Exhaust Emission Rates for Light-Duty
On-road Vehicles in MOVES2014
Final Report
&EPA
United States
Environmental Protection
Agency
-------
Exhaust Emission Rates for Light-Duty
On-road Vehicles in MOVES2014
Final Report
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.
&EPA
United States
Environmental Protection
Agency
EPA-420-R-15-005
October 2015
-------
Table of Contents
1 Gaseous Exhaust Emissions from Light-Duty Gasoline Vehicles (THC, CO, NO*)
Introduction 7
1.1 MOVES Background 7
1.1.1 Light-Duty Vehicles 7
1.1.2 Differences between MOVES and MOBILE 8
1.1.3 Overview 13
1.2 Emissions Sources (sourceBinID) and Processes (polProcessID) 14
1.2.1 The emissionRateByAge Table 16
1.2.1.1 Age Groups (ageGroupID) 17
1.3 Exhaust Emissions for Running Operation 18
1.3.1 Operating Modes (opModelD) 18
1.3.2 Scope 20
1.3.3 Emission-Rate development: Subgroup 1 (Model years through 2000) 20
1.3.3.1 Data Sources 20
1.3.3.1.1 Vehicle Descriptors 20
1.3.3.1.1.1 Track Road-Load Coefficients: Light-Duty Vehicles 21
1.3.3.1.2 Test Descriptors 22
1.3.3.1.3 Candidate Data Sources 22
1.3.3.2 Data Processing and Quality-assurance 25
1.3.3.3 Sample-design reconstruction (Phoenix only) 26
1.3.3.4 Source selection 27
1.3.3.5 Methods 28
1.3.3.5.1 Data-Driven Rates 28
1.3.3.5.1.1 Rates: Calculation of weighted means 28
1.3.3.5.1.2 Estimation of Uncertainties for Cell Means: 29
1.3.3.5.2 Model-generated Rates (hole-filling) 30
1.3.3.5.2.1 Rates 31
1.3.3.5.2.1.1 Coast/Cruise/Acceleration 32
1.3.3.5.2.1.2 Braking/Deceleration 36
1.3.3.5.2.2 Estimation of Model Uncertainties 37
1.3.3.5.2.3 Reverse transformation 38
-------
1.3.3.5.3 Table Construction 39
1.3.3.6 Verification and Adjustment for High-Power Operating modes 40
1.3.3.7 Estimating Rates for non-I/M Areas 50
1.3.3.8 Stabilization of Emissions with Age 61
1.3.3.8.1 non-I/M Reference Rates 66
1.3.4 Emission-Rate Development: Subgroup 2 (MY 2001 and later) 68
1.3.4.1 Data Sources 68
1.3.4.1.1 Vehicle Descriptors 68
1.3.4.2 Estimating I/M Reference Rates 68
1.3.4.2.1 Averaging IUVP Results 69
1.3.4.2.2 Develop Phase-In Assumptions 74
1.3.4.2.3 Merge FTP results and phase-in Assumptions 78
1.3.4.2.4 Estimating Emissions by Operating Mode 84
1.3.4.2.4.1 Running Emissions 84
1.3.4.2.5 Apply Deterioration 89
1.3.4.2.5.1 Recalculate the logarithmic mean 89
1.3.4.2.5.2 Apply alogarithmic Age slope 89
1.3.4.2.5.3 Apply the reverse transformation 92
1.3.4.2.6 Estimate non-I/M References 93
1.4 Exhaust Emissions for Start Operation 94
1.4.1 Subgroup 1: Vehicles manufactured in model year 1995 and earlier 94
1.4.1.1 Methods 94
1.4.1.1.1 Data Sources 94
1.4.1.1.2 Defining Start Emissions 95
1.4.1.1.3 Relationship between Soak Time and Start Emissions 95
1.4.2 Subgroup 2: Vehicles manufactured in MY1996 and later 97
1.4.3 Applying Deterioration to Starts 98
1.4.3.1 Assessing Start Deterioration in relation to Running Deterioration 98
1.4.3.2 Translation from Mileage to Age Basis 107
1.4.3.3 Application of Relative Multiplicative Deterioration 109
1.5 Incorporating Tier 3 Emissions Standards 115
1.5.1 Averaging FTP Results (Step 1) 116
1.5.2 Develop Phase-In Assumptions (Step 2) 118
-------
1.5.3 Merge Cycle Results and Phase-In Assumptions (Step 3) 122
1.5.4 Estimating Emissions by Operating Mode (Step 4) 126
1.5.5 Apply Deterioration (Step 5) 131
1.5.6 Recalculate the logarithmic mean 132
1.5.7 Apply a logarithmic Age slope 132
1.5.8 Apply the reverse transformation 133
1.5.9 Estimate non-I/M References (Step 6) 133
1.5.10 Start Emissions 133
1.6 Development of Emission Rates representing California Standards 133
1.6.1 Averaging IUVP Results 135
1.6.2 Develop Phase-In assumptions 135
1.6.3 Merge FTP Results and Phase-In assumptions 137
1.6.4 Scaling CA/177 rates to Federal Rates 139
1.6.5 Availability 141
1.6.6 Early Adoption of National LEV Standards 141
1.7 Replication of Rates 142
2 Particulate-Matter Emissions from Light-Duty Vehicles 143
2.1 Introduction and Background 143
2.1.1 Particulate Measurement in the Kansas City Study 144
2.1.2 Causes of Gasoline PM Emissions 147
2.2 New Vehicle or Zero Mile Level (ZML) Emission Rates 149
2.2.1 Longitudinal Studies 150
2.2.2 New Vehicle, or ZML Emission Rates and Cycle Effects 152
2.2.3 Aging or Deterioration in Emission Rates 156
2.2.3.1 Age Effects or Deterioration Rates 157
2.3 Estimating Elemental Carbon Fractions 159
2.4 Modal PM Emission Rates 163
2.4.1 Typical behavior in particulate emissions as measured by the Dustrak and
Photoacoustic Analyzer 164
2.5 Updates to PM2.s emission rates in MOVES2014 170
2.6 Incorporating Tier-3 Emissions Standards for Particulate Emissions 171
2.6.1 Assigning the Tier-2 Baseline 172
2.6.2 Apply Phase-in Assumptions 172
-------
2.6.3 Apply Scaling Fractions 172
2.6.4 Simulate the Extended Useful Life 175
2.7 Incorporating the LEV-III Standard for Particulate Matter 177
2.8 Conclusions 177
3 Gaseous and Particulate Emissions from Light-Duty Diesel Vehicles (THC, CO, NO*, PM)
179
4 Crankcase Emissions 180
4.1 Background 180
4.2 Modeling Crankcase Emissions in MOVES 180
4.3 Light-duty Gasoline Crankcase Emissions 180
4.4 Light-duty Diesel Crankcase Emissions 181
5 Nitrogen Oxide Composition 183
5.1 Light-Duty Gasoline Vehicles 183
5.2 Motorcycles 183
5.3 Light-duty Diesel Vehicles 184
6 References 216
-------
1 Gaseous Exhaust Emissions from Light-Duty Gasoline Vehicles
(THC, CO, N(Xr) Introduction
1.1 MOVES Background
The material presented in this document is a component of a much larger effort, including the
estimation of emission rates for heavy-duty vehicles, estimation of evaporation emissions,
estimation of usage and activity patterns for vehicles, the compilation and storage of all types of
input data in the MOVES database, and the algorithms that combine and process input
information during model runs, translating inputs and modeling assumptions into inventory
estimates.
Readers not familiar with MOVES may find it useful to access additional documentation
providing a broader view of MOVES, the rationale for its development as a replacement for
MOBILE6, and broad overviews of its design.
• The "Initial Proposar for MOVES describes the impetus behind the effort to design a
new inventory model from the ground up, with the goal of developing a tool both more
comprehensive and flexible than its predecessor.l
• A subsequent "Draft Design and Implementation Plan" describes the MOVES design and
introduces the reader to concepts and terminology developed for the new model.2
• Readers wishing to further understand the development of the modal design for running
emissions can consult the "Methodology for Developing Modal Emission Rates, "3 as well
as the "Shoot Ouf4 conducted among several candidate approaches.
• This document focuses on development of inputs to the MOVES Database. Readers
interested in further understanding the processes used by the model to process inputs into
inventory estimates can consult the MOVES Software Design Reference Manual
(SDRM).5
A large volume of additional documentary and supporting materials can be obtained at
http://www.epa.gov/otaq/models/moves/moves-reports.htm. In general, the most recent and
relevant materials are at the top of the page, with older material located further down. However,
as the previous references show, references posted throughout the page are still relevant to the
MOVES model and database in its most recent versions.
1.1.1 Light-Duty Vehicles
This chapter describes the technical development of emission rates for gaseous exhaust
pollutants for light-duty vehicles. These pollutants include total hydrocarbons (THC), carbon
monoxide (CO) and oxides of nitrogen (NO*). The resulting model inputs are included in the
MySQL database supporting the MOVES model.
Light-duty vehicles are defined as cars and trucks with gross vehicle weight ratings (GVWR) of
less than 8,500 Ibs. For purposes of emissions standards "cars" are designated as "LDV" or
"passenger cars" (PC), and are distinguished from "trucks" which are further subclassified as
"light light-duty trucks" (LLDT) and "heavy light-duty trucks" (HLDT), on the basis of GVWR
< 6000 Ibs and GVWR > 6000 Ibs, respectively. The two broad classes, LLDT and HLDT, are
further subdivided into LDT1/LDT2, and LDT3/LDT4. As these subdivisions are highly specific
-------
and technical, we do not describe them here. Interested readers can find more information at
http://www.epa.gov/otaq/standards/weights.htm. As MOVES pools all truck classes for
purposes of inventory estimation, we will refer to "cars" and "trucks" throughout. The
development of motorcycle emission rates in MOVES are covered in a separate report.6
Exhaust emissions from light-duty vehicles have contributed substantially to urban air pollution,
and have received a great deal of scientific, political and regulatory attention over the past forty
years. The Clean Air Act (CAA), passed in 1970 (and amended in 1977 and 1990), set "National
Ambient Air-Quality Standards" (NAAQS) for HC, CO and NOx. Carbon monoxide is targeted
for its respiratory toxicity, and HC and NOx largely for their roles in production of ground-level
ozone, another pollutant targeted under the CAA. Regulations designed to reduce automobile
emissions to facilitate achievement of compliance with the NAAQS include Tier-1 standards
introduced in the mid 1990's, followed by National Low-Emission Vehicle (NLEV) standards
starting in 2001, Tier-2 standards starting in 2004, and Tier-3 standards starting in 2017.
Concurrently, the state of California and additional states electing to adopt "California" in lieu of
"Federal" standards have implemented "LEV-I," "LEV-IT and "LEV-ID" standards.
In addition to introducing more stringent tailpipe standards, requiring introduction of oxygenated
gasolines, and modifying test procedures, the 1990 CAA Amendments expanded requirements
for Inspection-and-Maintenance programs (I/M). The role played by I/M programs in many
urban areas over the past twenty years means that accounting for the existence of such programs
is a primary consideration in modeling tailpipe emissions from light-duty vehicles.
Through a combination of regulation and improved technology, gaseous tailpipe emissions from
light-duty vehicles have declined substantially over the past several decades. Important
milestones in engine and emissions control technology have included the introduction of fuel
injection (replacing carburetion), positive crankcase ventilation (PCV), exhaust gas recirculation
(EGR), catalytic converters, electronic engine controls, and on-board diagnostic systems (OBD).
Development of emission rates thus largely involves constructing a "numerical" account of this
history. However, a detailed account of these developments is beyond the scope of this
document which will focus on the development of emission rates as inputs to the MOVES
database for the purposes of developing mobile-source emission inventories. However, this
history has been well described elsewhere, and we refer interested readers to the USEPA
website7 and to the peer-reviewed literature.8'9'10'11'12'13
1.1.2 Differences between MOVES and MOBILE
At the outset, it is useful to highlight four important differences between MOVES and MOBILE.
(1) While intending to estimate average emissions across the entire vehicle fleet, MOVES does
not distinguish between "normal" and "high emitters," (2) MOVES inverts MOBILE'S approach
to inspection and maintenance, (3) MOVES is a "modal" model, whereas MOBILE is "non-
modal," and (4) emission rates developed for MOVES are expressed in time-specific, rather than
distance specific terms, i.e., mass/time (g/hr), rather than mass/distance (g/mi, g/km).
1. A fundamental difference between MOVES and MOBILE is that MOVES does not classify
vehicles into "emitter classes." The MOBILE model(s) provided different sets of emission rates
for "normal" and "high" emitters. While arbitrary, this distinction made qualitative and practical
sense because the emission rates were themselves averages of FTP test results.
-------
We didn't attempt a similar approach in MOVES for several reasons, some conceptual, some
practical. The main conceptual reason is that in review of data, we did not see clear evidence of
distinct "high emitter" subpopulations, as might be evidenced by observation of bimodal
distributions. Rather, review of emissions data seems to show highly skewed but continuous
distributions, which we treat as log-normal for modeling purposes. Clearly, the vehicles in the
upper percentiles of the distributions make disproportionate contributions to the inventory,
assuming similar driving patterns to cleaner vehicles in the lower percentiles. Based on these
observations, our approach has been to capture the mean of the entire distribution, including the
"upper tail." We illustrate these concepts using two examples, based on aggregate cycle means
from the Phoenix I/M program, measured on the EVI147 cycle.
Figure 1-1 shows cumulative distributions of NO* emissions for "young" cars, aged 0-3 years,
representing two sets of emissions standards. The blue distribution represents "Tier 0" vehicles,
manufactured prior to 1994; the green represents "Tier 1" vehicles, manufactured in 1996-97,
and the red represents a mix of the two, during the Tier-1 phase-in period (1994-95). Note that
the combination of reduced standards and improved technology pushes the entire distribution
"leftward" or towards lower emission levels.
Figure 1-1. Cumulative distributions of running NO* for cars, Age 0-3, measured on the IM147
cycle (Source: Phoenix I/M program).
tiergroup
345
NOx Mass Rate, (g/mi)
Tier—0 Tier—Qffier—1
Tter-1
A similar example, Figure 1-2, shows NO* distributions for Tier-1 vehicles (MY1996-97) at two
different age levels, 0-3 and 8-9 years old, shown in blue and red, respectively. Qualitatively, the
picture looks very similar toFigure 1-1, except that in this case we can see the effect of age in
shifting the entire distribution "rightwards," towards higher emission levels. Note that the entire
distribution shifts, including the lower percentiles, not only the "high emitters" in the upper
percentiles.
-------
Figure 1-2. Cumulative distributions of running NO* for Tier-1 cars, at two age levels, measured
on the IM147 cycle (Source: Phoenix I/M program).
2345
NOx Mass Rate,
Age Group - 0—3 yr
9-9 yr
A pattern not necessarily apparent in the previous figure emerges if we view the same
distributions on a logarithmic scale, as shown in Figure 1-3. In the logarithmic view, we can see
that the distribution at 8-9 years is the same is that at 0-3 years, but shifted to the right; that is,
the shapes (variances) of the two distributions are very similar, but the means are shifted. These
figures illustrate the "logarithmic" or "multiplicative" scaling typical in emissions data. The
utility of logarithms in modeling follows from the fact that multiplicative patterns representing
actual changes can be represented and projected very conveniently as additive changes in
logarithmic space. These patterns obtain whether the data are analyzed with respect to
technology, age or power. The development of emission rates, as described in this chapter (and
for PM in Chapter 2), relies heavily on these concepts. Figure 1-4 shows a similar picture to
Figure 1-2, except for THC; what is notable is that the THC distributions are even more skewed
than the NO* distributions.
10
-------
Figure 1-3. Cumulative distributions of running NO* for Tier-1 cars, at two age levels, measured
on the IM147 cycle (LOGARITHMIC SCALE) (Source: Phoenix I/M program).
0.0001 0.0010 0.0100 0.1000
NOx Mass Rate, (g/rni)
Age Group 0-3 yr 8-9 yr
1.0000
10.0000
11
-------
Figure 1-4. Cumulative distributions of running THC for Tier-1 cars, at two age levels, measured
on the IM147 cycle (Source: Phoenix I/M program).
100:"
on
Qfl J
& :
70-
5- •
c
^ OU"
!T
- V\
it 50
01
•£ 40:
ca
i -v\-
£ >*J
5 2^
in -
o-
/-"
7
/
x- •
— ~- — — —
234567
THC Mass Rate, (g/mi)
Age Group - 0-3 yr - 8-9 yr
10
In addition to the conceptual reason just illustrated, there were practical reasons for not creating
one or more "high-emitter" classes: (1) a vehicle or test showing high emissions for one
pollutant need not show high emissions for other pollutants, (2) high emissions may be a
transitory phenomenon in many cases, i.e., vehicles with high results for one set of
measurements may not show similar results if re-measured; in such cases it is very difficult to
determine whether the apparent change is due to an actual change in the vehicle or the
notoriously high variability of emissions measurements, (3) given that rate development for
MOVES operating modes is not coupled to the FTP (or any particular cycle), convenient and
non-arbitrary definitions of "high emitter" are not readily available, and (4) distinction of emitter
classes would require that the intensive process of rate development be repeated for each class,
including the projection of emissions by age and power, and development of distinct adjustments
for temperature and fuel (performed separately). The detailed data required for these analyses
and their projection into the future is not available.
2. A second important difference between MOVES and MOBILE is that MOVES inverts
MOBILE'S approach to inspection and maintenance. That is, the emission rates provided with
MOBILE represented "non-I/M" conditions, and MOBILE represented I/M conditions by
making adjustments during model runs. By contrast, the MOVES input table contains two sets
of rates, representing "I/M reference" and "non-I/M" conditions, respectively. In development
of these rates, "I/M conditions" were assigned as the default case, and rates representing "non-
I/M" conditions were developed in relation to rates representing "I/M conditions." During model
runs, MOVES represents particular I/M programs as a function of both sets of rates, modified by
12
-------
adjustments calculated to represent the parameters of specific programs. These topics are
discussed in greater detail in 1.3.3.6.
3. A third major difference between MOVES and MOBILE is that MOVES is modal, whereas
MOBILE was not. This feature gives MOVES tremendous flexibility, allowing users to
represent any driving pattern, across a range of temporal and spatial scales. The modal emission
rates are applied consistently at the different analysis scales under which MOVES operates -
national, county and project.
4. Finally, emission rates in MOVES are expressed as "time-specific" rates (mass/time, g/hr), as
opposed to "distance-specific" rates (g/mi), as were rates in MOBILE. With respect to model
design, the purpose for this change was to introduce a measurement basis that would be
applicable to all emissions sources, processes, and operating modes, including those for which a
distance-specific basis is not applicable. Examples include all emissions for nonroad equipment,
which are expressed on a mass/work basis in the NONROAD model (g/hp-hr, g/kW-hr), and idle
or "hotelling" emissions for all sources, which occur while the source is stationary.
1.1.3 Overview
Section 1 describes the structure of the MOVES emissionRateByAge table, as it applies to
gaseous-pollutant emissions from gasoline-fueled light-duty vehicles. The values in this table
describe the "base rates" (meanBaseRate). These values represent mean emissions on the
MOVES reference fuel on a temperature range of 68-86 °F, and unadjusted for the effects of
temperature, humidity, air-conditioning and inspection-and-maintenance programs (I/M). The
adjustments for these factors, applied during MOVES runs, are described in a separate report.35
The emissionRateByAge table includes rates representing start and running operation, defined as
distinct "processes" in MOVES. Rates representing "running operation" are described in Section
1.3, and those for "start operation" are described in section 1.4.
For running emissions, section 1.3.3 describes the development of emission rates for vehicles
manufactured prior to model year 2000. Sub-sections 1.3.3.1 and 1.3.3.2 describe the processes
of data selection and quality assurance. Rates were generated either directly from available data
(see 1.3.3.5.1) or by development and application of statistical "hole-filling" models (see
1.3.3.5.2). These rates were derived using data from the Phoenix I/M program and represent
rates characteristic of a program with features similar to those in the Phoenix program.14 Because
the analyses described in sub-sections 1.3.3 and 1.3.4 relied on data collected on EVI240 and
EVI147 cycles, we thought it appropriate to evaluate the extrapolation with power to high levels
beyond those covered by the I/M cycles. The development and application of adjustments to
rates in operating modes at high power is discussed in sub-section 1.3.3.5.
As mentioned, the rates described in 1.3.3.5 represent emission rates for vehicles under the
requirements of an inspection-and-maintenance program, specifically the program in Phoenix,
AZ, during calendar years 1995-2005. For this reason, we refer to these rates as "I/M reference
rates." With respect to the I/M reference rates, we describe the approach taken to estimating
rates in non-I/M areas, designated as the "non-I/M reference rates", in 1.3.3.7. For runs
representing areas without an I/M program, MOVES uses the non-I/M reference rates. For runs
representing areas with I/M programs, MOVES adjusts the I/M reference rates to account for the
particular aspects of the program(s) represented. It is important to note that the I/M reference
rates assume full compliance with program requirements within the area. MOVES discounts
13
-------
estimated emissions for non-compliance during a model run, which is then represented in the
results.35
We have observed, as have other researchers, that emissions deterioration tends to follow
exponential, or log-linear trends over the first 8-9 years. However, after this point, the trends
enter a declining phase, during which increases in mean emissions continue at a reduced rate. For
the I/M reference rates, we assume that rates stabilize between 12 and 15 years of age. For the
non-I/M reference rates, we assume that they continue to increase at reduced rates through 20+
years of age. The analyses guiding these assumptions are described in 1.3.3.8.
For start emissions, we also applied different methods to different datasets to derive two sets of
rates. For vehicles manufactured in 1995 and earlier, the process of rate development is
described in 1.4.1. For vehicles manufactured in 1996 and later, the process of rate development
is described in 1.4.2. We assume that emissions deterioration affects start as well as running
emissions. Sub-section 1.4.3 describes how we estimate deterioration in start emissions in
relation to deterioration in running emissions.
Note that energy consumption rates for light-duty cars, trucks and motorcycles are documented
in a separate report.15
1.2 Emissions Sources (sourceBinID) and Processes (polProcessID)
In MOVES terminology, pollutants are emitted by "sources" via one or more "processes."
Within processes, emissions may vary by operating mode, as well as by age Group. The relevant
pollutants are the gaseous criteria pollutants: total hydrocarbons (THC), carbon monoxide (CO)
and oxides of nitrogen (NO*). The relevant processes are exhaust emissions emitted during
engine start and running processes, i.e., "exhaust start" and "exhaust running." Combinations of
pollutant and process relevant to this chapter are shown in Table 1-1. For start emissions, the
meanBaseRate is expressed in units of g/start, and for running emissions, the meanBaseRate is
expressed in units of g/hr, which MOVES terminology designates more specifically as "g/SHO,"
where SHO denotes "source-hours operating."
Note that this document describes only emission rates for exhaust hydrocarbons. Modeling of
emission rates for evaporative hydrocarbons is described in a separate report.76
For these pollutants and processes emissions sources include light-duty vehicles (cars and
trucks). Note that the engine-size and weight-class attributes are not used to classify vehicles.
For light-duty vehicles, these parameters are assumed not to influence emissions, as these
vehicles are required to meet applicable standards irrespective of size and weight.
In the emissionRateByAge table, the emissions source is described by a label known as the
"sourceBinID". This identifier is constructed as a "pattern variable" incorporating the attributes
shown in Table 1-2. Assignment of the attributes just described allows assignment of the source-
bin identifier. The identifier is a 19-digit numeric label, of the form "IffiteeyysssswwwwQQ"
where each component is defined as follows:
1 is the literal value "1," which serves as a leading value to set the magnitude of the entire
label,
^represents the fueltypelD,
14
-------
tt represents the engTechID,
ee represents the regClassID,
yy represents the shortModYrGrpID,
ssss represents the engSizelD,
wwww represents the weightClassID, and
00 is the literal value "00," which serves to provide two trailing zeroes at the end of the label.
The individual attributes are assembled in the proper sequence by constructing the sourceBinID
as a pattern variable, where
sourcebinID = 1 x 1018
+ fuelTypeIDx!016
+ engTechID xlO14
+ regClassID xlO12
+ shortModYrGroupID x 1010
+ engSizeID xlO6
+ weightClassID x 102
Equation 1-1
As an example, Table 1-3 shows the construction of sourceBin labels for light-duty gasoline
vehicles, manufactured in model years 1998 and 2010.
Table 1-1. Combinations of pollutants and processes for gaseous pollutant emissions.
pollutantName
HC
CO
NOX
pollutantID
1
2
3
processName
Running exhaust
Start exhaust
Running exhaust
Start exhaust
Running exhaust
Start exhaust
processID
1
2
1
2
1
2
polProcessID
101
102
201
202
301
302
15
-------
Table 1-2. Construction of sourceBins for exhaust emissions for light-duty vehicles.
Parameter
Fuel type
Engine Technology
Regulatory Class
Model-Year group
Engine Size Class
Vehicle Test Weight
MOVES Database Attribute1
fuelTypelD
engtechid
regClassID
shortModYrGroupID
engSizelD
weightClassID
Values
Gasoline = 01
Diesel = 02
E85 = 05
01= "Conventional internal
Combustion"
20 ="Car"(LDV)
30 = "Truck" (LOT)
Varies2
1 as used in the database table "emissionRateByAge."
2 as defined in the database table "modelYearGroup."
Table 1-3 Examples of sourceBinID construction for cars and trucks in model years 1998 and 2010.
fuelTypelD
1 (Gasoline)
1
1
1
engTechID
1 (conventional)
1
1
1
regClassID
20 (Car)
30 (Truck)
20 (Car)
30 (Truck)
shortModYrGroupID
98 (MY 1998)
30 (MY 2010)
98 (MY 1998)
30 (MY 2010)
sourceBinID
101 0120980000000000
101 0130980000000000
101 0120300000000000
101 0130300000000000
1.2.1 The emissionRateByAge Table.
The rates described in this document are stored in the MOVES emissionRateByAge table. This
table includes five fields, as shown in Table 1-4. Consistent with the MOVES modal approach,
the table contains mean base emission rates (meanBaseRate) and associated estimates of
uncertainty in these means for motor vehicles classified as "emissions sources" (sourceBinID),
and by "operating mode" (opModelD). The table includes rates for vehicles inside and outside of
Inspection-and-Maintenance Areas. The uncertainty estimates are expressed as coefficients of
variation for the mean (meanBaseRateCV); this term is synonymous with the "relative standard
error (RSE). In this section, we will describe the processes of data classification by source bin
and operating mode, calculation of mean emission rates, and statistical evaluation of the results.
16
-------
1.2.1.1 Age Groups (ageGroupID)
To account for emissions deterioration, MOVES estimates emission rates for vehicles in a series
of age ranges, identified as age groups (ageGroupID). Seven groups are used, as follows: 0-3,
4-5, 6-7, 8-9, 10-14, 15-19, and 20+ years. The values of the attribute ageGroupID for these
classes are 3, 405, 607, 809, 1014, 1519, and 2099, respectively. These groups assume that the
most rapid change in emissions as vehicles age occurs between 4 and 9 years.
Table 1-4. Description of the EmissionRateByAge table.
Field
SourceBinID
PolProcessID
opModelD
ageGroupID
meanBaseRate
meanBaseRateCV
meanBaseRatelM
meanBaseRatelM
CV
dataSourcelD
Symbol
•^cell
cvf
Description
Source Bin identifier. See Table 1-2
and Table 1-3 and Equation 1-1.
Combines pollutant and process. See
Table 1-1.
Operating mode: defined separately
for running and start emissions. See
Table 1-5.
Indicates age range for specific
emission rates. See 1.2.1.1.
Mean emission rates in areas not
influenced by inspection and
maintenance programs.
Coefficient of variation of the cell
mean (relative standard error, RSE),
for the meanBaseRate.
Mean emission rate in areas subject
to an I/M program with features
similar to the Phoenix program .
Coefficient of variation of the cell
mean (relative standard error, RSE),
for the meanBaseRatelM.
Numeric label indicating the data
source(s) and method(s) used to
develop specific rates.
17
-------
1.3 Exhaust Emissions for Running Operation
Running operation is defined as operation of internal-combustion engines after the engine and
emission control systems have stabilized at operating temperature, i.e., "hot-stabilized"
operation.
1.3.1 Operating Modes (opModelD)
For running emissions, the key concept underlying the definition of operating modes is "vehicle-
specific power" (VSP, Pv). This parameter represents the tractive power exerted by a vehicle to
move itself and its cargo or passengers.1? It is estimated in terms of a vehicle's speed and mass,
as shown in Equation 1-2
„ Avf + BvJ + Cvf + mvfaf _, . _
Pv t = '- '- '- '—!- Equation 1-2
m
In this form, VSP (Pv,/, kW/Mg) is estimated in terms of vehicles':
• speed at time t (vt, m/sec),
• acceleration at, defined as vt - VM, (m/sec2),
• - mass m (Mg) (usually referred to as "weight,"),
• - track-road load coefficients A, B and C, representing rolling resistance, rotational
resistance and aerodynamic drag, in units of kW-sec/m, kW-sec2/m2 and kW-sec3/m3,
respectively.3
This version of the equation does not include a term accounting for effects of road grade,
because the data used in this analysis was measured on chassis dynamometers. Note that during
model operation, MOVES does account for grade when characterizing vehicle activity. For a
description of this process, see the "Vehicle Population and Activity" report.18
On the basis of VSP, speed and acceleration, a total of 23 operating modes are defined for the
running-exhaust process (Table 1-5). Aside from deceleration/braking, which is defined in terms
of acceleration, and idle, which is defined in terms of speed alone, the remaining 21 modes are
defined in terms of VSP within broad speed classes. Two of the modes represent "coasting,"
where VSP < 0, and the remainder represent "cruise/acceleration," with VSP ranging from 0 to
over 30 kW/Mg. For reference, each mode is identified by a numeric label, the "opModelD."
18
-------
Table 1-5. Definition of MOVES operating modes for running-exhaust operation.
Operating
Mode
0
1
11
12
13
14
15
16
21
22
23
24
25
27
28
29
30
33
35
37
38
39
40
Operating Mode
Description
Deceleration/Braking
Idle
Coast
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Coast
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Vehicle-Specific
Power
(VSPt, kW/Mg)
VSP,< 0
0
-------
1.3.2 Scope
In estimation of energy consumption for MOVES2004, it was possible to combine data from
various sources without regard for the residence locations for vehicles measured. In contrast,
when turning attention to the regulated gaseous pollutants, it is essential to know with some
degree of confidence whether vehicles had been subject to inspect!on-and-maintenance (I/M)
requirements at or previous to the time of measurement. After reviewing data sources, it became
clear that the volumes of data collected within I/M areas vastly exceeded those collected in non-
I/M areas. We also concluded that I/M programs themselves could provide large and valuable
sources of data. In consideration of the demanding analytic tasks posed by the ambitious
MOVES design, we elected to estimate rates for vehicles in I/M areas first, as the "base-line" or
"default" condition. Following construction of a set of rates representing I/M "reference"
conditions, the plan was to estimate rates for non-I/M areas relative to those in I/M areas. This
approach is an inversion of that used in MOBILE6, in which "non-I/M" is the "default
condition" relative to which "I/M" emissions are calculated during a model run.
In addition, the rates described below represent emissions on the FTP temperature range (68 - 86
°F) to provide a baseline against which temperature adjustments would be applied during model
runs.
1.3.3 Emission-Rate development: Subgroup 1 (Model years through 2000)
1.3.3.1 Data Sources
For emissions data to be eligible for use in MOVES development, several requirements were
imposed:
• To derive rates for operating modes, it was essential to acquire data measured on
transient tests.
• Data had to be measured at a frequency of approximately 1 Hz, e.g., continuous or
"second-by-second" measurements.
• To make allowance for application of temperature adjustments (developed
separately), it was necessary to know the ambient temperature at the time of test.
• Vehicles were subj ect to I/M program requirements at the time of measurement.
1.3.3.1.1 Vehicle Descriptors
In addition to the requirements listed above, complete descriptive information for vehicles was
required. Vehicle parameters required for incorporation into MOVES are shown in Table 1-6.
20
-------
Table 1-6. Required vehicle parameters.
Parameter
VIN
Fuel type
Make
Model
Model year
Vehicle class
GVWR
Track road-load power
Units
Ib
hp
Purpose
Verify MY or other parameters
Assign sourceBinID, calculate age-at-test
Assign sourceBinID
Distinguish trucks from cars (LDV)
Calculate track road-load coefficients A, B and C
1.3.3.1.1.1 Track Road-Load Coefficients: Light-Duty Vehicles
For light-duty vehicles, we calculated the track load coefficients from the "track road load power
at 50 mph" (TRLP, hp), based on Equation 1-3.19
= PFn-
C = PF,,|TRLHP'C'
Equation 1-3
where:
PFy4 = default power fraction for coefficient^ at 50 mi/hr (0.35),
PFs = default power fraction for coefficient B at 50 mi/hr (0.10),
PFc = default power fraction for coefficient C at 50 mi/hr (0.55),
ci = a constant, converting TRLP from hp to kW (0.74570 kW/hp),
vso = a constant vehicle velocity (50 mi/hr),
ci = a constant, converting mi/hr to m/sec (0.447 m-hr/mi-sec)).
21
-------
In the process of performing these calculations, we converted from English to metric units, in
order to obtain values of the track road-load coefficients in SI units, as listed above. Values of
TRLP were obtained from the Sierra I/M Look-up Table.20
1.3.3.1.2 Test Descriptors
In addition, a set of descriptive information was required for sets of emissions measurements on
specific vehicles. Essential items for use in rate development are listed in Table 1-7.
Table 1-7. Required test parameters.
Parameter
Date
Time of day
Ambient temperature
Test Number
Test duration
Test result
Test weight
Units
°F
sec
pass/fail
Ib
Purpose
Determine vehicle age at test
Establish sequence of replicates
Identify tests on target temperature range
Identify 1st and subsequent replicates
Verify full-duration of tests
Assign tests correctly to pass or fail categories
Calculate vehicle-specific power
1.3.3.1.3
Candidate Data Sources
In addition to the parameters listed in Table 1-6 and Table 1-7, datasets with historic depth and
large sample sizes were highly desirable, to characterize the high variability typical of exhaust
emissions as well as trends with vehicle age.
At the outset, a large volume of emissions data was available, representing over 500,000 vehicles
when taken together (Table 1-8). In some cases they could be combined as broadly comparable
pairs representing I/M and non-I/M conditions. While not all available data could receive
detailed attention, due to limitations in time and resources, a selection of likely candidates was
subjected to a high degree of scrutiny and quality-assurance, after which some were excluded
from further consideration for specific reasons.
22
-------
Table 1-8. Datasets available for use in estimating running emissions from cars and trucks.
Dynamometer
I/M
AZ (Phoenix)
IL (Chicago)
MO (St. Louis)
NY (New York)
non-I/M
Remote-Sensing
I/M
AZ (Phoenix)
IL (Chicago)
MO (St. Louis)
Maryland/N Virginia
GA (Atlanta)
non-I/M
VA (Richmond)
GA (Augusta/Macon)
NE (Omaha)
OK (Tulsa)
Several remote-sensing datasets received consideration. However, we elected not to use remote-
sensing data directly to estimate rates, for several reasons: (1) For the most part, the remote-
sensing datasets on hand had very restricted model-year by age coverage (historic depth), which
severely limited their usefulness in assigning deterioration. (2) The measurement of
hydrocarbons by remote sensing is highly uncertain. The instruments are known to underestimate
the concentrations of many hydrocarbon species relative to other techniques, such as flame-
ionization detectors. In inventory estimation, a multiplicative adjustment of 2.0-2.2 is often
applied to allow comparison to HC measurements by other methods.21 (3) In MOVES, emissions
are expressed in terms of mass rates (mass/time). While fuel-specific rates (mass emissions/mass
fuel) can be estimated readily from remote-sensing data,22 mass rates cannot be calculated
without an independently estimated CO2 mass rate. It followed that remote-sensing would not
provide rates for any MY* Age combinations where dynamometer data were not available. In
these cases, remote-sensing would be dependent on and to some extent redundant with
dynamometer data. (4) Because remote-sensing measurements are typically sited to catch
vehicles operating under light to moderate acceleration, results can describe emissions only
selected cruise/acceleration operating modes. However, remote-sensing cannot provide
measurements for coasting, deceleration/braking or idle modes. For these reasons we reserved
the remote-sensing for additional roles, such as verification of results obtained from
dynamometer data.
23
-------
Table 1-9. Characteristics of candidate datasets.
Criterion
Type
Network
Exempt MY
Collects random
sample?
Program Tests
Fast-pass/Fast-fail?
Test type (for
random sample)
Available CY
Size (no. tests)
Chicago
Enhanced
Centralized
4 most recent
YES
Idle, IM240, OBD-
II
YES
IM240
2000-2004
8,900
Phoenix
Enhanced
Centralized
4 most recent
YES
Idle/SS, IM240,
IM147, OBD-II
YES
IM240, IM147
1995-1999
2002-2005
62,500
NYIPA
Basic/Enhanced
De-centralized
2 most recent
n/a
IM240
n/a
IM240
1999-2002
8,100
St. Louis
Enhanced
Centralized
2 most recent
NO
IM240
YES
n/a
2002-2005
2,200,000
Dynamometer datasets that received serious consideration are described below and summarized
in Table 1-9.
Metropolitan Chicago. We acquired data collected over four calendar years (2000-04) in
Chicago's centralized enhanced program. In addition to routine program tests, the program
performed EVI240 tests on two random vehicle samples. One is the "back-to-back" random
sample. This sample is relatively small (n ~ 9,000 tests), but valuable because each selected
vehicle received two full-duration EVI240 tests in rapid succession, obviating concerns about
conditioning prior to conduction of EVI240 tests. A second is the "full-duration" random sample,
in which selected vehicles received a single full-duration EVI240. This sample is much larger (n >
800,000) but less valuable due to the lack of replication. Despite its size, the full-duration
sample has no more historic depth than the back-to-back sample, and thus sheds little additional
light on age trends in emissions. Both samples were presumably simple random samples,
indicating that in the use of the data, users must assume that the samples are self-weighting with
respect to characteristics such as high emissions, passing/failing test results, etc.
St. Louis. Another large program dataset is available from the program in St. Louis. While a
large sample of program tests is available, this program differed from the others in that no
random evaluation sample was available. Because vehicles were allowed to "fast-pass" their
routine tests, results contained many partial duration tests (31 - 240 seconds). At the same time,
the lack of replication raised concerns about conditioning. Partial duration was a concern in itself
in that the representation of passing vehicles declined with increasing test duration, and also
because it compounded the issue of conditioning. In addition, while OBD-equipped vehicles
failing a scan received IM240s, those passing their scans did not. Because addressing the
interwoven issues of inadequate conditioning, "fast-pass bias" and "OBD-screening bias" proved
intractable, we excluded this dataset from further consideration.
Phoenix. At the outset, the random samples from the Phoenix program appeared attractive in
that they had over twice the historic depth of any other dataset, with model-year x age coverage
spanning 11 calendar years. Usage of these samples is somewhat complicated by the fact that no
random samples were collected for two years (2000-01) and by the fact that the sample design
employed changed in the middle of the ten-year period. During the first four years, a simple "2%
24
-------
random sample" was employed. During the last four years, a stratified design was introduced
which sampled passing and failing vehicles independently and at different rates. In the stratified
sample, failures were over-sampled relative to passing vehicles. Thus, using these data to
estimate representative rates and to combine them with the 2% sample, assumed to be self-
weighting, required reconstruction of the actual stratified sampling rates, as described below.
New York Instrumentation/Protocol Assessment (NYIPA). This dataset differs from the others in
that while it was collected within an I/M area in New York City, it is not an I/M program dataset
as such. It is, rather, a large-scale research program designed to establish correlation between the
EVI240 and an alternative transient test. It is not entirely clear whether it can be considered a
random sample, in part because estimation of representative averages was not a primary goal of
the study. All data that we accessed and used was measured on full-duration IM240s during a
four-year period. There was a high degree of replication in the conduction of tests, allowing
fully-conditioned operation to be isolated by exclusion of the initial test in a series of replicates.
While these data played a prominent role in development of energy consumption rates for
MOVES2004, the four-year duration of the program limits its usefulness in analysis of age
trends for gaseous pollutants.
1.3.3.2 Data Processing and Quality-assurance
We performed several quality-assurance steps to avoid known biases and issues in using I/M data
to estimate mean emissions. One source of error, "inadequate conditioning" can occur when
vehicles idle for long periods while waiting in line. To ensure that measurements used reflected
fully-conditioned vehicles we excluded either portions of tests or entire tests, depending on test
type and the availability of replicates. If back-to-back replication was performed, we discarded
the first test in a series of replicates. If replication was not performed, we excluded the first 120
seconds of tests (for IM240s only).
Another problem occurs when calculation of fuel economy for tests yields values implausible
enough to indicate that measurements of one or more exhaust constituents are invalid (e.g., 300
mpg). To identify and exclude such tests, we identified tests with outlying measurements for
fuel economy, after grouping vehicles by vehicle make, model-year and displacement.
An issue in some continuous or second-by-second datasets is that cases occur in which the
emissions time-series appears to be "frozen" or saturated at some level, not responding to
changes in power. We found that the occurrence of such problems was more or less evenly
distributed among the fleet regardless of age or model year, and that severe instances were rare.
We excluded tests in which 25% or more of the measurements were "frozen."
For a modal analysis assuming that emissions respond to power on short time scales, it is critical
that the emissions time-series be aligned to the power time-series. Consequently, we examined
alignment for all tests. As necessary, we re-aligned emissions time series to those for VSP by
maximizing correlation coefficients, using parametric Pearson coefficients for CO2 and NOX, and
non-parametric Spearman coefficients for CO and THC. For these two species, the trends with
respect to VSP were not linear, nor were distributions of emissions close to normal at any VSP
level. Consequently, we concluded that the Spearman coefficients, as measures of association,
rather than linear correlation, performed as well or better than Pearson coefficients for CO and
THC.
25
-------
1.3.3.3 Sample-design reconstruction (Phoenix only)
For data collected in Phoenix during CY 2002-05, we constructed sampling weights to allow use
of the tests to develop representative means. The program implemented a stratified sampling
strategy, in which failing vehicles were sampled at higher rates than passing vehicles.
It is thus necessary to reconstruct the sample design to appropriately weight failing and passing
vehicles in subsequent analyses. After selection into the random sample, vehicles were assigned
to the "failing" or "passing" strata based on the result of their routine program test, with the
specific test depending on model year, as shown in Figure 1-5. Within both strata, sample
vehicles then received three replicate EVI147 tests.
Based on test records, reconstructing sampling rates simply involved dividing the numbers of
sampled vehicles by the total numbers of vehicles tested, by model year and calendar year, for
failing (f) and passing (p) strata, as shown in Equation 1-4.
" p.MY.CY
Equation 1-4
Corresponding sampling weights indicate the numbers of vehicles in the general fleet represented
by each sample vehicle. They were derived as the reciprocals of the sampling fractions, as shown
in Equation 1-5.
wf,MY,cY - ,. WP,MY,CY - f Equation 1-5
/f,MY,CY /p,MY,CY
26
-------
Figure 1-5. Stratified sampling as applied in selection of the random evaluation sample in the
Phoenix I/M Program (CY 2002-05).
Official Test
MY 1980 and previous: Loaded-mode + Idle
MY 1981 -1995: IM147
MY 1996 and later: ODD II
Failing Stratum
Oversampled
"higher" sampling rate
Passing Stratum
"lower" sampling rate
Triplicate IM147
1.3.3.4
Source selection
After excluding the St. Louis dataset, and comparing the Phoenix, Chicago and NY datasets, we
elected to rely on the Phoenix dataset for purposes of rate estimation and to use the other
datasets, including selected remote-sensing data, for purposes of comparison. This course was
chosen for several reasons.
For our purposes, the greater historic depth of the Phoenix data was a tremendous advantage. It
was the only set deep enough to allow direct and independent assessment of deterioration. The
limited depth of the other datasets would have meant that the subset of calendar years that could
be covered by pooled data would have been relatively limited. Only a single calendar year,
2002, is covered by all three datasets. Several years would be covered by two out of three.
Calendar 1999 is covered by Phoenix and NY; 2000 and 2001 would have been covered by NY
and Chicago, and 2003 and 2004 by Chicago and Phoenix. The remaining years, 1996-98 and
2005 could have been covered only by Phoenix in any case.
In addition, pooling the three datasets would have raised several difficult technical issues that
may not be apparent at first glance. Table 1-9 shows that the datasets were of greatly differing
sizes. Thus, if the datasets were pooled without some type of relative weighting, Phoenix would
have exerted much stronger influence than the others in most shared calendar years. To rectify
disparities in influence by assigning the different datasets similar or proportional influence would
have required development of some sort of a weighting scheme, but a rational basis for such
relative weighting is not immediately apparent.
The question of pooling is further complicated by the fact that use of the Phoenix data collected
in CY 2002 to 2005 requires use of sampling weights for passing and failing tests (as described)
above), whereas the Chicago and NYIPA datasets are assumed to be self-weighting. Again, no
27
-------
rational basis for incorporating weighted and self-weighted tests from various programs in the
same CY was immediately apparent.
Finally, the selection of the Phoenix data provided a relatively consistent basis for specification
of a "reference fuel," and development of associated fuel adjustments.35
1.3.3.5
1.3.3.5.1
Methods
Data-Driven Rates
Where data was present, the approach was simple. We calculated means and other summary
statistics for each combination of sourceBinID, ageGroup and operating mode (i.e., table cell).
We classified the data by regulatory class (LDV="cars", LDT="trucks"), model-year group, age
group and operating mode (Table 1-5). The model-year groups used are shown in Table 1-10,
along with corresponding samples of passing and failing tests.
Table 1-10. Test sample sizes for the Phoenix random evaluation sample (n = no. tests).
Model-year
group1
1981-82
1983-85
1980-89
1990-93
1994-95
1996-98
1996
1997-98
1999-2000
Total
Cars
fail2
562
1,776
3,542
2,897
997
1,330
176
11,285
pass
539
2,078
6,420
8,457
4,422
3,773
753
26,478
Trucks
fail
340
1,124
1,745
1,152
703
526
858
136
6,589
pass
495
1,606
3,698
4,629
3,668
1,196
2,320
624
18,254
1 Note that these are the model-year groups used for analysis; NOT the
model-year groups used in the MOVES database.
2 Note that 'failure' can indicate failure for CO, HC or NO*, as applicable.
We calculated means and other summary statistics for each combination of sourceBinID,
ageGroupID and opModelD. For simplicity, we will refer to a specific combination of
sourceBinID, and opModelD as a "cell" to be denoted by label '/z'.
1.3.3.5.1.1 Rates: Calculation of weighted means
The emission rate (meanBaseRate) in each cell is a (Eh) simple weighted mean
28
-------
Eh = -^ Equation 1-6
where w, is a sampling weight for each vehicle in the cell, as described above, and R,,t is the
"second-by-second" emission rate in the cell for a given vehicle at a given second t.
1.3.3.5.1.2 Estimation of Uncertainties for Cell Means:
A new feature of MOVES is its ability to estimate uncertainty in emissions projections. In the
emissionRateByAge table, uncertainties for individual rates are stored in the
"meanBaseRateCV" fields (Table 1-4). To estimate sampling error for each cell, we calculated
standard-errors by weighted variance components. In estimating variances for cell means, we
treated the data within cells as effective cluster samples, rather than simple random samples. This
approach reflects the structure of the data, which is composed of sets of multiple measurements
collected on individual vehicles. Thus, measurements on a specific vehicle are less independent
of other measurements on the same vehicle than of measurements on other vehicles.
Accordingly, means and variances for individual vehicle tests were calculated to allow derivation
of between-test and within-test variance components. These components were used in turn to
calculate the variance of the mean for each cell, using the appropriate degrees of freedom to
reflect between-test variability.23 To enable estimation of variances under this approach, we
calculated a set of summary statistics, as listed below:
Test mean (Et): the arithmetic mean of all measurements in a given test on a specific vehicle in a
given cell.
Test sample size («/,), the number of individual tests represented in a cell.
Measurement sample size («;•): the number of measurements in a cell representing an individual
test on an individual vehicle.
Cell sample size (nh,i): the total number of individual measurements on all vehicles in a cell,
where each count represents a measurement collected at an approximate frequency of 1.0 Hz,
(i.e., "second-by-second").
Test variance (sj ): the variance of measurements for each test represented in a cell, calculated
as the average squared deviation of measurements for a test about the mean for that test. Thus,
we calculated a separate test variance for each test in each cell.
Weighted Between-Test variance component (sb ): the component of total variance due to
variability among tests in a cell, or stated differently, the weighted variance of the test means
about the cell mean, calculated as
29
-------
Equation 1-7
Weighted Within-Test Variance Component (sw ): the variance component due to variability
within tests, or the variance of measurements within individual tests (Ri,t) about their respective
test means, calculated in terms of the test variances, weighted and summed over all tests in the
cell:
Equation 1-8
2
Variance of the cell mean (•%): this parameter represents the uncertainty in the cell mean, and is
calculated as the sum of the between-vehicle and within-test variance components, with each
divided by the appropriate degrees of freedom.
s2 s2
2 *5u *5W
SE, = ' Equation 1-9
nh nhj
Coefficient-of-Variation of the Mean (CVsh)'. this parameter gives a relative measure of the
uncertainty in the cell mean, allowing comparisons among cells. It is calculated as the ratio of the
cell standard error to the associated cell mean
CV, =
Equation 1-10
Note that the term CVsh is synonymous with the term "relative standard error" (RSE).
1.3.3.5.2 Model-generated Rates (hole-filling)
Following averaging of the data, it was necessary to impute rates for cells for which no data was
available, i.e., "holes." With respect to vehicle age, empty cells occur for age Groups not
covered by available data (Figure 1-6). In the figure, "age holes" are represented by un-shaded
areas. Filling in these un-shaded areas required "back-casting" emissions for younger vehicles
for older model years, as well as "forecasting" deterioration of aging vehicles for more recent
model years. Empty cells occur as well in high-power operating modes not covered by the
EVI147 or EVI240, meaning operating modes with power greater than about 24 kW/Mg.
30
-------
Figure 1-6. Model-year by age structure of the Phoenix I/M random evaluation sample.
MY Vehicle Age at Test (yaars)
O 1 2 3 4 5 6 7 8 9 1O 11 12 13 14 15 16 17 18 19 2O 21 22 23 24
1.3.3.5.2.1 Rates
To estimate rates in empty cells (holes), we constructed statistical models of emissions data to
extrapolate trends in VSP and age. For this purpose, we generated a series of models based on
the MOVES operating-mode/ageGroup structure. Note that the extrapolated values were
modified on a case-by-case basis.
As a preliminary step, data were averaged for each test within a set of classes for VSP and speed.
We averaged emissions by model-year-group, regClass, age, VSP class, speed class and test.
Classes for VSP followed intervals of 3.0 kW/Mg ( e.g., 0-3, 3-6, ... 27-30, 30+). Speed classes
followed those used for the MOVES operating modes (e.g., 1-25 mph, 25-50 mph, 50+ mph).
The resulting dataset had a single mean for each test in each 6-way cell. The purpose for this
averaging was to give the resulting statistical model an appropriate number of degrees of
freedom for each of the class variables, i.e., the d.f. would be determined by the number of tests
rather than the number of individual "second-by-second" measurements. Note that the matrix
used for this purpose was finer than that represented in Table 1-5.
We fit separate models in three groups of operating modes. For all operating modes except
brake/deceleration and idle, we fit one model incorporating VSP. We call this group
"coast/cruise/acceleration." For braking/deceleration and idle, we fit two additional models not
incorporating VSP, as these modes are not defined in other terms (Table 1-5). Overall, we fit
three models for each combination of cars and trucks, for the model-year groups shown in Table
1-10, giving a total of 60 models.
Before fitting a model, we drew a sample of vehicle tests in each model-year group (n = 1,200 to
3,500, see Table 1-11). This sampling was performed to fit models on smaller volumes of data
that a standard desktop computer could handle at the time. The sample was stratified by test
result (pass, fail) and age, with allocation proportional to that in the sample pool. Within each
-------
result age stratum, tests were drawn using simple random sampling, and sampling frequencies
and weights,/strat and w strat, calculated as
«„,
I- _ ' '"strat
J strat ~ i T
1 N.
strat
/strat strat
Equation 1-11
where «stiat and TVstrat are the number of tests selected from a stratum and total number of tests in
the stratum, respectively. Then, for each test selected, a final weight was calculated as the
product of the stratum weight and the initial sampling weight (vcresuit,MY,CY), as shown in Equation
1-5.
~ WresulLMY,CYWstrat
Equation 1-12
Table 1-11. Sample sizes for statistical modeling, by regulatory class and test result.
Model-year
group
1981-82
1983-85
1980-89
1990-93
1994-95
1996-98
1996
1997-8
LDV
fail
645
569
375
260
406
663
pass
554
631
828
944
1,995
1,738
LOT
fail
476
508
343
209
378
346
671
pass
723
691
856
991
2,021
854
1,730
Each model included two sub-models, one to estimate means and one to estimate variances, as
described below.
1.3.3.5.2.1.1 Coast/Cruise/Acceleration
Means model
For the means sub-model, the dependent variable was the natural logarithm of emissions
In Eh = /?„ + /?jPv + /?2JPV2 + /?3JPV3 + 04a + fi5S + £!6PVS + J1tl + S Equation 1-13
32
-------
where :
• \nEh = natural-logarithm transform of emissions (in cell //),
• Pv, Pv2, Pv3 = first-, second- and third-order terms for vehicle-specific power
(VSP, kW/Mg),
• a = vehicle age at time of test (years),
• s = speed class (1 -25 mph, 25-50 mph and 50+ mph),
• t = test identifier (random factor)
• e= random or residual error
• ft = regression coefficients for the intercept and fixed factors Pv, a and s.
• 7 = regression coefficients for the random factor test.
The model includes first-, second- and third-order terms in Pvto describe curvature in the power
trend, e.g., enrichment for CO and the corresponding decline in NO* at high power. The age
term gives an In-linear trend in age. The speed-class term allows for a modified intercept in each
speed class, whereas the power/speed-class interaction allows slightly different power slopes in
each speed class. The random factor term for test fits a random intercept for each test, which
does not strongly affect the mean estimates but does affect the estimation of uncertainties in the
coefficients.
After fitting models, we performed basic diagnostics. We plotted residuals against the two
continuous predictors, VSP and age. We checked the normality of residuals across the range of
VSP and age, and we plotted predicted vs. actual values.
Variances model
The purpose of this sub-model was to model the variance of InE1/,, i.e., the logarithmic variance
si2, in terms VSP and age. To obtain a dataset of replicate variance estimates, we drew sets of
replicate test samples. Each replicate was stratified in the same manner as the larger samples
(Table 1-11). To get replicate variances, we calculated In-variance for each replicate within the
VSP/age matrix described above.
Models were fit on set of replicate variances thus obtained. The dependent variable was
logarithmic variance
sf = a{} + CCfl + a2Pv + a3Pva + S Equation 1-14
where Pv and a are VSP and age, as above, and a are regression coefficients. After fitting we
examined similar diagnostics as for the means model.
1.3.3.5.2.1.1.1 Model application
Application of the model involved several steps. The first step was to construct a cell matrix
including all emission rates to be calculated, as shown in Table 1-12.
33
-------
Table 1-12. Construction of emission-rate matrix for light-duty gasoline vehicles.
X
X
X
X
X
=
Count
1
2
10
21
7
3
9,660
Category
Fuel (gasoline)
Regulatory Classes (LDV,
LOT)
Model-year groups
Operating modes
Age Groups
Pollutant processes (running
HC, CO, NO,)
TOTAL cells
MOVES Database attribute
fuelTypeID = 01
regClassID = 20, 30
As in Table 1-11
opModeID= 11-16,21-30,33-
40
ageGroupID = 3, 405, 607,
809, 1014, 1519,2099
polProcessID = 101,201,301
Next, we constructed a vector of coefficients for the means sub-model (P) and merged it into the
cell matrix.
P = [A) A A A3 A4 A5(0-25) A5(25-50) A5(50+) A6
Equation 1-15
Then, for each table cell, we constructed a vector of predictors (X/,). Equation 1-16 shows an
example for an operating mode in the 1-25 mph speed class, e.g., the value for the 1-25 mph
class is 1 and the values for the 25-50 and 50+ speed classes are 0. To supply values for VSP
(Pv) and age group (a), cell midpoints were calculated and applied as shown in Table 1-13.
Equation 1-16
Xfc = 1 R, R, R a 1 0 0 R,
34
-------
Table 1-13. Values of VSP used to apply statistical models.
opModelD
11,21
12,22
13,23
14,24
15,25
16
27,37
28,38
29,39
30
40
33
35
Range
<0
0-3
3-6
6-9
9-12
12 +
12-18
18-24
24-30
30 +
30 +
<6
6-12
Midpoint
-2.0
-2.5
4.5
7.5
10.5
14.5
15.0
21.0
27.0
34.0
34.0
0.5
9.0
The final step was to multiply coefficient and predictor vectors, which gives an estimated
logarithmic mean (InEV) for each cell h.
Equation 1-17
The application of the variances model is similar, except that the vectors have four rather than
nine terms
Equation 1-18
Equation 1-19
Equation 1-20
Pv aPva]
Thus, the modeled logarithmic variance in each cell is given by
In some model-year groups, it was not always possible to develop plausible estimates for the age
slope /?4, because the data did not cover a wide enough range of calendar years. For example, in
the 99-00 model-year group, the available data represented young vehicles without sufficient
35
-------
coverage of older vehicles. We considered it reasonable to adapt the age slope for the 96-98
model-year group for cars, and the 1997-98 model-year group for trucks.
In the groups 83-85 and 81-82, the data covered vehicles at ages of 10 years and older but not at
younger ages. Simply deriving slopes from the available data would have given values that were
much too low, resulting in very high emissions for young vehicles. In these cases we considered
it more reasonable to adopt an age slope from a subsequent model year group. When making this
assumption, it is necessary to recalculate the intercept, based on the assumed slope and the
earliest available data point.
Intercepts, denoted as/?o*, were recalculated by rearranging Equation 1-13 to evaluate the model
in operating mode 24, using the age slope from the previous model-year group (/?4*) and an
estimate of In-emissions from the available dataset at the earliest available age (InEa*) at age a*.
In operating mode 24, the midpoint of the VSP range (6-9) is 7.5 kW/Mg and the speed class is
25-50 mph.
Pi =ln£a. -1.5/3, -1.52J32 -7.53/?3 -#a*-&(25_50) -1.5/36 Equation 1-21
On a case by case basis, age slopes were adopted from earlier or later model-year groups. In a
similar way, In-variance models or estimates could be adopted from earlier or later model years.
1.3.3.5.2.1.2 Braking/Deceleration
1.3.3.5.2.1.2.1 Means model
We derived models similar to those used for coast/cruise/acceleration. For these operating
modes, however, the models were much simpler, in that they did not include VSP or the speed
classes used to define the coast/cruise/accel operating modes. Thus, emissions were predicted
solely in terms of age, although random intercepts were fit for each test as before:
\nEh=P0+/3la + y1ti+S Equation 1-22
1.3.3.5.2.1.2.2 Variances model
In addition, we fit variances models for these operating modes, which were also simple functions
of age.
sf = a0 + a^a + £ Equation 1-23
1.3.3.5.2.1.2.3 Model_application
In these operating modes, rates were to be modeled for a total of 840 cells. This total is
calculated as in Table 1-12, except that the number of operating modes is 2, rather than 21. We
set up coefficient and predictor vectors, as before.
For the means and variances sub-models the vectors are
36
-------
and
= [ /?„ /?! ]
= [ 1 a ]
Equation 1-24
Equation 1-25
respectively.
For the variances model the coefficients vector is
tt =
! ]
Equation 1-26
and the predictor vector is identical to that for the means model.
As with coast/cruise/accel modes, we considered it reasonable in some model-year groups to
adopt a slope or In-variance from a previous or later model-year group. In model-year groups
where the purpose was to backcast rates for younger vehicles, rather than forecast rates for aging
vehicles, it was again necessary to recalculate the intercept based on a borrowed age slope and an
estimate of InEh calculated from the sample data for the youngest available age class. In this
case, Equation 1-27 is a rearrangement of Equation 1-22.
Pi = In Ea
-------
'4,0
crn
• <
cr,
6,4
r5(0-25)
Cr5(25-50)
'0,4'
Equation 1-28
Using the parameter vectors XA and the covariance matrix s/, the standard of error of estimation
for each cell was calculated as
— X,
Equation 1-29
The standard error of estimation in each cell represents the uncertainty of the mean estimate in
the cell, based on the particular values of the predictors defining the cell.24 The pre- and post-
multiplication of the covariance matrix by the parameter vectors represents the propagation of
uncertainties, in which the parameters represent partial derivatives of each coefficient with
respect to all others and the co-variances represent the uncertainties in each coefficient in relation
to itself and the others.
1.3.3.5.2.3
Reverse transformation
To obtain an estimated emission rate Eh in each cell, the modeled means and variances are
exponentiated as follows
EH =
Equation 1-30
The two exponential terms use the results of the means and variances sub-models, respectively.
The left-hand "means" term represents the geometric mean, or the center of the implied log-
normal distribution, whereas the right-hand "variance" term reflects the influence of the "high-
emitting" vehicles representing the tail of the distribution.
The estimate of In-variance could be obtained in several different ways. The first and preferred
option was to use the modeled variance as described above. A second option was to use an
estimate of variance calculated from the available sample of In-transformed data. A third option,
also based on available data, was an estimate calculated from averaged emissions data and the
mean and variance of In-transformed emissions data. This process involves reversing Equation
1-30 to solve for sj1. If the mean of emissions data is xa and mean of In-transformed data is xi,
then the logarithmic variance can be estimated as
38
-------
Equation 1-31
In practice one of these options was selected based on which most successfully provided model
estimates that matched corresponding means calculated from the data sample.
The uncertainties mentioned above represent uncertainties in InE1/,. Corresponding standard errors
for the reverse-transformed emission rate Eh were estimated numerically by means of a Monte-
Carlo process. At the outset, we generated a pseudo-random set of 100 variates of InE1/,, based on
a normal distribution with a mean of 0.0 and variance equal to sinE2. We applied Equation 1-30
to reverse-transform each variate, and then calculated the variance of the reverse-transformed
variates. This result represented the variance-of-the-mean for Eh ( SE ), as in Equation 1-9.
Finally, we calculated the CV-of-the-mean (CVsh) for each modeled emission rate, as in
Equation 1-10.
1.3.3.5.3 Table Construction
After compilation of the modeling results, the subset of results obtained directly from the data
(Equation 1-6 to Equation 1-10), shaded area in Figure 1-6) and the complete set generated
through modeling (Equation 1-13 to Equation 1-31) were merged. A final value was selected
for use in the model data table. The value generated from data was retained if two criteria were
met: (1) a subsample of three or more individual vehicles must be represented in a given cell («/,
> 3), and (2) the CVsh (relative standard error, RSE) of the data-driven Eh must be less than 50%
(CV^ < 0.50). Failing these criteria, the model-generated value was substituted. For purposes of
illustration, results of both methods are presented separately.
At this point, we mapped the analytic model-year groups onto the set of model-year groups used
in the MOVES database. The groups used in the database are designed to mesh with heavy-duty
standards and technologies, as well as those for light-duty vehicles. To achieve the mapping, we
replicated records as necessary, in cases where the analytic group was broader than the database
group. Both sets of groups are shown in Table 1-14.
39
-------
Table 1-14. Mapping "analytic" model-year groups onto MOVES-database model-year groups
"Analytic"
Cars
1981-82
1981-82
1983-85
1983-85
1986-89
1986-89
1990-93
1990-93
1994-95
1994-95
1996-98
1996-98
1996-98
1996-98
1996-98
Trucks
1981-82
1981-82
1983-85
1983-85
1986-89
1986-89
1990-93
1990-93
1994-95
1994-95
1996
1997-98
1997-98
1997-98
1997-98
"MOVES database"
1980 and previous
1981-82
1983-84
1985
1986-87
1988-89
1990
1991-1993
1994
1995
1996
1997
1998
1999
2000
modelYearGroupID
19601980
19811982
19831984
1985
19861987
19881989
1990
19911993
1994
1995
1996
1997
1998
1999
2000
shortModYrGroupID
1
61
62
85
63
64
90
65
94
95
96
97
98
99
20
1.3.3.6 Verification and Adjustment for High-Power Operating modes
The rates described were derived from data measured on IM240 or IM147 cycles, which are
limited in terms of the ranges of speed and vehicle-specific power that they cover. Specifically,
these cycles range up to about 50 mph and 24 kW/Mg for speed and VSP, respectively. Some
coverage does exist outside these limits but can be sporadic and highly variable. The operating
modes outside the I/M window include modes 28,29,30, 38, 39 and 40, which we'll refer to as
the 'high-power' operating modes. For these modes, the statistical models described above were
used to extrapolate up to about 34 kW/Mg.
Based on initial review and comment on this aspect of the analysis, we thought it advisable to
give additional scrutiny to the high power extrapolation. To obtain a framework for reference,
we examined a set independently measured data, collected on drive cycles more aggressive than
the EVI cycles, namely, the US06 and the "Modal Emissions Cycle" or "MEC." Much of the data
was collected in the course of the National Cooperative Highway Research Program (NCHRP)25
and the remainder on selected EPA programs, all stored in OTAQ's Mobile-Source Observation
Database (MSOD). Unlike the US06, which was designed specifically to capture speed and
acceleration not captured by the FTP, the MEC is an "engineered" cycle, designed not to
represent specific driving patterns, as does the FTP, but rather to exercise vehicles through the
ranges of speed, acceleration and power comprising the performance of most light-duty vehicles.
Several variants of the MEC were developed to provide a database to inform the development of
40
-------
the Comprehensive Modal Emissions Model (CMEM).25 Driving traces for the US06 and MEC
cycles are shown in Figure 1-7 and Figure 1-8. Both cycles range in speed up to over 70 mph
and in VSP up to and exceeding 30 kW/Mg.
Figure 1-7. Example speed traces for the US06 and MEC cycles.
80
200
400 600
800 1000 1200 1400 1600 1800 2000
Time (sec)
Drive CyclG rrec ^^~ us06
41
-------
Figure 1-8. Example vehicle-specific-power (VSP) traces for the US06 and MEC cycles.
40:
-30
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (sec)
Drive CyclG rrec ^^~ us06
Table 1-15 summarizes the numbers of available tests by regulatory class, model-year group and
drive cycle, with numbers of tests differing in each model-year group. Samples were somewhat
larger for cars for both cycles, which represented a broad range of model-years.
Table 1-15. Sample sizes for US06 and MEC cycles (No. tests).
Model-year group
1 980 & earlier
1981-85
1986-89
1990-93
1994-95
1996-99
Total
Car
US06
4
15
21
54
49
58
201
MEC
14
23
24
57
45
28
191
Truck
US06
8
13
22
22
56
121
MEC
6
19
31
36
30
17
139
Total
24
65
89
169
146
159
652
Figure 1-9, Figure 1-10 and Figure 1-11 show trends in emissions vs. VSP for CO, HC and NO*
for LDV and LDT by model year group. Both cycles were averaged and plotted as aggregates.
42
-------
Figure 1-9. CO emissions (g/sec) on aggressive cycles, vs. VSP, by regulatory class and model-year
group.
7-
6-
-10
10
20
VSP (kW/tonne)
30
Reg/MYG
— LDT-0060
LDT-9495
LCV-8683
1 LOT-8185 LOT- 868=1
1 LDT-9699 LDV-0080
LCV-9093 LDV-9495
40
LOT—9093
LCW-8185
LCW-9699
Figure 1-10. THC emissions (g/sec) on aggressive cycles, vs. VSP, by regulatory class and model-
year group.
0.15:
0.14:
0.13;
0.12-
0.11J
0.10:
0.09:
0.08:
0.07:
0.06:
0.05:
0.04:
0.03-
0.02:
0.01:
0.00:
Fteg/MYG
10 20
VSP (kW/tonne)
• LDT-Cmj
1 LDT-9495
LEW-8683
1 LOT-8185
1 LDT-aSi«
LCW-9093
LDT-8589
1 — LDV-OO80
1 — LDV-9495
LDT-9093
LDV-8185
LDV-9699
43
-------
Figure 1-11. NO* emissions (g/sec) on aggressive cycles, vs. VSP, by regulatory class and model-
year group.
0.21-
0,20-
0.19-
0,18-
0.17-
0.16-
0,15-
0,14-
0,13-
0,12-
0,11 -
0,10-
0.09-
ooe-
0,07-
0,05:-
0,05-
0,04-
0,03-
0,02-
0.01-
0.00-
Reg/MYG
• LDT-0080
• LDT-9495
20 30 40
VSP (kW/tonne)
• LOT-8185 LDT-8S89 — LDT-
1 LDT-9699 LDV-CC80 LCW-
LCW-90S3 LDV-9495 LCW-
To construct a basis for reference, we averaged the data by regulatory class, model-year group
and operating mode, using the model-year groups shown in Table 1-15. After averaging, we
calculated ratios from high-power operating modes to a selected reference mode. Specifically,
we selected two modes covered by the EVI cycles (27 and 37) to serve as reference points. The
midpoint VSP for each is -15 kW/Mg. With mode 27 as a reference, we calculated ratios to
modes 28, 29 and 30.
= —**-, for/ =28,29,30
Equation 1-32
and with mode 37 as a reference, we calculated ratios to modes 38, 39 and 40.
37=-, for/ =38,39,40
Equation 1-33
After calculating the ratios, we calculated ratio-based emissions estimates (ER) as the products of
their respective ratios and the initial rate for modes 27 or 37
44
-------
?™27fl/, or E*. =R,.37E'^jal Equation 1-34
respectively, where Ehimtial is the initial data-driven or model-generated rate calculated as
previously described.
The next step, the process by which ratio-based rates were selected as rates for particular
operating modes on a case-by-case basis has changed substantially for the final rates. In the draft,
we calculated upper and lower confidence limits for ER and replaced the initial rate with ER if it
fell outside the confidence band, i.e., if the initial rate was greater than the upper bound or lower
than the lower bound. Evaluation of the results of this approach showed, however, that it gave
spurious results in many cases. We found it impossible to assign a confidence level for the band
that would work in all cases, i.e., sufficiently sensitive to identify and correct problem cases, but
not so sensitive so as to make unnecessary modifications.
For the final rates, we developed a different logic for applying the ratio-based rates. One change
from the draft is that ratio-based rates were considered only for modes 29,30, 39 and 40, i.e.,
modes spanning the range of VSP beyond the EVI147. Modes 28 and 38 are partially covered by
the I/M cycles, and the differences among the data, model and ratios were generally much
smaller than for the four highest modes. The steps in the revised process are:
1) Identify acceptable candidate values (data, model or ratio). The data values were considered
acceptable if (1) a value was present, (2) it met the acceptability criteria (described above) and
(3) it was greater than the value in the next lowest mode. Similarly, predicted values were
acceptable if they exceeded the value for the preceding operating mode.
Following these evaluations, the final value was selected as the minimum of the acceptable
candidates. These criteria were applied sequentially to prevent declining emissions trends with
increasing power. As a first step, values were selected for operating modes 29 and 39, relative to
modes 28 and 38. In a successive step, values were selected for 30 and 40, relative to those
selected for 29 and 39, respectively. We present some examples below, showing differences
between the draft and final rates.
In the THC example (Figure 1-12), the final values are substantially reduced, particularly for
modes 29 and 30. In the draft (a), the initial rates fall outside the confidence intervals for the
ratio-based rates for three out of six possible cases, i.e., in modes 30, 39 and 40. The resulting
rate is higher for modes 30 and 40, but lower for 39. In the final rates, the results vary. For
modes 29 and 30, the data values meet the criterion of the minimum value giving an increasing
trend from mode 28 - 30. However, for modes 39 and 40, the ratio and the model give the
values meeting the criterion, as shown in (c).
The example for CO shows different behavior in the draft, but a similar outcome in the final
(Figure 1-13). In the draft (a), the initial values for modes 28-30 all fall within the confidence
intervals for the ratio-based value and are thus retained. The values for 39 and 40, fall outside the
band on the low side and are replaced by the ratio-based rates. For operating modes 29 and 30,
the data is selected as the minimum option available, as with HC. For modes 39 and 40, the
model is similarly selected. In the final rates, the ratio based values are not adopted for this
45
-------
example, as they had been in the draft, and the net result is a decrease in CO rates in the affected
operating modes.
Finally, in the NO* example (Figure 1-14), the initial rates are replaced in five out of six cases in
the draft (a). The initial values for 28-30 and 40 all fall below the lower confidence limit,
whereas that for 30 falls above the upper confidence limit. In the final, the ratio is used more
sparingly, as in the HC and CO examples. Model values are used in two cases (modes 30 and
40) and the ratio in one case (mode 39).
These examples highlight the uncertainty of projecting emissions at high power and of projecting
beyond the range of the IM147. Uncertainties are much smaller for opModes 28 and 38 than for
29, 30, 39 and 40. This pattern may be due to the fact that, for modes 28 and 38, the power range
for the EVI147 overlaps somewhat the range of the aggressive cycles. For this reason, the degree
of extrapolation is lower and the power trends are similar.
46
-------
Figure 1-12. THC emission rates (g/hr), vs. VSP for MY 1998 cars at ages 4-5 years: (a) options for
draft rates, (b) options for final model (data, model and ratio) and (c) options selected for final
rates.
U J
(a) Draft
2
..•i
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 2
E
/
/
/'
<
93
I
T
•yf
...-•*
0 33 35 37 38 39 40
60
% 40
:>
-a
8 20-
E
10
0
(b) Final: options available
'
i
i
f
f
, *
. * • . • * . * *
1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
opModelD
18
17
16
15
» 14-
.= 13
51 12
2 11 -
£ 10-
S. B-
£ a-
2 7
g 6
1 4-
3
2
1
0
(c) Final: options
' - . •
t
»
selected
•
•
i
'
. . •
•
t
t
»
•
>
1
n
•
. *
*
1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
opModelD
Source • • • Dora A A A Modol n r- n RatjO
47
-------
Figure 1-13. CO emission rates (g/hr), vs. operating mode for MY-1998 trucks at ages 6-7: (a)
options for draft rates , (b) options for final model (data, model and ratio and (c) options selected
for final rates.
n -
/
J
I
r
/ *
i />
»»»+»»*»»•»»»* **»*
•» Initial
— B — ratio
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
sooo:
4000:
f= 3000:
2000:
)F
ina
options i
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
opModelD
1400:
CD :
^ 900:
jD 800:
1 50°:
(c)
• • *
Fir
lal
OF
t *
itio
ns
se
ec
ted
• • *
4
1
-
1
1
i
i
1
L
*
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
opModelD
Source • * • Data A * ^ Model
48
-------
Figure 1-14. NO* emission rates (g/hr) vs. operating mode for MY-1995 Cars at ages 8-9: (a)
options for draft rates , (b) options for final model (data, model and ratio and (c) options selected
for final rates.
500
450
400
350
300
250
200
150
100
50
0
(a) Draft
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
400
- 300
(b) Final: options available
i i
• •
, •
t 4
t
i
t
1
•
. 1 •
i
•
\
A
i
t
.
'-
«
.
i
t
1
|
t
,
«
'
L
.
1 11 12 13 14 15 16 21 22 23 24 25 27 26 29 30 33 35 37 3S 39 40
opModelD
300-
§> 200
3
1
1
5s 100-
E
o-
(c) Final: options selected
1
'
»
<
• t m •
«
•
1
1
1
'
I
.*
.
*
.
1 V *
1 11 1Z 13 14 15 IB 21 ZZ 23 24 25 Z7 Z8 29 3D 33 35 37 33 39 40
opModel D
Source •» * EJalu * * * Model Kriiti
49
-------
1.3.3.7 Estimating Rates for non-I/M Areas
In modeling emission inventory for light-duty vehicles, it is necessary at the outset to consider
the question of the influence of inspect!on-and-maintenance (I/M) programs. In this regard a
fundamental difference between MOVES and MOBILE is that MOVES inverts MOBILE'S
approach to representing I/M. In MOBILE, the emission rates stored in the input data tables
represent non-I/M conditions. During a model run, as required, emissions for I/M conditions are
modeled relative to the original non-I/M rates.
In MOVES, however, two sets of rates are stored in the input table (emissionRateByAge). One
set represents emissions under "I/M conditions" (meanBaseRatelM) and the other represents
rates under "non-I/M conditions" (meanBaseRate). The first set, representing vehicles subject to
I/M requirements, we call the "I/M reference rates". The second, representing vehicles not
subject to I/M requirements, we call the "non-I/M reference rates."
For the I/M reference rates, the term "reference" is used because the rates represent a particular
program, with a specific design characteristics, against which other programs with differing
characteristics can be modeled. Thus, the I/M references are, strictly speaking, regional rates, and
not intended to be (necessarily) nationally representative. Development of the I/M reference rates
is discussed in 1.3.3.1 to 1.3.3.5. As the I/M references represent Phoenix, the program
characteristics implicitly reflected in them include:
• A four-year exemption period,
• transient tailpipe tests for MY 81-95,
• OBD-II for MY 96+,
• Biennial test frequency.
In addition, the Phoenix program provides a relatively stable basis against which to represent
other program designs and for application of fuel adjustments.
Our approach is to derive the non-I/M rates relative to the I/M references, by adjustment. One
reason for adopting this approach is that, as mentioned, the volumes of data available in I/M
areas vastly exceed those collected in non-I/M areas. An additional practical reason is that major
work-intensive steps such as "hole-filling" and projection of deterioration need only be
performed once.
In contrast to the I/M references, the non-I/M reference rates are designed to be nationally
representative. Broadly speaking, they are intended to represent all areas in the country without
I/M programs. In general, estimating the influence of I/M areas on mean emissions is not trivial,
and efforts to do so commonly follow one of two broad approaches. One approach is to compare
emissions for two geographic areas, one with and one without I/M, as shown in Figure 1-15(a).
A second and less common approach is to compare emissions between two groups of vehicles
within the same I/M area, but with one group representing the main fleet ostensibly influenced
by the program, and the second, far smaller, representing vehicles measured within the program
but presumably not yet influenced by the program, as shown in Figure 1-15(b).
50
-------
Figure 1-15. General approaches to estimating differences attributable to I/M programs: (a)
comparison of subsets of vehicles between two geographic areas, and (b) comparison within a
program area.
(a) Comparison between a program Area
and a non-program area
(b) Comparison within a program
For convenience, we refer to the first approach as the "between-area" approach, and the second
as the "within-area" approach. Neither approach attempts to measure the incremental difference
attributable to a program from one cycle to the next.
The approach we adopted emphasizes the "within-area" approach, based on a sample of vehicles
"migrating" into Phoenix. To lay the basis for comparison, the primary goal was to identify a set
of vehicles that had been measured by the program after moving into the Phoenix area, but that
had not yet been influenced by the program. The specific criteria to identify particular migrating
vehicles are presented in Table 1-16.
51
-------
Table 1-16. Criteria used to identify vehicles migrating into the Phoenix program.
logic
OR
AND NOT
AND
AND
Criterion
The vehicle comes from out-of-state
from a non-I/M county in AZ
from other I/M areas
receiving very first test in Phoenix program
selected for the random evaluation sample
After applying these criteria, we identified a sample of approximately 1,400 vehicles. The origin
of vehicles entering the Phoenix Area was traced by following registration histories of a set of
approximately 10,000 candidate vehicles. The last registered location of vehicles was identified
prior to registration in Phoenix or the vehicle's first test in the Phoenix program. Vehicles were
excluded if their most recent registration location was in a state or city with an I/M program.26
Figure 14 shows the distribution of incoming vehicles, by Census Region. Most vehicles
migrating to Phoenix came from the Midwest (47%), followed by the South (32%), the West
(20%) and the Northeast (1%). The low incidence from the NE may be attributable to the large
number of I/M programs in that region.
Figure 1-16. Geographic distribution of vehicles migrating into the Phoenix I/M area, 1995-2005.
WEST
Pacific
Mountain
MIDWEST
West East
North Central North Central
NORTHEAST
Mddte New
Atlantic England
52
-------
To assess the differences between migrating (non-I/M) and "local" (I/M) vehicles, we adopted a
simple approach. We calculated ratios between means for the migrating and local groups, as
shown in Equation 1-35. We used aggregate tests, after preliminary analyses suggested that the
ratios did not vary significantly by VSP. Because the sample was not large in relation to the
degree of variability involved, we also aggregated tests for cars and trucks in all model years.
However, we did calculate ratios separately for three broad age groups (0-4, 5-9, and 10+) years.
77
Ratio = ™"-I/M Equation 1-35
^I/M
For purposes of verification, we compared our results to previous work. An initial and obvious
comparison was to previous work based on an out-of-state fleet migrating into Phoenix that
provided a model for our own analysis.7 This previous effort identified a migrating fleet, and
analyzed differences between it and the program fleet for vehicles in model years 1984 - 1994
measured during calendar years 1995-2001. To adapt the previous results for our purposes, we
translated averages for migrating and program fleets into ratios as in Equation 1-35.
Another valuable source for comparison was remote-sensing data collected in the course of the
Continuous Atlanta Fleet Evaluation (CAFE) Program.27'28 Unlike our own analysis, this
program involves a comparison between two geographic areas. The "I/M area" is the thirteen-
county Atlanta area, represented by measurements for approximately 129,000 vehicles. The other
(the non-I/M area) is the twelve-county non-I/M area, surrounding Atlanta, represented by
measurements for approximately 28,000 vehicles. Both areas have been under a low-sulfur fuel
requirement since 1999. Results used for this analysis were collected during CY 2004. The non-
I/M : I/M ratios calculated from the remote-sensing are based on concentrations, rather than mass
rates.
A third source was an additional remote-sensing dataset collected in N. Virginia/D.C. area.
The I/M area was the "northern-Virginia" counties, and the non-I/M area was Richmond. The
I/M and non-I/M areas were represented by about 94,000 and 61,000 vehicles, respectively,
collected in CY 2004. In this case, the molar ratios were converted to mass rates, with use of
fuel-consumption estimates derived from energy-consumption rates in MOVES2004. After this
step, non-I/M : I/M ratios were calculated using the mass rates.
Results are shown in Figure 1-17. The charts show mean ratios for the three age groups for our
migrating vehicle analysis, as well as the remote-sensing studies. The diamonds represent
approximate values from Wenzel's earlier work with the Phoenix data. For our analyses (solid
bars) the ratios are generally lower for the 0-4 year age Group, and larger for the 5-9 and 10+ age
groups, but differences between the two older groups are small. The Atlanta results show a
similar pattern for HC and NO*, but not for CO, for which the ratios are very similar for all three
age groups. The Virginia results are the other hand, show increasing trends for CO and HC, but
not for NOx. The ratios in Atlanta are slightly higher than those for Phoenix in the 0-4 year age
group. This difference may be attributable to the shorter exemption period in Atlanta (2 years)
vs. the four-year period in Phoenix, but it is not clear that these differences are statistically
significant. In all three programs, ratios for the two older age classes generally appear to be
statistically significant.
53
-------
In interpreting the ratios derived from the Phoenix data, it is important to note that they assume
full program compliance. In the migrating vehicle analysis this is the case because all emissions
measurements were collected in I/M lanes. Thus, vehicle owners who evaded the program in one
way or another would not be represented. On the whole, results from multiple datasets, using
different methods, showed broad agreement.
If we calculate non-EVI reference rates from the I/M references by ratio, with the ratios constant
by model-year group and VSP, it follows that the absolute differences must increase with power.
Similarly, absolute differences increase with age, for two reasons. The first reason is the same as
that for VSP, that for a constant ratio, the absolute difference increases as emissions themselves
increase, and on top of this, the second reason is that the ratios themselves increase with age
(Figure 15). A third implication is the absolute differences would be smaller for successive
model-year groups as tailpipe emissions decline with more stringent standards.
54
-------
Figure 1-17. Non-I/M : I/M ratios for CO, HC and NO* for the Phoenix area (this analysis)
compared to remote-sensing results for Atlanta and N. Virginia, and previous work in Phoenix
(diamonds).
0-4
5-9
Age Class
10+
i AZ I/M ''/A GA RSD (CY04) = VA RSD (CY04)
lAZI/M _GARSD(CY04) • VA RSD (CY34)
1.60
CM 5-9 10+
Age Class
i AZ I/M • GA RSD (CMD4) • VA RSD (CM34)
A final practical step is to translate these results into terms corresponding to the MOVES age
groups. As mentioned, the program in Phoenix has a four-year exemption period for new
vehicles. However, it is not uncommon for other programs to have shorter exemptions; for
example, both the Atlanta and N. VA programs have two-year exemptions.
55
-------
An additional factor is that the coarser age groups used for the migrating-vehicle analysis don't
mesh cleanly with the MOVES age groups. It was therefore necessary to impute values to the
first two MOVES age groups (0-3 and 4-5 years). We achieved this step by linearly interpolating
the value for the 5-9 year age Group to a value of 1.0 at 0 years of age, as shown in Figure 1-18.
To anchor the interpolation, we associated the value of the ratio for the 5-9 year age group with
the midpoint of the group (7.5 years). Then, based on a straight line interpolation, we imputed
values for the 0-3 and 4-5 MOVES age groups, by taking the value on the line associated with
the midpoint of each class, 1.5 and 5 years, respectively.
Figure 1-18. Imputation of non-I/M ratios for the 0-3 and 4-5 year MOVES ageGroups by linear
interpolation from the midpoint of the 5-9 year analysis age group.
1.0-
0 - 4 years
5 - 9 years
1.5
5.0
7.5
0
h
|2 |3
0-3 years
|4
|5 |6
4-5 years
|7 |8
6-7 years
|9
8-9 years
Figure 1-19 shows final values of the non-I/M ratios for CO, THC and NO*, with error-bars
representing 95% confidence intervals. The values for each pollutant start at 5.0% and increase
with age, stabilizing at maximum values at 6 years (for NO*) and 10 years (for HC and CO).
56
-------
Figure 1-19. Final non-I/M ratios for CO, HC and NO*, by MOVES ageGroups, with 95%
confidence intervals.
1.60
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
-(a) GO
0-3 4-5 6-7 8-9 10-14
Age Class
15-19
20+
0.00
0-3
4-5
6-7
8-9 10-14
Age Class
15-19
20+
1.60'
1.40'
1.20'
1.00'
: 0.80'
0.60'
0.40'
0.20'
0.00'
--(c)NQx
0-3
4-5
6-7
8-9 10-14
Age Class
15-19
20+
57
-------
The ratios shown in Figure 17 are applied to the I/M reference rates to derive non-I/M reference
rates.
£/,,non-i/M = Ratio * Ek,iM Equation 1-36
The uncertainty in Eh,non-i/u was calculated by propagating the uncertainty in the Ratio with that
of the corresponding I/M rate EMU.
I r>F V I r^F V
2 _ ^/i.non-I/M 2 , ^/i.non-I/M 2
*S 77- — *S n ~r *S ^
-OA.non-IM C\n K 377 -^A.I/M „ . ^ ^_
^ Oft ) \ v^h,iM ) Equation 1-37
Thus, for any given cell //, the uncertainty in the non-I/M reference rate is larger than that for the
corresponding I/M reference rate, which is reasonable and appropriate given the additional
assumptions involved in developing the non-I/M reference rate.
Figure 1-20 shows an example of the reference rates vs. operating mode, for all three pollutants.
Note that not all the modes are shown, to allow examination of differences between non-I/M and
I/M rates at lower VSP. Figure 1-21 shows corresponding trends by age for two operating
modes. The first is opmode 11, (speed = 1-25 mph, VSP <0 kW/Mg) and 27 (speed = 25-50
mph, VSP = 12-18 kW/Mg). A clear observation from both plots is that the I/M difference is
much larger in the more aggressive mode (27) than in the less aggressive one (11), with the
inference that I/M differences will be more strongly expressed for more aggressive than less
aggressive driving, in absolute (but not relative), terms.
58
-------
Figure 1-20. Non-I/M and I/M reference rates by operating mode (example: cars, MY 1994, at 8-9
years of age).
'Cr fin
€
5 50
S
s.
c ^n
o -30
'g 20-
m 10 |
n
• I/M Reference •
• non-l/M Reference
(_\ TLJ/"*
a) THu i
•
• ;
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
opModelD
5 2500 :
01 :
c 1 *=;nn -
0 IOUU :
g 1000 :
m 500 :
0 1
*I/M Reference
• non-l/M Reference *
/u\ /~>O
(D) UU H
»*
m^ rr^ rr^
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
opModelD
300
^ 250
150
100
4 I/M Reference
non-l/M Reference
(c) NOx
0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
opModelD
59
-------
Figure 1-21. Non-I/M and I/M reference rates vs. age for two operating modes (example: cars,
MYG 1994).
40
35
30 i-
-• I/M Ref:opMode11
- - D- - • non-l/M Ref: Mode 11
I/M Ref: Mode 27
non-l/M Ref: Mode 27
10 15
Age (years)
800
700
600
500
400
300
200
100
0
-•«•-• I/M Ref: opMode 11
- D - • non-l/M Ref: Mode 11
I/M Ref: Mode 27
non-l/M Ref: Mode 27
10 15
Age (years)
uj 40
20
0
• -0- - • I/M Ref: opMode 11
• D- - • non-l/M Ref: Mode 11
-• I/M Ref: Mode 27
-•—non-l/M Ref: Mode 27
i a-
10 15
Age (years)
20
25
60
-------
1.3.3.8 Stabilization of Emissions with Age
One characteristic of the data is that fleet-average emissions do not appear to increase
indefinitely with age, but rather tend to stabilize at some point between 12 and 15 years of age.
This behavior is visible in datasets with enough historical depth for age trends to be observable,
including the Phoenix random sample and long-term remote-sensing studies.12 Figure 1-22 and
Figure 1-23 show age trends by model year for cars and trucks, respectively. The values shown
are aggregate mass rates over the EVI147 expressed as g/sec for CO, THC and NO*.
Incorporating stabilization of emissions with age is another departure with the approach used on
MOBILE, which allowed emissions to increase indefinitely.
From these figures, as well as Figure 1-6 (page 31), it is clear that no data were available at ages
older than 10 years of age for model years later than 1995, and that no data was available at ages
older than 15 years for model years older than 1990. Thus for model years more recent than
about 1995 it was necessary to project emissions for ages greater than 8-10 years.
However, it is not appropriate to simply extrapolate the statistical models past about 8-10 years.
As described above, emissions were modeled as In-linear with respect to age, which implies
exponential trends for reverse-transformed values. However, exponential trends will increase
indefinitely if extrapolated much beyond the range of available data, which obviously does not
describe observed patterns of fleet emissions. To compensate for this limitation, we employed a
simple approach to represent the decline and stabilization of the rates.
We calculated ratios of means between the 10-14 and 15-19 year ageGroups, each relative to the
8-9 year age group, using the 1986-89 and 1990-93 model-year groups, which contain data for
vehicles as old as 19 years. For this purpose we used Phoenix data averaged by MOVES model-
year and age groups, as shown in Figure 1-24. Data points in the figure represent aggregate tests
(g/mi). After averaging by model-year group and ageGroup, we calculated ratios of means for
the 10-14 and 15-19 ageGroups.
77 77
-^10-14 n -^15-19 ^ .• 1 *,o
- •> Kage ~ — Equation 1-38
8-9
We calculated modified rates for the 10-14 and 15-19 year ageGroups as the product of the rate
for the 8-9 year ageGroup and the corresponding ratio (.Rage). Assuming that emissions would be
fully stable by 20 years, we set the rate for the 20+ year ageGroup equal to that for the 15-19
year ageGroup. We calculated variances for the ratios as in Equation 1-37.
61
-------
Table 1-17. Ratios used to stabilize emission rates for the 10-14 and 15-19 year ageGroups,
calculated relative to the 8-9 year ageGroup.
Regulatory Class
Cars
Cars
Trucks
Trucks
ageGroup
10-14
15-19
10-14
15-19
Ratios (Rage)
THC
1.338
1.571
1.301
1.572
CO
1.226
1.403
1.220
1.479
NO,
1.156
1.312
1.156
1.312
Variances (VR)
THC
0.000000032
0.00000411
0.00000173
0.0000518
CO
0.000160
0.00268
0.000758
0.0666
NO,
0.00000009
0.00000261
0.00000138
0.0000499
62
-------
Figure 1-22. Aggregate IM147 emissions (g/sec) for cars, by model year and age, for the Phoenix
random evaluation sample.
LDV, WEIGHTED
CO vs. Age (years)
Vehicle age (years)
1S67
1333
1953
1991
1997
2003
LDV WEIGHTED
vs. Age (years), LDV
LDV WEIGHTED
NOx vs. Age (years), LDV
Vehicle age (years)
1391
1997
2003
63
-------
Figure 1-23. Aggregate IM147 Emissions (g/sec) for trucks, by model year and age, for the Phoenix
random sample.
LDT, WBGHTED
CO vs, Age (yaais)
LX»T. WBOhTTED
Tl C vs. Aga (yoars), LDT
»aj3™SJ8&KJ
wso
OH
frfJi
20XJ ' -*• ZIKM
LDT WBQhTTED
Nfix vs, Aga (yaars), LDT
(c) NOx
o - f<
ft R RJ K ft
SJD(B -"—'—I- HIM
64
-------
Figure 1-24. Aggregate IM147 emissions (g/mi) by model-year group and age group.
30.0
10 15
Age (years)
20
25
10 15 20
Age (years)
25
2.5
2.0 '- (c) NOx
10 15
Age (years)
20
25
-•-81-82
-A-83-85
-*-86-89
-*-90-93
-•-94-95
-•-99-00
—96-98
-•-81-82
-A-83-85
-*-86-89
HK-90-93
^^94-95
-•-99-00
-•-96-98
-•-81-82
-A-83-85
-X-86-89
HK-90-93
-•-94-95
-•-99-00
-•-96-98
65
-------
1.3.3.8.1 non-I/M Reference Rates
The ratios developed in 1.3.3.8 are assumed to apply in I/M areas, as the underlying data was
collected in the Phoenix I/M area. It is therefore plausible that the patterns observed may be
reflective of I/M areas. The program places some pressure on high-emitting vehicles to improve
their emissions, leave the fleet, leave the area, or, it could be added, evade requirements in some
way. However, in the absence of a program, high-emitting vehicles are not identified and owners
have less incentive to repair or replace them. Thus, the question arises as to whether
deterioration patterns would necessarily be identical in non-I/M as in I/M areas. Two plausible
scenarios can be proposed. In the first, the pattern of deterioration followed by stabilization is
similar in non-I/M as in I/M areas, but emissions stabilize at a higher level, and perhaps at a later
age. In the second, emissions continue to increase in non-I/M areas, but at a slower rate after 10-
15 years.
Data that sheds light on these questions are very limited, as the datasets with sufficient history
were collected within I/M areas. Thus, given the absence of information, we adopted an
assumption that, absent the existence of a program, emissions would increase after 19 years. We
applied this assumption by assuming that the ratio observed between the 10-14 and 15-19 year
ageGroups would persist in linear fashion from the 15-19 to the 20+ year ageGroups.
Table 1-18 shows the deterioration stabilization ratios for both the I/M and non-I/M references
rates. As mentioned above, all ratios are applied by multiplication by values for the 8-9 year
ageGroup in all operating modes. The ratios for I/M areas (7?age,i/M) are identical to those in Table
1-17. The center column shows the ratio of values of R&^,\iu for the 15-19 to the 10-14 year
ageGroups. Ratios for the non-I/M references (7?age,non-i/M ) are identical to those for I/M in the
10-14 and 15-19 year ageGroups. In the 20+ year ageGroup, the non-I/M ratio is equal to the
product of the 15-19 value and the ratio of the 15-19 and the 10-14 values.
66
-------
Table 1-18. Deterioration-stabilization ratios as applied to I/M and non-I/M reference rates.
Pollutant
THC
CO
NO,
Regulatory
Class
Cars
Trucks
Cars
Trucks
Cars
Trucks
ageGroup
10-14
15-19
20+
10-14
15-19
20+
10-14
15-19
20+
10-14
15-19
20+
10-14
15-19
20+
10-14
15-19
20+
Raffs.VM1
1.338
1.571
1.571
1.301
1.572
1.572
1.226
1.403
1.403
1.220
1.479
1.479
1.159
1.312
1.312
1.159
1.312
1.312
Ratio (15-19: 10-14)
1.174
1.206
1.144
1.213
1.132
1.132
-/Vage,non-I/M
1.338
1.571
1.845
1.301
1.572
1.898
1.226
1.403
1.606
1.220
1.479
1.795
1.159
1.132
1.486
1.159
1.132
1.486
1 Values in this column are identical to those in Table 1-17.
2 Calculated as the ratio of the values in the current and previous rows.
3 for 10-14 and 15-19 year ageGroups, values in this column identical to the I/M column; for the 20+ year ageGroup, values
in this column equal the product of the value in the previous row (15-19) and the value in the center column.
67
-------
1.3.4 Emission-Rate Development: Subgroup 2 (MY 2001 and later)
1.3.4.1
Data Sources
Data for vehicles in model years 2001 and later was acquired from results of tests conducted
under the In-Use Verification Program. This program, initiated in 2003, is run by manufacturers
and administered by EPA/OTAQ through the Compliance Division (CISD).
To verify that in-use vehicles comply with applicable emissions standards, customer-owned
vehicles at differing mileage levels are tested on an as-received basis with minimal screening.
Emissions are measured on the Federal Test Procedure, US06 and other cycles. The FTP is most
relevant to our purposes, but the US06 is also important.
1.3.4.1.1 Vehicle Descriptors
In addition to the parameters listed above in Table 1-7, the IUVP data provides test-group
(formerly engine family) information. Using test group, the IUVP files can be merged with
certification test records by model year. The certification test records provide information on
standard level and specific emissions standards applicable to each vehicle. The standard level
refers to the body of standards to which vehicles were certified (Tier 1, NLEV, LEV-I, LEV-II),
and the standards refer to specific numeric standards for HC, CO or NO*, where HC are
represented by non-methane hydrocarbons (NMHC) or non-methane organic gases (NMOG),
depending on combinations of standard level and vehicle class (LDV, LOT 1-4).
Table 1-19. Vehicle descriptors available in IUVP files and certification test records.
Parameter
VIN
Fuel type
Make
Model
Model year
Test group1
Tier
Emissions Standard
Source
IUVP
Y
Y
Y
Y
Y
Y
Cert. Records
Y
Y
Y
Y
Y
Y
Purpose
Verify MY or other parameters
Assign sourceBinID, calculate age-at-test
Assign Vehicle Class
'Formerly "engine family."
Combining data from both sources allows individual test results to be associated with the correct
standard level and emissions standard, allowing inference of the correct vehicle class.
1.3.4.2 Estimating I/M Reference Rates
The goal of this process is to represent I/M reference rates for young vehicles, i.e., the first
ageGroup (0-3 years). The rates are estimated by Tier, model year and regulatory class. The
process involves six steps, each of which is discussed in more detail in Section 1.3.4.2.1, below.
68
-------
1. Average IUVP results by standard level and vehicle class.
2. Develop phase-in assumptions for MY 2001 - 2017, by standard level, vehicle class and
model year.
3. Merge FTP results and Phase-in assumptions. For running emissions, calculate weighted
ratios of emissions in each model year to those for Tier 1 (MY2000). We assumed that the
emissions control at high power (outside ranges of speed and acceleration covered by the FTP)
would not be as effective as at lower power (within the range of speed and acceleration covered
by the FTP).
4. Estimate Emissions by Operating Mode. Then calculate emissions by operating mode in each
model year by multiplying the MY2000 emission rates by the weighted ratio for each model
year.
5. Apply Deterioration to estimate emissions for three additional age Groups (4-5, 6-7 and 8-9).
We assume thatNLEV and Tier-2 vehicles will deteriorate similarly to Tier-1 vehicles, when
viewed in logarithmic terms. We therefore apply In-linear deterioration to the rates developed in
steps 1-4. For the remaining three groups, emissions are assumed to stabilize as described above
on page 61.
6. Estimate non-I/Mreference rates. The rates in steps 1-6 represent I/M references.
Corresponding non-I/M references are calculated by applying the ratios applied to the Tier-1 and
pre-Tier-1 rates (Figure 1-19).
Each of these steps is described in greater detail in the sub-sections below.
1.3.4.2.1 Averaging IUVP Results
In using the IUVP results, "cold-start" emissions are represented as "Bag 1 - Bag 3" i.e., the
mass from the cold-start phase less that from the corresponding hot-start phase. Similarly, "hot-
running" emissions are represented by the "Bag 2," or the "hot-stabilized" phase, after the initial
cold-start phase has conditioned the engine.
The first step is to average the IUVP results by Tier and vehicle Class. Results of this process
are shown below. In the figures, note that the HC values represent non-methane hydrocarbons
(NMHC) for Tier 1 and non-methane organic gases (NMOG) for NLEV and Tier 2. Figure 1-25
shows FTP composite results in relation to applicable certification and useful-life standards. For
THC and NO*, the data show expected compliance margins in the range of 40-60% in most
cases. For CO, compliance margins are even larger, ostensibly reflecting the concomitant effects
of HC or NO* control on CO emissions.
Figure 1-26 shows results for separate phases of the FTP, to examine differential effects of
standards on start and running emissions. As mentioned, the "cold-start" emissions are
represented by the difference between Bags 1 and 3, divided by the nominal bag distance (3.59
miles) which expresses the values as a "start rate" in g/mi. The "hot-running" emissions are
represented by Bag 2 emissions, also divided by the appropriate distance to obtain an aggregate
rate, in g/mi. Additionally, Figure 1-27 shows composite, start and running values normalized to
their respective Tier-1 levels, which clearly displays the greater relative levels of control for
running as opposed to start emissions. Not surprisingly then, distinguishing start and running
69
-------
emissions shows that composite FTP values for HC and CO are strongly influenced by start
emissions. Starts are also important for NO*, but to a lesser degree. In any case, the results show
that sole reliance on composite results in projecting future emissions declines would give
misleading results in projecting either start or running emissions. Hence, the method described
below emphasizes treating them separately.
70
-------
Figure 1-25. Composite FTP Results for Tier 1, NLEV and Tier 2 passenger cars (LDV), as
measured by IUVP, in relation to corresponding certification and useful-life standards.
Oonn
1 0.250 -
2 n onn
§ u.zuu -
'«
w n 1 ^n
UJ
Q. n inn
u_
n nnn
10.000
.~ 9.000
£ 8.000
•2 7.000
c 6.000
•5; 5.000
~ 4.000
W 3.000
£ 2.000
"- 1.000
0.000
n vnn
~ 0.600 -
0 500
(A
§ 0.400 -
'in
E
LU n onn
D.
/a\ LJf^
• \"/ rlw - - Q - -Certification Standards (g/mi)
ci
1~\ . - Q - . Useful Life Standards (g/mi)
*, *t_ — A FTP Composite estimates (g/mi
^v D-.''- .•[1--
xr~ '-R'- -' ^ D Q---.
* — . "•"-'•p' A "a n
"^------^t---^* A A A ^ ^
T1 TLEV LEV ULEV binS bin? bin6 binS bin4 binS
T1 NLEV NLEV NLEV T2 T2 T2 T2 T2 T2
__
bin2
T2
(D) OU -- D- -• Certification Standards (g/mi)
- - D- - • Useful Life Standards (g/mi)
— A FTP Composite estimat
n n oX ,-',n D — n — -n.x
'*-Vrr'-'' '••"n n
^ y ' '•y Q - -
~~~- A i A
* A- — •* — ^ ^ i A ±_
T1 TLEV LEV ULEV binS bin? bin6 binS bin4 binS
T1 NLEV NLEV NLEV T2 T2 T2 T2 T2 T2
(f*\ M O Y
- - Q - • Certification Standards (g/n
Useful Lite Standards (g/m
1-1 i-i — A FTP Composite estimates (
\
4 A »
^sT ""••n-.
es (g/mi
- -n
--•a
A
bin2
T2
ii)
)
g/mi)
71
-------
Figure 1-26. Cold-start (Bag 1 - Bag 3) and hot-running (Bag 2) FTP emissions for Tier 1, NLEV
and Tier 2 passenger cars (LDV), as measured by IUVP (g/mi).
^ .4UU -
E U.ooU -
5 0300
w
o 0 250 -
w n 9nn
uj U.loU -
P- n 1 nn
u_
On^n
1 9 nnn
~ 1 o 000
3 Q nnn
•- R nnn
in 4 nnn
0.
t 9 nnn
0.000
1 nnn
Oonn
g U.oUU -
B) n ynn -
w
w n ?nn
i- n 9nn -
01 nn
Onnn
— * — Cold start (g/mi) — • — Hot Running (g/mi)
5^"—
(a) HC \ /^* — — * * *^^
_ ^^^
T1 TLEV LEV ULEV bin8 bin? bin6 bin5 bin4 bin3 bin2
T1 NLEV NLEV NLEV T2 T2 T2 T2 T2 T2 T2
^^^
^v > Cold start (g/mi) — • — Hot Running (g/mi)
\^
fh\ rn ^^ v»^
^V ^\^
^*^
T1 TLEV LEV ULEV bin8 bin? bin6 bin5 bin4 bin3 bin2
T1 NLEV NLEV NLEV T2 T2 T2 T2 T2 T2 T2
^^ 4 Cold start (g/mi) M Hot Running (g/mi)
\
\
(c) NOx ^_
' *~^-^_
^^^^.^^^
• •— . ^^^-^^^
~1~^~ — • • • * _ _ ^ * *
T1 TLEV LEV ULEV bin8 bin? bin6 bin5 bin4 bin3 bin2
T1 NLEV NLEV NLEV T2 T2 T2 T2 T2 T2 T2
72
-------
Figure 1-27. Composite, cold-start (Bag 1 - Bag 3) and hot-running (Bag 2) FTP emissions for Tier
1, NLEV and Tier 2 passenger cars (LDV), as measured by IUVP, normalized to respective Tier-1
levels.
1.000
0.900
0.800
0.700
0.600 -
0.500 -
0.400 -
0.300 -
0.200 -
0.100
0.000
1 .UUU
~~ n Qnn
> n onn
— ' 0 700
*•" 0 ROO -
|2 0.500 -
O U.4UU
** 0 300 -
*= 0.200 -
iv1 n -i nn
0 000 -
"^r^-^_
V ^
\
\
\v.
\
— 4 — Composite — •— Cc
X^
*\
^^A \. v^»^^
^>v \ ^S^^^^^M 1
^xN^^*^ ^^^^
Id Start A Hot Running
^-^
* -* ^ N? ? f- *--\.
(h)
T1
T1
CO
TLEV
NLEV
LEV ULEV bin8 bin? bi
NLEV NLEV T2 T2 T
-*^^S.
"*
16 bin5 bin4 bin3 bin2
2 T2 T2 T2 T2
0)
H
O
+J
o
«
a:
1.000
0.900
0.800 -
0.700 -
0.600 -
0.500 -
0.400 -
0.300 -
0.200 -
0.100 -
0.000
73
-------
1.3.4.2.2 Develop Phase-In Assumptions
To estimate emissions levels for specific model years, we developed assumptions describing the
phase-in of new emissions standards after model year 2000. For rates stored in the MOVES
default database, we developed assumptions intended to apply to vehicles sold in states where
Federal, rather than California standards applied. Thus, the phase-is designed to represent the
phase-in of National-Low-Emission-Vehicle (NLEV) and Tier-2 standards.
To achieve these steps, we obtained certification records and test results for a selection of model
years.29 These records contain information on certified vehicles, including model year, test
group (engine family), standard level (Tier-1, LEV, Bin 5, etc.), and sales area, as well as
numerical standards used for certification on the Federal Test Procedure (e.g., 0.05 g
NMOG/mile, etc.). For each engine family, we inferred the vehicle class (LDV, LDT1-LDT4)
based on combination of standard and numerical values. Examples illustrating this process are
shown in Table 1-20.
After compiling lists of engine families by standard, model year and vehicle class, we obtained
estimates of final sales from the EPA VERIFY database for MY 2001-2007.30 We merged the
certification records with the sales estimates, by model year and engine family.
Then to estimate the default "Federal" phase-in, we summed the sales by model year, standard
level and vehicle class, for a subset of sales areas in which Federal or California standards
applied, excluding those sales areas in which only California standards applied. Estimates of
numbers of engine families certified for various sales areas are listed in Table 1-21. Sales-
weighted phase-in scenarios for each vehicle class are shown in Figure 1-28 through Figure 1-31.
As noted, the results in the Figures reflect the certifications in the "Fed" or "Both" groups shown
in Table 1-21.
Proportions of each standard represent actual phase-in history for MY 2001-2007. We projected
phase-in assumptions through MY2010, after which we held assumptions constant, under
assumption that the Tier-2 phase-in would be complete.
The National LEV (NLEV) standards apply only to LDV, LDT1 and LDT2 vehicle classes, for
which Tier 1 certification ended in MY 2000. Certification to NLEV standards began in 2001
and ended in 2006, however, NLEV vehicles dominate the (Federal) fleet between 2001 and
2003. Tier 2 vehicles enter the fleet in 2003 and completely comprise new sales by 2010.
The phase-in for LDV, LDT1 and LDT2 are broadly similar in that LEV and Bin 5 vehicles
dominate certifications and sales. There are relatively small differences in that LDV-T1 contains
higher fractions of ULEV and Bin 8.
The phase-in for heavy light-duty trucks is simpler in that Tier-1 certifications continue through
2004, after which Tier-2 standards are introduced. After 2003, certifications are dominated by
Bin 8, Bin 5 and Bin 4.
74
-------
Table 1-20. Examples of information obtained from certification test records, with vehicle class
inferred from combinations of standard, and FTP certification values.
Standard
LEV
LEV
Tierl
Tierl
Engine Family
2HNXV02.0VBP
2MTXT02.4GPG
2CRXT05.95B2
2CRXT05.96BO
Sales Area
NLEV all states
NLEV all-states
Federal all-altitude
Federal all-altitude
FTP Standard
50,000-mi
0.075
0.100
0.32
0.39
100,000-mi
0.09
0.13
120,000-mi
0.46
0.56
Vehicle-Class
LDV, LDT1
LDT2
LDT3
LDT4
Table 1-21. Approximate numbers of engine families certified, by model year and age group, for
model years 2001-2007.
Sales Area
California
Clean Fuel Vehicle
California + NLEV
(all states)
Federal All Altitude
Federal + CA Tier 2
Clean Fuel Veh +
NLEV(ASTR)2
+ CA
NLEV (All States)
TOTAL
Code
CA
CF
CL
FA
FC
NF
NL
Group1
CA
Fed
Both
Fed
Both
Both
Fed
Model Year
2001
114
38
149
79
57
31
468
2002
116
46
140
75
56
47
480
2003
118
81
129
86
16
45
74
549
2004
240
76
209
81
606
2005
251
69
219
41
580
2006
275
61
271
33
640
2007
255
55
274
16
600
Total
1,369
426
418
1,213
187
158
152
3,923
1 "Fed" denotes areas for which vehicles were certified to Federal Tier 1 , NLEV or Tier 2 standards, "CA" denotes vehicles
certified to California LEV-I or LEV-II standards, including the "section 177" states, "Both" denotes vehicles certified for
Federal or California Sales Areas.
2 "ASTR" = "All-state trading Region."
75
-------
Figure 1-28. Phase-in assumptions for Tier 1, NLEV, and Tier 2 standards, for LDV and LDT1.
"c
0)
0)
0_
c
c
3
3 C
3 F
•
C
3 C
3 F
•
s| C"
3 C
3 F
T ^
3 C
3 F
•
r ir
3 C
3 F
i a
3 C
3 F
D r-
3 C
3 F
a
3 C
3 F
D 0
3 C
3 F
I
-> c
3 t-
3 F
3
3 F
c
3 F
si
s|
• Tier-2(Bin2)
• Tier-2(Bin 3)
Tier-2(Bin 4)
• Tier-2(Bin5)
• Tier-2(Bin 6)
• Tier-2(Bin7)
LEV-II (ULEV)
• LEV-II (LEV)
NLEV(ULEV)
NLEV(LEV)
• NLEV(TLEV)
Tier 1
Model Year
Figure 1-29. Phase-in assumptions for Tier 1, NLEV and Tier 2 standards, for LDT2.
90%
80%
n%
• Tier-2(Bin3)
Tier-2(Bin 4)
• Tier-2(Bin 5)
• Tier-2(Bin7)
NLEV(ULEV)
NLEV(LEV)
• NLEV(TLEV)
TieM
Model Year
76
-------
Figure 1-30. Phase-in assumptions for Tier 1 and Tier 2 standards, for LDT3.
70% -
20% -
n% -
• Tier-2(Bin4)
• Tier-2(Bin5)
• Tier-2(Bin8)
• Tierl
Model Year
Figure 1-31. Phase-in assumptions for Tier 1 and Tier 2 standards, for LDT4.
-------
1.3.4.2.3 Merge FTP results and phase-in Assumptions
The goal of this step is to calculate weighted averages of the FTP cold-start and running results
for all standards in each model year, with the emissions results weighted by applicable phase-in
fractions. We do this step for each vehicle class separately, then we weight the four truck classes
together using a set of fractions also derived from the weighted sales estimates. Through MY
2007, where we had actual history, these fractions vary by model year, but are held stable after
2008. See Figure 1-32.
Figure 1-33 shows an example of the Phase-in calculation for NO* from cars between model
years 2000 and 2010. The figure shows cold start and running FTP values for Tier 1, NLEV and
Tier 2 standards, as well as the phase-in fractions for each standard in each model year. Start and
running emissions in each model year are simply calculated as weighted averages of the
emissions estimates and the phase-in fractions. The resulting weighted start estimates are used
directly to represent cold-start emissions for young vehicles in each model year (ages 0-3). For
running emissions, however, the averages are not used directly; rather, each is expressed as a
ratio to the corresponding Tier-1 value.
Table 1-22 shows weighted average values for model-years 2001-2010 for simulated FTP
composites, cold-start and hot-running emissions. The start values, expressed as the cold-start
mass increment (g), are used directly in the MOVES emission rate table to represent cold-start
emissions (for operating mode 108). The composites and running emissions, expressed as rates
(g/mi), are presented for comparison. For running emissions, however, the averages shown in
the table are not used directly; rather, each is expressed as a ratio to the corresponding Tier-1
value, as shown in Figure 1-34 to Figure 1-36 below.
78
-------
Figure 1-32. Relative fractions of truck classes, by model year.
a)
0)
0_
• LDT1
• LDT2
• LDT3
• LDT4
Model Year
Figure 1-33. Example of phase-in calculation, for NO* from cars (LDV), for MY 2000-2010.
Standard
Cold start Hot Running
(8) (g'mi)
Phase-in by Model Year
Tierl
NLEV
Tier 2
LEV-II
Tierl
TLEV
LEV
IJLEV
bin8
bin7
bins
bin4
bin3
bin2
LEV
ULEV
0.888 0.127
0.888 0.127
0.566 0.040
0.566 0.040
0.418 0.035
0.364 0.052
0.165 0.008
0.090 0.005
0.071 0.004
0.067 0.000
0.165 0.008
0.071 0.004
(Start (g)
Running (g/mile)
RATIO to Tierl
1
0
0
0
0
0
0
0
0
0
0
0
0.838
0.127
1.00
0.011
0.052
0.801
0.136
0
0
0
0
0
0
0
0
0.586
0.046
0.36
0.004
0.018
0.752
0.226
0
0
0
0
0
0
0
0
0.573
0.042
0.33
0.002
0.011
0.613
0.192
0.115
0.017
0.049
0
0
0
0
0
0.530
0.039
0.31
0
0
0.175
0.042
0.251
0.004
0.491
0.016
0.008
0
0.0052645
0.0074988
0.314
0.022
0.17
0
0
0.110
0
0.163
0.005
0.682
0.021
0.009
0.010
0.000
0.000
0.248
0.016
0.13
0
0
0.132
0
0.095
0.004
0.698
0.033
0.003
0.011
0.000
0.024
0.237
0.015
0.12
0
0
0.103
0
0.002
0
0.799
0.042
0.013
0.014
0.000
0.026
0.199
0.011
0.087
0
0
0.070
0
0
0
0.830
0.050
0.010
0.015
0.000
0.025
0.185
0.010
0.079
0
0
0.035
0
0
0
0.855
0.060
0.010
0.015
0.000
0.025
0.170
0.009
0.070
B
0
0
0
0
0
0.890
0.060
0.010
0.015
0.000
0.025
0.156|
0.008
0.061
79
-------
Table 1-22. Weighted average FTP values for trucks and cars for MY 2001-2010.
regClass
MY
Reference1
2000
Trucks
Cars
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
CO
Comp.
(g/rni)
Start
(R)
Running
(g/mi)
1.62
11.4
0.805
1.43
1.41
1.47
0.923
0.783
0.697
0.664
0.647
0.632
0.618
0.8561
0.8206
0.8076
0.7141
0.6716
0.6566
0.6210
0.6114
0.6011
0.5915
12.6
12.4
12.7
7.92
7.05
6.12
5.85
5.75
5.67
5.58
7.68
7.27
7.05
6.16
5.91
5.85
5.63
5.55
5.47
5.38
0.566
0.552
0.586
0.393
0.315
0.296
0.281
0.270
0.260
0.251
0.287
0.284
0.300
0.298
0.274
0.257
0.234
0.232
0.230
0.229
THC
Comp.
(g/mi)
Start
fe)
Running
(g/mi)
0.126
1.53
0.0571
0.0965
0.0942
0.1004
0.0535
0.0440
0.0378
0.0361
0.0356
0.0350
0.0345
0.0361
0.0333
0.0340
0.0360
0.0358
0.0350
0.0341
0.0341
0.0339
0.0339
1.23
1.21
1.25
0.786
0.703
0.612
0.587
0.580
0.571
0.564
0.954
0.893
0.839
0.664
0.634
0.633
0.608
0.592
0.574
0.557
0.0400
0.0376
0.0424
0.0123
0.00574
0.00511
0.00490
0.00479
0.00470
0.00461
0.00508
0.00451
0.00462
0.00488
0.00477
0.00462
0.00443
0.00443
0.00442
0.00442
NO.
Comp.
(g/mi)
Start
fe)
Running
(g/mi)
0.209
0.888
0.127
0.171
0.169
0.181
0.0849
0.0596
0.0381
0.0315
0.0285
0.0258
0.0233
0.0948
0.0898
0.0824
0.0461
0.0351
0.0335
0.0271
0.0248
0.0224
0.0201
0.843
0.836
0.863
0.473
0.367
0.264
0.226
0.208
0.192
0.177
0.586
0.573
0.530
0.315
0.248
0.239
0.201
0.187
0.172
0.158
0.0876
0.0865
0.0934
0.0434
0.0291
0.0183
0.0148
0.0130
0.0115
0.0101
0.0457
0.0421
0.0394
0.0220
0.0161
0.0150
0.0112
0.0101
0.00896
0.00784
'The reference level for calculating ratios is MY 2000, representing cars (LDV) for Tier 1 .
80
-------
Figure 1-34. Weighted ratios for composite, start and running CO Emissions, for (a) trucks and (b)
cars.
0.5 H
0.4
0.3 -
0.2 -
0.1
0.0
-Composite
-Cold Start
-Hot Running
4-
-r
—I- I I I
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
0.290 0.288 0.286 0.284
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
81
-------
Figure 1-35. Weighted ratios for FTP composite, start and running THC emissions, for (a) trucks
and (b) cars.
re
{£.
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
(a) Trucks
» Composite
—•—Cold Start
—A—Hot Running
0.215
-T-
I 0.101 i G.U3U I U.USb I U.UB4 I U.US2 l U.US
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
ro
{£.
0.079 i 0.081 i 0.035 i 0.084 i 0.081 ' 0.078 ' 0.078 ' 0.077 ' 0.077
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
82
-------
Figure 1-36. Weighted ratios for FTP composite, start and running NO* emissions, for (a) trucks
and (b) cars.
re
{£.
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
ro
{£.
0.689 0.680
•Composite
Cold Start
-Hot Running
0.341
0.229
•"- i u.uau i u.U/9
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
• u.uou i u.U/U ' 0.062
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
83
-------
1.3.4.2.4 Estimating Emissions by Operating Mode
With the introduction of the NLEV standards, new emissions requirements were imposed, in
addition to standards defined in terms of the Federal Test Procedure. The new requirements,
under the "Supplemental Federal Test Procedure" (SFTP), imposed more stringent emissions
control under conditions of high speed and power (through the US06 cycle), and with air-
conditioning running (through the SC03 cycle). To project emissions for NLEV and Tier-2
vehicles, we divided the operating modes for running exhaust into two groups. These groups
represent the ranges of speed and power covered by the FTP standards (< -18 kW/Mg), and the
ranges covered by the US06 cycle. For convenience, we refer to these two regions as "the hot-
running FTP region" and "US06 region," respectively (See Figure 1-37). Data measured on the
SC03 cycle did not play a role in emission rate development.
To estimate emissions by operating mode, the approach was to multiply the emission rates for
MY 2000, representing Tier 1, by a specific ratio for each model year from 2001 to 2010, to
represent emissions for that year. For the FTP operating modes, we applied the "hot-running"
ratios shown in Figure 1-34 to Figure 1-36 above.
1.3.4.2.4.1 Running Emissions
For the "US06" operating modes, we followed a different approach from that described above in
1.3.4.2.3. At the outset, we noted that the degree of control in the FTP standards increases
dramatically between MY 2000 through MY 2010, following phase-in of the Tier-2 standards,
giving pronounced declines in emissions on the FTP, especially for the hot-running phase (Bag
2). For our purposes, we are referring specifically to declines in running emissions, as shown by
changes in Bag-2 emissions. However, it was not obvious that the degree of control would
increase as dramatically for the SFTP standards, as shown by the US06. Thus, in preparation of
the draft rates, we adopted a conservative assumption that emissions in the "US06" region would
not drop as sharply as those in the "hot-running FTP" region.
It was therefore necessary to estimate different sets of ratios. Two alternative approaches were
developed.
The first option involved returning to the Phoenix I/M data. To create pre- and post-SFTP
estimates, we pooled tests for two model-year groups, 1998-2000, representing Tier 1 vehicles
not subject to SFTP requirements, and 2001-2003, representing NLEV vehicles subject to the
SFTP. For each group, we calculated means for each pollutant for the US06 operating modes (as
a group), and calculated ratios between the two groups.
„ -£poll.SFTP.01-03 .
^SFTP = •= Equation 1-39
-^poll,SFTP,98-00
The resulting ratios were used for CO and HC, as shown in Figure 1-38.
The second approach involved compilation of results on the US06 cycle and calculation of ratios
in a manner similar to that used for FTP data as described in 1.3.4.2.3 above. It was possible to
obtain data representing US06 tests representing vehicles in MY 1996-97 from the Mobile-
Source Observation Database (MSOD), developed and maintained by EPA/OTAQ.31 For NLEV
and Tier-2 vehicles manufactured after MY2000, US06 results were available from the IUVP
program. Emissions results on the US06 by standard were weighted by the phase-in assumptions
84
-------
for MY 2001-2007 as with the FTP results. Resulting ratios for cars and trucks are shown in
Figure 1-38.
Figure 1-39 and Figure 1-40 show application of the ratios to the hot-running FTP and US06
operating modes in model years 2000 (the reference year), 2005, and 2010, both calculated with
respect to 2000. The sets of ratios shown in Figure 1-38 for cars are used for both sets of modes.
Note that the values for the SFTP modes are equal in 2005 and 2010 for HC and CO, because the
SFTP ratios are constant by model year. In these figures, the results are presented on both linear
and logarithmic scales. The linear plots display the differences in the high-power modes, but
obscure those in the low-power modes. The logarithmic plots supplement the linear plots by
making visible the relatively small differences between MY 2005 and 2010 in the lower power
modes.
Figure 1-37. Operating modes for running exhaust emissions, divided broadly into "hot-running
FTP" and "US06" regions.
30 +
IF 27-30
^
.£ 24-27
M 27-24
| 78-27
£ 15-18
u
-------
Figure 1-38. Weighted ratios for hot-running emissions, representing the "hot-running FTP
Region" (FTP) and the "US06 Region" (US06), for (a) CO, (b) THC and (c) NO*.
US06 Region (CARS)
—A_. US06 Region (TRUCKS)
» FTP Hot-Running Region (CARS)
—»—FTP Hot-Running Region (TRUCKS)
0.734
v 0.699 0.698 0.696 0.696 0.696 0.696
555 0.552 0.551 O.fco. 0.546 0.546 0.545 0.544 0.544 0.544
*«—.
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
1.2
0.9
0.7
°-6
0.5
0.4
0.3
0.2
0.1
0.0
1'°°7 0.970
%
% \
\\ (b)Th
\ ^^^^^^^^
\
V
1.049
\
1C *
— A —
•
—A-
* -—
0\649
**s 0.538
%,-•
\
\
%
\
JS06 Region (CARS)
:TP Hot-Running Region (CARS)
JS06 Region (TRUCKS)
TP Hot-Running Region (TRUCKS)
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
1.2
0.9
g0.7
'•s °-6
K0.5
0.4
0.3
0.2
0.1
0.0
0.941
0.959
0.914
~~A- .,„—•*_
- US06 Region (CARS)
FTP Hot-Running Region (TRUCKS)
— US06 Region (TRUCKS)
— FTP Hot-Running Region (TRUCKS)
0.366
"A-.^308 °-289 0.281 0.274 0.267
3%< ~""~A'""
(c) NOx
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Model Year
86
-------
Figure 1-39. Projected emission rates for cars, vs. VSP, for three model years (LINEAR SCALE).
(NOTE: rates pictured represent operating modes 21-30 for ages 0-3 years).
meanBaseRatelM (g/hr)
l—i l—i l—i l—i l_
N)-P*CTlOOON)-P*CTlO
8 OOOOOOOC
OOOOOOOC
O2000
• 2005
D2010
,
ft ft
(a)
A
CO
t_ J
>
1
H
•
a
•
B
10 15 20 25
Vehicle Specific Power (kW/Mg)
30
35
meanBaseRatelM (g/hr)
h-* h-* h-* h-* h
oro-P*cnoooN>-P*cno
42000
• 2005
D 2010
-r«-^
* *
, a, a
(b) THr
\M7
^-*-,
<
• D
>
i
*
•
H
*
H
10 15 20 25
Vehicle Specific Power (kW/Mg)
30
35
meanBaseRatelM (g/hr)
N) 4^ cn 00 O r-
O O O O O O C
*2000
• 2005
D2010
.
ft :
(c)l
r-m-,
urj
<
• x
*
D
^
10 15 20 25
Vehicle Specific Power (kW/Mg)
87
-------
Figure 1-40. Projected emission rates for cars, vs. VSP, for three model years (LOGARITHMIC
SCALE). (NOTE: rates pictured represent operating modes 21-30 for ages 0-3 years).
.c
I"
/hr) „ meanBaseRatelM (g/hr) meanBaseRate
3 § p
° 2 ° ^ S 8 ^ S 8
a0 j-ww.w
i"
Ol
ro 10 0
cc
Ol
1/1
(G
00
= 10
(G 1'U
Ol
E
0.1
*2000
• 2005
D2010
fl
o
a
5 0
42000
• 2005
D2010
* A
H
5 0
•
•
•
D
(a) CO
a
4
a
^
!
•
B
5 10 15 20 25 30 35 40
Vehicle Specific Power (kW/Mg)
2000
2005
2010
(b)l
H
FHC
D
B
!
Q
1
D
B
5 10 15 20 25 30 35 4
Vehicle Specific Power (kW/Mg)
•
n
(Or
m
«>,
<
*
a
-5
10 15 20 25 30
Vehicle Specific Power (kW/Mg)
35
40
88
-------
1.3.4.2.5 Apply Deterioration
Based on review and analysis of the Phoenix I/M data, we assume that deterioration for different
technologies is best represented by a multiplicative model, in which different technologies,
represented by successive model-year groups, show similar deterioration in relative terms but
markedly different deterioration in absolute terms. We implemented this approach by translating
emissions for the 0-3 age Group, as calculated above, into natural logarithms and applying
uniform logarithmic age trends to all model-year groups. We derived logarithmic deterioration
slopes for Tier-1 vehicles (MY 1996-98) and applied them to NLEV and Tier-2 vehicles. In this
process we applied the same logarithmic slope to each operating mode, which is an extension of
the multiplicative deterioration assumption.
1.3.4.2.5.1 Recalculate the logarithmic mean
Starting with the values of the arithmetic mean (xa) calculated above, we calculate a logarithmic
mean (xi), as shown in Equation 1-40. Note that this equation is simply a rearrangement of
Equation 1-30 (page 38).
f = lnxa — Equation 1-40
The values of the logarithmic variance are intended to represent values for young vehicles, as the
estimates for xa represent the 0-3 year age Group. The values of o/2used for this step were 1.30,
0.95 and 1.60 for CO, THC and NO*, respectively.
1.3.4.2.5.2 Apply a logarithmic Age slope
After estimating logarithmic means for the 0-3 age class (x/,o-3), we estimate additional
logarithmic means for successive age classes (x/,age), by applying a linear slope in In-space (mi).
= Xlft_3 + ml (age -1.5) Equation 1-41
j
The values of the logarithmic slope are adapted from values developed for the 1996-98 model -
year group. The values applied are shown in Table 1-23. When calculating the age inputs for
this equation, we subtracted 1.5 years to shift the intercept to the midpoint of the 0-3 year age
Group, as shown in Equation 1-41.
89
-------
Figure 1-41 shows an example of the approach, as applied to THC from LDV in the 1996-98
model-year group. The upper plot (a) shows InTHC vs Age, by VSP, where the VSP acts as a
surrogate for operating mode. The defining characteristics of the plot are a series of parallel
lines, with the gaps between the lines reflecting the magnitude of the VSP differences between
them. Similarly, the lower plot shows InTHC vs. VSP, by Age, where age acts as a surrogate for
deterioration. In this view, deterioration appears as the magnitude of the gaps between a family
of similar trends against power.
Table 1-23. Values of the logarithmic deterioration slope applied to running-exhaust emission rates
for MY following 2000.
pollutant
CO
THC
NO,
opMode Group
"hot-running FTP"1
"US06"2
"hot-running FTP"
"US06"
"hot-running FTP"
"US06"
Logarithmic slope (mi)
0.13
0.06
0.09
0.09
0.15
0.15
1 Includes opModelD = 0,1, 11-16, 21-25, 27, 33,35,37.
2 Includes opModelD = 28,29,30, 38,39,40.
90
-------
Figure 1-41. Example of logarithmic deterioration model for THC (cars, MYG 96-98): (a) InTHC
vs age, by VSP level (kW/Mg), and (b) InTHC vs. VSP, by age (yr).
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
-10.0
-3
6
•9
•12
•21
•30
-10.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0
VSP (kW/tome)
91
-------
1.3.4.2.5.3 Apply the reverse transformation
After the previous step, the values of x/,age were reverse-transformed, as in Equation 1-30. The
values of the logarithmic variance used for this step were adapted from the Phoenix I/M results
and are intended to represent emissions distributions for "real-world" vehicle populations,
meaning that the values are higher than the value used in step 1.3.4.2.5.1 and may vary with age.
Values of logarithmic variances for all three pollutants are shown in Table 1-24.
Table 1-24. Values of logarithmic variance used to calculate emissions deterioration by reverse
transformation of logarithmic means.
Age Group
0-3 years
4-5
6-7
8-9
Pollutant
CO
1.30
2.05
2.00
1.80
THC
0.95
1.50
1.70
1.90
NO,
1.60
1.60
1.40
1.40
No values are presented in the table for the 10-14, 15-19 and 20+ year age Groups. This
omission is intentional, in that we did not want to extrapolate the deterioration trend beyond the
8-9 year age Group. Extrapolation beyond this point is incorrect, as we assume that emissions
tend to stabilize beyond this age, while the In-linear emissions model would project an
increasingly steep and unrealistic exponential emissions trend. For the 10-14, 15-19 and 20+ age
Groups, the "stabilization of emissions with age" was estimated as described in section 1.3.3.8.
Figure 1-42 shows the same results as Figure 1-41, following reverse transformation. The
families of parallel logarithmic trends are replaced by corresponding "fans" of diverging
exponential trends. An implication of this model is that as deterioration occurs, it is expressed
more strongly (in absolute terms) at high power. Similarly, the relationship between emissions
and VSP becomes more pronounced, in absolute terms, with increasing age.
92
-------
Figure 1-42. Example of reverse transformation for THC (LDV, MYG 96-98): (a) THC vs. age, by
VSP level (kW/Mg), (b) THC vs. VSP, by age (yr).
110.0
314/V/tDme
614/V/tDme
914/V/tDme
1214/V/tDme
2H4/V/tome
3014/V/tDme
2.0
4.0
60 8.0
Age (Years)
10.0
120
14.0
200.0
180.0
5.0
10.0
15.0 20.0 25.0
VSP(kW/tonne)
30.0
35.0 40.0
1.3.4.2.6 Estimate non-I/M References
Completion of steps 1.3.4.2.1 - 1.3A.2.6 provided a set of rates representing I/M reference rates
for MY 2001-2021. As a final step, we estimated non-I/M reference rates by applying the same
ratios applied to the I/M references for MY 2000 and previous, as described above.
93
-------
1.4 Exhaust Emissions for Start Operation
1.4.1 Subgroup 1: Vehicles manufactured in model year 1995 and earlier
In EPA's previous emissions model (MOBILE6), start emissions for passenger cars and light-
duty trucks, were dependent upon three factors:
1. the (base) emissions of that vehicle at 75 degrees Fahrenheit,32
2. an adjustment factor based on the length of soak time,33 and
3. an adjustment factor based on the ambient temperature.34
Within the MOVES modal structure, operating modes for start emissions are defined in terms of
soak time (preceding an engine start). The following sections will discuss the development of
base rates for "cold starts" (operating mode 108), as well as those for "warm" or "hot" starts
following seven soak periods of varying length (operating modes 101-107).
Note that the development and application of temperature adjustments is discussed in a separate
report.35
1.4.1.1 Methods
1.4.1.1.1 Data Sources
Data used in these analyses were acquired from the following four sources:
1. EPA's Mobile Source Observation Database (MSOD) as of April 27, 2005. Over
the past decades, EPA has performed emission tests (usually the Federal Test
Procedure) on large numbers of vehicles under various conditions.
We identified (in the MSOD) 549 gasoline-fueled vehicles (494 cars and 55
trucks) that had FTPs performed at temperatures both within the normal FTP range
(68° to 86° Fahrenheit) as well as outside that range (i.e., either below 68° or above
86°). Aside from the differences in ambient temperature, the test parameters for
the paired FTPs on each vehicle were identical. The FTPs were performed at
temperatures from 16 through 111° F.
2. EPA's Office of Research and Development (ORD) contracted (through the Clean
Air Vehicle Technology Center, Inc.) the testing of five cars (model years 1987
through 2001). Those vehicles were tested using both the UDDS and the EVI240
cycle at temperatures of: 75, 40, 20, 0 and -20 °F.36
3. Southwest Research Institute (SwRI) tested four Tier-2 vehicles (2005 model year
car and light-duty trucks) over the UDDS at temperatures of: 75, 20, and 0 °F.37
4. During 2004-05, USEPA OTAQ and ORD, in conjunction with the Departments of
Energy and Transportation, conducted a program in the Kansas-City Metropolitan
Area. During this study, designed to measure paniculate emissions, gaseous
emissions were also measured on the LA92 cycle.38
94
-------
1.4.1.1.2 Defining Start Emissions
Using the FTP data described above, we estimated cold-start emissions as the difference in mass
between Bag 1 and Bag 3 (g). However, because Bag 1 follows a 12-hour (720 minute) soak and
Bag 3 follows a 10-minute soak, it is possible to use soak/time relationships to modify the Bagl-
Bag3 difference so as to account for the respective soak periods. The start/soak relationships we
applied were adapted from a study performed by the California Air Resources Board.39 Based on
these data, we derived a correction factor "A" as shown in Equation 1-42 and Table 1-25.
Cold - start Emissions =
(Bag 1-Bag 3)
l-A
Equation 1-42
Table 1-25. Correction factor A for application in Equation 39.
Vehicle Type
No Catalyst
Catalyst Equipped
Heated Catalyst
HC
0.37101
0.12090
0.05559
CO
0.34524
0.11474
0.06937
NO.
1.57562
0.39366
1.05017
1.4.1.1.3 Relationship between Soak Time and Start Emissions
In the MOVES input database, "operating modes" for start emissions are defined in terms of
soak time preceding an engine start. The "cold-start," as defined and calculated above, is
represented as opModeID=108. An additional seven modes are defined in terms of soak times
ranging from 3 min up to 540 min (opModelD = 101-107). To estimate start rates for the
additional seven modes, we applied the soak-time/start relationships mentioned above. The
specific values used are adapted from the MOBILE6 soak-effect curves for catalyst- and non-
catalyst-equipped vehicles.15 To adapt these relationships to the MOVES operating modes, the
soak time was divided into eight intervals, each of which was assigned a "nominal" soak time, as
shown in Table 1-26.
95
-------
Table 1-26. Operating-mode definitions for start emissions, defined in terms of soak time.
Nominal Soak Period
(min)
3
18
45
75
105
240
540
720
OpModelD
101
102
103
104
105
106
107
108
OpModeName
Soak Time < 6 minutes
6 minutes < Soak Time < 30 minutes
30 minutes < Soak Time < 60 minutes
60 minutes < Soak Time < 90 minutes
90 minutes < Soak Time < 120 minutes
120 minutes < Soak Time < 360 minutes
360 minutes < Soak Time < 720 minutes
720 minutes < Soak Time
We have adapted and applied soak time adjustments used in MOBILE6.2 for gasoline-fueled
vehicles, as shown in Table 1-27. Additionally, all pre-1981 model year passenger cars and
trucks use the catalyst equipped soak curve adjustments, although some of these vehicles are not
catalyst equipped.
Table 1-27. Calculated soak-time adjustments, derived from MOBILE6 soak-time coefficients for
catalyst-equipped vehicles.
opModelD
101
102
103
104
105
106
107
108
Soak
period
(min)
3
18
45
75
105
240
540
720
i
HC
0.051
0.269
0.525
0.634
0.645
0.734
0.909
1.000
\djustmen
CO
0.034
0.194
0.433
0.622
0.728
0.791
0.914
1.000
t
NO,
0.093
0.347
0.872
1.130
1.129
1.118
1.053
1.000
Model-year groups used to calculate start rates for vehicles in model year 1995 and earlier are
shown in Table 1-28. In some cases, model-year groups were adjusted to compensate for sparsity
of data in narrower groups. For example, the average NO* emissions for MY 1983-1985 trucks
are slightly negative. This result is possible, but is likely due to erratically behaving means from
96
-------
small samples. Thus, these model years were grouped with the 1981-1982 model years, which
for trucks had similar emission standards. In addition, the MY1994-1999 gasoline truck sample
includes a very high-emitting vehicle, which strongly influences the results for CO. To
compensate, these vehicles were grouped with the 1990-1993 model years. The values in the
table represent the difference of Bag-1 minus Bag-3, adjusted, as described above, to estimate
cold-start emissions.
Table 1-28. Cold-start emissions (Bag 1 - Bag 3,) for gasoline-powered cars and trucks.
Model-year
Group «
Years
Mean (g)
THC CO NOX
Standard deviation (g)
THC CO NOX
CV-of-the-Mean (RSE)
THC CO NOX
Cars
1960-1980 1,488
1981-1982 2,735
1983-1985 2,958
1986-1989 6,837
1990-1993 3,778
1994-1995 333
5.172 75.832 0.608
3.584 52.217 1.118
2.912 34.286 0.922
2.306 21.451 1.082
1.910 17.550 1.149
1.788 16.233 1.027
6.948 83.812 2.088
7.830 60.707 1.682
5.216 44.785 1.321
2.740 32.382 1.034
1.728 13.953 1.034
1.203 31.648 0.742
0.035 0.029 0.089
0.042 0.022 0.029
0.033 0.024 0.026
0.014 0.018 0.012
0.015 0.013 0.015
0.037 0.107 0.040
Trucks
1960-1980 111
1981-1985 910
1986-1989 1,192
1990-1995 1,755
9.008 115.849 0.155
4.864 94.608 0.0412
3.804 45.918 2.107
3.288 40.927 2.192
9.179 113.269 2.682
4.992 67.871 1.797
2.298 36.356 2.152
4.211 42.478 2.158
0.097 0.093 1.641
0.034 0.024 1.445
0.017 0.023 0.030
0.031 0.025 0.024
1.4.2 Subgroup 2: Vehicles manufactured in MY1996 and later
Start rates for vehicles manufactured in model year 1996 and later were estimated using data
from the EPA In-use Verification Program (IUVP), as with running rates for MY 2001 and later
(see 1.3.4 above).
For model years 1996-2000, rates for vehicles at 0-3 years of age (ageGroup=0003), are shown
above in Table 1-22, in the row for MY2000.
97
-------
For MY 2001 and later, cold-start rates (opModeID=108) were estimated as described in section
1.3.4 above, using the data and approaches described in steps 1-4 and step 6 (as described on
page 68). We applied the FTP averages as shown in Figure 1-26 and Figure 1-27, and the phase-
in assumptions shown in Figure 1-28, Figure 1-29, Figure 1-30, Figure 1-31 and Figure 1-32. As
with running emissions, Figure 1-33 illustrates the calculation of weighted average FTP results
by model year.
To estimate start emissions for the remaining seven operating modes, we applied "soak
fractions" to the "cold-start" emissions, as described above. The soak fractions were adapted
from the approach applied in the MOBILE model.20 Specifically, the piece-wise regression
equations used in MOBILE6 for "conventional catalyst" engines were evaluated at the midpoint
of the soak period for each operating mode. For each mode, the start rate is the product of the
cold-start rate and the corresponding soak fraction. Figure 1-43 shows the soak fractions for HC,
CO and NO*, with each value plotted at the midpoint of the respective soak period.
Figure 1-43. Soak fractions applied to cold-start emissions (opModelD = 108) to estimate emissions
for shorter soak periods (operating modes 101-107).
1.40
re
QL
0.00
120 240 360 480
Soak Time (minutes)
600
720
1.4.3 Applying Deterioration to Starts
1.4.3.1 Assessing Start Deterioration in relation to Running Deterioration
The large datasets used to develop rates for running emissions provided much information about
deterioration for hot-running emissions, but no information on deterioration for start emissions.
After some consideration, it occurred to us that the data from the IUVP program, used to develop
running rates for NLEV and Tier-2 vehicles, could also be useful to evaluate the relationship
between deterioration trends for start and running emissions. A valuable aspect of these data is
that they provide FTP results with the measurement phases separated. As before, we focused on
98
-------
cold-start emissions, calculated as Bagl - Bag3 (g), and hot-running emissions, represented by
Bag2 (g/mi). For this purpose, these data are also valuable because they provide emissions
measured over a wide range of mileage, up to 100,000 mi, although the corresponding range of
vehicle age is relatively narrow (0-5 years). Thus, we elected to evaluate trends in emissions vs.
mileage.
At the outset, we plotted the data for NMOG and NO* vs. odometer reading, on linear and
logarithmic scales. Scatterplots of start and running NMOG emissions are shown in Figure 1-44
and Figure 1-45; corresponding plots for InNMOG are shown in Figure 1-46 and Figure 1-47.
Similarly, scatterplots of start and running NO* emissions are shown in Figure 1-48 and Figure
1-49; corresponding plots for InNOx are shown in Figure 1-49 and Figure 1-50.
In viewing the data, some observations are apparent. The data are grouped, with one group
representing vehicles measured at less than 50,000 miles, centered around 10,000-20,000 miles,
and a second group representing vehicles measured at 50,000 to 100,000 miles. Given the
purpose of the IUVP program, the two groups are designed to assess compliance with
certification (< 50,000 mi) and useful-life (>50,000 mi) standards, respectively. As expected,
distributions of emissions are skewed, but with running emissions more skewed than start
emissions. On a logarithmic scale, the degree of skew is shown by the variability of the
transformed data, with the In(start) spanning 3-3.5 factors of e, and the In(running) spanning 6-7
factors of e. Finally, and of most relevance to this analysis, deterioration trends are visible in the
In plots, with the masses of points at >50,000 miles centered higher than those for < 50,000
miles.
To assess the presence of trends in emissions and mileage more rigorously, we ran linear
statistical models on the In-transformed data. To illustrate, we will focus on models run on
vehicles certified to LEV standards, as shown in Table 1-31 and Table 1-32. The model structure
includes a grand intercept for all vehicle classes (LDV, LDT1-4), and with separate intercepts for
each vehicle class. All parameters are highly significant, both for InNMOG and InNO*. A more
complex model structure was attempted, which included individual mileage slopes for different
vehicle classes. However, this model was not retained, as it did not improve the fit, nor were the
interaction terms themselves significant. The covariance structure applied was simple, in that a
single residual error variance was fit for all vehicle classes.
Models were fit to vehicles certified to other standards, such as ULEV and Tier-2/Bin-5, the
results for which are not shown here. The models for ULEV show very similar patterns to those
for LEV, whereas the models fit to Bin5 data were not considered useful as the range of mileage
covered for these more recent vehicles was not wide enough to demonstrate deterioration trends
(i.e., < 25,000 mi).
The models confirm the visual impression given by the plots of InNMOG and InNO*. Positive
trends in emissions do appear evident in these data, but the increase in emissions with mileage is
very gradual. The trends in InNOx are steeper than those for InNMOG, and the trends for
running emissions are steeper than those for start emissions. However, the differences between
the slopes for start and running are less pronounced for InNOx than for InNMOG. For InNOx, the
running slope is 1.25 times that for starts, and for InNMOG, the running slope is 1.65 times that
for starts.
99
-------
Figure 1-44. Cold-start FTP emissions for NMOG (g) vs. odometer (mi), for LEV vehicles, from the
IUVP program.
140000 150OB
Figure 1-45. Hot-running (Bag 2) FTP emissions for NMOG (g/mi) vs. odometer (mi), for LEV
vehicles, from the IUVP program.
i running
~ 0.16-
10000 20000 3DOGO 4COJO 50000 60000 70000 3CWQ
100000 110000 120000 130000 140000 150000
100
-------
Figure 1-46. Cold-start FTP emissions for In(NMOG) vs. odometer (mi), for LEV vehicles, from
the IUVP program (LOGARITHMIC SCALE).
10JSG 20000 30000 40000 50000 60XX) 7'KO] 90000 90000 100000 110000 120000 130000 140000 130000
D LUT2 00° LEVT1
MDV3 + •+ +• WDV4
Figure 1-47. Hot-running (Bag 2) FTP emissions for In(NMOG) vs. odometer (mi), for LEV
vehicles, from the IUVP program (LOGARITHMIC SCALE).
3000G &XKX) 70C03 9GOOO 50300 1QHOD 110000 120DOO 130000 140000 13DOOO
101
-------
Figure 1-48. Cold-start FTP emissions for NO* (g) vs. odometer (mi), for LEV and ULEV vehicles,
from the IUVP program.
1KBC 20000 30KB 40000 50X0 600X 700DC
90CKB 100000 11MCO 12OHM 130X0 14X00 15GCO3
D LDVT1 a i a MDV2 * * * MEfV3 + + -> MDV4
Figure 1-49. Hot-running (Bag 2) FTP emissions for NO* (g/mi) vs. odometer (mi), for LEV and
ULEV vehicles, from the IUVP program.
1C030 20DCC 30000 40030 50000 60000 70000 SGODQ QOOOO 100000 110000 1200GO 130000 140000 150000
vehdra aa a U3T2 aaa LDVT1 & & & MDV2 * * * MDV3 + + •+ MDV4 +-<• + MDV5
102
-------
Figure 1-50. Cold-start FTP emissions for In(NO^) vs. odometer (mi), for LEV vehicles (Source:
IUVP program).
10000 33000
40000 HOOD
TOCOJ 30000 90000 IQOKB 110000 120000 130000 140000 1KOOO
* MDVJ + + + MDV4 + t- t- MDV5
Figure 1-51. Hot-running (Bag 2) FTP emissions for In(NO^) vs. odometer (mi), for LEV vehicles
from the IUVP program.
20000 30000 -40000 3000D 60000 70000 30000 90X0 100000 110000 120000 130000 140DDD 150000
o L0VT1 " i ^ MDV2 * *= * MJV3 + + + MUVt + + +• MDV5
103
-------
Table 1-29. Model fit parameters for InNMOG, for LEV vehicles.
Parameter
Predictor
Estimate
Standard error
Denom. D.F.
lvalue
Pr>*
Cold-Start (Bag 1 - Bag 3) (residual error = 0. 1942)
Slope
intercept
intercept
intercept
intercept
Odometer (mi)
LDV-T1
LDT2
LDT3 (MDV2)
LDT4 (MDV3)
0.000004982
-1.9603
-1.7353
-1.5735
-1.2937
0.0
0.02224
0.02429
0.03520
0.03233
2,404
2,404
2,404
2,404
2,404
co
-88.14
-71.43
-44.70
-40.01
0.0001
0.0001
0.0001
O.0001
O.0001
Hot-Running (Bag 2) (residual error = 1.3018)
Slope
intercept
intercept
intercept
intercept
Odometer (mi)
LDV-T1
LDT2
LDT3 (MDV2)
LDT4 (MDV3)
0.000008237
-6.1604
-6.2554
-5.9018
-5.5949
0.0
0.05961
0.06577
0.09239
0.08766
2,225
2,225
2,225
2,225
2,225
co
-103.34
-95.11
-63.88
-63.83
O.0001
O.0001
O.0001
O.0001
O.0001
Table 1-30. Model fit parameters for InNO*, LEV+ULEV vehicles.
Parameter
Predictor
Estimate
Standard error
Denom. D.F.
f-value
Pr>*
Cold-Start (Bag 1 - Bag 3) (residual error = 0.68)
Slope
intercept
intercept
intercept
intercept
Odometer (mi)
LDV-T1
LDT2
LDT3 (MDV2)
LDT4 (MDV3)
0.000009541
-2.6039
-2.4538
-2.0769
-1.645
0.0
0.05231
0.06056
0.08173
0.08882
1,657
1,657
1,657
1,657
1,657
co
-50.74
-40.52
-25.41
-18.52
O.0001
O.0001
O.0001
O.0001
O.0001
Hot-Running (Bag 2) (residual error = 2.9643)
Slope
intercept
intercept
intercept
intercept
Odometer (mi)
LDV-T1
LDT2
LDT3 (MDV2)
LDT4 (MDV3)
0.000012
-4.7396
-4.9527
-4.3144
-4.1214
0.00000165
0.1092
0.1304
0.1740
0.1835
1,622
1,622
1,622
1,622
1,622
7.13
-43.40
-37.98
-24.80
-22.47
O.0001
O.0001
O.0001
O.0001
O.0001
Having drawn these conclusions, we developed an approach to apply them to emission rate
development. To begin, we applied the statistical models by calculating predicted values of
InNMOG and InNO* at mileages from 0 (the intercept) to 155,000 miles. We reverse-
transformed the models using Equation 1-30 (page 38) to obtain predicted geometric and
arithmetic means with increasing mileage, as shown in Table l-31for NMOG and Table 1-32 for
104
-------
We normalized the predicted means at each mileage to the value at 0 miles to obtain a
"deterioration ratio" Rdet, by dividing each predicted value at a given mileage by the predicted
value at 0 miles (i.e., the intercept); Rdet for the intercept =1.0 (Equation 1-43).
x
a,miles
Equation 1-43
We took this step to express start and running trends on a comparable relative multiplicative
basis, as trends in absolute running and start emissions are clearly not comparable.
Finally, to relate start and running trends, we calculated the ratio in Rdet for start to that for
running, designated as Rrs\
R,
-del, start
D
det, running
Equation 1-44
Values or Rdet and Rrs\ for NMOG and NO* are shown in Table 1-31 and Table 1-32,
respectively, with corresponding results shown graphically in Figure 1-52 and Figure 1-53,
respectively.
Table 1-31. Application of models for NMOG, representing emissions trends for LDV-T1 vehicles
certified to LEV standards.
Parameter
Odometer (mi, xlo,000)
0
7.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
Cold Start
InNMOG
Geometric mean
Arithmetic
mean
Deterioration
ratio (Rdet)
-1.960
0.141
0.156
1.000
-1.886
0.152
0.168
1.078
-1.836
0.159
0.176
1.133
-1.786
0.168
0.185
1.190
-1.736
0.176
0.195
1.251
-1.686
0.185
0.205
1.315
-1.636
0.195
0.215
1.382
-1.587
0.205
0.226
1.453
-1.537
0.215
0.238
1.527
Hot Running
InNMOG
Geometric mean
Arithmetic
mean
Deterioration
ratio (Rdet)
Relative Ratio
(Kiel)
-6.160
0.00211
0.00404
1.000
1.000
-6.037
0.00239
0.00458
1.132
0.9952
-5.954
0.00259
0.00497
1.229
0.922
-5.872
0.00282
0.00540
1.334
0.892
-5.790
0.00306
0.00586
1.449
0.864
-5.707
0.00332
0.00636
1.573
0.836
-5.625
0.00361
0.00691
1.708
0.809
-5.543
0.00392
0.00750
1.855
0.783
-5.460
0.00425
0.00815
2.014
0.758
105
-------
Table 1-32. Application of models for NO*, representing emissions trends for LDV-T1 vehicles
certified to LEV standards.
Parameter
Odometer (mi, xlQ,000)
0
7.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
Cold Start
InNO*
Geometric mean
Arithmetic mean
Deterioration
ratio (Rdet)
-2.604
0.0740
0.1039
1.000
-2.461
0.0854
0.1199
1.154
-2.365
0.0939
0.1319
1.269
-2.270
0.1033
0.1452
1.396
-2.175
0.1137
0.1597
1.536
-2.079
0.1250
0.1757
1.690
-1.984
0.1376
0.1933
1.859
-1.888
0.1513
0.2126
2.045
-1.793
0.1665
0.2339
2.250
Hot Running
InNO*
Geometric mean
Arithmetic mean
Deterioration
ratio (Rdet)
Relative Ratio (RKi)
-4.740
0.0087
0.0385
1.000
1.000
-4.560
0.0105
0.0461
1.097
0.964
-4.440
0.0118
0.0520
1.350
0.940
-4.320
0.0133
0.0586
1.522
0.918
-4.200
0.0150
0.0660
1.716
0.895
-4.080
0.0169
0.0745
1.935
0.874
-3.960
0.0191
0.0840
2.181
0.852
-3.840
0.0215
0.0947
2.460
0.832
-3.720
0.0242
0.1067
2.773
0.811
Figure 1-52. Deterioration ratios for cold-start and hot-running NMOG emissions.
2.5
2.0
1.5
8. 1.0
_
•Cold-start
Hot-running
•Start: Running
0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Mileage (mi, x 10,000)
7.0
8.0
9.0
106
-------
Figure 1-53. Deterioration ratios for cold-start and hot-running NO* emissions.
Cold-start
Hot-running
StartRunning
0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Mileage (miles, x 10,000)
7.0
8.0
9.0
For NOx (Figure 1-53) we decided to assign start NO* the same multiplicative relative
deterioration as running NO*. However, for HC, the difference between running and start
deterioration was greater, large enough that we reduce to reduce start deterioration relative to
running deterioration.
1.4.3.2 Translation from Mileage to Age Basis
The question remained, as to how the results derived from the IUVP data and presented above
could be applied during the generation of emission rates. At the outset, a question arises from the
fact that the results shown above were generated on the basis of mileage, whereas MOVES
assigns deterioration on the basis of age. It was therefore necessary to translate the RK\ from a
mileage basis to an age basis. We achieved the translation through a series of steps.
First, we assumed a rate of mileage accumulation of about 10,000 miles per year, from which it
follows that the Rrs\ at 125,000 miles would occur at about 12.5 years of age, or would be
represented by the 10-14 year ageGroup. Accordingly, we assigned midpoints to the 0-3 and 10-
14 year ageGroups of 2 and 12.5 years, respectively, and assume that Rrs\ declines linearly with
age. These assumptions allow calculation of a declining trend in the ratio with respect to age.
The slope of the trend is the change in ratio (ARrei) over the corresponding change in time
(Atime).
Afl, 0.675-1.0 -0.325
10.5
vrel
Atime 12.5-2
= -0.30952
Equation 1-45
The calculation of the slope lets us estimate a value of Rrs\ for each ageGroup.
107
-------
-mRldage
Equation 1-46
The results, as applied for hydrocarbons and CO, are shown in Table 1-33 and Figure 1-54. The
net result is a 15-40% reduction in multiplicative start deterioration, relative to running
deterioration.
Table 1-33. Relative deterioration ratios (Rre\), for NMOG (and CO), assigned to each ageGroup.
AgeGroup
0-3
4-5
6-7
8-9
10-14
15-19
20 +
Age (years)
2
5
7
9
12.5
17.5
23
Relative Ratio (Rrei)
1.000
0.845
0.783
0.721
0.613
0.613
0.613
108
-------
Figure 1-54. Relative deterioration ratios (Rrei), for NMOG (and CO), assigned to each ageGroup.
OJ
i
0
£
V^
_ns
£
I.UUU
Oonn
n Rnn -
Oynn
Ocnn
. jUU
01 nn
.4UU
O-inn
. 1 UU
Onnn
\
>^^
^•^-^^^
p^^^^
.UUU i i i i i i i i i i i i i i i i i i i i
0 5 10 15 20 2
Age (years)
1.4.3.3
Application of Relative Multiplicative Deterioration
An advantage of the modal approach is that any driving cycle can be represented as a weighted
average of the MOVES emission rates and the "operating-mode distribution" for the cycle. This
allows emissions from a vehicle driven on one drive cycle to be converted to another drive cycle
within the operating mode envelope. In this case, we applied an operating-mode distribution for
the "hot-running" phase of the FTP. We apply the FTP in this context because the start rates are
expressed in terms of the FTP cold-start and hot-start phases. This phase is 860 seconds long and
represents urban driving over a 3.86 mile route after the engine has stabilized at its normal
operating temperature. We estimated an operating-mode distribution using the "Physical
Emission Rate Simulator" (PERE).17 This distribution, shown in Table 1-34, represents a
"typical" car, with engine displacement and test weight of 2.73 L and 3,350 Ib. A corresponding
"typical" truck was represented with displacement and test weight of 4.14 L and 4,364 Ib,
respectively.
Combining emission rates for hot-running emissions with the operating-mode distributions, we
calculated aggregate cycle emissions for the hot-running phase of the FTP (g/mi), for all model-
year and age groups. Figure 1-55 and Figure 1-56 show resulting cycle aggregates for THC and
NOx. Note that the underlying rates for model years 1995 (representing Tier 0) and 2000
(representing Tier 1) were derived using data and methods described above in Section 1.3.3
(starting on page 20), and those for model years 2005 and 2010 were derived using data and
methods described above in Section 1.3.4 (starting on page 68).
It is important to note that this step is performed both for vehicles in inspection-and-maintenance
areas (I/M, using the meanBaseRateEVI) and for vehicles outside I/M areas (using
meanBaseRate). Because deterioration is represented differently for the non-I/M and I/M
reference rates (see 1.3.3.7.2, page 66), and this difference is carried into deterioration for the
start rates, the result is that the MOVES rates represent that I/M programs have effects on start as
well as running emissions.
109
-------
R,
-det,MYG,Age
_ -^FTP2,MYG,Age
Equation 1-47
^FTP2,MYG,0-3
Table 1-34. Operating-mode distributions for running emissions, representing a "typical" car and
truck on the hot-stabilized phase of the FTP (Bag 2).
opModelD
0
1
11
12
13
14
15
16
21
22
23
24
25
27
28
29
30
33
35
37
38
39
40
Cars (LDV)
Time in mode
(sec)
97
155
77
121
83
59
22
4
42
111
62
18
7
2
0
0
1
0
0
0
0
0
0
Time in mode
(%)
11.27
18.00
8.94
14.05
9.64
6.85
2.56
0.46
4.88
12.89
7.20
2.09
0.81
0.23
0.00
0.00
0.12
0.00
0.00
0.00
0.00
0.00
0.00
Trucks (LDT)
Time in mode
(sec)
97
155
74
112
88
66
19
7
41
102
69
21
7
2
0
0
1
0
0
0
0
0
0
Time in mode
(%)
11.27
18.00
8.59
13.01
10.22
7.67
2.21
0.81
4.76
11.85
8.01
2.44
0.81
0.23
0.00
0.00
0.12
0.00
0.00
0.00
0.00
0.00
0.00
110
-------
Figure 1-55. Cycle-aggregate THC emission rates by age, projected from MOVES running-exhaust
emission rates, for the hot-stabilized phase of the FTP, representing vehicles in I/M areas.
n
OL
a.
t
c
'E
c
OL
1.4
1.2
1 :
0.8
0.6
0.4
0.2
0
-1995
•2000
-2005
-2010
,•1 t I . ,», . I . ft , I ,^
10 15
Age (years)
20
25
Figure 1-56. Cycle-aggregate NO* emission rates by age, projected from MOVES running-exhaust
emission rates, for the hot-stabilized phase of the FTP, representing vehicles in I/M areas.
-1995
2000
-2005
•2010
10 15
Age (years)
20
25
For each model-year group, we divided cycle aggregate for each ageGroup (£FTP2,MYG,Age) by the
estimate for the 0-3 year ageGroup (EFTP2,MYG,o-3), to obtain a deterioration ratio (7?det,MYG,Age) as
shown in Equation 1-47. As examples, ratios for cars are shown for THC in Figure 1-57 (I/M)
and Figure 1-58 (non-I/M). Corresponding ratios for NO* are shown in Figure 1-59 (I/M) and
Figure 1-60 (non-I/M). The ratios show that, in relative multiplicative terms, the MOVES rates
represent greater deterioration for running exhaust THC than for NO*.
Ill
-------
Figure 1-57. Deterioration Ratios for THC, representing the hot-stabilized phase of the FTP (Bag
2), representing vehicles in I/M areas.
-1995
•2000
-2005
-2010
10 15
Age (years)
20
25
Figure 1-58. Deterioration Ratios for THC, representing the hot-stabilized phase of the FTP (Bag
2), representing vehicles in non-I/M areas.
TI
a;
n
GC
o
%
u>
_0
U
—4—1995
•^K-2000
-»—2005
-e—2010
10 15
Age (years)
20
25
112
-------
Figure 1-59. Deterioration Ratios for NO*, representing the hot-stabilized phase of the FTP (Bag
2), representing vehicles in I/M areas.
n
L_
0
•1995
2000
•2005
•2010
10 15
Age (years)
20
Figure 1-60. Deterioration Ratios for NO*, representing the hot-stabilized phase of the FTP (Bag
2), representing vehicles in non-I/M areas.
„ 4-5 :
-1995
-2000
•2005
•2010
10 15
Age (years)
20
25
At this point, projecting deterioration for start emissions is a simple matter of multiplying the
start rate for the 0-3 year ageGroup in each relevant operating mode (opModelD =101-108) by
the deterioration ratio (7?det) and the relative deterioration ratio (RK\) for each ageGroup. The
projected start rate in each agegroup (EWtage) is
113
-------
r
-'-'
'start, age
-'start, 0-3
p p
-3-1 Met, age-1 Vel, age
Equation 1-48
Note that for NO* the values of Rrs\ are 1.0 for all agegroups, i.e., relative multiplicative
deterioration for start emissions is the same as for running emissions. For THC and CO,
however, Rrs\ takes the values shown in Table 1-33, which reduces reduced relative start
emissions in comparison to relative running emissions. To illustrate the results, Figure 1-61 and
Figure 1-62 show deterioration for cold-start emissions (opModeID=108) for THC and NO*,
respectively.
Figure 1-61. Projected deterioration for cold-start THC emissions (opModeID=108), in four model
years, representing vehicles in I/M areas.
0
0
10 15
Age (years)
20
25
114
-------
Figure 1-62. Projected deterioration for cold-start NO* emissions (opModeID=108), in four model
years, representing vehicles in I/M areas.
-1995
2000
•2005
•2010
10 15
Age (years)
20
1.5 Incorporating Tier 3 Emissions Standards
Methods used to develop rates to represent emissions for vehicles certified to Tier-3 standards
were identical to those used to develop rates for vehicles certified to NLEV and Tier 2 standards,
with several specific modifications. Where no modifications to methods were made, we will
refer the reader to the appropriate section of this report. In particular, see Section 1.3.4.
As previously described, the goal of this process is to represent I/M reference rates for the 0-3
year ageGroup. The rates are estimated by Tier, model year and regulatory class. The process
involves six steps previously described, repeated below for convenience.
1. Average FTP results by standard level and vehicle class. As before, we made use of data
measured on the FTP cycle in the course of the In-use Verification Program (IUVP).
2. Develop phase-in assumptions for MY 2017 - 2031, by standard level, vehicle class and
model year.
3. Merge FTP results and Phase-in assumptions. For running emissions, calculate weighted
ratios of emissions in each model year relative to those for cars (LDV) in MY2000, which
represent Tier-1 LDV.
4. Estimate Emissions by Operating Mode. We calculated emissions by operating mode in each
model year by multiplying the MY2000 emission rates by the weighted ratio for each model
year. We assume that the emissions control at high power (outside ranges of speed and
115
-------
acceleration covered by Bag 2 of the FTP) would not be as effective as at lower power (within
the range of speed and acceleration covered by Bag 2).
5. Apply Deterioration to estimate emissions for three additional age Groups (4-5, 6-7 and 8-9).
We assume thatNLEV and Tier-2 vehicles will deteriorate similarly to Tier-1 vehicles, when
viewed in logarithmic terms. However, for Tier 3 vehicles, we modified deterioration to
represent an extended useful life of 150,000 miles, as opposed to the 120,000 mile duration
assumed for NLEV and Tier 2 vehicles. We therefore apply In-linear deterioration to the rates
developed in steps 1-4. For the remaining three age groups, emissions are assumed to stabilize
as previously described in 1.3.3.8 (page 61).
6. Estimate non-I/Mreference rates. The rates in steps 1-5 represent I/M references.
Corresponding non-I/M references are calculated by applying the ratios applied to the Tier-1 and
pre-Tier-1 rates.
We followed steps 1-6, with specific modifications to represent Tier 3 rates. In step 1, we
developed estimates of FTP results under Tier 3, including composite results, "cold-start"
emissions" (Bagl-Bag3) and "hot-running" emissions (Bag 2 FTP and US06). For step 2, we
developed phase-in assumptions representing the introduction of Tier-3 standards. Each of these
steps and modifications is described in greater detail in the sub-sections below.
1.5.1 Averaging FTP Results (Step 1)
Projecting emissions for Tier-3 vehicles is driven by the NMOG+NO* standard, set at 30 mg/mi.
However, because MOVES projects NO* and THC emissions separately, we apportioned the
aggregate standard into NMOG and NO* components, which we will refer to as the "effective
standards" for each pollutant. For purposes of apportionment, we assumed that NMOG control
would pose a greater technical challenge than NO* control. Accordingly, we assumed "effective
standards" for NMOG and NO* of 20 mg/mi and 10 mg/mi, respectively. To implement this
assumption, we further assumed that for NO*, vehicles would be effectively brought into Tier 2
Bin 2, and that for NMOG, vehicles would be brought to a level between Bin 2 and Bin-3, but
closer to Bin 2.
In addition, MOVES models start and running processes separately. It is therefore necessary to
translate the composite standard into start and running components. One component represents a
"cold start" on the FTP cycle, represented as "Bagl - Bag3" emissions. A second component
represents "hot-running" emissions, represented by the hot-running phase of the FTP (Bag 2). A
third component represents emissions on the US06 cycle, representing emissions at high speed
and power.
Estimated FTP and US06 emissions levels for hydrocarbons (NMOG and NMHC) are shown in
Table 1-35, for several Tier-2 Bins and for Tier 3. Values for all standards except Tier 3 are
identical to those used to develop rates in the default database. The values for Tier 3 are
calculated as a weighted average of those for Bins 2 and 3, using Equation 1-49.
116
-------
T3 = 0.775-B2 +0.225-B3
Equation 1-49
Table 1-35. Hydrocarbons (HC): useful-life FTP standards and associated cold-start and hot-
running results on the FTP and US06 cycles. Values for the FTP and US06 represent NMOG and
NMHC, respectively.
Bin
8
5
4
3
2
Useful-life Standard
(mg/mi)
125
90
70
55
10
FTP Composite1
(mg/mi)
41.3
35.5
24.8
21.5
5.6
FTP Cold Start1
(mg)
591
534
383
329
87
FTP hot Running1
(Bag 2)
(mg/mi)
3.56
2.63
2.28
1.74
0.42
US062
(mg/mi)
35.8
35.8
35.8
35.8
2.60
Tier 33
1 Values
2 Values
3 Values
20
9.2
142
0.7 10.0
represent "non-methane organic gases" (NMOG).
represent "non-methane hydrocarbons" (NMHC).
for Tier 3 calculated using Equation 1-49.
Under a general assumption that CO standards are not forcing, but that CO emissions tend to
track NMOG emissions, corresponding values for CO were calculated in the same manner, and
are presented in Table 1-36.
Table 1-36. CO: Useful-life FTP standards and associated cold-start and hot-running results on the
FTP and US06 Cycles.
Bin
8
5
4
3
2
Useful-life Standard
(mg/mi)
4,200
4,200
4,200
2,100
2,100
FTP Composite
(mg/mi)
861
606
537
463
235
Cold Start
(mg)
6,680
5,510
5,500
3,470
1,620
FTP hot Running
(Bag 2)
(mg/mi)
451
238
201
119
70
US06
(mg/mi)
2,895
2,895
2,895
2,895
948
2,100
286
2,040
81
1,390
Values for Tier 3 calculated using Equation 1-49
Corresponding results for NO* are presented in Table 1-37. In contrast to HC and CO, the values
for Bin 2 were adopted for Tier 3, as the FTP composite of 5.5 mg/mi suggests that Bin-2
vehicles gives a compliance margin of about 50% with respect to the "effective standard" of 10
mg/mi.
117
-------
Table 1-37. NO*: Useful-life FTP standards and associated cold-start and hot-running results on
the FTP and US06 cycles.
Bin
8
5
4
3
2
Useful-life Standard
(mg/mi)
200
70
40
30
20
FTP Composite
(mg/mi)
64.2
21.2
8.7
5.7
5.5
Cold Start
(mg)
418
165
90
71
67
FTP hot Running
(Bag 2)
(mg/mi)
35.1
8.2
4.7
3.8
0.4
US06
(mg/mi)
61.3
45.9
30.6
30.6
18.4
Tier 3
10
5.5
67
0.4
18.4
1.5.2 Develop Phase-In Assumptions (Step 2)
We designed phase-in assumptions so as to project compliance with the Tier-3 fleet average
NMOG+NOx requirements. The requirements are shown in Table 1-38 for cars and trucks. The
phase-in begins in model year 2017 and ends in model year 2025.
Table 1-38. Target NMOG+NO* fleet average requirements for the Federal Test Procedure.
Model year
2017
2018
2019
2020
2021
2022
2023
2024
2025
FTP Composite, NMOG+NO* (g/mi)
LDV/T1
0.086
0.079
0.072
0.065
0.058
0.051
0.044
0.038
0.030
LDT21
0.101
0.092
0.083
0.074
0.065
0.056
0.047
0.038
0.030
1 Throughout, these results applied to Federal truck classes
LDT2, LDT3 and LDT4.
These results are also pictured in Figure 1-63. Note the sharp drop in emissions at the outset of
the Tier-3 phase-in, also that the truck requirements (LDT2,3,4) are slightly higher than those for
the lighter vehicles (LDV-T1). After 2017, the reduction in the fleet average is linear, and at the
completion of the phase-in, the fleet averages for cars and trucks no longer differ.
118
-------
Figure 1-63. NMOG+NO* FTP fleet average requirements during phase-in of the Tier-3 exhaust
emissions standards for light-duty vehicles.
LDV/LDT1
LDT234
O 0.040
u
£; 0.020
"" 0.000
in
-------
Figure 1-64. Phase-in assumptions, by standard and bin, for LDV-T1 vehicles.
o
rxl
O
rxl
O
rxl
O
rxl
O
rxl
O
rxl
rxl
O
rxl
rxl
rxl
O
rxl
m
rxl
O
rxl
^~
rxl
O
rxl
in
rxi
O
rxl
Model Year
Figure 1-65. Phase-in assumptions, by standard and bin, for LDT2 vehicles.
100%
90%
Model Year
120
-------
Figure 1-66. Phase-in assumptions, by standard and bin, for LDT3 vehicles.
o
o
a.
o
u
01
-------
1.5.3 Merge Cycle Results and Phase-In Assumptions (Step 3)
The goal of this step is to calculate weighted averages of the FTP (cold-start and hot-running)
results for all standards in each model year, with the emissions results weighted by applicable
phase-in fractions. We do this step for each vehicle class separately, then weight the four truck
classes together using a set of fractions also derived from the weighted sales estimates.
Figure 1-68 shows an example of the phase-in calculation for NO* from cars between model
years 2016 and 2025. The figure shows cold-start and hot-running FTP values for Tier-1, Tier-2
and Tier-3 standards, as well as the phase-in fractions for each standard in each model year.
Start and running emissions in each model year are simply calculated as weighted averages of
the emissions estimates and the phase-in fractions. The resulting weighted start estimates are
used directly to represent cold-start emissions for young vehicles in each model year (ages 0-3).
For running emissions, however, the averages are not used directly; rather, each is expressed as a
ratio to the corresponding Tier-1 value.
Table 1-39 shows weighted average values for model-years 2016-2025 for simulated FTP
composites, cold-start and hot-running emissions. The start values, expressed as the cold-start
mass increment (g), are used directly in the MOVES emission rate table to represent cold-start
emissions (operating mode 108). The composites and running emissions, expressed as rates
(g/mi), are presented for comparison. For running emissions, however, the averages shown in
the table are not used directly; rather, each is expressed as a ratio to the corresponding Tier-1
value, as shown in Figure 1-69 to Figure 1-71 below.
Figure 1-68. Example of phase-in calculation, for NO* from LDV-T1, for MY 2016-2025.
Standard
Tier 1 Tier 1
Bin 8
Bin5
Tier 2 Bin 4
Bin 3
Bin 2
Tier 3 Tier 3
Cold Hot
Start Running
(mg) (mg/mi)
888.00 127.00
417.87 35.07
165.42 8.21
89.72 4.69
70.89 3.78
67.18 0.38
67.18 0.38
Phase-In by
2016
0
0
0.890
0.060
0.010
0.015
0.000
2017
0
0
0.407
0.027
0.016
0.007
0.543
2018
0
0
0.356
0.024
0.014
0.006
0.600
2019
0
0
0.305
0.021
0.012
0.005
0.657
2020
0
0
0.254
0.017
0.010
0.004
0.714
Model Year
2021
0
0
0.204
0.014
0.008
0.003
0.771
2022
0
0
0.153
0.010
0.006
0.003
0.829
2023
0
0
0.102
0.007
0.004
0.002
0.886
2024
0
0
0.058
0.004
0.002
0.001
0.935
2025
0
0
0
0
0
0
1.000
Cold Start (mg) 154.32 107.85 102.76
97.69
92.60
87.52
82.43
77.35
72.99
67.18
Hot Running (mg/mi) 7.64 3.74 3.32 2.90 2.48 2.06 1.64 1.22 0.86 0.38
RATIOtoTierl 0.0601 0.0295 0.0262 0.0228 0.0195 0.0162 0.0129 0.00961 0.00677 0.00299
122
-------
Table 1-39. Weighted average FTP values projected for trucks and cars for MY 2017-2025.
regClass
Ref.1
Trucks
Cars
MY
2000
2017
2018
2019
2020
2021
2022
2023
2024
2025
2017
2018
2019
2020
2021
2022
2023
2024
2025
CO
Composite
(mg/mi)
1,620
541
434
412
391
369
348
327
305
286
426
408
391
373
356
338
321
306
286
Start
(mg)
11,400
4,749
3,625
3,395
3,164
2,934
2,704
2,474
2,246
2,037
3,566
3,375
3,184
2,993
2,802
2,610
2,419
2,255
2,037
Running
(mg/mi)
805
213
155
144
134
123
112
101
91
81
149
140
132
123
115
106
98
91
81
HC
Composite
(mg/mi)
126
28.3
20.7
19.0
17.3
15.7
14.0
12.3
10.7
9.2
20.5
19.1
17.7
16.3
14.8
13.4
12.0
10.8
9.2
Start
(mg)
1,530
462
341
314
287
260
233
206
179
154
339
316
293
270
247
224
201
181
154
Running
(mg/mi)
57.1
3.82
2.76
2.54
2.32
2.10
1.88
1.66
1.45
1.25
2.70
2.52
2.34
2.16
1.97
1.79
1.61
1.46
1.25
NO*
Composite
(mg/mi)
209
19.6
13.1
12.0
10.9
9.82
8.72
7.62
6.53
5.54
12.0
11.2
10.4
9.60
8.77
7.96
7.16
6.46
5.54
Start
(mg)
888
154
114
108
101
93.9
87.0
80.2
73.4
67.2
108
103
97.7
92.6
87.5
82.4
77.4
73.0
67.2
Running
(mg/mi)
127
8.24
4.45
3.86
3.27
2.68
2.09
1.50
0.91
0.38
3.74
3.32
2.90
5.48
2.06
1.64
1.22
0.86
0.38
1 The reference level represents Tier-1 LDV-T1.
123
-------
Figure 1-69. Weighted ratios for composite, start and running CO emissions, for (a) trucks and (b)
cars.
0.45
(a) Trucks
» Composite
B Cold Start
A Hot Running
2017 2018 2019 2020 2021 2022 2023 2024 2025
Model Year
2017 2018 2019 2020 2021 2022 2023 2024 2025
Model Year
124
-------
Figure 1-70. Weighted ratios for composite, start and running THC emissions, for (a) trucks and
(b) cars.
0.35
(a) Trucks
•Composite
Cold Start
• Hot Running
O 0.20 --
& 0.15 +
0.10
0.05
0.00
u.uzs i 0.022
2017 2018 2019 2020 2021 2022 2023 2024 2025
Model Year
0.047 0.044
1 T
n n3« n nor
O.Q38 O.Q35 n n,1
I I I I I f ' I I.
2017 2018 2019 2020 2021 2022 2023 2024 2025
Model Year
125
-------
Figure 1-71. Weighted ratios for composite, start and running NO* emissions, for (a) trucks and (b)
cars.
0.08
0.06
0.04
0.02
0.00
2017 2018 2019 2020 2021 2022 2023 2024
Model Year
2017 2018 2019
2020 2021
Model Year
1.5.4 Estimating Emissions by Operating Mode (Step 4)
To project emissions for Tier-2 and Tier-3 vehicles, we divided the operating modes for running
exhaust into two groups. These groups represent the ranges of speed and power covered by the
hot-running phase (Bag 2) of the FTP standards (< -20 kW/Mg), and the ranges covered by the
126
-------
SFTP standards (primarily the US06 cycle). For convenience, we refer to these two regions as
"the hot-running FTP region" and "US06 region," respectively, as previously shown in Figure
1-37 (page 85). To estimate emissions by operating mode, the approach was to multiply the
emission rates for MY 2000, representing Tier 1, by a specific ratio for each model year from
2016 to 2025, to represent emissions for that model year. For the FTP operating modes, we
applied the ratios shown in Figure 1-69 to Figure 1-71 above.
To estimate rates for the US06 modes, we followed a procedure similar to that for the "FTP"
modes, but using the "US06" columns in Table 1-35 through Table 1-37. For HC and CO, we
used Equation 1-49, as before. For NO*, we applied the Bin-2 values. Figure 1-72 shows
fractional reductions in the US06 rates along with similar reductions in "FTP" hot-running
operation, relative to levels for MY 2000 cars (Tier-1 levels). Note that the trends for "hot-
running" are identical to those in Figure 1-69 through Figure 1-71.
Figure 1-73 and Figure 1-74 show application of the ratios for cars to the FTP and US06
operating modes in model years 2010, 2017, and 2025, representing fully phased-in Tier-2
standards, an interim year during the Tier-3 phase-in, and the fully phased-in Tier-3 standards,
respectively. Rates for all three years are calculated with respect to rates for cars in model-year
2000 (using reduction ratios shown above in Figure 1-70 through Figure 1-72), applied to
selected operating modes for running operation. In these figures, the results are presented on
both linear and logarithmic scales. The linear plots display the differences in the high-power
modes, but obscure those in the low-power modes. The logarithmic plots supplement the linear
plots by making visible the relatively small differences in the lower power modes. They also
illustrate the varying degrees of control between the "hot-running FTP" (< 20 kW/Mg) and the
"US06" (> 20 kW/Mg) modes. In addition, the logarithmic plots include the level for MY2000,
which represents Tier-1 standards. Thus, these plots display the degree of running-emissions
reduction between Tierl and Tier 2 (MY2000: MY2010), and between Tier 2 and Tier 3
(MY2010: MY2025), across the full range of vehicle-specific power. Note that for simplicity,
these figures represent rates for operating modes 21-30, covering a wide range of power for
vehicles operating at speeds between 25 and 50 mph.
127
-------
Figure 1-72. Weighted ratios for hot-running emissions for cars and trucks, representing the "hot-
running FTP Region" and the "US06 Region," for (a) CO, (b) THC, and (c) NO*.
0.8
0.7
0.5
0.595
(a) CO
—*— US06 Region (CARS)
—A-- US06 Region (TRUCKS)
» FTP Hot-Running Region (CARS)
--«-> FTP Hot-Running Region (TRUCKS)
kQ.459
V 0.431
*•"—... . 0.402
° . 0.374
__ A . 0.346
' *- , A . 0.317
0.376 •——' *- f— . 0
°-360 0.344 *— *
2017 2018 2019 2020 2021 2022 2023 2024 2025
Model Year
0.5 + 0.461
0.4
O
'«
(b) THC
—*— US06 Region (CARS)
» FTP Hot-Running Region (CARS)
^298
—A— US06 Region (TRUCKS)
—«— FTP Hot-Running Region (TRUCKS)
I
--•a... °-240
T-»-.
0.211
2017 2018 2019 2020 2021 2022 2023 2024 2025
Model Year
0.4
0.3 --
(c) NOx
—*— US06 Region (CARS)
» FTP Hot-Running Region (TRUCKS)
—A_. US06 Region (TRUCKS)
—»— FTP Hot-Running Region (TRUCKS)
2017 2018 2019 2020 2021 2022 2023 2024 2025
Model Year
128
-------
Figure 1-73. Projected emission rates for cars in operating modes 21-30, vs. VSP, in ageGroup 0-3
years, for three model years, for (a) CO, (b) THC and (c) NO* (LINEAR SCALE).
1,000
900
2
13
LU
5 10 15 20 25 30
Vehicle Specific Power (kW/Mg)
35
40
2
»
re
(A
(A
LJ
5 10 15 20 25
Vehicle Specific Power (kW/Mg)
_ 16 -(c)NOx
2
&
re
£
c
_0
'35
(A
E
LJ
18
16
14
12
10
8
6
-5
5 10 15 20 25 30
Vehicle Specific Power (kW/Mg)
35
40
129
-------
Figure 1-74. Projected emission rates for cars in operating modes 21-30, vs. VSP, in ageGroup 0-3
years, for four model years, for (a) CO, (b) THC and (c) NO* (LOGARITHMIC SCALE).
0 10 20 30
Vehicle Specific Power (kW/Mg)
40
u>
*£
re
K
c
.0
'55
M
E
LU
100.00
10.00
1.00
0.01
0 5 10 15 20 25
Vehicle Specific Power (kW/Mg)
30
35
40
0 5 10 15 20 25
Vehicle Specific Power (kW/Mg)
30
35
40
130
-------
1.5.5 Apply Deterioration (Step 5)
Based on review and analysis of data from the Phoenix Inspection-and-Maintenance Program,
we assume that deterioration for different technologies is best represented by a multiplicative
model, in which different technologies, represented by successive model-year groups, show
similar deterioration in relative terms but markedly different deterioration in absolute terms. We
implemented this approach by translating emissions for the 0-3 age Group, as calculated above,
into their respective logarithmic means and applying uniform logarithmic age trends to all
model-year groups. We derived logarithmic deterioration slopes for Tier-1 vehicles (MY 1996-
98) and applied them to Tier-2 vehicles. In this process we applied the same logarithmic slope to
each operating mode, which is an extension of the multiplicative deterioration assumption.
For Tier-3 vehicles, the deterioration assumptions were modified to represent an extension of the
full useful life (FUL), which is increased from 120,000 mi to 150,000 mi. The extension of the
useful life, which assumes improved durability, was expressed through a reduction in
deterioration, i.e., a somewhat gentler deterioration trend.
To represent improved durability, differential reductions were applied to logarithmic
deterioration for THC and NO*. These reductions were applied such that when FTP composites
were reconstructed from the resulting emission start and running emission rates, the reduction in
deterioration was proportional to the increase in the useful life.
Note that in the reconstruction of FTP composites for NMOG+NOx, values for NMOG were
estimated from those for THC. In this step the fraction of THC representing NMOG for start
emissions was higher than that for running emissions1. In addition, start deterioration for THC
(and NMOG) followed lower relative rates than running deterioration, whereas start and running
NO* emissions deteriorate at the same relative rates (See 1.4.3, page 98).
The modification of the deterioration trend is shown in Figure 1-75. In MOVES, deterioration is
expressed in terms of vehicle age, rather than mileage. However, assuming typical mileage
accumulation of 10,000 to 15,000 miles per year, we assume that vehicles reach the end of their
useful lives between 10 and 15 years of age, or between the 8-9 year and the 10-14 year
ageGroups. Accordingly, deterioration for THC and NO* was modified so that the value of
NMOG+NOx reached in the 8-9 year ageGroup under a 120K FUL assumption, would be
reached in the 10-14 year ageGroup under a 150K FUL assumption. Note that in the plot the
value for each ageGroup is assigned to the midpoint of the group, i.e., the value for the 8-9 year
ageGroup is shown at 9 years, and that for the 10-14 year ageGroup is shown at 12.5 years.
We do not assume that emission rates for young vehicles, i.e., in the 0-3 year ageGroup, would
differ under either assumption, as vehicles meet the same levels at certification. Thus, the
deterioration trends for the two cases diverge as the vehicles age. We assume that the most rapid
changes in fleet means occur during the first 10 years of life, or between the 0-3 year ageGroup
and the 8-9 year ageGroup. After that point, we assume that the rate of change declines as the
fleet means tend to stabilize (especially in I/M areas, which are represented in the plot). The
ratio between values for the 150K FUL and 120K FUL starts at 1.0 and declines, stabilizing at a
value of 0.85 in the 8-9 year ageGroup, and remaining constant thereafter. Accordingly, we
For start emissions, NMOG = 0.89xTHC, and for running emissions NMOG = 0.5?xTHC.
131
-------
assume that the around the end of the useful life, mean FTP emissions for vehicles certified at
150K mi would be approximately 85% of those certified at 120K mi.
Figure 1-75. Simulated FTP composites for NMOG+NO*, calculated from MOVES start and
running rates, under 120K and 150K FUL assumptions.
o
o
10 15
Age (years)
20
25
1.5.6 Recalculate the logarithmic mean
Starting with the values of the arithmetic mean (xa) calculated as described in step 4 above, we
calculate a logarithmic mean (xi), as previously described in 1.3.4.2.5.2 and shown in Equation
1-40 (page 89).
1.5.7 Apply a logarithmic Age slope
After estimating logarithmic means for the 0-3 age class (x/,o-3), we estimate additional
logarithmic means for successive age classes (x/,age), by applying a linear slope in In-space (mi),
using Equation 1-41 (page 89).
The values of the logarithmic slope are adapted from values developed for the 1996-98 model -
year group. The values applied to Tier-2 and Tier-3 vehicles are shown in Table 1-40. The
reduced slopes for Tier-3 were calculated by reducing the Tier-2 values by 27% for HC and CO
and by 14% for NO*. These values were estimated empirically so as to implement the
assumption of reduced deterioration for the extended useful life. When calculating the age inputs
for this equation, we subtracted 1.5 years to shift the intercept to the midpoint of the 0-3 year age
Group.
132
-------
Table 1-40. Values of the logarithmic deterioration slope applied to running-exhaust emission rates
for MY following 2000.
Pollutant
CO
THC
NO,
Operating-Mode Group
"hot-running FTP"1
"US06"2
"hot-running FTP"
"US06"
"hot-running FTP"
"US06"
Logarithmic Slope (mi)
Tier 2
0.13
0.06
0.09
0.09
0.15
0.15
Tier 3
0.0949
0.0438
0.0657
0.0657
0.129
0.129
1 Includes opModelD = 0,1, 11-16, 21-25, 27, 33,35,37.
2 Includes opModelD = 28,29,30, 38,39,40.
1.5.8 Apply the reverse transformation
After the previous step, the values of x/^ge were reverse-transformed, as shown in Equation 1-30
(page 38).
1.5.9 Estimate non-I/M References (Step 6)
Completion of the preceding steps provided a set of rates representing I/M reference rates for
MY 2016-2025. As a final step, we estimated non-I/M reference rates by applying the same
ratios applied to the I/M references for default rates (section 1.3.3.7, page 50).
1.5.10 Start Emissions
The values for "FTP Cold-start" shown in Table 1-39 (page 123) were used to represent cold-
start emissions (opModeID=108). Rates for "warm" or "hot" starts following a range of soak
periods were estimated as for the default rates (see 1.4.1.1.3, page 95 ). Deterioration was
applied to start emissions, relative to that for running emissions, also as described previously (see
1.4.3, page 98).
1.6 Development of Emission Rates representing California Standards
In general, the principle of pre-emption does not allow the states to promulgate or enact their
own vehicle emission standards. However, due to the unique severity of the air pollution issues
in Southern California, the Clean Air Act allows the state of California to seek waivers of
preemption. When granted by EPA, such waivers allow California to enact and enforce its own
emissions standards, under the condition that such standards are at least as stringent as applicable
Federal standards.
California has enacted several such programs, beginning with Tier 0 (c. 1977-1992) and Tier 1 in
1993. These were followed by the "Low Emission Vehicle" programs, beginning with "LEV-I"
133
-------
in 19942 and continuing with "LEV-IT and "LEV-ID" in 2001 and 2015, respectively. Under
the LEV programs, multiple standard levels were assigned, designated as "Transitional Low
Emission Vehicle" (TLEV), "Low Emission Vehicle" (LEV), "Ultra Low Emission Vehicle"
(ULEV) and "Super Ultra Low Emission Vehicle" (SULEV).
Although assigned the same labels, each standard level can be assigned different numeric values
for each vehicle class, i.e., LDV, LDT1, LDT2, LDT3 and LDT4. For simplicity, we have
assumed that the California "Medium-Duty" classes, MDV2 and MDV3, can be treated as
equivalent to Federal LDT3 and LDT4 classes, despite differences in loaded vehicle weights.
In addition, Section 177 of the Clean Air Act allows other states to adopt California emission
standards, with the proviso that adopted standards are identical to standards for which waivers
have been granted. States do not need approval from EPA to adopt California standards. As of
2015, 15 states had elected to adopt California LEV-II standards for emissions of criteria
pollutants from varying classes of light-duty motor vehicles.40 Collectively, these states will be
called the "C A/SI 77" states.3 In addition, 12 of these states have adopted the LEV-III
standards.41
Effectively, then, two sets of emission standards are in place throughout the United States. One
outcome of this situation is that many vehicles coming to market over the past 20 years have
been certified to both CA and Federal standards. The analysis described in this section
incorporates this reality by applying an assumption that the emissions behavior of vehicles with
multiple certifications would be governed by the "most stringent" certification. For example, a
vehicle certified to Tier-2/Bin-5 Federally, but certified to LEV-II/SULEV in California, is
assigned to "Bin-2" or "SULEV" for purposes of developing emission rates, rather than to Bin 5.
This section describes the process used to develop a set of emission rates representing the LEV
programs, covering model years 1994-2031. The set of rates is stored in a segment of the
emissionRateByAge table, available for use with MOVES2014, although not included in the
default database.
The methods used are similar to those used to develop rates representing vehicles under the
Federal standards (NLEV, Tier 2 and Tier 3). In general, as the implementation of LEV
standards involved higher fractions of vehicles at lower standard levels than under the
corresponding Federal standards, rates for a LEV program in a given model year are equal to or
lower than corresponding "Federal" rates.
To apply this assumption, we developed the C A/SI 77 rates by scaling down the Federal rates by
appropriate margins. The calculations were performed in a series of steps, with the first two
steps identical to those used to develop the Federal rates. The following discussion assumes that
the reader is familiar with the relevant sections of this report (See 1.3.4 (page 68), and 1.5 (page
115)). However, the final step differs from that used to generate the default rates, as described
below.
2 The "National LEV" (NLEV) program was a voluntary program modeled on the LEV-I program, and applicable to
LDV, LDT1 and LDT2 vehicles.
3 These states include Connecticut, Delaware, Georgia, Maryland, Maine, Massachusetts, New Jersey, New Mexico,
New York, North Carolina, Oregon, Pennsylvania, Rhode Island, Washington and Vermont.
134
-------
1.6.1 Averaging IUVP Results
The calculation of CA/S177 rates uses the same set of average IUVP results as the default rates.
Equivalencies between Federal and corresponding LEV standards is shown in Table 1-41. Note
that the equivalences listed in the table are not exhaustive; they are limited to the subset that
were applied in developing emission rates.
Table 1-41. Selected equivalencies between Federal and corresponding CA/S177 standards.
Program
Fed.
Tier I1
NLEV
Tier 22
CA/S177
Tier I1
LEV-I
LEV-IP
Vehicle Class
Fed.
LDV-T1
LDT2
LDT3
LDT4
LDV, LDT1
LDT2
LDV, LDT1,
LDT2,3,4
CA/S177
LDV-T1
LDT2
MDV2
MDV3
PC, LDT1
LDT2
PC, LDT1,
LDT2,3,4
Standard Level
Fed.
LDV-T1
LDT2
LDT3
LDT4
TLEV
LEV
ULEV
TLEV
LEV
ULEV
Bin 5
Bin33
Bin 2
CA/S177
LDV-T1
LDT2
MDV2
MDV3
TLEV
LEV
ULEV
TLEV
LEV
ULEV
LEV
ULEV3
SULEV
1 Under Tier 1, each vehicle class was assigned a specific standard.
2 Under this program, there was no assigned correspondence between vehicle class
and standard level for the FTP standards, however, such an assignment remains in
effect for the SFTP standards.
3 This equivalence is exact for HC and CO only, for NO*, LEV-II/ULEV is
equivalent to Bin 5 (LEV-II/LEV).
1.6.2 Develop Phase-In assumptions
Differences between the CA/S177 and Federal programs are expressed primarily through the
phase-in assumptions. For this step we developed phase-in assumptions representing the phase-in
of California Tier-1, LEV-I and LEV-II programs. These assumptions cover model-years from
1994 through 2016. Starting in model year 2017 for cars, and 2018 for trucks, Federal rates are
harmonized with CA rates during the Tier-3/LEV-III phase-in and thereafter.
As with the default Federal phase-in assumptions, the CA/S177 phase-in was based on fractions
of sales, grouped by standard level and model year. The LEV phase-in, however, is simplified in
135
-------
that, as in the LEV-II standards, the three largest truck classes, LDT2, 3 and 4, were consolidated
into a single class, which we will refer to as LDT234.
Phase-in assumptions for passenger cars (PC) and light trucks (LDT1) are shown in Figure 1-76.
In model year 2009 and later, the CA/S177 fleet is dominated by ULEV, SULEV and LEV
vehicles, in that order. The phase-in for trucks (LDT234) is shown in Figure 1-77.
As a final step, a distinct "simplified" Federal phase-in was also developed. In this version, the
truck classes LDT2, LDT3 and LDT4 were also pooled, to facilitate comparison to the CA/S177
version.
Figure 1-76. Phase-In assumptions for CA Tier-1, LEV-I and LEV-II standards for passenger cars
and light-trucks (PC, LDV, LDT1).
QJ
1
• Tier 1
• LEV-I/TLEV
• LEV-I/LEV
• LEV-I I/LEV
• LEV-II/ULEV
• LEV-I I/SU LEV
Model Year
Figure 1-77. Phase-In assumptions for CA Tier-1, LEV-I and LEV-II standards for trucks (LDT2,
LDT3, LDT4).
r LJ
T t
3 r
-. o
3 C
n c
3 ?
-\ f
1 ;
'•: >
1 >
3 F
-. o
3 f
n c
3 T-
-\ f
T •:
r L
1 U3
ITier 1
I LEV-I/TLEV
I LEV-I/LEV
I LEV-I/ULEV
ILEV-II/LEV
LEV-II/ULEV
ILEV-II/SULEV
Model Year
136
-------
1.6.3 Merge FTP Results and Phase-In assumptions.
In this step the FTP results and phase-in assumptions were merged so as to calculate weighted
average results for composites, cold-start and hot-running emissions, as described in 13.4.23
(page 78). However, as the truck classes were pooled and assigned a uniform phase-in,
calculating weighted averages by truck class did not play a role in these calculations as in the
default calculations.
This step was repeated for the CA phase-in and for the "simplified" Federal phase-in.
Sets of weighted averages by model year are shown in for FTP Composite Emissions (Figure
1-78), FTP cold-start emissions (Bag 1 - Bag 3) (Figure 1-79), and FTP hot-running emissions
(Bag 2) (Figure 1-80).
Figure 1-78. Weighted average FTP composite emissions for cars and trucks, for Federal and
CA/S177 standards.
THC, Trucks
THC, Cars
3s
|
8
;
N
N
v\
v
\
1
\
s
"•-"A,
«— CA/S177
3— Fed TO,T
LEV 1,11,111
,-.
JW
s
s
LL.
innnn^
\
,
\
-W«eg3
HB^,,
»— CA/S177
3— Fed TO,T
LEV 1,11,111
1,NLEV,T2
1990 1995 2000 2005 2010 2015 2020 1990 1995 2000 2005 2010 2015 20
CO, Trucks CO, Cars
" ,„
O "
i- 20
£ 2'°
51 c
Q. 1 n
11 "
\
\
1
«s
L
\
\
t"\
\ '
\
k_
Va-
^^
•— CA/S177
3— Fed TO,T
1,NLEV,T2
S 30
o ^^
i- 20
£ 2'°
1.5 •
a. 1 n
11 ;=
S3
_^^.
%
V
^^
»— CA/S177LEVI,II,III -
3— FedTO,Tl,NLEV,T2 -
1990 1995 2000 2005 2010 2015 2020 1990 1995 2000 2005 2010 2015 20
.._ _ . NOx, Cars
NOx, Trucks
£
O
5,
3
^-
V
\ ,
\s
\
i
Vr\
—
tS-T^
•— CA/S177LEVI,II,III -
a— FedTO,Tl,NLEV,T2 .
£ ;
O :
\
s
X
^
_
******
•— CA/S177LEVI,II,III -
B-FedTO,Tl,NLEV,T2 .
1990 1995 2000 2005 2010 2015 2020 1990 1995 2000 2005 2010 2015 20
137
-------
Figure 1-79. Weighted average FTP cold-start emissions, for Federal and CA/S177 standards.
THC, Trucks
THC, Cars
CO, Trucks
CO, Cars
NOx, Trucks
NOx, Cars
3"
(0
"° 0 8
3 ^
t -
a
^
«-l
V
v^
X '
s
\
t
"^V
S
—
^Sra-^,
Thin
•— CA/S177
3— FedTO.T
LEV 1,11, III
1,NLEV,T2
31'4
(0
3
\
^_ „ „ „ .
X,
I
\
-^
•— CA/S177
3— FedTO.T
LEV 1,11, III
1,NLEV,T2
138
-------
Figure 1-80. Weighted average FTP hot-running emissions (Bag 2), for trucks and cars, under
Federal and CA/S177 standards.
THC, Trucks
THC, Cars
2s
c
3
cc
£ 0.10
I
t
\
1
•*,
1
\
1
v^
—
•— CA/S177
3— FedTO.T
LEV 1,11,111
1,NLEV,T2
2§
c
3
CC
I
£0.05
n nn
_^S
^B D D D [
i
_
•— CA/S177 LEV 1,11, III
3— FedTO,Tl,NLEV,T2
CO, Trucks
CO, Cars
NOx, Trucks
"SB
no
3
\
\
^N
L
\
V
v^
X-
\
^
w^
^-^te
^***,
CA/S177 LEV 1,11,111
3— FedTO,Tl,NLEV,T2
• T T T T •
,_
NOx, Cars
"SB
no
2
Q
I
.
•Vgnni
s
L
X^
CA/S177 LEV 1,11, III
3— FedTO,Tl,NLEV,T2
1995
2005
2020
1990
2000
2010
2015
1.6.4 Scaling CA/177 rates to Federal Rates
At this point the next step in the calculation differs from the approach used to generate the
default Federal rates. As in the calculation of the default rates, we normalized hot-running
emissions for both FTP and US06 to Federal Tl levels, represented by MY1998. However, in
this calculation, we also performed this normalization for cold-start rates. The results were sets
of ratios relative to Tier 1 for both running and start emissions.
Next, we calculated ratios of the weighted CA ratio to its Federal counterpart, by model year, as
shown in Equation 1-50,
-"-CAiFed ~ "
vFed
Equation 1-50
139
-------
where ,RcA:Fed = the ratio of the C A/SI 77 weighted average to that for the Federal phase-in, and
Rfed and RCA are ratios of respective weighted averages to that for MY1998, in the CA/S177 and
Federal phase-ins, respectively. Note that if raw values of RcA-Fsd were > 1.0, they were
adjusted to 1.0, under than assumption that fleet averages under the LEV program(s) would be <
corresponding averages under the Federal program(s).
Values of ,RcA:Fed are presented in Figure 1-81 below. Note that ratios were calculated and
applied separately for each of the three gaseous emissions (HC,CO,NOX) and for start emissions
(opmodeid = 101-108), "FTP Bag-2" emissions (opmodeid = 0,1, 11-16, 21-27, 33-37) and
"US06" emissions (opmodeid = 28-30, 38-40).
In MY2017 and later, following the onset of the Tier-3/LEV-III phase-in, all ratios are set to 1.0,
to reflect an assumption that under T3, the Federal program is targeting the same NMOG+NOx
fleet average requirements as LEV-III. See 1.5, page 115.
Figure 1-81. Ratios of relative emission levels by model year under CA/S177 and Federal
standards, both individually normalized to "Tier-1" levels (See Equation 1-50).
1.0 q
THC, Trucks
Start
Running (FTP B2)
Running (US06)
1.0 q
THC, Cars
CO, Trucks
CO, Cars
NOx, Trucks
NOx, Cars
140
-------
1.6.5 Availability
A segment of the emissionRateByAge table containing the CA/S177 rates is available at
http://www.epa.gov/otaq/models/moves/tools.htm. See "Tools to develop special case MOVES inputs."
Guidance for application of the rates is also available at the same location.
1.6.6 Early Adoption of National LEV Standards
The National Low Emission Vehicle Standards program was adopted in 2001. However, a group of states
in the "Northeast Trading Region" (NTR) adopted the standards early, in 1999. Using an approach
identical to that used to develop the CA/S177 rates, we developed a table segment for the
emissionRateByAge table representing the adoption of NLEV rates in model years 1999 and 2000. As
with the national program, "early" NLEV applied only to the LDV, LDT1 and LDT2 vehicle classes.
As with the CA/S 177 rates, we developed phase-in assumptions specific to "early" NLEV. Figure 1-82
shows that fractions of Tier-1 vehicles start declining markedly in MY 1999, whereas in the default phase-
in, the fractions for Tier 1 are 100% until MY2001 for LDV-T1 and LDT2. The fractions shown apply to
LDT2, as well as to LDV-T1. Vehicle classes LDT3 and LDT4 remain in Tier 1 until the onset of Tier 2,
in MY2004.
The NTR rates were developed by scaling default rates for start and running emissions down
appropriately as implied by the differences in phase-in assumptions.
The table segment for early NLEV rates is stored at the same location as the CA/S 177 rates
(http://www.epa.gov/otaq/models/moves/tools.htm). It contains only model years 1999 and 2000, and can
be used as a user-input table.
Figure 1-82. Phase-in assumptions for early NLEV adoption, for LDV, LDT1 and LDT2.
Tier 1
ITLEV
I LEV
IULEV
1998
1999
2000
Model Year
2001
2002
141
-------
1.7 Replication of Rates
The rates developed as described in Section 1 represent gasoline-fueled conventional-technology
engines. For purposes of the inclusion in the emissionRateByAge table, we replicated these rates
to represent other fuels and technologies.
At the outset, we replicated the entire set of gasoline rates for high-level ethanol blends, i.e.,
"E77" through "E85." However, for lower-level ethanol blends (i.e., 0-20 vol.%), the effect of
ethanol (and other effects related to blending) is represented through fuel adjustments, rather
than through the base rates, as described in this document. The development and application of
fuel adjustments is described in a separate report.35
Table 1-42. Fuel types and engine technologies represented for gaseous-pollutant emissions from
light-duty vehicles.
Attribute
Fuel type
Engine Technology
sourceBin attribute
faelTypelD
engTechID
Value
01
02
05
01
30
Description
Gasoline
Diesel
Ethanol (E77, E85, etc.)
Conventional internal combustion (CIC)
Electric
142
-------
2 Particulate-Matter Emissions from Light-Duty Vehicles
2.1 Introduction and Background
A large body of research is available on the formation and measurement of particulate matter
(PM) emissions from combustion engines. This chapter describes the process by which
emissions measured in a subset of the PM research programs on light-duty gasoline vehicles was
employed to generate emission rates for MOVES. The emission rates developed by this
approach embody a strictly "bottom-up" method whereby emission rates are developed from
actual vehicle measurements following intensive data analysis, and are then structured as inputs
to the emissions inventory model. This is in contrast to a "top-down" approach which uses
measurements of ambient PM concentrations from local regions and may apportion these
emissions to vehicles (and other sources), which are then input into inventory models.
The primary study that this chapter relies on is the "Kansas City Characterization Study"
conducted in 2004-2005.42 The Environmental Protection Agency and several research partners
conducted this study to quantify tailpipe particulate-matter emissions from gasoline-fueled light
duty vehicles in the Kansas City Metropolitan Area. This study is the most comprehensive and
representative study of its kind. In the context of a rigorous recruitment plan, strenuous efforts
were made to procure a representative sampling of the fleet. During the summer and winter
phases, 261 and 278 vehicles were measured, respectively, with some overlap between the
phases. The measurements were conducted on a portable dynamometer using the LA92 driving
cycle under ambient temperature conditions.
Analyses of some of the data from this program are presented in the report: "Analysis of
Particulate Matter Emissions from Light-Duty Gasoline Vehicles in Kansas City."43 This
"analysis report" (which is the partner to this chapter) presented preliminary emission rates for
PM, elemental carbon fraction (EC) and organic carbon fraction (OC), as well as temperature
adjustment factors for start and hot-running emissions processes. These preliminary results form
the basis for the emission rates developed in this chapter. The rates in the analysis report are
based on aggregate or "bag" emissions measured on the filters, and are thus presented as
grams/start for start emissions and grams/mile for hot running operation.
The dataset included vehicles manufactured over several decades, measured at various ages
during CY2004-05. Thus, the program taken alone did not enable us to forecast emissions for
current vehicles as they age, or to backcast emissions of older vehicles when they were young.
This chapter describes the development of a deterioration model based on a comparison of
former PM studies with the 2005 Kansas City study. The rates from this deterioration model
allow both forecasting and backcasting as required by MOVES.
In addition, the previous analyses did not attempt to translate results measured on the LA92 cycle
(used in Kansas City) into terms of other cycles (such as the FTP) or to "real-world" driving. As
with the gaseous pollutants, MOVES has the capability to represent hot running "modal"
emission rates so that emissions vary depending on the driving pattern represented. The
operating modes defined for PM are the same as for the gaseous emissions (see Table 1-5, page
15). This chapter describes how the continuous PM measurements collected in the study were
used to populate the modal rates for MOVES. Because of the reliance on continuous PM
measurement, it is worth describing the measurement procedures used in this program.
-------
2.1.1 Particulate Measurement in the Kansas City Study
For measurements conducted on the dynamometer, vehicles were operated over the LA92
Unified Driving Cycle (see Figure 2-1). The LA92 cycle consists of three phases or "bags."
Phase 1 ("bag 1") is a "cold start" that lasts the first 310 seconds (1.18 miles). "Cold start" is
technically defined as an engine start after the vehicle has been "soaking" in a temperature
controlled facility (typically ~72°F) with the engine off. In the Kansas City study, the vehicles
were soaked overnight under ambient conditions. Phase 1 is followed by a stabilized Phase 2 or
"hot running" (311 - 1427 seconds or 8.63 miles). At the end of Phase 2, the engine is turned off
and the vehicle is allowed to "soak" in the test facility for ten minutes. At the end of the soak
period, the vehicle is started again, and is driven on the same driving schedule as Phase 1. This
Phase 3 is called a "hot start" because the vehicle is started when the engine and after-treatment
systems are still hot. Criteria pollutants were measured both in continuous and aggregate modes.
Particulate was collected during each of the three phases on 47 mm Teflon filters at 47°C ± 2°C.
Figure 2-1. Phases 1 and 2 of the LA92 Cycle, representing "cold-start" and "hot-running"
operation, respectively.
70.0
60.0
200
400
600
800
1000
1200
1400
time
In addition to the gaseous pollutants measured via the constant-volume sampler (CVS),
continuous measurements of total PM mass were taken using two instruments. The first was an
Booker Systems Model RPM-101 Quartz-crystal microbalance (QCM) manufactured by Sensors,
Inc.; the second was a Thermo-MIE Inc. DataRam 4000 Nephelometer. In addition to total
mass, estimated black carbon was measured continuously with a DRI photoacoustic instrument.
In addition, integrated samples were collected and analyzed by DRI for PM gravimetric mass,
elements, elemental and organic carbon, ions, particulate and semi-volatile organic compounds,
and volatile organic air toxics. All sampling lines were heated and maintained at 47°C ± 2°C.
The samples were extracted from the dilution tunnel through a low particulate loss 2.5 |j,m
144
-------
outpoint pre-classifier. Further details and a schematic of the sampling instrumentation are
shown in Figure 2-2 and Figure 2-3.
Figure 2-2. Schematic of the constant-volume sampling system used in the Kansas-City Study.
from vehicle
tailpipe
^
Diluted exhaust to aldehyde sample aldehyde
at 46 c flow comtroller i cartridge
/
to particle sample * heated i
flow controller i sample I I
particle filter
vent
background
V
Air Conditioner
«fc '^ sample line
^ \v
ution air \F~~
sater (46 C) E. J
Backgrd HC analyzer
vent
Llap 1 .
pump -^1
filter — *" |
•
flow ^~-J*'
measurement
and control
flow
measurement
and control
low CO
analyzer
* i *
igh CO NOx analyzer
nalyzer
high
CO2 analyzer
^
Positive
Displacement
Pump (POP)
Heated
HC analyzer
Dilute exhaust
collection bags
<-Dyno
CVS 10cm, 5
Figure 2-3. Continuous PM analyzers and their locations in the sample line.
PM2.5IMPACTORS
DusTrak
(light
scattering)
51pm
AJER
•ROL
i
i
i
i
i
HE ATE
CONTRC
1m, 4.51pm
QCM CART SYSTEM
mass)
Data RAM
(light
scattering)
PHOTO ACOUSTIC
(black carbon)
It is worth briefly describing the apparatus used to measure PM on a continuous basis. A more
thorough description may be found in the contractor's report.44 As of the date of this program,
145
-------
measuring continuous particulate was a daunting technical challenge. Each technique has
specific advantages and disadvantages. For this study, the cumulative mass as measured on the
Teflon filters was treated as a benchmark. Thus, prior to using the continuous measurements to
estimate modal emissions, the sums of the time series for the continuous measurements were
normalized to their corresponding filter masses to compensate for systematic instrument errors.
The Quartz Crystal Microbalance measures the cumulative mass of the PM deposited on a crystal
face by measuring the change in its oscillating frequency. It is highly sensitive to many artifacts
such as water vapor and desorption of lighter organic constituents. Due to the high degree of
noise in the continuous time series, the measurements were averaged over 10 seconds, thus
damping the temporal effects of transients. The QCM can accurately capture cumulative PM
over time, however, measurement uncertainties increase for successive points in time because the
values depend on a calculated difference between two sequential, and similar, measurements.
Due to the resulting high variability, including large and rapid fluctuations from positive to
negative emissions at any given instant, and vice versa, use of the QCM measurements was not
viewed as a practical option for use in emission rate development for MOVES, except as a check
on the other instruments.
The Dustrak and Dataram both work on light-scattering principles. As such, they have very
rapid response times and can measure larger PM volumes with reasonable accuracy. However,
their accuracy degrades when measuring low PM volumes. Since most PM mass lies within the
larger particles, the instruments should be able to capture most of the continuous mass
concentrations though it may miss a substantial portion of the smaller (nano) particles. To
provide a qualitative check on this supposition, the time-series for the QCM and optical
instruments were aligned and checked to ensure that significant mass was not missed. Based on
this analysis, the Dustrak instrument was observed to be the most reliable of the 3 instruments,
and mass correction at low loads was not judged to be worth the effort given the uncertainties
involved. This time-consuming analysis was done by eye for each test and the results are not
presented in this chapter.
The photoacoustic analyzer (PA) is unique among the continuous instruments in its ability to
capture only the soot or elemental carbon components of PM. The fast analyzer detects the
resonances coming off the carbon-carbon bonds in soot. Unfortunately, there were insufficient
Thermal Optical Reflectance (TOR) elemental carbon (EC) measurements from quartz filters to
normalize the PA data, but some comparisons are shown in the contractor's report44. In this
study, the PA data were compared qualitatively with the Dustrak and Dataram and found to be
consistent with expected ratios of elemental to total carbon during transient events, leading to the
conclusion that these instruments were largely consistent. These results are also not presented in
this chapter as every single trace was compared by eye. The data is used to determine the modal
relationship of elemental to total PM.
Due to the uncertainty of experimental measurement techniques for continuous PM at the time of
the Kansas City study, these instruments were employed only as a semi-qualitative/quantitative
means of determining modal emission rates, and the use of such data does not qualify them as
EPA recommended or approved devices or processes.
146
-------
2.1.2 Causes of Gasoline PM Emissions
In gasoline-powered spark-ignition engines, particulate matter is formed during incomplete fuel
and oil combustion (although the amount of oil consumed in combustion and its contribution to
PM varies greatly from vehicle to vehicle). During operation, numerous distinct technologies
used in vehicles are in various states of repair or disrepair which also affect PM emissions. Even
brand new vehicles emit PM from combustion but at very low levels. A complete description of
the causes of PM emissions and associated mechanisms is beyond the scope of this report, as
many aspects of the science that are still not fully understood. We will briefly summarize factors
that contribute to gasoline PM in the vehicle fleet in this section. Where appropriate, we will
make comparisons to the mechanisms of hydrocarbon (HC) formation, since parallels are often
drawn in the literature.
Simply put, particulate matter forms primarily during combustion when carbon-containing
molecules condense or otherwise agglomerate. This form of PM is generally composed of
higher molecular weight hydrocarbon compounds, some of which originate in the fuel or oil and
some of which are formed during combustion. Unlike diesel engines, elemental (molecular)
carbon or soot is not very prevalent with gasoline engines but does form in larger quantities
under relatively rich conditions, e.g., low airfuel ratios at cold start. The amount of elemental
carbon in PM varies from vehicle to vehicle (and, even for a given vehicle, varies depending on
operating conditions and state of repair). Studies conducted earlier in the 2000's concluded that
primary PM emissions from gasoline-fueled vehicles are dominated by organic carbon
emissions, and diesel emissions are dominated by elemental carbon emissions.45>46>47 However,
this pattern may be changing with the introduction of diesel particulate filters and emerging
technologies for gasoline vehicles, such as gasoline-direct injection.
Other compounds in the fuel or engine oil, such as trace levels of sulfur and phosphorus, form
sulfates and phosphates during combustion, both of which form particulate. The sulfur level in
gasoline is now very low, almost eliminating sulfate formation from gasoline sulfur content but
motor oil still contains higher fractions of sulfur (and phosphorus) compounds. Also, trace metal
constituents in gasoline and oil form PM in the combustion process as metallic oxides, sulfates,
nitrates, or other compounds. Catalyst attrition products from the substrate and trace amounts of
noble metals also form PM but not in the combustion process. The catalyst attrition products are
mechanically generated and are usually in larger size ranges compared to exhaust PM. Exhaust
PM as formed in the engine is generally very small in size (possibly much of it in nuclei mode
particulate in the range of 0.05 microns or smaller). In the exhaust system, including the muffler,
some of the PM agglomerates and increases in size.
The wide assortment of technologies used in vehicles can affect PM formation. These
technologies were mainly developed to control HC, CO and NO* emissions, but most have the
side benefit of reducing PM, since reducing exhaust HC generally also reduces exhaust PM
although not to the same extent. Older engines from the 1980s and earlier that deliver fuel
through a carburetor typically have poorer fuel droplet quality, as well as looser control of fuel-
air stoichiometry. These older vehicles are expected to produce more PM (on average) than their
fuel injected counterparts that followed in the late 1980s and early 1990s.
Among fuel-injected engines, throttle-body fuel injection (TBI) used in earlier engines with fuel
injection typically has poorer fuel atomization quality and airfuel ratio control than the port fuel-
injection (PFI) technology that replaced it; thus, one might expect older fuel-injected vehicles to
147
-------
have higher PM emissions (on average) than newer ones. Somewhat before the widespread
adoption of fuel injection, closed-loop control systems were developed in tandem with oxygen
sensors to improve the stoichiometric chemistry of combustion. These closed-loop controls
improved combustion as well as the effectiveness of the after-treatment system.
The after-treatment system on most vehicles consists of a 3-way catalyst. The 3-way catalyst was
designed for simultaneous control of hydrocarbons, carbon monoxide, and nitrogen oxides.
Vehicles with 3-way catalysts would meet more stringent hydrocarbon and carbon monoxide
emission standards while also meeting the first stringent nitrogen oxide standard. In oxidizing
hydrocarbons, these systems also improved PM control. These systems were utilized on almost
all gasoline-fueled vehicles beginning in the 1981 model year. On some vehicles manufactured
in the 1980s and a few more recently, a secondary air injection system was added between the
engine and oxidation portion of the catalyst in order to add supplementary air to the oxidation
reactions on the catalyst. These systems also helped oxidize PM (though probably not to the
extent as CO or HC). The deterioration of these technologies may affect PM and HC quite
differently. Throughout this chapter, there are parallels drawn between HC and PM formation as
well as control, however it should be noted that the correlation between these emissions is far
from perfect. Many examples of this relation are shown in the 2008 analysis report.43
Amounts of PM emitted are very sensitive to the amount of fuel in combustion as well as the
stoichiometry of the reaction. As mentioned above, over-fueled mixtures result in higher PM
formation and in some cases, also excess soot formation. Over-fueling can occur under several
different conditions. During cold start, engines often run rich to provide sufficient burnable fuel
(i.e. light ends that vaporize at colder temperatures) to start combustion when the cylinder walls
are still cold (which results in increased flame quench). Additionally, under cold conditions (e.g.,
below 20°F), additional enrichment of the fuel-air mixture is needed to start the engine and lasts
longer while the engine warms up.
When high acceleration rates or loads are encountered (such as wide-open throttle events), extra
fuel is often injected for greater power or for catalyst and component temperature protection.
Emission control systems since the late 1990s are better designed to control this enrichment.
Finally, engines can run rich when a control sensor (e.g. oxygen, MAP, MAP, or coolant
sensors) or the fuel system fails.
In addition to fuel, lubricating oil can get into the combustion chamber via several pathways.
Some engines may have poor tolerances for pistons and piston rings, thus the negative pressures
(engine intake vacuum) can pull oil through these larger gaps during the intake stroke.
Furthermore, engine components, such as valves, valve seals, piston rings, and turbochargers can
wear and deteriorate resulting in increasing emissions over time. In all gasoline automotive
engines, the crankcase (where the oil bathes the engine components) is vented back into the
combustion chamber through the intake manifold. This technology, known as Positive
Crankcase Ventilation (PCV), is required to remove and burn excess hydrocarbons in the hot
crankcase. Unfortunately, it can also introduce PM precursors and oil into the engine
combustion chamber. Because of the relatively small amount of oil consumption compared to
the volume of gasoline burned in a vehicle, the amount of HC from oil is typically small.
However, organic PM from oil consumption can be quite substantial because oil is composed of
high molecular-weight hydrocarbons, and more likely to persist as unburned droplets. Therefore,
as vehicles age, those that consume more oil will probably have very different emissions
behavior for HC than for PM, compared to when they were new. However, oil consumption can
148
-------
"poison" the catalyst substrate, reducing the effectiveness of the catalyst at oxidizing HC.
Analysis of the Kansas City program estimated that lubricating oil contributes from 13% to 37%
of the primary PM emissions, with the older vehicles contributing most of the oil-derived PM
emissions.48
The fuel itself may have properties that exacerbate PM formation, which may be affected by
concentrations of sulfur, lead, aromatics, and impurities. With the lower levels of lead and sulfur
in fuels, the first two are presumably less of a factor in the Kansas City program than aromatics.
In the calendar years that MOVES models, lead is not a significant portion of the inventory, and
is thus largely ignored.
Some of these PM forming mechanisms clearly affect HC emissions. A control technology or a
deterioration path for HC may or may not similarly affect PM depending on the source. It is also
likely that the processes that cause high PM may not be the same processes that cause organic
PM. Some of the mechanisms also form visible smoke. Smoke takes on a variety of
characteristics depending on the source, and can be due to oil consumption or over-fueling. The
smoke is visible because of the relative size of the particles compared to the light wavelengths
that are scattered. However, visible smoke is not necessarily a reliable indicator of high PM
emissions.
2.2 New Vehicle or Zero Mile Level (ZML) Emission Rates
In this section, we develop an approach to extend the PM results from the Kansas City Study to
estimate average emissions across the fleet. The section also compares the new vehicle results
from many different studies in order to estimate "zero mile" level (ZML) emission rates for all
model years. Before modeling deterioration, it is first necessary to capture ZML emission rates.
In constructing a model of emissions from the Kansas City data (Figure 2-4), the greatest
challenge is distinguishing model-year and age effects. As with most datasets, this issue arises
because the program was conducted over a two-year period, thus ensuring a direct
correspondence between model year and age. As a result, it is very difficult to distinguish the
reduction in emissions with model year from the increase in emission with age. Emissions tend
to decrease as technologies are introduced on vehicles (with later model years) in order to
comply with more stringent emissions standards. However, these technologies and vehicles tend
to deteriorate over time, thus for the same model year vehicle, older vehicles (greater age) will
have higher emissions (on average) than newer vehicles.
149
-------
Figure 2-4. Average particulate emission rates from the Kansas City study, by model year, shown
as cycle aggregates on the LA92.
I :
n -
i '
1
,
1
i
i
* KC measured
KC 5 yr measured avq
' "I III
J ' 1 ^ ' i ? , : i ,
1975
1980
1985 1990
Model Year
1995
2000
In concept, the most accurate means of quantifying emissions from vehicles over time is to
conduct a longitudinal study, where emissions are measured for the same vehicles over several
(or many) years. However, implementing such a study would be costly. Moreover, it is
impossible to obtain recent model year vehicles that have been significantly aged. In the
following sections, we will describe some limited longitudinal studies conducted in the past.
Then we will present our modeling methodology to isolate model year (technology) in this
chapter from age (deterioration) in the next.
2.2.1 Longitudinal Studies
There have been a few longitudinal studies conducted in the past that are relevant for PM
emissions. Unfortunately, they are all limited in their ability to conclusively distinguish model
from age effects.
Gibbs et al. (1979) measured emissions from 56 vehicles with mileage ranging from 0 to 55,000
miles (odometer) on 3 different cycles.49 Hydrocarbon emissions were analyzed, but
unfortunately, PM results were not reported as a function of mileage. The authors state that
"emission rates of measured pollutants were not found to be a consistent function of vehicle
mileage," however, the following figure shows that some increasing trend seems to exist for HC
(Figure 2-5).
150
-------
Figure 2-5. Hydrocarbon emissions as a function of mileage (Gibbs et al., 1979).
E
M>
±
3.5 -,
3 -
2.5 -
2 -
1.5 -
1 -
0.5 -
0 -
*
* +
+ +
+ *
*
*
11111
0 10 20 30 40 50 60
mileage ('1000)
Hammerle et al. (1992) measured PM from two vehicles over 100,000 miles.50 However, their
results for PM deterioration are somewhat inconclusive, as the following figure shows, since the
deterioration seems to occur mainly in the beginning of life, with very little occurring after
20,000 miles. Also, the study is limited to two specific vehicle models.
Figure 2-6. Particulate emissions as a function of odometer for two Ford vehicles (Hammerle et al.,
1992).
y=2&05< + 0.3218
y = SE-QQx + 1.4717
60000
odometer
Both of these studies assume that odometer is a surrogate for age. While there are some
deterioration mechanisms that worsen with mileage accumulation, there are others that
deteriorate with effects that occur over time, such as corrosion due to the elements, deposits and
impurities collecting in the gas tank and fuel system, etc. Therefore, we believe that any study
151
-------
that describes deterioration as a function of odometer (alone) may not account for all causes of
deterioration.
Whitney (2000) re-recruited 5 vehicles that had been measured in previous study 2 years prior
(CRC-E24).51 There are two significant limitations of this follow-up study: (1) the interval
between studies was only 2 years, though the odometers had increased 22,200 miles (on average)
and (2) these vehicles were tested on a different drive cycle, the LA92 compared to the previous
study, which used the FTP. We will explore the potential cycle differences on PM later, but
assuming the cycles give similar PM results, the PM emissions were only 8% higher (on
average). This increase is due to a single vehicle, which had significantly increased PM
emissions (the rest were the same or slightly lower). Unfortunately, this is not a large enough
sample and time period on which to resolve age effects, but it may be sufficient to conclude that
the differences between PM from the FTP and LA92 drive cycles are minimal for PM.
The three longitudinal studies described above are inconclusive, though they do hint that
deterioration does occur.
2.2.2 New Vehicle, or ZML Emission Rates and Cycle Effects
In order to isolate the effect of model year (technology) from age (deterioration), it is useful to
look at the model-year effect independently. This goal can be achieved by analyzing emissions
from new vehicles from historical studies. New vehicle emission rates tend to have lower
variability than older vehicles (in absolute terms) since they have lower emissions that comply
with more stringent HC standards. These standards, which decrease over time, tend to affect PM
emissions as well since many of the mechanisms for HC formation also form PM.
Several independent studies have measured PM emissions from nearly new vehicles. For our
purposes, we will define "new" as a vehicle less than 3 years old, i.e., vehicles within the 0-3
year age Group. Table 2-1 lists the 15 studies employed for this analysis.
Table 2-1. Historical gasoline PM studies including new vehicles at time of study.
Program
Gibbs et a/.49
Cadleetal52
Urban & Garbe53'54
Lang et a/.55
Volkswagen56
CARB57
Hammerleefa/., 199250
CRC E24-1 (Denver)58
CRC E24-2 (Riverside)59
CRC E24-3 (San Antonio)60
Chased a/.61
Whitney (SwRI)51
KC (summer)43-44
EPA (MSAT)62
Li et a/., 200663
Year of
Study
1979
1979
1979, 1980
1981
1991
1986
1992
1996
1997
1998
2000
1999
2004
2006
2006
No.
vehicles
27
3
8
8
7
5
2
11
20
12
19
13
4
3
Drive cycle
FTP
FTP
FTP
FTP
FTP
FTP
FTP
FTP
FTP
FTP
FTP
LA92
LA92
FTP
FTP, LA92
152
-------
Before we examine these emissions, we should convince ourselves that the LA92 driving cycle
will not give substantially different PM emissions than the FTP so that we can compare these test
programs directly. As described above, the results from Whitney (2000) seem to indicate little
difference between the two cycles. Even though the tests were conducted 2 years apart, one
would expect that the aging effects in combination with the slightly more aggressive LA92 cycle
(used later) would have given higher PM emissions. However, this was not the case, and only
one of the 5 vehicles showed significantly increased emissions.
Li et a/., (2006) measured three vehicles on both cycles at the University of California,
Riverside.63 The PM emissions from the LA92 were 3.5 time larger (on average) than the FTP
results. However, the HC emissions were only 1.2 times higher. These results seem rather
contradictory and inconclusive. The 3.5 factor also seems excessive in relation to other results,
such as the one conducted by Whitney (2000).
Finally, the California Air Resources Board conducted an extensive program over several years
comparing many different drive cycles. Unfortunately, PM was not measured in this program.
However Figure 2-7 shows the HC emissions compared for the two cycles. The trends indicate
little difference on average between the LA92 and FTP cycles for HC.
Figure 2-7. Hydrocarbon emissions on the LA92 versus corresponding results on the FTP cycle.
10
12
Based on these studies, we conclude that there is little difference in PM emissions between the
LA92 and FTP cycles on an aggregate basis (though their bag by bag emissions may differ). We
shall demonstrate that, for the purposes of ZML analysis, the overall results will be nearly
identical even if we omit the LA92 data, thus minimizing the significance of this issue.
Figure 2-8 shows the new-vehicle emission rates from the studies listed in Table 2-1. The data
points represent each individual test, and the points with error bars represent the average for each
source. The plot presents evidence of an exponential trend (fit included) of decreasing emissions
with increasing model year. The fit is also nearly identical if we omit the two programs that
employed the LA92 cycle. We will use this exponential ZML relationship as the baseline on
153
-------
which to build a deterioration model. However, the measurements from the older programs
primarily measured total particulate matter. These have been converted to PM10 (for the plot),
which is nearly identical (about 97% of total PM is PM10). We also assume that 90% of PM10 is
PM2.5 (EPA, 1981).64 For the older studies, we accounted for sulfur and lead directly if they
were reported in the documentation. In those cases where sulfur was not reported, the levels
were approximated using sulfur emission factors from MOBILE6 and subtracted as an
adjustment.
Unfortunately, many of the older studies used a variety of methods for measuring parti culate
matter. There were many differences in filter media, sampling temperature, sample length,
dilution, dynamometer load/settings etc. It is beyond the scope of this project to normalize all of
the studies to a common PM metric. It is likely that documentation is not sufficient to even
attempt it. Therefore no attempts at adjustment or normalization were made except for size
fraction, lead and sulfur, as described above.
Figure 2-8. Particulate emission rates for new vehicles compiled from 11 independent studies.
40
O)
E
35
30
25
20
15
all
• Gibbset al., 1979
o Cadleet al., 1979
n Urban&Garbe, 1980
A Langetal., 1981
x VW.1991
• GARB, 1986
• Hammerleetal., 1992
a E24-1 Denver
* E24-2UC Riverside
o E24-3 San Antonio
• Chrysler/Ford/GM
o SwRi'NREL
* KC-Summer LA92
+• MSAT-Tier2
• mean for each program
•Exponential Fit
10 15 20
Model Year (+1975)
25
30
To estimate the ZML emission rates from these data, the next step was to separate results for cars
and trucks, and to separate cold-start from hot-running emissions. Unfortunately, the historical
data does not present PM results by cycle phase. Therefore, the 2005 hot-running ZMLs for cars
vs. trucks were calculated from the KC dataset, and the model-year exponential trend from the
154
-------
aggregate trendline (-0.08136) is used to extend the ZMLs back to model year 1975. The base
hot running ZML emission rate for cars (LDV) (£HRO>) is:
F = F
HR,y ^H
,-0.814 y
Equation 2-1
where
y = model year - 1975, and
= hot running ZML rate for MY 2005.
To estimate equivalent rates for trucks, we multiplied this expression by a factor of 1.43. This
value is based on an average of all the studies with new vehicles from 1992 onward (before this
model year, there were no trucks measured). It is also multiplied by 0.898 to give hot running
bag 2 rates and 1.972 to give the cold start emission rate (here defined as bag 1-bag 3 in units of
g/mi). These values were estimated by running a general linear model of bag 2 and bag 1-3 with
respect to composite PM, respectively, using SPSS statistical software. The averages of these
ratios by model year are shown in Figure 2-9, in which no clear trend is discernible. The
parameters of the model are summarized in Table 2-2.
Figure 2-9. Ratios of hot-running/composite and cold-start/composite, Bag2 and Bagl-Bag3,
respectively, averaged by model year.
7
_
'
-------
Table 2-2. Best-fit parameters for cold-start and hot-running ZML emission rates.
Parameter
LDV hot-running ZML (g/mi)
Exponential slope
Truck/car ratio
Bag-2 coefficient
Cold-start coefficient
Value
0.01558
0.08136
1.42600
0.89761
1.97218
Figure 2-10 shows the ZML emission rates. The rates are assumed to level off for model years
before 1975 and again after 2005.
Figure 2-10. Particulate ZML emission rates (g/mi) for cold-start and hot-running emissions, for
LDV and LOT.
0.050
0.045
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
0.000
truck newbag1_3
car newbag1_3
-truck newbag2
-car newbag2
1975
1980
1985
1990
model year
1995
2000
2005
2.2.3 Aging or Deterioration in Emission Rates
In this section, a deterioration model is introduced that captures how new vehicles in all model
years deteriorate over time so that gasoline PM in any given calendar year can be modeled in
MOVES. The purpose of this model is to characterize the PM emissions from the fleet and to
backcast the past as well as forecast the future, as required in inventory models.
156
-------
2.2.3.1 Age Effects or Deterioration Rates
The ZMLs determined in the previous section represent baseline emissions for new vehicles in
each model-year group. By comparing the emissions from the "aged" Kansas City vehicles in
calendar year 2005, to the new rates determined earlier, we can deduce the "age effect" for each
corresponding age. However, simple an approach as this seems, there are many ways to connect
two points. This section describes the procedure and the assumptions made to determine the rate
at which vehicle PM emissions age.
We first break the data into age Groups. We use the MOVES age groups which correspond to the
following age intervals: 0-3 (new), 4-5, 6-7, 8-9, 10-14, 15-19, 20+.
As a first step, the bag measurements from all of the vehicles measured in Kansas City were
adjusted for temperature using the equation derived in the analysis report.43 The equation used is:
Equation 2-2
where EpM,72, is the adjusted rate at 72°F for cold-start or hot-running emissions, EPU,T is the
corresponding measured emissions for cold-start or hot-running, respectively, at temperature T,
respectively.
The temperature-adjusted measurements are the "aged" rates, i.e., the rates in each model-year
group represent emissions for that group at the age of measurement in 2004-05, at 72°F rather
than at the actual ambient temperature.
The method adopted is to ratio the aged rates with the new rates so that the changes with
deterioration rates are all proportional. This approach will be referred to as the "multiplicative
deterioration model," and is analogous to the approach used with the gaseous emissions (Chapter
1).
It is likely that some of the same mechanisms that cause HC and CO to increase over time would
also result in PM increases. These factors include deterioration in the catalyst, fuel control,
airfuel-ratio control, failed oxygen sensors, worn engine parts, oil leaks, etc. Figure 2-11 shows
trends in the natural logarithm of THC rates over approximately 10 years, based on random-
evaluation samples in the Phoenix I/M program. On a log-linear scale, the deterioration trends
appear approximately linear over this time period, suggesting that the deterioration rates are
exponential. This observation, combined with the approximate parallelism of the trends for
successive model years, implies that emissions follow a multiplicative pattern across model-year
or technology groups, calling for a multiplicative deterioration model. In such a model, the aged
rates and the new rates are converted to a logarithmic scale, after which the slopes are estimated
by fitting a general linear model. The average slope is estimated, with the ZMLs determined
earlier defining the j/-axis offsets. The result is a series of ladder-like linear trends in log scale as
shown in
Figure 2-12. The lines fan out exponentially on a linear scale as shown in Figure 2-13. The
dotted lines and the points with uncertainty bars represent the Kansas City data overlaid onto the
model and indicate that the model is consistent with the data.
157
-------
Figure 2-11. The natural logarithm of THC emissions vs. Age for LDV in the Phoenix (AZ)
Inspection and Maintenance program over a ten-year period (1995-2005).
LDV WEIGHTED
InfThC) vs. Age (yeara), LDV
Vehicle age (years)
Figure 2-12. The multiplicative deterioration model applied to PM results from Kansas City. The
y-axis offsets represent ZML rates. The dotted line represents the Kansas-City Data.
4.5
158
-------
Figure 2-13. The multiplicative deterioration model shown on a linear scale. The y-axis offsets
capture the new-vehicle ZML rates. The dotted lines and points with error bars represents the
Kansas-City results (with 95% confidence intervals).
100
19*]
10
15 20
Age (years)
25
30
Because the model is multiplicative, the deterioration factors can be applied directly to trucks,
cold start, hot-running, EC, and OC, since the order of operations does not matter. The start
process requires only a soak-time model to estimate remaining rates for starts other than the cold
start (opmodeID=101-107). Because no data is available describing how particulate start
emissions vary by soak time, we have used the HC soak curves shown previously (see p. 95).
Substantial analysis is yet required to fill modal particulate emission rates for
emissionRateByAge table in the MOVES input database. Because the simple multiplicative
model can be applied across the range of VSP, deteriorated rates by operating mode can be
directly generated, as described in the next section.
2.3 Estimating Elemental Carbon Fractions
After performing the analyses described above to estimate total particulate (PIVh.s), we
partitioned the total into components representing elemental carbon (EC) and non-elemental
carbon (nonECPM), respectively. Following this step, the values for EC and nonECPM were
loaded into the emissionRateByAge table, using the pollutant and process codes shown in Table
1-1. In MOVES2014, we represent 18 PM2.5 species.
We have assigned a new pollutant in MOVES2014 to represent particulate species other than
elemental carbon (EC). This pollutant, designated as "NonECPM," was added to maintain the
distinction between EC and "non-EC" components of PM in the emissionRateByAge table. This
159
-------
assignment (pollutant 112) is a refinement of that used in previous versions of the database, in
which PM2.5 emission rates were assigned into "elemental carbon" and "organic carbon"
components (pollutants 118 and 111). Although organic carbon is the dominant contributor to
nonECPM, quantifying the "nonECPM" component more accurately accounts for the additional
non-carbonaceous species in exhaust particulate. The speciation of PM2.5 is discussed in greater
detail in the located in the TOG and PM Speciation Report.65
Table 2-3. Combinations of pollutants and processes for particulate emissions.
pollutantName1
Primary PNfc.s - Non-elemental
carbon particulate matter
Primary PNfc.s - Elemental Carbon
pollutantID1
118
112
processName2
Running exhaust
Start exhaust
Running exhaust
Start exhaust
processID2
1
2
1
2
polProcessID3
11801
11802
11201
11202
1 as shown in the database table "pollutant." Note that MOVES will reaggregate the particulate components to construct
"Primary Exhaust PMio" (pollutantID 100) and "Primary Exhaust PM2.5" (pollutantID 110).
2 as shown in the database table "emissionProcess."
3 as shown in the database table "emissionRateByAge."
The initial analysis of the EC composition of the light-duty exhaust is reproduced and adapted
from the Kansas-City analysis report.43 Exhaust particulate matter consists of many different
chemical species, including elemental carbon (EC), organic carbon (OC), sulfates, nitrates, trace
metals and elements. The majority of the PM emissions is in the form of EC or OC. Elemental
carbon, also known as soot or black carbon, is produced during combustion when fuel or fuel
droplets are pyrolyzed (or carbonized) under low oxygen levels. In this process hydrogen is
stripped from the carbon atoms in the hydrocarbon, and carbon soot residue remains. Elemental
carbon is formed in gasoline engines primarily when the fuel air mixture is rich (even in
localized portions of the air/fuel mixture of the engine). The hot oxygen-starved and fuel-rich
environment favors pyrolysis reactions.
We might expect to see higher EC fractions in gasoline engines following engine starts, or when
during enrichment mode such as under heavy engine load. These fine soot particles are generally
nonreactive in the atmosphere, though they may act as agglomerization centers for particle
growth both in the exhaust stream and in the atmosphere. In other words, other compounds
including organic carbon adsorb onto the surface of the elemental carbon. In turn, these
adsorbed organic carbon compounds can react in the atmosphere, generally in oxidation
reactions.
160
-------
Organic carbon forms clusters of organic molecules that agglomerate and grow throughout
combustion, as the exhaust cools, and finally as it disperses into the atmosphere. In gasoline
engines OC can be formed normally during combustion from the fuel or the lubricating oil.
Sulfate emissions have largely been controlled through fuel sulfur controls, and, previously, by
the closer control of airfuel ratio necessary for the three-way catalyst to effectively function.
We expect the sulfate emissions to be much lower than past studies.
It is important to separate EC and OC in inventory estimation since photochemical models treat
these fractions of paniculate separately. Also, the ratios are helpful for comparing emissions
(and air quality) models to source apportionment studies. Finally, EC is easier to measure and
more stable in the atmosphere than OC, therefore it is useful to track for a variety of purposes.
In the Kansas City study, EC was measured using two different methods. The first was the
technique of thermal optical reflectance (TOR). This procedure also measured OC and total PM,
but unfortunately, not all the vehicles in the study were measured using this technique.
Elemental carbon was also measured using the photoacoustic analyzer, which measures EC on a
continuous basis. More information can be found on these techniques and their calibration and
comparison results in the contractor's report44 and Fujita et al. (2006).66 The former reference
indicates that the photacoustic analyzer has good correlation with TOR EC measurement
especially at higher PM levels, however, at lower levels (in bag 3 for example), the correlation is
poorer. This is not surprising since all instruments have limited ability to measure small signals.
To accentuate the full range of operation, Figure 2-14 shows a plot of a comparison of the two
instruments on a natural-log scale. The plot reinforces the excellent agreement between the two
instruments in bag 1 of the test, when emissions levels are at their highest. The correlation (and
slope) is also good for the high values in bag 2, however, as the measurements get smaller there
is relatively more variability (in log-space) between the two measurements.
161
-------
Figure 2-14. Comparison of photoacoustic to TOR EC measurements on a logarithmic scale.
» bag 1
• bag 2
A bag 3
— Linear (bag 1)
— Linear (bag 2)
— Linear (bag 3)
y = 0.982'ix- 0.2107
R2 = 0.9417
ln(TOR EC)
We explored the EC/PM fraction for the four measurement techniques employed in the Kansas-
City study: photoacoustic analyzer (PM, continuous EC), Dustrak analyzer (DT, continuous
optical PM), gravimetric filter (PM), and thermal optical reflectance (TOR, which measured both
EC and total carbon, TC). Table 2-4 shows the comparison of the 3 different fractions using
results from these instruments. The values were calculated as fractions of average values in the
numerator and denominator. The TOR fractions have two major limitations: the ratios are
unexpectedly high and, after eliminating bad data points, only 75 valid measurements remain.
Due to the latter condition (primarily), the TOR fractions will not be used in subsequent analysis.
The photoacoustic to dustrak ratios present a reasonable approach, however, since the Dustrak
and PM are not strongly correlated,44 we elected to use the photo-acoustic to gravimetric filter
ratios for EC/PM fraction estimation.
Table 2-4. Elemental to total PM ratio for 4 different measurement techniques.
Instruments
PA/DT
PA/PM
EC/TC
(TOR)
all
0.128
0.197
0.382
start
0.188
0.340
0.540
running
0.105
0.164
0.339
162
-------
In MOVES2014, we updated the EC/PM fractions for light-duty gasoline vehicles to be
consistent with detailed PM2.5 speciation profiles developed for all the measured PM species in
the Kansas City Study.67 The EC/PM fractions are estimated using the photoacoustic analyzer to
filter-based PM emissions.
The speciation analysis confirmed previous analysis that the EC/PM fraction is relatively
consistent across the range temperatures measured in Kansas City study, and across the ranges of
model years in the study. For this reason, no differentiation in the EC/PM fraction is modeled in
relation to temperature or model year of vehicles in MOVES2014.
In developing speciation profiles for light-duty gasoline vehicles from the KCVES,67 we
discovered high concentrations of silicon in the particulate matter samples. Upon further
investigation, we determined that the silicone rubber couplers used in the sampling system
probably contributed to the filter-measured mass. The resulting contamination of filter masses
with silicon substantially impacted the Bag 2 PM2.5 emission rates, which had the highest
exhaust temperatures. No significant contribution of silicon was found in the PM2.5 start
emissions.
The silicon contamination in these measurements resulted in a positive bias in the values for OC.
In consequence, the EC and nonECPM emission rates in MOVES2014 were revised to account
for the updated data analyses used to derive the PM2.5 profile (e.g. VMT-weighted means), and to
compensate for the silicon contamination in the PM2.5 emission rates. Upon removal of the
silicon contamination, the EC/PM fractions are not significantly different between light-duty cars
and trucks. The data from cars and trucks were pooled as documented in the speciation
analysis.67 The updated EC/PM2.5 fractions in MOVES2014 are compared to the previous
fractions in MOVES2010 in Table 2-5. The EC/PM2.5 ratio is constant across all operating
modes for start and running processes.
Table 2-5. Updated EC/PM2.5 fractions by start and running emissions processes.
Vehicle
Type
Car
Truck
Start Emissions
MOVES2010
35.0%
33.0%
MOVES2014
44.4%
Running Emissions
MOVES2010
18.0%
7.0%
MOVES2014
14.0%
2.4 Modal PM Emission Rates
As mentioned earlier, the continuous emissions measurements from the Kansas City study were
examined at great length, after which we determined that the Dustrak gave the most reliable
second-by-second PM time-series data when compared to the quartz-crystal microbalance
(QCM) and the nephelometer. In the following sections, we describe some of the trends in
continuous PM for "typical" normal-emitting and high-emitting vehicles. We conclude by
describing the procedure by which results from the Dustrak were used to develop emission rates
by operating mode.
163
-------
2.4.1 Typical behavior in particulate emissions as measured by the Dustrak and
Photoacoustic Analyzer
After looking at over 500 second-by-second traces, it became apparent that most of the vehicles
fell into certain general patterns. The most common behavior involved highly non-linear PM
emissions increases with increasing engine load. This pattern led to a monolithic "spike" in
emissions during the most aggressive acceleration event in the LA92 drive cycle during the 2nd
(hot-running) bag, at around 900 seconds. This peak is captured in Figure 2-15, which includes
two plots. The higher emissions prior to 300 seconds can be attributed to cold start, during
which the engine is still cold and the fuel:air mixture tends to be rich. The plot on the bottom
confirms this supposition since it indicates that elemental carbon is relatively high during the
start. The hydrocarbons are overlaid on the bottom plot merely for comparison, and provide a
loose and qualitative comparison to organic PM emissions. Some vehicles had variations on this
spike where it was much larger than even the cold start emissions, but this pattern is more typical
of the newer vehicles tested on the warmer days.
On the following series of plots the dustrak (most prominent), nephelometer and QCM are
overlaid on the top chart, while the photoacoustic analyzer, hydrocarbon and speed are overlaid
on the bottom chart. Ordinate values are all relative and not absolute. "Shifted" means time-
aligned, "Temp" means ambient temperature and the filter measurements as well as vehicle type
and model year are written above the figures.
Figure 2-15. A typical time-series plot of continuous particulate emissions as measured by several
instruments.
Xlo' Test[B4714] Model [STATION WAGON] MY [1994] Bag1 PM |B2.09 mg/m] Bag2 PM [6.69 mg/m] Temp [51.5 F]
i 3 -
600 BOO
Time, seconds
QCM (raw) DustTrak (shifted, normalized)'
x10
L
DataRAM (shifted, normalized) PA (g, shifted)
400
1000
600 BOO
Time, seconds
" PA (g, shifted) Speed (mph) —HC (g. normalized to bag 2 PA)
164
-------
The next series of two figures shows how in some cases, the cold-start emissions appear to be
persist into the "hot-running" phase of the cycle (bag 2). Figure 2-16 shows an older vehicle
(MY1976) tested at 54°F, for which one might expect the cold start emissions to have a longer
duration than a newer vehicle. In this case, the cold start emissions seem to end at around 550
seconds (based on the HC trace). However, such cases where large portions of the cold start
emissions "leak" during bag 2 were rare in the dataset, and thus they were not "corrected".
Additional discussion of the impact of cold start on bag 2 emissions is included in a separate
MOVES report.68
Figure 2-16. Continuous particulate emissions from a 1976 Nova measured at 54°F.
Test [B4712] Model [NOVA] MY [1976] Bagl PM [354.43 mg/m] Bag2 PM [13.71 mg/m] Temp [54.3 F]
3
200
400
1000
1200
600 800
Time, seconds
QCM (raw) DustTrak (shifted, normalized) DataRAM (shifted, normalized) PA (g, shifted)
1000
- PA (g, shifted)
BOO
Time, seconds
Speed (mph) HC (g, normalized to bag 2 PA)
1200
Figure 2-17 shows a similar but slightly more commonly seen effect for a newer vehicle. The
difference is that the cold start seems to end at around 250 seconds in bag 1, but then is high
again when bag 2 starts at around 350 seconds. Here the HC is low, but the EC (as indicated by
the PA) is relatively high hinting at a slightly fuel rich mixture. It is uncertain at this time, why
these vehicles would go into enrichment during this relatively mild acceleration.
165
-------
Figure 2-17. Measured Particulate time series for a recent model year vehicle.
x 10 * Test [84828] Model [TRAIL BLAZER] MY [2002] Bagl PM [64.21 mg/m] Bag2 PM [12.17 mg/m] Temp [50 F]
11
xlO
200
400
1000
1200
600 800
Time, seconds
QCM (raw) DustTrak (shifted, normalized) DataRAM (shifted, normalized) PA (g, shifted)
200
1000
600 800
Time, seconds
• PA (g, shifted) Speed (mph)'" HC (g, normalized to bag 2 PA)
1200
1400
The traces shown so far can be considered as "normal emitters" during hot running operation, i.e.
they did not have unusually high emissions during bag 2. These vehicles represent the bulk of
the data. However, some vehicles do exhibit higher or otherwise unusual hot-running PM
emissions.
Figure 2-18) shows a more typical high PM emitter, where the bag 2 emission rate is 266 g/mi.
Here the EC does mirror the high emissions seen in the other instruments. Even the HC
measurements are saturated. This trace, representing a 1978 MG, is an indicator of poor fuel
control, as might be expected with an older (1978) carbureted engine.
166
-------
Figure 2-18. Continuous particulate (and HC) time series for a 1978 MG.
Test [84277] Model [MG] MY [1978] Bagl PM[251.11 mg/m] Bag2 PM [266.15 mg/m] Temp [69.7 F]
, 0.01
k 0.008
200
100
1000
1200
600 800
Time, seconds
QCM (raw) - DustTrak (shifted, normalized) -- DataRAM (shifted, normalized) - PA (g, shifted)
-WO
- PA (g, shifted)
1000 1200
Time, seconds
Speed (mph) HC (g, normalized to bag 2 PA)
At this point it is possible to classify the emission rates into operating modes based on speed,
acceleration and vehicle-specific power (VSP) (Table 1-5). The following two figures show
Dustrak PM emissions binned by VSP and classified by model year Groups. Figure 2-19 shows
this relationship on a linear scale and Figure 2-20 shows the relationship on a logarithmic scale.
It is clear from the latter plot that VSP trends for PM tend to be exponential with VSP load, i.e.
they are approximately linear on a log scale, showing similar patterns to the gaseous emissions,
particularly CO. Thus we assume smooth log-linear relations when calibrating our VSP based
emission rates.
167
-------
Figure 2-19. Particulate emissions, as measured by the Dustrak, averaged by VSP and model-year
group (LINEAR SCALE).
x10
Cars
5 101520253035404550
VSP. kw/tonne
Figure 2-20. Particulate emissions, as measured by the Dustrak, averaged by VSP and model-year
group (LOGARITHMIC SCALE).
-5
-6
-7
| -8
o>
-9
§5-10
-11
-12
-13
Cars
1983
-1989
1996
-2000
-2001
-2004
10 15
20 25 30
VSP. kw/tonne
35 40 45 50
In order to calculate VSP-based modal rates, we followed seven steps:
1. The LA92 equivalent hot-running emission rate (g/mi) is calculated for each age group
within each model-year group, using the deterioration model described in section 2.2.
168
-------
2. Continuous emission rates (g/sec) are calculated from the dustrak measurements for cars
and trucks. These trends are then extrapolated to higher VSP levels where data is
missing.
3. The VSP operating-mode distribution is calculated for bag 2 of the LA92 drive cycle for
cars and trucks separately - this step is equivalent to determining the number of seconds
in each mode.
4. The set of continuous measurements (Step 2) are then classified into the operating-mode
distribution and summed to give an aggregate emission rate representing Bag 2 of the
LA92.
5. The results from Step 4 are divided by those from Step 1 to calculate a ratio for each
combination of the model-year and age groups. The ratios are used to normalize the
modal emission rates to the aggregate filter measurements.
6. The rates from step 5 are then apportioned into EC and nonEC components to give final
rates for the hot-running process. These rates are stored in the emissionRateByAge table
under polProcessID 11201 and 11801, respectively.
The output from step 3 (operating-mode distribution) for cars and light trucks is shown in Figure
2-21. For operating-mode definitions, see Table 1-5 (Page 19).
Figure 2-21. Operating-mode distribution for cars and light trucks representing the hot-running
phase (Bag 2) of the LA92 cycle.
160
140 -•
1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
VSP Bin
169
-------
The output of step 5 for the ZML (0-3 year age Group) in each model year is shown in Figure
2-22.
Figure 2-22. Particulate emissions for passenger cars (LDV) from Kansas City results, by model
year Group, normalized to filter mass measurements.
ID
14
"to"1
~cn in -
QJ
"oi
c °
g
CO
•- R
E 6
LJJ
n -
-
M.
*
X
.,
• 1
• * . -
n „„--«• ™™S«n"-* !<••*
•1960-1930
1981-1982
1983-1934
<198S
« 1986-1 937
M 988-1 989
-1990
-1991-1993
1994
1995
1996
1997
1998
1999
2000
-2001
2002
•2003
2004
10
15
20 25
VSP bin
30
35
40
45
After the rates were calculated, a quality check was performed to ensure that the aged rates in
any particular mode were not too high. A multiplicative model with exponential factors risks
excessively high emission rates under extreme conditions. For example any rate over 100 g/sec
was considered too high, this would be an extremely high-smoking vehicle. This behavior was
corrected in only two cases bins in operating mode 30, representing values for cars and trucks in
the 1975 model-year Group. In these cases, the value from operating mode 29 was copied into
mode 30.
2.5 Updates to PM2.5 emission rates in MOVES2014
We corrected the PIVb.s light-duty gasoline emission rates in MOVES2014 to account for the
silicon contamination measured in the Kansas City study, using our best available estimates. As
stated earlier, the PIVb.s emission rates in MOVES are based on a meta-analysis of multiple
studies and programs. The Kansas City study was used to estimate deterioration from the
estimated zero-mileage emission rates, to estimate the modal PIVb.s emission rates, and the PM2.5
temperature dependency. In MOVES2014, we maintain the temperature relationship, the relative
deterioration, and the power trends developed for MOVES2010. However, we reduced the
running PM2.5 emission rates across all age groups and operating modes by the values shown in
Table 2-6.
Table 2-6 contains the estimated contribution of silicon to the start (bag 1-bag 3) and the running
(bag 2) PM2.5 emissions measured in Kansas City. The silicone rubber contains silicon, oxygen,
170
-------
carbon, and hydrogen which contribute to the measured particulate and organic carbon mass. We
estimated the contribution of the silicon to the PIVh.s emission rates by using the elemental
silicon emission rates from the set of 102 tests analyzed for elements. Additionally, we estimated
that the silicone rubber contributed particulate mass equal to 4.075 times the measured silicon
emission rates, as documented in the speciation profile analysis by Sonntag et al. (2013).6? We
applied these estimates to average silicon emission rates measured for each model year group,
and for trucks and cars. The trucks have a higher silicon contribution which is expected due to
higher exhaust temperatures and larger exhaust tailpipes which expose more silicone rubber to
the hot exhaust. The updated emission rates in MOVES2014 reflect both the reduction in total
PM from the silicon in Table 2-6 and the revised EC/PM ratios in Table 2-5.
Table 2-6. Reductions to PM2 5 in MOVES2014 compared to MOVES2010b due to silicon
contamination.
Stratum
1
2
3
4
5
6
7
8
Vehicle
Type
Truck
Car
Model group
pre-1981
1981-1990
1991-1995
1996-2005
pre-1981
1981-1990
1991-1995
1996-2005
Start
0%
0%
0%
0%
0%
0%
0%
0%
Running
35.3%
25.3%
34.5%
19.1%
14.6%
3.5%
6.1%
8.5%
2.6 Incorporating Tier-3 Emissions Standards for Particulate Emissions
Under the Tier-3 exhaust emissions standards, finalized in April, 2014, the FTP standard for
particulate emissions was reduced from its level under the Tier-2 standard (10.0 mg/mi) to a new
value of 3.0 mg/mi.
Developing rates to represent particulate emissions from gasoline-fueled vehicles under the Tier-
3 standards involved scaling down rates representing vehicles under the Tier-2 standard to a level
that assumes a reasonable compliance margin with respect to the lower standard. More
specifically, we assumed that average FTP emissions for new light-duty vehicles (age 0-3 years)
would be 1.5 mg/mi in MY 2025, when the new standard was fully phased in. This assumption
is independent of engine and fuel-injection technology. The reduced rates assume that
additional controls would be needed to meet the new standard for vehicles employing gasoline
direct-injection technologies, but not for the declining fraction of vehicles in the market
employing port-fuel-injection. The ZML analysis above shows that new PFI vehicles start at
about this level and thus can virtually meet the new standard without modification.
171
-------
Additionally, as with the gaseous emissions, the regulatory useful life was increased from
120,000 to 150,000 miles. The concomitant assumption of increased durability was expressed
through a reduction in the assumed deterioration rate.
We accomplished these modifications in a series of four steps.
2.6.1 Assigning the Tier-2 Baseline
The first step was to establish an assumed baselines for particulate emissions on the FTP cycle,
before and after Tier-3 control is applied. The starting point was to simulate FTP composites
combining hot-running and start rates for PM in MY 2016, after discounting for "silicon
contamination," as described above. The revised estimates for FTP composites were 2.53 and
2.61 mg/mi for cars and trucks, respectively. Note that these values incorporate both the EC
(pollutant 118) and non-EC (pollutant 112) components of PM.
As the simulated composites for cars and trucks are similar, the value for cars was used to
calculate reduction fractions to represent the phase-in. Under the Tier-3 standard, we assumed
that vehicles would achieve a reduction of 41.6% on average relative to the pre-Tier-3 baseline
level. Applying this reduction to the baseline gives an assumed average of 1.48 mg/mi under the
3.0 mg/mi standard, corresponding to a compliance margin of approximately 50%.
2.6.2 Apply Phase-in Assumptions
The second step was to apply the phase-in assumptions applicable to PM. The phase-in begins in
2017 for cars (LDV) and in 2018 for trucks (LOT), and ends in 2021 for both cars and trucks.
Fractions of new vehicles meeting the new standard during the phase-in are shown in Table 2-7.
The table also shows simulated FTP composites (for cars) during the phase-in. These projections
were simply calculated as averages of the Tier-2 and Tier-3 baselines (2.53 and 1.48 mg/mi),
with the phase-in fractions used as weights. Finally, the table shows fractional reductions
relative to the Tier-2 baseline.
Table 2-7. Phase-in Fractions and simulated FTP composites projected for the introduction of the
Tier-3 exhaust particulate-matter standard.
Model Year
2016
2017
2018
2019
2020
2021
Fraction meeting Tier-3 Standard
Cars (LDV)
0.0
0.10
0.20
0.40
0.70
1.00
Trucks (LOT)
0.0
0.0
0.20
0.40
0.70
1.00
Simulated FTP
Composite (mg/mi)
2.53
2.42
2.32
2.11
1.79
1.48
Reduction
^ed, %)
0.0
4.2
8.3
16.6
29.1
41.6
2.6.3 Apply Scaling Fractions
The third step was to apply the fractions to the emission rates for running and start emissions in
the EC and non-EC pollutant processes (11201, 11202, 11801, 11802). The fractions were
applied uniformly to rates in all operating modes, for both cars and trucks.
172
-------
Figure 2-23 shows an example of scaling, for a subset of non-elemental-carbon (nonECPM,
11801) rates for three model years, 2016, 2019 and 2021. Model year 2016 represents Tier 2
standards prior to the onset of the phase-in, 2021 shows fully phased-in Tier 3 standards, and
2019 shows an intermediate year during the phase-in period. In (a), the rates are shown on a
linear scale to show the steepness and non-linearity of the trends against power, whereas in (b),
rates are shown on a logarithmic scale to make clear that the multiplicative scaling is uniform
across the power range. Although not pictured, note that rates for elemental-carbon (ECPM,
11201) show an identical scaling pattern. Note also, that for convenience, emissions in the plot
are presented in mg/hr, whereas rates in the emissionRateByAge table are provided in g/hr.
The uniformity of the multiplicative scaling is also clear if the rates for a single model year are
viewed against age for a set of operating modes, as shown in Figure 2-24. The plot shows rates
for six modes for running operation, including idle (mode 1), with the remaining five modes
spanning a range from low to moderate power. As previously described in 2.2.3 (page 156), the
deterioration trends are exponential (or log-linear).
173
-------
Figure 2-23. Non-elemental-carbon (nonECPM) running rates for cars vs. vehicle-specific power
for three model years on (a) linear, and (b) logarithmic scales (NOTE: rates are presented for
operating Modes 21-30, with each mode represented by VSP at its respective midpoint).
"^
DO
«
t£.
Si mn
CO
c
E
—
s
•-20 16
•-20 19
*-2021
(a)
Linea
-^T r- -7 -I '
r Seal
^
e
/
//
y
/
/
.X
f
M
/
/*
-5 0 5 10 15 20 25 30 3
Vehicle Specific Power (kW/Mg)
10,000
1,000
£ wo
re
CO
c
re
10 =
(b) Logarithmic Scale
5 10 15 20 25
Vehicle Specific Power (kW/Mg)
35
174
-------
Figure 2-24. Non-elemental-carbon rates for trucks vs. Age for selected running operating modes
in model year 2016, presented on (a) linear and (b) logarithmic scales.
450.00
400.00
(a) Linear Scale
10 15
Vehicle Age (Years)
20
25
10,000.00
M 1,000.00
cu
1/1
DO
C
100.00
10.00
-1
13
-21
24
•27
(b) Logarithmic Scale
10 15
Vehicle Age (Years)
20
25
2.6.4 Simulate the Extended Useful Life
The fourth and final step was to reduce deterioration for vehicles under T3, relative to those for
Tier 2. As with the gaseous emissions (1.5.5, page 131), the deterioration trends were scaled
down such that the fleet is 1.25 times as old when a given emissions level is reached under the
extended useful life as under the original useful life. The value of the fraction, 1.25, was
calculated as 150,000 mi/120,000 mi, or 15/12.
The reduction in the deterioration trend is illustrated in Figure 2-25, which shows age trends for
cold-start non-elemental-carbon before and during the phase-in period. The upper pane (a)
shows the moderation of the exponential trend, whereas the lower pane (b) shows the reduction
175
-------
in the logarithmic slope starting in model year 2017. As before, these rates are presented in
mg/start, as opposed to g/start in the database table. Note again that a similar chart for elemental
carbon would show an identical pattern.
Figure 2-25. Elemental-carbon rates for cars vs. Age for cold-start emissions in six model years,
presented on (a) linear, and (b) logarithmic scales.
10 15
Vehicle Age (Years)
100.00
OJ
re
10.00
re
to
c
re
-2016 (b) Logarithmic Scale
-2017
-2018
-2019
-2020
10 15
Vehicle Age (Years)
176
-------
2.7 Incorporating the LEV-III Standard for Particulate Matter
The Tier-3 and LEV-III standards are harmonized with respect the light-duty standard for
paniculate matter through MY 2024, at which point, a 3.0 mg/mi FTP standard will be fully
phased in. However, between 2025 and 2028, the LEV-III program goes further, enacting a
further reduction to a 1.0 mg/mi FTP standard. This reduction is incorporated into a segment of
the emissionRateByAge table applicable to the CA/SI77 states.
The assumptions used to express the transition from rates at the 3.0 mg/mi level and the 1.0
mg/mi level are shown in Table 2-8. We assume a linear phase-in over three years. The
calculations assume a 50% compliance margin with respect to the 3.0 mg/mi standard in MY
2024, transit!oning to a 20% compliance margin in MY 2028.
These assumptions were applying to default Federal rates in the model years listed, by applying
the reduction fractions shown in the right-most column. These fractions were applied uniformly
to start and running emissions of EC- and non-EC PM, for cars and trucks, across all operating
modes.
The table segment including these rates is available at the same location as the corresponding
C A/SI 77 rates for the gaseous emissions (http://www.epa. gov/otaq/model s/moves/tool s .htm).
See 1.6, starting on page 133.
Table 2-8. Phase-in assumptions and reduction fractions used to represent a transition to the 1.0
mg/mi PM standard under LEV-III.
Model Year
2024
2025
2026
2027
2028
Phase-in Fraction
At 3.0 mg/mi
1.00
0.75
0.50
0.25
0.00
At 1.0 mg/mi
0.00
0.25
0.50
0.75
1.00
FTP Composite
(mg/mi)
1.50
1.33
1.15
0.98
0.80
Reduction Fraction1
1.000
0.883
0.767
0.650
0.533
1 Applied to default rates in listed model years.
2.8 Conclusions
The previous discussion describes analyses of particulate-matter emissions designed to develop
modal emission rates for use in the MOVES emissionRateByAge table, discounting the effect of
temperature, and incorporating the effects of model year, age, vehicle-specific power. Rates
were estimated for two components of PM, designated as "elemental carbon," (EC) and "non-
EC" (nonECPM). Speciation of the non-EC component is discussed in detail in a separate
report.65
The supporting analyses are crucial for understanding how PM emissions have declined over
time with the introduction of new technologies, and how rates for new vehicles are projected to
deteriorate over time.
177
-------
Zero-mile level (ZML) rates for older technologies were estimated by analyzing results from
historical studies that measured PM emissions for these vehicles when new. The trends indicate
that emissions have been decreasing exponentially with model year as the engine and fuel
controls have improved and after-treatment devices have been installed. Naturally, ZML rates
for trucks are higher than those for cars.
The effect of age on mean emissions (deterioration) was estimated by comparing the new vehicle
rates to the results from the Kansas City Study. Based on patterns observed for the gaseous
emissions, we have assumed that emissions deteriorate exponentially with the age of the vehicle.
We also found that PM emission increase exponentially with VSP (or road or engine load).
For MOVES2014, we have updated the PM2.5 rates to be consistent with the increased speciation
capabilities of MOVES. The PM2.5 emission rates, and EC/PM ratios have been updated to be
consistent with the light-duty speciation profiles developed from the Kansas City study.
Additional work will be needed to capture the impact of changing light-duty gasoline
technologies on PM emission rates and composition.
178
-------
Gaseous and Particulate Emissions from Light-Duty Diesel
Vehicles (THC, CO, NO^, PM)
In MOVES, emission rates are calculated for each operating mode. However, for the diesel-
fueled passenger cars (LDV) and light-duty trucks (LDT), we lack the necessary continuous or
"second-by-second" measurements to directly calculate emission rates for running emissions in
relation to vehicle-specific power.
Upon additional review, we concluded that the diesel rates developed for draft MOVES and
retained in MOVES2010 were not plausible in relation to corresponding rates for gasoline
vehicles. We concluded that these rates were not adequate to retain in MOVES2014. However,
we also did not consider it a tenable option to release MOVES2014 without rates representing
diesel vehicles.
Consequently, we decided to allow rates for light-duty gasoline vehicles to represent those for
light-duty diesel vehicles. While not an exact parallel and not desirable from a technical
standpoint, we considered it an acceptable solution, as vehicles running on both fuels would be
certified to similar standards. Also, as there are very few light duty diesel vehicles in the U.S.
fleet, their contribution to the inventory is very small.
However, in contrast to the gasoline rates, we did not incorporate a difference in the base rates
attributable to Inspection and Maintenance. That is to say, values for meanBaseRate (non-I/M
condition) were substituted for both the meanBaseRate and meanBaseRatelM.
The level of detail for the rate substitution is shown in Table 3-1.
Table 3-1. Level of detail for substitution of light-duty gasoline Rates onto light-duty diesel rates.
Parameter
Pollutant
Process
Regulatory Class
Model-year Group
Data Source
Description
THC
CO
NO,
EC-PM
Non-EC-PM
Running Exhaust
Start Exhaust
Passenger Car (LDV)
Light Truck (LDT)
All
Replicated from corresponding
Rates for light-duty gasoline
Identifier
1
2
3
112
118
1
2
20
30
1960-2031
4910
179
-------
4 Crankcase Emissions
4.1 Background
In an internal combustion engine, the crankcase is the housing for the crankshaft. The enclosure
forms the largest cavity in the engine and is located below the cylinder block. During normal
operation, a small amount of unburned fuel and exhaust gases escape around the piston rings and
enter the crankcase, and are referred to as "blow-by." These unburned gases are a potential
source of vehicle emissions.
To alleviate this source of emissions, the Positive Crankcase Ventilation (PCV) system was
designed as a calibrated air leak, whereby the engine contains its crankcase combustion gases.
Instead of the gases venting to the atmosphere, they are fed back into the intake manifold where
they reenter the combustion chamber as part of a fresh charge of air and fuel. A working PCV
valve should prevent virtually all crankcase emissions from escaping to the atmosphere.
PCV valve systems have been mandated in all gasoline vehicles, since model year 1969.
4.2 Modeling Crankcase Emissions in MOVES
Crankcase emissions are calculated by chaining a crankcase emissions ratio to the calculators for
start, running, and extended-idle processes. Crankcase emissions are calculated as a fraction of
tailpipe exhaust emissions, which are equivalent to engine-out emissions for pre-1969 vehicles.
Crankcase emissions are calculated for selected pollutants, including THC, CO, and NOx, and the
elemental-carbon and non-elemental-carbon paniculate fractions of PIVb.s. For each of these
pollutants, ratios are stored in the CrankcaseEmissionRatio table.
For vehicles with working PCV valves, we assume that emissions are zero. Based on EPA
tampering surveys, MOVES assumes a failure rate of 4% for PCV valves.69 Consequently, for
fuelType/model-year combinations equipped with PCV valves, we assume a crankcase ratio of
0.04; i.e., emission fractions for the crankcase process are estimated as 4% of the emission
fractions assumed for uncontrolled emissions. While this 4% estimate may be pessimistic for
new vehicles, and optimistic for old vehicles, available data does not support a more detailed
estimate. As older vehicles have higher overall emissions due to deterioration effects, use of the
aggregate rates may understate the impacts of crankcase emissions.
4.3 Light-duty Gasoline Crankcase Emissions
Very little information is available on crankcase emissions, especially those for gasoline
vehicles. A literature review was conducted to identify available data sources for emission
fractions for gasoline vehicles (Table 4-1).
Table 4-1. Selected Sources of published data on hydrocarbon crankcase emissions from gasoline
vehicles.
Authors
Heinen and Bennett70
Bowditch71
Montalvo and Hare72
Year
1960
1968
1985
Fuel
Gasoline
Gasoline
Gasoline
No.
Vehicles
5
9
Estimate
33
70
1.21-1.92
Units
% of exhaust
% of exhaust
g/mi
180
-------
Based on these sources, we estimated emission fractions for model years without mandated PCV
valves. In absence of better information, gasoline emission fractions are a reflection of diesel
research, with the exception of the gasoline HC ratio. Given that the diesel vehicles studied are
largely heavy duty, and that most gasoline vehicles are light duty, there is a potential mismatch
between the data sources, which is unavoidable due to the paucity of data. As noted previously,
model years with PCV valves were assigned emission fractions calculated as 4% of the fractions
shown in Table 4-2.
Table 4-2 Emission fractions for vehicles without PCV systems (percent of exhaust emissions).
Pollutant
HC
CO
NOX
PM (all species)
Gasoline
(uncontrolled,
pre-1969)
0.33
0.013
0.001
0.20
Gasoline (1969 and
later)
0.013
0.00052
0.00004
0.008
The crankcase emission fractions for HC, CO and NO* may underestimate emissions. These
percentages of exhaust emissions are generally based on engine- out, uncontrolled exhaust,
which is not estimated by MOVES. MOVES produces exhaust estimates based on a number of
control technologies (such as catalytic converters). Uncontrolled exhaust in the 1970s was
considerably higher than current tailpipe exhaust.
4.4 Light-duty Diesel Crankcase Emissions
After 2001, all light-duty vehicles, including diesels, are required to avoid venting crankcase
emissions into the atmosphere.73 This requirement differs from turbocharged and supercharged
heavy-duty diesel engines, which are allowed to vent crankcase emissions, as long as the
crankcase emissions are included in the certification tests. As such, we modeled crankcase
emissions from light-duty diesel emissions with two model-year groups, pre-2001, and post-
2001. The values used for the pre-2001 are the same as the pre-2007 heavy-duty diesel fractions.
For 2001 and later, we multiply the pre-2007 by 4% (our assumed PCV failure rate). These
crankcase emission ratios are located in Table 4-3.
181
-------
Table 4-3. Light-duty diesel crankcase emission fractions (% of exhaust emissions).
Pollutant
HC
CO
NO,
PM2 5 (all species)
Light-duty diesel
1960-2000)
0.037
0.013
0.001
0.2
Light-duty diesel
(2001-2050)
0.00148
0.00052
0.00004
0.008
182
-------
5 Nitrogen Oxide Composition
Nitrogen oxides (NOX) are defined as NO + NO2. In MOVES, NO* includes NO, NO2, and a
small amount of HONO. The rationale for including HONO in NO* emissions is discussed in the
heavy-duty report.74 Currently, the HONO/NO* ratio is estimated as 0.8% of NOX emissions
based a study that measured concentrations of NO* and HONO from a highway tunnel in
Europe.75 The NO/NO* and NO2/NOx fractions were developed from a report by Sierra
Research.6
5.1 Light-Duty Gasoline Vehicles
The NOxand HONO fractions for light-duty gasoline vehicles are presented in Table 5-1. The
HONO fraction of NO*, was subtracted from the original NO2 fraction, because the HONO likely
interferes with the estimated NO2 fraction when measured with a chemiluminescent analyzer, as
discussed in the heavy-duty report.
Table 5-1. NOX and HONO fractions for light-duty gasoline vehicles.
Model Year
1960-1980
1981-1990
1991-1995
1996-2050
Running
NO
0.975
0.932
0.954
0.836
NO2
0.017
0.06
0.038
0.156
HONO
0.008
0.008
0.008
0.008
Start
NO
0.975
0.961
0.987
0.951
NO2
0.017
0.031
0.005
0.041
HONO
0.008
0.008
0.008
0.008
5.2 Motorcycles
The NO/NO2 fractions for motorcycles were also developed by Sierra Research.6 The values are
based on measurements on light-duty gasoline vehicles, but apply to different model year groups,
to correspond to similar exhaust emission control technologies. The NO2 fractions reported by
Sierra Research were adjusted to account for the HONO measurements. Development of the
x, CO, HC, and PM, emission rates for motorcycles, is documented in the same report.6
Table 5-2. NOX and HONO fractions for motorcycles.
Model Year
1960-1980
1981-2000
2001-2005
2006-2009
2010-2050
Running
NO
0.975
0.932
0.939
0.947
0.954
NO2
0.017
0.06
0.053
0.045
0.038
HONO
0.008
0.008
0.008
0.008
0.008
Start
NO
0.975
0.961
0.97
0.978
0.987
NO2
0.017
0.031
0.022
0.014
0.005
HONO
0.008
0.008
0.008
0.008
0.008
183
-------
5.3 Light-duty Diesel Vehicles
The NOx and HONO fractions for light-duty diesel vehicles are the same as those for for heavy-
duty diesel. Discussion of the heavy-duty diesel fractions is presented in the corresponding
report.74 These values are presented in Table 5-3 for completeness.
Table 5-3. NOX and HONO fractions for Light-duty Vehicles.
Model Year
1960-2006
2007-2009
2010-2050
NO
0.935
0.764
0.594
NO2
0.057
0.228
0.398
HONO
0.008
0.008
0.008
184
-------
Appendix A: Peer-Review Comments for MOVES2014
Sections of this report that were submitted to peer-review for MOVES2010 were not submitted
for review a second time for release of MOVES2014. In addition, revisions to emission rates for
MOVES2014 that involved straightforward extensions of methods used to develop rates for
MOVES2010 were also not peer-reviewed again. This statement is applicable to the following
report sections:
1. Updates to NLEV and Tier-2 light-duty gasoline rates for gaseous emissions,
2. Development of rates for CA/Section 177 states,
3. Revising the light-duty diesel emission rates to be the same as gasoline emission rates
until we have further information,
4. Updates to the light-duty diesel crankcase emission rates, adopted in view of updates to
the heavy-duty-diesel crankcase emission rates. However, the light-duty diesel crankcase
emissions rates account for the differences in regulations on crankcase emissions between
heavy-duty and light-duty diesel vehicles.
The revisions to light-duty emission rates used to estimate the impact of the Tier 3 Vehicle
Emission and Fuel Standards were incorporated into the version of MOVES2010 used in the
rulemaking analyses. These revisions were documented in the Regulatory Impact Analysis,76 as
well as in a memorandum to the rulemaking docket,77 and were subject to public comment
during the rulemaking process.
In addition, since the fmalization of the Tier-3 standards, the light-duty emission rates for
paniculate matter were updated to account for silicon contamination detected when developing
speciation profiles from the results of the Kansas City study.65 The peer-reviewers charged with
reviewing the MOVES2014 TOG and PM Speciation Report, were also charged with peer-
reviewing the related updates to the PIVh.s light-duty gasoline emission rates documented in this
report, comprising:
Section 2.3 Estimating Elemental Carbon Fractions
Section 2.5 Updates to PM2.5 emission rates in MOVES2014
Comments from the peer-review of the MOVES2014 PIVh.s light-duty gasoline emission rates
are included below. Related comments regarding PIVh.s speciation and emission rates (including
Elemental Carbon, and silicon correction rates) are included in the peer-review comments of the
TOG and PM Speciation Report.
Comments from: Dr. Tom Durbin (University of California at Riverside):
Section 2.3 Estimating Elemental Carbon Fractions
The photoacoustic instrument should provide relatively good measurements for EC over a range
of concentrations. The 2.4 mg/mi differences between the TOR and the photoacoustic seems a bit
high. How [do] these two measurements compare to the total PM mass on the filter would be a
good question to address here. Also, how high are the PM mass emission rates, where the 2.4
185
-------
mg/mi offset would be small considered to be a small fraction of [PM]. Seems like 2.4 mg/mi
would be a big number in comparison to emission rates of typical modern vehicles.
Response: We have removed the discussion regarding the 2.4 mg/mile from the text. We
agree that that discussion was potentially misleading, and a 2.4 mg/mi le would be a very
large offset in terms of emissions from modern vehicles. As presented in the supporting
information ofSonntag et al. (2013), the offset (or intercept from a linear fit) between the
two measures is typically much less than 0.5 mg/milefor vehicles classified according to
technology strata and the seasons of the Kansas City Study.
Re: Comparisons of EC andBC to PM. As shown in Table 2-5, the BC represents less
than 50% of the filter-measured PM mass emissions. We agree that comparing the
variability between the BC and EC to the filter-measured mass is of interest, and Figure
S6 in Sonntag et al. (2013) evaluates the BC/PMandEC/PMratio among the different
vehicle strata and seasons measured in the Kansas City Study. However, that information
is outside the scope of the documentation of the EC/PM emission rates used in MOVES.
Section 2.5 Updates to PIVb.s emission rates in MOVES2014
The issue of silicon contamination is probably something that needs further consideration. I think
that some rational should be given in this description as to where the 4.075 factor comes from. In
fact, I looked through the referenced ES&T paper and did not find anything either, unless there
was a error with the reference numbering. This issue further emphasizes points raised above
[located in the Speciation report comments65] that EPA probably is using too narrow a focus in
the data sets that it considers.
RESPONSE:
We have made sure that the report correctly references Sonntag et al., 2013,6? which includes
the derivation of the 4.075 factor used in the silicon correction.
We agree with reviewer regarding the importance of incorporating additional studies in either
the derivation of the emission rates, or as a source of validating results from a limited number of
studies. However, for the time-frame ofMOVES2014, we were unable to incorporate newer
studies into the data analysis of the light-duty gasoline PM emission rates.
186
-------
Appendix B: Peer-Review Comments and Response for MOVES2010:
Reviewer 1
Reviewer 1: John M. German, International Council on Clean Transportation, Washington,
D.C.
Mr. German is a Senior Fellow and Program Director with the International Council on Clean
Transportation. He is a highly qualified expert in the areas of automotive engineering and
emissions control, whose career includes experience with both the industry and the USEPA. His
experience with the EPA includes managing the development of the US06 and SC03 test cycles
used to implement the Supplemental Federal Test Procedure, oversight of the development of the
Comprehensive Modal Emissions Model, a project conducted by engineers at the University of
California at Riverside, management of the cold-temperature CO Rule, and development of
facility cycles used in the MOBILE6 model. His industry experience includes power-train
engineering at both Chrysler and Honda over a period of 19 years.
This document contains comments received from Mr. German following conclusion of his
review of the draft report. Following each comment, I have included our specific response,
describing whether we have accepted the comment and made corresponding revisions in the final
report, or whether we have offered a rebuttal or otherwise declined to make revisions.
Note that page and paragraph numbers listed in the comments refer to the draft document: Development of
Emission Rates for Light-Duty Vehicles in the Motor Vehicle Emissions Simulator (MOVES2009): Draft
Report. A copy of this document is included in the peer-review records, and is also available at
http://www3.epa.gov/otaq/models/moves/techdocs/420p09002.pdf.
187
-------
Comment Summary
In general, the MOVES draft did an excellent job of assessing emissions. This is a very difficult
task, especially considering the lack of data in many areas. Of course, given the lack of data and
the multitude of assumptions that have to be made, there were a number of places where
different approaches may yield better results, as discussed in my comments.
My detailed comments were written as annotations on the draft report itself. I did not have the
time or facilities to print out all 124 pages of the draft report with my annotated comments.
Thus, the detailed comments are submitted only in electronic form. It is perfectly OK if EPA
wants to print out these detailed comments.
PEER REVIEW CHARGE QUESTIONS:
1. For the most part, the report provides adequate description of selected data sources to allow
the reader to form a general view of the quantity and representativeness of data used in
development of emission rates. In some cases information about the sources of data used in
the report were lacking. My detailed comments note these places and asks for additional
description of the data.
2. The description of the analytic methods and procedures was generally excellent and allowed
the reader to understand the steps taken and assumptions made. There were a few cases
where additional explanation would be helpful, as noted in my detailed comments. The
examples chosen for tables and figures were also generally excellent.
3. Most of the methods and procedures employed were technically appropriate and reasonable.
However, there were some areas where alternative approaches might better achieve the goal
of developing accurate and representative model inputs. These are listed in my detailed
comments. The major areas of concern are as follows (these duplicate the summary at the
beginning of my detailed comments):
a. I did not see anything on ambient temperature adjustments. This is a very large
factor.
b. The use of EVI data for pre 2000 vehicles needs to be validated. If I recall correctly,
there are offsets between EVI data and the FTP and the correlation is not all that good.
While the EVI data is needed to determine deterioration rates over time, it may not be
a good idea to use it directly for baseline emissions.
c. Diesels used yet another source of data - FTP data instead of IUVP or EVI data.
Should establish a correlation between FTP and IUVP data and apply this as an offset
to the FTP data.
d. The use of bag 2 for running emissions is not appropriate. Running emissions plus
start emissions should equal the FTP. Using only bag 2 completely ignores the
running emissions from the 505 (bag 1/3). Running emissions should be determined
by subtracting start emissions from total FTP emissions and dividing by 7.5 miles.
188
-------
4. In most cases, EPA's assumptions when applicable data is meager or unavailable were
reasonable. My detailed comments note areas where different assumptions might be better.
The major area of concern is that the modeling of PM deterioration implicitly assumed that
PM correlated with HC (this is also in the summary at the beginning of my detailed
comments). However, CO is a much better predictor of air/fuel ratio than HC (reasons
explained in my detailed comments on page 88). I would investigate how well your
individual PM test results correlate with CO, or with a combination of HC and CO, instead of
just assuming they correlate with HC. If a reasonable correlation can be established, this
would be a much better way to establish PM cold start and running emissions and to assess
PM deterioration.
5. In general, the model inputs were appropriate and are reasonably consistent with physical and
chemical processes involved in exhaust emissions formation and control. Cases where better
inputs could be used are listed in my detailed comments.
My most important recommendation is outside the scope of reviewing the draft report. EPA
desperately needs to have better data upon which to base the MOVES model. Collection of
consistent data across the variety of vehicles and operation conditions would allow creation of a
much better model and help avoid all the assumptions needed in the current version.
189
-------
Overall comments.
1. I did not see anything on ambient temperature adjustments. This is a very large factor.
RESPONSE:
The base rates described in this report represent the temperature range of 68-86 °F, i.e., the
"FTP temperature range ". They are not designed to represent the effect of temperature.
The revised report mentions this fact and refers readers to the appropriate report describing
adjustments for temperature (and other factors).
2. The use of IM data needs to be validated. If I recall correctly, there are offsets between
EVI data and FTP and the correlation is not all that good. While the EVI data is needed to
determine deterioration rates over time, it may not be a good idea to use it directly for
baseline emissions.
RESPONSE:
The I/M data used were not used as cycle aggregates. Rather, second-by-second data were
used, after being classified into operating modes on basis of vehicle-specific power and
speed, as described in 1.3. Breaking down the cycle in this way neutralizes the differences
that would be expected had we used cycle aggregate values, as in MOBILE6.
The potential offset is made worse by the use of FTP data for diesels. Need to establish a correlation
between FTP and IUVP data and apply this as an offset to the FTP data.
RESPONSE:
As with the data for gasoline vehicles, the FTP data for diesels was not used directly, as in
MOBILE, but rather to develop scaling factors that were applied to modal emission rates for
light-heavy-duty diesels so as to represent light-duty-dies els.
3. The use of bag 2 for running emissions is not appropriate. Running emissions plus start
emissions should equal the FTP. Using only bag 2 completely ignores the running
emissions from the 505 (bag 1/3). Running emissions should be set by subtracting start
emissions from total FTP emissions and dividing by 7.5 miles.
190
-------
RESPONSE:
We did not use FTP bag 2 emissions directly to assign emission rates for running operation.
Rather, we used them to derive relative changes in hot-running emissions for vehicles in
different standard levels relative to Tier 1. These relative changes, or ratios, were used to
scale down Tier 1 modal emissions appropriately to represent vehicles certified to NLEV
and Tier 2 standards. We usedBag-2 emissions for this purpose because we are confident
that the engine is conditioned before Bag 2 commences, and because we lack a way to
readily separate start and running emissions in Bags 1 and 3, i.e., none of the data available
to use included Bag 1 run under hot stabilized conditions ("hot-running505 "). In the
revised report, this process is described in 1.3.4.2.
We explicitly avoided the use of FTP composites, which include both start as well as running
emissions. While the FTP standard represents the effects of control of both start and running
emissions, the relative levels of control for start and running differ both from each other and
from that for the composite. Generally, start emissions decline less than the standard would
suggest, and running emissions decline more. We elected to treat start and running
separately to account for these differences in levels of control. Thus, to represent start
emissions, we followed the common practice of estimating the cold start as the difference
between Bagsl and 3. We selected Bag 2 to represent hot-running emissions because unlike
Bags 1 and 3, it does not contain a "start increment". We did not use Bags 1 or 3 to
represent running because it is not possible to isolate the "running component" from the
"start increment" in these bags, except for the cold start, as described.
4. Modeling of PM deterioration implicitly assumed that PM correlated with HC. However,
CO is a much better predictor of air/fuel ratio than HC (reasons explained in my comments
on page 88). I would investigate how well your individual PM test results correlate with
CO, or with a combination of HC and CO, instead of just assuming they correlate with HC.
If a reasonable correlation can be established, this would be a much better way to establish
PM cold start and running emissions and to assess PM deterioration.
RESPONSE:
We agree that PM, as well as CO, responds to enrichment. More generally, though, a major
component ofPMis HC consisting of higher molecular weight compounds (about C10-12)
including semi-volatiles and non-volatile compounds, both of which are emitted in the
par ticulate phase, with parti culate being formed from unburned and partially burned fuel
components. In the cylinder, processes such as wall quenching, particularly during cold
starts, tends to create both HC and par ticulate. In addition, since the introduction of NLEV
or LEV-I standards, targeted reductions in HC have yielded associated reductions in PM,
but have not driven compliance for CO, which the manufacturers have easily achieved.
Taking these factors in combination, we suggest that while a relationship between CO and
PM exists, the corresponding relationship between HC andPMis stronger. According to the
EPA report titled "Analysis of Paniculate Matter Emissions From Light-Duty Gasoline
Vehicles in Kansas City, " PM correlated better with HC than it did with CO on the LA92
drive cycle (page39 ofEPA420-R-08-010). However, it is entirely possible that with a more
aggressive cycle including more opportunities for vehicle enrichment, CO may prove to have
stronger correlations.
191
-------
Chapter 1. Light-Duty Gasoline Criteria Exhaust Emissions (HC/CO/NO*)
1. PageS, 1.5.1 :
Are you considering only IM data? If so, you should state so. If not, should state what the
data sources are and why FTP data was not included.
RESPONSE:
We did not restrict consideration to I/Mdata as such, although we did require that data was
measured on vehicles subject to I/M requirements. For example, data from the New York
Instrumentation Protocol Assessment (NYIPA) received serious attention. These data are
not I/M program data, but were measured on vehicles subject to the I/M program in New
York City between 1998 and 2002. Nor did we exclude the FTP as such. Had datasets
measured on the FTP been available and met all requirements, we would have considered
using them.
2. Page 9, 1.5.1.1.1, Re: Eq 1-2:
How were these equations derived? Should describe or include a reference. Also, are A, B,
and C derived directly from these equations? As A is proportional to v, B to v sq., and C to
v cubed, it seems like these equations are missing a step.
RESPONSE:
We have cited a reference for these equations in the final report. They are taken from the
"IM240 andEvap Technical Guidance", April, 2000, EPA420-R-00-007. Note that the
squared and cubic exponents in the denominators of the equations for B and C lead to
correct proportionality in the resulting units.
3. Page 10, 1.5.1.3, 2nd para:
Table 1-6 only lists about 80,000 vehicles. If the large majority of these "several million"
are remote-sensing or poor data, this statement probably should be revised.
RESPONSE:
The data sources listed in Table 1-5 in the draft report represent data determined to be
available and potentially suitable, before we began examination to verify suitability and
quality. Table 1-6 lists datasets that did received detailed scrutiny, and with the exception of
the St. Louis I/M data, were also confirmed to be suitable. The corresponding Tables in the
final report are Tables 8 and 9.
4. Page 11, 1.5.1.3, Table 1-6:
What about British Columbia, Denver, Indiana, Ohio, and Wisconsin listed in Table 1-5?
Why were they not discussed and/or excluded?
192
-------
RESPONSE:
These data were not considered or discussed due to a combination of quality issues or lack
of time and resources needed to process them. In the final report, we have removed
references to these datasets in Table 8 (formerly Table 1-5).
5. Page 18, 1.5.4.2.2, first para:
To help the reader, might want to mention here that extrapolation to high VSP bins was
modified based on actual high VSP data in section 1.5.5.
RESPONSE:
We have added a sentence in this paragraph to refer the reader to the description of the
adjustments in 1.3.3.5. (formerly 1.5.5).
6. Page 26, Table 1-11:
Emission rates for 81-82 are very different than for 1980 and previous. Probably not
appropriate to use 81-82 to represent older vehicles.
RESPONSE:
While acknowledging the differences between the two groups of vehicles, our difficulty with
the 1980 and older vehicles is that we lacked sufficient data to backcast their emissions to
young ages. Given this difficulty, and the negligible influence of the 1980 and older model-
year group on inventory, we considered the substitution reasonable.
1. Page 26, 1.5.5, 2nd para:
Should provide a reference for this program [NCHRP].
RESPONSE:
In the final report, we have cited a reference for this program: "Development of a
Comprehensive Modal Emissions Model: Final Report", NCHRP Project 25-11, April,
2000.
8. Page 26, 1.5.5, 2nd para:
Should say something about why it [MEC cycle] was developed and by whom.
RESPONSE:
In the final report, we have added a brief description of the context and purposes for
development of the MEC cycles.
9. Page 27, 1st para, bottom of page:
Typo here [incomplete sentence].
193
-------
RESPONSE:
We have corrected the incomplete sentence.
10. 1.5.6, Page 35, 1st para:
Were the vehicle compositions the same for the migrating and local I/M vehicles? Need to
demonstrate that the comparison did not include effects of different mix of vehicle types
(car, light trucks) and sizes within vehicle type. If there are effects, this should be corrected.
This is especially important, considering that the migrating vehicles had lower HC and NOx
than the local vehicles.
RESPONSE:
At the time lacked a means to verify standard levels or vehicle class, without information on
engine family for measured vehicles. Since release of the draft, we have made progress in
this area. It may be possible to revisit the analysis in terms of truck classes for revisions to
be considered during the next 24 months.
11. Section 1.5.6, Page 35, 2nd para:
Can this data be sorted by the same age groups as the new data? This would help the
comparisons.
RESPONSE:
Unfortunately, we are unable to distinguish age groups in these data, which we acquired
from a published source.
12. Section 1.5.6, Page 35, 3rd para:
Were the locations controlled so that they produced similar vehicle speeds and accelerations
at the point of [remote] sensing? Some discussion of why these data are compatible should
be included. Also, were the fleet mixes similar?
RESPONSE:
For the data collected in the Atlanta area, the researchers attempted to select multiple sites
with similar driving characteristics, as described in the 2004 Biennial Evaluation Report.
"remote-sensing sampling sites are selected to ensure physically consistent but demographically
diverse characteristics. Single straight lines of traffic with an average 35 mile-per-hour velocity are
sought to facilitate single vehicle measurements and speeds that maximize measurement
opportunities. Driver behavior and driving maneuvers are also observed at each site to ensure that
remote sensing measurements would not be biased high by acceleration or low by coasting. "
(Reference 10 in the Final Report). In the 2004 comparisons, the areas used were
geographically contiguous, to account for the existence of a new low-sulfur fuel requirement
in the 25-county Atlanta area (13 I/M and 12 non-I/M counties), which suggests broad
similarity in the composition and age of the two fleets. In addition, before generating
aggregate fleet-level results, we compared the I/M and non-I/M remote-sensing data on a
model-year basis, to control for potential differences in fleet age.
194
-------
13. Page 37, Figure 1-13:
It is troubling that the migratory vehicles had lower emissions than the local vehicles in
Phoenix for the 0-4 age group. Need to explore possible biases in the data, such as vehicle
composition, and correct if possible.
RESPONSE:
It is possible that fleet composition could contribute bias. As mentioned above, we lacked a
way of assigning vehicle class and standard level while performing these analyses. Recent
developments in this area may allow to reevaluation of this question in the future. In any
case, the differences shown in the figure are not statistically significant for HC or CO, and
perhaps marginally so for NOX.
14. Page 38, 1st para:
Does this mean that you ignored the data from the 0-4 year age group? If so, should explain
why and if not, should explain how the 0-4 year age data was used.
Also, how were the three data sets combined to determine the value at 7.5 years?
Also, why is the midpoint of the 4-5 MOVES age group 5 years instead of 4.5?
RESPONSE:
We did not combine the three datasets, but rather assigned the ratios on the basis of our
analysis of the Phoenix data. Accordingly, we used the other two datasets for verification.
We did tend to discount the 0-4 yr age group in the case of Phoenix. Because this program
had afour-yr exemption period, it did not make sense to assume that the program was
achieving benefits for vehicles that were exempt from testing. For Phoenix we thus assumed
a ratio of 1.0 during the exemption period. However, the development of the non-I/M
reference rates must allow for the fact that many programs have exemption periods shorter
than those in Phoenix. For this reason we did not think it reasonable to assign no I/M
difference in the MOVES 0-3 year ageGroup, and performed the interpolation to estimate a
difference for this group. In the interpolation, the value at 7.5 years was taken as the value
for the 5-9 year group from the Phoenix analysis.
When the midpoint of the 4-5 year ageGroup was set at 5 rather than 4.5 years is because
this age group spans two full years. Vehicles enter this group when they turn 4, and leave it
when they turn 6. When they turn 5 they have been in the group for one year, and will
remain in it for another year. When they turn 4.5, they have been in the group only 0.5 year,
but will remain in the group for another 1.5 years. On this basis, we concluded that 5,
rather than 4.5, is the effective midpoint of the ageGroup.
15. Page 38, 2nd para:
This implies that the 10+ data was also ignored.
195
-------
RESPONSE:
No. The statement reflects the fact that the ratios in Figure 1-13 show that the ratios in the
10+ age class are very similar to those in the 5-9 year age class, suggesting that the ratio
has stabilized by 10 years of age.
You need to add explanations of how the three data sets were combined and how the data
was translated into the lines in Figure 1-14 and the bars in 1-15.
RESPONSE:
As mentioned above, the datasets were not combined; the ratios were assigned based on the
Phoenix data, as modified by the interpolation, with the remote-sensing data used for
verification.
It is potentially troubling that the non-I/M ratios stabilized at only 6 years of age. I would
expect that emissions from vehicles in I/M areas would stabilize after a certain age due to
being required to be fixed, but I would NOT expect emissions from vehicles in non-I/M
areas to stabilize with age. So, the ratios should continue to increase beyond 6 years.
RESPONSE:
Note that the fact that the ratio stabilizes does not mean that the absolute emissions in the
non-I/M area stabilize at six years, because the absolute emissions in the I/M area continue
to increase until after 10 years. Therefore, the absolute emissions in the non-I/M areas also
continue to increase.
16. Page 40, 3rd para:
I don't understand this statement. The non-I/M % increases appear to be the same for both
modes. Mode 27 simply has much higher emissions whether the vehicle is from an I/M or
non-I/M area. The I/M factor has the same effect on both.
RESPONSE:
In percentage terms the effect is the same, but in absolute terms the effect is greater in
opMode 27 than in 11, simply because the emission rate is correspondingly higher.
17. Page 43, 1st para:
Again, this seems to be reasonable for vehicles in I/M areas, due to requirement to maintain
emissions. But is it also true in non-I/M areas? Do you have data supporting this?
RESPONSE:
We revised assumptions for non-I/M areas between the draft and final releases. The revised
assumptions are described in the final report and in the response to comment #20 below.
196
-------
18. Page 46, 1st para:
For NOx, the average emissions for the 15-19 age group were lower than for the 10-14 age
group for your 1986-89 model year control group. Is it appropriate to lower NOx emissions
for higher age vehicles for model years after 1990? Probably better to assume that they
don't change after 10-14 years.
RESPONSE:
The apparent decline with increasing age can be seen in various datasets. It is uncertain
whether it is a real effect, or due to erratic behavior in sub-samples of decreasing size. We
have thus assumed that rates do not decline after stabilizing.
19. Page 46, Table 1-14:
These ratios don't match the data in graphs 1-15, which shows that the NOx for 15-19 year
old vehicles was lower than 10-14. However, the ratios in Table 1-14 are the reverse.
RESPONSE:
Table 1-14 has been obviated by revisions since release of the draft. Its counterpart in the
final report is Table 17. The approach used to stabilize emissions has been modified, as
described in 1.3.3.7 of the final report.
20. Page 48, 2nd para:
In section 1.5.6, you presented extensive analyses for non-I/M areas based upon Phoenix
I/M data. You need to explain what that data doesn't work for this section.
RESPONSE:
The migrating vehicle sample used to develop the non-I/M reference rates was not sufficient
in itself to allow assessment of age trends in non-I/M areas past about 10 years. For this
reason, it was not useful to inform modeling of emissions stabilization in non-I/M areas.
I think this is a poor assumption. Catalyst problems identified by HC and CO monitors will
also lower NOx emissions. On the other hand, fixing air/fuel ratio problems may well
increase NOx emissions. Many malfunctions are being identified and repaired - and these
repairs have impacts on NOx emissions. This is not similar to a non-I/M area. For I/M areas,
the emission increase was not the same for all pollutants. Why would it be the same for non-I/M
areas?
RESPONSE:
After some consideration, we have come to agree with this comment. We have accordingly
revised the assumptions used to represent trends in emissions for vehicles over 15 years of
age in the non-I/M reference rates. In the final report, the revised assumptions are
described in 1.3.3.7.2. In the revisions, we assume that the relative trend observed between
the 10-14 and 15-19 year ageGroups will persist from the 15-19 year ageGroup to the 20+
ageGroups. Thus, in non-I/M areas, rates continue to rise after ten years, but at lower rates
than before ten years.
197
-------
21. Page 49, 1.5.9, 1st para:
How are start emissions calculated? I don't see any discussion or reference.
RESPONSE:
Discussion of start emissions, especially for vehicles manufactured prior to 1996, was
inadequate in the draft report. In the final document material has been added; the expanded
discussion has been inserted in the new section 1.4.
22. Page 49, 2nd para:
Where? [is the definition of start rates?].
RESPONSE:
In the final report, start emissions are defined in 1.3.4, and in 1.4.1.1.2.
23. page 50, para 1:
What simulated FTPs? All you have talked about is bag 2. Also, what relationship does bag
2 have to start emissions? This doesn't make sense.
RESPONSE:
In this paragraph, the "simulated FTPs" would have been more accurately referred to as
"simulated FTP Bag 2. " The Bag 2s were simulated to describe the relative deterioration
rate for running emissions, as described in [draft] Table 1-16. The relative deterioration
trend for starts was then assessed in relation to the relative deterioration trend for running
emissions, as described in the follow ing paragraphs.
24. page 50, para 2:
Are your "rates" multiplicative or additive?
RESPONSE:
The rates are multiplicative. For light-duty gaseous emissions, deterioration is applied
multiplicatively, and any adjustments or modifications are also multiplicative.
25. page 52, 1.6, 1st para:
How does the FTP data from the IUVP program correlate with the IM data used for pre-
2000 vehicles? If there is an offset between the EVI and FTP data, this will lead to
discontinuities in your assessment.
Not to mention the simple correlation between EVI data and real world data. You need to
validate that EVI second-by-second data correlates with FTP data, or apply an offset factor to
the EVI data.
RESPONSE:
We address this question in our responses to overall comments 2 and 3 above.
198
-------
26. Page 53, 5th para (step 4):
This may be true for vehicles certified before SFTP phase-in. But for vehicles certified to
the SFTP, it may not be a valid assumption.
RESPONSE:
Based on supplementary analyses performed since release of the draft, we believe that the
assumption holds, although uncertainty remains in estimating differing degrees of control.
27. Page 53, 1.6.2.1, 1st para:
This means that you are throwing away running emissions on bags 1 and 3. Would be more
appropriate to subtract cold start emissions from total FTP emissions, then calculated the
running emissions over the entire drive cycle, not just bag 2.
RESPONSE:
While it would be a desirable step, we are unaware of a way to distinguish running from
start emissions in Bags 1 and 3, to allow calculation of running emissions over the total FTP
as you suggest. See our response to overall comment 3 above.
28. Page 53, 1.6.2.1, 2nd para:
It's not HC control, its NOx control. Manufacturers have found that catalysts are most
efficient when the air/fuel control is precisely at stoich, instead of cycling from slightly rich
to slightly lean. The elimination of the slightly rich events has reduced CO, as well as fast
catalyst lightoff.
RESPONSE:
While we don't disagree that prevention of rich events and promotion of fast lightoff would
reduce CO, as well as NOX, we would suggest that as CO and exhaust hydrocarbons are
both products of incomplete hydrocarbon combustion, and that the overall control strategy
is to drive oxidation towards completion, CO control is more fundamentally linked to HC
control than to NOX control.
29. Page 56, 1st para:
Or there are running emissions in bags 1 and 3, which were not evaluated.
RESPONSE:
There are running emissions in Bags 1 and 3, of course. Incorporating them would not
fundamentally change the relationships shown in the figure 1-23, as most of the mass of Bag
1 is attributable to the start increment, with the imputed running component making a
relatively small contribution.
199
-------
30. Page 56, 1.6.2.2, 3rd para:
This is not correct. LDT3/4 had interim standards that had the same phase-in as
LDV/LDT1/LDV2. However, the phase in to the final Tier 2 standards for LDT3 and 4 was
50% in 2008 and 100% in 2009.
RESPONSE:
Since release of the draft model, we have fundamentally revised the phase-in assumptions
based on certification records and sales figures, as described in 1.3.4.2.2 in the final report.
31. Page 58, 1.6.2.3, 1st para:
LDTls are rapidly disappearing; although it probably doesn't matter, as they are becoming
LDT2s which have the same standards.
RESPONSE:
Revised phase-in assumptions project low fractions ofLDTl, on basis of certification
records and sales.
32. Page 64, 1st para:
Your "FTP region" INCLUDES bagl/3 driving, but your running emissions don't. This
mismatch needs to be fixed.
RESPONSE:
We agree that the text at this point is unclear, and not reflective of what the rates represent.
In the final report we redesignate the "FTP region" as the "hot-running-FTP" region, and
the "SFTP region" as the "US06 region." Under this designation, the "hot-running-FTP"
region includes the speed and power ranges covered by the FTP Bag 2 (or the
IM240/IM147). It is not intended to include the somewhat more aggressive driving
represented in Bags 1/3. The revised discussion is in 1.3.3.2.4 in the final report.
33. Page 64, 3rd para:
You can justify this - you don't have to just assume it. The SFTP standards were calibrated
to the Tier 1 and NLEV FTP standards. When the FTP standards were increased in
stringency with Tier 2, the SFTP standards were NOT increased correspondingly. Instead,
the SFTP standards are still calibrated to NLEV levels. Thus, the SFTP standards are not as
stringent and the expected reductions are less.
RESPONSE:
We agree on this point.
34. Page 64, 5th para:
You have SFTP data for 2001-2003? If so, you should discuss the data, similar to how you
discussed the data for vehicles prior to 2000 in section 1.5.5 and Table 1-12. Is the vehicle
composition for 2001-3 similar to 1998-2000?
200
-------
RESPONSE:
In revisions to the rates for MOVES2010, SFTP results from the IUVP program were
applied for the NOX rates, but not for the HC or CO rates. For each pollutant, we adopted
the approach that appeared to most improve verification against external data.
35. Page 65, 2nd para:
If I follow this correctly, this means that SFTP emissions are the same for 2010 vehicles as
for 2005 vehicles. If this is accurate, should state this explicitly.
RESPONSE:
This is correct. In the final report, this point is made clear in Figure 34.
While I agree that the reduction in SFTP emissions should not track FTP reductions, as
SFTP standards were not reduced in conjunction with Tier 2 FTP reductions, it may not be
reasonable to assume that there is no reduction in SFTP emissions for Tier 2 vehicles.
RESPONSE:
In the rates released with MOVES2010, we assume some reduction in Tier 2 vehicles for
NOX, but not for HC or CO. Revisions made to the draft rates are described in 1.3.4.2.4.1.
36. Page 69, Table 1-19:
You have a minor problem here. You define cold start emissions as > 720 min soak period
minus a 10 minute soak period. But you then apply start emissions for a 10 minute soak.
This implies that your cold start emissions are understated and should be corrected by
adding the 10 minute soak start emissions to the emissions from bag 1 minus bag 3.
RESPONSE:
We have neglected the component associated with the 10-min soak in the cold start rates. We
have assumed this was reasonable in that for Tier-1 and later vehicles, we lacked data to
estimate the start for a "0-min " soak. We preferred not to apply the soak curve relationships
(Figure 39) for this purpose as to use the relationships to estimate the cold start and then to
reuse them to estimate warm starts would introduce circular reasoning into the process.
37. Page 69, 2nd para:
Some information on the vehicles used to derive these soak fractions would be very useful. The NOx
start fractions suggests that these are from older vehicles - I would not expect to see higher NOx for an
intermediate start on an NLEV or Tier 2 vehicle.
RESPONSE:
It is true that these relationships were derived for older vehicles, manufactured before 1995.
In the final report we cite the study, released by CARB, from which they were derived:
"Methodology for Calculating and redefining Cold and Hot Start Emissions ", California Air
Resources Board, El Monte, CA, March, 1996 (Reference 22).
201
-------
38. Page 69, 3rd para:
How do you handle ambient temperatures other than 75 degrees? Colder temperatures will
dramatically change the soak fractions as a function of soak time. For example, at 20 F a
soak time of 240 to 360 minutes should result in a complete cold start and even a 30 minute
soak will have a completely cold catalyst.
RESPONSE:
Emissions for temperatures outside the "FTP range " of68-86°Fare estimated through the
application of temperature adjustments. Adjustments are described in a separate report:
"MOVES2010 Highway Vehicle Temperature, Humidity, Air Conditioning, and Inspection
and Maintenance Adjustments " EPA-420-R-10-027.
(http://www.epa. gov/otaq/models/moves/420rl 002 7.pdf). At present, the model does not
have the sophistication to apply interactions of soak time and temperature. Soak
relationships and temperature relationships are assumed to be independent.
39. Page 74, 1.7, 2nd para:
Replicating gasoline data for ethanol and advanced gasoline technologies, including hybrids,
is reasonable. However, it is NOT reasonable to replicate the gasoline data for diesels.
Diesels are inherently lean-burn, which changes both the emissions from new vehicles and
how they deteriorate.
RESPONSE:
Lacking better data on light-duty diesel vehicles under Tier 2, we have retained this
assumption at the present.
Chapter 2 Light-Duty Gasoline Participate Exhaust Emissions
1. Page 84, para 3:
Should note that at colder temperatures, additional enrichment is needed and the enrichment
lasts longer.
RESPONSE:
We have added a sentence to this effect.
TDorra 88 9n t->ot-o •
rage oo, / para.
Not a good assumption. LA92 is a higher load cycle which induces more enrichment. The
additional enrichment should be expected to cause additional PM emissions, so there might
not be any deterioration effects.
The effects of enrichment can be analyzed using CO emissions. There is a direct
relationship between enrichment and CO from the engine - virtually all of the excess fuel is
emitted as CO, not HC. In fact, air/fuel ratio can be calculated from engine-out CO
202
-------
emissions. Of course, the catalyst interferes with this relationship, but during rich operation
CO conversion efficiency drops faster than HC conversion efficiency. Thus, tailpipe CO is
still a reasonable predictor of the amount of enrichment.
So, for example, you could see if the single vehicle with significantly increased PM also had
large increases in CO emissions. If there were large increases in CO, then it could have
been an enrichment effect caused by the additional load on the LA92 - especially if the
vehicle had a lower power to weight ratio. If the CO did not increase dramatically, then you
can be more confident that this is actually deterioration of PM.
RESPONSE:
Notwithstanding the aggressiveness of the LA92 with respect to the FTP, the data we
reviewed showed reasonably good and direct correlation between the two, when working
with cycle composites, that would probably not be expected if comparing individual bags.
Perhaps the inclusion of the longer start bag in the FTP compensates for the aggressiveness
of the hot-running bag of the LA92.
3. Page 89, para 2:
Not necessarily. Again, if the LA92 caused additional enrichment, the higher PM could be
from the enrichment. Comparing the CO emission rates, not the HC emission rates, will
give you a better handle on this.
RESPONSE:
See our response to overall comment #4 above.
4. Page 89, para 3:
Again, you shouldn't compare just HC. Enrichment is reflected in CO emissions, not HC
emissions. The CO comparison is more important.
RESPONSE:
See our response to overall comment #4 above.
5. Page 90, para 1:
You can't conclude this just looking at HC. It is more important to compare CO.
RESPONSE:
See our response to overall comment #4 above.
6. Page 91, para 2:
Instead of this general PM methodology, you should evaluate how well your PM test results
correlate with CO and/or HC emissions, after adjusting for fuel effects. If you can establish
a reasonable correlation, then you can adjust PM rates based on HC and CO emissions. This
would be especially useful in estimating PM deterioration.
203
-------
RESPONSE:
At the outset, it is not clear to us how attempting to analyze PM through correlations with
HC and CO (both of which are fairly poor), and introducing associated uncertainty, would
improve the resulting model inputs when PM measurements are available. In addition,
adjustment for fuel effects simply not possible with the older data, as the needed fuel
parameter information is not available.
1. Page 91, para 2:
Again, bag 2 is not representative of hot running emissions, which should be the average of
bag 2 and bag 3.
RESPONSE:
Bag 3 in the LA92 includes a hot-start component. As with the gaseous emissions, we lack a
way to separate the start and running components in Bag 3. For this reason, we have
focused on Bag 2, rather than Bag 3, to represent hot-running emissions.
8. Page 94, 2nd para:
You discuss this later, but it would help the reader if you would state here that the "aged"
data are affected both by vehicle age and model year.
RESPONSE:
We have added text to clarify this point: "the rates in each model-year group represent
emissions for that group at the age of measurement..."
9. Page 96, 3rd para:
Again, CO correlates better with air/fuel ratio than HC and air/fuel ratio has a strong impact
onPM.
RESPONSE:
See our response to overall comment #4 above.
10. Page 98, Figure 2-15:
How does a multiplicative model based on CO instead of HC look? One based on both HC
and CO?
RESPONSE:
Multiplicative patterns for CO look very similar to those for HC. For reference, see Figures
1-18, 1-19 in the draft report (Figures 18 and 19 in the final report).
11. Page ,2.3.1, 1st para:
What about CO (the air/fuel surrogate)? How does this compare with PM emissions?
204
-------
RESPONSE:
See our response to overall comment #4 above.
12. Page 100, 1st para:
Again, HC is not the proper metric to determine when the cold start ends. CO is directly
proportional to the air/fuel ratio and is a better predictor of closed-loop air/fuel control.
RESPONSE:
See our response to overall comment #4 above.
13. Page 101, 1st para:
Again, CO will tell you exactly when the vehicle runs rich and by how much.
RESPONSE:
See our response to overall comment #4 above.
14. Page 106, Figure 2-23:
Note that the PM emissions by VSP look a lot more like the CO emissions by VSP on page
28 than the HC emissions by VSP on page 29. An indication that PM may correlate more
strongly with air/fuel ratio (CO) than with HC.
RESPONSE:
See our response to overall comment #4 above.
15. Page 109, 1st para:
This wasn't "determined" for PM. It was determined for other pollutants and assumed to be
the same for PM.
RESPONSE:
The comment is well taken. We have revised the text to reflect the assumption that
deterioration trends for PM would show patterns similar to those for the gaseous emissions.
205
-------
Chapter 3 Light-Duty Diesel Criteria Exhaust Emissions (HC, CO, NOx)
NOTE: In MOVES2014, all content related to light-duty diesel rates has been revised completely, as
described in Chapter 3 (page 179). Accordingly, the peer review comments previously located here are
no longer relevant and have been removed. The documentation and peer review for the rates used in
MOVES2010 are still available in the previous version of this report, available at:
http://www.epa.gov/otaq/models/moves/documents/420rl 101 l.pdf.4
Chapter 4 Crankcase Emissions
1. Page 119, 5th para:
Should note that the emissions are actually a percentage of engine-out emissions, but that
vehicles before 1969 did not have catalysts, so the tailpipe correlation works.
RESPONSE:
We have added text to clarify this point.
2. Page 119, 6th para:
Is it appropriate to assume the rate is 4% for all vehicles? Should be virtually zero for newer
vehicles, especially those still under warranty, and be higher than 4% for old vehicles, with
some function in between.
RESPONSE:
Text has been added to clarify this point. While this 4% estimate may be pessimistic for new
vehicles, and optimistic for old vehicles, current data does not support a more detailed
estimate. As older vehicles have higher overall emissions due to deterioration effects, this
may understate the impacts ofcrankcase emissions. Should additional data become
available, this may be a candidate for future updates.
3. Page 121, 1st para:
Probably not a good assumption. Diesels have much higher compression ratios and are likely to have
higher blow-by rates.
RESPONSE:
Text has been added to clarify this point. Diesel engines have both higher compression
ratios and require a tighter seal in order to operate. Otto cycle engines potentially allow a
greater proportion of combustion gases to escape to the crankcase. As a result, it is difficult
to predict whether diesel engines have higher or lower crankcase emissions.
That being said, we agree with the commenter that it would be preferable to have data
specific to gasoline engines.
4 USEPA Office of Transportation and Air Quality. Development of Emission Rates for Light-Duty Vehicles in the
Motor Vehicle Emissions Simulator: Final Report. EPA-420-R-11-011. Assessment and Standards Division, Ann
Arbor, MI. August, 2011.
206
-------
4. Page 121, Table 4-2:
Why is HC crankcase emissions 16.5 times larger for gasoline than diesel? Deserves
explanation.
Text has been added to clarify this point. See previous response.
207
-------
Appendix C: Peer-Review Comments and Response for MOVES2010:
Reviewer 2
Reviewer 2: Robert A. Harley, PhD., University of California at Berkeley.
Dr. Harley is a professor in the Department of Civil and Environmental Engineering. He has
conducted research and published extensively in the field of automotive emissions measurement
and control.
This appendix contains comments received from Dr. Harley following conclusion of his review
of the draft report. Following each comment, we have included our specific response, describing
whether we have accepted the comment and made corresponding revisions in the final report, or
whether we have offered a rebuttal or otherwise declined to make revisions.
Note that page and paragraph numbers listed in the comments refer to the draft document: Development of
Emission Rates for Light-Duty Vehicles in the Motor Vehicle Emissions Simulator (MOVES2009): Draft
Report. A copy of this document is available at:
http://www3.epa.gov/otaq/models/moves/techdocs/420p09002.pdf.5
5 USEPA Office of Transportation and Air Quality. Development of Emission Rates for Light-Duty Vehicles in the
Motor Vehicle Emissions Simulator (MOVES2009): Draft Report. EPA-420-P-09-002. Assessment and Standards
Division, Ann Arbor, MI. August, 2009.
208
-------
Review of Draft Report
Development of Emission Rates for Light-Duty Vehicles in the
Motor Vehicle Emissions Simulator (MOVES2009)
June 2009 version
Reviewed by
Robert Harley, Ph.D
Department of Civil and Environmental Engineering
University of California, Berkeley, CA 94720-1710
Prepared for
U.S. EPA Office of Transportation and Air Quality
September 2009
209
-------
1. The draft report is missing introductory and concluding text.
Add Introduction. An introduction should be included, briefly describing motivation for
developing MOVES. It may not be obvious or known to all readers that MOVES is intended to
estimate emissions from both on-road and some off-road mobile sources, replacing the existing
MOBILE and OFFROAD modeling tools. A discussion of why a new modeling approach is
needed should be added (this might be simply a reference to other documents where more details
are available). Where should one look for other MOVES-related documents and information?
Clearly the present report is part of a larger effort, but that context is missing here.
RESPONSE:
We have added material to orient a reader to the broader context of MOVES and its
development, and to refer them to available sources of more detailed information.
The importance of light-duty (LD) vehicles as a source of air pollution should be summarized.
Readers may not be aware of an earlier related report on development of MOVES2004 for
estimating greenhouse gas emissions from LD vehicles. Some relevant background and
findings on mobile source emission trends from recent studies are that:
(a) LD vehicle emissions of carbon monoxide (CO), hydrocarbons (HC), nitrogen oxides
(NOx), and paniculate matter (PM) have declined substantially in recent years [Harley et al.,
2006; Bishop and Stedman, 2008; Ban-Weiss et al., 2008]
(b) in-use deterioration rates have declined for newer vehicles [Bishop & Stedman 2008]
(c) the effect of variations in engine load/driving conditions on emissions is not as large
as in the past. For example, Bishop and Stedman [2008] show that plots of exhaust emission
factors versus vehicle specific power (VSP) are flatter (i.e., there is less variation in emissions
with changes in engine load) for newer vehicles.
(d) the relative importance of other mobile sources of air pollution has increased as LD
vehicle emission control efforts have progressed. For example for NOx, diesel trucks now
dominate total on-road emissions, and there are also significant contributions to NOx from off-
road diesel-powered equipment [Harley et al., 2005; Ban-Weiss et al., 2008].
RESPONSE:
We have added a brief summary of light-duty exhaust emissions and their control. Given the
length of the document, we did not attempt a lengthy or comprehensive discussion, but cite
several sources in the peer-reviewed literature.
Provide Methodology Overview Key features of the modeling approach for LD vehicles
emissions that should be summarized at the outset include use of g/hr (rather than g/km or g/kg)
emission factors, binning of emission factors with vehicle specific power (VSP) serving as a
master variable, changes in how I/M vs. non-I/M areas are modeled, and reliance on emissions
data from Phoenix for pre-2000 vehicles, and manufacturer-conducted in-use vehicle emissions
testing for post-2000 vehicles.
210
-------
RESPONSE:
We have added an overview describing the content of sections and subsections in Chapter 1,
and briefly discussing important differences between MOVES and MOBILE, including
changes with respect to time-based rather than distance-based emission rates, and changes
in the approach to modeling I/M.
Add Recommendations/Conclusions What are the data needs, and how will new data be
incorporated into the MOVES model to update it? What are key areas of uncertainty that would
benefit from additional study?
Most importantly, how will MOVES be verified against independent data that were not used in
model development? In addition to doing overall comparisons of MOVES against older
MOBILE/OFFROAD models, I recommend efforts to evaluate/verify each component of
MOVES as part of the model development process. Documenting the modeling approach and
input data can serve as a starting point for this, but the evaluation itself (or at least plans for such
an evaluation) is missing from the present LD vehicle emissions report. If this were a manuscript
being reviewed for publication in a scientific journal, verification of the model predictions using
independent observations would be an essential component. Given the potential future
importance of MOVES predictions to national air pollution control policy, I believe that
similarly high standards should apply with respect to including model evaluation/verification in
this report.
RESPONSE:
Since the release of the draft model and database, we have made substantial efforts to verify
several aspects of MOVES with respect to independent data. Of course, it is very difficult
and labor intensive to verify evaporative and start emissions, which is unfortunate, given
their importance. Nonetheless, we have verified running emissions for light-duty (and heavy-
duty) emissions and have identified several areas that merit attention and improvement. The
results of these analyses will be made available in a separate report, within which data and
research needs will be discussed.
2. There is unnecessary/imprecise use of acronyms and jargon where plain language would
be clearer and more accurate.
Light-duty vehicles (LDV) include cars as well as light trucks. The report often uses the LDV
acronym in text, figures, and tables when referring only to cars. In those situations, I
recommend using the term "cars". Use of LDV when you mean cars is confusing: for me LDV
includes light trucks as well. The report title uses the phrase "Light-Duty Vehicles" which
emphasizes the point that both cars and LD trucks are included under the heading LDV.
211
-------
RESPONSE:
We have added a paragraph describing the technical designations for "LDV" and "LDT. "
Throughout the rest of the document, we have simplified the discussion by substituting "car "
and "truck" for "LDV" and "LDT. "
I recommend changing the first chapter title to be "Gaseous" rather than "Criteria" pollutant
emissions. Criteria pollutant is confusing jargon. Isn't PM also a criteria pollutant? Also I
suggest merging the discussion of gaseous emissions from LD diesel vehicles (currently
section 3) together with section 1. This would provide a more consistent parallel report
structure between gaseous and PM emissions, and between gasoline and diesel engines.
RESPONSE:
The comment is well taken, given that parti culate matter is also a criteria pollutant.
Accordingly, we have revised the document to refer to HC, CO andNOx as the "gaseous
exhaust pollutants."
3. Discussion of evaporative emissions and weather effects on all emissions is missing.
I did not find any discussion of evaporative emissions. If a separate report is planned to discuss
methods for representing evaporative emissions, that should be mentioned. Also climate
variables such as temperature and humidity affect vehicle emissions through increases in cold
start emissions as ambient temperature decreases, increased evaporative emissions with
increasing diurnal temperature range, and increased use of vehicle air conditioning on hot days. I
did not find any discussion of how changes in weather affect vehicle emissions, or how such
effects are modeled in MOVES. Ambient temperature affects gaseous as well as PM emissions,
including cold start effects and gas/particle partitioning of semi-volatile organics present in the
exhaust.
RESPONSE:
Both evaporative emissions and adjustments to exhaust emissions are discussed in separate
reports. The revised report refers readers to these additional documents.
A potential problem with using emissions data from Phoenix to represent all pre-2000 model
year vehicles nationwide is that mild winters in Phoenix may extend vehicle lifetimes and
reduce in-use emission deterioration rates relative to other parts of the country that
experience more severe weather.
RESPONSE:
777/5 issue was raised and considered in the FACA MOVES Review workgroup. It is difficult
to address because datasets broad and deep enough to assess deterioration are difficult to
locate, and where they do exist (such as remote-sensing), observed differences potentially
attributable to "climate " are likely to be confounded by several other factors. Prominent
212
-------
confounders can include measurement differences related to instrumentation and
calibration, differences in fueI composition, differences in I/M requirements, differences in
the degree of representativeness, and random error. Nonetheless, we attempted to evaluate
the issue by comparing the Phoenix data to evaluation data from the Chicago I/M program,
as well as data collected during the New York Instrumentation Protocol Assessment
(NYIPA). Both the Chicago and New York data represent fleets operating in colder climates
with harsher winters than in Phoenix. Aside from differences ostensibly attributable to I/M
requirements, e.g., the lack of a NOX requirement in Chicago, the three programs appeared
comparable enough to suggest that "climate " per se was not a major issue.
4. Reliance on IUVP Data Raises Concerns
Going forward, the pre-2000 vehicle model year data from Phoenix will play a minor role in
determining LD vehicle emissions, as those vehicles are 10+ years old already, and they will
constitute a declining fraction of the in-use vehicle fleet in assessments of future year emissions.
For emissions from 2000 model year and newer vehicles, the MOVES model relies on data
from the In-Use Verification Program (IUVP), a program that started in 2003 which is
administered by EPA and run by the vehicle manufacturers. Relying on IUVP data to model
vehicle emissions is a questionable approach. The IUVP appears to have been instituted as a
regulatory compliance program, and as such may not be well-suited to capturing the full range
of vehicles, operating conditions, and emissions that are relevant to the MOVES model and
developing emission inventory estimates. For example, will the vehicle sample in IUVP be
large and random enough to ensure that major emission contributions from malfunctioning/
high-emitting vehicles will be captured? How will high-emitting vehicles be represented in
MOVES? As fleet-average emissions decline, the remaining emissions are increasingly
dominated by contributions from high-emitting vehicles (Bishop and Stedman, 2008).
RESPONSE:
We agree that the IUVP is not necessarily designed to obtain representative data for the
entire "real world" fleet. But it is important to remember that similar issues apply equally
to most sources of emissions data, including high-quality laboratory studies reported in the
peer-reviewed literature, which very frequently use relatively small vehicle samples. This
situation is entirely understandable, given the difficulty and expense of measurement using
dynamometers or portable instruments combined with the difficulty of acquiring
representative samples. Despite these questions, we found the IUVP to be a very valuable
source, in that it provided information allowing assignment of standard level to individual
vehicles, which we found indispensable in projecting NLEV and Tier-2 emissions. However,
we did not take the representativeness of IUVP entirely for granted, and the approaches we
adopted compensated in three ways. (1) We used IUVP only to estimate rates for "young"
vehicles, aged 0-3 years. (2) we used the IUVP to develop scaling factors that we applied to
results from Phoenix I/M that represented Tier-1 rates. Thus the rates developed using the
IUVP data incorporate a direct link to "real-world" results, that probably more effectively
represent "high-emitting " or "mal-functioning " vehicles. (3) In projecting deterioration
from NLEV and Tier-2 vehicles, we used logarithmic variances "borrowed" from the
Phoenix I/M data, which are higher than those obtained directly from the IUVP. Because
this parameter represents the degree of skew in the distribution, increasing it effectively
213
-------
represents an increase in the fraction of "high-emitting" vehicles, with associated increases
in mean emission rates.
5. PMEmissions
MOVES relies on data from the 2005 Kansas City PM study which focused on LD gasoline
vehicles. This was a comprehensive and well-conducted study and the resulting emissions data
are relatively current. Cold start as well as running emissions were measured. A limitation of
this study is the lack of information on how vehicle emissions for a given model year increase
with vehicle age/odometer reading.
On pp. 91-92, following equation 2-1, there is excessive precision and no associated uncertainty
reported for the multiplicative adjustment factors (0.898 and 1.972) used to give hot running and
cold start emission rates for LD trucks. Excessive precision and lack of uncertainty estimates
concern also applies to values presented in Table 2-2.
There is no discussion of PM emission rates from LD diesel vehicles; the Kansas City study
was for LD gasoline vehicles only. Chapter 3 on LD diesel vehicle emissions covers only the
gaseous pollutant emissions, not exhaust PM. Also non-exhaust PM emissions (e.g., tire wear,
brake wear) from LD vehicles are not discussed in this report.
RESPONSE:
Chapter 3 does cover PM emissions for light-duty diesel vehicles, as it does for
HC/CO/NOx.
As a model evaluation case study, EPA staff may wish to consider long-term LD vehicle PM
emission trends reported by Ban-Weiss et al. (2008) at the Caldecott tunnel in California. LD
vehicle fleet-average PM2.5 mass emission rates were measured to have decreased by 36±17%
over a 9-year time period between 1997 and 2006, due to model year effects on zero mile levels
and/or deterioration rates. Both VSP and average vehicle age (i.e., calendar year-average model
year) were similar between the two field campaigns. Cold start emissions were not measured by
Ban-Weiss et al.
RESPONSE:
We thank the reviewer for pointing out the availability of this dataset as an opportunity to
verify PM predictions. It could make a good candidate for a future verification effort.
References
Ban-Weiss, G.A.; McLaughlin, J.P.; Harley, R.A.; Lunden, M.M.; Kirchstetter, T.W.; Kean,
A.J.; Strawa, A.W.; Stevenson, E.D.; Kendall, G.R. (2008). Long-Term Changes in Emissions of
Nitrogen Oxides and Particulate Matter from On-Road Gasoline and Diesel Vehicles.
Atmospheric Environment 42, 220-232.
214
-------
Bishop, G.A.; Stedman, D.H. (2008). A Decade of On-Road Emissions Measurements.
Environmental Science & Technology 42, 1651-1656.
Harley, R.A.; Marr, L.C.; Lehner, J.K.; Giddings, S.N. (2005). Changes in Motor Vehicle
Emissions on Diurnal to Decadal Time Scales and Effects on Atmospheric Composition.
Environmental Science & Technology 39, 5356-5362.
Harley, R.A.; Hooper, D.S.; Kean, A.J.; Kirchstetter, T.W.; Hesson, J.M.; Balberan, N.T.;
Stevenson, E.D.; Kendall, G.R. (2006). Effects of Reformulated Gasoline and Motor Vehicle
Fleet Turnover on Emissions and Ambient Concentrations of Benzene. Environmental Science &
Technology 40, 5084-5088.
215
-------
6 References
1 USEPA Mobile Source Emissions Modeling Workgroup. EPA's New Generation Mobile Source Emissions
Model: Initial Proposal and Issues. EP A420-R-01 -007. Office of Air and Radiation, Office of Research and
Development, Region 4. April, 2001. (http://www.epa.gov/otaq/models/ngm/r01007.pdf).
2 USEPA Office of Transportation and Air Quality. Draft Design and Implementation Plan for EPA 's Multi-Scale
Motor Vehicle and Equipment Emission System (MOVES). EPA420-P-02-006. Assessment and Standards Division,
Ann Arbor, MI. October, 2002. (http://www.epa.gov/otaq/models/ngm/p02006.pdf).
3 Computational Laboratory for Energy, Air, and Risk. Methodology for Developing Modal Emission Rates for
EPA 's Multi-Scale Motor Vehicle and Equipment Emission System. EPA420-R-02-027. Department of Civil
Engineering, North Carolina State University, Raleigh, NC. October, 2002.
(http://www.epa.gov/otaq/models/ngm/r02027.pdf).
4 USEPA Office of Transportation and Air Quality. EPA 's Onboard Emissions Analysis Shootout: Overview and
Results. EPA420-R-02-026. Assessment and Standards Division, Ann Arbor, MI. October, 2002.
(http://www.epa.gov/otaq/models/ngm/r02026.pdf).
5 USEPA (2014). MOVES2014 Software Design Reference Manual. EPA-420-B-14-056. Office of Transportation
and Air Quality. US Environmental Protection Agency. Ann Arbor, MI. December, 2014.
http://www.epa.gov/otaq/models/moves/documents/420bl4056.pdf.).
6 USEPA (2012). Use of Data from "Development of Emission Rates for the MOVES Model, " Sierra Research,
March 3, 2010. EPA-420-R-12-022. Office of Transportation and Air Quality. US Environmental Protection
Agency. Ann Arbor, MI. August, 2012. http://www.epa.gov/otaq/models/moves/documents/420rl2022.pdf.).
7 USEPA Office of Transportation and Air Quality. Cars and Light Trucks: Vehicle Standards and Regulations.
http://www.epa.gov/otaq/standards.htm.
8 Calvert, J.G.; Heywood, J.B.; Sawyer, R.F.; Seinfeld, J.H. 1993. Achieving acceptable air quality: some
reflections on controlling vehicle emissions. Science. 261:37-44.
9 Sawyer, R.F.; Harley, R.A.; Cadle, S.H.; Norbeck, J.M.; Slott, R.; Bravo, H.A. 2000. Mobile sources critical
review: 1998 NARSTO assessment. Atmospheric Environment. 34:2161-2181.
10 Harley, R.A.; Marr, L.C.; Lehner, J.K.; Giddings, S.N. 2005. Changes in motor vehicle emissions on diurnal to
decadal time scales and effects on atmospheric composition. Environ. Sci. Technol. 39:5356-5362.
11 Harley, R.A.; Hooper, D.S.; Kean, A.J.; Kirchstetter, T.W.; Hesson, J.M.; Balberan, N.T.; Stevenson, E.D.;
Kendall, G.R. 2006. Effects of Reformulated Gasoline and Motor Vehicle Fleet Turnover on Emissions and
Ambient Concentrations of Benzene. Environ. Sci. Technol. 40: 5084-5088.
12 Bishop, G.A.; Stedman, D.H. 2008. A decade of on-road emissions measurements. Environ. Sci. Technol.
42:1651-1656.
13 Ban-Weiss, G.A.; McLaughlin, J.P.; Harley, R.A.; Lunden, M.M.; Kirchstetter, T.W.; Kean, A.J.; Strawa, A.W.;
Stevenson, E.D.; Kendall, G.R. 2008. Long-Term Changes in Emissions of Nitrogen Oxides and Paniculate Matter
from On-Road Gasoline and Diesel Vehicles. Atmospheric Environment. 42: 220-232.
14 Wenzel, T. 2001. Reducing emissions from in-use vehicles: an evaluation of the Phoenix inspection and
maintenance program using test results and independent emissions measurements. Environmental Science & Policy.
4:359-376.
15 USEPA Office of Transportation and Air Quality. Greenhouse Gas and Energy Consumption Rates for On-road
Vehicles: Updates for MOVES2014. Assessment and Standards Division, Ann Arbor, MI. September, 2015.
(http:/www.epa.gov/otaq/models/moves/moves-reports.htm).
216
-------
16 USEPA Office of Transportation and Air Quality. Evaporative Emissions from On-road Vehicles in
MOVES2014. EPA-420-R-14-014. Assessment and Standards Division, Ann Arbor, MI. September, 2014.
17 USEPA Office of Transportation and Air Quality. Fuel Consumption Modeling of Conventional and Advanced
Technology Vehicles in the Physical Emission Rate Estimator (PERE). EPA420-P-05-001. Assessment and
Standards Division, Ann Arbor, MI. February, 2005. (http://www.epa.gov/otaq/models/ngm/420p05001.pdf).
18 USEPA Office of Transportation and Air Quality. Population and Activity of On-road Vehicles in MOVES2014:
Draft Report. EPA-420-D-15-001. Assessment and Standards Division, Ann Arbor, MI. July, 2015.
(http://www.epa.gov/otaq/models/moves/moves-reports.htm).
19 USEPA Office of Air and Radiation, Office of Transportation and Air Quality. IM240 and Evap Technical
Guidance. EPA420-R-00-007. Transportation and Regional Programs Division, Ann Arbor, MI. April, 2000.
(http://www.epa.gov/oms/regs/im/r00007.pdf). (page 106).
20 Heirigs, P.; Dulla, R.; Crawford, R.W. Processing ofIM240 Data for Use in MOVES. SR007-05-02. Sierra
Research, Inc., Sacramento, CA. May, 2007.
21 Singer, B.C.; Harley, R.A.; Littlejohn, D.; Ho, J.; Vo, T. 1998. Scaling of infrared remote sensor hydrocarbon
measurements for motor vehicle emission inventory calculations. Environ. Sci. Technol. 32(21)3241-3248.
22 Bishop, G.A.; Stedman, D.H. On-road Remote Sensing of Automobile Emissions in the Denver Area: Year 6,
January 2007. Department of Chemistry and Biochemistry, University of Denver, Denver, CO. June 2007.
23 Kish, L. Survey Sampling. John Wiley & Sons, New York. 1965.
94
Neter, J.; Kutner, M.H.; Nachtsheim, C.J.; Wasserman, W. Applied Linear Statistical Models. Irwin, Chicago.
Fourth Edition. 1996.
25 Barth, M.; An, F.; Younglove, T.; Scora, G.; Levine, C.; Ross, M.; Wenzel, T. Development of a Comprehensive
Modal Emissions Model: Final Report. National Research Council, Transportation Research Board, National
Cooperative Highway Research Program. NCHRP Project 25-11. April, 2000.
(http://www.cert.ucr.edu/cmem/docs/NCHRP_Final_Report.pdf).
McClement, D.; Dulla, R.G. Identification ofNon-I/M Vehicles in I/MProgram Vehicle Emission Datasets.
Draft Report, Sierra Research, Sacramento, CA. October, 2007.
27 Air Quality Group, Aerospace, Transportation and Advanced Systems Laboratory, Georgia Technical Research
Institute. Biennial Evaluation of the Emissions Reduction Effectiveness of the Atlanta Vehicle Inspection and
Maintenance Program for 2003-2004. Prepared for the Air Protection Branch, Environmental Protection Division,
Georgia Department of Natural Resources, Atlanta. September, 2007.
28 DeHart-Davis, L.; Corley, E.; Rodgers, M.O. 2002. Evaluating vehicle inspection/maintenance programs using
on-road emissions data. Evaluation Review. 26(2) 111-146.
29 Annual Certification Test Results & Data: Cars and Light Trucks, http://www.epa.gov/otaq/crttst.htm.
30 Engine and Vehicle Compliance System (VERIFY): http://www.epa.gov/otaq/verify/index.htm.
31 USEPA, Office of Transportation and Air Quality. Mobile Source Observation Database (MSOD): User Guide
and Reference. Assessment and Standards Division, Ann Arbor, MI. EPA420-B-08-017. December, 2008.
(http://www.epa.gov/otaq/models/msod/420b08017.pdf).
32 Enns, P.; Brzezinski, D. Comparison of 'Start Emissions in the LA92 andST01 Test Cycles. EPA420-R-01-025.
(M6.STE.001), USEPA Office of Transportation and Air Quality, Ann Arbor, MI. April, 2001.
33 Glover, E.; Carey, P. Determination of Start Emissions as a Function of Mileage and Soak Time for 1981-1993
Model-year Light-Duty Vehicles. EPA420-R-01-058 (M6.STE.003). USEPA Office of Transportation and Air
Quality, Ann Arbor, MI. November, 2001.
34 Glover, E.; Brzezinski, D. Exhaust Emission Temperature Correction Factors for MOBILE6: Adjustments for
Engine Start and Running LA4 Emissions for Gasoline Vehicles. EPA420-R-01-029 (M6.STE.004). USEPA Office
of Transportation and Air Quality, Ann Arbor, MI. April, 2001.
217
-------
35 USEPA Office of Transportation and Air Quality. Emission Adjustments for Temperature, Humidity, Air
Conditioning, and Inspection and Maintenance for On-road Vehicles in MOVES2014. EPA-420-R-14-012.
Assessment & Standards Division, Ann Arbor, MI. December, 2014.
(http://www.epa.gov/otaq/models/moves/moves-reports.htm).
36 Stump, F.D.; Dropkin, D.L.; Tejada, S.B.; Loomis, C.; Pack, C. Characterization of Emissions from
Malfunctioning Vehicles Fueled with Oxygenated Gasoline-Ethanol (E-10) Fuel-Part III. EPA/600/R-01/053
(NTIS PB2004-106735). USEPA National Exposure research Laboratory (NERL). July, 2002.
37 Whitney, K.A. Collection ofIn-Use Mobile Source Emission Samples for Toxicity Testing. 08.02602. Southwest
Research Institute, San Antonio, Texas. October, 2000.
39 Sabate, S. Methodology for Calculating and Redefining Cold and Hot Start Emissions. California Air
Resources Board, El Monte, CA. March, 1996. As cited in: Glover, E.; Carey, P. Determination of Start Emissions
asaFunction of Mileage and Soak Time for 1981-1993 Model-year Light-Duty Vehicles. EPA420-R-01-058
(M6.STE.003). November, 2001 (Reference 15).
40 USEPA Office of Transportation and Air Quality. Sales of California Vehicles for 2011 Model Year and Beyond
(Cross-Border Sales Policy). Memorandum CISD-11-06 (LDV/LDT/HD). Compliance and Innovative Strategies
Division, Ann Arbor, MI. May 3, 2011. (http://iaspub.epa.gov/otaqpub/display_file.jsp?docid=24724&flag=l).
41 USEPA Office of Transportation and Air Quality. Control of Air Pollution from Motor Vehicles: Tier 3 Motor
Vehicle Emission and Fuel Standards; Final Rule. Federal Register. Vol. 79, No. 81. April 28, 2014. Page 23417.
43 Nam E.; Fulper, C.; Warila, J.E.; Somers, I; Michaels, H.; Baldauf, R.; Rykowski, R.; Scarbro, C. Analysis of
P'articulate MatterEmissions from Light-Duty Vehicles in Kansas City. EPA420-R-08-010. USEPA, Office of
Transportation and Air Quality, Ann Arbor, MI; Office of Research and Development, Research Triangle Park, NC.
April, 2008.
44 Kishan, S.; Burnette, A.; Fincher, S.; Sabisch, M.; Crews, W.; Snow, R.; Zmud, M.; Santos, R.; Bricka, S.;
Fujita, E.; Campbell, D.; Arnott, P. Kansas City PM Characterization Study: Final Report. EPA420-R-08-009.
USEPA Office of Transportation and Air Quality, Ann Arbor, MI. April, 2008.
45 Zielinska, B.; Sagebiel, I; McDonald, J. D.; Whitney, K.; Lawson, D.R. Emission rates and comparative
chemical composition from selected in-use diesel and gasoline fueled vehicles. J. Air Waste Manage. Assoc. 2004.
54:1138-1150.
46 Fujita, E. M.; Zielinska, B.; Campbell, D. E.; Arnott, W. P.; Sagebiel, J. C.; Mazzoleni, L.; Chow, J. C.; Gabele,
P. A.; Crews, W.; Snow, R.; Clark, N. N.; Wayne, W. S.; Lawson, D. R.. Variations in speciated emissions from
spark-ignition and compression-ignition motor vehicles in California's South Coast Air Basin. J. Air & Waste
Manage. Assoc. 2007. 57:705-720.
47 Lough, G. C.; Christenson, C. C.; Schauer, J. J.; Tortorelli, J.; Bean, E.; Lawson, D. R.; Clark, N. N.; Gabele, P.
A. 2007. Development of molecular marker source profiles for emissions from on-road gasoline and diesel vehicle
fleets. J. Air Waste Manage. Assoc. 2007a. 57:1190-1199.
48 Sonntag, D. B.; Bailey, C. R.; Fulper, C. R.; Baldauf, R. W. Contribution of lubricating oil to particulate-matter
emissions from light-duty gasoline vehicles in Kansas City. Environ. Sci. Technol. 2012. 46(7): 4191-4199.
49 Gibbs, R.E.; Wotzak, G.P.; Byer, S.M.; Kolak, N.P. Sulfates and P articulate Emissions from In-Use Catalyst
Vehicles. Regulated/Unregulated Emissions and Fuel Economy. EPA-600/9-79-047. US Environmental Protection
Agency, 1979.
50 Hammerle, R.H.; Korniski, T.J.; Weir, J.E.; Cladek, E.; Gierczak, C.A.; Chase, R.E.; Hurley, R.G. Effect of
Mileage Accumulation on Particulate Emissions from Vehicles Using Gasoline with Methylcyclopentadienyl
Manganese Tricarbonyl. SAE920731. Society of Automotive Engineers. 1992.
51 Whitney, K.A. Collection of In-Use Mobile Source Emission Samples for Toxicity Testing. SwRI 08.02602.
Southwest Research Institute, San Antonio, TX. October, 2000.
218
-------
52 Cadle, S.H.; Nebel, G.J.; Williams, R.L. Measurement of Unregulated Emissions from General Motors' Light-
Duty Vehicles. SAE790694. Society of Automotive Engineers. 1979.
53 Urban, C.M.; Garbe, R.J. Regulated and Unregulated Exhaust Emissions from Malfunctioning Automobiles.
SAE790696. Society of Automotive Engineers. 1979.
Urban, C.M.; Garbe, R.J. Exhaust Emissions from Malfunctioning Three-Way Catalyst Equipped Automobiles.
SAE800511. Society of Automotive Engineers. 1980.
55 Lang, J.M.; Snow, L.; Carlson, R.; Black, F.; Zweidinger, R.; Tejeda, S. Characterization ofParticulate
Emissions from In-Use Gasoline-Fueled Motor Vehicles. Society of Automotive Engineers. SAE 811186. 1981.
56 Volkswagen AG. Unregulated Motor Vehicle Exhaust Emission Gas Components. Volkswagen Aktien
Gesellschaft, Forschung und Entwicklung [Research and Development], Wolfsburg, Germany. 1991.
57 California Air Resources Board. F'articulate Exhaust Emissions: Gasoline Powered Vehicles. Emissions
Studies Section, El Monte, CA. Project No. 2R8618. 1986.
58 Cadle, S.H.; Mulawa, P.; Hunsanger, E.G.; Nelson, K.; Ragazzi, R.A.; Barrett, R.; Gallagher, G.L.; Lawson,
D.R.; Knapp, K.T.; Snow, R. Measurement of Exhaust Particulate Matter Emissions from In-use Light-duty Motor
Vehicles in the Denver, Colorado Area. CRCE-24-1. Coordinating Research Council, Alpharetta, GA. NTIS
PB98-136401. March, 1998.
59 Norbeck, J.M.; Durbin, T.D.; Turex, T. J. Measurement of Primary Particulate Matter Emissions from Light-
Duty Motor Vehicles. 98-VE-RT2A-001-FR. University of California, College of Engineering, Center for
Environmental Research and Technology, Riverside, CA. CRC E-24-2. Coordinating Research Council, Alpharetta,
GA. NTIS PB99-151755. 1998.
60 Whitney, K. Measurement of Primary Exhaust Particulate Matter Emissions from Light-duty Motor Vehicles.
CRCE-24-3. Coordinating Research Council, Alpharetta, GA. 1998. NTISPB99-121279. November, 1998.
61 Chase, R.E.; Duszkiewicz, G.J.; Jensen, T.E.; Lewis, D.; Schlaps, E.J.; Weibel, A.T.; Cadle, S.; Mulawa, P.
Particle mass emission rates from current-technology light duty gasoline vehicles. J Air & Waste Manage. Assoc.
2000: (50)930-935.
62 Southwest Research Institute. VOC/PM Cold Temperature Characterization and Interior Climate Control
emissions/Fuel Economy Impact. SwRI 03.11382.04. San Antonio, TX. Prepared for USEPA, Office of
Transportation and Air Quality, Ann Arbor, MI. October, 2005.
63 Li, W.; Collins, J.F.; Norbeck, J.M.; Cocker, D.R.; Sawant, A. Assessment of Particulate Matter Emissions from
a Sample of In-Use ULEV and SULEV Vehicles. Society of Automotive Engineers. SAE 2006-01-1076. 2006.
64 U.S. Environmental Protection Agency. Study of Particulate Matter Emissions from Motor Vehicles. 1981.
65 USEPA (2014). Speciation of Total Organic Gas and Particulate Matter Emissions from On-road Vehicles in
MOVES2014. EPA-420-R-14-020. Ann Arbor, MI. October, 2014. http://www.epa.gov/otaq/models/moves/moves-
reports.htm).
66 Fujita, E.M.; Campbell, D.E.; Zielinska, B. Chemical Analysis of Lubrication Oil Samples from a Study to
Characterize Exhaust Emissions from Light-Duty Gasoline Vehicles in the Kansas City Metropolitan Area. Desert
Research Institute, Division of Atmospheric Sciences, Reno, NV. CRCE-69a. Coordinating Research Council,
Alpharetta, GA. December, 2006.
67 Sonntag, D. B., R. W. Baldauf, C. A. Yanca and C. R. Fulper (2013). Particulate matter speciation profiles for
light-duty gasoline vehicles in the United States. Journal of the Air & Waste Management Association, 64 (5), 529-
545.
68 USEPA (2014). Emission Adjustments for Temperature, Humidity, Air Conditioning, and Inspection and
Maintenance for On-road Vehicles in MOVES2014. EPA-420-R-14-012. Assessment and Standards Division.
Office of Transportation and Air Quality. US Environmental Protection Agency. Ann Arbor, MI. 2014.
http://www.epa.gov/otaq/models/moves/moves-reports.htm.).
69 US Environmental Protection Agency. EPA Motor Vehicle Tampering Survey. EPA 420-A-90-001. 1990.
219
-------
70 Bennet, P. A. et al. Reduction of Air Pollution by Control of Emissions from Automotive Crankscase. SAE Paper
142A, January 1960.
71 Bowditch, F. W. The Automobile and Air Pollution. SAE Paper No. 680242. Presented at SAE Mid-Year
Meeting, Milwaukee, WI. May, 1968.
72 Montalvo and Hare, 1985. Crankcase Emissions with disabled PCV systems EPA 460/3 -84-011. Southwest
Research Inst, San Antonio, TX. Environmental Protection Agency, Emission Control Technology Div.
73 Title 40: Code of Federal Regulations. Part 86- Protection of Environment. Control of Emissions from New and
In-Use Highway Vehicles and Engines. 86.1810-01 Subpart S—General Compliance Provisions for Control of Air
Pollution From New and In-Use Light-Duty Vehicles, Light-Duty Trucks, and Complete Otto-Cycle Heavy-Duty
Vehicles. General standards; increase in emissions; unsafe conditions; waivers.
74 USEPA (2015). Exhaust Emission Rates for Heavy-Duty On-mad Vehicles in MOVES2014. EPA-420-R-15-004.
Assessment and Standards Division. Office of Transportation and Air Quality. US Environmental Protection
Agency. Ann Arbor, MI. 2015. http://www.epa.gov/otaq/models/moves/moves-reports.htm.
75 Kurtenbach, R., K. H. Becker, J. A. G. Gomes, J. Kleffmann, J. C. Lorzer, M. Spittler, P. Wiesen, R. Ackermann,
A. Geyer and U. Platt (2001). Investigations of emissions and heterogeneous formation of HONO in a road traffic
tunnel. Atmos Environ, 35 (20), 3385-3394.
76 USEPA Office of Transportation and Air Quality. Regulatory Impact Analysis, Control of Air Pollution from
Motor Vehicles: Tier 3 Motor Vehicle Emission and Fuel Standards, Final Rule (Chapter 7). EPA-420-R-14-005.
Assessment and Standards Division, Washington, D.C., Ann Arbor, MI. February, 2014. (Also available at
www.regulations.gov. Docket EPA-HQ-OAR-2011-0135, Item 5214).
77 USEPA Office of Transportation and Air Quality. Developing Emission Rates Representing the Effect of
Proposed Tier-3 Standards. Memorandum to Docket EPA-HQ-OAR-2011-0135, Item 0531. Assessment and
Standards Division, Ann Arbor, MI. December, 2011.
220
------- |