Air Toxic Emissions from On-road
Vehicles in MOVES2014
£% United States
Environmental Protect
Agency
-------
Air Toxic Emissions from On-road
This technical report does not necessarily represent final EPA decisions or
positions. It is intended to present technical analysis of issues using data
that are currently available. The purpose in the release of such reports is to
facilitate the exchange of technical information and to inform the public of
technical developments.
Vehicles in MOVES2014
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
NOTICE
4>EPA
United States
Environmental Protection
Agency
EPA-420-R-16-016
November 2016
-------
6.1.1 Vapor Venting, Fuel Leaks, and Refueling Emission Processes 64
6.1.2 Permeation 66
6.2 Diesel Vehicles 67
7 Crankcase Emissions 68
7.1 Volatile Organic Compounds 68
7.2 Polycyclic Aromatic Hydrocarbons 68
7.3 Metal and Dioxin Emissions 68
References 69
Appendix A Development of Motor Vehicle Emission Factors for
Chromium 74
Appendix B Development of Motor Vehicle Emission Factors for
Mercury 79
Appendix C Responses to Peer-Review Comments 84
2
-------
1 Introduction: Air Toxics
Through MOVES, users can estimate inventories for selected compounds identified as air toxics
in the National Emission Inventory (NEI) and National Air Toxics Assessment (NATA), and for
which adequate data are available to develop emissions estimates. This document describes the
data and methods used to estimate emissions of toxic compounds emitted from highway vehicles
in the MOVES2014 database and model. The current release of the MOVES database
(MOVES2014) includes substantial updates to inputs and structures used to estimate emissions
of air toxics, incorporating data from recent programs conducted on new vehicles employing
current technologies. It also includes the capability to estimate emissions for ethanol blends
containing more than 10% ethanol, including E15, E20, and E85 (70-100% ethanol).
The toxics included in MOVES are classified into four categories:
1) Volatile Organic Compounds (VOC): EPA defines VOC as any compound of carbon,
excluding carbon monoxide, carbon dioxide, carbonic acid, metallic carbides or
carbonates, and ammonium carbonate, which participates in atmospheric photochemical
reactions, except those designated by EPA as having negligible photochemical reactivity1
2) Polycyclic aromatic hydrocarbons (PAHs): This category is defined as hydrocarbons
containing fused aromatic rings. These compounds can be measured in the gaseous phase,
particulate phase, or both, depending on properties of the compound, particle
characteristics and conditions in the exhaust stream or the atmosphere.
3) Dioxins and furans: This category includes polychlorinated organic compounds which are
persistent in the environment and considered bioaccumulative in aquatic and terrestrial
food chains.
4) Metals: This category includes metals or metal-containing compounds in elemental,
gaseous and particulate phases.
Specific compounds in each category are listed in Table 1 through Table 4. Note that each
compound is identified by its "pollutantID" in the MOVES database. With the exception of the
metal species in Table 4, each compound is also identified by its Chemical Abstracts Service
Registry number (CAS number).2 For most other compounds, the identifier for the National
Emissions Inventory (NEIPollutantCode in the table "Pollutant") is identical to the CAS number
(minus the dashes).
3
-------
Table 1. Hydrocarbons and Volatile Organic Compounds included in MOVES2014.
Pollutant
pollutantID
CAS Number
Benzene
20
71-43-2
Ethanol
21
64-17-5
1,3-Butadiene
24
106-99-0
Formaldehyde
25
50-00-0
Acetaldehyde
26
75-07-0
Acrolein
27
107-02-8
Methyl-Tertiary-Butyl Ether (MTBE)
22
1634-04-4
2,2,4-Trimethylpentane
40
540-84-1
Ethyl Benzene
41
100-41-4
Hexane
42
110-54-3
Propionaldehyde
43
123-38-6
Styrene
44
100-42-5
Toluene
45
108-88-3
Xylene(s)1
46
1330-20-7
1 This species represents the sum of emissions from three isomers of xylene, i.e., ortho-,
meta-, and para-xylene.
Table 2. Polycyclic Aromatic Hydrocarbons included in MOVES2014.
Pollutant
pollutantID
CAS Number
(gaseous phase)
(particulate
phase)
Acenaphthene
170
70
83-32-9
Acenaphthylene
171
71
208-96-8
Anthracene
172
72
120-12-7
Benz(a)anthracene
173
73
56-55-3
Benzo(a)pyrene
174
74
50-32-8
Bcnzo(/))fluoranthcne
175
75
205-99-2
Benzo(g,/2,/)perylene
176
76
191-24-2
Benzo(£)fluoranthene
111
77
207-08-9
Chrysene
178
78
218-01-9
Dibenzo(a, /i)anthraccnc
168
68
53-70-3
Fluoranthene
169
69
206-44-0
Fluorene
181
81
86-73-7
Indeno( 1,2,3 ,c, c/)pyrcnc
182
82
193-39-5
Naphthalene
185
23
91-20-3
Phenanthrene
183
83
85-01-8
Pyrene
184
84
129-00-0
4
-------
Table 3. Dioxins and Furans included in MOVES2014
Pollutant
pollutantID
CAS Number
2,3,7,8 -T etrachlorodibenzo-p -Dioxin
142
1746-01-6
1,2,3,7,8-Pentachlorodibenzo-p-Dioxin
135
40321-76-4
1,2,3,4,7,8-Hexachlorodibenzo-p-Dioxin
134
39227-28-6
1,2,3,6,7,8-Hexachlorodibenzo-p-Dioxin
141
57653-85-7
1,2,3,7,8,9-Hexachlorodibenzo-p-Dioxin
130
19408-74-3
1,2,3,4,6,7,8-Heptachlorodibenzo-p-Dioxin
132
35822-46-9
Octachlorodibenzo-p-dioxin
131
3268-87-9
2,3,7,8 -T etrachlorodibenzofiiran
136
51207-31-9
1,2,3,4,6,7,8-Heptachlorodibenzofuran
144
67562-39-4
1,2,3,4,7,8,9-Heptachlorodibenzofuran
137
55673-89-7
1,2,3,4,7,8 -Hexachlorodibenzofuran
145
70648-26-9
1,2,3,6,7,8-Hexachlorodibenzofuran
140
57117-44-9
1,2,3,7,8,9-Hexachlorodibenzofuran
146
72918-21-9
1,2,3,7,8-Pentachlorodibenzofuran
139
57117-41-6
2,3,4,6,7,8-Hexachlorodibenzofuran
143
60851-34-5
2,3,4,7,8-Pentachlorodibenzofuran
138
57117-31-4
Octachlorodibenzofuran
133
39001-02-0
Table 4. Metals included in MOVES2014.
Pollutant
pollutantID
Mercury (elemental gaseous)
60
Mercury (divalent gaseous)
61
Mercury (particulate)
62
Arsenic compounds
63
Chromium (Cr6+)
65
Manganese compounds
66
Nickel compounds
67
This report has gone through two revisions since first released for MOVES2014 (EPA-420-R-14-
0213). The first revision of this report4 was made to document corrections made to the toxics
emission rates in MOVES2014a. MOVES2014a corrected the values for five of the dioxin &
furan emission rates and for the benzene diesel refueling emission rate. MOVS2014a enables
toxics to be calculated from pre-2001 model year E85 vehicles, and to be calculated from
evaporative and refueling emissions from all model years of E85 vehicles. MOVES2014a also
corrected the assignment of toxic ratios from diesel auxiliary power unit (APU) exhaust. This
report is the second revision, which we released with the 'MOVES2014a November 2016 patch',
which changed the units of the dioxin emission rates to grams, as discussed in Section 2.4.
5
-------
1.1 Methods
Toxics are emitted through exhaust, crankcase and evaporative processes, and by both light-duty
and heavy-duty vehicles, operating on gasoline, diesel and compressed natural gas (CNG) fuels.
While MOVES attempts to estimate emissions from vehicles representing relevant combinations
of technology and fuel, the availability and quality of data acquired and used varied widely.
Consequently, the methods and approaches used to develop model inputs also varied as
necessary.
During model runs, emissions of toxic compounds (except for metals and dioxins/furans), are
estimated as fractions of the emissions of VOC, or for toxic species in the particulate phase,
fractions of total organic carbon <2.5 [j,m (OC2.5) Emissions of VOC are themselves calculated
from emissions of total hydrocarbon (THC). All toxic fractions are mass-based (as opposed to
using molar-ratios).
For some compounds, the toxic emissions are estimated using fractions that vary with levels of
other fuel properties, such as ethanol, aromatics or Reid Vapor Pressure (RVP). Fractions that
vary according to fuel properties are termed "complex" by MOVES. For other sets of
compounds, "simple" fractions are used, meaning that the fractions are constants and do not vary
with fuel properties. Note that the generalizations made here apply to evaporative as well as to
exhaust emissions. In addition, in some cases, available data were sufficient to model emission
as a function of two different combustion processes, e.g., start and running exhaust emissions.
However, in other cases, available data were not adequate for this purpose, with the result that
single sets of inputs are used to represent both start and running emissions. Similarly, for
evaporative emissions, inputs were developed so as to distinguish "permeation" and "non-
permeation" processes. Finally, fractions vary with level of emission control (e.g. pre-Tier 2
versus Tier 2), and for old vehicles, catalyst type and fuel delivery system.
The approach differs for estimation of emissions of metals and dioxin/furans. These species are
estimated directly through application of emission rates that are assumed to be independent of
operating mode. Rates for metals and dioxins/furans are expressed on a distance-specific basis
(g/mile).
It should be noted that metals and dioxin emission rates are only produced from the 'running'
exhaust emission process with the g/mile rates. We do not estimate their emissions explicitly
from other exhaust emission processes such as start, extended idle, auxiliary power unit usage,
and crankcase processes. In fact, for extended idle, auxiliary power unit usage, and crankcase
emissions we do not have data on these emissions. However, in some cases the start emissions
for these pollutants are included in the driving cycle used to derived distance-based emision
factors as discussed in the report.
Finally, a uniform approach was used to develop single sets of inputs to estimate emissions of
toxics from gasoline fuels containing ethanol at levels of 70-100 volume percent (vol.%). The
data used for this purpose were typically measured on "E85" blends, containing 70-85 vol.%
ethanol.
It is important to note that the inputs used to estimate emissions of toxics do not vary by
temperature, i.e., the ambient temperature simulated during a run. However, inventories of toxic
compounds estimated by the model may vary by ambient temperatures for specific runs because
6
-------
VOC and OC2.5, do vary by temperature, and as described above, emissions of toxics compounds
are estimated as fractions of VOC or OC2.5 emissions.
1.2 Overview of the Report
The report first considers exhaust emissions from gasoline vehicles, covered in Section 2. The
data used to develop the emission rates are based on light-duty gasoline vehicles. However the
light-duty gasoline emission rates are applied to all gasoline vehicles, including motorcycles and
heavy-duty gasoline trucks. For volatile organic compound toxic emissions, the rates are derived
from two broad groups of gasoline vehicles, incorporating differences in vehicle technologies,
emission-control technologies and emissions standards, as well as subsets of available data and
analytic methods. These two groups are defined as "model year 2000 and earlier," and "model
year 2001 and later." The two technologies groups are used to distinguish emissions starting with
light-duty gasoline vehicles regulated under the National Low Emission Vehicle (NLEV)
program, which began with 2001 model year vehicles, followed by the Tier 2 Light-duty vehicle
emission standards, which began with 2004 model year vehicles.
For other toxic emissions from gasoline vehicles (PAHs, metals, and dioxins), we estimated
fleet-average toxic emission ratios, with no distinction for vehicle technology or model year, as
discussed in Sections 2.2, 2.3, and 2.4.
Next, the report considers exhaust emissions from diesel vehicles, covered in Sections 3 and 4.
The development of inputs for diesel vehicles are defined as "pre-2007" and "model year 2007
and later" based on technology and emissions standards for heavy-duty vehicles. This distinction
is made because emission controls on 2007 and later engines have a substantial effect on
composition of emissions. In addition, due to a lack of applicable data, the toxic emission rates
developed from heavy-duty trucks are also used to represent light-duty diesel vehicles, as well as
diesel engines used as auxiliary power units, as noted in Section 3
Section 5 contains the derivation of the toxic emission rates for CNG-powered transit buses in
MOVES. At present, MOVES only models CNG fuel usage within transit buses. Toxic
emissions from evaporative emission processes and crankcase emission processes are addressed
in Section 6 and 7.
7
-------
2 Gasoline Exhaust
2.1 Volatile Organic Compounds
2.1.1 Vehicles Operating on Fuel Blends Containing 0-20% Ethanol
2.1.1.1 2000 and Earlier Model Year Vehicles
For three sets of compounds, Table 5 summarizes the methods used to estimate toxic fractions.
The specific data and methods used for each are described in further detail below.
Table 5. Calculation Methods for VOC
Compound
Fraction Type
Basis for Estimation
Benzene
complex
Complex Model
1,3-Butadiene
complex
Complex Model
Acetaldehyde
complex
Complex Model
Formaldehyde
complex
Complex Model
Methyl-tert-butyl ether
complex
Derived from Complex Model Database
2,2,4-Trimethylpentane
Simple
SPECIATE profile
Acrolein
Simple
SPECIATE profile
Ethylbenzene
Simple
SPECIATE profile
n-Hexane
Simple
SPECIATE profile
Propionaldehyde
Simple
SPECIATE profile
Styrene
Simple
SPECIATE profile
Xylene(s)
Simple
SPECIATE profile
Ethanol
Simple
4 test programs outlined in Section 2.1.1.1.4
2.1.1.1.1 Use of Equations Developedfor the Complex Model
For the first four compounds listed in Table 5, "complex" toxic fractions of VOC were estimated
through application of equations developed for the Complex Model for Reformulated Gasoline.5
The equations are based on about 1,800 observations collected on vehicles equipped with three-
way or three-way-plus-oxidation catalysts.51 The equations were developed by stratifying the
light-duty gasoline fleet into ten technology groups and fitting statistical models to subsets of
data for each group. The resulting sets of equations are known collectively as the
"unconsolidated Complex Model." The ten groups were assigned as combinations of fuel system,
catalyst type, air injection (yes/no), exhaust-gas recirculation (EGR), and normal/high emitter
status. The first nine groups were intended to represent only "normal-emitting" vehicles. The
tenth group represents all "high emitters," regardless of technology. In application, the equations
are consolidated by weighting them together using model-year specific weights based on the mix
of technologies in the sales fleet for each model year, as obtained from MOBILE6.
a While more recent emissions data are available for Tier 1 and earlier vehicles, such as data from the Kansas test program
mentioned earlier, testing was not done on a matrix of fuels which enable development of a fuel effects model.
8
-------
The Complex Model equations are applied to running, start and extended idle emissions for
gasoline-fueled vehicles for all 2000 and earlier model years for the first four pollutants listed in
Table 5 (acetaldehyde, formaldehyde, benzene and 1,3-butadiene). While MOBILE6 applied
separate equations for older technologies not included in the Complex Model, such as vehicles
without catalysts or vehicles equipped only with oxidation catalysts, these equations were not
included in MOVES since these vehicles now comprise an extremely small and ever shrinking
portion of the fleet. For 1974 and earlier model years, 1975 weightings are used. In addition,
while MOBILE6.2 relied on very limited data from heavy-duty gasoline vehicles, MOVES
applies Complex Model effects to both light-duty and heavy-duty gasoline vehicles. This step
was taken because the very limited data specific to heavy-duty gasoline vehicles are not adequate
to account for effects of fuel properties
2.1.1.1.2 Overview of the Complex Model
The Complex Model is so called because it was designed to model the "complex" behavior of
selected emissions in relation to changes in a set of selected fuel properties.
The underlying dataset included measurements collected on sample of vehicles manufactured in
model year (MY) 1990 or earlier, and reflecting "Tier 0" standards over a variety of gasoline
formulations.
The Complex Model is composed of sets of models for each pollutant. The models are statistical
models fit to sets of emissions measurements on a set of fuels with widely varying properties.
For each pollutant, 10 models were fit, with each representing a specific combination of fuel-
delivery, catalyst, air injection and emissions-control technology. The technology groups are
described in Table 6. As an aggregate, these sets of models are referred to as the "unconsolidated
Complex Model."
In fitting the Complex Model, the measurements for all fuel properties were "centered," meaning
that the mean of all measurements for the property was subtracted from each individual
measurement. This step aids in scaling the dataset so that each fuel property is centered on a
mean of 0.0. Thus, if ln7 is the natural logarithm of a specific compound, such as acetaldehyde,
the model is fit as shown in Equation 1, using terms for oxygenate (wt.%), aromatics (vol.%) and
RVP (psi) as examples.
In Y = p0 + /?oxy (x,i::y; - xoxy ) + - *arom )+••• + /?RVP (xRVP-i - xRVP ) Equation 1
The mean values used for centering all individual fuel-property values are presented in Table 7.
Sets of coefficients (fi values in Equation 1) for models by technology group are presented for
acetaldehyde, formaldehyde, benzene and 1,3-butadiene in Table 8 to Table 11. Dashes in table
cells indicate no coefficient was fit for that property. It should be noted that the sulfur effects
terms in the original Complex Model were not included when the model was adapted for
inclusion in MOVES; rather, sulfur effects on toxic emissions are assumed to be proportional to
the effects of sulfur on total VOC, as estimated by MOVES.
9
-------
Table 6. Technology Groups included in the Complex Model.
Technology Group
Fuel System1
Catalyst2
Air Injection
Exhaust-gas
Recirculation
1
PFI
3-Way
No
Yes
2
PFI
3-Way
No
No
3
TBI
3-Way
No
Yes
4
PFI
3-Way + Oxy
Yes
Yes
5
PFI
3-Way
Yes
Yes
6
TBI
3-Way
Yes
Yes
7
TBI
3-Way + Oxy
Yes
Yes
8
TBI
3-Way
No
No
9
carburetor
3-Way + Oxy
Yes
Yes
10 ("High Emitters")
ALL
ALL
ALL
ALL
1 Fuel System: PFI = port fuel injection, TBI = throttle body injection.
2 Catalyst type: "3-way" = three-way catalyst, "Oxy" = oxidation catalyst.
Table 7. Mean Fuel-Property Values used for Centering Terms in the Complex Model.
Property
Units
Mean Value
Aromatics
Vol. %
28.26110
Olefins
Vol. %
7.318716
Methyl-tertiary-butyl-ether (MTBE)1
Wt.%
0.947240
Ethyl-tertiary-butyl-ether (ETBE)1
Wt.%
0.023203
Ethanol (EtOH)1
Wt.%
0.314352
Tertiary-amyl-methyl-ether (TAME)1
Wt.%
0.016443
Oxygenate2
Wt.%
1.774834
RVP
Psi
8.611478
E200
%
46.72577
E300
%
85.89620
1 Species-specific values used in the aldehyde models.
2 Aggregate value used for the butadiene and benzene models.
10
-------
Table 8. Complex Model Coefficients for Acetaldehyde, by Technology Group.
Technology Group
Fuel Property
75
O
"o3
a
2
is
u
w
m
W
m
H
Q
<
RVP
o
o
(N
o
o
m
<
U
w
W
H
W
W
1
-0.05548
-
-0.03646
0.316467
0.249326
-
-
-
-0.01216
2
-0.05548
-
-
0.316467
0.249326
-
-
-
-0.01216
3
-0.05548
-
-
0.316467
0.249326
-
-
-
-0.01216
4
-0.05548
-
-
0.316467
0.249326
-
0.24230
-
-0.01216
5
-0.05548
-
-
0.316467
0.249326
-
-
-
-0.01216
6
-0.05548
-
-
0.316467
0.249326
-
-
-
-0.01216
7
-0.05548
-
-
0.316467
0.249326
-
-
-
-0.01216
8
-0.05548
-
-
0.316467
0.249326
-
-
-
-0.01216
9
-0.05548
-
-
0.316467
0.249326
-
-
-
-0.01216
10
-0.05548
-
-0.05598
0.316467
0.249326
-
-
-
-0.01216
Table 9. Complex Model Coefficients for Formaldehyde, by Technology Group.
Technology Group
Fuel Pro
perty
Aromatics
Olefins
MTBE
ETBE
EtOH
TAME
RVP
E200
E300
1
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
2
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
3
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
4
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
5
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
6
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
7
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
8
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
9
-0.00717
-
0.046213
-
-
-
-
-
-0.01023
10
-0.00717
-0.03135
0.046213
-
-
-
-
-
-0.01023
11
-------
Table 10. Complex Model Coefficients for Exhaust Benzene, by Technology Group.
Technology Group
Fuel Property
Aromatics
Olefins
Oxygenate
Fuel
Benzene
RVP
E200
E300
1
0.02588
-
-
0.222318
-
-0.00948
-
2
0.02588
-
-
0.222318
-
-
-
3
0.02588
-
-
0.222318
-
-0.00578
-
4
0.02588
-
-
0.222318
-
-
-
5
0.04859
-
-
0.222318
-
-
-
6
0.02588
-
-
0.222318
-
-
-
7
0.02588
-
-
0.222318
-
-
-
8
-
-
0.222318
-
-
-
9
0.02588
-
-
0.222318
-
-
-
10
0.01188
-
-0.09605
0.222318
-
-
0.011251
Table 11. Complex Model Coefficients for 1,3-Butadiene, by Technology Group.
Technology Group
Fuel Property
Aromatics
Oxygenate
Olefins
E200
E300
1
-0.00401
-
0.028238
-0.00731
-0.01678
2
-0.00401
-
0.028238
-0.00731
-0.01678
3
-0.00401
-
0.028238
-0.00731
-0.00625
4
-0.00401
-
0.028238
-0.00731
-0.01678
5
-0.00401
-
0.028238
-0.00731
-0.01678
6
-0.00401
-
0.028238
0.005786
-0.01678
7
-0.00401
-
0.028238
-0.00731
-0.01678
8
-0.00401
-
0.028238
-0.00731
-0.01678
9
-0.00401
-
0.028238
-0.00731
-0.01678
10
-0.00401
-0.06077
0.043696
-0.00731
-0.00806
For each compound, the model equations as shown in Equation 1, are evaluated for a "base" and
a "target" fuel. We assume that vehicles were running on a specific fuel when the data
underlying the base emission rates were measured. We refer to these fuels as "base" fuels and
use them as reference points to estimate the effects of "target" fuels simulated during MOVES
runs.21 The "target" fuels are represented by specific sets of properties and represent fuels "in-
use" in the geographic area(s) and season(s) being modeled in MOVES.
Initially, an adjustment for the difference in emissions of the compound modeled on the target
fuel relative to the base fuel is calculated. If the model, as shown in Equation 1, can be
conveniently expressed, using matrix notation, as XPtarget and XPbase for estimates on the target
and base fuels, then the fractional difference in emissions is given by
fp(XP target)
exr
adj
ex
P(XPbaJ
-1.0
Equation 2
12
-------
The expression in Equation 2 is evaluated for target and base fuels for each of the ten technology
groups. A mean value of the adjustment is then calculated for each model year from 2000 back to
1970, as a weighted average of the fraction of sales in each group in each model year, for the
groups, as shown in Equation 3. The weights are shown in Table 12. The weights represent the
sales fractions for the ten vehicle technologies defined in Table 6 above.
Note that the use of varying weights in applying the Complex Model in MOVES differs from the
original application in which the weights were invariant. The application of Equation 3 to each of
the 30 ages listed in the table gives a set of 30 adjustments, with each applied to a single model
year, which represents a specific age with respect to the calendar year simulated.
10 10
./dj,mean = E ^Group/ad,Group i E ^ Group =^0 Equation 3
Group=1 Group=1
The mean adjustments calculated in Equation 3 are then applied to estimate emissions of the
toxic on the target fuel (/relative,toxic), representing the effect on the emissions of the toxic due to
the changes in fuel properties between the target and base fuels. If the target and base fuels were
identical, the values of/adj ,mean would be 0.0.
"^relative,toxic "^base, toxic 1 ^ -^adj.mean ) I',(| llil 11()I1 4
The calculations in Equation 1 to Equation 4 are also applied to VOC emissions, ending with the
generation of a value of /^eiativc.voc. This value for VOC is then combined with that for each
toxic to calculate a fraction of VOC used to estimate the total mass of emissions for each toxic
during a model run. These fractions are denoted as /toxic and calculated as shown in Equation 5.
f
, = relative,toxic Equation 5
J toxic TP
relative,VOC
As a final step, the mass emissions of each toxic (/toxic) during a model run are estimated by
multiplying the mass of VOC emissions estimated by MOVES (/voc ) by the values of/oxic.
Zoxic = /ox,c/voc Equation 6
The equations and parameters presented are used to estimate the fuel impacts for both Tier 0 and
Tier 1 gasoline vehicles. This approach is based on the assumption that the proportional
responses of air toxic emissions to changes in fuel properties are similar for vehicles certified to
both sets of standards.
13
-------
Table 12. Weights Applied to Complex Model coefficients for Technology Groups, by Age (Vehicle Age 0
represents model year 2000).b
Age
Technology Group
1
2
3
4
5
6
7
8
9
10
0
0.2360
0.2829
0.1806
0.1814
0.0290
0.0042
0.0556
0.0
0.0203
0.0100
1
0.2339
0.2803
0.1789
0.1797
0.0287
0.0042
0.0551
0.0
0.0201
0.0190
2
0.2315
0.2774
0.1771
0.1779
0.0284
0.0041
0.0546
0.0
0.0199
0.0290
3
0.2272
0.2723
0.1738
0.1746
0.0279
0.0041
0.0536
0.0
0.0196
0.0470
4
0.2229
0.2672
0.1706
0.1713
0.0274
0.0040
0.0525
0.0
0.0192
0.0650
5
0.2189
0.2623
0.1675
0.1682
0.0269
0.0039
0.0516
0.0
0.0188
0.0820
6
0.2148
0.2574
0.1644
0.1651
0.0264
0.0038
0.0506
0.0
0.0185
0.0990
7
0.2110
0.2529
0.1614
0.1621
0.0259
0.0038
0.0497
0.0
0.0182
0.1150
8
0.2072
0.2483
0.1585
0.1592
0.0254
0.0037
0.0488
0.0
0.0178
0.1310
9
0.2036
0.2440
0.1558
0.1565
0.0250
0.0036
0.0480
0.0
0.0175
0.1460
10
0.2000
0.2397
0.1530
0.1537
0.0246
0.0036
0.0471
0.0
0.0172
0.1610
11
0.1967
0.2357
0.1505
0.1512
0.0241
0.0035
0.0464
0.0
0.0169
0.1750
12
0.1934
0.2317
0.1479
0.1486
0.0237
0.0035
0.0456
0.0
0.0166
0.1890
13
0.1903
0.2280
0.1456
0.1462
0.0234
0.0034
0.0448
0.0
0.0164
0.2020
14
0.1872
0.2243
0.1432
0.1438
0.0230
0.0033
0.0441
0.0
0.0161
0.2150
15
0.1843
0.2209
0.1410
0.1416
0.0226
0.0033
0.0434
0.0
0.0159
0.2270
16
0.1814
0.2174
0.1388
0.1394
0.0223
0.0032
0.0428
0.0
0.0156
0.2390
17
0.1786
0.2140
0.1366
0.1372
0.0219
0.0032
0.0421
0.0
0.0154
0.2510
18
0.1760
0.2109
0.1346
0.1352
0.0216
0.0031
0.0415
0.0
0.0151
0.2620
19
0.1736
0.2080
0.1328
0.1334
0.0213
0.0031
0.0409
0.0
0.0149
0.2720
20
0.1712
0.2052
0.1310
0.1315
0.0210
0.0031
0.0403
0.0
0.0147
0.2820
21
0.1688
0.2023
0.1291
0.1297
0.0207
0.0030
0.0398
0.0
0.0145
0.2920
22
0.1664
0.1994
0.1273
0.1279
0.0204
0.0030
0.0392
0.0
0.0143
0.3020
23
0.1643
0.1969
0.1257
0.1262
0.0202
0.0029
0.0387
0.0
0.0141
0.3110
24
0.1624
0.1946
0.1242
0.1248
0.0199
0.0029
0.0383
0.0
0.0140
0.3190
25
0.1602
0.1920
0.1226
0.1231
0.0197
0.0029
0.0378
0.0
0.0138
0.3280
26
0.1602
0.1920
0.1226
0.1231
0.0197
0.0029
0.0378
0.0
0.0138
0.3280
27
0.1602
0.1920
0.1226
0.1231
0.0197
0.0029
0.0378
0.0
0.0138
0.3280
28
0.1602
0.1920
0.1226
0.1231
0.0197
0.0029
0.0378
0.0
0.0138
0.3280
29
0.1602
0.1920
0.1226
0.1231
0.0197
0.0029
0.0378
0.0
0.0138
0.3280
30
0.1602
0.1920
0.1226
0.1231
0.0197
0.0029
0.0378
0.0
0.0138
0.3280
2.1.1.1.3 Estimating Emissions of Methyl-tertiary-butyl-ether (MTBE)
As of calendar year 2008, MTBE (pollutantID = 22) has been almost completely phased-out of
the fuel supply in the United States due to concerns related to contamination of ground water.
Thus, its inventory levels as predicted by MOVES based on default inputs should be very small
if not zero in future years. It is presently in the MOVES model as a legacy pollutant for calendar
b Note that in the MOVES database, these weights are stored in the table FuelModelWtFactor.
14
-------
years 1990 and 1999 - 2005c. However, the MTBE fuel volume is a user input, and MOVES has
the capability to calculate MTBE emissions for any calendar year.
For MTBE, a fuel-effects model based on the Complex Model database and applied in
MOBILE6.2 was used.6'7 This model is based on equations fit to data representing nearly 900
observations. However, instead of using model equations directly, MOBILE6.2 was run at
different fuel MTBE volumes (Fmtbe). Using the results of the MOBILE6.2 runs, the MTBE
fractions of VOC were calculated and related to MTBE fuel levels using a simple least-squares
regression. A quadratic equation fixed at the origin was selected, and gives results consistent
with the original parameterization in MOBILE6.2. The parameters are shown in Table 13. The
same equation is used for both start and running processes and is shown in Equation 7.
{ — AV I m/2 Equation 7
./MTBE - MTBE ~r MTBE
The coefficients A and B take the values shown in Table 13. As with the other toxic emissions,
the fraction/mtbe is multiplied by the mass of VOC to estimate MTBE emissions, as shown in
Equation 6.
Table 13. Exhaust Calculation Coefficients for MTBE (see Equation 7 ).
Pollutant Process
polProcessID
A (coeffA)
B (coeffB)
Running Exhaust
2201
0.00007809
0.00007537
Start Exhaust
2202
0.00007809
0.0007809
Data were not available to develop emission effects for ETBE and TAME blends; thus, the
equations for ethanol-oxygenated gasoline were used for ETBE blends, and those for MTBE-
oxygenated gasoline were used for TAME blends.
2.1.1.1.4 Simple Fractions of VOC
Table 14 lists toxic fractions of VOC for a set of additional compounds designed to represent
toxic emissions for several fuel blends containing different oxygenates. With the exception of
ethanol, for gasoline fuels containing 0 and 10% ethanol (E0 and E10), fractions were developed
by Sierra Research using speciation profiles estimated from EPA's SPECIATE 4.2 database.8
The fractions for E10 are also used to represent blends in which the oxygenate is ethyl-tertiary-
butyl-ether (ETBE) at levels of 5 vol.% or greater.
For blends containing methyl-tertiary-butyl ether (MTBE), however, fractions were adopted
from the National County Database for the National Mobile Inventory Model (NMIM). The
fractions used in NMIM were derived for the 1999 National Emission Inventory (NEI) for
hazardous air pollutants (HAPS), version 3, and summarized in Volume 1, Appendix D, Table 1
of the documentation. These fractions were based on older speciation profiles than the E0 and
E10 data. One set of fractions represents winter fuels containing MTBE at 12 vol. % or greater,
or tertiary-amyl-methyl-ester (TAME) at levels of 13% or more (winter). A second set
represents reformulated gasoline fuels containing MTBE at levels between 5.0 and 13.0 vol.% or
TAME at levels between 5.0 and 13.0 vol.% (RFG). These fractions are provided in Table 15.
c MOVES does not currently explicitly model calendar years 1991-1998
15
-------
Emissions of ethanol in exhaust are estimated for gasoline blends containing ethanol at levels of
0 to 10 vol.%. For vehicles running on 10% ethanol, ethanol was estimated to comprise 2.39% of
exhaust VOC. This estimate is based on results measured on nine vehicles in four test
programs.9'10' u'12 The fraction of ethanol in exhaust VOC for blends containing 5.0% and 8.0%>
ethanol is estimated by interpolating linearly between the fractions for 0.0%> and 10.0%> ethanol.
No data exist for 2000 and earlier vehicles running on El 5 or E20. These emissions comprise a
minor fraction of the inventory, as conventional vehicles do not have an EPA waiver to operate
on ethanol fractions higher than 10%>13, and flex-fuel vehicles were manufactured in only the
1999 and 2000 model years. For pollutantlDs 40 - 46, we used toxics ratios for 2001 and later
vehicles, found in Table 41. For acrolein and ethanol, we simply extended the E10 toxic fractions
as shown in Table 14.
Table 14. Toxic Fractions of VOC for Selected Air Toxics, Representing Gasoline and Ethanol Blends.
Compound
pollutantID
Fuel Blend (by Ethanol Level)
0% (E0)
10% (E10)
15%
(E15)
20%
(E20)
Ethanol
21
0
0.0239
0.0239
0.0239
Acrolein
27
0.000628
0.000628
0.000628
0.000628
2,2,4-
T rimethylpentane
40
0.01823
0.01849
Table 41
Ethyl Benzene
41
0.02147
0.01932
Hexane
42
0.01570
0.01593
Propionaldehyde
43
0.00086
0.00086
Styrene
44
0.00108
0.00097
Toluene
45
0.09619
0.08657
Xylene
46
0.07814
0.07032
Table 15. Toxic Fractions for Selected Air Toxics of VOC, Representing Gasolines containing MTBE.
Compound
pollutantID
MTBE
Winter
RFG
Acrolein
27
0.0006
0.0006
2,2,4-Trimethylpentane
40
0.04327
0.04327
Ethyl Benzene
41
0.01398
0.01484
n-Hexane
42
0.00861
0.00888
Propionaldehyde
43
0.00073
0.00073
Styrene
44
0.00328
0.00340
Toluene
45
0.09873
0.10494
Xylene
46
0.05557
0.05910
16
-------
In the MOVES database, these inputs are stored in the table "minorHAPratio." In the label, the
term "HAP" refers to "hazardous air pollutant." A description of the table is provided in Table
16.
Table 16. Description of the Database Table "minorHAPRatio."
Field
Description
RelevantValues
polProcessID
Identifies the pollutant (1st two
digits and Emissions Process
(last two digits).
Pollutants are identified in the table above;
Relevant processes include:
"Running Exhaust" (processID =1)
"Start Exhaust" (processID = 2)
fuelTypelD
Identifies broad classes of fuels,
e.g., "gasoline." "diesel."
1 = "Gasoline"
2 = "Diesel"
5 = "Ethanol"
fuelSubTypelD
Identifies specific fuel classes
within the fuelTypelD
10 = "Conventional Gasoline"
11 = "Reformulated Gasoline"
12 = "Gasohol (E10)"
13 = "Gasohol (E8)"
14 = "Gasohol (E5)"
15 = "Gasohol (E15)"
18 = "Gasohol (E20)"
51 = "Ethanol (E85)"
52 = "Ethanol (E70)"
modelY earGroupID
Identifies a set of model years
covered by a specific value of
atRatio.
1960-1970
1971-1977
1978-1995
1996-2003
2004-2050
atRatio
Fraction, or "ratio" of the toxic
relative to total VOC.
atRatioCV
"Coefficient of Variation of the
Mean" or "relative standard
error" of the atRatio.
dataSourcelD
Indicates source data and
methods used to estimate
atRatio.
2.1.1.2 2001 and later model year vehicles
For vehicles manufactured in MY2001 and later, and certified to NLEV or Tier 2 standards,
recently-collected data were available. As before, toxic emissions are estimated as fractions of
VOC, with toxic fractions for various compounds estimated using differing datasets and
methods. For some compounds and processes, models were developed to estimate "complex"
fractions (responding to fuel properties), whereas for others, "simple" fractions were estimated
(not responding to fuel properties). An additional feature for these fractions is that in some cases,
different fractions could be estimated for the start and running emission processes. For the
compounds included in MOVES, data sources and estimation methods are summarized in Table
17.
17
-------
Table 17. Data Sources and Estimation Methods Used in Estimation of Toxic Fractions for VOCs
Compound
Process
Fraction Type
Basis for Estimation
Acetaldehyde
Start
complex
application of EPAct models1
Running
complex
application of EPAct models
Formaldehyde
Start
complex
application of EPAct models
Running
complex
application of EPAct models
Acrolein
Start
complex
application of EPAct models
Running
simple
Data from EPAct Project (Phase 3)2
Ethanol
Start
complex
application of EPAct models
Running
complex
application of EPAct models
Benzene
Start
complex
application of EPAct models
Running
simple
Data from EPAct Project (Phase 3)
1,3 -Butadiene
Start
complex
application of EPAct models
Running
simple
Data from EPAct Project (Phase 3)
2,2,4-Trimethylpentane
Both
simple
Speciation Profile (EPAct Phase l)3
Ethylbenzene
Both
simple
Speciation Profile (EPAct Phase 1)
N-Hexane
Both
simple
Speciation Profile (EPAct Phase 1)
Propionaldehyde
Both
simple
Speciation Profile (EPAct Phase 1)
Styrene
Both
simple
Speciation Profile (EPAct Phase 1)
Xylene(s)
Both
simple
Speciation Profile (EPAct Phase 1)
1 Derived from models fit to data from EPAct Phase 3 Results.
2
Derived from data collected in EPAct Phase 3.
3
Derived from data collected in EPAct Phase 1.
2.1.1.2.1 Application of the Results of the EPAct Program
Since the initiation of the MOVES project, it was clear that application of the Complex Model to
2001 and later vehicles, as in MOVES 2010b and MOBILE6.2, was no longer appropriate.
Thus, an updated fuel-effects model representing Tier-2 certified vehicles was needed. To meet
this goal, EPA entered a partnership with the Department of Energy (DOE) and the Coordinating
Research Council (CRC) to undertake the largest fuels research program conducted since the
Auto/Oil program in the early 1990's, aimed specifically at understanding the effects of fuel
property changes on exhaust emissions on recently manufactured Tier 2 vehicles. The resulting
research program was dubbed the "EPAct/V2/E-89" program (or "EPAct" for short), with the
three components of the label denoting the designation given to the study by the EPA, DOE and
CRC, respectively.
The program was conducted in three phases. Phases 1 and 2 were pilot efforts involving
measurements on 19 light-duty cars and trucks on three fuels, at two temperatures. These
preliminary efforts laid the groundwork for design of a full-scale research program, designated as
Phase 3.
Initiated in March 2009, the Phase 3 program involved measurement of exhaust emissions from
fifteen high-sales-volume Tier-2 certified vehicles. The vehicles were selected so as to represent
the latest technologies in the market at the time the program was launched (2008). The vehicles
18
-------
were to reflect a majority of sales for model year 2008. In addition, the vehicles were to conform
primarily to Tier-2 Bin-5 exhaust standards, and to reflect a variety of emission-control
technologies, as realized through the selection of a range of vehicle sizes and manufacturers. The
vehicle sample is summarized in Table 18.
Table 18. Test Vehicles for the Phase-3 EPAct Program (all vehicles in MY2008).
Make
Brand
Model
Engine Size
Tier 2 Bin
LEVTI
Std
Odometer
GM
Chevrolet
Cobalt
2.2L 14
5
NA
4,841
GM
Chevrolet
Impala FFV
3.5L V6
5
L2
5,048
GM
Saturn
Outlook
3.6L V6
5
L2
5,212
GM
Chevrolet
Silverado FFV
5.3L V8
5
NA
5,347
Toyota
Toyota
Corolla
1.8L 14
5
U2
5,019
Toyota
Toyota
Camry
2.4L 14
5
U2
4,974
Toyota
Toyota
Sienna
3.5L V6
5
U2
4,997
Ford
Ford
Focus
2.0L 14
4
U2
5,150
Ford
Ford
Explorer
4.0L V6
4
NA
6,799
Ford
Ford
F150 FFV
5.4L V8
8
NA
5,523
Chrysler
Dodge
Caliber
2.4L 14
5
NA
4,959
Chrysler
Jeep
Liberty
3.7L V6
5
NA
4,785
Honda
Honda
Civic
1.8L 14
5
U2
4,765
Honda
Honda
Odyssey
3.5L V6
5
U2
4,850
Nissan
Nissan
Altima
2.5L 14
5
L2
5,211
The study used a total of twenty-seven test fuels spanning wide ranges of five fuel properties
(ethanol, aromatics, vapor pressure, and two distillation parameters: T50 and T90). The numbers
of test points and values of each property are shown in Table 19. The properties of the test fuels
were not assigned to represent in-use fuels, but rather to allow development of statistical models
that would enable estimation of relative differences in emissions across the ranges of fuel
properties expected in commercially available summer fuels in the U.S. (5111 to 95111 percentiles
for each property).
Table 19. Levels Assigned to Experimental Factors (Fuel parameters) for the Phase-3 EPAct program.
Factor
No. Levels
Levels
Low
Middle
High
Ethanol (vol.%)
4
0
10, 15
20
Aromatics (vol.%)
2
15
35
RVP (psi)
2
7
10
T50 (°F)
5
150
165, 190, 220
240
T90 (°F)
3
300
340
The LA92 test cycle was used with emissions measured over three phases analogous to those in
the Federal Test Procedure (FTP), at an ambient temperature of 75°F. Note that throughout this
chapter, the terms "start," "cold start" and "Bag 1" will be treated as synonymous, and similarly,
the terms "running," "hot-running" and "Bag 2" will also be treated as synonymous.
The experimental design embodied in the fuel set is the product of an iterative process involving
balancing among research goals, fuel-blending feasibility and experimental design. As fuel
properties tend to be moderately to strongly correlated, and as the goal was to enable analysis of
fuel effects as though the properties were independent (uncorrected), it was necessary to address
19
-------
these issues in design and analysis. Accordingly, the fuel set was designed using a computer-
generated optimal design, as modified by additional requirements such as the total number of
fuels and specific properties for subsets of fuels. In addition, to generate the design, it was
necessary to specify the fuel effects to be estimated by the resulting model. The fuel set was
designed to allow estimation of linear effects for the five properties shown in Table 19, plus two-
way interactions of ethanol and the other five properties, as shown in Equation 8, in which
represents a linear coefficient for each effect.
Y = J30 + /?,etOH + fi2 Arom + J33 RVP + J3J50 + J35 T90 +
/?6T502 +/?netOH2 Equadon8
/?7etOH x Arom + /?setOH x RVP + /?,etOH x T50 + /?„,etOH x T90 +
In the equation, the linear terms (e.g., /?ietOH, etc.) describe linear associations between
emissions (7) and the value of the fuel property. The quadratic terms are used to describe some
degree of curvature in the relationship between emissions and the fuel property. Note that a
minimum of 3 test levels for a property is needed to assess curvilinear relationships and that the
design included such effects only for ethanol and T50. Two-way interaction terms indicate that
the relationship between emissions and the first fuel property is dependent on the level of the
second fuel property. For example, if an etOHxArom interaction is included in a model, it
implies that the effect of ethanol on the emission Y cannot be estimated without accounting for
the aromatics level, and vice versa. Note that inclusion of the 11 effects in the design does not
imply that all effects will be retained in all models following the fitting process. Properties for
each of the test fuels are shown in Table 20.
Emissions measured include carbon dioxide (CO2), carbon monoxide (CO), THC, methane
(CH4), oxides of nitrogen (NO,;), and PM2.5. In addition, hydrocarbons were speciated for
subsets of vehicles and fuels, allowing calculation of derived parameters such as non-methane
organic gases (NMOG) and non-methane hydrocarbons (NMHC). Speciation also allowed
independent analyses of selected toxics including acetaldehyde, formaldehyde, acrolein,
benzene, 1,3-butadiene and ethanol.
Due to limitations in budget, the entire study design was not applied to speciated hydrocarbons,
including those discussed in this chapter. For the speciated compounds, the volume of data
collected varies by Bag, compound and vehicle. For selected compounds, measurements for Bag
1 were taken for all vehicles over the entire fuel set, thus encompassing the entire study as
designed, including replication. However, for the remaining compounds in Bag 1 and for all
compounds in Bags 2, measurements were taken for a smaller number of vehicles over a reduced
set of fuels, without replication. The combinations of fuels and vehicles included for each
compound analyzed are summarized in Table 21.
Throughout this chapter, the complete set of 27 fuels will be denoted as the "full design," as it
includes all the fuel parameter points for which the design was optimized. Similarly, the set of 11
fuels will be denoted as the "reduced design," as it covers a set of fuel parameter points narrower
than that for which the design was originally optimized. Note that Table 20 also identifies the
subset of fuels included in the reduced design.
Phase 3 data collection was completed in June 2010. Dataset construction and analysis was
conducted between January 2010 and November 2012. This process involved ongoing
20
-------
collaboration among EPA staff, DOE staff and contractors, and CRC representatives. Following
the completion of data collection, construction of the dataset involved intensive evaluation and
quality assurance. The analysis involved several iterations between analysis and additional
physical and chemical review of the data. Successive rounds of statistical modeling were applied
to the data to achieve several goals, including identification of potential candidate models,
identification and review of outlying observations, identification and review of subsets of data
from influential vehicles, and identification of models including subsets of terms that best
explain the results obtained. The EPAct exhaust research program and analysis are extensively
documented in the "EPAct Test Program Report14" and "EPAct Analysis Report.15"
This document describes how the data and statistical models developed during the EPAct study
are applied in the MOVES model (MOVES2014).
Table 20. Measured Parameters for Fuels in the Phase-3 EPAct Program
Fuel1
etOH (vol.%)
Aromatics (vol.%)
RVP (psi)2
T50 (°F)
T90 (°F)
1
10.03
15.4
10.07
148.9
300.2
2
0
14.1
10.2
236.7
340.1
33
10.36
15.0
6.93
217.5
295.9
4
9.94
15.5
10.01
221.9
337.5
5
0
34.7
6.95
237.0
300.0
63
10.56
15.0
7.24
188.5
340.4
73
0
17.0
7.15
193.1
298.4
8
0
15.7
10.2
221.1
303.1
9
0
35.8
10.30
192.8
341.8
103
9.82
34.0
7.11
217.1
340.2
11
10.30
35.0
9.93
189.3
298.6
12
9.83
34.8
10.13
152.2
339.8
133
0
34.1
6.92
222.5
337.9
143
0
16.9
7.14
192.8
338.5
15
0
35.3
10.23
189.7
299.4
16
10.76
35.6
7.12
218.8
300.6
20
20.31
15.2
6.70
162.7
298.7
213
21.14
35.5
7.06
167.6
305.0
22
20.51
15.0
10.21
163.2
297.3
233
20.32
15.9
6.84
162.5
338.2
24
20.51
15.3
10.12
165.1
338.1
25
20.03
35.2
10.16
166.9
337.9
26
15.24
35.6
10.21
160.3
338.7
273
14.91
14.9
6.97
221.5
340.3
283
14.98
34.5
6.87
216.6
298.8
30
9.81
35.5
10.23
152.9
323.8
313
20.11
35.5
6.98
167.3
325.2
1 Note that numbering of fuels is not entirely sequential throughout.
2 This parameter was measured as "DVPE," but for simplicity, will be referred to as "RVP" in this
document.
3 These fuels included in the "reduced design."
21
-------
Table 21. Features of the Study Design Applied to Speciated Compounds Selected for Analysis.
Compound
Bag 1
Bag 2
No. vehicles
No. Fuels
replication
No. vehicles
No. Fuels
replication
Acetaldehyde
15
27
YES
5
11
NO
Formaldehyde
15
27
YES
5
11
NO
Acrolein
15
27
YES
5
11
NO
Ethanol
15
27
YES
5
11
NO
Benzene
15
11
NO
5
11
NO
1,3-Butadiene
15
11
NO
5
11
NO
Ethane
15
11
NO
5
11
NO
2.1.1.2.2 Standardizing Fuel Properties
In model fitting, as well as in applying the resulting sets of coefficients, it is necessary to first
"center" and "scale" the properties of fuels, also known as "standardization." This process
simply involves first "centering" the measured fuel properties by subtracting the sample mean
from the given value, and then "scaling" by then dividing the centered values by their respective
standard deviations, as shown in Equation 9. Note that the means and standard deviations are
calculated from the fuel set used for the program (see Table 20). The result is a "Z score,"
representing a "standard normal distribution" with a mean of 0.0 and a standard deviation of 1.0.
y — xj ~ * Equation 9
For the linear effects in the model, standardization is performed using the values of each fuel
property, each in their respective scales (vol. %, psi, °F.). Using aromatics as an example, the
standardization of the linear term is shown in Equation 10.
^ _ ^arom—^arom_ Equation 10
arom £
c
arom
For second-order terms, however, the process is not performed on the values of the fuel
properties themselves. Rather, quadratic and interaction terms are constructed from the Z scores
for the linear terms, and the process is repeated. Using the quadratic term for ethanol as an
example (etOHxetOH), the standardized value, denoted by ZZetoHxetoH, is calculated as shown in
Equation 11, where mz«oHZe,oH and fe.oHZe.oH are the mean and standard deviation of the quadratic
term constructed from the Z score for the linear effect.
ZZetOHxetOH = ZetQHZetOH ~ ^ZetoHZetoH Equation jj
$z z
A etOHxetOH
Standardized terms for interaction effects are constructed similarly. For example, Equation 12
shows the standardization of an interaction term between ethanol and aromatics.
22
-------
ZZ n„ A = ZetollZAmm—^etPH^rom Equation 12
etOHxeArom "
^Z Z
¦^etOH^Arom
Means and standard deviations for relevant model terms are shown in Table 22. Note that the
means and standard deviations shown in the table are calculated from the fuel set itself as shown
in the table; in this calculation the properties are not weighted for numbers of replicates on each
fuel and emission combination. In this way, the process is simplified by using the same
standardization in fitting all models, as well as in subsequent applications of the models. Note
also that the reduced fuel set is standardized using a different set of parameters than the full fuel
set.
The process of standardization is illustrated for three test fuels in Table 23. Overall, the process
applied here is similar to the "correlation transformation" sometimes applied in multiple
regression. One difference in this case is that the standardization is applied only to the predictor
variables, whereas it is also possible to apply it to the response variable.16
Table 22. Means and Standard deviations for Fuel Properties, based on Fuel Matrices for the Full and
Reduced Designs.
Model Term
Ethanol (%)
Aromatics (%)
RVP (psi)
T50 (°F)
T90 (°F)
etOH x etOH
T50x T50
etOH x Arom
etOH x RVP
etOH x T50
etOH x T90
Full Design1
Mean
Standard
deviation
10.3137
7.87956
25.6296
10.0154
8.5178
1.61137
190.611
28.5791
320.533
19.4801
0.962963
0.802769
0.962963
0.739766
-0.03674
0.978461
-0.0992352
0.999615
-0.541342
0.769153
0.0163277
0.972825
Reduced Design2
Mean
Standard
Deviation
11.0182
8.05925
24.3909
9.92426
197.000
23.4536
323.527
19.6015
1 Applies to models fit with data for 15 vehicles measured on 27 fuels.
2 Applies to models fit with data for 5 or 15 vehicles measured on 11 fuels. Note that these
models have no linear term for RVP and no 2nd order terms.
23
-------
Table 23. Examples of One-Stage and Two-Stage Standardization for Three Test Fuels (1, 5 and 20).
Fuel
etOH
Arom
RVP
T50
T90
etOH
T50
etOH
etOH
etOH
etOH
(vol.%)
(vol.%)
(psi)
(°F)
(°F)
X
X
X
X
X
X
etOH
T50
Arom
RVP
T50
T90
Fuel Properties
1
10.03
15.4
10.07
148.9
300.2
5
0.00
34.7
6.95
237.0
300.0
20
20.31
15.2
6.70
162.7
298.7
Mean1
10.314
25.630
8.518
190.6
320.5
Std.
Dev.1
7.880
10.015
1.611
28.6
19.5
One-Stage Standardized Values (Z) (Equation 10)
Ze
za
Zr
z5
z9
1
-0.036
-1.021
0.963
1.460
1.044
5
-1.309
0.906
0.973
1.623
1.054
20
1.269
-1.041
1.128
0.977
1.121
Mean
0.9630
0.9630
-0.0367
-0.0992
-0.5413
0.1633
Std.
Dev.
0.8028
0.7398
0.9785
0.9996
0.7692
0.9728
Two-Stage Standardized Values (ZZ) (Equation 11, Equation 12)
ZZee
ZZ55
ZZea
ZZer
ZZe5
ZZe9
1
-1.198
1.578
0.075
0.065
0.772
0.022
5
0.935
2.260
-1.174
1.373
-2.058
1.401
20
0.805
-0.012
-1.313
-1.332
-0.907
-1.478
1 Mean and Standard Deviations of fuel properties for the entire fuel set. See Table 22.
2 Mean and Standard Deviations of 2nd order terms values for the entire fuel set, constructed from the one-stage Z values.
2.1.1.2.3 Model Fitting
Throughout model fitting, the response variable was the natural logarithm transformation of the
emissions results (In7), and the predictor variables were the one- or two-stage standardized fuel
properties, as shown in Table 23. Thus, the model to be fit includes some subset of the 11
candidate terms shown in Equation 13.
In Y = J30 +
P\Ze + P'7-a + P7-r + PJ-5 + +
/?6ZZ55+/?7ZZee+ Equation 13
A^Zea + (39zzer + ft0ZZt5 + pnzze9 +
£
A model containing all potential candidate terms is referred to as a "full model," whereas a
model containing some subset of the candidate terms is referred to as a "reduced model." The
goal of model fitting is to identify a reduced model by removing terms from the full model that
do not contribute to fit.
24
-------
When the available data were sufficient, "mixed models" were fit, in which the terms listed in
Table 22 were included as "fixed" terms. In addition, a "random intercept" was fit for each
vehicle, which represents the high degree of variability contributed to the dataset by the vehicles
measured. One way of understanding this distinction that the fuel properties are "fixed" because
the fuels studied span the entire range of properties under study, and because the goal of the
analysis is to estimate the effect of these parameters on the mean levels of emissions. On the
other hand, "vehicle" is treated as a "random" factor because the sample of vehicles measured is
but one of many samples that could have been measured. In the analysis, the emission levels of
the specific vehicles are not of interest per sc\ but rather the degree of variability contributed to
the analysis by the different vehicles. Analyses were performed using the MIXED procedure in
the Statistical Analysis System (SAS®), version 9.2.17
When data were not sufficient for the mixed-model approach, models were fit by "Tobit
regression." This technique was used when specific datasets were affected by low-end
"censoring." For some measurements, the sample ostensibly obtained from the vehicle exhaust
was lower than that attributable to background levels. In these cases, we assumed that a small but
detectable mass was not measured accurately due to limitations in the sampling technique. In the
Tobit model, the fitting method (maximum likelihood) is modified so as to compensate for the
absence of the censored measurements. As with the mixed models, individual intercepts were fit
for each vehicle; however, as the Tobit procedure does not distinguish "fixed" and "random"
factors, vehicles were entered into the model as fixed factors (i.e., "dummy" variables). The
Tobit models were fit using the LIFEREG procedure in SAS 9.2.18
Model fitting was conducted by backwards elimination, in which all terms in the full model were
included at the outset. In fitting successive models, terms not contributing to fit were removed
based on results of likelihood-ratio tests (LRT).19 Note that the LRT were used for model
selection because all models were fit using "maximum-likelihood" (rather than "least-squares")
methods.
Model fitting results for acetaldehyde, formaldehyde, acrolein and ethanol are shown in Table 24
through Table 27. Note that these four models represent "Bag 1" or "start" emissions on the
LA92 cycle, based on datasets incorporating the full design. Also note that in fitting these
models, an additional six terms beyond the original 11 design terms were included in the full
models. These terms included one quadratic term (T90xT90), three interaction terms for
aromatics, one interaction for RVP, and one interaction for the distillation parameters
(T50xT90). However, none of these additional terms were retained as significant, with the single
exception of the T50xT90 term.
During MOVES runs, emissions of toxics are estimated as fractions of volatile organic
compounds in exhaust (VOC). To allow estimation of VOC, it was necessary to develop models
for non-methane organic gases (NMOG). NMOG is equivalent to VOC, plus the mass of ethane
and acetone.d It is calculated in MOVES from non-methane hydrocarbons (NMHC) by
correcting for the mass of oxygenated compounds not fully measured by the flame ionization
detector used to determine NMHC.20 EPA and CARB regulations set NMOG emission standards
for motor vehicles, so NMOG is an important model output. The model representing start
emissions for NMOG, fit using the full design, is shown in Table 28. This model was fit using
d Note that acetone was treated as negligible for purposes of these calculations.
25
-------
the same methods as that for total hydrocarbons (THC), as described in the Fuel Effects
Report.21
Table 24. Acetaldehyde (Bag 1): Coefficients and Tests of Effect for the Full and Reduced Models.1
Effect
Intercept
Z,
zzei
zz«
zze(
zze,
ZZe.5
ZZep
zza,
ZZa5
ZZa_
ZZpp
7.7.59
ZZri
Full Model
Estimate
Std.Err.
d.f.
lvalue
Pr>?
-5.2324
0.08802
15
-59.4
0.000000
0.8250
0.01297
898
63.6
0.000000
0.03999
0.009279
898
4.31
0.000018
-0.03667
0.01297
898
-2.83
0.0048
0.09927
0.01826
898
5.44
0.000000
0.04235
0.01115
898
3.80
0.00016
-0.1716
0.01548
898
-11.09
0.000000
0.07115
0.01314
898
5.42
0.000000
0.03016
0.01304
898
2.31
0.021
0.02020
0.008769
898
2.30
0.021
-0.01614
0.01673
898
-0.965
0.33
-0.01486
0.01072
898
-1.39
0.17
0.01738
0.01618
898
1.07
0.28
0.004828
0.01729
898
0.28
0.78
0.008759
0.008852
898
0.99
0.32
0.01270
0.01503
898
0.84
0.40
0.02718
0.01132
898
2.49
0.013
-0.0206
0.009971
898
-2.07
0.039
0.1154
0.08743
Reduced Model
Estimate
Std.Err.
d.f.
lvalue
Pr>f
-5.2323
0.08785
15
-59.6
0.000000
0.8145
0.01020
898
79.9
0.000000
0.03484
0.008249
898
4.22
0.000027
-0.04170
0.008833
898
-4.72
0.000003
0.08670
0.01063
898
8.16
0.000000
0.03801
0.007764
898
4.90
0.000001
-0.1669
0.007849
898
-21.3
0.000000
0.06665
0.007993
898
8.34
0.000000
0.01840
0.007777
898
2.37
0.018
0.02194
0.007845
898
2.80
0.0053
0.03959
0.008256
898
4.80
0.000002
0.1149
0.08850
1 See 9.2.2 and 8.7.3 in the Project Report.1
26
-------
Table 27. Ethanol (Bag 1): Coefficients and Tests of Effect for Full and Reduced Models.
Effect
Intercept2
Z,
ZZel
ZZ55
ZZel
ZZe 5
ZZe9
ZZe,
ZZa5
ZZa9
ZZ99
ZZ59
ZZa,
ZZ19
Full Model
Estimate
Std.Err.
d.f.
t-
value
Pr>?
15
1.4759
0.07240
15
20.38
<0.00001
-0.0067
0.04327
15
-0.16
0.88
-0.05004
0.04316
15
-1.16
0.26
0.1050
0.03806
15
2.76
0.015
-0.1261
0.03701
15
-3.47
0.0034
-0.4787
0.06014
15
-7.96
<0.00001
0.1261
0.05018
15
2.51
0.024
-0.005952
0.03881
15
-0.15
0.88
0.02820
0.05277
15
0.54
0.60
0.0008509
0.06491
15
0.0090
0.99
0.03237
0.05103
15
0.64
0.53
0.03318
0.03212
15
1.04
0.32
-0.01143
0.03461
15
-0.33
0.74
-0.5112
0.04523
15
-1.13
0.28
0.05311
0.04341
15
1.22
0.24
0.04136
0.02855
15
1.45
0.17
-0.008676
0.04644
15
-0.20
0.85
0.5697
Reduced Model
Estimate
Std.Err.
d.f.
t-
value
Pr>f
-4.9081
1.4643
0.07115
15
20.56
<0.00001
-0.05990
0.02940
15
-2.06
0.057
0.07188
0.02964
15
2.37
0.032
-0.09990
0.03574
15
-2.78
0.014
-0.4967
0.05229
15
-9.51
<0.00001
0.1121
0.03826
15
2.90
0.011
0.1283
0.05739
1 See 9.2.2 in the Project Report.15
2 Not fit by the Tobit model, manually recalculated from intercepts for individual vehicles.
29
-------
Table 28. NMOG (Bag 1): Coefficients and Tests of Effect for Full and Reduced Models.1
Effect
Intercept
Ze
Za
Zr
z5
z9
zzee
zz«
ZZea
ZZer
ZZei
ZZS9
2
veil
Full Model
Estimate
Std.Err.
d.f.
t-
value
Pr>?
-0.9520
0.09077
15
-10.49
<0.0001
0.07981
0.01326
941
6.02
<0.0001
0.08789
0.00929
941
9.46
<0.0001
-0.04595
0.01053
941
-4.36
<0.0001
0.1344
0.01329
941
10.12
<0.0001
0.01593
0.00925
941
1.72
0.0855
0.04594
0.01760
941
2.61
0.00918
0.07680
0.01336
941
5.75
<0.0001
0.01635
0.00906
941
1.80
0.0714
-
-
-
-
-
0.04754
0.01893
941
2.51
0.0122
0.01961
0.00902
941
2.17
0.0300
0.1224
0.07538
Reduced Model
Estimate
Std.Err.
d.f.
lvalue
Pr> t
-0.9521
0.09089
15
-10.48
<0.0001
0.08019
0.01330
941
6.027
<0.0001
0.08782
0.00932
941
9.424
<0.0001
-0.04224
0.01046
941
-4.037
<0.0001
0.1345
0.01333
941
10.09
<0.0001
0.04432
0.01764
941
2.513
0.012
0.07579
0.01340
941
5.656
<0.0001
0.01693
0.00909
941
1.862
0.063
0.04653
0.01898
941
2.452
0.014
0.1224
0.07538
1 See 9.1.2 in the Project Report
15
2.1.1.2.4 Model development under the Reduced Design
Recall that, as previously discussed, the "reduced design" involved the measurement of 11 fuels
on 5 or 15 test vehicles, whereas the "full design" involved measurement of 27 fuels on 15
vehicles.
As shown in Table 21, measurements of two compounds in Bag 1, and all compounds in Bag 2,
were performed under the reduced design. Supplementary analyses suggested that the reduced
design was not adequate to support model fitting as described in 2.1.1.2.3 above. These results
suggested that in these cases, full models retaining all four linear terms would perform as well or
better than corresponding reduced models, many of which would retain only single terms. Thus,
this sub-section presents results for full models under the reduced design.
Models representing start (Bag 1 on LA92) emissions are presented for benzene, 1,3-butadiene,
non-methane organic gases (NMOG) and ethane in Table 29 through Table 32. These models
were fit using subsets of data incorporating 15 vehicles measured over 11 fuels.
Similarly, models representing hot-running (Bag 2 on LA92) emissions are presented for
acetaldehyde, formaldehyde, ethanol, NMOG and ethane in Table 33 through Table 37. These
models were fit using subsets of data incorporating five vehicles measured over 11 fuels.
The development of these models is described in greater detail in sub-section 9.2.1 of the EPAct
analysis report.
30
-------
Table 29. Benzene (Bag 1): Coefficients and Tests of Effect for the Full Model
(fit under the reduced design, with 15 vehicles, 11 fuels).1
Effect
Intercept
Z,
Full Model
Estimate
Std.Err.
d.f.
f-value
Pr>f
-4.1019
0.1392
15
-29.48
<0.0001
-0.004685
0.03704
161
-0.126
0.90
0.4056
0.03389
161
11.97
<0.0001
0.04142
0.03789
161
1.09
0.28
0.01133
0.03255
161
0.35
0.73
0.2741
0.1873
1 See 9.2.2 and Appendix 0.3 to the Project Report.
15
Table 30. 1,3-Butadiene (Bag 1): Coefficients and Tests of Effect for the Full Model
(fit under the Reduced Design, with 15 vehicles, 11 fuels).1
Effect
Intercept
Full Model
Estimate
Std.Err.
d.f.
/-value
Pr>f
-5.8371
0.1235
15
-47.28
1.06xl0"17
-0.01729
0.03071
160
-0.56
0.57
0.02673
0.02730
160
0.98
0.33
0.01247
0.03031
160
4.11
0.000062
0.10036
0.02657
160
3.78
0.00022
0.2192
0.1089
'See 9.2.2 in the Project Report.
15
Table 31. NMOG (Bag 1): Coefficients and Tests of Effect for the Full Models
(fit under the Reduced Design, 15 vehicles, 11 fuels).1
Effect
Intercept
Z,
Full Model
Estimate
Std.Err.
d.f.
/-value
Pr>f
-0.8943
0.08668
15
-10.32
0.000000033
0.1040
0.01921
362
5.411
0.00000011
0.09435
0.01697
362
5.559
0.000000053
0.1527
0.01890
362
8.079
0.000000000
0.02127
0.01648
362
1.290
0.198
0.1091
0.08907
'See 9.2.2 in the Project Report.
15
31
-------
Table 32. Ethane (Bag 1): Coefficients and Tests of Effect for the Full Models
(fit under the Reduced Design, with 15 vehicles, 11 fuels).1
Effect
Intercept
Full Model
Estimate
Std.Err.
d.f.
/-value
Pr>f
-4.308
0.09833
15.0
-43.81
2.84xl0"17
0.1204
0.02075
160
5.805
3.37xl0"8
-0.1728
0.01844
160
-9.373
6.51xl0"17
0.2169
0.02047
160
10.59
3.30xl0"20
0.09531
0.01795
160
5.311
3.60xl0"7
0.1407
0.04970
1 See 9.2.2 in the Project Report.
15
Table 33. Acetaldehyde (Bag 2): Coefficients and Tests of Effect for the Full Models
(fit under the Reduced Design, with 5 vehicles, 11 fuels).1
Effect
Intercept
Z5
z9
Full Model
Estimate
Std.Err.
d.f.
f-value
Pr>f
-9.4189
0.1177
5
-80.1
0.000000
0.1520
0.06080
58
2.50
0.0152
0.07991
0.05279
58
1.51
0.136
-0.02997
0.05957
58
-0.503
0.617
-0.07836
0.05153
58
-1.52
0.134
0.05654
0.3814
1 See 9.2.2 and Appendix K.3 to the Project Report.
15
Table 34. Formaldehyde (Bag 2): Coefficients and Tests of Effect for the Full Model
(fit under the Reduced Design, with 5 vehicles, 11 fuels).1
Effect
Intercept
Z,
Full Model
Estimate
Std.Err.
d.f.
f-value
Pr>f
-8.6574
0.1372
5.01
-63.10
<0.00001
0.08456
0.05937
58.04
1.424
0.16
0.01575
0.05154
58.05
0.306
0.76
0.01863
0.05815
58.03
0.320
0.75
-0.08138
0.05031
58.16
-1.62
0.11
0.08205
0.3762
1 See 9.2.2 and Appendix L.4 to the Project Report.
15
32
-------
Table 35. Ethanol (Bag 2): Coefficients and Tests of Effect for the Full Model
(fit under the Reduced Design, with 5 vehicles, 11 fuels).1
Effect
Intercept1
Z,
Full Model
Estimate
Std.Err.
d.f.
/-value
Pr>f
-9.3072
0.6333
5
-15.45
0.000021
0.9233
0.2824
5
3.27
0.022
-0.3772
0.28499
5
-1.32
0.24
-.01910
0.2091
5
-0.091
0.93
-0.3017
0.2416
5
-1.25
0.27
0.3707
1.0889
1 See 9.2.2 and Appendix N.4 to the Project Report.
15
Table 36. NMOG (Bag 2): Coefficients and Tests of Effect for the Full Model
(fit under the Reduced Design, with 5 vehicles, 11 fuels).1
Effect
Intercept1
Z,
Z9
Full Model
Estimate
Std.Err.
d.f.
/-value
Pr>/
-4.777
0.4784
5
-9.99
0.00017
0.01778
0.03574
124
0.497
0.62
0.03320
0.03117
124
1.07
0.29
0.04258
0.03494
124
1.22
0.23
0.09051
0.03038
124
2.98
0.0035
1.1405
0.1026
1 See 9.2.2 in the Project Report.
15
Table 37. Ethane (Bag 2): Coefficients and Tests of Effect for the Full Model
(fit under the Reduced Design, with 5 vehicles, 11 fuels).1
Effect
Intercept1
Z^
Full Model
Estimate
Std.Err.
d.f.
/-value
Pr>/
-7.724
0.7325
5
-10.54
0.00013
0.07345
0.05873
57
1.251
0.22
-0.1260
0.05151
57
-2.447
0.018
0.1815
0.05727
57
3.168
0.0025
0.1322
0.04994
57
2.647
0.010
2.6712
0.1476
1 See 9.2.2 and Appendix Q.4 to the Project Report.
15
2.1.1.2.5 Application of EPAct Statistical Models
The approach for toxics estimates the emissions of the toxic as a fraction of emissions for VOC,
on the same fuel. So, to model the behavior of the fraction with respect to changes in fuel
33
-------
properties, it was necessary to develop models for NMOG and ethane, as well as the toxics,
because VOC is estimated as NMOG minus ethane.®
The models generated using EPAct results allow estimation of emissions effects related to the
five fuel properties included in the study design: ethanol content (vol.%), aromatics content
(vol.%), RVP (psi), T50 (°F) and T90 (°F), as well as selected interaction terms among these five
parameters.
The statistical models generated from the EPAct data follow the general structure shown in
Equation 14 below, which uses the model for acetaldehyde as an example (see Table 24). Note
that the subsets of the potential terms vary by emission and process, depending on the results of
model fitting, as described in the previous two sub-sections.
Emissions (g/mi) = exp
'/?0+/?eZe+/?aZa-
= exp /?eeZZee+/?55ZZ55
V Asa^ea + Aer^er "
f —5.23 + 0.814Z. 4
P7-< + /?5z5 + /?9z9
" O-X^veh +Ss)
y
Equation
14
= exp
0.0348Z. —0.0417Z + 0.0867Z,
0.0380Z9 —^
0.1669ZZee +0.0667ZZ55 +
0.0184ZZea +0.0219ZZer +0.5(0.1149 + 0.08850)
When the data were sufficient, two sets of exhaust fuel effect coefficients were employed for
each pollutant; one set representing cold start emissions and a second set representing hot-
running emissions. In some cases fuel effects estimated for these two processes differed
substantially, as the effects of fuel properties on start emissions are dominated by changes in
combustion and catalyst warm-up, while the impact of running emissions is dictated by catalyst
efficiency when fully operational. Thus, using convenient matrix notation, the expressions
XPtoxic, XaNMOG and XGethane represent models for a selected toxic compound, NMOG and
ethane, respectively, calculated by applying Equation 14 to each compound for a specified fuel.
The toxic emissions as a fraction of VOC emissions (/toxic) are given by
Toxic Fraction = f =
J toxic
Xa.K
,X0„
Equation 15
For all compounds, the calculation shown in Equation 15 is applied in the
GeneralFuelRatioExpression table. In calculating toxic fractions, we elected to use models for
NMOG and ethane fit using study designs and datasets similar to those for the toxic compounds.
That is to say, if the toxic model was fit with the reduced design, we combined it with the
NMOG and ethane models also fit with the reduced design. We followed this approach to
prevent the calculation and propagation of artifacts in the estimated fractions resulting from
differing levels of information and complexity in the numerator and denominator in Equation 15.
In this context we considered it important to apply "information parity" to the toxic model in the
numerator and the NMOG model in the denominator, as the vast majority of VOC mass is
e In MOVES, VOC is typically calculated as NMOG - ethane - acetone, but for this purpose, acetone was considered negligible,
and was not subtracted.
34
-------
represented by NMOG, with ethane comprising only a small fraction. Table 38 summarizes the
combinations of models used to calculate toxic fractions for start and running emissions.
Note that for three compounds in Bag 2, levels of "left censoring," were high enough that
modeling was not considered feasible. Again, "censoring" occurs when background levels of the
compounds under study were as high or higher than levels ostensibly measurable in vehicle
exhaust. Estimation of "simple" toxic fractions for these compounds is covered in the following
sub-section.
Table 38. References to Tables containing Coefficients for Models
used to Calculate Toxic Fractions of VOC (see Table 17, page 18)
Compound
Acetaldehyde
Formaldehyde
Acrolein
Ethanol
Benzene
1,3-butadiene
Start Emissions (Bag 1)
Toxic
NMOG
Ethane
Table 24
Table 28
Table 32
Table 25
Table 28
Table 32
Table 26
Table 28
Table 32
Table 27
Table 28
Table 32
Table 29
Table 31
Table 32
Table 30
Table 31
Table 32
Running Emissions (Bag 2)
Toxic
NMOG
Ethane
Table 33
Table 36
Table 37
Table 34
Table 36
Table 37
NO MODEL
Table 35
Table 36
Table 37
NO MODEL
NO MODEL
2.1.1.2.6 Estimating Simple Fractions of VOC for Running Emissions
As noted in Table 21, models for running emissions are not available for three compounds:
acrolein, benzene and 1,3-butadiene. For these compounds, the relevant subsets of data were
inadequate to allow model fitting. Therefore, for these compounds, running emissions were
represented as "simple" (constant) fractions of VOC, with values derived from the available data.
Thus, for acrolein, benzene and 1,3-butadiene, the values of the toxic fractions were 0.00077,
0.047 and 0.0, respectively. These values were derived as "ratios of means" (ROM), in which the
toxic and VOC values were averaged first by vehicle and then across vehicles, as described
below. The ROM approach is generally preferred as it provides an unbiased estimator of the true
fraction as the sample size increases22.
For benzene, results were available for four vehicles, differing widely in their benzene and VOC
levels, and also in numbers of available measurements, as shown in Table 39. The averaging was
performed in two steps so that the vehicle(s) with the greatest numbers of measurements would
not dominate the overall mean. In the first step, the benzene and VOC values were averaged for
each vehicle. In the second step, the four vehicle means were averaged to give an overall mean.
Finally, the overall mean for benzene was divided by that for VOC to give a simple ratio
estimator for benzene as a fraction of VOC.
35
-------
Table 39. Benzene (Running): Derivation of a Ratio-of-Means Estimator for Benzene as a Fraction of VOC.
Vehicle
n
Benzene (mg)
VOC (mg)
Ratio of means (ROM)1
Corolla
2
0.053752
2.2694
F150
10
2.2241
28.427
Impala
3
0.10825
10.670
Silverado
4
0.29381
16.216
| All vehicles | 4 | 0.669971 | 14.396 | 0.0465 |
1 This value is a simple average of the means for all four vehicles, as listed above.
The VOC fraction for acrolein was derived similarly (Table 40). For this compound results were
available for five vehicles. Values for acrolein are considerably lower than for benzene, so
results are expressed in jag, rather than mg. The resulting fraction is two orders of magnitude
lower than that for benzene.
Table 40. Acrolein (Bag 2): Derivation of a Ratio-of-Means Estimator for Acrolein as a Fraction of VOC.
Vehicle
n
Acrolein (|ig)
VOC (ng)
Ratio of means (ROM)
Civic
3
5.4190
3,038.9
Corolla
5
2.8934
2,929.6
F150
5
8.3558
24,321
Impala
6
8.0180
10,408
Silverado
10
19.662
17,192
| All vehicles | 5 | 8.86961 | 11,578 | 0.0007661 |
1 This value is a simple average of the means for all five vehicles, as listed above.
For 1,3-butadiene in hot-running operation, measurements were extremely low; in fact, we
considered the dataset so heavily affected by "left-censoring" that we did not consider it
adequate for either model fitting or development of ratio estimators. Accordingly, for modeling
purposes, we have adopted an assumption that this compound is not emitted during hot-running
operation, i.e., the ROM estimator is 0.0.
2.1.1.2.7 Post-Model Adjustments
For two compounds, benzene and 1,3-butadiene, additional refinements were applied to
supplement the study design of the EPAct fuel set. These adjustments are applied to both start
and running emissions.
For benzene, the issue is that the fuel matrix included aromatics generally, but not benzene
specifically. As we considered it inadequate to model benzene in exhaust without explicitly
accounting for benzene levels in fuel, we developed a "post-model" refinement using data
external to the EPAct program. In this case, the source was a program conducted in support of
the 2007 MSAT2 rule. This program performed measurements on nine Tier-2 certified vehicles
on fuels with benzene levels ranging from 0.6 to 1.1 percent by weight.23,24 With benzene
represented as a fraction of VOC (as in Equation 15) denoted as ^benzene, a value modified to
account for benzene levels in different fuels (J*benzene) is calculated as shown in Equation 16
where Xbenzene is the benzene level for the fuel modeled (weight percent), A is the mean benzene
level in the EPAct exhaust program fuel set (0.66 weight percent), and B is an empirical
coefficient, taking a value of 0.24.
36
-------
/be
benzene
[(-*".
benzene
A)-B-fb
benzene
]+ ./l:
benzene
Equation 16
Similarly, given the importance of olefins to estimation of emissions for 1,3-butadiene, and that
the EPAct exhaust program study design did not incorporate olefins as a factor, we considered it
appropriate to develop a post-model adjustment explicitly accounting for olefin level. This
adjustment was derived by varying olefin levels in the Complex Model and fitting a polynomial
trend to the results.25 Starting with an unadjusted toxic fraction for 1,3-butadiene (/buta), the
modified fraction f buta is calculated using Equation 17, in which x0iefm is the olefin level, and A,
B, C and D are coefficients, taking values of 0.000008, 0.0002, 0.0069 and 0.008823,
respectively.
For fuel blends with 0%, 10% and 15% ethanol, composite speciation profiles developed from
the results of EPAct (Phase 1) were used to develop toxic fractions of VOC for the hazardous air
toxics listed in Table 41.f These profiles were based on averaging results of tests from 3
vehicles.26'27 Toxic fractions for E10 are used for all gasolines containing ethanol levels of 5
vol.% or greater. For fuel blends containing 20% ethanol, fractions were developed using a
composite speciation profile developed using results from the EPAct (Phase 3) program. The
fractions are also presented in Table 41. The values shown in Table 41 are stored in the database
table minorHAPRatio (see Table 16). For blends containing MTBE, no data were available for
Tier 2 vehicles; thus the toxic to VOC ratios for Tier 1 and earlier vehicles were used (See Table
_ (Ax, + Bx, + C
^ olelin olenn
Equation 17
2.1.1.2.8 Additional Air Toxics Estimated from EPAct Speciation Profiles
15).
fPhase 1 testing was done using fuels more representative of in-use fuels, in contrast to the orthogonal matrix used for EPAct
Phase 3.
37
-------
Table 41. Toxic fractions of VOC for Selected Compounds, Representing Model years 2001 and Later.
Pollutant (pollutantID)1
2,2,4-Trimethylpentane (40)
Ethyl Benzene (41)
Hexane (42)
Propionaldehyde (43)
Styrene (44)
Toluene (45)
Xylene(s) (46)
Fuel Blends (Gasoline and Ethanol'
0% (E0)
10% (E10)**
15% (E15)
20% (E20)
0.03188
0.01227
0.02198
0.004625
0.01683
0.01660
0.01568
0.022199
0.002790
0.02911
0.0110
0.02497
0.00122
0.00054
0.0005984
0.0006607
0.00085
0.00083
0.004588
0.004096
0.07542
0.07440
0.0727
0.09646
0.06127
0.06047
0.06902
0.09302
1 For fuels containing 0-20% ethanol, fractions for ethanol, benzene, acetaldehyde, formaldehyde, 1,3-butadiene,
and acrolein were estimated using methods described in 2.1.1.2.1.
**Values also applied for fuels containing 5% and 8% ethanol, (E5 and E8).
2.1.2 Vehicles Operating on Fuel Blends Containing 70-100% Ethanol
2.1.2.1 2000 and Earlier Model Year Vehicles
Major HAP emissions for 2000 and earlier model year vehicles operating on fuel blends
containing 70-100% ethanol are estimated using toxic fractions of VOC. The toxic fractions were
derived from data for four flexible-fuel vehicles running on E85 gasoline, collected during the
EPAct program (Phase 3) and are displayed in Table 42. Since no measurements were obtained
on an E70 blend, more typically used in winter, or blends above E85, the same toxic to VOC
fractions are used for all ethanol-gasoline blends containing 70-100% ethanol. These ratios are
applied to older technology (2000 and earlier vehicles), even though data were collected from
Tier 2 vehicles®. The 2000 and earlier HAP emission rates are stored in the database table
"ATRatioNonGas" (see Table 43)h.
g Because the data used to derive the E85 emission rates are based on Tier 2 vehicles, there is more uncertainty in the emission
rates from 2000 and older technology vehicles running on high ethanol blends in MOVES. However, pre-2001 flex-fuel vehicles
are minor portion of the light-duty gasoline fleet. For example, the default MOVES2014 population indicates that less than 1-3%
of the 1998-2000 model year light-duty gasoline vehicles are flex-fuel vehicles, and MOVES2014 doesn't include any flex-fuel
vehicles earlier than 1998. Additionally, few flex-fueled vehicles actually use high-ethanol blend fuels, making pre-2001, high-
ethanol blend fueled vehicles an even smaller portion of the vehicle emissions inventory
hThe October release of MOVES2014 contained an error, in that it did not produce NMOG and VOC emissions from pre-2001
vehicles fueled on E85. Subsequently, it did not produce toxics that were calculated as a fraction of VOC. This error has been
corrected in MOVES2014a.
38
-------
Table 42. E70/E85 Major HAP VOC Fraction for 2000 and Earlier Model Year Vehicles.
Pollutant (pollutantID)
Toxic Fraction
Benzene (20)
0.0170
Ethanol (21)
0.3724
1,3-butadiene (24)
0.0011
Formaldehyde (25)
0.0291
Acetaldehyde (26)
0.1644
Acrolein (27)
0.0010
Table 43. Description of the Database Table "ATRatioNonGas," as Applied to Light-Duty Vehicles.
Field
Description
RelevantValues
polProcessID
Identifies the pollutant (1st two
digits and Emissions Process
(last two digits).
Pollutants are identified in the table above;
Relevant processes include:
"Running Exhaust" (processID =1)
"Start Exhaust" (processID = 2)
"Extended Idle Exhaust" (processID = 90)
"Auxiliary Power Exhaust" (processID = 91)
sourceTypelD
Identifies types of vehicles,
classified by function
Motorcycle (11)
Passenger Car (21)
Passenger Truck (31)
Light Commercial Truck (32)
fuelSubTypelD
Identifies specific fuel classes
within the fuelTypelD
51 = "Ethanol (E85)"
52 = "Ethanol (E70)"
modelY earGroupID
Identifies a set of model years
covered by a specific value of
atRatio.
atRatio
Fraction, or "ratio" of the toxic
relative to total VOC.
atRatioCV
"Coefficient of Variation of the
Mean" or "relative standard
error" of the atRatio.
dataSourcelD
Indicates source data and
methods used to estimate
atRAtio.
2.1.2.2 2001 and Later Model Year Vehicles
For major HAPs in 2001 and later model year vehicles, we conducted a more comprehensive
analysis than for the older model year vehicles. Instead of deriving toxic fractions of VOC, we
developed adjustment factors that were compatible with the EPAct toxic ratios derived for
gasoline 2001 and later model year vehicles discussed in the Section 2.1.1.2. The toxic
adjustment factors were developed based on the analysis of EPAct (Phase 3) program, National
Renewable Energy Laboratory (NREL) E4028, Coordinating Research Council (CRC) E-8029,
and the PM Speciation Program30. All programs measured emissions from LA92 test cycle on
both E10 and E85, except CRC E-80 which tested E6 and E85. Only the vehicles tested on both
39
-------
E10 (E6) and E85 were included in the analysis. Numbers of vehicles in each program are
summarized in Table 44.
Table 44. Number of Vehicles included in the Analysis of Major HAPs
Test Program
Number of Vehicles
EPAct (phase 3)
4
NRELE40
9
CRC E-80
7
PM Speciation
2
Consistent emission trends were observed across datasets; thus, all available datasets were
pooled to examine the effect of E85 on emissions compared to E10. First, the test of significance
of differences between E10 and E85 was performed using Student's paired ^-tests. Next, when
there was a statistically significant difference in emissions between E10 and E85, the adjustment
factors were calculated using Equation 18. The adjustment factor was set to zero when the
differences in emissions were not statistically different (i.e., acrolein).
ToxicsFR5
£85 adjustment factor = Equation 18
J ' Toxicse10
VOCE10
The resulting adjustment factors are shown in Table 45, and are stored in the database table,
"GeneralFuelRatioExpression" for fuelTypelD = 5. The E10 to E85 adjustments are used to
estimate major HAP emissions for all 2001 model year vehicles and later.
Table 45. E70/E85 Adjustment Factors for Major HAPs for 2001 and Later Model Year Vehicles
Pollutant (pollutantID)
Adjustment Factor for E70/E85
Benzene (20)
0.6672
Ethanol (21)
7.587
1,3-butadiene (24)
0.2167
Formaldehyde (25)
1.572
Acetaldehyde (26)
7.126
Acrolein (27)
0
2.1.2.3 Air Toxics Fractions that Apply to All Model Year Vehicles
Fractions for the remaining air toxic compounds modeled in MOVES were developed from the
four flexible-fuel vehicles tested during the EPAct program (Phase 3) for all model year vehicles
running on fuels containing 70-100% ethanol. A single emission test program was used for these
pollutants, because they were not involved in the updated analysis discussed in the previous
section (2.1.2.2). As stated earlier, the vehicles were tested on a single E85 gasoline fuel. These
ratios are applied to older technology (2000 and earlier vehicles) as well as the modern
technology vehicles in the test program; thus, there is more uncertainty in emission estimates for
older technology vehicles running on high ethanol blends than for newer vehicles.g The VOC
fractions shown in Table 46 are stored in the database table "minorHAPRatio" (see Table 16).
40
-------
Table 46. Toxic Fractions of VOC for Vehicles Running on E70/E85 for All Model Year Vehicles
Pollutant (pollutantID)
Toxic Fraction of VOC
2,2,4-Trimethylpentane (40)
0.0078
Ethyl Benzene (41)
0.0055
Hexane (42)
0.0045
Propionaldehyde (43)
0.0025
Styrene (44)
0.0003
Toluene (45)
0.0177
Xylene(s) (46)
0.0185
2.2 Poly cyclic Aromatic Hydrocarbons (PAHs)
2.2.1 Vehicles Operating on Fuel Blends Containing 0-20% Ethanol
Emissions of PAHs are estimated through the use of fractions in a manner similar to that used for
VOCs as described in the previous section. However, for PAHs, the process is complicated by
the fact that exhaust and crankcase emissions of these compounds are emitted in both the
gaseous and particulate phases. Accordingly, emissions in the gaseous phase are estimated as
fractions of total VOC, and emissions in the particulate phase as fractions of organic carbon <
2.5 [j,m (OC2.5).
The PAH emission fractions for gasoline vehicles are estimated from a set of 99 vehicles
measured in the Kansas City Light-duty Vehicle Emissions Study (KCVES).31 These vehicles
were included in a subsample selected for chemical speciation. For each vehicle, emissions of
THC and particulate matter 2.5 microns in diameter or less (PM2.5) were measured. Fleet-average
fractions of PAH/THC and PAH/PM2.5 were calculated with each sample weighted by total
emissions1, vehicle-miles traveled (VMT), and an equal weight between summer and winter.34
We used a VOC/THC fraction of 0.86 developed from the total organic-gas speciation profile
developed from the Kansas City program (8750a), in estimation of PAH/VOC fractions. We
adjusted the PAH/PM2.5 fraction by the fraction of OC measured in the start (42.6%) and running
emission processes (55.7%) to produce PAH/OC2.5 emission fractions.34 Because OC/PM
fractions differ for start and running, we have separate PAH/OC toxic fractions for start and
running.
The partitioning of PAH emissions between gaseous and particulate phases is assigned on the
basis of average temperature and dilution conditions at the time of measurement, i.e., in the
sample train and constant-volume sampler. Thus, the partitioning reflected in the emission
fractions does not reflect cooling and dilution occurring in the "real world" after the exhaust
leaves the tailpipe. The sampling conditions set forth in EPA regulations for particulate and
hydrocarbon measurement differ for light-duty and heavy-duty vehicles, which affects the phase
partitioning of PAH emissions obtained from both engine types. In preparing inputs for MOVES,
we developed one set of phase allocation factors for gasoline sources and another for diesel
1 Each sample contained emissions from one to five vehicles.
41
-------
sources in order to streamline data processing, and to be consistent with the measurement
conditions reflected in the PAH measurements.
The allocations of PAHs into gaseous and particulate phases for gasoline vehicles are based on
measurement samples analyzed by Desert Research Institute (DRI) on a subset of vehicles in the
KCVES that were measured with dilution air at both low and high dilution temperatures.32 One
of the purposes of this follow-up study was to examine the impact of sampling conditions on
PAH emission measurements. DRI measured PAH species with Teflon-impregnated glass filters
(TIGF) and backup glass cartridges with Amberlite XAD-4 adsorbent resins over the LA-92
cycle. Relative concentrations of individual PAH were measured on the TIGF and the XAD with
sampling line and dilution temperatures of 20°C and 47°C for four composite samples, with each
composite sample containing one to three vehicles. Table 47 reports the TIGF/XAD phase
allocation factors measured at 47°C (which was the measurement temperature for the Kansas
City Light-duty Vehicle Emissions Study), for the composite sample referred to as the 'medium-
emitters.' This class contained a 1989 Camry and 1992 Voyager. In MOVES2014, we used the
PAH phase-partitioning of this sample to estimate the relative gas and particle portioning of all
gasoline-source emissions. Clearly, this sample may not adequately represent phase-partitioning
of PAH emissions from the current in-use fleet; however, it was deemed the most representative
of the breadth of gasoline vehicles sampled in the KCVES. Note that the PAH species
partitioning was heavily dependent on molar mass (molecular weight); compounds with lighter
molar masses (e.g., naphthalene) were measured almost entirely in the gaseous phase, whereas
compounds with heavier molar masses were measured almost entirely in the particulate phase
(e.g., dibenzo(a,h)anthracene).
42
-------
Table 47. Gasoline PAH Phase Allocation Factors.
PAH species
Molar Mass
(g/mol)
Phase Fraction
Gaseous
Particulate
Naphthalene
128
0.9996
0.0004
Acenaphthylene
152
0.9985
0.0015
Acenapthene
154
1.0000
0.0000
Fluorene
166
1.0000
0.0000
Anthracene
178
0.9915
0.0085
Phenanthrene
178
0.9953
0.0047
Fluoranthene
202
0.9822
0.0178
Pyrene
202
0.9831
0.0169
Benz(a)anthracene
228
0.6721
0.3279
Chrysene
228
0.7307
0.2693
Benzo(a)pyrene
252
0.0426
0.9574
Benzo(b)fluoranthene
252
0.5546
0.4454
Benzo(k)fluoranthene
252
0.5546
0.4454
Bcnzo(g.h.i)pcrvlcnc
276
0.0000
1.0000
Indeno( 1,2,3 -cd)pyrene
276
0.0000
1.0000
Dibenzo(a,h)anthracene
278
0.0000
1.0000
The PAH/VOC and PAH/OC emission fractions used in MOVES2014, are calculated by
multiplying the PAH/VOC, and PAH/OC fractions calculated from the Kansas City Vehicle
Emission Study (KCVES) by the gas/particle partitioning factors in Table 19. The calculation is
displayed with Equation 19 and Equation 20 for each PAH, i= 1:16.
PAH¦ PAH-
'¦(Table 48) = ' (KCVES) x Gaseous Fraction^ (Table 47)
VOC
VOC
Equation 19
PAH¦ PAH-
'¦(Table 48) = ' (KCVES) x Particulate Fractiorii (Table 47)
OC
OC
Equation 20
Within MOVES, the PAH fractions in Table 20 are applied to all gasoline fuels with ethanol
content less than 20%. In the MOVES database, these fractions are stored in two tables.
Fractions for the gaseous and particulate phases are stored in the tables "pahgasratio" and
"pahparticleratio," respectively. The two tables have the same structure, which is presented in
Table 49.
43
-------
Table 49. Description of the Database Table "pahGasRatio" and "pahParticleRatio"
Field
Description
Relevant Values
polProcessID
Identifies the pollutant (1st two
digits and Emissions Process
(last two digits).
Pollutants are identified in the table above;
Relevant polprocesses include:
18501 = "Naphthalene gas, running exhaust"
18502 = "Naphthalene gas, start exhaust"
fiielTypelD
Identifies broad classes of fuels,
e.g., "gasoline." "diesel."
1 = "Gasoline"
2 = "Diesel"
3 = "CNG"
5 = "Ethanol"
modelY earGroupID
Identifies a set of model years
covered by a specific value of
atRatio.
1960-1970
1971-1977
1978-1995
1996-2006
2007-2050
atRatio
Average PAH/VOC emission
ratio for a combination of
process, fuel type, sourceType
and modelYearGroup.
meanBaseRate CV
"Coefficient of Variation of the
Mean" or "relative standard
error" of the meanBaseRate.
dataSourcelD
Indicates source data and
methods used to estimate
atRAtio.
2.2.2 Vehicles Operating on Fuel Blends Containing 70-100% Ethanol
Hays et al. (20 1 3)33 reported speciated filter-collected semi-volatile organic compound (SVOC)
measurements from three Tier 2 compliant vehicles tested using EO, E10 and E85 fuels.
Reductions in total PAH between EO and E85 in total measured filter-collected PAHs ranged
between 22% and 93% depending on the temperature and phase of the LA-92 cycle. They found
that E85 significantly reduced the lighter PAHs, including naphthalene, fluorene, anthracene,
phenanthrene, fluoranthene, pyrene, benzo(a)anthracene and chrysene. However, no significant
effect was observed for the heavier PAHs, including benzo(a)pyrene, benzo(£)fluoranthene,
benzo(g/z/)perylene, and indeno(l,2,3-cJ)pyrene.
Because Hays et al. (2013) reported only the filter-collected PAH emissions, and the results were
conducted on a limited number of vehicles, we used the results to adjust the fleet-average PAH
ratios derived from the Kansas City Vehicle Study tested on EO fuel. We reduced the VOC phase
PAH ratios by 74%, assuming that (1) the annual average ethanol content of high ethanol fuels
is 74%), and (2) the PAH in the gaseous phase are reduced proportionally to the gasoline content
reductions. The 74%> reduction is within the range of reductions observed by Hays et al. (2013)33
for total PAHs. Because Hays et al. (20 1 3)33 observed no significant decrease of the heavier
PAHs for which MOVES assumes exist primarily in the particle-phase (Table 47), we assume
45
-------
the E85 particle PAH/OC fractions are the same as the E0-E20 fractions derived from the Kansas
City Light-duty Vehicle Emissions Study. The resulting fractions are presented in Table 50.
Table 50. Toxic Fractions for PAH species for Vehicles Running on High-Ethanol Blends by Process.
PAH species
PAH/VOC
PAH/OC25
Start
Running
Naphthalene
5.38xl0"4
1.68xl0"4
1.29xl0"4
Acenaphthylene
4.71xl0"5
5.01 xlO"5
3.83xl0"5
Acenaphthene
1.04xl0"5
0.0
0.0
Fluorene
2.10X10"5
0.0
0.0
Anthracene
8.70xl0"6
5.19xl0"5
3.97xl0"5
Phenanthrene
5.57xl0"5
1.81 xlO"4
1.39xl0"4
Fluoranthene
1.45xl0"5
1.83 xlO"4
1.40 xlO"4
Pyrene
1.66X10"5
1.98xl0"4
1.52xl0"4
Benz(a)anthracene
1.41xl0"6
4.76 xlO"4
3.64 xlO"4
Chrysene
1.57xl0"6
4.02 xlO"4
3.08X10"4
Bcnzo(fl)pyrcnc
7.65 xlO"8
1.19xl0-3
9.13 xlO"4
B c 11 /o (/>) fl 110 ra 111 lie 11c
1.04xl0"6
5.81 xlO"4
4.45 xlO"4
B e 11 /o (A') fl no ra nt 1 ic 11c
1.04X10"6
5.81 xlO"4
4.45 xlO"4
B c 11 /o (#/? /) pc rv 1 c 11c
0.0
3.23 xlO"3
2.47 xlO"3
Indeno( 1.2.3.«/)pyrcnc
0.0
1.21 xlO"3
9.28xl0"4
D ibc nzo(fl/? )a nthraccnc
0.0
2.79xl0"5
2.13 xlO"5
2.3 Metals
Emissions of metals in vehicle exhaust result from trace-level contamination of fuel and engine
oil, as well as attrition from engine, exhaust system, and emission-control components.
MOVES2014 models two groups of metal emissions, 1) metals that are used for air quality
modeling, and 2) metals that are included due to their known toxicity. The metals that are
included for air quality modeling, which include metals such as iron, aluminum and calcium are
discussed in the MOVES2014 Speciation report.34 Emissions of these metals are estimated as
fractions of PM2.5 emission rates.
This report covers seven metal species included due to their known toxicity, including five
metals and three forms of mercury, as listed in Table 4. The toxic metal emissions are estimated
using distance-specific emission rates (g/mile). Manganese is the only metal that is required for
both purposes, and is estimated using the g/mile approach. In the database, these rates are stored
in the table metalEmissionRate, described in Table 52. Note that while the table contains a field
for "fuel type," the emission rates listed in the table do not vary among fuel types.
Emission rates for magnesium and nickel were developed from the 99 vehicles sampled for
chemical composition in the KCVES. The mean rates are calculated as weighted averages of
metal measured on Bag 2 of the LA92, using weights designed to represent the on-road vehicle
fleet.34 The use of Bag 2 emissions in the averaging helps ensure that the emission rates for
these metals are consistent with the PM2.5 emission profile for running emissions discussed in the
MOVES2014 TOG and PM Speciation Report.34 These approaches were adopted because while
46
-------
PM2.5 emissions are much lower during hot-stabilized running conditions, PM2.5 emissions are
more enriched in metals during hot-stabilized running conditions than during start emissions. We
compared the g/mi emission rates from Bag 2 to the average of the entire LA92; the difference in
the Bag 2 emission rates from the average of the LA92 is 38% and -16% for manganese and
nickel. Thus, in using Bag 2 emission rates for metal emission rates, the approach is both
consistent with the PM2.5 speciation running emission profile and provides a likely upper limit (in
the case of manganese) when compared to the cycle average.
Hexavalent chromium was estimated using data collected at U. S. EPA's National Vehicle
Emissions Laboratory and analyzed at the Wisconsin State Laboratory of Hygiene at the
University of Wisconsin-Madison. These data were collected on a single vehicle, a 2008
Chevrolet Impala flexible-fuel vehicle. They are the only available data with direct measurement
of hexavalent chromium from a highway vehicle. Development of a gasoline vehicle emission
rate from these data is detailed in Appendix A. Eighteen percent of chromium was assumed to
be hexavalent, based on combustion data from stationary combustion turbines burning diesel
fuel.35
Emission factors for arsenic were developed from data reported for tunnel tests.36 These data
were collected in two Milwaukee tunnels in 2000/2001, using inductively-coupled plasma mass
spectrometry (ICP-MS) and a chemical mass balance model was used to apportion
concentrations to sources. Emission factors for mercury were obtained from a 2005 test program
at EPA's National Exposure Research Laboratory (NERL). In this program mercury samples in
raw exhaust were collected from 14 light-duty gasoline vehicles and two heavy-duty diesel
vehicles. Documentation describing development of these emission factors can be found in
Appendix B.
Table 51. Metal Emission Rates for Gasoline Motor Vehicles.
Pollutant
Emission Rate (g/mi)
Chromium, hexavalent (6+)
1.20xl0"8
Manganese
1.33xl0"6
Nickel
1.50xl0"6
Mercury, Elemental (Gaseous
Phase)
l.lOxlO"7
Mercury, Reactive (Gaseous
Phase)
9.90xl0"9
Mercury, Particulate Phase
4.00X10"10
Arsenic
2.30xl0"6
Fleet-average metal emission rates were derived for vehicles running on gasoline and gasoline-
ethanol blends. Since metal emissions can result from trace level contamination of fuel and
engine oil, as well attrition from exhaust emission components, there is no way to estimate metal
emissions for vehicles running on E85 or E70 fuel in the absence of data. Thus metal emission
rates were assumed to remain unchanged from those applicable to conventional gasoline vehicles
(see Table 51, page 47).
47
-------
Table 52. Description of the Database Table "metalEmissionRate"
Field
Description
RelevantValues
polProcessID
Identifies the pollutant (1st two
digits and Emissions Process
(last two digits).
Pollutants are identified in the table above;
Relevant processes include:
1 = "Running Exhaust"
fuelTypelD
Identifies broad classes of fuels,
e.g., "gasoline." "diesel."
1 = "Gasoline"
2 = "Diesel"
5 = "Ethanol"
sourceTypelD
Identifies vehicle types,
classified by function
Motorcycles (11)
Passenger Cars (21)
Passenger Trucks (31)
Light Commercial Trucks (32)
modelY earGroupID
Identifies a set of model years
covered by a specific value of
atRatio.
1960-1970
1971-1977
1978-1995
1996-2006
2007-2050
Units
Identifies units in which the
meanBaseRate is expressed.
grams/mile
meanBaseRate
Average emission rate for a
combination of process, fuel
type, sourceType and
modelYearGroup.
meanBaseRate CV
"Coefficient of Variation of the
Mean" or "relative standard
error" of the meanBaseRate.
dataSourcelD
Indicates source data and
methods used to estimate
atRAtio.
2.4 Dioxins and Furans
2.4.1 Vehicles Operating on Fuel Blends Containing 0-20% Ethanol
The MOVES model estimates emissions for 17 dioxin and furan congeners. The emissions are
estimated using distance-specific emission rates as shown in Table 54. These emission rates
were obtained from a tunnel study and used in EPA's dioxin assessment.37'38 The emission rates
from the tunnel study did not vary among fuel types; in MOVES we are applying theses rates to
gasoline vehicles. The rates are stored in the database table "dioxinEmissionRate," which is
described in Table 55.
Before the 'MOVES2014a November 2016 Patch', the rates for dioxins/furans in MOVES2014
were not expressed in terms of mass directly. Rather, dioxins and furans were expressed in terms
of "toxic equivalents" (TEQ), which effectively resolves the emissions of all dioxin and furan
48
-------
congeners into a single "species," represented by the two most carcinogenic congeners, 2,3,7,8-
tetrachlorodibenzo-p-dioxin and 1,2,3,7,8-pentachlorodibenzo-p-dioxin. The emissions of the
other congeners were expressed as equivalent masses of these two congeners, by multiplying the
mass-based emission rates by the toxic equivalency factors (TEFs) shown in Table 53. In the
MOVES2014a November 2016 patch, we are now reporting the absolute mass of emission rates
of dioxins and furans, so the magnitude of the emission rates has increased by the inverse of the
TEFs shown in Table 53.
Table 53. Dioxin/Furan Toxic Equivalency Factors (World Health Organization)
Pollutant
TEF
2,3,7,8-TCDD TEQ
1.0
1,2,3,7,8-Pentachlorodibenzo-p-Dioxin
1.0
1,2,3,4,7,8-Hexachlorodibenzo-p-Dioxin
0.10
1,2,3,6,7,8-Hexachlorodibenzo-p-Dioxin
0.10
1,2,3,7,8,9-Hexachlorodibenzo-p-Dioxin
0.10
1,2,3,4,6,7,8-Heptachlorodibenzo-p-Dioxin
0.01
Octachlorodibenzo-p-dioxin
0.0003
2,3,7,8-Tetrachlorodibenzofuran
0.10
1,2,3,7,8-Pentachlorodibenzofuran
0.030
2,3,4,7,8-Pentachlorodibenzofuran
0.3
1,2,3,4,7,8-Hexachlorodibenzofuran
0.1
1,2,3,6,7,8-Hexachlorodibenzofuran
0.1
1,2,3,7,8,9-Hexachlorodibenzofuran
0.1
2,3,4,6,7,8-Hexachlorodibenzofuran
0.1
1,2,3,4,6,7,8-Heptachlorodibenzofuran
0.01
1,2,3,4,7,8,9-Heptachlorodibenzofuran
0.01
Octachlorodibenzofuran
0.0003
49
-------
Table 55. Description of the Database Table "DioxinEmissionRate"
Field
Description
RelevantValues
polProcessID
Identifies the pollutant (1st two
digits and Emissions Process
(last two digits).
Pollutants are identified in the table above;
Relevant processes include:
1 = "Running Exhaust"
fuelTypelD
Identifies broad classes of fuels,
e.g., "gasoline." "diesel."
1 = "Gasoline"
2 = "Diesel"
5 = "Ethanol"
modelY earGroupID
Identifies a set of model years
covered by a specific value of
atRatio.
1960-2050
1960-2006
2007-2009
2010-2050
Units
Identifies units in which the
meanBaseRate is expressed.
grams/mile
meanBaseRate
Average emission rate for a
combination of process, fuel
type, sourceType and
modelYearGroup.
meanBaseRate CV
"Coefficient of Variation of the
Mean" or "relative standard
error" of the meanBaseRate.
dataSourcelD
Indicates source data and
methods used to estimate
atRAtio.
In the absence of additional data, the fractions for more recently-manufactured vehicles were
assumed to be the same as those for vehicles employing older technologies (see page 50). Of
course, this extrapolation from one set of technologies to another involves some degree of
uncertainty.
2.4.2 Vehicles Operating on Fuel Blends Containing 70-100% Ethanol
No emissions data exist for dioxin and furan emissions from vehicles running on E85 or E70.
Thus dioxin emission factors for E85 and E70 were estimated by multiplying fractions for
vehicles running on E0 fuels (Table 54) by the fraction of gasoline in the fuel, assuming no
emission of dioxins or furans resulting from the combustion of ethanol. Resulting ratios are
given in Table 56.
51
-------
3 Diesel Exhaust: Pre-2007
Toxic fractions, dioxin and metal emission rates were developed for exhaust emissions from
heavy-duty diesel vehicles and applied to all diesel vehicle categories. The pre-2007 diesel toxic
fractions for VOCs and PAHs are applied to auxiliary power unit exhaust for all model year
vehicles, because auxiliary power units are not subject to the same stringency of control as
highway engines. There are no separate emission ratios or factors for diesel engines running on
biodiesel fuels or synthetic diesel fuels, due to limited data. Biodiesel vehicles use the same toxic
ratios and factors as regular diesel. The toxic emission data are based on heavy-duty testing but
are applied to light-duty diesel with the same model year distinctions (pre-2007 and post-2007).
3.1 Volatile Organic Compounds
The composition of VOC emissions for heavy-duty diesel engines lacking the advanced control
technologies applied in more recently-manufactured vehicles differs substantially from earlier
technologies. Thus, we developed one set of toxic fractions for pre-2007 diesel engines and
another set for engines manufactured in 2007 and later.
To estimate toxic fractions of VOC for vehicles in the pre-2007 model-year group, EPA relied on
a database compiled for the Coordinating Research Council and the National Renewable Energy
Laboratory (NREL) (CRC E-75).39 This database was developed from a literature survey and
compiled data collected in 13 different studies. The studies included were conducted in a
number of different countries, included heavy-duty and light-duty engines, a variety of diesel and
biodiesel fuels, and a number of different operating modes and cycles.
For 2,2,4-trimethylpentane, hexane, propionaldehyde, and toluene, toxic fractions of VOC were
developed by Sierra Research. Their analysis of CRC E-75 data is described in detail in the
technical report.39 Data from tests using non-conventional diesel fuel (Fischer-Tropsch, bio-
diesel, ethanol-Diesel blends, emulsified fuel, European blends, and other obvious research
fuels) were excluded, as were data from light-duty engines. The fractions are provided in Table
57. Toxic fractions for other compounds in Table 57 were developed by EPA from the E-75
database. We relied on data collected in the United States from heavy-duty diesel engines
running on conventional diesel fuels, collected on test-cycles representative of real world
operation. Some studies reported results on a distance-specific basis (g/mi) whereas others
reported results on a brake-specific basis (g/hp-hr). For both subsets of data, we calculated mean
emissions for each toxic and for VOC, and then calculated mean fractions for each reporting
basis. We then calculated an overall mean fraction using the respective sample sizes to weight
the two fractions.
53
-------
Table 57. Toxic Fractions of VOC for Pre-2007 Diesel Engines.
Pollutant
Toxic fraction
1,3-Butadiene
0.002918
2,2,4-Trimethy lpentane
0.001808
Acetaldehyde
0.035559
Acrolein
0.006622
Benzene
0.007835
Ethyl Benzene
0.002655
Formaldehyde
0.078225
n-Hexane
0.00197
Propionaldehyde
0.00468
Styrene
0.001312
Toluene
0.00433
Xylenes
0.003784
Since extended idle emissions associated with auxiliary power units (APUs) are not subject to
2007 standards, toxic to VOC ratios for pre-2007 diesel engines were used for the APU VOC
toxic emission rates for all model years.40'k
3.2 Polycyclic Aromatic Hydrocarbons
As with gasoline emissions, PAH mass emissions from diesel engines were apportioned into
gaseous and particulate phases, using a single set of allocation factors for all temperature
conditions. The partitioning factors for diesel PAHs were developed by Sierra Research41 using
estimates from EPA's SPECIATE 4.2 database42 and information on compounds' physical and
chemical properties. The allocations from SPECIATE were based on medium-duty diesel engine
data.43 The phase-partitioning factors are shown in Table 58. Compared to the partitioning for
gasoline (Table 47), the fraction of PAH in the particulate phase is higher for diesel emissions,
which is consistent with the higher concentrations of particles in diesel exhaust. However, it
should be noted that the data used represent partitioning in the sampled diluted exhaust, which is
not representative of partitioning in the atmosphere.
k MOVES2014 used the 2007+ exhaust diesel value for 2007+ model year APU exhaust. MOVES2014a has corrected the toxic
ratios to be consistent with the documentation, and applies the pre-2007 exhaust toxic fractions to all model years of APU
exhaust.
54
-------
Table 58. Phase-Partition Fractions for Emissions of Polycyclic Aromatic Hydrocarbons from Diesel
Engines.
PAH species
Molar Mass
(g/mol)
Phase Fraction
Gaseous
Particulate
Naphthalene
128
1.0
0.0
Acenaphthylene
152
1.0
0.0
Acenapthene
154
1.0
0.0
Fluorene
166
0.785
0.215
Anthracene
178
0.534
0.466
Phenanthrene
178
0.665
0.335
Fluoranthene
202
0.484
0.516
Pyrene
202
0.448
0.552
Benz(a)anthracene
228
0.277
0.723
Chrysene
228
0.177
0.823
Bcnzo(fl)pyrcnc
252
0.0
1.0
B c nzo (h) fl no ra lit lie 11c
252
0.0
1.0
B c iizo (k) fl no ra ntlie nc
252
0.0
1.0
B c nzo (w/? /) pc r\lc 11c
276
0.227
0.773
Indeno( 1,2,3 -6c/)pyrcnc
276
0.0
1.0
Dibenzo(a/0 anthracene
278
0.0
1.0
Emissions of PAH in the gaseous and particulate phases were estimated as fractions of total
VOC and OC2.5, respectively. Toxic fractions were calculated using results from the E-75
database. For the particulate phase, a fraction was first calculated with respect to total PM2.5, and
then converted to a fraction of total OC2.5 using estimates of OC as a fraction of total PM2.5.
Note that the OC:PM fractions differed by emissions process, with separate fractions applied for
start, running and extended-idle emissions.
In estimating fractions, we relied on data collected in the United States on heavy-duty diesel
engines running on conventional diesel fuels, measured on test-cycles representative of real
world operation. It should be noted that for some compounds, substantially more data were
available than for others; thus the level of confidence in emission rates varies among individual
compounds. For instance, while data from 66 tests were available for acenaphthene, data from
only two tests were available for dibenz(c//?)anthracene. Table 59 shows fractions for PAH
emissions relative to OC and VOC, by emissions process.
55
-------
Table 59. Toxic Fractions for PAH Species, by Phase and Process, for pre-2007 Diesel Vehicles
PAH
PAH/VOC
PAH/OC2.5
Start/Idle
Running
Extended Idle
Naphthalene
9.05 xlO"3
0.0
0.0
0.0
Acenaphthylene
5.01 xlO"4
0.0
0.0
0.0
Acenaphthene
2.98xl0"4
0.0
0.0
0.0
Fluorene
4.85 xlO"4
2.80xl0"4
8.49xl0"4
2.54xl0"4
Anthracene
2.35xl0"4
1.63 xlO"4
4.94xl0"4
1.48xl0"4
Phenanthrene
7.08xl0"4
6.44 xlO"4
1.96xl0"3
5.86xl0"4
Fluoranthene
3.55xl0"4
6.24 xlO"4
1.90X10"3
5.68xl0"4
Pyrene
4.27 xlO"4
9.02 xlO"4
2.74xl0"3
8.21xl0"4
Benzo(a)anthracene
4.36xl0"5
3.23 xlO"4
9.81xl0"4
2.94xl0"4
Chrysene
1.70 xlO"5
2.04 xlO"4
6.20xl0"4
1.86xl0"4
Bcnzo(fl)pyrcnc
0.0
1.21 xlO"4
3.69xl0"4
1.10 xlO"4
B c nzo (h) fl no ra lit lie nc
0.0
3.60 xlO"5
1.10 xlO"4
3.28xl0"5
B c nzo (k) fl no ra nt lie nc
0.0
5.08xl0"6
1.54xl0"5
4.62xl0"6
B c nzo (w/? /) pc rv lc nc
8.3xl0"7
5.78xl0"6
1.75xl0"5
5.26xl0"6
Indeno( 1,2,3 -6c/)py rene
0.0
9.24 xlO"6
2.81xl0"5
8.41xl0"6
D ibc nz( ah)n nthracc nc
0.0
4.85 xlO"6
1.47 xlO"5
4.41xl0"6
The PAH Toxic fractions in Table 59 are applied to exhaust emission for 2006 and earlier model
year diesel vehicles in MOVES. The extended idle toxic fractions are applied to auxiliary power
unit (APUs) exhaust for all model year vehicles in MOVES (1960-2050), because the APUs are
not subject to the same control as exhaust from the highway engines.
3.3 Metals
Emission rates for selected metals representing pre-2007 heavy-duty diesel engines were based
on data from the CRC E-75 program, with the exception of rates for hexavalent chromium,
mercury and arsenic. The hexavalent chromium emission rate was obtained by multiplying the
gasoline vehicle emission rate by the ratio of total chromium in diesel exhaust to that in gasoline
exhaust. The total chromium estimates came from the previously cited CRC E-75 and Kansas
City test programs, respectively. More details are provided in Appendix A. The pre-2007 diesel
emission rate for arsenic is the same as for gasoline vehicles and obtained from the same study
(see Table 51). It does not vary with emission control technology. The mercury emission rates
for pre-2007 diesels is calculated from emission tests conducted on two heavy-duty diesel
vehicles, as documented in Appendix B. Table 60 provides metal emission factors for pre-2007
diesel vehicles.
56
-------
Table 60. Emission Rates for Selected Metals for Pre-2007 Diesel Vehicles
Pollutant
Emission Rate (g/mi)
Chromium VI
2.0xl0"8
Manganese
8.0xl0"6
Nickel
1.4xl0"5
Mercury, Elemental Gaseous Phase
6.2xl0"9
Mercury, Reactive Gaseous Phase
3.2xl0"9
Mercury, Particulate Phase
1.6xl0"9
Arsenic
2.3xl0"6
3.4 Dioxins and Furans
To represent emissions of dioxins and furans from pre-2007 heavy-duty diesel engines,
emissions rates for 17 congeners were calculated from the results of an EPA diesel dioxin/furan
study of legacy engines.44 In this study, dioxin emissions from three heavy-duty engines
manufactured prior to 1994 were measured. These engines included a 1985 GM 6.2 L, a 1987
Detroit Diesel 6V92 and 1993 Cummins L10. The emission factors in mg/mi are shown in Table
61. Since these engines are older than most of the pre-2007 fleet, dioxin emissions for pre-2007
engines may be overestimated.
Table 61. Emission Rates for Dioxin/Furan Congeners for Pre-2007 Diesel Vehicles
Congener
Emission Rate
(mg/mi)
2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD)
2.23 x 10"10
1,2,3,7,8-Pentachlorodibenzo-p-dioxin
0.0
1,2,3,4,7,8-Hexachlorodibenzo-p-dioxin
0.0
1,2,3,6,7,8-Hexachlorodibenzo-p-dioxin
1.03 x 10"10
1,2,3,7,8,9-Hexachlorodibenzo-p-dioxin
4.78 x 10"10
1,2,3,4,6,7,8-Heptachlorodibenzo-p-dioxin
4.18 x 10"9
Octachlorodibenzo-p-dioxin
1.61 x 10"8
2,3,7,8-Tetrachlorodibenzofuran
6.50 x 10"9
1,2,3,7,8-Pentachlorodibenzofuran
1.39 x 10"9
2,3,4,7,8-Pentachlorodibenzofuran
2.23 x 10"9
1,2,3,4,7,8-Hexachlorodibenzofuran
8.02 x 10"10
1,2,3,6,7,8-Hexachlorodibenzofuran
4.24 x 10"10
1,2,3,7,8,9-Hexachlorodibenzofuran
0.0
2,3,4,6,7,8-Hexachlorodibenzofuran
3.03 x 10"10
1,2,3,4,6,7,8-Heptachlorodibenzofuran
2.16 x 10"9
1,2,3,4,7,8,9-Heptachlorodibenzofuran
0.0
Octachlorodibenzofuran
1.85 x 10"9
57
-------
4 Diesel Exhaust: MY 2007 and later
4.1 Volatile Organic Compounds
For heavy-duty diesel engines manufactured in 2007 and later, advanced emission controls
change the composition of VOCs. For these engines, we relied on speciated emissions data from
the Advanced Collaborative Emissions Study (ACES), directed by the Health Effects Institute
and Coordinating Research Council, with participation from a range of government and private-
sector sponsors.45 In this study detailed emissions measurements were performed on four engines
operated on low-sulfur diesel fuel over several test cycles. We made use of data from the 16-hour
transient cycle which is composed of FTP and CARB 5-Mode cycles, developed specifically to
gain sufficient mass of toxics emitted at low concentrations, and to capture diesel particulate
filter regeneration events. The ACES measurements for the selected VOC emissions in MOVES
were background corrected using background dilution air.45 Toxic fractions of VOC calculated
from the ACES data are provided in Table 62.
Table 62. Toxic Fractions of VOC for 2007 and later Diesel Vehicles.
Pollutant
Toxic fraction
1,3-Butadiene
0.00080
2,2,4-Trimethylpentane
0.0078
Acetaldehyde
0.06934
Acrolein
0.00999
Benzene
0.01291
Ethyl Benzene
0.0063
Formaldehyde
0.21744
N-Hexane
0.0054
Propionaldehyde
0.0031
Styrene
0.00000
Toluene
0.03
Xylenes
0.038
4.2 Poly cyclic Aromatic Hydrocarbons
For heavy-duty diesels manufactured in 2007 and later, advanced emission controls reduce the
total mass of PAH emitted and change the composition of these compounds. For these engines,
we relied on speciated emissions data from the ACES study. The PAH emissions measured in the
ACES study were uncorrected for background concentrations.45 Toxic fractions applicable to
these engines are shown in Table 63, in which the fractions are differentiated by phase but not by
emissions process. We used the same phase fractions presented in Table 58. For the particulate
phase, a single fraction is provided for all processes (similar to HC) because the OC/PM fraction
in MOVES for 2007+ diesel is a single fraction for all emission processes. The OC/PM fraction
is derived from measurements made on a 16-hour drive cycle that comprises multiple driving
modes, as documented in the MOVES2014 TOG and PM Speciation Report.20
58
-------
Table 63. Toxic Fractions for Polycyclic Aromatic Compounds, by Phase, for 2007 and later Diesel Vehicles
PAH
Gaseous Phase
(PAH/VOC)
Particulate Phase
(PAH/OC2.5)
Naphthalene
1.63 xlO"2
0.0
Acenaphthylene
8.53 xlO"5
0.0
Acenaphthene
5.26xl0"5
0.0
Fluorene
1.96 xlO"4
2.41xl0"4
Anthracene
3.04 xlO"5
1.19xl0"4
Phenanthrene
8.51 xlO"4
1.92xl0"3
Fluoranthene
4.57xl0"5
2.18xl0"4
Pyrene
3.79 xlO"5
2.09xl0"4
Benzo(a)anthracene
3.00 xlO"7
3.58xl0"6
Chrysene
5.00xl0"7
1.12 xlO"5
Benzo(a)pyrene
0.0
1.48xl0"5
Benzo(b)fluoranthene
0.0
6.27 xlO-6
Benzo(k)fluoranthene
0.0
6.27 xlO-6
B enzo (ghi)pery lene
2.00 xlO"7
8.96xl0"7
Indeno( 1,2,3 -cd)pyrene
0.0
2.24xl0"6
Dibenz(a,h)anthracene
0.0
4.48xl0"6
4.3 Metals
Emissions rates for manganese and nickel representing diesel engines manufactured since 2007
were developed using data from the ACES program. The ACES metal emission rates were
uncorrected for background concentrations.45 The emission rate for arsenic is identical to the
emission rate used for gasoline vehicles and pre-2007 diesels (Table 51, page 47). The emission
rates for mercury are the same as those derived for pre-2007 diesel engines, as discussed in
Appendix B. The hexavalent chromium emission rate was obtained by multiplying the gasoline
vehicle emission rate by the ratio of total chromium from diesel and gasoline engines. The total
chromium estimates came from the previously cited Kansas City and ACES test programs,
respectively. More details are provided in Appendix A. Metal emission rates are presented in
Table 64.
Table 64. Emission Rates for Metals, for 2007 and Later Diesel Vehicles
Pollutant
Emission Rate (g/mi)
Chromium VI
5.8xl09
Manganese
5.5xl0"7
Nickel
6.5xl0"7
Mercury, Elemental Gaseous Phase
6.2xl0"9
Mercury, Reactive Gaseous Phase
3.2xl0"9
Mercury, Particulate Phase
1.6xl0"9
Arsenic
2.3xl0"6
59
-------
4.4
Dioxins and Furans
The data used to calculate the emission rates for engines manufactured in 2007 and later were
obtained from the EPA diesel dioxin study of 2007 and later engines.46 The results represent
measurements of transient tests conducted on a MY2008 Cummins ISB engine over 48 replicates
on the FTP cycle in a 1:23 cold:hot start ratio, combined with several emission-control
technologies. To represent emissions from engines manufactured between 2007-2009 the results
for the diesel oxidation-catalyst plus catalyzed diesel particulate filter were used. For engines
manufactured in 2010 and later, the results for the diesel oxidation catalyst plus catalyzed diesel
particulate-filter coupled with flow-through copper zeolite selective catalytic reduction and urea
and ammonia slip catalyst were used. Rates are presented in Table 65.
Table 65. Emission Rates for Dioxins and Furans, for 2007 and Later Diesel Vehicles (mg/mi)
Congener
2007 - 2009
2010 and later
2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD)
0.0
0.0
1,2,3,7,8-Pentachlorodibenzo-p-Dioxin
0.0
0.0
1.2.3.4.7.8-Hcxachlorodibcn/o-/)-Dioxin
0.0
0.0
1.2.3.6.7.8-Hc\achlorodibcn/o-/)-Dioxi n
0.0
0.0
1.2.3.7.8.9-Hcxachlorodibcnzo-/?-Dioxin
4.11 x 10"11
0.0
1,2,3,4,6,7,8-Heptachlorodibenzo-p-Dioxin
2.58xlO"10
1.05xl0"9
Octachlorodibenzo-p-dioxin
9.30xl0"10
6.98xl0"9
2,3,7,8-Tetrachlorodibenzofuran
0.0
5.09xl0"n
1,2,3,7,8-Pentachlorodibenzofuran
0.0
1.07X10-10"
2,3,4,7,8-Pentachlorodibenzofuran
6.30X10"11
3.24xlO"10
1,2,3,4,7,8-Hexachlorodibenzofuran
0.0
2.20xl0"10
1,2,3,6,7,8-Hexachlorodibenzofuran
0.0
2.43xlO"10
1,2,3,7,8,9-Hexachlorodibenzofuran
0.0
0
2,3,4,6,7,8-Hexachlorodibenzofuran
0.0
1.80xl0"10
1,2,3,4,6,7,8-Heptachlorodibenzofuran
3.00xl0"10
9.94xlO"10
1,2,3,4,7,8,9-Heptachlorodibenzofuran
0.0
5.81xl0"n
Octachlorodibenzofuran
7.06xl0"10
1.74xl0"9
60
-------
5 Compressed Natural Gas (CNG) Transit Bus Exhaust
MOVES2014 estimates emissions of toxics from transit buses fueled by compressed natural gas.
This section describes the development of toxic emission inputs for this class of vehicles.
5.1 Volatile Organic Compounds
We used speciated hydrocarbon measurements sponsored by the California Air Resources
Board.47 These measurements were taken on a 2000 MY Detroit Diesel Series 50G engine with
and without an oxidation catalyst, measured on the Central Business District (CBD) cycle. As
discussed in the MOVES2014 heavy-duty emission rates report59, we used the uncontrolled
results to represent speciation from pre-2002 CNG transit buses, and the results with oxidation-
catalyst to represent 2002 and later buses. The use of the CBD cycle is also consistent with the
results used for criteria-pollutant emissions.
The toxic fractions of VOC derived from this set of measurements are displayed in Table 66. The
total VOC emission rates are reduced by 70% from pre-2002 levels. As shown in the table,
formaldehyde emissions are preferentially reduced by the oxidation catalyst. Formaldehyde
contributes over 50% of the VOC emissions for the uncontrolled CNG bus, but only 16.2% of
the VOC emissions for the CNG bus equipped with an oxidation catalyst. The MOVES toxics
not measured in this study are assumed to be negligible, and are modeled as 0.
Table 66. Toxic Fractions of VOC for CNG Transit Buses.
No control
(pre-2002)
With oxidation
catalyst (2002+)
1,3 Butadiene
0.000234
0.0
Benzene
0.00135
0.00253
Toluene
0.000691
0.00786
Ethylbenzene
0.0000841
0.00131
Xylenes
0.000823
0.00634
Formaldehyde
0.517
0.162
Acetaldehyde
0.0305
0.138
Acrolein
0.00235
0.0
Propionaldehyde
0.0153
0.0
5.2 Poly cyclic Aromatic Hydrocarbons
The PAH toxic fractions for compressed natural gas are derived from tests on a MY2000 DDC
Series 50G engine on a New Flyer CNG transit bus tested by the California Air Resources Board
(CARB).48 This engine had no catalyst, but the emission fractions are used to represent both
catalyst and non-catalyst engines. Emissions were measured in two stages (the bus was re-tested
after 3 months of service in the Los Angeles County Metropolitan Transit Authority). The PAH
emissions were measured in the semi-volatile phase using PUF-XAD, and measured in the
particulate phase on Teflon-coated glass-fiber filters. VOC emissions are derived from the
NMHC and speciated hydrocarbon emissions. The OC emissions rates were provided to EPA by
CARB. We estimated the volatile PAH emissions by calculating PAH/VOC fractions from the
PUF-XAD measurements, and particle-phase PAH/OC fractions using the filter-based
61
-------
measurements for both stages of the study. For use in MOVES, we averaged the ratios estimated
from both stages of the testing. The average ratios are displayed in Table 67.
Table 67. PAH Fractions of Volatile Organic Carbon (Volatile PAHs), and of Organic Carbon (Particle-Phase
for CNG Transit Buses
Compound
VOC fraction
OC fraction
Naphthalene
9.554xl0"6
2.114xl0"5
Acenaphthylene
4.230xl0"6
ND
Acenaphthene
1.243 xlO"6
1.886xl0"5
Fluorene
2.986xl0"6
3.301xl0"5
Anthracene
1.164xl0"6
1.644 xlO"6
Phenanthrene
8.356xl0"6
2.043 xlO"5
Fluoranthene
1.936xl0"6
2.874xl0"5
Pyrene
3.743 xlO"6
5.350xl0"5
Benz(a)anthracene
1.682xl0"7
9.390xl0"6
Chrysene/triphenylene
2.441xl0"7
1.911xl0"5
Bcnzo(fl)pyrcnc
ND
ND
B c nzo (/>) fl no ra lit lie 11c
ND
ND
B c iizo (/c) fl no rant lie 11c
ND
ND
Indeno( 1,2,3 -«/)pyrcnc
ND
ND
B c rizo (ghi) pc rslc 11c
ND
5.502xl0"6
D ibc nz( ah)i\ nthracc 11c
ND
ND
ND = not detected, fractions set to 0.
5.3 Metals
We used the nickel emission rates reported from an uncontrolled 2000 MY DDC Series 50G.49
We used the uncontrolled bus to be consistent with the PM2.5 speciation profile. The hexavalent
chromium emission rate was obtained by multiplying the gasoline emission rate by the ratio of
total chromium from the DDC Series 50G CNG engine and total chromium from gasoline
engines in the previously cited Kansas City test program. More details are provided in Appendix
A.
Results for the other metals predicted by MOVES were not available in the published literature.
Thus, we used the same emission rates as for gasoline vehicles. The rates are presented in Table
68.
62
-------
Table 68. Metal Emission Rates and Sources used for CNG Transit Buses
Pollutant
Emission Rate (g/mi)
Source
Chromium 6+
2.1xlO"10
University of Wisconsin
(2010) and Okamoto et al.
(2006)
Manganese
1.33xl0"6
Same as gasoline
Nickel
l.OOxlO"8
Okamoto et al. (2006)
Elemental Gas Phase Hg
l.lOxlO"7
Same as gasoline
Reactive Gas Phase Hg
9.90xl0"9
Same as gasoline
Particulate Hg
4.00xl0"10
Same as gasoline
Arsenic
2.30xl0"6
Same as gasoline
5.4 Dioxins and Furans
No published dioxin and furan emission rates for CNG vehicles were available. Thus, we are
using the dioxin emission rates for gasoline reported in Table 54.
63
-------
6 Evaporative Emissions
Emissions of toxics emitted through evaporation of unburned fuel are estimated as fractions of
total evaporative VOC. MOVES estimates toxic emission ratios for each evaporative processes
from gasoline vehicles (including gasoline-ethanol blends), and for refueling emissions from
diesel vehicles. Currently, MOVES does not estimate evaporative emissions (e.g. refueling
natural gas leaks) from CNG vehicles as discussed in the evaporative emission report.50 This
section documents the source of the toxic ratios used for evaporative emissions from gasoline
and diesel vehicles.
6.1 Gasoline Vehicles
6.1.1 Vapor Venting, Fuel Leaks, and Refueling Emission Processes
MOVES estimates evaporative emissions from gasoline vehicles using toxic fractions that
pertain to the evaporative emission processes. In addition, the toxic fractions for some
compounds are estimated as complex fractions based on fuel properties such as oxygenate
content and vapor pressure. For other compounds, simple fractions are estimated. For the
compounds modeled, fraction types and data sources are summarized in Table 69.
Expressions used to generate complex fractions were adapted from those used in MOBILE6.2.51
These equations were adapted to compensate for a lack of data from newer vehicles collected in
the context of appropriate experimental designs. However, as the conceptual basis for modeling
evaporative emissions has changed in MOVES, the equations are applied to the emission
processes considered most closely analogous. Thus, equations for hot soak in MOBILE6.2 are
used for vapor venting and refueling vapor loss, and equations for running loss are used for fuel
leaks and refueling spillage loss. The equations are applied for fuels containing up to 20%
ethanol, and are presented in Table 70. MOVES has fields for evaporative naphthalene, but all
values in the model are zero. E0 data used for MOBILE6.2 had very low but detectable
naphthalene, and it is often measured at very low levels in gasoline. However, we decided to not
include naphthalene emissions from evaporative processes in MOVES since it is inconsistently
measured in detectable quantities in evaporative emission testing
Simple fractions for other air toxics in evaporative non-permeation emissions were obtained
from profiles developed for EPA by Environ Corporation, using data from the Auto/Oil program
conducted in the early 1990's.52 The fractions for these compounds are the same for all pollutant
processes (except permeation) and are presented in Table 71.
The ratios for 10% ethanol are used for all fuels with greater than or equal to 5% ethanol and less
than 12%). Conventional gasoline ratios are also used for MTBE oxygenated gasoline.
For vehicles operating on fuels containing 15% ethanol (El 5), no data describing evaporative
emissions are available. For the vapor-venting and spillage emission processes, emission rates
calculated from E15 and E10 fuel speciation data from the EPAct Program were used to adjust
the E10 evaporative emissions speciation.53 Resulting toxic fractions are provided in Table 71.
For vehicles containing 20% ethanol, toxic fractions were developed for fuel speciation profiles
created from data collected in the EPAct program. Average fractions by weight were calculated
as a composite of data from the seven E20 blends included in the fuel matrix. Resulting fractions
are shown in Table 71.
64
-------
For vehicles operating on fuels containing high levels of ethanol, ranging from 70 to 100%, the
toxic fractions were developed using results of two-day diurnal tests on four 2007 model year
flex-fuel vehicles from CRC E-80 program.29 Following typical speciation procedures, the
fraction of each compound in a test was first calculated by dividing its emission rates for each
compound by the sum of all rates for that test. The percentages for each compound were then
averaged across all tests to form the composite profile. The resulting fractions are presented in
Table 71.
Table 69. Data Sources and Estimation Methods Used in Estimation of Toxic Fractions for Evaporative
VOCs
Compound
Process
Fraction Type
Basis for Estimation
Benzene
Vapor venting/refueling (vapor)
complex
Adapted from MOBILE6.2
Fuel leaks/spillage
complex
Adapted from MOBILE6.2
MTBE
Vapor venting/refueling (vapor)
complex
Adapted from MOBILE6.2
Fuel leaks/spillage
complex
Adapted from MOBILE6.2
2,2,4-trimethylpentane
All (except permeation)
simple
Speciation profile
Ethylbenzene
All (except permeation)
simple
Speciation profile
N-Hexane
All (except permeation)
simple
Speciation profile
Propionaldehyde
All (except permeation)
simple
Speciation profile
Toluene
All (except permeation)
simple
Speciation profile
Xylenes
All (except permeation)
simple
Speciation profile
Ethanol
All (except permeation)
simple
Speciation profile
Table 70. Complex Fractions of VOC for Evaporative Emissions of Two Compounds Applied for Fuels
Containing up to 10% Ethanol.
Pollutant
Process
Equation for Toxic Fraction
Benzene
Vapor venting/Refueling (vapor)
(-0.03420*OXY - 0.080274*RVP + 1.4448)*BNZ/100
Fuel Leaks/Spillage
(-0.03420*OXY - 0.080274*RVP + 1.4448)*BNZ/100
MTBE
Vapor Venting/Refueling (vapor)
(24.205 - 1.746*RVP)*MTBE/1000
Fuel Leaks/Spillage
(17.8538 - 1.6622*RVP)*MTBE/1000
OXY = oxygen content (wt%)
RVP = Reid Vapor Pressure (psi)
BNZ = benzene content (vol.%)
MTBE = methyl-tertiary-butyl ether content (vol.%).
65
-------
Table 71. Toxic Fractions for Evaporative VOC Emissions, for Vapor-venting and Refueling-spillage
Processes.
Pollutant
Ethanol Level
0.0% (EO)
10% (E10)
15% (El5)
20% (E20)
70-100% (E85)1
Ethanol
0.00000
0.11896
0.1935
0.2227
0.61042
2,2,4-Trimethy lpentane
0.01984
0.03354
0.05313
0.0430
0.00830
Ethyl Benzene
0.02521
0.01721
0.01662
0.0155
0.00124
N-Hexane
0.02217
0.02536
0.007478
0.0186
0.01276
Toluene
0.09643
0.14336
0.1406
0.0874
0.01608
Xylenea
0.07999
0.06423
0.05735
0.0711
0.00733
Benzene
Table 70
0.02758
0.0073
0.00664
6.1.2 Permeation
The composition of VOCs emitted through permeation differs substantially from that of
hydrocarbons emitted through other processes. Work to better characterize these permeation
emissions was recently conducted by Southwest Research Institute for EPA and the Coordinating
Research Council in the CRC E-77-2b and E-77-2c test programs.54'55 Data from 3-day diurnal
tests on vehicles meeting Tier 1 and near-zero evaporative emission standards were used.
Fractions representing emissions of toxic compounds relative to total VOC were estimated for
EO, E10 and E20 fuels by averaging data from fuel formulations with varying vapor pressures.
Fractions are presented in Table 72, for all compounds except benzene. To estimate toxic
fractions for vehicles operating on fuels containing 15% ethanol, the fractions for E10 and E20
fuels were linearly interpolated for ethanol levels of 15%. Toxic fractions are shown in Table 72.
For benzene, the diurnal emissions equation from MOBILE6.2 was used to calculate the
permeation fraction^benz,permeation, since it accounts for changes in oxygenate, vapor pressure and
fuel benzene levels, as shown in Equation 21.56 However, a study of permeation emissions
suggests that the fraction of benzene from permeation is about 1.77 times higher than the ratio
associated with evaporation.57 Thus the diurnal emissions algorithm was multiplied by 1.77.
/benz,permeation =l-77[(-0.02895 OXY-0.080274 RVP + 1.3 75 8) benz /100] Equation21
In the absence of data on permeation emissions for MTBE, a complex fraction/mtbe,permeation is
calculated using the resting-loss algorithm from MOBILE6.2 ( Equation 22).
/mtbe,permeatlon = (22.198 -1.746RVP)MTBE /1,000 Equation 22
1MOVES2014 incorrectly did not produce NMOG and VOC emissions from evaporative and refueling emissions from vehicles
fueled on E85. Subsequently, it did not produce toxics that were calculated as a fraction of VOC. This error has been corrected in
MOVES2014a.
66
-------
Table 72. Toxic Fractions Representing Permeation Emissions as Components of Total VOC Emissions, by
Ethanol Level (Source: CRC E-77-2b and CRC E-77-2c).
Pollutant
Ethanol Level
0.0% (E0)
10% (E10)
15% (E15)
20% (E20)
70-100% (E85)p
Ethanol
0.000
0.202
0.2694
0.3296
0.61042*
2,2,4-Trimethylpentane
0.036
0.024
0.0172
0.0107
0. 00830*
Ethylbenzene
0.003
0.001
0.0017
0.0019
0.00124*
Hexane
0.050
0.065
0.0472
0.0308
0.01276*
Toluene
0.110
0.101
0.0666
0.0354
0.01608*
Xylene(s)
0.016
0.011
0.0127
0.0140
0.00733*
Benzene
Equation 21
0.0236
0.0244
0.00664*
* Identical to fractions for the vapor-venting process, based on CRC E-80 program (Table 71).
For ethanol levels of 70-100%, no permeation data were available. Thus, the toxic fraction for
non-permeation evaporative emissions was also applied to permeation.
6.2 Diesel Vehicles
For diesel-fueled vehicles, evaporative emissions are estimated for the refueling-spillage process
only. As no results describing the speciation of spilled diesel fuel, we developed toxic fractions
of total VOC based on a diesel "headspace" profile, in which the "headspace" is the empty space
above the liquid fuel in a tank. The profile used was No. 4547 from the SPECIATE database.42
The fractions are shown in Table 73.
Table 73. Toxic Fractions for the fuel-spillage Process, for Diesel fuel.
Pollutant
Toxic fraction
2,2,4-Trimethylpentane
0.00974
Ethyl Benzene
0.00324
N-Hexane
0.01076
Toluene
0.01419
Xylene
0.01222
Benzene
0.00410m
m The benzene emission ratio was missing in the MOVES2014 database. This ratio has been included in MOVES2014a.
67
-------
7 Crankcase Emissions
Crankcase emissions are modeled as a ratio of the exhaust emissions. Discussion of the ratios
used to estimate THC, CO, NOx, and PM crankcase emissions can be found in the light-duty58
and heavy-duty59 emission rate reports. In general, toxic crankcase emissions that are calculated
as a ratio from VOC or from PM are computing as a fraction of the toxic exhaust emissions. The
details on crankcase emissions are discussed in the following sections.
7.1 Volatile Organic Compounds
Table 1 lists the VOC toxics modeled in MOVES2014, which are also modeled from crankcase
emission processes. MOVES2014 models the crankcase emissions from these toxics by
multiplying the exhaust emissions of these species by the THC crankcase emission fraction listed
in the light-duty and heavy-duty emissions reports. For example, the THC crankcase/exhaust
fraction for light-duty gasoline (1969 and later model year) is 0.013. Thus, crankcase emissions
for 1,3-butadiene are calculated as 1.3% of the exhaust emissions of 1,3-butadiene. Similar
calculations are applied to all VOC toxic emissions. The crankcase emission ratios are stored in
the MOVES table crankCaseEmissionratio, which differentiates the factors according to
pollutant, process, model year range, source type and fuel type.
7.2 Poly cyclic Aromatic Hydrocarbons
The PAH fractions for exhaust emissions are also applied to crankcase emissions. The gaseous
PAHs are modeled in a similar fashion as the VOC toxic emissions. The PAH crankcase
emissions are modeled as a fraction of the tailpipe exhaust gaseous PAH emissions, with factors
stored in the crankCaseEmissionRatio table. The PAH crankcase emission factors are the same
as the THC crankcase emission factors (e.g. 0.013 for 1969 and later gasoline vehicles).
To estimate crankcase particulate PAH emissions, MOVES applies the PAH/OC fractions
developed for exhaust emissions to the crankcase OC emissions. The PAH/OC ratios are stored
in the pahParticleRatio table for the crankcase emission processes (15, 16, and 17). The OC/PM
speciation can be substantially different between crankcase emissions and exhaust emissions. For
example, because conventional diesel crankcase emissions has a higher OC/PM composition
than the tailpipe exhaust emissions, MOVES models elevated particulate PAH emissions in
crankcase emissions compared to tailpipe PAH emissions. Research on conventional diesel
vehicles validates that PM emissions from the crankcase are more enriched with PAHs than
emissions from the exhaust.60
7.3 Metal and Dioxin Emissions
MOVES models crankcase metal emissions for the metal species included in the PM2.5 speciation
profiles, such as iron and aluminum. Details on speciation of crankcase emissions are included in
the speciation report.34 MOVES does not produce crankcase emission rates for metals that are
not included in the speciation profiles such as arsenic, mercury and other metals listed in Table 4.
Similarly, MOVES does not estimate dioxin and furan emissions from crankcase emissions. We
are assuming that the emissions from crankcase are negligible compared to exhaust emissions.
68
-------
References
1 Code of Federal Regulations, 40: Chapter 1, Subchapter C, Part 51, Subpart F, 51100, 1986.
2 Chemical Abstracts Service, A division of the American Chemical Society.
http://www.cas.org.
3 USEPA (2014). Air Toxic Emissions from On-road Vehicles in MOVES2014. EPA-420-R-14-
021. Assessment and Standards Division. Office of Transportation and Air Quality. US
Environmental Protection Agency. Ann Arbor, MI. December, 2014.
https ://cfpub. epa.gov/si/si_public_record_report. cfm?dirEntryId=263 653.
4 USEPA (2015). Air Toxic Emissions from On-road Vehicles in MOVES2014. EPA-420-R-15-
021. Office of Transportation and Air Quality. US Environmental Protection Agency. Ann
Arbor, MI. October 2015. https://www.epa.gov/moves/moves-technical-reports.
5 USEPA. (1993). Final Regulatory Impact Analysis for Reformulated Gasoline. December 13,
1993. Available at http://www.epa.gov/otaq/regs/fuels/rfg/.
6 Cook, Richard; Glover, Edward L. (2002). Technical Description of the Toxics Module for
MOBILE6.2 and Guidance on Its Use for Emission Inventory Preparation. EPA420-R-02-
011. Assessment and Standards Division, Office of Transportation and Air Quality, Ann
Arbor, MI. http://www.epa.gov/otaq/models/mobile6/r02029.pdf
7 Wyborny, L. (1998). Methyl Tertiary Butyl Ether (MTBE) Emissions from Passenger Cars.
Draft Technical Report. U. S. Environmental Protection Agency, Office of Mobile Sources.
April, 1998.
8 Sierra Research (2010). Development of Emission Rates for the MOVES Model. Prepared for
the U. S. Environmental Protection Agency by Sierra Research, Sacramento, CA, July 2,
2010. Sierra Research Report No. SR2010-07-01.
9 Southwest Research Institute (2007). Flex Fuel Vehicles (FFVs): VOC/PM Cold Temperature
Characterization When Operating on Ethanol (E10, E70, E85). Prepared for U. S.
Environmental Protection Agency.
10 Environment Canada (2007). Comparison of Emissions from Conventional and Flexible Fuel
Vehicles Operating on Gasoline andE85 Fuels. ERM Report No. 05-039. Emissions
Research Division.
11 Durbin T. D., Miller J. W., Younglove T., Huai T., Cocker K. (2007). Effects of fuel ethanol
content and volatility on regulated and unregulated exhaust emissions for the latest
technology gasoline vehicles. Environmental Science and Technology 41, 4059-4064.
12 USEPA (1995). MOVES Toxic Inputs 5.xls. Office of Research and Development,
unpublished data.
13 USEPA. El5: Frequently Asked Questions, http://epa.gov/otaq/regs/fuels/additive/el5/el5-
faq.htm
14 USEPA (2013). EPAct/V2/E-89: Assessing the Effect of Five Gasoline Properties on Exhaust
Emissions from Light-Duty Vehicles certified to Tier-2 Standards: Final Report on Program
Design and Data Collection. EPA-420-R-13-004. Assessment and Standards Division,
69
-------
Office of Transportation and Air Quality, Ann Arbor, MI; National Renewable Energy
Laboratory, Golden, CO; Coordinating Research Council, Alpharetta, GA. April, 2013.
15 USEPA (2013). Assessing the Effect of Five Gasoline Properties on Exhaust Emissions from
Light-Duty Vehicles certified to Tier-2 Standards: Analysis of Data from EPAct Phase 3
(EPAct/V2/E-89). Final Report. EPA-420-R-13-002. Assessment and Standards Division,
Office of Transportation and Air Quality, Ann Arbor, MI. April, 2013.
16 Neter, J.: Kutner, M.H.; Nachtsheim, C.J.; Wasserman, W. (1996). Applied Linear Statistical
Models. Fourth Ed., Irwin, Chicago. (Section 7.5).
17 West, Brady T.; Welch, Kathleen B.; Galecki, Andrzej T. (2007). Linear Mixed Models: A
Practical Guide Using Statistical Software. Taylor and Francis CRC Press. 376 pp.
18 SAS Institute Inc. (1990). SAS/STAT User's Guide, Version 6, Fourth Ed., Vol. 2. Chapter 25,
The LIFEREG Procedure. Pp. 997-1026. Cary, NC.
19 Neter, J.; Kutner, M.H.; Nachtsheim, C.J.; Wasserman, W. (1996). Applied Linear Statistical
Models. Fourth Ed., Irwin, Chicago.
20 USEPA (2014). Speciation of Total Organic Gas and Particulate Matter Emissions from On-
road Vehicles inMOVES2014. EPA-420-R-14-020. Ann Arbor, MI. October, 2014.
http://www.epa.gov/otaq/models/moves/moves-reports.htm.
21 USEPA (2016). Fuel Effects on Exhaust Emissions from On-road Vehicles in MOVES2014.
EPA-420-R-16-001. Assessment and Standards Division. Office of Transportation and Air
Quality. US Environmental Protection Agency. Ann Arbor, MI. 2016.
https://www.epa.gov/moves/moves-technical-reports.
22 Zarnoch, S.J., and W.A. Bechtold (2000). Estimating mapped-plot forest attributes with
ratios of means. Canadian Journal of Forest Research. 30:688-697.
23 USEPA (2007). Final Regulatory Impact Analysis: Control of Hazardous Air Pollutants from
Mobile Sources. EPA420-R-07-002; U.S. Environmental Protection Agency; Office of
Transportation and Air Quality: Ann Arbor, MI. Chapter 6.11.
http://www.epa.gov/otaq/toxics.htm.
24 U SEP A (2011). Spreadsheet"Bern Buta Adjustments.xlsx." EPA Docket EP A-HQ-OAR-
2011-0135.
25 USEPA (2011). Spreadsheets "Olefins versus Butadiene.xls" and "Bern Buta
Adjustments.xls." EPA Docket EPA-HQ-OAR-2011-0135.
26 USEPA (2009). Exhaust Emission Profiles for EPA SPECIATE Database: Energy Policy Act
(EPAct) Low-Level Ethanol Fuel Blends and Tier 2 Light-Duty Vehicles. EPA-420-R-09-
002. Office of Transportation and Air Quality, Assessment and Standards Division, Ann
Arbor, MI.
27 USEPA (2012). Data Collected in EPAct Fuel Effects Study Pilot Phases. Memorandum
from Aron Butler to Docket EPA-HQ-OAR-2011-0135.
28 Yanowitz, J., Knoll, K., Kemper, J., Luecke, J., and McCormick R. (2013) Impact of
Adaptation on Flex-Fuel Vehicle Emissions When Fueled with E40. Environmental Science
& Technology, 2013, 47(6), 2990-2997.
70
-------
29 Haskew, H.M., and Liberty, T.F. (2011) Exhaust and Evaporative Emissions Testing of
Flexible-Fuel Vehicles. Final Report. Coordinating Research Council. CRC Report No. E-
80, August, 2011. Available at www.crcao.org.
30 Long, T., Herrington, J., Hays, M., Baldauf, R., and Snow, R. (2 01 1) A ir Toxic Emission from
Passenger Cars Operating on Ethanol Blend Gasoline. 104th Air and Waste Management
Association Annual Conference and Exhibition, 2011, #118, 3053-3058.
31 USEPA (2008). Kansas City PM Characterization Study. Final Report. EPA420-R-08-009.
Office of Transportation and Air Quality. Assessment and Standards Division, .
http://www.epa.gov/oms/emission-factors-research/420r08009.pdf.
32 Fujita, E. M.; Campbell, D. E.; Zielinska, B.; Chow, J. C. (2006). Kansas City Gasoline
Vehicle Characterization Study - Effect of Sampling Temperature on Chemical Speciation.
NREL Subcontract Number ACI-5-55500-01. NREL Project ES04-2. May 31, 2006.
33 Hays, M. D., W. Preston, B. J. George, J. Schmid, R. Baldauf, R. Snow, J. R. Robinson, T.
Long and J. Faircloth (2013). Carbonaceous Aerosols Emitted from Light-Duty Vehicles
Operating on Gasoline and Ethanol Fuel Blends. Environ Sci Technol, 47 (24), 14502-
14509. DOI: Doi 10.1021/Es403096v.
34 USEPA (2014). Speciation of Total Organic Gas and Particulate Matter Emissions from On-
road Vehicles inMOVES2014. EPA-420-R-14-020. Ann Arbor, MI. August, 2014.
http://www.epa.gov/otaq/models/moves/documents/420rl4020.pdf.
35 Taylor, M. (2003). Memorandum: Revised HAP Emission Factors for Stationary Combustion
Turbines. Docket ID: OAR-2002-0060-0649. Prepared by Alpha-Gamma Technologies, Inc.
for Sims Roy, EPA OAQPS ESD Combustion Group. August, 2003. Access via
http://www.regulations.gov.
36 Schauer, J. J., Lough, G. C., ShaferM. M., Christensen W. F., Arndt, M. F., DeMinter, J.T.,
Park, J-S. (2006). Characterization of Metals Emittedfrom Motor Vehicles. Health Effects
Institute Research Report Number 133. (http://pubs.healtheffects.org).
37 Gertler, A. W., J. C. Sagebiel, W. A. Dippel and R. J. Farina. (1998). Measurement of dioxin
and furan emission factors from heavy-duty diesel vehicles. J. Air and Waste Manage.
Assoc. 48: 276-278.
38 USEPA (2000). Draft Final Assessment: Exposure and Human Health Reassessment of
2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and Related Compounds. Volume 2: Sources
of Dioxin-like Compounds in the United States. Office of Research and Development,
National Center for Environmental Assessment, Washington, D.C. Report No. EPA/600/P-
OO/OOlBb, September 2000.
39 Hsu ,Y., and Mullen, M. (2007). Compilation of Diesel Emissions Speciation Data. Prepared
by E. H. Pechan and Associates for the Coordinating Research Council. CRC Contract No.
E-75, October, 2007. Available atwww.crcao.org.
40 USEPA (2011). Final Rulemaking: Greenhouse Gas Emissions Standards and Fuel Efficiency
Standards for Medium- and Heavy-Duty Engines and Vehicles. Final Regulatory Impact
Analysis, Chapter 5 (RIA) (PDF) (553 pp, EPA-420-R-11-901, August 2011)
71
-------
41Sierra Research, 2010. Development of Emission Rates for the MOVES Model. Prepared for
the U. S. Environmental Protection Agency by Sierra Research, Sacramento, CA, July 2,
2010. Sierra Research Report No. SR2010-07-01.
42USEPA (2009). SPECIATE 4.2 Speciation Database Development Documentation, Office of
Research and Development. Report No. EPA/600-R-09/038, June 2009.
43 Schauer, J.J., M.J. Kleeman, G.R. Cass, and B.R.T. Simoneit. (1999) Measurement of
Emissions from Air Pollution Sources, 2. C1-C30 Organic Compounds from Medium Duty
Diesel Trucks. Environmental Science and Technology, vol. 33, no. 10, pp. 1578-1587,
1999.
44 Laroo, A. Christopher, Charles R. Schenk, L. James Sanchez, Joseph McDonald, Peter L.
Smith (2012). Emissions of PCDD/Fs, PCBs, and PAHs from legacy on-road heavy-duty
diesel engines. Chemosphere 89: 1287 - 1294.
45Khalek, I., Bougher, T., and Merritt, P. M. (2009). Phase 1 of the Advanced Collaborative
Emissions Study. Prepared by Southwest Research Institute for the Coordinating Research
Council and the Health Effects Institute, June 2009. Available at www.crcao.org.
46 Laroo, A. Christopher, Charles R. Schenk, L. James Sanchez, and Joseph McDonald. (2011).
Emissions of PCDD/Fs, PCBs, and PAHs from a Modern Diesel Engine Equpped with
Catalyzed Emission Control Systems. Environmental Science & Technology 45: 6420 -
6428.
47 Ayala, A., Gebel, M., Okamoto, R., Rieger, P. et al. (2003). "Oxidation Catalyst Effect on
CNG Transit Bus Emissions," Society of Automotive Engineers, SAE Technical Paper
2003-01-1900, 2003. (http://papers.sae.org/2003-01-1900).
48 Kado, N. Y.; Okamoto, R. A.; Kuzmicky,, P. A.; Kobayashi, R.; Ayala,A.; Gebel, M. E.;
Rieger, P. L.;Maddox, C.; Zafonte, L.; (2005). Emissions of Toxic Pollutants from
Compressed Natural Gas and Low Sulfur Diesel-Fueled Heavy-Duty Transit Buses Tested
over Multiple Driving Cycles. Environmental Science & Technology 2005 39 (19), 7638-
7649
49 Okamoto, R. A.; Kado, N. Y.; Ayala, A.; Gebel, M.; Rieger, P.; Kuzmicky, P. A.; Kobayashi,
R.; (2004) Chemical and Bioassay Analyses of Emissions from Two CNG Buses with
Oxidation Catalyst, http://www.arb.ca.gov/research/veh-emissions/cng-diesel/cng-
diesel.htm.
50 USEPA (2014). Evaporative Emissions from On-road Vehicles inMOVES2014. EPA-420-R-
14-014. Assessment and Standards Division. Office of Transportation and Air Quality. US
Environmental Protection Agency. Ann Arbor, MI. September, 2014.
https://www.epa.gov/moves/moves-technical-reports.
http://www.epa.gov/otaq/models/moves/documents/420rl4014.pdf
51 USEPA (2002). Technical Description of the Toxics Module for MOBILE6.2 and Guidance
on Its Use for Emission Inventory Preparation. EPA420-R-02-011. U.S. Environmental
Protection Agency; Office of Transportation and Air Quality: Ann Arbor, MI. Available at
http ://www. epa. gov/otaq/m6. htm.
72
-------
52C.E. Lindhjem (2008). Emission Profiles for EPA SPECIATE Database, EPA Contract No.
EP-C-06-094, Work Assignment No. 1-7, ENVIRON International Corporation, January 31,
2008.
53 USEPA (2013). EPAct/V2/E-89: Assessing the Effect of Five Gasoline Properties on Exhaust
Emissions from Light-Duty Vehicles Certified to Tier 2 Standards: Final Report on Program
Design and Data Collection. EPA Report No. EPA-420-R-13-004, March, 2013.
54Southwest Research Institute (2009). Evaporative Emissions Breakdown Including
Permeation Effects and Diurnal Emissions on Aging Enhanced Evaporative Emissions
Certified Vehicles (CRC E-77-2b). Prepared by Harold M. Haskew and Thomas F. Liberty,
Harold Haskew and Associates, Inc.; submitted to U. S. EPA, December, 2009. Available at
www.crcao.org.
55 Coordinating Research Council (2010). Study to Determine Evaporative Emission Breakdown,
Including Permeation Effects and Diurnal Emissions, Using E20 Fuels on Aging Enhanced
Evaporative Emissions Certified Vehicles. Report No. E-77-2c.
http://www.crcao.org/reports/recentstudies2011/E-77-2c/E-77-
2c%20Final%20Report%20for%20sure%201-28-11 .pdf.
56 USEPA (2002). Technical Description of the Toxics Module for MOBILE6.2 and Guidance
on Its Use for Emission Inventory Preparation. EPA420-R-02-011. U.S. Environmental
Protection Agency; Office of Transportation and Air Quality: Ann Arbor, MI. Available at
http ://www. epa. gov/otaq/m6. htm.
57Haskew, H. M., Liberty, T. F., and McClement, D. (2004) Fuel Permeation from Automotive
Systems. Prepared for the Coordinating Research Council by Harold Haskew and
Associates and Automotive Testing Laboratories, Inc. September 2004. CRC Project No.
E-65. http://www.crcao.com.
58 USEPA (2015). Exhaust Emission Rates for Light-Duty On-road Vehicles inMOVES2014.
EPA-420-R-15-005. Assessment and Standards Division. Office of Transportation and Air
Quality. US Environmental Protection Agency. Ann Arbor, MI. October, 2015.
https://www.epa.gov/moves/moves-technical-reports.
59 USEPA (2015). Exhaust Emission Rates for Heavy-Duty On-road Vehicles inMOVES2014.
EPA-420-R-15-015a. Assessment and Standards Division. Office of Transportation and Air
Quality. US Environmental Protection Agency. Ann Arbor, MI. November, 2015.
https://www.epa.gov/moves/moves-technical-reports.
60 Zielinska, B., D. Campbell, D. R. Lawson, R. G. Ireson, C. S. Weaver, T. W. Hesterberg, T.
Larson, M. Davey and L. J. S. Liu (2008). Detailed Characterization and Profiles of
Crankcase and Diesel Particulate Matter Exhaust Emissions Using Speciated Organics.
Environ Sci Technol, 42 (15), 5661-5666. DOI: 10.1021/es703065h.
73
-------
Appendix A Development of Motor Vehicle Emission Factors for
Chromium
The emission rate for gasoline vehicles and trucks in MOVES 2010b (EPA's Motor Vehicle
Estimation Simulator) for hexavalent chromium, or chromium 6+ (Cr(VI)) is 8.9xl0"7
grams/mile.1 This gasoline emission factor (EF) remained unchanged from the value used in
NMIM (National Mobile Inventory Model) and was obtained from a paper by Ball, 1997.2 The
Ball (1997) test program and other testing from motor vehicles included only total chromium
measurements, therefore Cr(VI) concentrations were estimated based on combustion data from
stationary combustion turbines that burn diesel fuel which showed eighteen percent of chromium
was hexavalent.3
An updated total chromium emission rate for gasoline vehicles and trucks was recently
developed for MOVES based on data from the Kansas City test program.4 The Kansas City test
program sampled 99 vehicles for chemical composition from which a total chromium emission
factor of 4.07xl0"6 grams/mile was developed.5 This average grams/mile rate was calculated by
averaging the metal measured in Bag 2 of the LA92 driving schedule test (described below), with
a weighted-average computed using vehicle miles traveled (VMT).
In 2010, the EPA's National Vehicle and Fuel Emissions Laboratory (NVFEL) collected
particulate matter (PM) and volatile organic compound (VOC) exhaust samples, as well as CO,
NOx, CO2, and CH4 samples from a 2008 3.5L V6 Chevrolet Impala flex fuel light-duty gasoline
vehicle. This testing also included direct Cr(VI) measurements.
The Impala had a beginning odometer reading of 38,934 miles and was tested using E10
gasoline. The vehicle test procedure used four sample bags and the LA92 "unified"
dynamometer driving schedule.6 The bags in this study represent the following conditions:
Bag 1 - concentrated cold start compared to FTP (Federal Test Procedure); short
distance, low speeds
Bag 2 - hot and running; longer distance and higher speeds than FTP (represents realistic
real world driving)
Bag 3 - hot start; short distance, low speeds
Bag 4 - hot and running; long distance
PM was collected on four (labeled A-D) pre-cleaned and prepared filter media per bag. The PM
filter samples labeled D were sent to the Wisconsin State Laboratory of Hygiene at the
University of Wisconsin-Madison for chromium metal speciation. Total and hexavalent
chromium was measured in extracts of filter-collected PM sent from NVFEL. Detection limits
were in the <0.2 ng/filter range. A comparison of 47mm filter collection substrates was
performed using Polyvinyl Chloride (PVC) and bicarbonate-impregnated Mixed Cellulose Ester
(MCE) filters. Total chromium was analyzed by SF-ICPMS (Sector Field Inductively Coupled
Plasma Mass Spectrometry) and Cr(VI)was analyzed by Inductively Coupled (IC)-post-column
derivation. The Cr(VI) results obtained using PVC collection substrates were below the detection
74
-------
limit, with the exception of the tunnel blanks, and thus not listed in this memo. The extractable
total chromium levels in the filters and bicarbonate were at such a level that swamp any signal
from the PM, making the ICPMS data useless. However, the Cr(VI) data from the MCE filters
analyzed by IC could be used to develop new emission rates as described below.
Spike and blank studies were performed. Spike studies had a recovery between 93-104%,
indicating the matrix did not interfere with the chromium results. The Cr(VI) MCE filter results
were blank corrected by subtracting the mean background value of 0.298 ng/filter (standard
deviation±0.098 ng/filter; 95% confidence interval±0.157). The 95% confidence interval was
calculated from student's ^-distribution as a function of the probability and degrees of freedom
and multiplied by the standard deviation over the square root of the number of blanks.
Cr(VI) speciation results and emission rates are reported in Table A-l along with the
corresponding distance driven per sample. The emission rates were calculated by dividing the
blank-corrected Cr(VI) MCE mass/filter by the distance driven per sample and multiplying by a
factor representing the CVS (constant volume sampler) volume over the individual filter sample
volume (NVFEL filter sample D was used for each bag). This factor was used because all
exhaust was not passed through the collection filter during the test.
blankcorrected Cr(VI)MCE mass/filter CVS volume
Emission Rate = x
distance Sample volume
The overall emission rate in Table A-l is a composite average of the total Cr(VI) measured
divided by the total distance of the test and then multiplied by the sum of CVS volumes/sum of
filter sample volumes.
Table A-l. Cr(VI) Emission Rates From an On-road Gasoline Engine
Sample/bag
number
Cr(VI)
(ng/filter)
Mean IC
Blank± Std
Deviation
(ng/filter)
Blank-
corrected
Cr(VI)
(ng/filter)
CVS
Volume (scf
at 68°F)
Sample
Volume (scf
at 68°F)
Distance
(miles)
Emission
Rate
(g/mile)
1
0.792
0.298±0.098
0.49
1666.87
7.675
1.194
8.9x10 s
2
0.493
0.298±0.098
0.20
6280.73
28.815
8.612
5.1xl0"9
3
0.488
0.298±0.098
0.19
1682.74
7.711
1.186
3.5xl0"8
4
0.508
0.298±0.098
0.21
6281.82
28.894
8.620
5.3xl0"9
Overall
1.1
15912.2
73.10
19.61
1.2x108
75
-------
Direct Cr(VI) emission factors were not measured from a diesel engine. To develop on-road
diesel emission factors, the overall gasoline emission factor from Table A-l is multiplied by the
ratio of total chromium from diesel engines verses gasoline engines. Emission factors are
calculated for diesel engines based on the most recent estimates from engines before7 and after8
implementation of EPA's 2007 heavy-duty highway rule which reduced PM emissions from
heavy-duty diesel vehicles. The total chromium emission factor for gasoline comes from the
Kansas City Particulate Matter Characterization Study (4.07xl0"6 g/mi).9
Cr(VI) Pre-2007 On-road Diesel Emission Factor
Total Cr FR, , a 6.8x10 6 a
EF = Gasoline Cr(VI)EF x ^ = 1.2xl0"8 x = 2. OxlO"8
Total Cr EFgasoiine mi 4.07xl0—tifti
mi
Cr(VI) 2007 and Later On-road Diesel Emission Factor
—6 _9_
mi
Total Ct EFgasonne mi 4.07xl0-6-^
, Total Cr EFdiesel ag 1.94x10 6„
EF = Gasoline Cr(VI)EF x ^ = 1.2xl0"8 — x ^ = 5.8xl0"9
mi
A Cr(VI) emission factor for transit buses using compressed natural gas is calculated by
multiplying the overall Cr(VI) emission factor from Table A-l by the ratio of total chromium
from CNG transit buses10 verses gasoline light-duty vehicle engines (from the Kansas City
study).
Cr(VI) Transit Bus Compressed Natural Gas (CNG) Emission Factor
-8 _9_
_ mi _
Total Cv EFgasonne mi 4.07xl0—6
Total CrEFCNG Q g 7.0x10 g
EF = Gasoline Cr(VI)EF x = 1.2xl0"8 x ^ = 2. lxlO"10
mi
Non-road emission factors for gasoline engines are presented in grams per gallon and calculated
from the 2008 Chevrolet Impala based on a city fuel economy of 18 miles per gallon.11
Cr(VI) Non-road Gasoline Emission Factor
o .Q mi , fl
EF = 1.2x10 — X 18 = 2.2X10"7 —
mi gal gal
Non-road pre-2007 and 2007 and later diesel emission factors are calculated from the overall
gasoline emission factor from Table A-l. The above gram per gallon gasoline emission factor is
multiplied by the ratio of total chromium emission factors from diesel engines (before and after
2007) verses gasoline engines (from the Kansas City study) to obtain the nonroad diesel engine
gram per gallon emission factor.
Cr(VI) Pre-2007 Non-road Diesel Emission Factor
a 6.8xl(T6^ a
EF = 2.2xl0~7 X = 3. 7xl0~7
9al 4.07xl0-6 9al
mi
76
-------
Cr(VI) 2007 and Later Non-road Diesel Emission Factor
„ 1.94xl0~6 -^r a
EF = 2.2xl0~7 — X ^ = 1. OxlO"7 —
3al 4.07xl0~6 -^r 9al
mi
A summary of the results for Cr(VI) emission factors is presented in Table A-2. While these
results are based on measured Cr(VI), the results are limited by the following:
• Emissions from only one vehicle were measured, so the data do not provide information
regarding variability among vehicles
• No measurements have been made for diesel and CNG vehicles or engines
Table A-2. Summary: Cr(VI) Emission Factors
Emission
Factor Units
On-road gasoline (MY2008)
1.2xl0"8 grams/mile
On-road diesel (pre-2007)
2.0xl0"8 grams/mile
On-road diesel (2007 and later)
5.8xl0"9 grams/mile
CNG Transit Buses
2.1xlO"10 grams/mile
Non-road gasoline (MY2008)
2.2xl0"7 grams/gallon
Non-road diesel (pre-2007)
3.7xl0"7 grams/gallon
Non-road diesel (2007 and later)
l.OxlO"7 grams/gallon
a calculated by dividing bags 2 and 4 blank-corrected Cr(VI) MCE mass/filter by the distance
driven for bags 2 and 4, then multiplied by the bag 2+4 sum of CVS volumes/sum of sample volumes
1 http://www.epa.gov/otaq/models/moves/documents/420bl2029a.pdf
2 Ball, James C. Emission Rates and Elemental Composition of Particles Collected From 1995 Ford
Vehicles Using the Urban Dynamometer Driving Schedule, the Highway Fuel Economy Test, and the
US06 Driving Cycle. 97FL-376. Society of Automotive Engineers, Inc. 1997.
Table 1. MCE filter, Test# 20100024028
3 Taylor, M. Memorandum: Revised HAP Emission Factors for Stationary Combustion Turbines,
Prepared by Alpha-Gamma Technologies, Inc for Sims Roy, EPA OAQPS ESD Combustion Group.
August, 2003.Docket ID: OAR-2002-0060-0649. Access via http://www.regulations.gov
4 Kansas City Particulate Matter Characterization Study. Final Report, EPA420-R-08-009. Assessment
and Standards Division Office of Transportation and Air Quality U.S. Environmental Protection Agency
Ann Arbor, MI,
77
-------
5 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. DOI: 10.1080/10962247.2013.870096.
6 http: //www. epa.gov/otaq/standards/light-dutv/la92 .htm
7 Hsu, Y., and Mullen, M. 2007. Compilation of Diesel Emissions Speciation Data. Prepared by E. H.
Pechan and Associates for the Coordinating Research Council. CRC Contract No. E-75, October, 2007.
Available at www.crcao.org.
8 Khalek, I., Bougher, T., and Merritt, P. M. 2009. Phase 1 of the Advanced Collaborative Emissions
Study. Prepared by Southwest Research Institute for the Coordinating Research Council and the Health
Effects Institute, June 2009. Available at www.crcao.org.
9 Kansas City Particulate Matter Characterization Study. Final Report, EPA420-R-08-009. Assessment
and Standards Division Office of Transportation and Air Quality U.S. Environmental Protection Agency
Ann Arbor, MI.
10 Okamoto et al. 2006. Unregulated Emissions from Compressed Natural Gas (CNG) Transit Buses
Configured with and without Oxidation Catalyst. Environ. Sci. Technol. Vol. 40, 332-341 (value
obtained from page 338, Table 6)
11 http://www.fueleconomv.gov/feg/Find.do?action=sbs&id=24696
78
-------
Appendix B Development of Motor Vehicle Emission Factors for
Mercury
B.l Calculation of Mercury Emission Factors from Vehicle Tests
In 2005, the USEPA National Exposure Research Laboratory (NERL) collected mercury (Hg)
samples in the raw exhaust from 14 light-duty gasoline vehicles and two heavy-duty diesel
vehicles. The work plan for this project includes details of the methods used that are not
reproduced here including quality assurance and quality control for Hg collection and analysis.
This information can be obtained from EPA upon request. Briefly, mercury and regulated
pollutant data were collected during two sets of three consecutive LA92 drive cycles for each
vehicle. The morning set of LA92 cycles began with one 'cold start' and the afternoon set of
three LA92 cycles began with a 'hot start'. The intake air was filtered through charcoal to greatly
reduce background mercury concentrations entering the vehicle intake. Separate sample lines
were used for gaseous and particulate mercury species. Samples analyzed for mercury were
drawn from raw exhaust at a constant flow rate and fixed dilution. Carbon dioxide measurements
were also taken in the exhaust stream where mercury samples were collected.
Mercury samples were collected in the raw exhaust since previous data suggested that mercury
levels might be sufficiently low to challenge mercury detection limits. This sampling method
imposed a challenge in calculating emission factors since it assumes that the exhaust flow rate
from the vehicle is constant. Calculation of exhaust flow and its application to the development
of mercury emission rates is described below.
Evaporative losses of mercury from motor vehicles and loss of mercury during refueling were
not measured. The emission of mercury through evaporative processes is expected to be
negligible compared with that expected from exhaust emissions.
A description of the vehicles tested for which data were used in developing emission rates is
provided in Table B-l. The data collected from these vehicles in diluted exhaust in the constant
volume sampler (CVS) included THC, carbon dioxide (CO2), nitrogen oxides (NOx), methane
(CH4), and carbon monoxide (CO). In raw, undiluted exhaust, data collected included elemental
and total gas-phase mercury, particulate mercury and CO2. Gas-phase mercury was also
measured in the intake air. Total air flow was measured for all sampling systems and corrected to
standard temperature and pressure conditions. The data streams had different reporting
frequencies, all due to the nature of the instrumentation. The dilute measurement of the standard
emission gases (THC, CO2, NOx, CH4, and CO), CVS flows, and vehicle speed were reported at
1 Hertz. The gas-phase mercury samples were analyzed at 2.5 minute intervals and particle-
phase mercury samples were collected cumulatively for the duration of three consecutive LA92
cycles. Gas-phase elemental mercury in the engine intake air was measured at five-minute
intervals.
79
-------
Table B-l. Vehicles tested for Mercury Emissions
Model Year
Make
Model
Fuel Type
Odometer
(mi)
Cylinders
Displacement
(L)
2005
MERCURY
GRAND MARQUIS LS
Gasoline
9,953
8
4.6
2005
FORD
MUSTANG
CONVERTIBLE
Gasoline
5,424
6
4.0
2003
SATURN
L 200
Gasoline
29,667
4
2.2
2002
HONDA
ACCORD EX
Gasoline
51,824
4
2.3
2001
HONDA
ACCORD EX
Gasoline
88,611
4
2.3
2001
CHRYSLER
PT CRUISER
Gasoline
54,010
4
2.4
2000
CHEVROLET
SUBURBAN
Gasoline
39,787
8
6.0
2000
JEEP
CHEROKEE SPORT
Gasoline
48,468
6
4.0
1999
FORD
F250 XLT
Diesel
113,897
8
7.3
1999
FORD
F250 XLT SD
Diesel
109,429
8
7.3
1998
HONDA
CIVIC DX
Gasoline
204,983
4
1.6
1994
CHEVROLET
SILVERADO
Gasoline
129,521
8
5.7
1992
CHEVROLET
S10 BLAZER
Gasoline
162,249
6
4.3
1991
HONDA
ACCORD EX
Gasoline
143,289
4
2.2
1987
CHRYSLER
FIFTH AVENUE
Gasoline
72,573
8
5.2
1984
FORD
F150 PICKUP
Gasoline
36,727
8
5.8
Exhaust flow was integrated at the same reporting frequency as the mercury exhaust values for a
particular test and then used to calculate total, elemental, and reactive gas-phase mercury mass
emissions. The intake air mercury values were typically collected at half the frequency of the
mercury exhaust values and used to correct exhaust measured values that are reported at higher
frequencies. The particulate matter measurements were filter-based, test-level measurements and
were corrected in that manner.
B. 2 Calculation of Emission Rates
Emission rates were calculated separately for elemental gas-phase mercury, reactive gas-phase
mercury and particulate mercury. Elemental gas-phase mercury in the exhaust was corrected for
the intake air concentration of elemental mercury. To estimate the gas-phase mercury
concentration in dilute exhaust from the measured mercury in raw exhaust, the dilution factor
was applied. For light-duty gasoline vehicles, the dilution factor equation found in 40 CFR
90.426 (d) was used:
Dilution factor = 13.4 / ([C02%] + ([THC, ppm] + [CO, ppm])* 0.0001)
Exhaust flow = (CVS flow / dilution factor)
Exhaust flow calculation was initiated when the analytical equipment indicated that the dilute
exhaust CO2 concentration was greater than the background CO2 concentration.
To calculate exhaust flow for the diesel vehicles, the dilution factor was calculated by simply
dividing CO2 in the raw exhaust by CO2 in the CVS. This method was used because diesel
engines operate across a very wide range of fuel to air mixtures and the CFR method described
above was not appropriate.
80
-------
B. 3 Determination of Reactive Gas Mercury Mass in Exhaust
Reactive gas-phase mercury (RGM) was calculated by subtracting elemental gas-phase mercury
measurements from total gas-phase mercury measurements. RGM values were typically small
and therefore influenced by the variability in the elemental mercury measurements. Negative
RGM values for a given measurement period were observed. Values for which there was not a
positive RGM measurement were treated as non-detects and were nulled in the aggregation of
RGM values for the test. The measurement uncertainty for gas-phase elemental mercury was
estimated from quantitative recovery of injections of known amounts of mercury into the
sampling system. The uncertainty in measuring elemental mercury was applied to the total gas-
phase and elemental gas-phase measurements to determine when the RGM value was above the
measurement uncertainty. Values within the measurement uncertainty were not included in the
emission factor calculation.
B.4 Calculating Weighted Emission Test Results
Highway vehicles were tested on the LA92 cycle, a more aggressive chassis-dynamometer test
similar in concept to the Federal Test Procedure's (FTP) UDDS or LA4. Like the FTP, the LA92
includes a cold start, a hot start, and a hot stabilized phase using identical drive schedules for the
starts. We considered it appropriate to calculate a weighted emission factor (representing cold
start and hot start driving) for each vehicle in the same manner as the FTP, using the equation
below for each test (a test consisting of all six LA92 cycles performed on each vehicle).
We summed the gas-phase mercury mass emissions for the first phase (300 seconds) of the
morning test and last phase (1,135 seconds) of the individual LA92 drive schedules for all the
tests (e.g., 'hot stabilized emissions'), divided by the total distance covered in these phases and
multiplied by 0.43. We also summed the sum of the mass gas-phase mercury emissions of the
first phase of the afternoon test and last phase (1,135 seconds) of all the tests, divided by the total
distance covered in these phases and multiplied by 0.57. The two terms were summed to
calculate a test level emission rate for each of the gasoline powered vehicles.
The equation used to calculate test-level emission rates is as follows:
£Hg=0.43
( C + R ^
\Cm+Rm J
-0.57
r H+R ^
1-f 4- R
\ m m J
Where:
Eug = mean aggregate emission rate (g/mi),
C = mercury mass collected in the first 300 seconds of the first morning test ('cold start', g ) ,
Cm = distance covered in the cold start phase (mi),
R = mercury mass collected in the last 1,135 seconds of all six cycles of the LA92 ('hot
stabilized', g)
Rm= cumulative distance covered in all six cycles of the LA92 ('hot stabilized ', mi)
H = mercury mass collected in the first 300 seconds of the first afternoon test ('hot start', g )
Hm = distance covered by the hot start (mi)
It should be noted that the 'hot start' in the afternoon typically occurred after the vehicle had
been off for at least 1 hour, making this start closer to a 'cold start' than 'hot start'. Since the true
cold start emissions were slightly higher than hot start emissions, it is expected that this approach
81
-------
would bias the emission factors high by a small amount, relative to the value expected for a cycle
composite.
Particulate mercury emissions could not be apportioned into modes of operation in similar
manner because filters were collected across all three LA92 cycles and could not be parsed into
the three phases. A test-level composite emission rate was calculated by multiplying the morning
particulate mercury emission rate by 0.43 and the afternoon particulate mercury emission rate by
0.57 and adding the two values together.
The average of emission factors across vehicles was calculated for each form of mercury and is
reported in Table B-2. A simple average was used since the data did not suggest that mercury
concentrations varied by vehicle age, mileage, displacement or other factors.
Mercury emission factors for on-road diesel engines were obtained from the first 715 seconds of
the morning and afternoon tests on the Ford F250 XLT SD; data from the second diesel vehicle
could not be used. The first 715 seconds is approximately half of the first of the three LA92
drive cycles that made up a single test. The truncation of the test was due to sample flow
problems in the mercury sampling manifold due to particulate matter restricting flow across the
particulate matter filters. Graphical analysis of exhaust flow indicated that they appeared
nominal during the first LA92 cycle. We decided that only using measurements collected before
715 seconds in both tests provided the most reliable data.
Nonroad grams per gallon emission factors in Table B-2 were calculated from the on-road
factors using a fuel economy estimate of 17 miles per gallon for the gasoline vehicle and 19 for
the diesel vehicle.
82
-------
Appendix C Responses to Peer-Review Comments
C. 1 Adequacy of Selected Data Sources
Does the presentation give a description of selected data sources sufficient to allow the reader to
form a general view of the quantity, quality and representativeness of data used in the
development of emission rates? Are you able to recommend alternate data sources might better
allow the model to estimate national or regional default values?
C.l.l Dr. TomDurbin
No comments specific to the Toxics Report.
C.1.2 Dr. Allen Robinson
I thought that the report did not do a good job of providing in text citations to the data sources.
Often when the report referred to a data source there was not an in text citation. For example, on
page 14 — they were taken directly from the Complex Model Spreadsheet "CMFinal.xls". Need
a reference for this spreadsheet. This is just one example.
RESPONSE: We removed the reference to the Complex Model Spreadsheet in the text,
which referenced Equation 4, and the value used are presented within the Report in
Table 8,Table 9,Table 10, and Table 11. We also added text citations for data sources
(The number of cited references increased from 49 to 59 in the main report) .
Pre2000 vehicles (Section 2.1) This model is based on old Tier 0 data, which is applied to a
large fraction of Tierl vehicles. There is a lot of speciated data for Tier 1 vehicles from the
KCVES. Why was a model not developed based on that data? The proposed model should be
tested against the KCVES Tier 1 data to demonstrate that it is applicable to those vehicles. At a
minimum this needs to be discussed.
RESPONSE: We addedfootnote "a" on page 8, which states: "While more recent
emissions data are available for Tier 1 and earlier vehicles, such as data from the
Kansas test program mentioned earlier, testing was not done on a matrix of fuels which
enable development of a fuel effects model".
C. 2 Clarity of Analytical Methods and Procedures
Is the description of analytic methods and procedures clear and detailed enough to allow the
reader to develop an adequate understanding of the steps taken and assumptions made by EPA
to develop the model inputs? Are examples selectedfor tables andfigures well chosen and
designed to assist the reader in understanding approaches and methods?
C.2.1 Dr. TomDurbin
p. 6 - 2nd paragraph discusses pre-2001 vehicles and 2004+ vehicles, but does not address 2001-
2004 vehicles. 4th paragraph - what two fuel properties are used for evaporative emissions.
84
-------
RESPONSE: Vehicles in MY 2001-2004 are represented by inputs for Tier 2 vehicles.
We have modified the text to clarify this point in Section 1.2. We moved the discussion
regarding fuel properties accountedfor in modeling evaporative emission processes to
Section 6.
p. 15 - its not clear what is meant by the phrase that "relations of air toxic emissions to changes
in fuel properties has remained stable from Tier 0 to Tier 1"
RESPONSE: We added a sentence and edited the sentence on page 14 to state: "The
equations and parameters presented are used to estimate the fuel effect for both Tier 0
and Tier 1 gasoline vehicles. This approach is based on the assumption that the
proportional responses of air-toxic emissions to changes in fuel properties are similar for
vehicles certified to both sets of standards. "
p. 17 - There is a reference to modeling 2000 and earlier vehicles on E15-E20, but not
discussion on factors that would be used for such fuels. It would be useful to at least reference
the section where this will be discussed.
RESPONSE: We added columns to Table 14 to include or reference the toxic fractions
used for El5 and E20. We also added text on page 16 discussing the source of the data.
For section 2.2.1 see suggestions for the report "Gasoline Fuel Effects for Vehicles Certified to
Tier-2 Standards". Then on page 32, it talks about the "full" vs. "reduced" design. The fact that
the reduced design represents 5 vehicles and 11 fuels (as opposed to 5 vehicles by 27 fuels)
should be discussed in the 1st paragraph, rather than the 2nd. Then the 2nd paragraph talks about
Table 30 and 31 before these tables are introduced in the 3rd paragraph, so the 2nd paragraph
seems out of place. It should at least be mentioned here that acrolein, benzene, and 1,3 butadiene
are not modeled for hot running emissions in this section (even though it is discussed in the next
section). The approach using "information parity" appears to the reasonable for NMOG and
ethane.
RESPONSE: We added background on the EPAct program and our analysis of the
results in Section 2.1.1.2, including descriptions of the data used to fit models for each
combination of pollutant and test phase (bag) (Table 11). In addition, we amplified the
explanation of the full and reduced designs (Section 2.1.1.2.1).
Section 2.1.3 - It should be mentioned at the start of the paragraph that metals are represented
both with these metals and the metals presented in the PM2.5 emission profile. Also,
"conservative" is probably too weak a term to describe using the bag 2 emission rates, since its
actually more of an upper limit estimate (although this only appears to be the case for
manganese).
RESPONSE: We have added the text in the beginning of Section 2.3 that mentions the two
ways MOVES models metals (using speciation profiles, and gram/mile emissions).
We also removed the term "conservative " and mentioned that using bag-2 is a likely
upper limit estimate.
p. 42 - A recent study by CARB/UC Riverside/UC Davis should provide some information
related to biodiesel emission factors.
RESPONSE: We will consider these studies in future updates of the model.
85
-------
p. 42/43 seems like final paragraph on 42 and 1st paragraph on 43 could be combined, since the
three different references to Table 39 in these paragraphs is a little confusing.
RESPONSE: The change was made so that all the references to Table 57 (old Table 39)
are contained in one paragraph.
p. 46 - section 2.3.4 - It seems like dioxin emissions might be overestimated using a data set
with such older vehicles. This might be worth mentioning in the text.
RESPONSE: We have added text to mention this point in Section 3.4.
p. 47 - section 2.4.2 - It's not clear what the basis of the particulate to gaseous phase split is for
the PAHs. If it is discussed previously, it should be reiterated here.
RESPONSE: we have added text in Section 4.2 to reference the source of the
apportionment (Table 58).
p. 53 - 3rd paragraph on 20% ethanol. It is unclear what fuel speciation data was used here. Was
this from in-use fuels? Since the test fuels were not necessary representative of average fuels, but
rather represent the extremes of in-use fuels. Table 51 (now Table 69) is useful.
RESPONSE: The text has been modified to clarify that we used data for blends
containing 20 vol. % ethanol from the EPAct program. These data are not representative
of in-use fuels, but are the best available information.
C.2.2 Dr. Allen Robinson
The report commonly uses the word "fraction" or "toxic fraction". You need to define fraction
of what - VOC, NMOG, THC, etc (presumably each of these is defined using standard EPA
definitions). For tables actually defining in header as was done for Table 20 (now Table 48) is
useful. Also is this a mass or a mole fraction?
RESPONSE: Section 1.1 defines the term toxic fraction (as a function of VOC) and OC
for particulate compounds, and that all fractions are mass-based. We have reviewed the
report to make sure we are clear about the definition of the fraction we are discussing in
each section.
Please make sure that all variables are defined - a nomenclature table with units should be added
to the report.
RESPONSE: We have carefully reviewed the document to ensure that all variables were
properly defined in the text, prior to first use. We have ensured that the units for the
emission rates for metals, and dioxin/furans are defined in each table the results are
presented (e.g. Table 61), and that the definition of the toxic ratios are clear (see
previous comment).
Centering data (page 10) (now page 9) - It appears that you are using a different centering
approach for older data than for the new model (e.g. eqn 8) (now Equation 9). Why were
different approaches used?
86
-------
RESPONSE: The Complex Model and the EPAct models were developed in separate
research efforts by different authors at different times. Not surprisingly, the approaches
used are similar to some degree but not identical. The EPAct study reflects
improvements in computer-optimized study design and analytic methods developed and
introduced between 1993 (Complex Model) and 2008 (EPAct). The approach used in the
Complex Model ("centering") effectively shifts the means of all fuel properties to the
origin, but leaves each property in its native units, i.e., each property is scaled
differently. The approach used in EPAct ("standardization ") centers the fuel properties,
and goes one step further to express all properties in the same scale, i.e., each property is
expressed in units of its own standard deviation.
What is meant by model year specific weightings (page 10)? [Table 12], What do these weights
represent? Fraction of vehicles for a given year?
RESPONSE: We added text on page 13 clarifying that the weights represent the sales mix
of technologies within a given model year.
Equation 1 - what are the units of the different variables?
RESPONSE: The primary purpose for including the equation was to illustrate the
"centering" approach used in the analysis. Nonetheless, we have specified units in the
text for the terms shown in the equation. In addition, units for all terms are specified in
Table 7.
Table 8 - Complex Model coefficients - these are beta's in equation (1).
RESPONSE: Yes. The ft are regression coefficients for the centeredfuel-property terms.
We have added text after Equation 1 to make this point explicit.
Page 13 "For each compound, the model equations as shown in Equation 1, are evaluated for a
"base" and a "target" fuel." This base fuel resides in MOVES? Is this the same as the average
fuel listed in Table 7?
RESPONSE: Table 7 does not describe an "average fuel. " Rather, this table lists the
set ofproperties included in the Complex Model, and lists the mean value of each
property for the fuel set used in the analysis. The mean values are used in Equation
1.
The "base fuel" is stored in the MOVES database. Several base fuels are used by MOVES, with
each applied to a different set of model years. Base fuels are applied to represent the fuel
implicitly reflected in the base emission rates. Thus, because "in-use" fuels applying to model
runs for specific locations differ from the base fuels, MOVES calculates and applies fuel
adjustments, relative to the base fuel, to represent corresponding fuel effects.We reference the
MOVES2014 Fuel Effects report in this section which provides more detail on the use of base
fuels in MOVES. Page 14 - equation 3. It was not clear how the weights are being applied. You
are trying to derive one adjustment factor for all pre2000 vehicles? Are you driving a separate
factor for the 10 different technology classes? This needs to be clarified.
RESPONSE: We have added text to clarify the meaning of the weights: "The weights
represent the sales fractions for the ten vehicle technologies defined in Table 6. Note
87
-------
that the use of varying weights in applying the Complex Model in MOVES differs
from the original application in which the weights were invariant. The application of
Equation 3 to each of the 30 ages listed in the table gives a set of 30 adjustments,
with each applied to a single model year, which represents a specific age with
respect to the calendar year simulated. "
Table 12 (page 14) — According to the text these weights represent prevalence for a given
technology year. Prevalence means what? Fraction of vehicles based on number, VMT? I am
confused that Table 12 lists weights based on "age" as opposed to model years? Is this age
relative to 2000? It would be clear to define a base year to calculate age.
RESPONSE: Prevalence indicates the fraction of new vehicle sales in a given model
year. We added text clarifying that vehicle age 0 represents the simulation year for
which an inventory is calculated. The other ages represent older model years
relative to the simulation year.
Equation 6 — What is Ivoc? Where does the value come from? The standard moves code.
RESPONSE: These two terms are defined in the paragraph immediately preceding
their use in Equation 6.
Post2000 organic emissions are based on models derived from the EPAct data. It was not clear if
these models are the same as those in the EPAct report. I assumed that they were. If so, the
Toxic report needs to specifically acknowledge that. In addition, it should provide specific
references to which models are being used as the EPAct report describes a whole bunch of
models. Please provide in text citations for the EPAct report.
RESPONSE: We have substantially revised this section (Section 2.1.1.2.1) of the
report, adding material and tables to better describe the origins of the EPAct models
and provide appropriate references to the project report.
Table 1 - Are all these hydrocarbons? There are compounds that contain elements other H and
C, which I don't consider to be hydrocarbons.
RESPONSE: We have altered the text to note that the list includes volatile organic
compounds, which is inclusive of the organic gases in the toxics report.
When you use the term "start" please define it as either cold (e.g. bag 1 of LA92 with appropriate
preconditioning) or hot start (bag 3 of LA92).
RESPONSE: In this context, "start" is synonymous with "cold start, " or Bag 1 of the
LA92. We have added text to make this usage explicit on page 20.
Page 6 "algorithms" -are these really curve fits as opposed to algorithms?
RESPONSE: We have removed the terms algorithms from the report and substituted with
terms such as "statistical models fit to these data "
Page 8 "Toxics inputs for MOVES are not explicitly designed to vary by temperature." Not sure
what this means? The outputs do not vary with temperature? What does temperature refer to?
Ambient? Cold versus hot start?
88
-------
RESPONSE: Text was added in Section 1.1 to clarify that the coefficients and other
inputs used to estimate emissions of toxics do not vary by temperature, but that resulting
emissions estimates may, because they are modeled as a function ofVOC and OC2.5,for
which estimated emissions are adjustedfor temperature.
"In addition, while MOBILE6.2 relied on very limited data from heavy-duty gasoline vehicles,
MOVES applies Complex Model algorithms to both light-duty and heavy-duty gasoline
vehicles" Is there a basis for this extensions. Have additional heavy duty gasoline vehicle data
been obtained? If not why is MOVES being extended to heavy duty gas while MOBILE did
not?
RESPONSE: The approach taken in MOVES differs from that in MOBILE because the
data from heavy-duty gasoline vehicles used in MOBILE was so limited that it did not
allow for estimation of differences in emissions attributable to changes in fuel properties.
We included this rationale within the text in Section 2.1.1.1.1.
Page 16 (last sentence of first paragraph) Does MOVES have representative fuel data for
different regions and simulations year? Given the focus of fuel dependence of emissions
providing the user with a robust set of default fuel values (year and region) would be helpful.
RESPONSE: The MOVES database does provide a set offuels designed to represent
typical commercially available fuels throughout the nation. This set of inputs is
designated as the "fuel supply" and is described in a separate MOVES2014 report21.
Equation 7 - what is V and what are its units? Equation 7 and associated parameters in Table 13
were derived by fitting MOBILE output. Why not fit directly the original data or use the original
parameterization in MOBILE? You claim this equation provides the best fit. What are statistics
of fit?
RESPONSE: The variable V is defined in the paragraph immediately preceding its use in
Equation 7. The equation used is consistent with the best fit parameterization originally
developedfor MOBILE and has the same two terms. We have simplified coding in
MOVES by developing a quadratic regression that gives results consistent with the
original model used inMOBILE6.2.
Table 12 — What do these weight represent? The distribution of different classes of vehicles in
different model years? It seems like the minimum age of 2000 vehicle is 13 years (if running a
present day simulation).
RESPONSE: The weights represent the sales fractions of each technology group in
vehicle sales for a given model year. The table lists sets of weights for ages 0-30 within a
broad model-year group, "1960-2000. " Thus, each age represents a single model year
within the broader group. The models described in this section are applied to vehicles in
model year 2000 and earlier, so in a present-day simulation, it is correct that vehicles in
MY2000 would be 12-14 years old.
[Equation 7] "It should be noted that the sulfur effects terms in the equations were not included;
rather, sulfur effects on toxic emissions were assumed to be proportional to the sulfur impacts on
total VOC estimated by MOVES." Sulfur effects in what equations? There is no sulfur in
equation 7 (which is the equation that this sentence seems to refer to).
89
-------
RESPONSE: The reviewer is correct that this sentence is not relevant to Equation 7. It
has accordingly been moved to Section 2.1.1.1.2 and revised to clarify that the sulfur
terms in the original Complex Model were not included when the model was adapted for
application in MOVES. Rather, sulfur effects on toxic emissions are assumed to be
proportional to the effect of sulfur on total VOC, as estimated by MOVES.
Table 16 exists in Pre-2000 section (Section 2.1.1.1.4) but appears to apply more generally.
Move into a more general section of the report?
RESPONSE: This table does also apply to vehicles manufactured after 2000, but we
thought it sufficient to describe the table in its current location in the report and then
reference it as appropriate in later sections, (e.g. Section 2.1.1.2.8).
Do you really want to call ethanol blends gasohol? When I hear gasohol I think of Brazil.
RESPONSE: We agree that this term seems out of date. However, as it is currently used
in the MOVES database, in the table "FuelSubType", we have retained it for the present
in Table 16. However, it is a good candidate for replacement with a more current term,
such as "ethanol blend", which is used consistently throughout the text of the report.
Page 31: "one set representing start emissions and a second set representing hot-running" start
emissions is hot start (LA92 bag 3) or cold start (LA92 bag 1, with appropriate conditioning)?
RESPONSE: In revising Section 2.1.1.2.1, we have added text to clarify that "start"
refers to "cold-start, " as represented by LA92 Bag 1, and that "running" represents
LA92 Bag 2.
There are table reference problems (e.g. see page 32, 35, 38, 40, ...). There are other instances
of this.
RESPONSE: We have modified and updated table references as needed.
Table 27, 28, etc. Are these parameters from the EPAct report. If so provide citation. Please
cite the specific model from the EPAct report, not just the general report.
Page 40: What is OC2.5 VOC?
RESPONSE: This combination of terms was simply a typographical error that we have
corrected.
Page 41—dioxins and furans - "to be similar" You are assuming them to be the same not just
similar. Seems like these estimates are very uncertain since they are based on very old vehicles.
RESPONSE: We have combined the previous two sections discussing gasoline dioxin and
furan emission rates into a single section (Section 2.4) of the revised report, to help the
reader understand that we are using fleet-average emission rates for dioxins andfurans.
We modified the language in Section 2.4 to be more precise regarding the use of the data
for newer vehicles and have noted the uncertainty involved in this extrapolation.
Diesel PAH data [Section 3.2]- Similar problems with the partitioning estimates. Partitioning in
Schauer study is biased compared to atmosphere. This needs to be explicitly noted in the report.
90
-------
There is a "higher concentration of particles in diesel exhaust" compared to gasoline exhaust in
the CVS or plume, but not in the atmosphere. Concentrations in the atmosphere not exhaust is
what matters for partitioning.
RESPONSE: Please see our response to the similar question in C.3.2, pages 92-93.
Table 49 - Particle phase naphthalene? That must be a measurement artifact.
RESPONSE: As mentioned in the response C.3.2 pages 92-93, the gas-particle
partitioning is meant to be representative of the sampling conditions from which the
emissions are measured, not atmospheric conditions. We left the gas-phase partitioning
in Table 47 as reported by Fujita et al. (2013), and assume that 99.96% of the
naphthalene is from the gas phase, and 0.04% is in the particle-phase. Whether we used
100% or 99.96% as the gas-phase fraction will have a trivial impact on the total
naphthalene estimated by MOVES.
C. 3 Appropriateness of Technical Approach
Are the methods and procedures employed technically appropriate and reasonable, with respect
to the relevant disciplines, including physics, chemistry, engineering, mathematics and statistics?
Are you able to suggest or recommend alternate approaches that might better achieve the goal of
developing accurate and representative model inputs? In making recommendations please
distinguish between cases involving reasonable disagreement in adoption of methods as opposed
to cases where you conclude that current methods involve specific technical errors.
C.3.1 Dr. TomDurbin
Overall, the Complex Model provides a robust framework for modeling acetaldehyde,
formaldehyde, benzene, and 1-3 butadiene, especially with its recent updates.
Table 7 - the mean value for centering the sulfur at 204 ppmw is relatively high compared to
current sulfur levels. Will this potentially be modified going into the future?
RESPONSE: The sulfur terms in the Complex Model were not retained when the
equations were adaptedfor use in MOVES, as sulfur effects were modeled using a
different approach. For this reason, the mean value for sulfur in Table 7 is irrelevant in
MOVES and has been removedfrom the table.
Tables 8 to 11 - What do the dashes in the table represent? Is that where the data show no effect
or are insufficient? For example, there is no sulfur effect on formaldehyde.
RESPONSE: We added text on page 10 stating that the dash means the data show no
effect for a given term. Stated differently, the term for the fuel property was not
significant or did not contribute to fit.
For MTBE, the model applied previously in MOBILE6.2 should be adequate, especially since
MTBE use is essentially historical. Similarly, in section 2.2.2.1.1,[now Section 2.1.1.2.8] the use
of Tier 1 and earlier vehicles for Tier 2 vehicles appears reasonable.
91
-------
Section 2.1.2[now Section 2.2]: Its not clear what samples are being used to estimate the PAHs.
It talks about a set of 99 samples being used for the fractions in the second paragraph and how the
fractions are determined in terms of PAH/THC and PAH/OC2.5. Then it talks about the
partitioning into gaseous and particulate phases in the 3rd and 4th paragraph that appears to be based
on 2 vehicles in the medium emitter category, which was selected from 4 samples collected at two
temperatures. Why was the "medium emitter" sample selected? How significant were the
differences between the samples collected at 20°C and 47°C? If there were big differences
wouldn't that make a big difference in the partitioning for the PAH/THC and PAH/OC2.5 for the
other 99 samples? Then its unclear what Table 20 [now Table 48] represents, since it is multiplying
fractions (PAH/THC and PAH/OC2.5) by fractions (Table 19)[Table 47] in a seemingly strange
was. Where do the absolute emission rates for the individual species play in here?
RESPONSE: We added a paragraph, Equation 19 and Equation 20 to demonstrate how
we are using the KCVES emission rates, with the phase-partitioning values in Table 48
from the follow-on KCVES study, to derive the PAH ratios used in MOVES.
We also added text in Section 2.2.1 to emphasize that the gas-particle partitioning is not
indented to be representative of atmospheric conditions, but of the measurement
conditions from which VOC and PM emission factors are calculated.
We recognize that using the 'medium emitter 'for phase-partitioning may not be
representative of all the vehicles measured in the Kansas City study or for Tier 2
vehicles, but it was deemed the most representative for phase-partitioning the PAH
measurements made in KCVES. We added text in Section 2.2.1 to explain our rationale.
"Clearly, this sample may not adequately represent phase-partitioning of PAH emissions
from the current in-use fleet; however, it was deemed the most representative of the
breadth of gasoline vehicles sampled in the KCVES. "
Fujita et al. (2006) didfind that the dilution tunnel had an impact on the PAH speciation,
and PM emissions. However, the impact was not always intuitive (e.g. They observed
higher OC emission rates at the higher dilution temperature). We used the phase-
partitioning at 47°C because the dilution tunnel was operated at that temperature during
the main study (from which the PAH, THC, and PM measurements were made).
We also added Table 49 to provide information on the structure of the database that
contains the PAH/VOC and PAH/OC ratios.
Page 37 - Although benzene can be a function of fuel benzene, it can also be a function of other
low weight aromatics, especially toluene. In the EPA study on benzene, how did toluene levels
vary between the fuels?
RESPONSE: We added a footnote in Section 2.1.1.2.7 stating that the toluene levels
were constant. The only difference between the fuels was the benzene level.
Section 2.3 - Developing the air toxics factors from the E-75 database appears to be a reasonable
approach. Its unclear how these factors might account for states with low levels of aromatics, such
as California. Also, its unclear why the partitioning for the PAHs was made based on a medium-
duty diesel engine. Maybe just one sentence to clarify this.
92
-------
RESPONSE: Since data were not adequate to develop a fuel-effects model for diesel,
results will not account for impacts of low aromatics diesel fuel on toxic emissions. We
added text to Section 3.2 stating that the PAH partitioning was done using data from a
medium-duty diesel engine, because it was the best available data for the purpose at the
time of analysis.
The ACES study provides a good data set for the development of the air toxics factors for the 2007
and new engines, p. 49 section 2.4.4 - Would be interested to see how backgrounds were dealt
with in this study. At such levels backgrounds would be important in terms of not overestimating
emissions.
RESPONSE: We added information regarding background corrections in Sections 4.1,4.2,
and 4.3. The ACES program background corrected the VOC measurements, but not the
PAH or metal measurements. Details on the background correction are available at the
cited ACES Phase I report45.
Section 2.6 - CNG emissions - For the PAHs, is there any consideration given to how the oxidation
catalyst would reduce PAHs?. It appears that the estimates were based on measurements without
an oxidation catalyst, but that these are applied to both technology categories, p. 51-
RESPONSE: In Section 5.2 we only used the PAH emissions from the CNG transit bus
without an oxidation catalyst (Okamoto et al. 2006) to simplify modeling of this relatively
small source. We decided to use the non-catalyst equipped PAH emission rates as a way
to be environmentally conservative.
Section 2.6.3 - By using the only the data where chromium and nickel were detected, this would
presumably overestimate emissions. Were the metal rates from heavy-duty engines also considered
before deciding to use the gasoline emission rates[?].
RESPONSE: In Section 5.3, the heavy-duty diesel emission rates were also considered as
a surrogate for the CNG emission rates. We chose to use gasoline rates, because both fuel
types employ spark-ignition engines, and gasoline is a lighter fuel than diesel.
Section 3 - Some more details should be provided for why the hot soak and running loss algorithms
from MOBILE6.2 are applied to MOVES for the non-permeation factors. The methodologies for
the permeation factors appear reasonable.
RESPONSE: We added text in section 6.1.1 stating that these algorithms were adopted
due to a lack of relevant data from vehicles with more recent technologies measured over
a fuel set with an applicable range ofproperties.
Appendix A [now Appendix B]- the fleet of vehicles used for this study appears to be too heavily
weighted towards older vehicles. Were the results for the different vehicles [used] to provide a
profile that was more representative of the modern fleet?
RESPONSE: The data did not suggest that mercury concentrations varied by vehicle age,
mileage, displacement or other factors.
Using an average exhaust flow might tend to underestimate emissions, since often periods of
higher emissions also can be periods with higher exhaust flow.
RESPONSE: Proportional sampling was not used and the Hg sample was extractedfrom
raw exhaust. The sample flow rate was held constant, although the total exhaust flow is
93
-------
varies during the emision test. Thus, the sampler under sampled at high exhaust flow
rates and over sampled at low exhaust flow rates. We did not have a way to correct for
this, and it is a source ofHg measurement uncertainty (that we acknowledged in the
report) in the test program.
Last paragraph [in Appendix B.4] - by using only the first 715 seconds, would this over represent
cold start emissions?.
RESPONSE: Yes, however the data from the diesel tested represents the best available
information. We mention the uncertainty regarding using the only the first 715 seconds in
Appendix B.4.
C.3.2 Dr. Allen Robinson
It is not clear why the demarcation for the gasoline vehicles is MY2000 - it seems like the years
in which tier 1 or tier 2 vehicles were introduced would make alot more sense. In contrast, the
MY2007 distinction for diesel vehicles makes alot more sense than the apparently arbitrary split
for gasoline vehicles.
RESPONSE: In calendar year 2001, the national low emission vehicle program (NLEV)
went into effect, andfuel effects are better represented by Tier 2 vehicles tested in the
EPAct program.
Page 19 section 2.1.1.2 It seems very problematic to be using emissions data from EPAct for a
new Tier 2 vehicle to apply to these older vehicles to simulate emissions from high ethanol fuel
operations from a pre2000 vehicle. The uncertainty must be very large. Can you run older
vehicles on E85? There seems to be little basis for this extrapolation - it seems like you are
simply trying to be comprehensive. Ideally a quantitative estimate of uncertainty should be
provided for this estimate. At a minimum MOVES should flag the value as massively uncertain.
RESPONSE: We have added a footnote (g) in Section 2.1.2.1 and Section 2.1.2.3 pointing
out the uncertainty inherent in the emission rates, while also understanding that this is a
minor contribution to the uncertainty of the total inventory due to the small number of
pre-2001 vehicles operating on high-ethanol blended gasoline.
Phase partitioning of PAH (page 21). This applies to all vehicles (pre2000 and post2000).
However it is in the pre2000 section. I found this confusing. Why not have one section that says
PAH emissions of all gasoline vehicles estimated using this approach.
RESPONSE: We revised the outline of the report to have one section (Section 2.2,2.3 and
2.4) for gasoline PAH, dioxin, and metal emissions, respectively, which do not have
separate inputs for pre- and post-2001 vehicles. Additionally, we added text to clarify
that we used a fleet-average PAH emission rates from a sample of vehicles with model
years ranging from 1968 to 2004.
More PAH: There is a paragraph that provides the caveat that "gas-particle partitioning of PAHs
emission in the atmosphere depends on particle and gas concentrations, exhaust temperature and
other factors." It is good to state this. However, presumably the relevant temperature for
atmosphere partitioning is atmospheric temperature (not exhaust). This paragraph implies, but
does not specifically state, that the gas particle partitioning measured in source test is not
94
-------
representative of atmospheric conditions (or at least not all atmospheric conditions). I think that
this caveat needs to be explicitly stated. "The gas particle partitioning of PAHs measured in
source tests and implemented in MOVES is likely not representative of atmospheric
partitioning."
RESPONSE: We also added text in Section 2.2 to emphasize that the gas-particle
partitioning is not intended to be representative of atmospheric conditions, but of the
measurement conditions from which VOC and PM emission factors are calculated.
More PAH: The model use results for composite class, "medium emitters," to estimate gas
particle partitioning of all PAHs. Why was a medium-emitters class used? I also suspect that the
conditions inside the CVS during the test of these old vehicles (esp. PAH concentrations, PM
concentrations, BC concentrations) are not representative of atmospheric conditions (or the
newer Tier 2 vehicles). This likely biases phase partitioning towards particle phase. EPA should
choose a test in which the conditions concentration and temperature inside the CVS were within
the envelope of conditions that likely occur in the atmosphere. This likely would be a test for a
cleaner vehicles. An even better approach would be to review the literature of ambient gas-
particle partitioning measurements of these compounds and use those values (as opposed to
values from a source test). Finally, if the phase partitioning of PAHs is an important output for
some of MOVES uses then it is not difficult to implement a gas-particle partitioning model.
RESPONSE: We added/edited the text in Section 2.2.1 to address that the MOVES
emission rates are developed to be consistent with average measurement test conditions,
rather than atmospheric conditions.
We also added text in Section 2.2.1 explaining the use of the phase-partitioning from the
medium emitter. "Clearly, this sample may not adequately represent phase-partitioning
of PAH emissions from the current in-use fleet; however, it was deemed the most
representative of the breadth of gasoline vehicles sampled in the KCVES. "
Lastly, the phase partitioning of PAHs is not viewed as an important output of PAH
emissions. In the National Emission Inventory, the gas-phase and particle-phase PAH
valued are summedfor each PAH species.
MOVES reports emissions as measured in emission test programs. Substantial work and
research is needed if MOVES is changed to estimate the gas-particle partitioning of
emissions as emitted into the atmosphere. We agree that this is an important area of further
research.
Table 20 -The same PAH emissions ratios appear to be applied to all vehicles, which are based
on some sort of fleet average from the entire KCVES (or just the pre-2001 vehicles)? It is not
clear why this approach was adopted. With this approach you are locking in the emissions based
on a fleet that was 10 years old today. How constant were these ratios across the fleet? If they
are not constant, why not stratified the emissions into classes (at least Tierl, Tier2) which will
allow the model to better forecast future emissions?
95
-------
RESPONSE: The PAH emissions are based on a fleet-average of emissions that contain
measurements of vehicles with model years ranging from 1968 to 2004. They
measurements are fleet-weighted, so the newer vehicles contribute according to their
expected contribution to VMT in the KC metropolitan area. No Tier 2 vehicles were
tested as part of the KCVES. A single-fleet average PAH emission factor was derived to
be consistent with the fleet-average PM speciation profile developed from the Kansas
City study, andfor the reasons given in the Speciation report. These include: 1)
avoiding over-fitting data to model year groups, and 2) underestimation of high-emitters
within the newer model year groups. We added text referencing the Kansas City profile to
the text, which references the TOG and PM Speciation Profiles.
C.4 Appropriateness of Assumptions
In areas where EPA has concluded that applicable data is meager or unavailable, and
consequently has made assumptions to frame approaches and arrive at solutions, do you agree
that the assumptions made are appropriate and reasonable? If not, and you are so able, please
suggest alternative sets of assumptions that might lead to more reasonable or accurate model
inputs while allowing a reasonable margin of environmental protection.
C.4.1 Dr. TomDurbin
p. 9 at the top The EPA assumption that metals should be independent of temperature appears
reasonable. It might be useful to examine metal emissions as a function of operation mode,
however, for example, comparing more vs. less aggressive driving, although perhaps not for the
metals included in Table 4.
RESPONSE: We agree that such an analysis would be useful, but data are too limitedfor
this type of analysis.
Page 16 developing regressions for ETBE and TAME from algorithms for ethanol and MTBE
appears to be a reasonable assumption, especially as these fuels are not at all prevalent.
p. 37 - When modeling 1,3 butadiene as 0.0 for hot-running operation, the impact of olefins should
be considered. Later on the page - CRC E-83 can be considered for olefins, although these values
were near background levels as well.
RESPONSE: EPA will consider data from CRC E-83 for future updates of the inputs for
1,3-butadiene.
Section 2.2.2.2 - Overall, the assumptions used in this section appear to be reasonable, as E85 data
are not available for some of the toxics being measured. The section does use a range of different
descriptions of higher ethanol levels from E70 to E85 to 74% ethanol without clearly describing
when all of these different conditions are applied. For example is the same factor used for E70 and
E85? Also, on page 40, the approach that ethanol contributes no PAHs should be verified. A UC
Riverside/CEC/SCAQMD study will be completed next year that will provide some data in that
area.
RESPONSE: We added text to clarify that the PAH fractions developed in Table 50 apply
to all high-ethanol blends (E70-E100).
96
-------
In response to this comment, we examined PAH emissions data collected on E85 vehicles
tested by Hays et al. (2013). We found there was insufficient evidence to model a reduction
in the heavier PAHs with the use ofE85. We didfind sufficient evidence to model reductions
in the lighter PAHs, that exist primarily in the gaseous phase in measurement testing. For
modeling in MOVES we updated the PAH particle ratios to reflect no difference to the EO-
E2 0 PAH particle ratios in MO VES. The text is included in Section 2.2.2.
Section 3 - For section 3.1.1, when using the fuel speciation from the EPAct study to make
estimates for El 5 and E20, was the volatility of the species considered? This would not necessarily
be an essential change.
RESPONSE: We did not account for volatility of the species.
C.4.2 Dr. Allen Robinson
In this chapter/report there is wider range of data quality compared to other reports and chapters.
Some of the models are based on pretty robust data sources (e.g. basic gaseous organic air
toxics), but others are based on data that, at best, are loosely related to the source (Why should
fraction of hexavalent chromium emissions from a stationary turbine be representative of onroad
vehicles? Or why should emissions from a tier 2 E85 vehicle be representative of emissions
from much older vehicle operating on high ethanol blends). I understand the desire for the
model to be comprehensive as possible, but the uncertainty of the predictions will vary widely.
It does not seem like the model user will have any idea about the quality of the predictions.
Ideally each MOVES prediction would provide a quantitative estimate for every prediction. At a
minimum the model should provide a grade (e.g. similar to AP42) for each pollutant. For
pollutants with robust models, the grade will be high (e.g. A). For less robust models (e.g.
hexavalent chromium), the grade would be poor (e.g. F).
RESPONSE: We agree in concept that it would be desirable to provide uncertainty
estimates with MOVES predictions. In fact, MOVES was originally designed to include a
Monte Carlo simulation feature to estimate uncertainty in model runs by repeating
scenarios with random variations. However, given the scope and complexity of the
model, applying the uncertainty feature has become infeasible for most users, and not
relevant to their goals. Notwithstanding these points, development and application of
quality levels to at least some inputs could be considered.
In some cases there are important sources of data that have not been utilized (e.g. KCVES to
estimate pre2000 vehicle air toxics emissions or PAH emissions for post2000 vehicles).
RESPONSE: For all toxics except the four included in the Complex Model for Reformulated
Gasoline, the data used to develop toxic emission estimates were in fact obtainedfrom the
Kansas City Test Program (KCVES) which contains measurements from vehicles ranging from
1968-2004 model years. However, for benzene, 1,3-butadiene, formaldehyde and acetaldehyde,
we relied on the Complex Model because data were collected on a matrix of different fuels which
enabled modeling impacts of changes in fuel properties. CNG buses - It seems like there is more
data available. WVU has done a bunch of testing on transit buses. Aerodyne research also did a
bunch of chase studies of CNG powered transit buses in which they measured high formaldehyde
emissions.
97
-------
RESPONSE: As shown in Table 66, the formaldehyde emission fraction composes a
large percentage of the VOC emissions. We are aware of additional studies being
conducted on CNG-fueled vehicles equipped with three-way catalysts, including those
from WVU, but unfortunately, these were not available to us at the time we were
developing inputs for the current MOVES update. We plan to continue to update these
rates in future versions of MOVES.
Section 2.1.3 Metals -You assume constant emission rates across fleet (which seems plausible,
much more so than for PAHs). However, if there were systematic variations in metals emission
rates across the fleet why not stratify the model to capture them. What is the quality of the metal
emissions? Presumably metal emissions will be sensitive to lube oil therefore it is not clear how
widely applicable the data are.
RESPONSE: The metal emission rates were developed as fleet averages to be consistent
with the fleet-based PM speciation profile cited in Section 2.3.
Hexavalent chromium - The speciation is based on stationary combustion turbine testing. Is
there any reason to think that is applicable to on-road vehicles? If not, why even report it. At
best the results will be highly uncertain. I think this an example of where the model predictions
are not supported by robust data.
RESPONSE: We have replaced the emission factors for hexavalent chromium with test
data from a motor vehicle, as discussed Sections 2.3, 0, 4.3, 5.3, and Appendix A
Page 25 - Why are dioxins and furans expressed as TEQs as opposed to not mass. I am not
familiar with dioxins but it struck me as strange. The quality of the dioxins data seemed low.
RESPONSE: This convention is commonly used with dioxins and furans, to resolve the
multiple congeners into a single "species, " by expressing all compounds as equivalents
of the most toxic congener, e.g., the "2,3,7,8" congener for the dioxins. However, we
have changed emission rates to mass rather than TEQs in the most recent update to the
model.
C. 5 Consistency with Existing Body of Data and Literature
Are the resulting model inputs appropriate, and to the best of your knowledge and experience,
reasonably consistent with physical and chemical processes involved in emissions formation and
control? Are the resulting model inputs empirically consistent with the body of data and
literature that has come to your attention?
C.5.1 Dr. TomDurbin
Overall, the methodologies selected and applied for this report appear to be providing reasonable
input to the MOVES model. As additional data sets become available, they should also be
considered for incorporation into the model, as discussed above.
C.5.2 Dr. Allen Robinson
The report does not provide sufficient information to assess this.
98
-------
¦ page 51- 1st paragraph under section 2.6.3, end of last sentence in paragraph has extra space?
¦ page 52- 1st paragraph after 3.1 (evaporative?); later
¦ page 55- under eq. 18 linearlyinterpolated
¦ Appendix A - p. 61 2nd paragraph 1st sentence "in the raw exhaust"; p. 62 last paragraph the end of
the 1st sentence is no clear, and should have a comma after powc^ "; p. 63 last sentence "The
Eequation.."
RESPONSE: These clarifications and grammatical errors have been addressed.
C.6.2 Dr. Allen Robinson
Compared to the other reports there were more typos, broken links, placeholders like "???" in the
text, and many typos (e.g. superscripts for references and on numbers, e.g. see Table 47) in this
report.
RESPONSE: These clarifications and grammatical errors have been addressed.
100