided by the 2014 source type populations (see Section 4.1) to determine the age 0
fractions. The other fractions for ages 1-30 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
41
-------
the age distributions (i.e., using the sales values as age 0 counts) because the IHS data set is only
used in MOVES to determine vehicle distributions, not for vehicle populations.
Figure 6-1 shows the fraction of vehicles by age and source type for calendar year 2014, which
formed the basis for forecasting and back-casting age distributions as described in the following
sections. Please note that since all vehicles age 30 and older are grouped together, there is an
uptick in this age bin for most source types.
8%
r1
oq'
"
L\
I
D
c
* J
(
10 20 30
Source Type
—1 Motorcycle
-- Passenger Car
Passenger Truck
— Light Commercial Truck
V /^\
Buses
a
S3
Li.
0) 12%-j
bO
<
8% -
4% -
0%-
10
20
30
A
C/J
3"
o
\a
1
C
•H
-t
c
o
7T
cw
y
Source Type
-- Intercity Bus
Transit Bus
— School Bus
Source Type
— Refuse Truck
Single Unit (Short- & Long-haul)
-- Motor Home
10
20
30
O
o
3
4
3
§
1 A
A
o
a
H
C
o
7T
-
c*a
10 2
0 30
Source Type
-- Combination Short-haul Truck
-- Combination Long-haul Truck
Age
Figure 6-1 2014 age distributions by source type in MOVES3
42
-------
6.1.2. Historic Age Distributions
The 1999-2013 age distributions were backcast from the 2014 base age distribution using
historic population and sales estimates. Age distributions are calculated from population counts,
if the populations are known by age:
f —El Equation
^ ~ Py 6-1
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:
Equation
Py-1 — Py Ny + Ry-i ^
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-i 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 29 and 30+ fractions with the base year age 29 and 30+ fractions
and renormalize the new age distribution to sum to 1 while retaining the original age 29
and 30+ fractions.
7. This results in the previous year age distribution (/y_i). If this algorithm is to be
repeated, fy-1 becomes fy for the next iteration.
The fraction of age 30+ vehicles is kept constant because most source types have a sizeable
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. However, lacking better data, we decided to keep the age 30+ bin at a constant
fraction for all historic age distributions.
43
-------
Age 29 is additionally retained because when the number of scrapped vehicles are calculated, a
large proportion of them come from the age 30 bin. In reality, these scrapped vehicles have a
distribution well beyond age 30, 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 29 bin. To prevent this from happening, the base year age 29 fractions are
also retained in each backcasted year.
Please see Appendix C, Detailed Derivation of 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 2015-2060 age distributions from the 2014 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+1 — Py Ry + Ny+1
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 30+ fraction with the base year age 30+ fraction and renormalize the
new age distribution to sum to 1 while retaining the original age 0 and age 30+ 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 30+ 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 29 vehicles when projecting forward. Therefore, the age 29 bin is calculated as
the others are instead of being retained from the base age distribution.
Please see Appendix C, Detailed Derivation of Age Distributions
In addition to producing the default projected age distributions, this algorithm was implemented
in the Age Distribution Projection Tool for MOVES2014.33 We anticipate developing a similar
tool for future versions of MOVES. This tool can be used to project future local age
Equation
6-3
44
-------
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.
6.2.Relative Mileage Accumulation Rate
For emission calculations, MOVES needs to estimate the miles travelled by each age and source
type. MOVES uses a relative mileage accumulation rate (RMAR) in combination with source
type populations (see Section 4) and age distributions described in Section 6.1 to distribute the
total annual miles driven by each HPMS vehicle type (see Section 3) to each source type and age
group. Using this approach, the vehicle population and the total annual vehicle miles traveled
(VMT) can vary from calendar year to calendar year, but the proportional travel by an individual
vehicle of each age will not vary.
The RMAR is determined from the mileage accumulation rate (MAR) within each HPMS
vehicle classification such that the annual mileage accumulation for a single vehicle of each age
of a source type is relative to the mileage accumulation of all of the source types and ages within
the HPMS vehicle classification. For example, passenger cars, passenger trucks and light
commercial trucks are all within the same HPMS vehicle classification (Light-duty vehicles,
HPMSVTypelD 25). As described below in Section 6.2.2, new (age 0) passenger trucks and
light commercial trucks are defined to have a RMAR of one (1.0)> and new passenger cars have a
RMAR of 0.885. This means that when tMOVES 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 88.5 percent of the annual VMT assigned to a passenger truck
or light commercial truck of age 0. The RMAR values used in MOVES3 are shown in Figure
6-2.
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.
45
-------
age
Figure 6-2. Relative Mileage Accumlation Rates (RMAR) by HPMS Class and SourceTypelD
The deivati on of the RMAR values for each sourcetype and HPMS class are discussed in the
following subsections. The RMAR values for heavy-duty vehicles in MOVE:S3 have been
updated from MOVES2014 as described below. The RMAR values of light-duty vechicles for
MOVES3 are not changed from MOVES2014.
46
-------
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.110
6.2.2. Passenger Cars, Passenger Trucks ami Light-Commercial
Trucks
The RMAR values for passenger cars, passenger trucks and light commercial trucks
(sourceTypelD 21, 31 & 32) were taken from a NHTSA report on survivability and mileage
schedules.111 In the NHTSA analysis, annual mileage by age was determined for cars and for
trucks using data from the 2001 National Household Travel Survey. In this NHTSA analysis,
vehicles that were less than one-year old at the time of the survey were classified as "age 1", etc.
NHTSA used a simple cubic regression to smooth the VMT by age estimates. We used NHTSA's
regression coefficients to extrapolate mileage to ages 26 through 30 not covered by the report.
Passenger cars, passenger trucks and light commercial trucks are grouped together as light-duty
vehicles (HPMSVTypelD 25). The NHTSA data for light-duty trucks were used for both the
passenger truck and commercial truck source types. Since the trucks had a higher MAR than
passenger cars, each source type's mileage by age was divided by truck mileage at age 1 to
determine a relative MAR. For consistency with MOVES age categories, we then shifted the
RMARs such that the NHTSA age 1 ratio was used for MOVES age 0, etc. Analysis of the data
determined that new passenger cars (age 0) accumulate only 88.5 percent of the annual miles
accumulated by new light-duty trucks.
We conducted a preliminary analysis of the impact of updating the MARs based on results from
the 2009 National Household Travel Survey. While the 2009 values may not fully represent
current trends in vehicle usage due to the economic downturn in that year, the use of 2009 values
resulted in changes to the MOVES allocation of VMT by one percent or less for each of the
vehicle categories covered by the survey. Consequently, we feel that the MARs developed from
the 2001 survey are still reasonable for use in MOVES3. However, this is an area where
additional data collection and analysis would be useful.
Table 6-1 shows the original raw data values from the 2001 NHTSA survey. The regression
values provide a "smooth" curve for annual mileage by age that avoids anomalous values, such
as the average mileage accumulation by 29 year old trucks, that are likely the result of very small
sample sizes.
47
-------
Table 6-1 NHTSA Vehicle Miles Traveled from 2001 National Househo
Annual Vehicle Miles Traveled
Vehicle Age
Passenger Cars
Light Trucks
1
14,417
15,806
2
13,803
15,683
3
13,692
15,859
4
13,415
15,302
5
13,183
14,762
6
12,301
13,836
7
12,253
13,542
8
11,709
13,615
9
11,893
12,875
10
11,855
12,203
11
10,620
11,501
12
9,986
10,815
13
10,248
11,391
14
9,515
10,843
15
9,168
10,378
16
8,636
9,259
17
8,941
8,358
18
7,267
9,371
19
8,890
7,352
20
8,759
8,363
21
6,878
6,999
22
7,242
7,327
23
6,350
6,969
24
5,745
6,220
25
4,130
6,312
26
6,745
27
9,515
28
6,635
29
12,108
30
5,067
31
4,577
32
6,923
d Travel Survey
6.2.3. Buses
The transit bus (sourceTypelD 42) annual mileage accumulation rate 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.34 The MOBILE6 results were
extended to calculate values for ages 26 through 30.
48
-------
The definition of sourceTypelD 41 has changed (see Section 5.1.4) from MOVES2014. In
MOVES2014, this source type was defined as an "intercity bus" with a constant RMAR. For
MOVES3, we have redefined source type 41 as "other bus" (sourceTypelD 41) and assigned the
same RMAR as the transit bus (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 Book1^. In MOVES3, we updated the RMAR for
school buses to be based on the transit bus RMAR, adjusted down such that year 0 is based on
the 9,939 miles per year from the School Bus Fleet Fact Book. The same relatie shape is evident
in of the Bus RMAR in Figure 6-2
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).35 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.
49
-------
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 per vehicle and no trend. After that, the average annual miles per vehicle 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 to age 30. The relative mileage
accumulation rate at age 30 were derived from the 1992 Truck Inventory and Use Survey (TIUS)
as documented in the ARCADIS report.36
Mileage accumulation rates for these vehicles were determined for each age from 0 to 30 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 1992 TIUS relative mileage accumulation rate for age 30. 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
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.
50
-------
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 0-3a
30,437
23,250
37,069
61,240
116,591
Intercept13
30,437
23,250
37,069
61,240
116,591
Slopeb
-1,361
-1,368
-2,476
-4,092
-6,418
Age 30 RMAR
0.027
0.0115
0.086
0.015
0.052
Notes:
a Average sample annual miles traveled for ages 0 through 3.
b Intercept at age 3; slope from ages 4 through 16.
The RMAR values for heavy-duty were updated in MOVES3. MOVES2014 included minor
miscalculation that used inconsistent baseline mileages for heavy-duty RMAR rates, and they
were fixed in MOVES3. The updated resulting relative mileage accumulation rates are shown in
Table 6-5 below and Figure 6-2 above. As in previous versions of MOVES, the first four ages
(age 0 to 3) are identical and then decline linearly to age 16 and then linearly to age 30 with a
different slope.
6.2.5. Motor Homes
In MOVES2014, the RMAR for motor homes (sourceTypelD 54) was a constant value based on
a year 2000 owner survey.37 For MOVES3, we have updated the RMAR values and added a
decreasing trend with age. Data from the 2017 National Household Travel Survey38 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. 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 6-2 above. Note that first four ages are identical and then decline linearly to age 30
since the 2017 National Household Travel Survey has data available from age 0 to 30.
51
-------
Table 6-5 Relative mileage accumulation rates for heavy-duty trucks in MOVES3
agelD
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
52
-------
7. VY1T 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 SHOAllocFactor found in the ZoneRoadType table.
The national default distribution of VMT to source type for each road type in MOVES3 were
derived to reflect the VMT data included in the 2017 National Emission Inventory (NEI) Version
2.39 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 statistics40 when
state supplied estimates are not available. The FHWA road types mapped to the MOVES road
type ID 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 IV
OVES road types
FHWA Road Type
MOVES
Road Type ID
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.
53
-------
Table 7-2 MOV
S3 road type distribution by source type
Road Type3
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
Note:
a RoadTypelD = 1 (Off Network) is assigned no VMT.
54
-------
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 s
peed 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 updated in
MOVES3 using the telematics data.
8.1. Description of Telematics Dataset
In a study done by the Coordinating Research Council (CRC A-100)41, the GPS data collected by
StreetLight Data was 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 distance and time
55
-------
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 data 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 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 analyses42, 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 1/16th 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.41
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 platform43. The following
section describes the procedure to generate the average speed distributions included in
MOVES3.
56
-------
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 (z) for each source type (ST) - road type (RT) combination in each
county (Co) and dividing by the corresponding annual average speed:
. , „„„ lli=fuel Annual VMTSTi RT i Co
Annual SHOSTRTCo =
Annual Average SpeedST RT Co
miles
miles/hour.
Equation
8-1
Then, we aggregate over all counties z 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
i=co ' 8-2
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:
I
ASDhdST RT —
Equation
AverageSpeedFractioni h d STRT Co x Annual SHOSTRTCo 3.3
i=16 National Annual SHOSTRT
57
-------
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 used in MOVES3, we used the same
mapping of telematics data to MOVES source type used in the NEI to maintain consistency. For
buses, refuse trucks, 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
MOVES Source Type ID
MOVES 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
Intercity 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 and comparison with
MOYES2014
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.
58
-------
HD commercial
MD commercial
Personal
c
o
o
Cfl
u_
"O
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 type see Table 8-2.
A comparison between the average speed distributions in MOVES3 and those included in
MOVES2014 for several source types is shown in
59
-------
MOVES2014
0.3-
0.2-
0.1 -
0.0-
0.3-
g
tj 0.2-
T3
0)
0)
Q.
to
CI)
>
TO
0.1 -
o.o-
0.3-
0.2-
0.1 -
0.0^
20
40
60
MOVES3
-Weekday
-Weekend
80 0
20
40
60
80
Average Speed (mph)
Figure 8-2, for urban restricted roads at 5pm. For weekdays, the major differences are seen for
Passenger Cars and Combination Short-haul Trucks, whereas Light Commercial Trucks remain
with similar distributions. For weekends, we see differences for the three example source types
where in all cases the new profiles assign more time at speeds between 60 and 70 mph for Light-
Commercial Trucks and Combination Short-haul Trucks (i.e. any source type mapped to the
telematics Medium-Duty Commerical or Heavy-Duty Commercial distributions) and more time
above 60 mph for Passenger Cars (i.e. any source type mapped to the telematics Personal
distribution).
60
-------
0.3-
0.2-
0.1 -
0.0
0.3
MOVES2014
o
o 0.2
"O
a)
(D
ro
0.0
0.3
0.2-
0.1
0.0-I
MOVES3
-Weekday
-Weekend
20
40
40
60
80
60 80 0 20
Average Speed (mph)
Figure 8-2 MOVES2014 and MOVES3 average speed distributions for 5pm on urban restricted roads for
Passenger Cars, Light Commercial trucks and Combination Short-haul Trucks.
The MOVES2014 data also came from telematics sources (TomTom GPS data), however, that
dataset was based on data largely from light-duty vehicles. As can been seen in the graph above,
the MOVES2014 speed distributions for combination truck on restricted access roads were based
on this light-duty data, but were adjusted to have an eight percent lower average speed.44 The
new StreetLight data improves on these 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.
61
-------
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 MOVES3 are unchanged from 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
Drive Schedule Assoc 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
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.
k However, as described in Section 10, recent data suggests that drive schedules miss a substantial fraction of real-
world idling. MOVES3 has been updated to better account for the idling that was not captured in previous versions
of the model.
62
-------
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
Table 9-2 MOVES driving cycles for other buses (41
ID
Cycle Name
Average
Speed
Unrestricted access
Restricted access
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
MD 40mph 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
63
-------
ID
398
401
404
201
405
202
402
203
204
403
205
206
251
252
253
254
255
397
ID
398
501
301
302
303
304
305
306
351
352
353
354
355
396
Table 9-3 MOVES driving cycles for transit and school buses (42, 43)
Cycle Name
Average
Speed
Unrestricted access
Rural
Urban
Restricted access
Rural
CRC E55 HHDDT Creep
U
X
X
X
Bus Low Speed Urban
3.1
X
X
New York City Bus
3.7
X
X
MD 5mph Non-Freeway
4.6
X
WMATA Transit Bus
3.3
X
X
MD lOmph Non-Freeway
10.7
X
Bus 12mph Non-Freeway
11.5
X
X
MD 15mph Non-Freeway
15.6
X
MD 20mph Non-Freeway
20i
X
Bus 30mph Non-Freeway
21.9
X
X
MD 25mph Non-Freeway
24.5
X
MD 30mph Non-Freeway
31.5
X
MD 30mph Freeway
34.4
X
MD 40mph Freeway
44.5
X
MD 50mph Freeway
55.4
X
X
X
MD 60mph Freeway
60.4
X
X
X
MD High Speed Freeway
72.8
X
X
X
MD High Speed Freeway Plus 5mph
Hi
X
X
X
Table 9-4 MOVES c
riving cycles for refuse trucks (51)
Cycle Name
Average
Speed
Unrestricted access
Restricted access
Rural
Urban
Rural
CRC E55 HHDDT Creep
U
X
Refuse Truck Urban
2.2
X
X
HD 5mph Non-Freeway
5i
X
HD lOmph Non-Freeway
11.2
X
X
X
HD 15mph Non-Freeway
15.6
X
X
X
HD 20mph Non-Freeway
19.4
X
X
X
HD 25mph Non-Freeway
25.6
X
X
X
HD 30mph Non-Freeway
32.5
X
X
X
HD 30mph Freeway
34.3
X
X
X
HD 40mph Freeway
47.1
X
X
X
HD 50mph Freeway
54.2
X
X
X
HD 60mph Freeway
59.4
X
X
X
HD High Speed Freeway
71.7
X
X
X
HD High Speed Freeway Plus 5mph
77.S
X
X
X
64
-------
Table 9-5 MOVES driving cycles for single-unit trucks and motor homes (52, 53, 54)
ID
Cycle Name
Average
Speed
Unrestricted access
Restricted access
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
MD 40mph 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 9-6 MOVES driving cycles for combination trucks
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
61, 62)
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.001.45 "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,46 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 2004.110 These cycles were selected to best
cover the range of road types and average speeds modeled in MOVES.
65
-------
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"47 and "WMATA Transit Bus"48 drive schedules are included for urban driving that includes
transit-type bus driving behavior. The "CRC E55 HHDDT Creep" 49 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 Arbor Transit Authority buses instrumented in Ann Arbor, Michigan.50 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, we revised the handling of bus speeds. In MOVES2014, the derived bus cycles
401, 402 and 403, were associated with the average speed of 15, 30 and 45 mph, respectively,
even though the actual average speed of the cycles were 3.1, 11.5 and 21.9 mph, respectively.
This was done assuming that the input average speed for buses on unrestricted access roadways
was based on the traffic speed, while the actual speed was lower due to bus stops. In MOVES3,
we changed the driving cycle mapping in the DriveSchedule table to be the actual speed in
MOVES3 for all bus drive cycles. Consistent with our changes, users should input the actual
average speed distribution for transit buses, rather than the traffic speed.
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.51 The drive cycle data was segregated into restricted access and
unrestricted access driving for medium- and heavy-duty vehicles and then further stratified
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
66
-------
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)52 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
percentage of activity in the highest power, high speed operating mode bins.11 This may be
reasonable because the manufactur-run testing for the HDIUT data set 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.
9.2. Modeling of Ramps in MON KS
For MOVES3, we simplified the modeling of emissions on restricted access roadways by
removing the option to explicitly model emissions from ramp road types at the national and
county-scale. Based on an analysis of instrumented real-world vehicles operating on highways
with a variety of ramp configurations, we determined that the added complexity of modeling
ramps separately from restricted access highways was not justified for county and national scale
runs. The ramp fraction field that existed in prior versions of MOVES has been removed.
Modeling ramps as part of highway driving reduced mobile-source emissions inventories by less
than 3 percent for NOx and less than 1 percent for HC, CO and Primary PM2.5 exhaust.
Brakewear particulate was reduced by less than 9 percent. For more details on this analysis, see
Appendix G, Freeway Ramp Contribution at the County-Scale
In addition to reducing run time and complexity, this approach eliminates the need for users to
estimate the ramp fraction of highway driving and removes the need for MOVES to extrapolate
from limited data default operating mode distributions for ramps for each vehicle source type.
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.
However, 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 emissions53 and
67
-------
brake emissions. Users can continue to estimate ramps as individual links in project-scale.
Preferably, project-level users can characterize the operating mode or driving cycle of the ramps
they are evaluating.
68
-------
10. Off-Network Idle Activity
With the exception of extended idle emissions from combination long-haul trucks (see Section
11), all vehicle running emissions in MOVES2014 are assigned to the four "real" road types;
vehicle idle emissions occur only during the driving schedules and vary by average speed by
road type. However, recent data has shown that MOVES driving schedules substantially under-
predict the amount of idle time that occurs during vehicle trips. To put this into perspective, the
percentage of operating time spent idling (total idle fraction) in MOVES2014 (national default)
is around 14 percent for sourceTypelDs 21 and 31, compared to 18-31 percent as derived from
Verizon Telematics data described below. The difference is partially due to drive cycle
development approaches that intentionally excluded activity in drive-ways, parking lots, queues
and during delivery operations. In addition, the driving schedules in MOVES2014 may not have
accounted for the increased amounts of congestion in recent years. Telematics data can capture
these idle times.
To better account for observed levels of idling, we have added a new emission calculation to
MOVES3 for County and National Scale runs1 allowing the model to estimate idle emissions that
occur off the road network (i.e., on roadTypeID=l) for all soucetypes. This section summarizes
the new calculation methodology employed by MOVES3 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. MOVES3 defines "idle" as any seconds in the driving schedules where the speed is less
than one mile per hour (opModeID=l) during engine operation. Using the fraction of vehicle
operation hours that are opModeID=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 new 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. In MOVES3, the additional ONI hours are assigned to the running exhaust process
(processID=l) for the off-network road type (roadTypeID=l).
1 In Project Scale, MOVES3 does not adjust activity to account for off-network idling. Instead, the user can provide
location-specific idling activity as appropriate.
69
-------
In MOVES2014, total SHO is calculated from vehicle miles traveled (VMT) and average speed
for all onroad roadTypelDs 2, 3, 4 and 5. In MOVES3, we are renaming 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:
SHO = (V SHOi) + ONI
^—>i=2
Where i = roadTypelD
Equation 10-1
Source hours idle (SHI) then is the total hours of idle, excluding diesel long-haul combination
truck hotelling idle:
SHI = ( > SHI, 1 + ONI Equation 10-2
,5
n=2
Where i = roadTypelD
= (Y SHIt) +
^—>i=2
All running exhaust activity for roadTypeID=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:
(ld=2SHhJ+ ONI Equation 10-4
~ ffi=2SHOi)+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,
MOVES3 can calculate the hours for off-network idle (ONI):
Si=2 SH°i ) X TIF - Ei=2 SHI, Equation 10-5
(1 - TIF)
Where i = roadTypelD
As an example, the default values of TIF for light-duty vehicles in idleRegionID=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.
70
-------
Off-network idle emissions are calculated for each hour by using the corresponding emission rate
(grams per hour) for opModeID=l for that hour. All of the adjustments (e.g., fuel effects, air
condition effects) made to the emission rates for opModeID=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 (processID=l) for road type "off-
network" (roadTypeID=l).
10.2. Light-Duty Off-Network Idle
10.2.1. Verizon Telematics Data
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 was collected August 2015 through August 2016 using on-board
diagnostic data loggers under contracts with State Farm insurance, Mercedes-Benz and
Volkswagen. The data set 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.54 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 was
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 (I/M) program or not.
71
-------
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.
Ta
)le 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%
Notes:
* 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.
72
-------
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
VI
pi
5000
u
2
>
4000
Ct-x
O
N
Q*>
3000
—
=
s
2000
1000
0
J
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 MOVES3 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 commercial trucks
73
-------
(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 110 off-network idle.
Table 10-2: 2014 Vehicle populations of the idle regions80.
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
o
Vancouver
a
California
Colorado
Ottawa Montrd
o o A
ToronH
2-1 ii :atjc« Detroit,
Illinois
¦boston
New Jersey
Rphfladelphia
*tngton
Georgia
MA
0
MEXICO
Guadalajara
I Mi am i
Ha-; ana
< Ul!A
Esn. HERE. Ga... •
Figure 10-2: Default Regions for Weighting Light-Duty Activity111
In addtion to region, the Verizon Telematics data analysis suggested that the following factors
are important when estimating total idling fraction:
m Note, Alaska is associated with Colorado. Hawaii, Puerto Rico and the Virgin Islands are associated with
California.
74
-------
• 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 Equation
+ monthIDm + idleRegionIDl x monthIDm + n 10-6
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 MOVES3, 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., countyTypeID=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 idleRegionID=101 (New Jersey).
The model fit TIF values apply to all calendar years in MOVES3. 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).
75
-------
—TIF (Modeled) - 0 -
— • —TIF (Actual)-0 -
TvpclD=2l, idleRej>ionlD=101
—TIF (Modeled) - 1 0 represents countyTypeID=0
-TIF (Actual) - 1 1 rePresents countvTypeID=l
sourceType!D=31, idleRej{ionID= 101
2 3 4 5 6 7 8 9 1011 12| I 2 3 4 5 6 7 8 9 10 li 12
monlblD/daylD
sourccTypel D=21. idleRegionl I)=103
0.34
0J2
0.3
B 0-28
S
| 0.26
c
£ 0.24
i 0-22
|
0.2
0.18
0.16
sourceTypeID=3l, id!eRegionID=103
•= 0.22
1 234567 89 1011 12 1 2 3 4 5 6 7 8 9 1011 12
2 | S
inonthlD/ilaylD
sourccTypel I>=21. id It-Region ID-102
sourccTypelD=21. idlcRcgionID=!05 sourceTypelD=31, idleRegionlD=105
sourceTypelD=31, idleRegioul D= 102
sourccTypel R=31, id leRcgionl I)= 104
Figure 10-3: TIF model results compared to the values from the Verizon Telematics data
76
-------
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. Recently, the
University of California Riverside's Bourns College of Engineering Center for Environmental
Research and Technology (CE-CERT) concluded their data collection for a study to evaluate the
selective catalytic reduction (SCR) behavior of heavy-duty vehicles. We plan 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 database55 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 network56) 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 set 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
77
-------
idle emission rates, due to differences in congestion, topography and regional policies11,57.
However, as presented in the NREL project report58, 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
sourceTypelD
Source Type Name
Number of Vehicles
in Fleet DNA
Number of States
with Recorded
Activity
41
Other Buses (non-school, non-transit)
0
0
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
CO
Note: The number of trucks operating in each US state is listed in the NREL project report
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.58
11 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
78
-------
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
79
-------
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 intend 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 was 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-average activity.0
° 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 1-3 (Method 3 "normalized sum over sum") in Appendix I.
Appendix I includes an overview of each approach and a comparison between calculation approaches.
80
-------
Equation 10-7 shows the calculation of the total idle fraction for each source type and specific
day type (weekday or weekend).
Y /idle hourSi / \
Idle fractions a = ^~r- z
' Y /operating hoursi / \
V ! day Si) Equation 10-7
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 expect to have more of the
source types and day ID'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.
81
-------
Idle Fraction, Weekends
¦ Off-Network
a Extended Idle
Vehicle Description I 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
¦ Off-Network
a Extended Idle
Vehicle Description / MOVES SourceType
Figure 10-5 Weekday idle fractions for heavy-duty vehicle sourceTypes based on data from
NREL's Fleet DNA database
82
-------
Table 10-5 Idle fraction values for heavy-duty sourceTypes based on data from NREL's Fleet DNA
database
SourceType
Vehicle Description
Weekend It
le Fractions
Weekday It
le 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.
83
-------
daylD: 2
0.5
0.4 i
0.3
0.2
c 0-1
o
S0.0H
0.5
0.41
0.3
0.2
0.1
0.01
daylD: 5
T3
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 estimated in MOVES for an Urban
County in the Midwestern Region using MOVES3.
ll.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 of the long-haul combination trucks are diesel-fueled. All
source use types other than long-haul combination trucks have hotelling activity fractions set to
zero.
84
-------
11.1. Hotellliig 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 11-1 Hotelling activity operating modes in MOVES
OpModelD
Description
200
Extended Idling of Main Engine
201
Hotelling Diesel Auxiliary Power Unit (APU)
203
Hotelling Battery or AC (plug in)
204
Hotelling All Engines and Accessories Off
Previously, MOVES assumed drivers required power for the entire duration of hotelling, which
was supplied by either 100 percent idling (prior to 2010) or a combination of idling and APU use
(for 2010+). For MOVES3, we updated the model's hotelling activity distributions to be
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.59
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.60 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. For MOVES3,
we assumed the drivers did 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.
The HotellingActivityDistribution, 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
updated the 100 percent extended idling assumption to account for the 20 percent of time we
assumed drivers would 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 Standards61 and a fraction of the time that previously
was assigned to extended idle is now assigned to opModelD 201 (the use of APUs). In model
years 2021, 2024 and 2027, 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 adoption of 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
2.62 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.
85
-------
Table 11-2 Default hotelling activity distributions
beginModelYearlD
endModelYearlD
opModeFraction for given opModelD
200
201
203
204
Idle
APU
Electric
Off
1960
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
0.08
0.20
2024
2026
0.40
0.32
0.08
0.20
2027
2060
0.36
0.32
0.12
0.20
Based on peer-review comments on the above analysis 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 consistent with two fleet surveys:
NACFE 2018 Annual Fleet Fuel Study63'p and Shoettle et al. (2016)64'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 MOVES 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.
p NACFE (2018) reported increasing diesel APU and and battery penetration rates 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 considered
representative of the entire fleet.
q Shoettle et al. reports that 3 8.7 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.
86
-------
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 = - —- Equation 11-1
Total Restricted Miles Traveled
Where: Total Restricted Miles Traveled is the total miles traveled by diesel long-haul
combination trucks on rural and urban restricted access roads (freeways) in MOVES.
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 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).
In MOVES2014, the national default hotelling rate was based on hours-of-service regulations
from the Federal Motor Carrier Safety Administration (FMCSA).65 For every 10 hours of
driving, MOVES2014 assumed that the trucks spent 8 hours in hotelling activity. This was
believed to be a conservatively high estimate for at least a couple reasons, including: 1):
hotelling is not require at trip ends, including trips less than 10 hours and 2) team drivers can
switch drivers rather than hotelling the truck. In addition, MOVES2014 used only the VMT on
rural restricted roads as the surrogate for allocating total hotelling hours.
MOVES3 expands the hotelling VMT to include urban restricted roads to capture the truck
traffic around cities. Also, for MOVES3, we updated the national default hotelling rate based on
data collected and analyzed by the National Renewable Energy Laboratory (NREL) Fleet DNA58
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
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.
87
-------
Hotelling Rate =
Extended Idle Hours
¦)G
Total Miles Traveled
j) G
Hotelling Hours
Total Miles TraveledJ \Restricted Access Miles TraveledJ \Extended Idle Hours)
3.45 w 1
= ,3A5_-), 1 wj_i
VlOOO/ \0.34 + 0.26/ Vo.8/
= (111) (J.) (A.)
VlOOO/V0.6/V0.8/
7.2 Hotelling Hours
1000 Restriced Access Miles Traveled
Equation
11-2
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 266 and two other studies. Lutsey et
al.67 presented data from a nationwide truck surveyr and NCHRP 08-10168 conducted an analysis
of an instrumented truck dataset with 300 truckss.
20.00 -
15.00 -
U)
S
=5
O
H
5
m
27.3
=
¦
11.0
1
5
1
7.2
I
6.3
¦
0.00
MOVES2014 MOVES3 Lutsey (UC Davis) et NCHRP 08-101
(2014 NEI v2) (NREL Fleet DNA) al. (2004)
Figure 11-1. Hotelling hours per 1000 miles driven on freeways compared across different datasets.
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.
88
-------
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
89
-------
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-
duty10 and heavy-duty11 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.
MOVES3 uses the following set of tables in the default database to determine the default number
of starts, soak times and their temporal distributions:
• StartsPerDayPerVehicle
• Starts Age Adjustment
• StartsHourFraction
• StartsMonth Adjust
• 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)
In MOVES2014, starts varied only by source type and day type. In MOVES3, starts also vary by
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 30 are lower
than the fleet-average starts per day.
MOVES3 accounts for the effect of age using the ageAjustment factors stored in the
Starts Age Adjustment 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
MOVES3 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 absolute values in
90
-------
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 month Adjustment is used as a raw multiplicative factor,
with values greater and less than one. Unlike the startsageadjustment table, MOVES does not
scale the month Adjustment 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.
For the purpose of estimating vehicle start activity, the data described here fully replace the data
in the SampleVehicleTrip table used in MOVES2014, except for a few noted instances (Section
12.3). However, as discussed in Section 13.4, the MOVES2014 SampleVehicleTrip table is still
used in MOVES3 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. While we think the impact of these inconsistencies is small,
we plan to address this conflict in future versions of MOVES.
12.1. Light-Duty Start Activity
For MOVES3, light-duty start activity are calculated from the same sample of vehicles from the
Verizon Telematics data discussed in Section 10.2.1.
91
-------
12.1.1. Starts Per Bay Per Vehicle
The vehicle starts input format has been substantially updated for MOVES3 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.
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
Source Type
Source-
TypelD
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
92
-------
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.
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 MOVES3 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-lFigure 12-1, the age in MOVES 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
MOVES 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.
93
-------
National Average Starts per Vehicle per Day
30
= ^ (Starts per Day Per Vehicle)age X ageFractionage Equation
age=0 12-2
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
MOVES3. 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.
94
-------
Table 12-2. Calculation of the National Average Starts per Vehicle per Day for Passenger Cars
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
95
-------
12.1.2. Temporal Distributions
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 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 new 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.
MOVES 3 MOVE S3 - - MOVES2014 - - MOVES2014
Weekday Weekend Weekday Weekend
0.12
0.1
0.08
|
| 0.06
la
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
Hour ED
Figure 12-2 Start distribution for source type 21: MOVES3 derived from Verizon data vs.
MOVES2014
12.1.2.2. Monthly Distribution
For MOVES3, 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
96
-------
table discussed in Section 13.1. Light-duty vehicles and all other source types (except
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.
Figure 12-3 shows the MOVES2014 defaults for engine soak time distribution for a weekday for
passenger cars (sourceTypelD 21) using trip information from a set of instrumented vehicles.
The new MOVES3 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 12-4 illustrates the MOVES3 national default soak distribution for a weekday
for passenger cars (sourceTypelD 21). The new soak distribution is similar to the data used in
MOVES2014, but much smoother given the much larger dataset.
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
97
-------
¦ 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
HourED
Figure 12-3 MOVES2014 default engine soak time distribution for source type 21 and weekday
(dayID=5)
98
-------
¦ 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 12-4 MOVES3 national average engine soak distribution for source type 21 and weekday
MOVES3 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 set 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. In MOVES2014,
start activity for heavy-duty vehicles was derived from a small sample of instrumented heavy-
duty trucks and extrapolated to different source types using assumptions about numbers of starts
per day. For MOVES3, data that covers a wider range of heavy-duty vocations was available.
99
-------
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 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.1
Y (startSi , \
Starts Per Daysd= —
n
Equation 12-3
i = Vehicle ID within a given sourceType, s
daysi = days within a given daylD, d, when vehicle, is instrumented
n = number ofVehiclelDs withing a given sourceType, s
1 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.
100
-------
rstartshi . \
^V ' daysj
Start fractionhsd = ,starts \~
^ V ' daysj
Equation 12-4
h = hour of the day
i = Vehicle ID within a given sourceType, s
daysi = 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.
soak time = engine stop time — engine start time Equation 12-5
Every start was assigned a soak opModelD based on the definitions in Table 12-3" We then
calculated the average soak fraction, using a normalized sum-over-sum approach like we did for
the start fraction.1 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 day ID, hourlD and sourceTypelD.
yi fstartsh i o j \
^ ^ V /daysj
Soak fractwnh 0 Sid - ,starts v
^ V ' daysj
Equation 12-6
h = hour of the day
i = Vehicle ID within a given sourceType, s
o = operating mode/soak length
daysi = days within a given day ID, d, vehicle, is instrumented
u The first start identified for each vehicle was not considered when calculating soak time due to lack of a previous
recorded stop time.
101
-------
12.2.2.
Starts Per Vehicle Per Day
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. We expect data collected by CE-CERT to cover more of the
source types and daylDs when it becomes available, and we plan 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 day ID. 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 (dayID 2),
assuming the frequency of school bus starts differed between weekends and weekdays.
Figure 12-5 and Figure 12-6 show the starts-per-day values for weekends and weekdays,
respectively.
Daily Starts, Weekends
Vehicle Description / MOVES SourceType
Figure 12-5 Weekend starts per day for heavy-duty vehicle sourceTypes in MOVES3 based on data
from NREL's Fleet DNA database
102
-------
Daily Starts, Weekdays
Combo,
Long
62
Vehicle Description / MOVES SourceType
Figure 12-6 Weekday starts per day for heavy-duty vehicle sourceTypes in MOVES3 based on data
from NREL's Fleet DNA database
As shown in Figure 12-6, 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 12-7 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 MOVES3,
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.
103
-------
0
Q
"0
o
03
CL
Daily Starts by Vocation, Weekdays
Single-Unit Short-Haul Trucks
&
0
0
>
>
"0
"0
Q
Q
c
~o
0
0
c
0
—I
LL
CD
C
0
0
o
c
o
o
0
>
"0
Q
0
"O
"O
0
_c
CO
c
03
I-
0
C
03
Q_
O
o
Q_
E
Q
Vehicle Vocation
Figure 12-7 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 12-8 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.
104
-------
Single-Unit Trucks
Single-Unit Trucks
Single-Unit Trucks
Weekdays
- 1 Weekend
10 20
vehicle age
Figure 12-8. 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 12-8 and the 2014 heavy-duty default
age distributions in MOVES. 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
MOVES3 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. Table 12-4.
Table 12-4 National Average Starts Per Day Per Vehicle for Heavy-duty Vehicles based on data
from NE
tEL'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
105
-------
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
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 MOVES3 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 this report.
We expect data collected by CE-CERT to cover more of the source types and daylDs and when
it becomes available, we plan 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 12-9
shows the resulting starts distribution for these two bus types.
106
-------
Transit Buses (sourceType 42) & Other Buses (sourceType 41)
0.12
0.10
c 0.08
o
-4—'
£ 0.06
t
ro
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 12-9 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 12-10. The refuse trucks did not operate on
weekends from 7:00 PM to 4:00 AM and no data was 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
Ell
~ Weekends
¦ Weekdays
Mi
107
-------
School Bus, Weekdays
0.25
0.20
0.15
0.10
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
0.20
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 12-10 Approach for renormalizing the start fraction results to avoid zeros hours when no
data was 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 12-11 and Figure 12-12 show the resulting start fractions by hour for school buses and
refuse taicks, respectively. Note that, in MOVES3, 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.
108
-------
o
TO
TO
-t—'
CO
0.25 i
0.20 -
2 0.15 -
0.10 -
0.05 -
0.00
School Buses (sourceType 43)
ll n«
~ 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 12-11 Start fractions by hour for school buses (sourceType 43) based on data from NREL's
Fleet DNA database
CO
0.25 n
0.20 -
S 0.15 -
O
03
TO 0.10
0.05 -
0.00
a
Refuse Trucks (sourceType 51)
~ Weekends
¦ Weekdays
IliJjm iiiiiJiij
nl nB rl
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 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-unit short-haul taicks (sourceType 52).
109
-------
Figure 12-13 shows the start distribution applied to both single-unit truck types for weekends and
weekdays.
0.12
0.10
c 0.08
o
'•*—'
£ 0.06
t
ro
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 12-13 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 was 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 12-14 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 12-15 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)
~ Weekends
¦ Weekdays
r-|
n
r-i
p.
I
-
I
-
r~i
J
¦
n.
¦
I
1
1
1
1
¦
¦
110
-------
0.14 -
0.12 -
§ 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 12-14 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
0.12 -
0.10 -
c 0.08 -
o
¦4—'
£ 0.06 -
t
03
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 12-15 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
PI
¦ Weekdays
-
-1
p.
.1
1
1
¦
pi
i-i
p.
n
ill
nl
nl
ni
na
nl
Wi
1
1
1
1
1
Combination Trucks (sourceTypes 61 & 62)
~ Weekends
¦ Weekdays
111
-------
12.2.3.2.
Monthly Distribution
In MOVES3, 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 MOVES3. 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. We expect data collected by CE-CERT to cover more of the source types and
daylDs and when it becomes available, we plan 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.
Throughout the Fleet DNA dataset, some vehicles did not have every soak time OpMode
represented for each hour of the day. To avoid tables with zero OpMode values for those hours,
we replaced those zeros with a very small, nonzero value of 0.0001 and renormalized values for
that hour to sum to 1.0.
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 12-16
and Figure 12-17 show the resulting starts distributions for these two bus types on weekends and
weekdays, respectively.
112
-------
Transit & Other Buses (sourceTypes 42 & 41), Weekends
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 12-16 Weekend soak distributions transit buses (sourceType 42) and other buses
(sourceType 41) based on data from NREL's Fleet DNA database
Transit & Other Buses (sourceTypes 42 & 41), Weekdays
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 Weekday soak distributions transit buses (sourceType 42) and other buses
(sourceType 41) based on data from NREL's Fleet DNA 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 12-18 shows the soak distribution
applied to school buses for both weekends and weekdays.
113
-------
School Buses (sourceType 43), Weekdays & Weekends
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-18 Weekend and weekday soak distributions for school buses (sourceType 43) based on
data from NREL's Fleet DNA database
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 was 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 12-19 and Figure 12-20 show the soak distributions for refuse
trucks on weekends and weekdays, respectively.
Refuse Trucks (sourceType 51), Weekends
100%
90%
80%
c 70%
¦§ 60%
n)
it 50%
| 40%
W 30%
20%
10%
0%
Hour
Figure 12-19 Weekend soak distributions for refuse trucks (sourceType 51) based on data from
NREL's Fleet DNA database
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
114
-------
Refuse Trucks (sourceType 51), Weekdays
!¦!
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-20 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 12-21 and Figure 12-22 show the soak distributions applied to both single-unit truck types
for weekends and weekdays, respectively.
Single-Unit Trucks (sourceTypes 52 & 53), Weekends
100%
90%
80%
c
70%
o
o
60%
2
LL
50%
ol
o
40%
00
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 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
115
-------
Single-Unit Trucks (sourceTypes 52 & 53), Weekdays
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-22 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 DNA database
Figure 12-23 shows the weekend soak distribution that was applied in MOVES for combination
short-haul trucks (sourceType 61). Figure 12-24 shows the weekday soak distribution for the
same vehicles.
Combination Short-Haul Trucks (sourceType 61), Weekends
100%
90%
80%
c
70%
o
CJ
60%
ra
LL
50%
ro
o
40%
w
30%
20%
10%
0%
123456789
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-23 Weekend soak distributions for combination short-haul trucks (sourceType 61) based
on data from NREL's Fleet DNA database
116
-------
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 12-24 Weekday soak distributions for combination short-haul trucks (sourceType 61) based
on data from NREL's Fleet DNA database
As mentioned in the start di stribution 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 tmcks on
weekends are shown in Figure 12-25. The weekday soak distrbituions are in Figure 12-26.
Combination Long-Haul Trucks (sourceType 62), Weekends
100% |
90%
OpMode
80%
¦ 108
c
70%
¦ 107
o
o
60%
¦ 106
2
¦ 105
LL
50%
o
40%
¦ 104
CO
30%
¦ 103
20%
¦ 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 12-25 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
117
-------
Combination Long-Haul Trucks (sourceType 62), Weekdays
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 12-26 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, MOVES3 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.
118
-------
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 (SourceTypelD
11)
Motorhomes
(SourceTypelD 54)
startsPerDayPerV ehicle
Starts from Table 13-8
adjusted to represent CY
2014 age distribution
Table 13-8 adjusted to
represent CY 2014 age
distribution
startsHourF racti on
Passenger Cars (21)
Passenger Trucks (31)
startsOpmodeDistribution
(soaks)
Passenger Cars (21)
Passenger Trucks (31)
startsMonth Ad] ust
Table 13-2
Table 13-1
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 12-27.
O 1.5
JZ
>
>
TO
Q
£0.5
CL
CO
r
ro _ _
yj 'J.u
Motorcycle
0.6
0.4-
0.2-
Motor Home
dayName
Weekdays
Weekend
10
20
10
20
30
0.0^
30 0
vehicle age
Figure 12-27. 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 12-27Figure 12-8 and the 2014
heavy-duty default age distributions in MOVES. 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
119
-------
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 Motor
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
Homes
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 starts is
assumed to be the same as for passenger trucks , both of which are estimated from the Verizon
database.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.
120
-------
13.Temporal 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).69 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 for MOVES3.
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, for MOVES3, 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
MOVES3 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. EPA plans to update the temporal allocations currently in MOVES using more
recent data sources, such as telematics data, as they become available.
121
-------
13.1. VY1T Distribution by Mouth of tie 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 Month VMTFraction
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 2010.70 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.
122
-------
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 Type of Bay
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.69
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
123
-------
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 tie Bay
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.
124
-------
Table 13-4 MOV]
LS 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
125
-------
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. Recent analysis by CRC has made county specific hourly VMT distributions
available for calendar year 2014.42
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
126
-------
updating the activity data in these tables was beyond the scope of MOVES3.v 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.71 This change is described in greater
detail in the report describing evaporative emissions in MOVES3.72
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 (priorTripID) 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 SampleVehicleDay table
Source Type
Number of Records
sourceTypelD
Description
Weekday (daylD 5)
Weekend (daylD 2)
11
Motorcycle
2214
983
21
Passenger Car
821
347
31
Passenger Truck
834
371
32
Light Commercial Truck
773
345
41
Other Bus
190
73
42
Transit Bus
110
14
43
School Bus
136
59
51
Refuse Truck
205
65
52
Single-Unit Short-Haul Truck
112
58
53
Single-Unit Long-Haul Truck
123
50
54
Motor Home
5431
2170
61
Combination Short-Haul Truck
130
52
62
Combination Long-Haul Truck
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 beyond the scope of this
update.
127
-------
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.73'74 The data comes from a variety of instrumented vehicle studies, summarized in Table
13-6. This data was 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 MOBILE675 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.
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
128
-------
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, both trips,
engine soak periods and parking hours would be consistent. The important changes made in the
activity for engine starts will need to be reconciled with the parking hours in future versions of
MOVES. 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.
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.74
129
-------
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.w 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 MOVES2014, 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, Roadway-
Specific Driving Schedules for Heavy-Duty Vehicles51 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.
w 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.
130
-------
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
An estimate of the distribution of truck hotelling duration times is derived from a 2004 CRC
paper76 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.
131
-------
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.
132
-------
Table 13-11 Calculation of hourly distributions of
lotelling
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
activity
Note:
*Assumes every trip ends 10 hours after it starts, such that all trips are 10 hours long. For the first hour of hotelling in each hour
bin, the column sum is reduced by 60 percent to account for trip ends in a column that are not a full hour.
133
-------
The distribution calculated using this method is similar to the behavior observed in a
dissertation77 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
.1 0.06
+J
-Q
b 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.68 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.78 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 in
MOVES.x
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.
134
-------
14. Geographical Allocation of Activity
MOVES is designed to model activity at a "domain" level and then allocate that activity to
"zones." The MOVES default database is populated for a domain of the entire United States
(including Puerto Rico and the Virgin Islands) and the default zones correspond to individual
counties. The MOVES design only allows for one set of geographic allocations to be stored in
the default database. While geographic allocations clearly 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.79 The MOVES geographic allocation factors are
stored in two tables, Zone and ZoneRoadType. The current geographic allocations in MOVES3
are based on the 2017 NEI.80 All allocations are based on the distribution of vehicle miles
traveled (VMT) in counties.
In MOVES3, 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.
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
SHOAllocFactor 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 2017 NEI VMT was aggregated into the annual sum for the four MOVES road types in each
county and nationally and used to calculate the SHO AllocFactor using Equation 14-1.
run mi r +. CountyVMTRoadTypeID Equation
SHOAllocFactorRoadTypeID =
The county allocation values for each roadway type sum to one (1.0) for the nation. The same
SHO AllocFactor set is the default for all calendar years at the National scale. County- and
Project-level calculations do not use the default SHO AllocFactor allocations at all. Instead,
County and Project scales require that the user input all local activity.
135
-------
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 SHPAllocFactor 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.
The county allocation values for parking hours sum to one (1.0) for the nation. The same
SHPAllocFactor set is the default for all calendar years at the National scale. County- and
Project-level calculations do not use the default SHPAllocFactor allocations at all. Instead,
County and Project scales require that the user input all local activity.
SHP Alloc Factori = CountyVMT,
Equation
14-2
iel
136
-------
15. Vehicle Mass ami 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 rule81 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:
VSP =
STP =
Equation
Av + Bv2 + Cv3 + M ¦ (a + g ¦ sinO) ¦ v 15-1
M
Av + Bv2 + Cv3 + M ¦ (a + g ¦ sinO) ¦ v Equation
fscale
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 fscale's 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
fscale is similar, but not equal to the average source mass of the vehicle source type (fscaie ^
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
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
y MOVES does not model grade at the national and county scale. Road grade may be entered at the project scale.
137
-------
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-duty10 and heavy-duty11 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 ami 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 m 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 MOVES3, the source masses of light-duty vehicles were unchanged from MOVES2014, as
presented in Table 15-1 and documented in Appendix F.
138
-------
Table 15-1. Average Vehicle Weight and Mass for Motorcycles, Light-duty Vehicles, and Light-
duty Trucks Regulatory Classes
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.82 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).83 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.
139
-------
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.84 These vehicles are assumed to have an
average source mass of 45,645 lbs, based on several studies of in-use refuse truck activity.85 86 87
88
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; however, for combination trucks, increases in source masses were expected as a
byproduct of trailer and engine improvements made to those 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 Analysis89 and in the docket for the Phase 2 rule.90'91
140
-------
Table 15-4 MHD and HHP Changes in Vehicle Weight by Model Year
Source Type1
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
2014-2017
-321
2018-2020
-298
2021-2023
-278
2024-2026
-278
2027+
-278
Combination Long-haul Truck
2014-2017
-400
2018-2020
-260
2021-2023
-201
2024-2026
-106
2027+
-40
Note:
1 No change in vehicle weights is expected for other sourcetypes.
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).92 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).20 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
141
-------
specifications for the most popular vehicle models in the NTD with a bus type of cutaway.
Specifically, we assumed the following:
• 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 MOVES3 are listed in Table K-l in
Appendix K.
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.
142
-------
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:93'94
A = 0.088 ¦ M Equation
15-3
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 Equati on 15 - 8:95
0.7457
A ~ 50_0~447 0-35 TRLHP@s°mPh
0.7457
B ~ (50 ¦ 0.447)2 0'10 TRLHP@SOmPh
0.7457
C ~ (50 ¦ 0.447)3 0-55 ™^^@50mph
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)96 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.97 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.
Equation
15-6
Equation
15-7
Equation
15-8
143
-------
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. Petrushov,98 as 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 the individual source types2. 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. In MOVES3, the road load parameters for
combination trucks have been revised for model years 1960-2060 using the methods described in
Section 15.2.2.2. The revised road load coefficients for heavy-duty source types other than
combination trucks for model years 2014-2060 are described in Section 15.2.2.3
Table 15-6 Road Load Coefficients for MY 1960-2013 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\
A( m )
0.0996 -M
0.0875 -M
0.0661 -M
0.0643 -M
Bfwf)
\ mz J
0
0
0
0
c(kwf)
\ J
0.00147 +
5.22 X 10'"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
z The A and C coefficients were derived in MOVES2010 based on the equations in Table 15-6 and the population
fraction of regulatory classes within the sourcetypes in MOVES2010. In MOVES2014 and MOVES3, we updated to
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.
144
-------
15.2.2.1. Incorporation of Heavy-Duty Greenhouse Gas
Standards In MOVE S3
EPA set greenhouse gas (GHG) emission standards for heavy-duty vehicles in two separate
rulemakings, refered to in this report as the Phase l" and Phase 2100 HD GHG rules. The Phase 1
rulemaking became effective for the 2014 model year, and was incorporated into MOVES2014.
The Phase 2 rulemaking became effective for the 2018 model year for trailers and becomes
effective in 2021 model year for other heavy-duty truck types and is fully phased-in in 2027
model year.aa
The road load coefficients in MOVES3 have been updated to reflect the projected improvements
to the vehicles in different model year groups. The first model year group includes model years
1960-2013 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 combination tractor and
trailer standards are phased-in using model year groups 2018-2020, 2021-2023, 2024-2026 and
2027-and-later. The Phase 2 standards for the source types other than combination trucks are
grouped into 2021-2023, 2024-2026 and 2027-and-later groups. 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 MOVES3 as a result of Phase 2
HD GHG rules.
The aerodynamic drag force, Faero as a function of speed is represented as:
1 2 Equation
raero ~ ^ P d^f^air 15-9
where p is the density of air, C& is the aerodynamic drag coefficient, At is the frontal area of the
vehicle and vatr 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:
STPaero = (j—) "\PCdAfV3
Vscale' ^
Equation
15-10
aa On October 27, 2017 the Truck Trailer Manufacturers' Association was granted their request to provisionally stay
the trailer provisions of the greenhouse gas standards that were slated to go into effect in January 2018. MOVES3
reflects the federal regulations with the trailer provisions in place. Adjustments will be made as needed in future
versions of MOVES if changes are made to the regulations.
145
-------
Thus, the C road load coefficient can be represented as:
1 Equation
c = 15.n
The quantity CdAf shortened to CdA, is called the drag area and is used to characterize the overall
aerodynamic drag forces for a vehicle.
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.
Section 15.2.2.2 describes the analysis to update road load coefficients for combination long-
haul (sourceTypelD 62) and short-haul (sourceTypelD 61) trucks in MOVES3. Section 15.2.2.3
describes the updates applied to heavy-duty source types other than combination trucks to
account for HD GHG Phase 1 and Phase 2 rulemakings. The details on the discussion of
incorporating Phase 1 and Phase 2 energy reductions from engine technology improvements into
MOVES3 can be found in the MOVES3 Heavy-Duty Emission Rate Report.11
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.11 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.
15.2.2.2. Combination Trucks for Model Years 1960-2060
MOVES3 includes updates to both the aerodynamic and rolling resistance components of the
overall road load reflecting the greenhouse gas emissions standards for combination trucks. A
new aerodynamic assessment of all model years of combination trucks was conducted to utilize a
consistent method in MOVES3, and the aerodynamic values were updated for all model years to
reflect the aerodynamic technology analysis and projections in HD GHG Phase 2 rulemakings.
The average road load coefficients are updated by source type and regulatory class through the
beginModelYearlD and endModelYearlD fields in the SourceUseTypePhysics table.
Appendix J describes how the aerodynamic improvements were developed as part of the
rulemaking and how they were used to update MOVES.
146
-------
15.2.2.3. Heavy-Duty Source Types other than Combination
Trucks for Moid Years 2014-2060
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.101
The final road load coefficients for all regulatory classes and sourcetypes in MOVES3 are shown
in Table K-l in Appendix K.
147
-------
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 have not been updated for MOVES3. More information on air
conditioning effects is provided in the MOVES technical report on adjustment factors.102
16.1. AC PenetrationFraction
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 MOBILE6.103 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 was 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.
148
-------
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.
149
-------
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
16.3. ACActivityTerms
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 MOBILE6,103 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.
150
-------
able 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.
85 90 95
Heat Index (F)
Figure 16-1 Air conditioning activity demand as a function of heat index
110
151
-------
17.Conclusion ami 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 MOVES3 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 MOVE3 were developed for a 2017 base year and much of the source data is from 2017 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 and gasoline heavy-duty vehicles 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;
• Incorporation of existing data from a recent CRC study41 that provided local data for
hourly speeds and VMT distributions by MOVES source use types—this data could be
summarized nationally to update the MOVES default distributions;
152
-------
• 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, particularly extended engine idling and APU use;
• 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;
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.
153
-------
Appendix A Fuel Type ami Regulatory Class Fractious from Previous
Versions of MOVES
Fuel type and regulatory class distributions for most source types are described in Section 5.2. In
MOVES3, 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.
Al. Distributions for Model II
The fuel type distributions between 1960 and 1981 for each source type have been summarized
in Table and Table . Truck diesel fractions in Table were derived using the 1999 IHS vehicle
registrations and the 1997 VIUS,104 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
Model
Year
Passenger
Trucks
(31)
Light
Commercial
Trucks
(32)
Refuse
Trucks
(51)
Single-Unit
Trucks
(52 & 53)
Short-Haul
Combination
Trucks
(61)
Long-Haul
Combination
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
Note:
* All other trucks are assumed to be gasoline-powered. Motor homes values were estimated as
described in Section 5.2.
For the non-truck source types, school bus fuel type fractions were reused from MOBILE6,
originally based on 1996 and 1997 IHS data,105 and passenger cars were split between gasoline
and diesel for 1960-1981 using the 1999 IHS vehicle registrations data set. As in previous
versions of MOVES, motorcycles were assumed to be all gasoline.
154
-------
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
Note:
* All other vehicles are assumed to be gasoline-powered. Values for Transit Buses and
Other Buses were estimated as described in Section 5.2.
The 1960-1981 regulatory class distributions were derived from the 1999 IHS data set 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 .
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%
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).
155
-------
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%
Note:
* For these source types, all remaining trucks are in the HHD regulatory class (regClassID 47)
A2. Distributions for Mod I irs 191 1 I
VIUS was our main source of information for determining fuel and regulatory class fractions for
these model years. Table A-5 summarizes how the VIUS2002 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 (PRIMARYTRIP) 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.
156
-------
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
#1
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)
TRIPPRIMARY
in (5,6)
Any
Any
Any
AXLE CONFIG
<=21
TRIPPRIMARY
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
TRIPPRIMARY
in (5,6)
Any
Any
Any
Notes:
* 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.
Source Type Definitions
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.
Ugihfi I 'I If neks
Light-duty trucks include pickups, sport utility vehicles (SUVs) and vans.23 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.
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 (VIUSGVW) to
157
-------
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 MOVES3. 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.
Single-Ur icks
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 (TRIPPRIMARY) 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 of MOVES were not retained in MOVES3.
Combination Tricks
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.106
Fuel Type ai a lory Class Distributions
The SampleVehiclePopulation table fractions were developed by EPA using the sample vehicle
counts data set, 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 set were
generated by multiplying the 2011 IHS vehicle populations by the source type allocations from
VIUS.
158
-------
While VIUS provide source type classifications, we relied primarily on the 2011 IHS vehicle
registration data set 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.
Table 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
Note:
'After the 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 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,107
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.
159
-------
Table A-7 Fractions used to distribute Class 2a and 2b trucksbb
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. The Table A- 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- 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.
bb Note, the values from the ORNL report were applied incorrectly in MOVES2014, leading to an overestimate in
the fraction of gasoline and Class 2a trucks and an underestimate in the fraction of diesel and Class 2b trucks.
160
-------
Table A-8
ist 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 was first
assigned to MOVES source types using the constraints in Table 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-l.
161
-------
Table A-l Map
ping 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
/c, the source type population fraction / for a specified source type I will be the number of VIUS
trucks JV 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:
f(VIUS)Uk, =
Nu,
j,k,l
X N'-i
t—'leL
Equation A-l
i.i.k.l
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:
NCSyP);j,fc,j = P(Polk)i jXi ¦ f (VIUS)Equation A-2
162
-------
where JV is the number of vehicles used to generated the SampleVehiclePopulation table, P is the
2011 IHS vehicle populations and / is the source type distributions from VIUS.
Polk 2011
Vehicle Category
Model Year
Fuel Type
GVWR
Household Units
Work Units
INTERCITY BUSES
TRANSIT BUSES
SCHOOL BUSES
Interim Polk
Interim VIUS
r »
VIUS 2002
SAMPLEID
AXLECONFIG
TRIPPRIMARY
OPCLASS
FUEL
\TUS_GVW
ADMMODELYEAR
ADM_GVW
TAB TRUCKS
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
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
163
-------
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-Duty Tricks
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
MOVES3. 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-2. 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.108 The school bus
regulatory class distributions were also derived from 2011 FHWA data109 as listed in Table A-3 ,
which were applied to model years prior to 2000 for both gasoline and diesel.
164
-------
Table A-2 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-3 Regulatory c
ass 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 V.2.2.5 Sing! i nn «\ niiibiiialion 'I
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 (regClassIDs 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.
165
-------
Appendix B 1990 Age Distributions
In MOVES3, the 1990 age distributions were unchanged from previous versions of the model.
This appendix describes their derivation; details on the derivations of the other age distributions
in MOVES3 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.
112. 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.
;ks
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 . 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-11-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."
166
-------
Table B-l VIUS1997 codes used for distinguishing truck source types
Source Type
Axle Arrangement
Primary Area of
Operation
Body Type
Major Use
Passenger Trucks
2 axle/4 tire (AXLRE=
1,5,6,7)
Any
Any
personal
transportation
(MAJUSE=20)
Light Commercial
Trucks
2 axle/4 tire (AXLRE=
1,5,6,7)
Any
Any
any but personal
transportation
Refuse Trucks
Single-Unit
(AXLRE=2-4, 8-16)
Off-road, local or
short-range
(AREAOP <=4)
Garbage hauler
(BODTYPE=30)
Any
Single-Unit Short-
Haul Trucks
Single-Unit
(AXLRE=2-4, 8-16)
Off-road, local or
short-range
(AREAOP<=4)
Any except garbage
hauler
Any
Single-Unit Long-
Haul Trucks
Single-Unit
(AXLRE=2-4, 8-16)
Long-range
(AREAOP>=5)
Any
Any
Combination Short-
Haul Trucks
Combination
(AXLRE>=17)
Off-road, local or
medium
(AREAOP<=4)
Any
Any
Combination Long-
Haul Trucks
Combination
(AXLRE>=17)
Long-range
(AREAOP>=5)
Any
Any
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.
115. School Buses and Mot mes
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. isit 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 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
167
-------
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.
Table B-2 Curve fit equations for registration distribution data by age
Vehicle
Age
Equation
1-17
(r age \ 12.53214119n
y = 3462 * e (Ai7.i690947sJ )
18-25+
24987.0776 * e-°-2000*age
168
-------
Appendix C Detailed Derivation of Age Distributions
Since purchasing registration data for all calendar years is prohibitively costly for historic years,
the base age distribution described in Section 6.1 and presented below 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 MOVES3, 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. Generic Survival Males
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.110 Survival rates
for passenger cars, passenger trucks and light commercial trucks came from NHTSA's
survivability Table 3 and Table 4.111 These survival rates are based on a detailed analysis of IHS
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 for vehicles up to age 30 (with all older vehicles lumped
into the age 30 category), but NHSTA car survival rates were available only to age 25.
Therefore, we extrapolated car rates to age 30 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
represent a one-dimensional array of sa values at each permissible age a as described in
Equation C- through Equation C-3 below:
169
-------
Age 0:
Equation C-l
Age 1:
s1 = l-
2(1-ff2)
Equation C-2
3
Ages 2-30:
Equation C-3
With limited data available on heavy-duty vehicle scrappage, survivability for all other source
types came from the Transportation Energy Data Book112 We used the heavy-duty vehicle
survival rates for model year 1980 (TEDB37, Table 3.14). 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 can be 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.
The resulting survival rates are listed in the default database's SourceTypeAge table, shown
below in Table C-l. Please note that since MOVES3 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 e.
170
-------
Table C-l Vehicle survival rate by age
Age
Motorcycles
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
0.300
0.300
0.300
0.300
C2. Vehicle Sales by Soun
Knowing vehicle sales by source type for every calendar year is essential for estimating age
distributions in both historic and projected years. Since MOVES3 doesn't calculate age
distributions at run time, this information isn't stored in the default database.00 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.
00 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 Os in the MOVES3 default database.
171
-------
Historic motorcycles sales came from the Motorcycle Industry Council's 2015 Motorcycle
Statistical Annual,113 which contains estimates of annual on-highway motorcycle sales going
back to 1989. Sales for calendar year 2015 and 2016 beyond 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 TEDB37 Table 4.6 estimate for total new retail car
sales.
Historic light truck sales came from the TEDB37 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 and 2019 publications of School Bus Fleet
Fact Book19 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)20 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 5.1.4. 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.
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.
„ , Sales school Sales transit „ „ „ .
~ ~ ¦ Popothey Equation C-4
Pschool *OPtransit
Historic sales for heavy-duty trucks were derived from the TEDB37 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 the 2014 IHS data set. Since the 2014
IHS data set 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.
172
-------
Projected sales for all source types were derived from AEO2019. 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 AEO2019 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.
Table C-2 Mapping AEO categories to source types for projecting vehicle populations
AEO Sales Category Groupings
MOVES Source Type
Total Car Sales1
11 - Motorcycle
21 - Passenger Car
Total Light Truck Sales1
+
Total Commercial Light Truck Sales11
31 - Passenger Truck
32 - Light Commercial Truck
Total Sales111
41 - Other Bus
42 - Transit Bus
43 - School Bus
Light Medium Subtotal Sales111
+
Medium Subtotal Sales111
51 - Refuse Truck
52 - Single-Unit Short-haul Truck
53 - Single-Unit Long-haul Truck
54 - Motor Home
Heavy Subtotal Sales111
61 - Combination Short-haul Truck
62 - Combination Long-haul Truck
I From AEO2019 Table 39: Light-Duty Vehicle Sales by Technology Type
II From AEO2019 Table 45: Transportation Fleet Car and Truck Sales by Type and Technology
III From AEO2019 Table 50: Freight Transportation Energy Use
173
-------
Age
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
25
26
27
28
29
30+
lias- i ,trA,". I i itributions
Tab e C-3 2014 age fractions by MOVES source type
11
21
31
32
41
42
43
51
52
53
54
61
0.048749
0.067075
0.066767
0.066767
0.072837
0.087654
0.069836
0.037865
0.062272
0.062272
0.016299
0.055691
0.042924
0.061964
0.047716
0.119023
0.047215
0.068418
0.039713
0.033829
0.036611
0.036611
0.020135
0.050905
0.045758
0.056152
0.044172
0.083282
0.042178
0.078347
0.041803
0.034106
0.049487
0.049487
0.010521
0.047936
0.031549
0.043031
0.046773
0.066891
0.038117
0.066915
0.045112
0.025332
0.037268
0.037268
0.019962
0.024696
0.024357
0.044877
0.038382
0.043327
0.041974
0.090937
0.043380
0.018752
0.018998
0.018998
0.003384
0.019356
0.053659
0.041199
0.029116
0.032828
0.047949
0.091708
0.056681
0.034343
0.025040
0.025040
0.005989
0.029012
0.066182
0.052143
0.051139
0.061667
0.048592
0.085745
0.058875
0.029047
0.053999
0.053999
0.022737
0.023073
0.081538
0.057332
0.055358
0.055807
0.050785
0.060813
0.052376
0.084356
0.059404
0.059404
0.038799
0.081495
0.079157
0.053820
0.056978
0.058020
0.063279
0.054740
0.051189
0.068844
0.067473
0.067473
0.052469
0.054521
0.071324
0.053307
0.060561
0.049878
0.038752
0.042457
0.046074
0.058608
0.057204
0.057204
0.041156
0.053846
0.058046
0.049173
0.062020
0.045780
0.038427
0.046542
0.052596
0.049756
0.044208
0.044208
0.063954
0.030149
0.062351
0.050226
0.057092
0.040942
0.050263
0.049449
0.039994
0.054893
0.039326
0.039326
0.048349
0.032315
0.050151
0.048462
0.055007
0.036421
0.047094
0.047076
0.048330
0.049993
0.037384
0.037384
0.045693
0.024980
0.041655
0.045002
0.048183
0.034160
0.054325
0.044969
0.055483
0.053075
0.044271
0.044271
0.030069
0.041563
0.033072
0.045704
0.044937
0.031612
0.063892
0.031786
0.050152
0.064437
0.047490
0.047490
0.056193
0.057629
0.024850
0.036964
0.039505
0.027008
0.038284
0.021421
0.027986
0.052779
0.043121
0.043121
0.087104
0.044710
0.018282
0.030852
0.031213
0.019471
0.031023
0.011808
0.026992
0.033098
0.023479
0.023479
0.039411
0.033750
0.014802
0.026554
0.028363
0.018999
0.028111
0.005410
0.024274
0.021538
0.026495
0.026495
0.065630
0.031220
0.013367
0.020137
0.020277
0.013165
0.021638
0.007407
0.021346
0.027190
0.020353
0.020353
0.034423
0.034261
0.010992
0.019016
0.019572
0.013398
0.021781
0.001345
0.023788
0.030628
0.025485
0.025485
0.037894
0.045554
0.009109
0.014037
0.016683
0.011014
0.016510
0.002512
0.012167
0.018851
0.017215
0.017215
0.039647
0.031652
0.008085
0.011117
0.011640
0.007811
0.014453
0.000366
0.014562
0.014524
0.013776
0.013776
0.023489
0.023816
0.005866
0.009004
0.008614
0.006198
0.008894
0.000544
0.013275
0.011540
0.011089
0.011089
0.022851
0.016466
0.004800
0.007487
0.007305
0.005465
0.007729
0.000445
0.017004
0.014326
0.011776
0.011776
0.015131
0.015985
0.004978
0.006083
0.006600
0.005063
0.010913
0.000544
0.017892
0.015966
0.013918
0.013918
0.022977
0.018710
0.005475
0.005086
0.006762
0.005230
0.013515
0.000277
0.009126
0.011579
0.012477
0.012477
0.028532
0.015744
0.005422
0.004188
0.005667
0.004675
0.008607
0.000109
0.009563
0.012034
0.011621
0.011621
0.025315
0.015033
0.006760
0.003785
0.004271
0.003559
0.005702
0.000030
0.008774
0.010690
0.009758
0.009758
0.022812
0.011995
0.009409
0.003289
0.004336
0.003829
0.004765
0.000010
0.006657
0.007608
0.009643
0.009643
0.013770
0.009313
0.008320
0.002669
0.003155
0.003184
0.004757
0.000020
0.004937
0.006995
0.008551
0.008551
0.014229
0.009199
0.059011
0.030264
0.021833
0.025526
0.017637
0.000198
0.010062
0.013417
0.060808
0.060808
0.031076
0.015427
174
-------
C4 Historic A .tributions
The base 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 29 and 30+ fractions with the base year age 29 and 30+ fractions
and renormalize the new age distribution to sum to 1 while retaining the original age 29
and 30+ 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):
Py—± = Py ~ Ny + Ry_1 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
historic populations shown in Section 4.1 leveled off during the recent recession; 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 (o) 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~i = ky_± ¦ (l — S0) o Py_1 Equation C-5
Substituting Equation C-5 into Equation 6-2 yields Equation C-6:
Py_! = Py — Ny + ky_t ¦ (l — S0) ° Py_ 1 Equation C-6
To solve for ky_1, Equation C-6 can be transformed into Equation C-7 using known total
populations and sales:
175
-------
Py-1 — Py Ny + ky_ 1 ¦ ^ ^(l S0) O Py-^j
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_1 is
approximated by applying the base age distribution fy to the population of the previous year
Py-i- The scaling factor ky_1 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_ 1. 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 1 x 10"6, which
occurred within 10 iterations for most source types and calendar years.
This algorithm was then repeated for each historic year from 2013 to 1999 and for each source
type using the following data sources:
• Total populations Py and Py-1 as described in Section 4.
• Generic survival rates S0 as described in Section CI.
• Vehicle sales Ny as described in Section C2.
• Base age distributions f2014 as described in Section 6.1.1. All other fy come from the
fy—\ of the previous iteration.
With all of this information, the age distributions were algorithmically determined for years
1999-2013 and are stored in the SourceTypeAgeDistribution table of the default database.
C5. 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 30+ fraction with the base year age 30+ fraction and renormalize the
new age distribution to sum to 1 while retaining the original age 0 and age 30+ 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.
176
-------
This is mathematically described with the following equation (reprinted from Section 6.1.3for
reference):
Equation
6-3
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:
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+1- The resulting age distribution fy+1 is then calculated using the known Py+1-
This algorithm was then repeated for each projected year from 2015 to 2060 and for each source
type using the following data sources:
• Total populations Py and Py+1 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 f2014 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
2015-2060 and are stored in the SourceTypeAgeDistribution table of the default database. An
illustration of passenger car age distributions is presented in Figure C-l. For clarity, only four
years are shown: 2014, 2020, 2030 and 2040.
Py+1 — Py ky 1 (1 Sq) ° Py + Ny+1
Equation C-8
Equation
C-9
177
-------
0%
0 5 10 15 20 25 30 35
Passenger Car Age
Figure C-l Selected age distributions for passenger cars in MOVES3
178
-------
Appendix I)
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, MOVES3 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 MOVES3. 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.
179
-------
Table D-l MOVES3 default driving schedule statistics
drive
schedule id
drive schedule name
avg
speed
max
speed
idle
time
(sec)
percent of
time idling
miles
time (sec)
minutes
hours
101
LD Low Speed 1
2.5
10.00
280
46.5%
0.419
602.00
10.03
0.167
153
LD LOS E Freeway
30.5
63.00
5
1.1%
3.863
456.00
7.60
0.127
158
LD High Speed Freeway 3
76.0
90.00
0
0.0%
12.264
581.00
9.68
0.161
201
MD 5mph Non-Freeway
4.6
24.10
85
29.0%
0.373
293.00
4.88
0.081
202
MD lOmph Non-Freeway
10.7
34.10
61
19.6%
0.928
311.00
5.18
0.086
203
MD 15mph Non-Freeway
15.6
36.60
57
12.6%
1.973
454.00
7.57
0.126
204
MD 20mph Non-Freeway
20.8
44.50
95
9.1%
6.054
1046.00
17.43
0.291
205
MD 25mph Non-Freeway
24.5
47.50
63
11.1%
3.846
566.00
9.43
0.157
206
MD 30mph Non-Freeway
31.5
55.90
54
5.5%
8.644
988.00
16.47
0.274
251
MD 30mph Freeway
34.4
62.60
0
0.0%
15.633
1637.00
27.28
0.455
252
MD 40mph Freeway
44.5
70.40
0
0.0%
43.329
3504.00
58.40
0.973
253
MD 50mph Freeway
55.4
72.20
0
0.0%
41.848
2718.00
45.30
0.755
254
MD 60mph Freeway
60.1
68.40
0
0.0%
81.299
4866.00
81.10
1.352
255
MD High Speed Freeway
72.8
80.40
0
0.0%
96.721
4782.00
79.70
1.328
301
HD 5mph Non-Freeway
5.8
19.90
37
14.2%
0.419
260.00
4.33
0.072
302
HD lOmph Non-Freeway
11.2
29.20
70
11.5%
1.892
608.00
10.13
0.169
303
HD 15mph Non-Freeway
15.6
38.30
73
12.9%
2.463
567.00
9.45
0.158
304
HD 20mph Non-Freeway
19.4
44.20
84
15.1%
3.012
558.00
9.30
0.155
305
HD 25mph Non-Freeway
25.6
50.70
57
5.8%
6.996
983.00
16.38
0.273
306
HD 30mph Non-Freeway
32.5
58.00
43
5.3%
7.296
809.00
13.48
0.225
351
HD 3 Omph Freeway
34.3
62.70
0
0.0%
21.659
2276.00
37.93
0.632
352
HD 40mph Freeway
47.1
65.00
0
0.0%
41.845
3197.00
53.28
0.888
353
HD 5 Omph Freeway
54.2
68.00
0
0.0%
80.268
5333.00
88.88
1.481
354
HD 60mph Freeway
59.7
69.00
0
0.0%
29.708
1792.00
29.87
0.498
355
HD High Speed Freeway
71.7
81.00
0
0.0%
35.681
1792.00
29.87
0.498
396
HD High Speed Freeway Plus 5mph
76.7
86.00
0
0.0%
38.170
1792.00
29.87
0.498
397
MD High Speed Freeway Plus 5mph
77.8
85.40
0
0.0%
103.363
4782.00
79.70
1.328
180
-------
Table D-l MOVES3 default driving schedule statistics
drive
schedule id
drive schedule name
avg
speed
max
speed
idle
time
(sec)
percent of
time idling
miles
time (sec)
minutes
hours
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
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
181
-------
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)
0.209770278
dayID5
0.01126165
Applicable when dayID=5
sourceTypeID31
0.001328731
Applicable when sourceTypeID=31
countyTypelD 1
0.030580086
Applicable when equation is used for an urban county
(countyTypeID= 1)
idleRegionID 104
0.021341588
Applicable when idleRegionID=104
idleRegionID 102
0.026097089
Applicable when idleRegionID=102
idleRegionID 103
0.054609956
Applicable when idleRegionID=103
idleRegionID 101
0.057215976
Applicable when idleRegionID=101
monthID2
0.002789102
Applicable when monthID=2
monthID3
-0.004290649
Applicable when monthID=3
monthID4
-0.006087151
Applicable when monthID=4
monthID5
-0.004123423
Applicable when monthID=5
monthID6
-0.002637001
Applicable when monthID=6
monthID7
0.002913621
Applicable when monthID=7
monthID8
-0.000662777
Applicable when monthID=8
monthID9
-0.002960034
Applicable when monthID=9
monthlDIO
0.007288183
Applicable when monthID=10
monthlDl 1
0.005849819
Applicable when monthID=l 1
monthID12
0.007585819
Applicable when monthID=12
idleRegionID 104 :monthID2
-0.014777342
Applicable when monthID=2 and idleRegionID=104
idleRegionID 102 :monthID2
-0.006638333
Applicable when monthID=2 and idleRegionID=102
idleRegionID 103 :monthID2
-0.017303092
Applicable when monthID=2 and idleRegionID=103
idleRegionID 101 :monthID2
-0.015947997
Applicable when monthID=2 and idleRegionID=101
idleRegionID 104 :monthID3
-0.026662158
Applicable when monthID=3 and idleRegionID=104
idleRegionID 102 :monthID3
-0.01167098
Applicable when monthID=3 and idleRegionID=102
idleRegionID 103 :monthID3
-0.043578722
Applicable when monthID=3 and idleRegionID=103
idleRegionID 101 :monthID3
-0.033397602
Applicable when monthID=3 and idleRegionID=101
idleRegionID 104 :monthID4
-0.028548744
Applicable when monthID=4 and idleRegionID=104
idleRegionID 102 :monthID4
-0.011944882
Applicable when monthID=4 and idleRegionID=102
idleRegionID 103 :monthID4
-0.047593842
Applicable when monthID=4 and idleRegionID=103
idleRegionID 101 :monthID4
-0.038414264
Applicable when monthID=4 and idleRegionID=101
182
-------
Table E-l Total idle fraction regression coefficients for light-duty vehicles trucks in urban counties
for weekt
ays (Continued)
Variable
Coefficients
Comments
idleRegionID 104 :monthID5
-0.040105796
Applicable when monthID=5 and idleRegionID=104
idleRegionID 102 :monthID5
-0.014531686
Applicable when monthID=5 and idleRegionID=102
idleRegionID 103 :monthID5
-0.057127644
Applicable when monthID=5 and idleRegionID=103
idleRegionID 101 :monthID5
-0.046499987
Applicable when monthID=5 and idleRegionID=101
idleRegionID 104:monthID6
-0.04388419
Applicable when monthID=6 and idleRegionID=104
idleRegionID 102:monthID6
-0.012980897
Applicable when monthID=6 and idleRegionID=102
idleRegionID 103 :monthID6
-0.057285679
Applicable when monthID=6 and idleRegionID=103
idleRegionID 101 :monthID6
-0.050253407
Applicable when monthID=6 and idleRegionID=101
idleRegionID 104:monthID7
-0.049352207
Applicable when monthID=7 and idleRegionID=104
idleRegionID 102:monthID7
-0.013796675
Applicable when monthID=7 and idleRegionID=102
idleRegionID 103 :monthID7
-0.064939617
Applicable when monthID=7 and idleRegionID=103
idleRegionID 101 :monthID7
-0.055021202
Applicable when monthID=7 and idleRegionID=101
idleRegionID 104: monthID 8
-0.045892406
Applicable when monthID=8 and idleRegionID=104
idleRegionID 102: monthID 8
-0.01495486
Applicable when monthID=8 and idleRegionID=102
idleRegionID 103 :monthID8
-0.060514513
Applicable when monthID=8 and idleRegionID=103
idleRegionID 101 :monthID8
-0.050001647
Applicable when monthID=8 and idleRegionID=101
idleRegionID 104:monthID9
-0.04806906
Applicable when monthID=9 and idleRegionID=104
idleRegionID 102:monthID9
-0.021947448
Applicable when monthID=9 and idleRegionID=102
idleRegionID 103 :monthID9
-0.060010652
Applicable when monthID=9 and idleRegionID=103
idleRegionID 101 :monthID9
-0.04850918
Applicable when monthID=9 and idleRegionID=101
idleRegionID 104 monthID 10
-0.05048841
Applicable when monthID=10 and idleRegionID=104
idleRegionID 102 monthID 10
-0.032213346
Applicable when monthID=10 and idleRegionID=102
idleRegionID 103 monthID 10
-0.068309965
Applicable when monthID=10 and idleRegionID=103
idleRegionID 101 monthID 10
-0.052869353
Applicable when monthID=10 and idleRegionID=101
idleRegionID 104 monthID 11
-0.02092116
Applicable when monthID=l 1 and idleRegionID=104
idleRegionID 102 monthID 11
-0.026195031
Applicable when monthID=l 1 and idleRegionID=102
idleRegionID 103 monthID 11
-0.045139401
Applicable when monthID=l 1 and idleRegionID=103
idleRegionID 101 monthID 11
-0.046514269
Applicable when monthID=l 1 and idleRegionID=101
idleRegionID 104 monthID 12
-0.00750439
Applicable when monthID=12 and idleRegionID=104
idleRegionID 102 monthID 12
-0.025582194
Applicable when monthID=12 and idleRegionID=102
idleRegionID 103 monthID 12
-0.042625551
Applicable when monthID=12 and idleRegionID=103
idleRegionID 101 monthID 12
-0.047243005
Applicable when monthID=12 and idleRegionID=101
183
-------
Table E-2Table E-2 shows a sample calculation of MOVES3 default total idle fractions using the
coefficients for passenger cars (sourceTypeID=21) in rural counties (countyTypeID=0) in
idleRegionID=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.
Table E-2 Example total idle fractions for rural New Jersey passenger cars
sourceTypelD
monthID
daylD
idleRegionID
county TypelD
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
184
-------
Appendix F
Source Masses from Previous Versions of MOVES
In MOVES3, the source masses of light-duty vehicles were unchanged from MOVES2010b and
MOVES2014. 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 weightClassID. 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 weightClassID. 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.
M =
Equation
F-l
185
-------
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 was 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 1997 and for 1999. Values
for 2000-and-later model years are based on model year 2000 certification data.
186
-------
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%: 5001bs. -7001bs.
280+ cc (9)
0.792
0.792
0.908
0.928
30%: 500 lbs.-700 lbs.
70%: > 7001bs.
F2. Passenger 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 was 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.
F3. Light-Duty Tricks
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.114 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):
• 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.
• VIUS 1997 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.
187
-------
• 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
VIUS 1997 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.
188
-------
Appendix G Freeway Ramp Contribution at tie County-Scale
MOVES3 removed the capability to model ramp emissions separately from freeways (Rural
restricted and Urban restricted roadtypes). This appendix contains summary of the analysis used
to evaluate the emission consequences of removing the ramp roadtype from MOVES.
We analyzed vehicle activity on ramps and freeways from a study using portable activity
measurement systems (PAMS) conducted in the Detroit metropolitan area on 12 light-duty
vehicles115. From the PAMS measurements, we calculated MOVES running operating mode
distributions for each of the 62 highway trips using two scenarios: 1) we included the on and off
ramp as part of each highway trip 2) we excluded the ramp activity from the highway trips.
Using MOVES2014a, we calculated the emission rates (g/hr and g/mile) from the two scenarios.
The overall emission rates calculated from all 62 trips (in both g/hr and g/mile) ramps are higher
than emissions estimated from MOVES highway driving cycles for all speeds greater than 20
mph. Thus, it is reasonable to expect that removing ramps could decrease the g/mile estimates
for exhaust pollutants, which it did for HC, CO and PM. Whereas, NOx and C02 were only
increased slightly (<1.1 percent), which may be attributed to the lower g/mile emission rates
observed on off-ramps compared to highway driving.
For estimating the impact of removing ramp in MOVES, the g/hr difference is used. This is
because MOVES estimates emissions by multiplying emission rates (g/hr) by source hours
operating (SHO). The calculation of SHO on restricted access highways is not affected with the
removal of ramps in MOVES3, because the inputs to calculate SHO on restricted access
roadways (VMT and average speed) in both MOVES2014 and MOVES3 include all the activity
on restricted access roadways, including the ramp activity.
Brake wear emissions exhibit a different behavior than the tailpipe emissions. The brakewear
emissions from the trips that exclude ramps are 44 percent (g/hr) and 33 percent (g/mile) lower
than the trips that contain the on and off ramp activity. These results are intuitive as off-ramps
should contain a large percentage of the deceleration that occurs on each highway trip. Tire wear
emissions were not estimated from the two scenarios, but are anticipated to differ only slightly,
because MOVES tire wear emissions are a function of speed and not acceleration.
189
-------
0
0 20 40 60
AVG Cycle Speed (mph)
Figure G-l . Comparison of g/hr/vehicle and g/mile/vehicle across cycle average speed estimated
from MOVES for vehicles operating on ramps measured in the Detroit PAMS study on on-ramps
(red), off-ramps (green) and interchange ramps (orange) The MOVES highway (black line) plots
the estimated emissions using the default MOVES driving cycles which do not include ramp
activity.
We estimated the impact of excluding ramps from onroad mobile source emissions inventories
for three urban counties across five different calendar years. We first estimated the mobile
emissions by roadtype using MOVES2014a without any ramp activity (ramp fraction = 0). Then
we adjusted the restricted access roadtype emissions to account for ramp activity based on the
g/hr values in Figure G-l estimated from the Detroit Light-duty PAMS study. As stated earlier,
we used the g/hr values because we assume the average speed and VMT by MOVES user is
190
-------
unchanged for restricted access roadtypes, to isolate the impact of changing only the operating
mode distribution of the roadtypes. We applied the percentage differences to all sourcetypes,
assuming that the values derived from light-duty vehicles can be extended to all vehicle types.
Using these assumptions, we calculated the emissions impact of excluding ramp activity from the
highway driving cycles as shown in Table G-l.. By treating ramp VMT as non-ramp freeway
VMT, the mobile-source emissions inventories are reduced by less than 3 percent for NOx and
less than 1 percent for HC, CO and Primary PM2.5 exhaust. Brakewear particulate is reduced by
<9 percent.
Table G-l. Estimated Emissions Inventory impact from excluding ramp activity from highway
driving cycles
Pollutant
County
2011
2015
2020
2025
2030
HC
A
0.24%
0.24%
0.22%
0.21%
0.19%
B
0.40%
0.39%
0.33%
0.31%
0.30%
C
0.19%
0.18%
0.15%
0.14%
0.13%
CO
A
0.39%
0.40%
0.40%
0.40%
0.40%
B
0.69%
0.73%
0.74%
0.75%
0.76%
C
0.37%
0.39%
0.42%
0.43%
0.42%
NOx
A
2.48%
2.63%
2.73%
2.71%
2.64%
B
3.00%
3.05%
3.01%
2.91%
2.78%
C
1.95%
2.00%
1.97%
1.92%
1.82%
Primary Exhaust
PM2.5
A
0.26%
0.27%
0.26%
0.25%
0.23%
B
0.33%
0.33%
0.32%
0.30%
0.29%
C
0.19%
0.20%
0.19%
0.18%
0.16%
Brake wear
Particulate
A
-5.73%
-5.99%
-5.95%
-5.92%
-5.88%
B
-8.51%
-8.73%
-8.74%
-8.74%
-8.72%
C
-4.66%
-4.75%
-4.72%
-4.70%
-4.69%
191
-------
Appendix H NREL Fleet DNA Preprocessing Steps
Appendix H 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 H-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
! ~ Vehicle
Day
Time.json
Latitude.json
Longitude.json
Engine Speed.json
Wheel Speed.json
Figure H-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
192
-------
length is too long, starts and vehicle soaks may be missed. A possible scenario resulting in a
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 H-2 and Figure H-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 H-2 and Figure H-3 to demonstrate what effect
changes in gap length might have. Figure H-2 provides the distributions for source type 62 which
consists of combination long-haul trucks that have very few starts per day and Figure H-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.
193
-------
ID 62 Day 5 Gap: 0.02min
ID 62 Day 5 Gap: 6.0min
ID 62 Day 5 Gap: 30.0min
3 6 9 12 15 18 21 24
Hour ID
Figure H-2. Start fraction weights soak distribution weighted by gap length: source type 62
194
-------
ID 52 Day 5 Gap: 0.02min
ID 52 Day 5 Gap: 6.0min
ID 52 Day 5 Gap: 30.0min
ssiil
Op Modes
101
102
m 103
104
105
| H 106
107
108
fr=
Hi
0.08
0.06
0.04
3 6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 60.0min
x
£ 0.04
B Op Mc
"'l. =
,
i
I c
6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 9Q.0min
6 9 12 15
Hour ID
18 21 24
ID 52 Day 5 Gap: 12Q.0min
0.00
.2 0.08
0.06
.2 0.04
6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 360.0min
0.02
0.00
9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 720.0min
t 0.06
(13
(/)
X
o 0.04
0.02
0.00
aflilll
it. =
Op Modes
101
102
103
104
105
Hi
106
107
108
I
X
3 6 9 12 15 18 21 24
Hour ID
ID 52 Day 5 Gap: 1800.Omin
0.08
0.04
S 0.02
6 9 12 15 18 21 24
Hour ID
9 12 15 18 21 24
Hour ID
6 9 12 15 18 21 24
Hour ID
Figure II-3 Start fraction weights soak distribution weighted by gap length: source type 52
195
-------
Appendix I 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 day ID (weekend and weekday) to illustrate the strengths
and weaknesses of each of the methods.
luated Methods
Initally, we used Method 1 (Equation 1-1) to average the idle fractions across all vehicles within
the same sourcetypelD and day ID (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 i = individual vehicle ID Equation 1-1
of Ratios" n = vehicles sampled within each sourcetype
s = source type ID
d = day type ID
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 1-2, 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 hours i, will yield the total idle hours, £ idle hours^
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 a =
Method 2 - ' Z operating hourst
"Sum over Equation 1-2
gum„ i = individual vehicle ID
s = source type ID
d = day type ID
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
196
-------
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 data sets 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 set 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
1-3. 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 1-3) 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
2('
operating hoursi
/ daysj)
i = individual vehicle ID
day Si = # of days vehicle, is instrumented
s = source type ID
d = day type ID
Equation
1-3
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.
Start fractionhsd
Method 1 - h= hour of the day
"Average of / = individual vehicle ID
Ratios" n = # of sampled vehicles
s = source type ID
d = day type ID
Y, Start fractionhi
n
Equation
1-4
197
-------
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 = source type ID
d = day type ID
Start fractionhsd =
£ startsi
rstartshi . \
^ V ' daysj
£ ^startsi
/ daysj)
h= hour of the day
i = individual vehicle ID
day Si = days vehicle, is instrumented
s = source type ID
d = day type ID
Equation
1-5
Equation
1-6
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 = source type ID
d = day type ID
Soak fractionh o s d =
h= hour of the day
i = Vehicle ID
o = operating mode (soak length)
s = source type ID
d = day type ID
Y, Soak fractionhi 0
n
2
^ startshi
Soak fractionh o s d =
yi fstartsh i o i \
^ V ' day Si)
rstartsh i
/ daysj)
h= hour of the day
i = Vehicle ID
o = operating mode (soak length)
day Si = days vehiclei is instrumented
s = source type ID
d = day type ID
Equation
1-7
Equation
1-8
Equation
1-9
198
-------
12. Comparison of Evaluated Methods
Figure 1-1 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.
0.5
0.4
c
o
"£03
ru
OJ
2 0.2
0.1-
0.0
Weekend | Weekday
Extended
Workday
#
<,
-------
0.10
Single-Unit Short-Haul | Weekday
C
o
fj 0.08
u
s 0.06
IS)
c
¦° 0.04
u
(C
g 0.02
0.00
6 9 12 15 18 21 24
Hour of the Day
Single-Unit Short-Haul | Weekday
6 9 12 15 18 21 24
Hour of the day
Method 1 "Average of Ratios'"
Method 3 "Normalized Sum over Sum'
Figure 1-2. Start fraction and Soak fraction calculated using Method 1 and Method 3.
13. 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
soureetypes. 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.
Ftowever, 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.
200
-------
Method 4 -
"Vocation
and
Activity
Weighted
fraction"
Idle fraction^ d =
/idle hourSi
day Si
x w.
'v^j
v, foperatinq hours; /
/daySi Wl
V J
i = individual vehicle ID
day Si = # of days vehicle, is instrumented
v = vehicle vocation
wv= (population/sample size of each vocation, v)
s = source type ID
d = day type ID
Equation
1-10
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, manufacturering 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.
201
-------
Appendix J Road Load Coefficient for Combination Trucks in ill) GHG
Rule
In the HD GHG rules, certification test procedures were developed to evaluate the aerodynamic
performance of tractors and trailers. 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. In Phase 2, trailers were added to the program
and new test procedures were developed 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 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 programs, the drag value is represented by the
aerodynamic drag area, (\iA. In the trailer program, the drag value is represented as a reduction
in drag area, ACdA, 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 or trailer. 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 program 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 AC117 The trailer bin structure is common to all box van trailer types.
Table J-l. Phase 2 GHG Aerodynamic Drag Area Bin Structure for Tractors [m2]
High-roof Sleeper Cab
High-roof Day Cab
Low-roof Sleeper &
Day Cabs
Mid-roof Sleeper &
Day Cabs
Tractor
CdA Bin
CdA range
CdA input
CdA range
CdA input
CdA range
CdA input
CdA range
CdA input
I
>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
202
-------
Tab
e J-2. Phase 2 GHG Aerodynamic Drag Area Bin Structure for Box Van Trailers
Trailer ACdA Bin
ACdA range
ACdA input for GEM
Midpoint of ACdA range
I
<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
[m2
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. In Phase 2,
that trailer is equipped with a trailer skirt. 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 J-3.
Table J-3. Tractor aerodynamic technology adoption rates by model year groups
Tractor
Tractor Bin
1960-2013
Phase 1 GHG
Phase 2 GHG
Phase 2 GHG
Phase 2 GHG
Bin
CdA input [m2]
2014-2020
2021-2023
2024-2026
2027+
I
7.15
25%
0%
0%
0%
0%
II
6.55
70%
10%
0%
0%
0%
—
a
u
III
5.95
5%
70%
60%
40%
20%
u
v
IV
5.40
0%
20%
30%
40%
30%
&
v
QJ
V
4.90
0%
0%
10%
20%
50%
*4-
VI
4.40
0%
0%
0%
0%
0%
o
s
VII
3.90
0%
0%
0%
0%
0%
¦§)
Mean CdA (w/ skirt) [m2]
6.67
5.9
5.68
5.52
5.26
n
Skirt effect [nf]
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
203
-------
I
7.45
25%
0%
0%
0%
0%
II
6.85
70%
30%
0%
0%
0%
C*5
pD
a
it*
III
6.25
5%
60%
60%
40%
30%
IV
5.70
0%
10%
35%
40%
30%
-------
However, since trailers were not regulated in the Phase 1 GHG rulemaking, 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 2008.118 Therefore, for trailers, we
split the model year groups prior to 2018, the year that the Phase 2 GHG rule takes effect for
trailers. The model years between 1960-2007 reflect the time period prior to the use of trailer
aerodynamic improvements. The model year groups of 2008-2014 and 2014-2018 reflect
voluntary improvements to trailer aerodynamics. As a result, the following trailer technology
adoption rates were used to determine the average AC
-------
Table J-4 Trailer aerodynamic technology adoption
Trailer Bin
1960-
2007
2008-
2013
2014-
2017
2018-
2020
Phase 2
GHG
2021-2023
Phase 2
GHG
2024-2026
Phase 2
GHG
2027+
I
100%
65%
55%
0%
0%
0%
0%
II
0%
0%
0%
0%
0%
0%
0%
Q£
e
es
>
X
III
0%
35%
40%
100%
0%
0%
0%
IV
0%
0%
5%
0%
100%
0%
0%
O
pD
V
0%
0%
0%
0%
0%
100%
30%
fi
o
VI
0%
0%
0%
0%
0%
0%
70%
N-1
VII
0%
0%
0%
0%
0%
0%
0%
Average
ACdA [m2]
0
0.1925
0.2625
0.55
0.85
1.2
1.48
I
100%
100%
100%
100%
0%
0%
0%
II
0%
0%
0%
0%
100%
0%
0%
fi
III
0%
0%
0%
0%
0%
100%
40%
s*
M
IV
0%
0%
0%
0%
0%
0%
60%
pfi
t:
o
-fi
V
0%
0%
0%
0%
0%
0%
0%
VI
0%
0%
0%
0%
0%
0%
0%
t/5
VII
0%
0%
0%
0%
0%
0%
0%
Average
ACdA [m2]
0
0
0
0
0.25
0.55
0.73
oe
I
100%
100%
100%
0%
0%
0%
0%
a
>
II
0%
0%
0%
0%
0%
0%
0%
o
-C
III
0%
0%
0%
100%
100%
100%
100%
W)
fi
IV
0%
0%
0%
0%
0%
0%
0%
o
o
V
0%
0%
0%
0%
0%
0%
0%
VI
0%
0%
0%
0%
0%
0%
0%
3
VII
0%
0%
0%
0%
0%
0%
0%
€
Average
ACdA [m2]
0
0
0
0.55
0.55
0.55
0.55
OA
fi
I
100%
100%
100%
100%
0%
0%
0%
eS
>
X
o
pfi
II
0%
0%
0%
0%
100%
100%
100%
III
0%
0%
0%
0%
0%
0%
0%
t:
o
IV
0%
0%
0%
0%
0%
0%
0%
pfi
QC
V
0%
0%
0%
0%
0%
0%
0%
'«
&
VI
0%
0%
0%
0%
0%
0%
0%
I
13
VII
0%
0%
0%
0%
0%
0%
0%
¦c
a
a.
Average
ACdA [m2]
0
0
0
0
0.25
0.25
0.25
rates by model year groups
206
-------
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 J-5 was
used. Trailers in the non-aero category are incompatible with aerodynamic improvements and
standards are based on tire technologies in the Phase 2 regulations. These trailers 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 J-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 J-6. Trailer aerodynamic improvements
are calculated using the trailer distribution shown in Table J-5 and the adoption rates of Table J-
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 J-3.
Table J-6 Average trailer ACdA values by tractor category and model year group [m
—-Model^ears
Category ——
Pre-2008
2008-2013
2014-2017
2018-2020
2021-2023
2024-2026
2027+
High-roof sleeper cab
0
0.140
0.191
0.400
0.687
1.023
1.276
High-roof day cab
0
0
0
0.199
0.358
0.358
0.358
The resulting drag values that include aerodynamic improvements from tractors and trailers are
shown below.
207
-------
Table J-7 Drag area, CdA [m2], by tractor-trailer subcategory and model year group
—MaxIcKcars
Category ——
Pre-2008
2008-2013
2014-2017
2018-2020
2021-2023
2024-2026
2027+
High-roof sleeper cab
7.2200
7.0798
6.2589
6.0495
5.5431
5.0467
4.5339
High-roof day cab
7.5200
7.4505
6.9250
6.7263
6.1966
6.0116
5.8566
Mid-roof
7.0000
7.0000
6.4225
6.4225
6.3250
6.2500
6.2100
Low-roof
6.0000
6.0000
5.3450
5.3450
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 J-9 was calculated using the distribution from Table
J-8 and the drag areas from Table J-7.
Table J-8 Roof height distribution within cab types
Roof height
Sleeper Cab
Day Cab
Low-roof
5%
47%
Mid-roof
15%
0%
High-roof
80%
45%
Vocational
0%
8%
Table J-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
7.0139
6.2377
6.0702
5.6452
5.2328
4.8146
Day cab (sourceType 61)
6.6840
6.6840
6.1390
6.0496
5.7313
5.6104
5.5219
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
dshown in Table J-10.
Table J-10 C coefficients [kW-s3/m3] of source types 61 and 62
)y model year group
Pre-2008
2008-2013
2014-2017
2018-2020
2021-2023
2024-2026
2027+
Sleeper cab (sourceType 62)
0.00435
0.00428
0.00381
0.00370
0.00344
0.00319
0.00294
Day cab (sourceType 61)
0.00408
0.00408
0.00374
0.00369
0.00350
0.00342
0.00337
The Phase 1 and Phase 2 GHG emission standards also project improvements to the tire rolling
resistance. MOVES3 reflects these improvements through revisions to 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
J-l
where g is the gravitational acceleration.
208
-------
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 while Phase 2 will lead to
lower rolling resistance of trailer tires.119120 The overall rolling resistance of the vehicle is a
weighted average of rolling resistance over axle based on axle loading.
n _ n Msteer , n Mdrive „ Mtrailer Equation J-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 J-8. Rolling resistance distributions, based on tire rolling resistance levels
from the GHG rules are shown in Table J-l 1.
Table J-11 Crr by axle and tractor type
Tire Crr
Tire Crr value
Pre-2014
Phase 1
Phase 1
Phase 2
Phase 2 GHG
Phase 2
level
[kg/metric
ton]
GHG
2014-2017
2018-
2020
GHG
2021-2023
2024-2026
GHG
2027+
Base
7.8
100%
10%
10%
5%
5%
5%
a
¦-0
1
6.6
0%
70%
70%
35%
15%
10%
i-H
a
Base
8.1
100%
10%
10%
5%
5%
5%
s.
0>
—
1
6.9
0%
70%
70%
35%
15%
10%
2
6.0
0%
20%
20%
50%
60%
50%
o
o
•c
Q
3
5.0
0%
0%
0%
10%
20%
35%
•-
1
JS
.Sf
Avg Crr
kg/metric ton]
8.1
6.84
6.84
6.32
6.04
5.845
a
1
6.5
0%
0%
0%
0%
0%
0%
H
2
6.0
100%
100%
0%
0%
0%
5%
i-H
s.
Avg Crr
kg/metric ton]
7.8
6.87
6.87
6.04
5.91
5.785
—
a
Base
8.1
100%
30%
30%
15%
10%
5%
«4-
o
1
6.9
0%
60%
60%
35%
25%
10%
o
•-
2
6.0
0%
10%
10%
50%
65%
85%
•c
Q
3
5.0
0%
0%
0%
0%
0%
0%
a
Avg Crr
kg/metric ton]
8.1
7.17
7.17
6.63
6.435
6.195
s
a
1
6.5
100%
100%
100%
0%
0%
0%
¦
-------
Table J-ll (Continued) Crr
)y axle ant
tractor type
Tire Crr
Tire Crr value
Pre-2014
Phase 1
Phase 1
Phase 2
Phase 2 GHG
Phase 2
level
[kg/metric
tonl
GHG
2014-2017
2018-
2020
GHG
2021-2023
2024-2026
GHG
2027+
Base
7.8
100%
30%
30%
5%
5%
5%
M
1
6.6
0%
60%
60%
35%
15%
10%
vh
a>
2
5.7
0%
10%
10%
50%
60%
50%
2
6.0
0%
10%
10%
50%
60%
50%
s
•C
Q
3
5.0
0%
0%
0%
10%
20%
35%
-C
W)
Avg Crr
[kg/metric tonl
8.1
7.17
7.17
6.32
6.04
5.845
M
1
6.5
0%
0%
100%
0%
0%
0%
2
6.0
100%
100%
0%
0%
0%
0%
V-l
2
6.0
0%
10%
10%
50%
65%
85%
o
-
•C
Q
3
5.0
0%
0%
0%
0%
0%
0%
£
Avg Crr
[kg/metric tonl
8.1
7.29
7.29
6.63
6.435
6.195
-J
1
6.5
100%
100%
100%
0%
0%
0%
2
6.0
0%
0%
0%
0%
0%
0%
$-H
0)
3
5.1
0%
0%
0%
100%
100%
0%
2
H
4
4.7
0%
0%
0%
0%
0%
100%
Avg Crr [kg/metric ton]
6.5
6.5
6.5
5.1
4.7
4.7
Base
7.8
100%
10%
10%
5%
5%
5%
1
6.6
0%
70%
70%
35%
15%
10%
5-H
2
6.0
0%
20%
20%
50%
60%
50%
•C
Q
3
5.0
0%
0%
0%
10%
20%
35%
9*
(J
Avg Crr
[kg/metric tonl
8.1
6.84
6.84
6.32
6.04
5.845
>
1
6.5
6.5
100%
100%
100%
0%
0%
2
6.0
6.0
0%
0%
0%
0%
0%
O
3
5.1
5.1
0%
0%
0%
100%
100%
2
H
4
4.7
4.7
0%
0%
0%
0%
0%
Avg Crr [kg/metric ton]
6.5
6.5
6.5
5.1
4.7
4.7
210
-------
The average Crr values of each tire type were weighted based on a typical loading of a heavy-
duty vehicle - 42.5 percent over the trailer axle, 42.5 percent over the drive axle and 15 percent
over the steer axle.dd The result is shown in Table J-12.
Table J-12 Crr [kg/metric ton! by tractor category
Pre-2014
2014-2017
2018-2020
2021-2023
2024-2026
2027+
High-roof sleeper cab
7.163
6.438
6.056
5.590
5.432
5.352
High-roof day cab
7.163
6.628
6.245
5.590
5.432
5.324
Low and Mid-roof sleeper cab
7.375
6.840
6.245
5.891
5.619
5.498
Low-roof day cab
7.375
6.909
6.314
5.891
5.619
5.498
Vocational tractor
7.375
6.909
6.314
5.935
5.679
5.515
Using the roof height distributions in Table J-8, the resulting Crr values are:
Table J-13 Crr [kg/metric ton] values by model year group
Pre-2014
2014-2018
2018-2020
2021-2023
2024-2026
2027+
Sleeper cab (sourceType 62)
7.2050
7.2050
6.5185
6.0935
5.6499
5.4690
Day cab (sourceType 61)
7.2794
7.2794
6.7826
6.2832
5.7589
5.5393
To calculate the A coefficient, Equation J-l was used in combination with the source mass
values and Crr values from Table J-13. Resulting A coefficients by model year group are shown
in Table J-14.
Table J-l
A coefficient values [kW-s/m] by model year group
Pre-2014
2014-2018
2018-2020
2021-2023
2024-2026
2027+
Sleeper cab (sourceType 62)
1.739
1.562
1.464
1.358
1.317
1.298
Day cab (sourceType 61)
1.641
1.519
1.408
1.291
1.242
1.215
dd This distribution is equivalent to the federal over-axle weight limits for an 80,000 GVWR 5-axle tractor-trailer:
12,000 pounds over the steer axle, 34,000 pounds over the tandem drive axles (17,000 pounds per axle) and 34,000
pounds over the tandem trailer axles (17,000 pounds per axle).
211
-------
Appendix K MO¥ES3 Source 1 seTypePhysics Table
Ta
}le K-l MOVES3 SourceUseTypePhysics Table
sourceTypelD
regClassID
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)
11
10
1960
2060
0.0251
0
0.0003
0.285
0.2850
21
20
1960
2060
0.1565
0.0020
0.0005
1.479
1.4788
30
1960
2060
0.2211
0.0028
0.0007
1.867
1.8669
31
41
1960
2009
0.2211
0.0028
0.0007
3.402
2.0598
2010
2060
0.2211
0
0.0007
3.402
5
30
1960
2060
0.2350
0.0030
0.0007
2.060
2.0598
32
41
1960
2009
0.2350
0.0030
0.0007
3.402
2.0598
2010
2060
0.2350
0
0.0007
3.402
5
1960
2009
1.2952
0
0.0037
5.684
2.0598
41
2010
2013
1.2952
0
0.0037
5.684
5
2014
2060
1.2304
0
0.0037
5.684
5
1960
2009
1.2952
0
0.0037
7.782
2.0598
2010
2013
1.2952
0
0.0037
7.782
5
42
2014
2020
1.2304
0
0.0037
7.782
5
2021
2023
1.0065
0
0.0037
7.782
5
2024
2026
0.9745
0
0.0037
7.782
5
2027
2060
0.9265
0
0.0037
7.782
5
1960
2009
1.2952
0
0.0037
11.367
17.1
41
2010
2013
1.2952
0
0.0037
11.367
7
46
2014
2020
1.2304
0
0.0037
11.367
7
2021
2023
1.0065
0
0.0037
11.367
7
2024
2026
0.9745
0
0.0037
11.367
7
2027
2060
0.9265
0
0.0037
11.367
7
1960
2009
1.2952
0
0.0037
15.603
17.1
2010
2013
1.2952
0
0.0037
15.603
10
47
2014
2020
1.2304
0
0.0037
15.603
10
2021
2023
1.0065
0
0.0037
15.603
10
2024
2026
0.9745
0
0.0037
15.603
10
2027
2060
0.9265
0
0.0037
15.603
10
1960
2009
1.0944
0
0.0036
7.782
2.0598
2010
2013
1.0944
0
0.0036
7.782
5
42
42
2014
2020
1.0397
0
0.0036
7.782
5
2021
2023
1.0397
0
0.0036
7.782
5
2024
2026
1.0397
0
0.0036
7.782
5
2027
2060
0.9139
0
0.0036
7.782
5
212
-------
Table K-l MOVES3 SourceUseTypePhysics table (Continued)
sourceTypelD
regClassID
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)
1960
2009
1.0944
0
0.0036
11.367
17.1
2010
2013
1.0944
0
0.0036
11.367
7
46
2014
2020
1.0397
0
0.0036
11.367
7
2021
2023
1.0397
0
0.0036
11.367
7
2024
2026
1.0397
0
0.0036
11.367
7
42
2027
2060
0.9139
0
0.0036
11.367
7
1960
2009
1.0944
0
0.0036
15.603
17.1
2010
2013
1.0944
0
0.0036
15.603
10
47, 48
2014
2020
1.0397
0
0.0036
15.603
10
2021
2023
1.0397
0
0.0036
15.603
10
2024
2026
1.0397
0
0.0036
15.603
10
2027
2060
0.9139
0
0.0036
15.603
10
1960
2009
0.7467
0
0.0022
5.684
2.0598
41
2010
2013
0.7467
0
0.0022
5.684
5
2014
2060
0.7094
0
0.0022
5.684
5
1960
2009
0.7467
0
0.0022
7.782
2.0598
2010
2013
0.7467
0
0.0022
7.782
5
42
2014
2020
0.7094
0
0.0022
7.782
5
2021
2023
0.6377
0
0.0022
7.782
5
2024
2026
0.6037
0
0.0022
7.782
5
2027
2060
0.5696
0
0.0022
7.782
5
1960
2009
0.7467
0
0.0022
11.367
17.1
43
2010
2013
0.7467
0
0.0022
11.367
7
46
2014
2020
0.7094
0
0.0022
11.367
7
2021
2023
0.6377
0
0.0022
11.367
7
2024
2026
0.6037
0
0.0022
11.367
7
2027
2060
0.5696
0
0.0022
11.367
7
1960
2009
0.7467
0
0.0022
15.603
17.1
2010
2013
0.7467
0
0.0022
15.603
10
47
2014
2020
0.7094
0
0.0022
15.603
10
2021
2023
0.6377
0
0.0022
15.603
10
2024
2026
0.6037
0
0.0022
15.603
10
2027
2060
0.5696
0
0.0022
15.603
10
213
-------
Table K-l MOVES3 SourceUseTypePhysics table (Continued)
sourceTypelD
regClassID
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)
1960
2009
1.5835
0
0.0036
3.574
2.0598
41
2010
2013
1.5835
0
0.0036
3.574
5
2014
2060
1.5043
0
0.0036
3.574
5
1960
2009
1.5835
0
0.0036
5.768
2.0598
2010
2013
1.5835
0
0.0036
5.768
5
42
2014
2020
1.5043
0
0.0036
5.768
5
2021
2023
1.5043
0
0.0036
5.768
5
2024
2026
1.5043
0
0.0036
5.768
5
2027
2060
1.3223
0
0.0036
5.768
5
1960
2009
1.5835
0
0.0036
13.800
17.1
51
2010
2013
1.5835
0
0.0036
13.800
7
46
2014
2020
1.5043
0
0.0036
13.800
7
2021
2023
1.5043
0
0.0036
13.800
7
2024
2026
1.5043
0
0.0036
13.800
7
2027
2060
1.3223
0
0.0036
13.800
7
1960
2009
1.5835
0
0.0036
20.704
17.1
2010
2013
1.5835
0
0.0036
20.704
10
47
2014
2020
1.5043
0
0.0036
20.704
10
2021
2023
1.5043
0
0.0036
20.704
10
2024
2026
1.5043
0
0.0036
20.704
10
2027
2060
1.3223
0
0.0036
20.704
10
214
-------
Table K-l MOVES3 SourceUseTypePhysics table (Continued)
sourceTypelD
regClassID
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)
1960
2009
0.6279
0
0.0016
3.574
2.0598
41
2010
2013
0.6279
0
0.0016
3.574
5
2014
2060
0.5965
0
0.0016
3.574
5
1960
2009
0.6279
0
0.0016
5.768
2.0598
2010
2013
0.6279
0
0.0016
5.768
5
42
2014
2020
0.5965
0
0.0016
5.768
5
2021
2023
0.5583
0
0.0016
5.766
5
2024
2026
0.5583
0
0.0016
5.763
5
2027
2060
0.5357
0
0.0016
5.761
5
1960
2009
0.6279
0
0.0016
13.800
17.1
52
2010
2013
0.6279
0
0.0016
13.800
7
46
2014
2020
0.5965
0
0.0016
13.800
7
2021
2023
0.5583
0
0.0016
13.798
7
2024
2026
0.5583
0
0.0016
13.795
7
2027
2060
0.5357
0
0.0016
13.793
7
1960
2009
0.6279
0
0.0016
25.048
17.1
2010
2013
0.6279
0
0.0016
25.048
10
47
2014
2020
0.5965
0
0.0016
25.048
10
2021
2023
0.5583
0
0.0016
25.046
10
2024
2026
0.5583
0
0.0016
25.044
10
2027
2060
0.5357
0
0.0016
25.041
10
215
-------
Table K-l MOVES3 SourceUseTypePhysics table (Continued)
sourceTypelD
regClassID
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)
1960
2009
0.5573
0
0.0015
3.574
2.0598
41
2010
2013
0.5573
0
0.0015
3.574
5
2014
2060
0.5294
0
0.0015
3.574
5
1960
2009
0.5573
0
0.0015
5.768
2.0598
2010
2013
0.5573
0
0.0015
5.768
5
42
2014
2020
0.5294
0
0.0015
5.768
5
2021
2023
0.4849
0
0.0015
5.765
5
2024
2026
0.4590
0
0.0015
5.757
5
2027
2060
0.4590
0
0.0015
5.750
5
1960
2009
0.5573
0
0.0015
13.800
17.1
53
2010
2013
0.5573
0
0.0015
13.800
7
46
2014
2020
0.5294
0
0.0015
13.800
7
2021
2023
0.4849
0
0.0015
13.797
7
2024
2026
0.4590
0
0.0015
13.789
7
2027
2060
0.4590
0
0.0015
13.782
7
1960
2009
0.5573
0
0.0015
25.048
17.1
2010
2013
0.5573
0
0.0015
25.048
10
47
2014
2020
0.5294
0
0.0015
25.048
10
2021
2023
0.4849
0
0.0015
25.045
10
2024
2026
0.4590
0
0.0015
25.038
10
2027
2060
0.4590
0
0.0015
25.031
10
1960
2009
0.6899
0
0.0021
3.574
2.0598
41
2010
2013
0.6899
0
0.0021
3.574
5
2014
2060
0.6554
0
0.0021
3.574
5
1960
2009
0.6899
0
0.0021
5.768
2.0598
2010
2013
0.6899
0
0.0021
5.768
5
42
2014
2020
0.6554
0
0.0021
5.768
5
2021
2023
0.5191
0
0.0021
5.768
5
54
2024
2026
0.5191
0
0.0021
5.768
5
2027
2060
0.4935
0
0.0021
5.768
5
1960
2009
0.6899
0
0.0021
13.800
17.1
2010
2013
0.6899
0
0.0021
13.800
7
46
2014
2020
0.6554
0
0.0021
13.800
7
2021
2023
0.5191
0
0.0021
13.800
7
2024
2026
0.5191
0
0.0021
13.800
7
2027
2060
0.4935
0
0.0021
13.800
7
216
-------
Table K-l MOVES3 SourceUseTypePhysics table (Continued)
sourceTypelD
regClassID
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)
1960
2009
0.6899
0
0.0021
25.048
17.1
2010
2013
0.6899
0
0.0021
25.048
10
54
47
2014
2020
0.6554
0
0.0021
25.048
10
2021
2023
0.5191
0
0.0021
25.048
10
2024
2026
0.5191
0
0.0021
25.048
10
2027
2060
0.4935
0
0.0021
25.048
10
1960
2009
1.6406
0
0.0041
14.012
17.1
2010
2013
1.6406
0
0.0041
14.012
7
2014
2017
1.5190
0
0.0037
13.867
7
46
2018
2020
1.4078
0
0.0037
13.877
7
2021
2023
1.2908
0
0.0035
13.886
7
2024
2026
1.2416
0
0.0034
13.886
7
2027
2060
1.2151
0
0.0034
13.886
7
1960
2009
1.6406
0
0.0041
24.830
17.1
2010
2013
1.6406
0
0.0041
24.830
10
61
2014
2017
1.5190
0
0.0037
24.684
10
47
2018
2020
1.4078
0
0.0037
24.695
10
2021
2023
1.2908
0
0.0035
24.704
10
2024
2026
1.2416
0
0.0034
24.704
10
2027
2060
1.2151
0
0.0034
24.704
10
1960
2013
1.6406
0
0.0041
24.830
17.1
2014
2017
1.5190
0
0.0037
24.684
17.1
49
2018
2020
1.4078
0
0.0037
24.695
17.1
2021
2023
1.2908
0
0.0035
24.704
17.1
2024
2026
1.2416
0
0.0034
24.704
17.1
2027
2060
1.2151
0
0.0034
24.704
17.1
217
-------
Table K-l MOVES3 SourceUseTypePhysics table (Continued)
sourceTypelD
regClassID
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)
1960
2009
1.7388
0
0.0043
14.012
17.1
2010
2013
1.7388
0
0.0043
14.012
7
2014
2017
1.5615
0
0.0038
13.831
7
46
2018
2020
1.4635
0
0.0037
13.894
7
2021
2023
1.3585
0
0.0034
13.921
7
2024
2026
1.3173
0
0.0032
13.964
7
2027
2060
1.2976
0
0.0029
13.994
7
1960
2009
1.7388
0
0.0043
24.830
17.1
2010
2013
1.7388
0
0.0043
24.830
10
62
2014
2017
1.5615
0
0.0038
24.648
10
47
2018
2020
1.4635
0
0.0037
24.712
10
2021
2023
1.3585
0
0.0034
24.739
10
2024
2026
1.3173
0
0.0032
24.782
10
2027
2060
1.2976
0
0.0029
24.812
10
1960
2013
1.7388
0
0.0043
24.830
17.1
2014
2017
1.5615
0
0.0038
24.648
17.1
49
2018
2020
1.4635
0
0.0037
24.712
17.1
2021
2023
1.3585
0
0.0034
24.739
17.1
2024
2026
1.3173
0
0.0032
24.782
17.1
2027
2060
1.2976
0
0.0029
24.812
17.1
218
-------
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, Moves 3 Technical Guidance: Using MOVES to Prepare Emission Inventories for
State Implementation Plans and Transportation Conformity, Ann Arbor, MI
3 USEPA (2015). U.S. Environmental Protection Agency Peer Review Handbook. EPA/100/B-
15/001. Prepared for the U.S. Environmental Protection Agency under the direction of the EPA
Peer Review Advisory Group. Washington, D.C. 20460. October 2015.
https://www.epa.gov/sites/production/files/2020-
08/documents/epa_peer_review_handbook_4th_edition.pdf.
4 USEPA (2017). Population and Activity of Onroad Vehicles in MOVES201X - Draft Report.
Draft report and peer-review documents. Record ID 328870. EPA Science Inventory. September
2017. https://cfpub.epa.gov/si/si_public_record_report.cfm?dirEntryId=328870.
5 USEPA (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. December 2019.
https ://cfpub. epa.gov/si/si_public_record_report. cfm?dirEntryId=34713 5.
6 US FHWA, Table VM-1, "Annual Vehicle Distance Traveled in Miles and Related Data,"
Highway Statistics, 1990-2017, Washington, DC: March 2019,
https://www.fhwa.dot.gov/policyinformation/statistics/2017/vml.cfm.
7 US EPA, Frequently Asked Questions about Heavy-Duty "Glider Vehicles" and "Glider Kits",
EPA-420-F-15-904, Ann Arbor, MI: July 2015,
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=Pl 00MUVI.PDF.
8 USEPA (2020). Fuel Supply Defaults: Regional Fuels and the Fuel Wizard in MOVES3. EPA-
420-R-20-017. Office of Transportation and Air Quality. US Environmental Protection Agency.
Ann Arbor, MI. November 2020. https://www.epa.gov/moves/moves-technical-reports.
9 US EPA, Development of Gasoline Fuel Effects in the Motor Vehicle Emissions Simulator
(MOVES2009), EPA-420-P-09-004, Ann Arbor, MI: August 2009,
http://www.epa.gov/otaq/models/moves/techdocs/420p09004.pdf.
219
-------
10 USEPA (2020). Exhaust Emission Rates for Light-Duty Onroad Vehicles in MOVES3. EPA-
420-R-20-019. Office of Transportation and Air Quality. US Environmental Protection Agency.
Ann Arbor, MI. November 2020. https://www.epa.gov/moves/moves-technical-reports.
11 U SEP A (2020). Exhaust Emission Rates of Heavy-Duty Onroad Vehicles in MO VES3. EP A-
420-R-20-018. Office of Transportation and Air Quality. US Environmental Protection Agency.
Ann Arbor, MI. November 2020. https://www.epa.gov/moves/moves-technical-reports.
12 US FHWA, 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: August 2011,
http://www.fhwa.dot.gov/ohim/vml_methodology_2007.pdf.
13 US EPA, Total Activity Generator, October 2020, https://www.epa.gov/moves/moves-
algorithms#moduletotal-activity-generator
14 US Energy Information Administration (EIA), Annual Energy Outlook 2019, Washington, DC:
January 2019, https://www.eia.gov/outlooks/aeo/pdf/AEO2019.pdf.
15 US Federal Highway Administration (FHWA), Office of Highway Policy Information (OHPI),
Table MV-1, "State Motor-Vehicle Registrations," Highway Statistics, 1990-2017, Washington,
DC: January 2019, https://www.fhwa.dot.gov/policyinformation/statistics/2017/mvl.cfm.
16 IHS, Inc. (formerly R.L. Polk & Co.), National Vehicle Population Profile®, Southfield, MI;
1999, 2011, and 2014; https://www.ihs.com/btp/polk.html.
17 IHS, Inc. (formerly R.L. Polk & Co.), Trucking Industry Profile TIP® Vehicles in Operation,
Southfield, MI: 1999, 2011, and 2014; https://www.ihs.com/btp/polk.html.
18 US EPA, MOVES2010 Highway Vehicle Population and Activity Data, EPA-420-R-10-026,
Washington, DC: November 2010,
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=Pl OOABRO.pdf.
19 Bobit Publications, School Bus Fleet Fact Book, Torrance, CA: 1992, 2002, 2004-2019,
http://www.schoolbusfleet.com.
20 US Federal Transit Administration (FTA), National Transit Database, 2002-2017,
https://www.transit.dot.gov/ntd.
220
-------
21 Koupal, J., T. DeFries, C. Palacios, S. Fincher and D. Preusse (2014). Motor Vehicle
Emissions Simulator Input Data. Transportation Research Record: Journal of the Transportation
Research Board, 2427, 63-72. DOI: doi: 10.3141/2427-07.
22 Yoon, S., Georgia Institute of Technology, A New Heavy-Duty Vehicle Visual Classification
and Activity Estimation Method for Regional Mobile Source Emissions Modeling (student
thesis), Atlanta, GA: August 2005,
https://smartech.gatech.edu/bitstream/handle/1853/7245/seungju_yoon_200508_phd.pdf.
23 US FHWA, "Vehicle Type Codes and Descriptions," Highway Performance Monitoring
System Field Manual, Washington, DC: last updated 4 April 2011,
http://www.fhwa. dot.gov/ohim/hpmsmanl/chapt3. cfm.
24 Weatherrby, M., Jackson, D., Koupal, J., and DenBleyker, A., Eastern Research Group, Inc.
(ERG), 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: July, 2017.
25 US FTA, National Transit Database, "NTD Glossary," last updated 19 February 2015,
http://www.ntdprogram.gov/ntdprogram/Glossary. htm.
26 US FHWA, Office of Planning, Environment, and Realty, "Planning Glossary," updated 21
March 2012, http://www.fhwa.dot.gov/planning/glossary/glossary_listing.cfm?sort=definition.
27 Stanard, A., Fincher, S., Kishan, S., and Sabisch, M., Eastern Research Group, Inc. (ERG),
Data Analyses on Drayage Heavy-Duty Vehicles. EPA EP-C-12-017, Work Assignment 0-2,
Austin, TX: 7 December 2012.
28 US EIA, "Light-Duty Vehicle Sales by Technology Type," reference case, Annual Energy
Outlook 2018, released February 2018, http://www.eia.gov/oiaf/aeo/tablebrowser.
29 US EIA, "Transportation Fleet Car and Truck Sales by Type and Technology," reference case,
Annual Energy Outlook 2018, released February 2018,
http://www.eia.gov/oiaf/aeo/tablebrowser.
30 US Government Publishing Office (GPO), Code of Federal Regulations, Title 40 - Protection
of Environment, Vol. 19, CFR 86.091-2, "Definitions," 1 July 2012, http://www.gpo.gov/fdsys.
221
-------
31Eastern Research Group, MOVES Input Improvements for the 2011 NEI: CRC Project A-88
October 2014, http://crcsite.wpengine.eom/wp-content/uploads/2019/05/CRC-A88-Final-Report-
102114.pdf
32 Brakora, J. 2018. Estimated Glider Vehicle Production for Use in EPA MOVES Models.
Jessica Brakora. Cleaner Trucks Initiative - Docket EPA-HQ-OAR-2019-0055. Draft
memorandum to the docket.
33 The Age Distribution Projection Tool for MOVES2014 is a macro-enabled spreadsheet and is
available for download at http://www.epa.gOv/oms/models/moves/tools.htm#fleet-2014.
34 Browning, L., Chan, M., Coleman, D., and Pera, C., ARCADIS Geraghty & Miller, Inc.,
Update of Fleet Characterization Data for Use inMOBILE6: Final Report, EPA420-P-98-016,
Mountain View, CA: 11 May 1998,
http://nepis. epa.gov/Exe/ZyPDF.cgi?Dockey=P 1001ZUK.PDF.
35 US Census Bureau, 2002 Vehicle Inventory and Use Survey, EC02TV-US, Washington, DC:
December 2004, http://www.census.gov/svsd/www/vius/2002.html.
36 Browning, L., Chan, M., Coleman, D., and Pera, C., ARCADIS Geraghty & Miller, Inc.,
Update of Fleet Characterization Data for Use inMOBILE6: Final Report, EPA420-P-98-016,
Mountain View, CA: 11 May 1998,
http://nepis. epa.gov/Exe/ZyPDF.cgi?Dockey=P 1001ZUK.PDF.
37 47 TL Enterprises, Inc., Good Sam Club, Highways Member Study 2000, Ventura, CA: 2000.
38 National Household Travel Survey, 2017 NHTS Average Annual Vehicle Miles of Travel Per
Vehicle (Best Estimate), https://nhts.ornl.gov.
39 US EPA, "The National Emission Inventory (NEI)," , https://www.epa.gov/air-emissions-
inventories/national-emissions-inventory-nei.
40 U.S. Department of Transportation Federal Highway Administration Office of Highway Policy
Information Highway Statistics Series:
https://www.fhwa.dot.gov/policyinformation/statistics.cfm
41 CRC Project A-100, Improvement of Default Inputs for MOVES and SMOKE-MOVES.
http://crcsite.wpengine.eom/wpcontent/uploads/2019/05/ERG_FinalReport_CRCA100_28Feb20
17.pdf.
222
-------
42 U.S. 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.
43 National Emissions Inventory Collaborative. (2019). "2016beta Emissions Modeling
Platform." Retrieved from http://views.cira.colostate.edu/wiki/wiki/10197.
44 USEPA (2016). Population and Activity of On-road Vehicles inMOVES2014. EPA-420-R-
16-003a. Office of Transportation and Air Quality. US Environmental Protection Agency. Ann
Arbor, MI. March 2016. https://www.epa.gov/moves/moves-technical-reports., Appendix F.
45 Sierra Research, Inc., Development of Speed Correction Cycles, M6.SPD.001, EPA 68-C4-
0056, Sacramento, CA: 26 June 1997, http://www.epa.gov/oms/models/mobile6/r01042.pdf.
46 Hart, C., Koupal, J., and Giannelli, R., US EPA, EPA's Onboard Analysis Shootout: Overview
and Results, EPA420-R-02-026, Ann Arbor, MI: October 2002,
http://epa.gov/oms/models/ngm/r02026.pdf.
47 DieselNet, "Emission Test Cycles: New York Bus," last accessed 8 June 2015,
https://www.dieselnet.com/standards/cycles/nybus.php.
48 Melendez, M. T., J; Zuboy, J (2005). Emission Testing of Washington Metropolitan Area
Transit Authority (WMATA) Natural Gas and Diesel Transit Buses. NREL/TP-540-36355.
December 2005., http://www.afdc.energy.gov/pdfs/36355.pdf.
49 Clark, N. and Gautam, M., West Virginia Research Corporation, Heavy-Duty Vehicle Chassis
Dynamometer Testing for Emissions Inventory, Air Quality Modeling, Source Apportionment
and Air Toxics Emissions Inventory, CRC Project E55/59, Morgantown, VW: 12 July 2005,
http://www.crcao.org/reports/recentstudies2005/E55-2%20FINAL%20REPORT%20071205.pdf.
50 Sensors, Inc., On-RoadEmissions Testing of 18 Tier 1 Passenger Cars and 17 Diesel Powered
Public Transport Buses, QT-MI-01-000659, Saline, MI: 22 October 2002,
http://www.epa.gov/oms/models/ngm/r02030.pdf.
51 Eastern Research Group, Inc., Roadway-Specific Driving Schedules for Heavy-Duty Vehicles,
EPA420-R-03-018, Austin, TX: August 2003, http://nepis.epa.gov.
52 US EPA, Greenhouse Gas Emissions Model (GEM) for Medium- and Heavy-Duty Vehicle
Compliance, Simulation Model v2.0.1, May 2013, http://www.epa.gov/otaq/climate/gem.htm.
223
-------
53 Liu, H., D. Sonntag, D. Brzezinski, C. R. Fulper, D. Hawkins and J. E. Warila (2016).
Operations and Emissions Characteristics of Light-Duty Vehicles on Ramps. Transportation
Research Record: Journal of the Transportation Research Board, 2570, 1-11. DOI:
doi: 10.3141/2570-01.
54 US EPA. Amendments to Vehicle Inspection Maintenace Program Requirements
Incorporating the Onboard Diagnositc Check. 66 FR 18156, April 5, 2001
55 Walkowicz, K., K. Kelly, A. Duran and E. Burton (2014). Fleet DNA Project Data. National
Renewable Energy Laboratory, http://www.nrel.gov/fleetdna.
56 Controller Area Network (CAN) Overview. National Instruments, http://www.ni.com/white-
paper/2732/en/August 1, 2014.
57 ATRI. (2015). Compendium of Idling Regulations. Updated August 2015.
https://truckingresearch.org/research/idling/ATRI_Idling_Compendium
58 Kotz, A. and K. Kelly (2019). MOVES Activity Updates Using Fleet DNA Data: Interim
Report. NREL/TP-5400-70671. National Renewable Energy Laboratory. Golden, CO. January
2019. https://www.nrel.gov/docs/fyl9osti/70671.pdf.
59 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.c.v "Idle
Reduction", page 73593.
60 National Academies Press. "Commercial Motor Vehicle Driver Fatigue, Long-Term Health,
and Highway Safety: Research Needs." Aug 12, 2016. Available online:
https://www.ncbi.nlm.nih.gov/books/NBK384967/
61 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.
62 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.c.v "Idle
Reduction", page 73593.
63 NACFE (2018). 2018 Annual Fleet Fuel Study. North American Council for Freight
Efficiency, https://nacfe.org/annual-fleet-fuel-studies/.
224
-------
64 Schoettle, B., M. Sivak and M. Tunnell (2016). A Survey of Fuel Economy andFuel Usage by
Heavy-Duty Truck Fleets. SWT-2016-12. The University of Michigan. Sustainable Worldwide
Transportation. October 2016. http://umich.edu/~umtriswt/PDF/SWT-2016-12.pdf.
65 US Government Publishing Office (GPO), Code of Federal Regulations, Title 49 -
Transportation, Vol. 65, No. 85, CFR Parts 350, 390, 394, 395 and 398, 2 May 2000,
http://www.gpo.gov/fdsys.
66 USEPA(2018). 2014 National Emissions Inventory, version2: Technical Support Document.
Office of Air Quality Planning and Standards. Research Triangle Park, North Carolina. July
2018. https://www.epa.gov/sites/production/files/2018-
07/documents/nei2014v2_tsd_05jul2018.pdf.. Section 6.8.4.
67 Lutsey, N., C.-J. Brodrick, D. Sperling and C. Oglesby (2004). Heavy-Duty Truck Idling
Characteristics: Results from a Nationwide Truck Survey. Transportation Research Record,
1880 (1), 29-38. DOI: 10.3141/1880-04.
68 TRB (2019). Guide to Truck Activity Data for Emissions Modeling. Transportation Research
Board. 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.
69 Festin, S., US FHWA, Summary of National and Regional Travel Trends: 1970-1995,
Washington, DC: May 1996. http://www.fhwa.dot.gov/ohim/bluebook.pdf.
70 Motorcycle crash data from NHTSA, Fatality Analysis Reporting System (FARS),
http://www.nhtsa.gov/FARS; raw 2010 data last updated 11 December 2012,
ftp://ftp.nhtsa.dot.gov/fars.
71 Guensler, R., Yoon, S., Li, H., and Elango, V., Georgia Institute of Technology, 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: April 2007,
http://nepis.epa.gov/Adobe/PDF/P100AE2E.pdf.
72 USEPA (2020). Evaporative Emissions from Onroad Vehicles in MOVES 3. EPA-420-R-20-
012. Office of Transportation and Air Quality. US Environmental Protection Agency. Ann
Arbor, MI. November 2020. https://www.epa.gov/moves/moves-technical-reports.
225
-------
73 Sierra Research, Inc., 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: 27 March 27, 2006.
74 US EPA, Development of Evaporative Emissions Calculations for the Motor Vehicle
Emissions Simulator MOVES2010, EPA-420-R-12-027, Ann Arbor, MI: September 2012,
http://www.epa.gov/otaq/models/moves/documents/420rl2027.pdf.
75 US EPA, MOBILE6, https://www3.epa.gov/otaq/models/mobile6/m6tech.htaL
76 Lutsey, N., Brodrick, C., Sperling, D., and Oglesby, C., "Heavy-Duty Truck Idling
Characteristics: Results from a Nationwide Truck Survey," Transportation Research Record
(1880), pg. 29-38, Washington, DC: January 2004, http://trijournalonline.trb.org.
77 Indale, G., University of Tennessee, 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: May 2005,
http://trace.tennessee.edu/utk_graddiss/2085.
78 FHWA. Jason's Law Truck Parking Survey Results and Comparative Analysis.
https://ops.fhwa.dot.gov/freight/infrastructure/truck_parking/jasons_law/truckparkingsurvey/ch3.
htm
79 MOVES User Guide: https://www.epa.gov/moves/moves2014a-latest-version-motor-vehicle-
emission-simulator-moves
80 USEPA (2018). 2014 National Emissions Inventory, version 2: Technical Support Document.
Office of Air Quality Planning and Standards. Research Triangle Park, North Carolina. July
2018. https://www.epa.gov/sites/production/files/2018-
07/documents/nei2014v2_tsd_05jul2018.pdf.https://www.epa.gov/sites/production/files/2020-
04/documents/nei2017_tsd_full_30apr2020.pdf. Section 6.8.4.
81 Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty
Engines and Vehicles—Phase 2, 81 FR 73478, October 25, 2016.
82 National Research Council, Technologies and Approaches to Reducing the
Fuel Consumption of Medium- and Heavy-Duty Vehicles. Table 2-1. Washington, DC: The
National Academies Press. 2010. https://doi.org/10.17226/12845.
226
-------
83 US FHWA, Vehicle Travel Information System. Table W-3: 2013,
http://www.fhwa. dot. gov/ohim/ohimvti s. cfm.
84 US FHWA, Bridge Formula Weights, FHWA-HOP-06-105, August 2006.
85 Sandhu, G. S., H. C. Frey, S. Bartelt-Hunt, and E. Jones. 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/j.atmosenv.2014.04.036.
86 Sandhu, G. S., H. C. Frey, S. B artel t-Hunt, and E. Jones. 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, 2015, pp. 306-323.
https://doi.org/10.1080/10962247.2014.990587
87 Sandhu, G. S., H. C. Frey, S. Bartelt-Hunt, and E. Jones. Real-World Activity, Fuel Use, and
Emissions of Diesel Side-Loader Refuse Trucks. Atmospheric Environment, Vol. 129, 2016, pp.
98-104. https://doi.Org/10.1016/j.atmosenv.2016.01.014.
88 Sandhu, G. S. Evaluation of Activity, Fuel Use, and Emissions of Heavy-Duty Diesel and
Compressed Natural Gas Vehicles. Ph.D. North Carolina State University, Raleigh, NC, USA,
2015. http://www.lib.ncsu.edu/resolver/1840.16/10222
89 US EPA, 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, MI: August 2011,
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=Pl 00O7VJ.pdf.
90 Memorandum to the Docket "FRM - Tractor-Trailer Inputs to MOVES" Docket No. EPA-HQ-
OAR-2014-0827-2221.
91 Memorandum to the Docket "FRM - Vocational Inputs to MOVES" Docket No. EPA-HQ-
OAR-2014-0827-2222.
92 American Public Transportation Association, An Analysis of Transit Bus Axle Weight Issues,
Table 7: November 2014.
93 US Government Publishing Office (GPO), Code of Federal Regulations, Title 40 - Protection
of Environment, Vol. 17, CFR 86.529-78, "Road load force and inertia weight determination," 1
July 2004, http://www.gpo.gov/fdsys.
227
-------
94 Steven, H., United Nations Economic Commisssion for Europe, 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: 28 December
2002, http://www.unece.org.
95 Warila, J., "Derivation of Mean Energy Consumption Rates within the MOVES Modal
Framework," 14th Coordinating Research Council On-Road Vehicle Emissions Workshop
(poster), San Diego, CA: 29-31 March 2004, http://www.crcao.org/workshops/index.html.
96 US EPA, IM240 andEvap Technical Guidance, EPA420-R-00-007, Ann Arbor, MI: April
2000, http://www.epa.gov/otaq/regs/im/r00007.pdf.
97 Final Rule for Model Year 2017 and Later Light-Duty Vehicle Greenhouse Gas Emissions and
Corporate Average Fuel Economy Standards, 77 FR 62624, October 15, 2012.
98 Petrushov, V., "Coast Down Method in Time-Distance Variables," SAE International, SAE
970408, Detroit, MI: 24 February 1997, http://www.sae.org.
99 US EPA, 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, MI: August 2011,
http://www.epa.gov/otaq/climate/documents/420rl 1901 .pdf
100 Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty
Engines and Vehicles— Phase 2. Federal Register Volume 81, No. 206 (October 25, 2016)
Pages 73587, 73608-73611.
101 US EPA, 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, MI: August 2011
102 US EPA, Emission Adjustments for Temperature, Humidity, Air Conditioning, and Inspection
and Maintenance for On-road Vehicles inMOVES2014, EPA-420-R-14-012, Ann Arbor, MI:
December 2014, http://www.epa.gov/otaq/models/moves/documents/420rl4012.pdf.
103 Koupal, J., Air Conditioning Activity Effects in MOBILE6, M6. ACE.001, EPA420-R-01-054,
Ann Arbor, MI: November 2001, http://www.epa.gov/otaq/models/mobile6/r01054.pdf.
104 US Census Bureau, 1997 Vehicle Inventory and Use Survey, EC97TV-US, Washington, DC:
October 1999, https://www.census.gov/svsd/www/vius/1997.html.
228
-------
105 Browning, L., Chan, M., Coleman, D., and Pera, C., ARCADIS Geraghty & Miller Inc.,
Update of Fleet Characterization Data for Use in MOBILE6 - Final Report, EPA420-P-98-016,
Mountain View, CA: 11 May 1998, http://www.epa.gov/otaq/models/mobile6/m6flt002.pdf.
106 Stanard, A., Fincher, S., Kishan, S., and Sabisch, M., Eastern Research Group, Inc. (ERG),
Data Analyses on Drayage Heavy-Duty Vehicles. EPA EP-C-12-017, Work Assignment 0-2,
Austin, TX: 7 December 2012.
107 Davis, S. and Truitt, L., ORNL, Investigation of Class 2b Trucks (Vehicles of8,500 to 10,000
lbs GVWR), ORNL/TM-2002.49, Oak Ridge, TN: March 2002, http://www-
cta. ornl. gov/cta/Publications/Reports/ORNL_TM_2002_49 .pdf.
108 Union of Concerned Scientists (personal communication), http://www.ucsusa.org.
109 Data from 2011 was provided by US FHWA to replace Table II.2 and II.3 in the 1997 Federal
Highway Cost Allocation Study (http://www.fhwa.dot.gov/policy/hcas/final/index.htm)
specifying vehicles by weight (email correspondence), January 2013.
110 US EPA, Use of Data from "Development of Emission Rates for the MOVES Model, " Sierra
Research, March 3, 2010, EPA-420-R-12-022, Ann Arbor, MI: August 2012,
http://www.epa.gov/oms/models/moves/documents/420rl2022.pdf.
111 National Highway Traffic and Safety Administration (NHTSA), "Vehicle Survivability and
Travel Mileage Schedules," DOT HS 809 952, Springfield, VA: January 2006, http://www-
nrd.nhtsa.dot.gov/Pubs/809952.pdf.
112 Davis, S., Diegel, S., and Boundy, R., Oak Ridge National Laboratory (ORNL), Center for
Transportation Analysis, Transportation Energy Data Book (TEDB) Edition 37, Oak Ridge, TN:
2019, http://cta.ornl.gov/data/index. shtml.
113 Motorcycle Industry Council, Motorcycle Statistical Annual, Irvine, CA: 2015,
http s J/www. mi c. org/ StatAnnual. aspx.
114 Ward's Automotive Inc., http://www.wardsauto.com.
115 Liu, H., D. Sonntag, D. Brzezinski, C. R. Fulper, D. Hawkins and J. E. Warila (2016).
Operations and Emissions Characteristics of Light-Duty Vehicles on Ramps. Transportation
Research Record: Journal of the Transportation Research Board, 2570, 1-11. DOI:
doi: 10.3141/2570-01.
229
-------
116 US Government Publishing Office (GPO), Code of Federal Regulations, Title 40 - Protection
of Environment, Vol. 36, CFR 1037.520, "Modeling C02 emissions to show compliance for
vocational vehicles and tractors," 25 October 2016, http://www.gpo.gov/fdsys.
117 US Government Publishing Office (GPO), Code of Federal Regulations, Title 40 - Protection
of Environment, Vol. 36, CFR 1037.515, "Determining CO2 emissions to show compliance for
trailers," 25 October 2016, http://www.gpo.gov/fdsys.
118 North American Council for Freight Efficiency, 2015 Annual Fleet Fuel Study. 6 May 2015,
http://nacfe.org/wp-content/uploads/2015/05/NACFE-2015-Annual-Fleet-Fuel-Study-Report-
050115.pdf
119 Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty
Engines and Vehicles. Federal Register Volume 76, No. 179 (September 15, 2011) Page 57211.
120 Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty
Engines and Vehicles— Phase 2. Federal Register Volume 81, No. 206 (October 25, 2016)
Pages 73587, 73608-73611.
230
------- |