MOVES2010 Fuel Adjustment and Air

            Toxic Emission Calculation Algorithm —

            Development and Results
&EPA
United States
Environmental Protection
Agency

-------
                MOVES2010 Fuel Adjustment and Air
                Toxic Emission Calculation Algorithm -
                          Development and Results
                                Assessment and Standards Division
                               Office of Transportation and Air Quality
                               U.S. Environmental Protection Agency
                  NOTICE

                  This technical report does not necessarily represent final EPA decisions or
                  positions. It is intended to present technical analysis of issues using data
                  that are currently available. The purpose in the release of such reports is to
                  facilitate the exchange of technical information and to inform the public of
                  technical developments.
&EPA
United States
Environmental Protection
Agency
EPA-420-R-11-009
July 2011

-------
1.     Introduction

       This document describes in detail the default fuels, fuel adjustment and air toxic
calculation algorithms contained in the MOVES2010 model released in December 2009.
It also describes the minor changes made for MOVES2010a, released in September 2010.
The MOVES2010 algorithms were substantially redesigned from those contained in
previous versions of the MOVES model. The revisions make the MOVES model more
robust in terms of fuel modeling, easier to use, and removes a number of coding errors
and problems.

       This document discusses the impact of fuel properties on emissions, particularly
the fuel adjustments applied to base emission results in MOVES. It details the fuel
adjustment model methodology, and presents limited results pertaining to the following
pollutants and their associated exhaust processes. These include the 'running', 'start' and
'extended idle' processes. It also includes evaporative processes for Benzene, MTBE and
Ethanol where simple ratios are used.  This document does not cover the impact of
gasoline properties on the total hydrocarbon evaporative emission process algorithms
such as permeation, vapor venting or leaks, which are described in the MOVES draft
technical report on evaporative emissions (Development of Evaporative Emissions
Calculations for the Motor Vehicle Emissions Simulator (Draft MOVES2009) EPA-420-
P-09-006, August 2009,
http://www.epa.gov/otaq/models/moves/techdocs/420p09006.pdf)

       In general, the MOVES  model calculates detailed "base emissions" on a base fuel,
sums the emissions to develop total  emissions for a given model year, source type and
fuel type, and then applies a fuel adjustment to this aggregate emission quantity.  The
fuel adjustment is a weighted average for the fuels  in a given area, as defined by their fuel
properties and the market share listed in the MOVES fuel supply table. Air toxic
emissions are calculated as a ratio to the emission of other pollutants (usually total
hydrocarbons), and the ratio may or may not vary with fuel properties.

       Different algorithms are used to calculate fuel effects for different fuel properties
and for different air pollutants.  Each is covered in  a separate section of this report:

   •   Complex Model Algorithms for Gasoline Vehicle Air Toxic Fuel Effects - this
       covers benzene; 1, 3 butadiene, acetaldehyde and formaldehyde
   •   MTBE Complex Model (revised between December, 2009 and September, 2010)
       provides algorithms for emissions of methyl tertiary-butyl ether from gasoline
       fueled vehicles
   •   Algorithms for other toxics from gasoline vehicles in MOVES2010, covering
       ethanol, acrolein, and naphthalene

-------
   •   Complex Model Algorithms for Carbon Monoxide
   •   Predictive Model for Total Hydrocarbon (THC) and Oxides of Nitrogen (NOx)
       Fuel Effects
   •   MOBILE6.2 Fuel Sulfur Effects Model applies to start and running THC, CO and
       NOx emissions.
   •   Combining the Complex/Predictive Models with the Fuel Sulfur Effects Model
   •   Bio-Diesel Criteria Pollutant Emission Effects
   •   Toxic Emissions for Diesel Vehicles
   •   Hydrocarbon Speciation Adjustments

       The report ends with a description of the data sources for MOVES default fuels
and with a section on results.

       The algorithms used in MOVES2010 for gasoline and diesel vehicles were
developed over 10 years ago.  EPA is currently completing analysis of more recent data,
representing more extensive testing, modern vehicles and engines, and fuels more
representative of what is currently being sold. EPA is planning a more comprehensive
fuels update that will include data from test programs being completed as part of the
analysis mandated by the 2005 Energy Policy Act (EPACT)
(http://www.epa.gov/otaq/fuels.htm ) However, this document only covers analysis, code
and data that were analyzed and implemented in the MOVES2010 and MOVES2010a
version of the model.  None of the EPAct study results were available for the
MOVES2010a version of September, 2010, and are not discussed here.

       As noted by the nomenclature of the above list, various existing fuel models were
incorporated into the MOVES algorithm. These models are:

1.      EPA Complex Model - This model was published in 1993 and is based on
       emissions from 1990 technology light duty vehicles.  It estimates fuel effects on
       exhaust emissions of volatile organic  compounds (VOC), carbon monoxide (CO),
       NOx and air toxics; and evaporative emissions of HC, including benzene. Normal
       emitters and high emitters are treated separately. The output of the model
       compares emissions of a test fuel to the Base Fuel defined in the Clean Air Act
       Amendments of 1990.  See the link below to download the Complex Model.

       http://www.epa.gov/otaq/rfg.htm.
2.      EPA Predictive Model - This model was originally created to evaluate a
       California request for a waiver of reformulated gasoline requirements. It is based
       on more recent data than that used for the Complex Model but only predicts VOC
       and NOx. The model was revised when EPA evaluated the Renewable Fuel
       Standard (RFS). It is described in some detail in an EPA report (EPA 420 R-07-

-------
       004, "Regulatory Impact Analysis: Renewable Fuel Standard Program ", USEPA,
       OTAQ, Assessment and Standards Division, April 2007).
3.      ARB Predictive Model - Last updated in 2006, this model predicts emissions of a
       test fuel compared to California Air Resources Board (ARB) Phase 3 gasoline.  It
       estimates exhaust emissions of non-methane organic gases (NMOG), NOx, and
       air toxics; and evaporative emissions of NMOG and benzene.  The time period of
       the estimate is 2015, and all light duty vehicles (cars and light trucks) were
       considered in developing the model.  No distinction is made in the model between
       normal and high emitters. This model was not explicitly included in the
       MOVES2010a version. The link for additional documentation is:
       www. epa.gov/otaq/regs/fuel s/rfg/rO 1016 .pdf
       MOBILE6/MOBILE6.2 - This model is a precursor to MOVES and predicts
       inventory emissions for mobile sources.  Within MOBILE6 is a fuels module that
       estimates how emissions will change as a function of Reid Vapor Pressure (RVP),
       oxygenate and sulfur content.  The sulfur portion of MOBILE6.2 is described in
       an EPA report. (EPA420-R-01-039, "Fuel Sulfur Effects on Exhaust Emissions,
       Recommendations for MOBILEff\ Venkatesh Rao, July 2001).
       Both the Complex and Predictive models rely on the concept of a Base Fuel(s).
These are specific fuel formulations at which the underlying vehicle tests that were used
to generate the basic emission factors in MOVES were assumed "on average" to be
conducted.  In the MOVES model, a Base Fuel will produce a MOVES fuel correction
factor of unity. The emission effects of other 'target' fuels are "ratioed" to the Base Fuel,
and the subsequent product is the MOVES fuel correction factor for that 'target' fuel.
The specific MOVES Base Fuels were chosen because they best represent the typical in-
use fuel during the period in which the MOVES test vehicle emission data were
collected.  This choice was based on local fuel survey data of the Phoenix area during the
1995 through 2002 time period when the primary MOVES vehicle dataset was tested.
(see Section 2.0 for a more complete discussion of the Base Fuels).

-------
2.      Complex Model Algorithms for Gasoline Vehicle Air
        Toxic Fuel Effects
       This section details the calculations used to compute the air toxic emission rates
in MOVES for selected air toxic compounds.  These calculations were taken from the
EPA Complex Model developed in the early 1990's. The air toxic compounds which use
these algorithms are:

       Benzene
       Acetaldehyde
       Formaldehyde
       1,3 Butadiene

       The Complex Model algorithms are applied to running, start and extended idle
emissions for gasoline fueled vehicles for all 1975 and later model years for these four
pollutants. While MOBILE6 included algorithms for older technologies not included in
the Complex Model, such as non-catalyst and oxidation catalyst vehicles, these
algorithms are not in MOVES since these vehicles now comprise such a small portion of
the fleet.  In addition, while MOBILE6.2 relied on very limited data from heavy-duty
gasoline vehicles to develop toxic to VOC ratios for that vehicle type, MOVES applies
Complex Model algorithms to both light-duty and heavy-duty gasoline vehicles.  The
general structure of this section and all subsequent ones will be to describe and discuss
the calculation algorithms in mathematical terms, show the underlying equation
coefficients and present some limited results in graphical form for selected cases (All
results are shown in a Results section at the end of the document).

       MOVES2010a also models methyl tertiary butyl ether (MTBE), ethanol, acrolein
and naphthalene.  In MOVES2010a, acrolein is not affected by fuel properties and is
modeled with a simple acrolein/VOC ratio as discussed later in this document.  MTBE is
modeled using algorithms developed from the same dataset as the Complex Model.
Ethanol and naphthalene are modeled with simpler algorithms described later in this
report.

2.1     Complex Model Mathematical Overview

       Complex Model (CM) algorithms are used to compute the fuel effects for the
pollutants benzene (exhaust and evaporative), 1,3-butadiene (exhaust), formaldehyde
(exhaust) and acetaldehyde (exhaust). These algorithms were based on testing of a
sample of Tier 0 vehicles (1990 and earlier model years) over a wide variety of gasoline
fuel formulations. A complex statistical analysis using multivariate regression techniques
was performed on the primary data set, and a set of statistical models were developed.
These statistical models were originally programmed into the Unconsolidated Complex
Model spreadsheet, and have now been programmed into the MOVES model and
described in this document.  The reader who is interested in the original test program

-------
design, the data processing, and/or the details of the statistical analysis should refer to the
original primary documents of the study.1

       The underlying goal of the CM algorithm is to compute an "air toxic ratio".  This
is the ratio (or proportion) of the particular air toxic pollutant to the amount of VOC in
the exhaust.  In MOVES, this air toxic ratio is multiplied by the MOVES VOC estimate.
Mathematically, this is shown in Equation 2-1.
AT Emission (g/hr)   =      AT Ratio * VOC Emission (g/hr)                 Eq 2-1

       "AT Emission" is the final emission estimate for the air toxic pollutant as
reported by MOVES, and "AT Ratio" is the multiplicative factor calculated by the CM
algorithm that reflects the fuel properties. The "VOC Emission" variable is the
underlying VOC emissions calculated by MOVES.

       In its native form, the CM model produces the (AT Ratio) by comparing the air
toxic emission of a given "target" fuel to those produced by a "base" gasoline fuel (see
Table 2.1). Mathematically, this relationship is shown in Equation 2-2.
AT Ratio     =     1.0 + [ (ATTarget - ATBase) / ATBase ]                 Eq 2-2
2.1.1   Base Fuel Concept

       The 'base' fuel in MOVES2010a is a specific fuel at which the basic emission
rates in MOVES remain uncorrected for fuel effects. MOVES2010a contains three
'base' fuels. There are shown in Table 2-1. Base Fuel A represents gasoline vehicles
with model years 2001 and later, Base Fuel B represents  gasoline vehicles with model
years 2000 and earlier.  Both Base Fuel A and B are used in the calculation of HC, CO
and NOx emission fuel effects. Base Fuel C is used as the base fuel in the air toxic
equations for all model years.

       Base Fuels A and B were chosen because they best represent the typical in-use
fuel during the period in which the MOVES test vehicle emission data were collected.
The underlying test vehicle emission data is a large sample of in-use vehicles as they
received their state transient EVI240  or EVI147 test.  Since fuel properties were not
measured on individual test vehicles during their emission test, it is presumed that the
base (uncorrected) emissions best represent "on average" the presence of these fuels and
their properties.  The assumption of 'best represent' is based on local fuel survey data of
the Phoenix area during the 1995 through 2002 time period when the primary MOVES
vehicle dataset was tested.

       Over the time period of the emission testing, the general in-use sulfur level  varied
due to EPA fuel sulfur regulations.  As a result, MOVES2010a contains two base fuels
1 U. S. EPA.  1993.  Final Regulatory Impact Analysis for Reformulated Gasoline.
December 13, 1993. http://www.epa.gov/otaq/regs/fuels/rfg/ (ria.zip)

-------
with the same fuel properties except fuel sulfur. Base Fuel A (a lower sulfur fuel) is used
to model the 2001 and later model year vehicles, and Base Fuel B is used to model the
2000 and earlier model year vehicles. The properties of each fuel are shown in Table 2-1
below.

       Base fuel C is the base fuel used only in the air toxic ratio calculations. It is a fuel
which was prevalent in the early 1990's, and is the base fuel for the Complex model.
This fuel formulation is central to the "centering values" and other coefficients and
algorithms in the Complex model, and was the baseline fuel specified in the 1990 Clean
Air Act Amendments for air toxic calculations.
2.1.2   Target Fuel Concept

       The "target" gasoline is the gasoline which is to be evaluated for its effect on air
toxic emissions.  The target gasoline may vary by county, year and month.
MOVES2010a contains a large set of fuel formulations and associated fuel market share
fractions for each of the 3,222 United States counties, for each month and for calendar
years  1990 and 1999 through 2012. These fuel data were assembled under EPA contract
with EH Pechan who analyzed in-use fuel surveys from 2005 and made projections based
statistics and EPA fuel regulations.  In addition, to the large set of built-in fuel data, the
user may enter their own fuel formulations and market share information, if available.
Note,  that the use of a 'base' fuel as a 'target' fuel will produce a fuel correction of unity.
Table 2-1
MOVES Base Fuel Properties
Fuel Property
Name
Fuel Sub-Type
RVP
Sulfur Level
ETOH Volume
MTBE Volume
ETBE Volume
TAME Volume
Aromatic Content
Olefin Content
Benzene Content
E200
E300
Volume to percent
Oxygen
Base Fuel A
Conventional Gas
6.9 psi
30.0 ppm
0.0 %
0.0 %
0.0 %
0.0 %
26.1%
5.6%
1.0%
41.1%
83.1%
0.0 %
Base Fuel B
Conventional Gas
6.9 psi
90.0 ppm
0.0 %
0.0 %
0.0 %
0.0 %
26.1 %
5.6 %
1.0%
41.1 %
83.1 %
0.0 %
Base Fuel C
Conventional Gas
8.7 psi
338.0 ppm *
0.0 %
0.0 %
0.0 %
0.0 %
26.4 %
11.9%
1.64%
50.0 %
82.0 %
0.0 %
• Air toxic computation is not a function of sulfur level

-------
       For both the "Target" and "Base" fuels, the overall Complex model consists often
individual, statistically derived models (see Section 2.2 below for more details). Each of
these ten fuel models predicts emissions as an exponential function of fuel parameters
and associated coefficients.  Each has a weighting factor which varies by vehicle model
year that is used to weight the contribution of all ten models into a composite result.

       The MOVES algorithm to compute these emission effects is complex and is
detailed in the next section.  In summary, for each of the ten Complex Model sub-
models, MOVES sums the coefficients for each model parameter, calculates the
difference between the fuel value and an average fuel value, multiples the coefficient sum
by this difference, computes the  sum of the resulting values across the complex model
parameters and computes the exponential function of this value.

       MOVES then calculates a ratio between the exponential values for the target and
base fuels for each of the Complex Model sub-models. The sub-model  ratios are
weighted together to create a weighted average fractional difference in air toxic emissions
between the base and target fuels. This information is used to predict the toxics
emissions for each target fuel. Similar calculations are used to predict the VOC
emissions for the same fuels, and the resulting ratios of the predicted  toxics and VOC are
calculated.
2.2    Complex Model Mathematical Algorithm

       The Complex Model algorithm is presented in this section in a series of nine
calculation steps.  Each variable is defined in the steps and where and in some cases an
example calculation is provided.
2.2.1          Step 1

       This series of steps are used only for pollutants:

       Benzene,
       1,3-butadiene,
       Formaldehyde
       Acetaldehyde


       They are performed for each of the MOVES emission processes:

       running
       start
       evaporative processes

-------
       There are up to 64 different Complex Model parameter (cmpID) variables that
represent the effects of individual fuel parameters and cross products of individual fuel
parameters determined from statistical analysis of the data.  Each of the ten Complex
Model sub-models has coefficients that must be added. The addition for each parameter is
done across each model.

       The variable baseSumCoeff is the sum of coefficients for the base fuel, and the
variable targetSumCoeff is the sum of coefficients for the target fuels.  The structure of
the table containing the coefficients is given in Table 2-2.
basesumcoeff
targetsumcoeff
              coeffl + coeff2
              coeffl + coeff2
Eq 2-3 a
Eq 2-3b
       The calculations are performed for each fuel-type (i.e., gasoline and E-85 only),
for each fuel formulation, for each fuel model year combination and for all cmpID
variables (denoted as cmpID in Table 2-2).  Currently, the variable "coeffS" is not used
in the calculation. It is reserved for possible future use.
Table 2-2 ComplexModelParameters
Fields
polProcessID
fuelModellD
cmpID
coeffl
coeff2
coeffi
Datatype
smallint(6)
smallint(6)
smallint(6)
float
float
float
Description
PK: standard polProcessID variable
PK: basic fuel and technology model index
PK: all fuel property variable
First order coefficients used in calculations
Second order coefficients used in calculations
Third order coefficients (placeholder) used in
calculations. None are currently used.
2.2.2a
Step 2a
       In this step 2a, the baseSumCoeff variable computed in Step 1 along with two
additional variables "baseValue" and "centeringValue" are used to calculate a subsequent
variable BaseProd. This calculation is performed only for the 'base' fuel (an analogous
calculation is performed for the target fuel). Like Step  1 this calculation is performed for
each pollutant / process / fuel-type / fuel formulation (base only) / fuelModellD and
cmdID combination.  Mathematically,  the equation is given by Eq 2-4a.

-------
Baseprod
basesumcoeff * (base Value - centering Value)
Eq2-4a
       The variable "baseValue" is found in Table 2-1 for all of the parameters. The
'baseValue' variable and the "centeringValue" variable are listed. The table
MeanFuelParameters has the following structure as used in MOVES.

fueltypeid
modelyeargroupid
fuelparameterid
fuelparametername
basevalue
centeringvalue
       The baseValue variable is the actual value for the base fuel (i.e., sulfurlevel, RVP,
E200, etc).  The "centering value" a mean value used in the calculations.

       Some cmdIDs describe the product of two fuel parameters such as oxygen and
RVP. This special case is illustrated in Eq2-4b for a baseprod for the cmpID that is the
product of oxygen and RVP.
Baseprod    =
basesumcoeff * (Oxygen - Oxygen centering value) *
       (RVP-RVP centering value)                     Eq2-4b
Table 2-3 MeanFuelParameters
Fields
polProcessID
fuelTypelD
model YearGroupID
fuelParameterlD
baseValue
centeringValue
stdDevValue
Datatype
smallint(6)
smallint(6)
int(ll)
Smallint(6)
float
float
float
Description
PK: standard polProcessID variable
PK: fuelTypelD = 1 only
PK: two new myglDs added
PK: major fuel property variable
Baseline fuel properties.
Fuel parameter mean value used in calculations.
Fuel parameter standard deviation used only in
HC and NOx calculations.

-------
2.2.2b        Step 2b

       In this step 2b, the targetSumCoeff variable (calculated in Step 1) along with the
variables "fuelParameter" and "centeringValue" are used to calculate a subsequent
variable TargetProd.  This calculation is performed for each of the target fuels which are
being evaluated.  Like Step 1 this calculation is performed for each pollutant / process /
fuel-type / fuelformulationID (target fuels only) / fuelModellD and all possible cmdID
combinations. Mathematically, the equation is given by Eq 2-5a. The variable
"centeringValue" comes from the MeanFuelParameters.xls table. The fuelparameter
variable is the particular fuel parameter for the target fuel (i.e., sulfurlevel, RVP, E200,
etc) that the user wishes to evaluate. The calculation is performed separately for each
fuel parameter.
Targetprod   =      targetSumCoeff * (fuelparameter - center Value)        Eq 2-5 a

       In some cases for the Target fuel, a particular cmdID may include a fuel
parameter which is a product of two fuel parameters. This special case is illustrated in
Eq2-5b for a Targetprod that is the product of oxygen and RVP.
Targetprod   =             Targetumcoeff * (Oxygen - Oxygen centering value) *
                           (RVP-RVP centering value)                     Eq2-5b
2.2.3         Step 3

       In this step the variables eBaseProdSum and eTargetProdSum are calculated for each
pollutant, process, fuel-typelD, fuel-formulationID (base and target), and fuelModellD
combination.  The variables eBaseProdSum and eTargetProdSum are the exponential sum of
each of the individual BaseProd and TargetProd variables.  The sums are across all of the
possible cmdID (1 to 64) groups.

eBaseprodSum = EXP [SUM(Baseprod) where cmdID = 1 to 64 ]              Eq2-6a

eTargetprodSum = EXP [SUM(Targetprod) where cmdID = 1 to 64 ]           Eq2-6b
In cases where the SUM(Baseprod) or SUM(Targetprod) equals zero, the value of
eBaseprodSum or eTargetprodSum is set to zero rather than calculated as unity.
                                       10

-------
2.2.4
Step 4
       In this step the variable Ratiol is calculated for each pollutant, process,
fueltypelD, fuelModellD, fuelformulationID (the unchanging base fuel and variable
target fuel) and fuelModellD combination using Equation 2-7.  There are eleven
fuelModellD's for the primary exhaust air toxic pollutants.
Ratiol =     ((eTargetprodSum / eBaseprodSum ) - 1.0 )
                                                            Eq2-7
2.2.5
StepS
       In this step, the variable Ratio2 is computed by multiplying Ratiol by a
fuelModelWtFactor variable. This accounts for the mix of technologies and high emitters
by model year. Each of the ten fuel models has an associated weighting factor which is a
function of agelD, and two broad model year groups (1960-2000 and post-2000).  The
structure of the table is shown in Table 2-4, and the names of the individual fuel models
for air toxic emissions are provided in Table 2-5.  The individual fuel models represent
fuel metering, EGR, catalyst and air injection technology groupings, as well as high
emitters.  The weighting factors account for the prevalence of each type of vehicle.

       The weighting factors which are used in MOVES differ from those used in the
original Complex model.  They vary by model year group because of the changing
important of technology groupings (the original does not vary), and there is now less
emphasis on so called "High emitters". The original Complex model gave a 55 percent
weighting to high emitters (i.e., fuel model = 10). The fraction of high emitters now
varies with age. It ranges from 0.01 percent at age zero to 32.8 percent at 30 years.
Table 2-4 FuelModel Weighting Factors
Fields
fuelModellD
model YearGroupID
agelD
fuelModelWtF actor
Datatype
smallint(6)
int(ll)
smallint(6)
float
Description
PK: basic fuel and technology model index
Only two groups
Age 0 through 30
Weight factor that sums to unity for each
fuelModellD and modelYearGroupID
                                       11

-------
Table 2-5 FuelModel Names
Fuel Model ID
1
2
3
4
5
6
7
8
9
10
11
Fuel Model Name
PFI & Sway & No Air & EGR
PFI & Sway & No Air & No EGR
TBI & Sway & No Air & EGR
PFI & 3way+Ox & Air & EGR
PFI & Sway & Air & EGR
TBI & Sway & Air & EGR
TBI & 3way+Ox & Air & EGR
TBI & Sway & No Air & No EGR
GARB & 3way+Ox & Air & EGR
All High Emitters
All Tier2
Ratio2
       Ratio 1 * fuelModelWtFactor
Eq2-8
Note that FuelModel = 11 (the All Tier2 model) is not part of the calculation or the
weighting (it currently receives a weighting factor of zero). It is reserved for future use.
2.2.6
Step 6
       In this step, the Ratio2 variables are summed across the ten fuel models.  This
produces the variable atDifferenceFraction. It is a weighted sum of all of the fuel models
and the base and target fuel effects. In essence, it is the fractional difference in air toxic
emissions between the base and target fuels.  It is calculated for each combination of
pollutant, process, fueltypelD, agelD  and fuelformulationlD.
atDifferenceFraction  =      SUM (Ratio2)  where i = 1,max(fuelmodels)
                                                             Eq2-9
fuelmodel [i]
max (fuelmodels)
              10    for air toxic pollutants
2.2.7
Step 7
       In this step the variable RelATEmissions is calculated for the Target fuel using
Equation 2-10. This variable represents the air toxic emissions on the target fule with no
other adjustments. The variable RelATEmissions includes the air toxic emissions effect
of the change in fuel properties of the Target fuel versus the Base fuel.  If the base fuel
were identical to the Target fuel, the variable atDifferenceFraction would be zero and the
RelATEmissions would be the same as the atBaseEmissions.
                                        12

-------
       The atBaseEmissions are the air toxic emission rates in units of grams per mile,
and are a function of month of the year.  They were taken directly from the Complex
Model Spreadsheet "CMFinal.xls".  They are the base air toxic emissions for the Base
Fuel in the air toxics model. These base emissions were calculated from air toxic
emission data from the Tier 0 vehicles in the original air toxic / fuel effects studies on
which the Complex Model was based. They are used in conjunction with  similar VOC
emissions (see Step 8 below) to compute a final air toxic to VOC emission ratio which is
used in the MOVES model.  The MOVES air toxic algorithm is based on the assumption
that within a fuel model group (i.e., technology group) the air toxic pollutant to VOC
ratio has remained constant even as vehicle emission standards have been  lowered from
TierO to Tier2 levels. When the results of the EPACT test program are fully available,
this assumption will be re-evaluated and new ratios / fuel models may be developed for
Tier2 vehicles.  The "All Tier2" fuel model is  currently a placeholder for these new
values / coefficients.
  RelATEmissions = atBaseEmissions + atBaseEmissions * atDifferenceFraction     Eq2-10
2.2.8          Step 8

       Steps 1 through 7 are repeated for the VOC pollutant.  Only the final equation
where the RelVOCEmissions variable is shown in Equation 2-11 is repeated. It is the
denominator for the final air toxic to VOC ratio.  This is NOT the computation of the
standard VOC pollutant in MOVES. These steps compute a VOC value for the variables
atDifferenceFraction and RelATEmissions. Like the air toxic pollutant, the
atBaseEmission value for VOC was also based the original testing of TierO vehicles.

  RelVOCEmissions = VOCBaseEmissions + VOCBaseEmissions * atDifferenceFraction     Eq2-l 1
2.2.9   Step 9

       This step in Equation 2-12 calculates the final air toxic to VOC ratio which is
used in the MOVES model. This algorithm is used for processes of running and start and
for pollutantTD (20, 24, 25 and 26}. It is used only for gasoline fueled vehicles. It shall
be computed for each pollutant-process,  fuelformulationID, and monthGroupID.

atRatio       =     RelATEmissions  /RelVOCEmissions                   Eq2-12
       The variable atRatio is used in MOVES to compute the air toxic emissions for benzene,
1,3-butadiene, formaldehyde and acetaldehyde. This is done by multiplying the appropriate
atRatio by the final MOVES VOC emission rate after fuel, temperature and I/M effects have
been applied.
                                       13

-------
2.3 MTBE Complex Model Algorithm

2.3.1   MTBE Exhaust Model

       As of calendar year 2008, MTBE (MOVES pollutantTD = 22) has been almost
completely phased-out of the fuel supply of the United States due to ground water
contamination concerns. Thus, its inventory levels in the MOVES output from default
inputs should be very small if not zero in future years.  It is present in the MOVES model
as mostly a legacy pollutant for calendar years 1990, 1999 - 2005.  However, the MTBE
fuel volume is a user input, and MOVES has the capability to calculate MTBE emissions
for any calendar year.

       The MTBE Exhaust model was developed from a simple empirical analysis of
MOBILE6.2 outputs. MOBILE6.2 relied on a draft MTBE fuel effects model.2'3 This
process included several detailed MOBILE6.2 runs, and a simple regression of the
exhaust MTBE / VOC ratio versus MTBE fuel volume. A quadratic equation fixed at the
origin was found to be the best fit. The equation is shown in Equation 2-13 and the
parameters are shown in Table 2-6.  The same equation is used for both start and running
processes.

MTBE/VOC ratio    =      A * MTBE Volume+ B*(MTBEVolume)A2        Eq2-13

Where

A     =      ComplexModelParameters.coeffA
B     =      ComplexModelParameters.coeffB
Table 2-6
MOVES MTBE Exhaust Calculation Coefficients
polProcessID
2201
2202
Process Description
Running Exhaust
Start Exhaust
Coeff A
0.00007809
0.00007809
Coeff B
0.00007537
0.0007809
       Like the other air toxic ratios, the MTBE / VOC ratio is multiplied by the
MOVES VOC emission factors to produce a MOVES MTBE emission factor in the
2 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.

3 Rich Cook and Edward L. Glover. 2002. Technical Description of the Toxics Module for MOBILE6.2
and Guidance on Its Use for Emission Inventory Preparation. Assessment and Standards Division, Office
of Transportation and Air Quality, Ann Arbor, MI. Report No. EPA420-R-02-011.
http://www.epa.gov/otaq/models/mobile6/r02029.pdf

                                        14

-------
appropriate units (i.e., grams per hour, grams per start, grams per mile). In MOVES the
MTBE / VOC ratio ranges from 0.015 to 0.15 in cases where MTBE is present in the
fuel. The MTBE ratio is 0.000 in cases where no MTBE is present in the fuel.
2.3.2   MTBE Evaporative Model

       MTBE ratios are estimated using algorithms originally developed for MOBILE
6.2, using the MTBE fuel effects model cited above. However, evaporative emissions
processes for MOVES differ from those in MOBILE6.2.  Thus, algorithms for hot soak in
MOBILE6.2 are used for vapor venting and refueling vapor loss in MOVES, and
algorithms for running loss are used for fuel leaks and refueling spillage loss. The
MOBILE6.2 algorithm for resting loss is used for permeation.
       The equation for MTBE fuel vapor venting and refueling displacement vapor loss:

MTBE / VOC = MTBEVolume*(24.2050 - 1.7460*RVP)/1000.0


       The equation for MTBE fuel leaks and refueling spillage loss:

MTBE/VOC = MTBEVolume*(17.8538 - 1.6622*RVP)/1000.0

       The equation for MTBE Permeation:

MTBE / VOC = MTBEVolume*(22.1980 - 1.7460*RVP)/1000.0


2.4 Benzene Evaporative Model Algorithm

       The equations for Benzene evaporative emissions are simpler that the ones used
for exhaust Benzene, and there are no Base fuels in the equations.


       The equation for Benzene Permeation is:

Benzene / VOC = (BenzeneContent * (-0.02895*OXY - 0.080274*RVP + 1.3758)7100.0) +
                    (0.77*(benzeneContent*(-0.02895*ETOHVolume*0.3488-
                    0.080274*RVP + 1.3758)/100.0)*ETOHVolume/10.0)

This equation is based on the diurnal emissions algorithm from MOBILE6.2.  The diurnal
emissions equation accounts for impacts of changing oxygenate, RVP and fuel benzene
levels.  However, a study of permeation emissions suggests that the ratio of benzene from
permeation to total VOC is about 1.77 times higher than the ratio associated with
evaporation.4 Thus the ratio predicted by the diurnal emissions  algorithm was multiplied
by 1.77.
4 Haskew, 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.

                                       15

-------
      For other evaporative processes, equations for hot soak in MOBILE6.2 are used:

Benzene / VOC = (BenzeneContent * (-0.03420*OXY - 0.080274*RVP + 1.4448)7100.0)


      Where OXY is:

OXY    =   ETOHVolume*0.3488 + MTBEVolume*0.1786 + ETBEVolume*0.1533
            + TAMEVolume*0.1636
                                    16

-------
3.      Algorithms for Ethanol, Acrolein and Naphthalene from
       Gasoline Vehicles in MOVES2010

3.1    Ethanol

Ethanol emissions in MOVES are calculated by applying an Ethanol / VOC ratio to the
calculated VOC emissions in MOVES.  Equation 3-1 is used.
Ethanol
Ethanol/VOC Ratio * VOC
Eq3-l
              For vehicles running on 10% ethanol, 2.39 percent of exhaust VOC is
estimated to be ethanol.  This estimate is based on tests on 9 vehicles from 4 test
programs.5'6'7'8 MOVES also has ethanol to VOC ratios for E5 and E8, based on linear
interpolation from the E10 value (0.01195 for E5 and 0.01912 for E8). The exhaust
ethanol / VOC ratios for gasoline fuels (i.e. pure gasoline or E10 or less) are shown in
Table 3-2.
Table 3-2
Exhaust Ethanol / VOC Ratios for Gasoline Fuels
Ethanol Fuel Volume (%)
0
5
8
10
Ethanol
/ VOC Ratio
0.00000
0.01195
0.01912
0.02390
Toxic to VOC ratios for E85 and E70 are given in Table 3-3. E85 exhaust ratios for all
pollutants except naphthalene were obtained from data on seven vehicles from a 1995 test
program in EPA's Office of Research and Development, along with data from a 2007 test
program at Southwest Research Institute and a 2005 test program at Environment
  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.

Environment Canada, 2007. Comparison of Emissions from Conventional and Flexible Fuel Vehicles
Operating on Gasoline and E85 Fuels. ERM Report No. 05-039, Emissions Research Division.

7Durbin 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.

8 U. S. EPA, 1995. Office of Research and Development, unpublished data.
                                        17

-------
Canada.9'10'11 E70 exhaust ratios are from the 2007 test program at Southwest Research
Institute cited above.9 E85 benzene evaporative emission ratios were derived from
speciation data collected as part of the Auto/Oil test program in the early 1990's.12E85
benzene permeation ratios are from the CRC E-65 test program.4  Ratios for evaporative
and permeation emissions from E70 fuel were estimated by multiplying the E85 ratios by
70/85.
Table 3-3
Ethanol / VOC Ratios for High Ethanol Fuels
Ethanol Fuel Volume
E85
E85
E85
E85
E85
E85
E85

E70
E70
E70
E70
E70
E70
E70
Process
Running
Start
Permeation
Fuel Vapor Venting
Fuel Leaks
Refueling Vapor Loss
Refueling Spillage

Running
Start
Permeation
Fuel Vapor Venting
Fuel Leaks
Refueling Vapor Loss
Refueling Spillage
Ethanol / VOC Ratio
0.5348
0.5348
0.5940
0.6123
0.6123
0.6123
0.6123

0.7036
0.7036
0.4892
0.5042
0.5042
0.5042
0.5042
3.2    Acrolein and Naphthalene

       The toxic to VOC ratio for acrolein carries over from MOBILE6.2. Acrolein is
found only in exhaust.  Naphthalene is carried over from the National Mobile Inventory
                13
Model (NMEVI).   Exhaust naphthalene and other PAH emissions are estimated as a ratio
 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.  Available in Docket EPA-HQ-OAR-2005-0161.

10 Environment Canada. 2007. Comparison of Emissions from Conventional and Flexible Fuel Vehicles
Operating on Gasoline and E85 Fuels.  ERM Report No. 05-039, Emissions Research Division. Available
in Docket EPA-HQ-OAR-2005-0161.

11 Graham,  L. A.; Belisle, S. L. and C. Baas. 2008. Emissions from light duty gasoline vehicles
operating on low blend ethanol gasoline and E85. Atmos. Environ. 42: 4498-4516.

12 Auto/Oil Air Quality Improvement Research Program.  1990. Phase 1 Working Data Set (published in
electronic form).  Prepared by Systems Applications International, San Rafael, CA.

13 Michaels, H., Brzezinski, D., Cook, R. 2005.  EPA's National Mobile Inventory Model (NMIM), A
Consolidated Emissions Modeling System for MOBILE6 and NONROAD. U. S. EPA, Office of
                                           18

-------
to PM in NMTM.  This approach is used, even though naphthalene is found in the gas,
semi-volatile and particle phase, because there is generally reasonable correlation
between PAH and PM emissions.  However, for future versions of MOVES, PAH
exhaust emissions will be apportioned into the gas and particle phase, and gas phase
PAHs such as naphthalene will be estimated using toxic/VOC ratios and particle phase
PAHs will be estimated using toxic to OC2.5 (organic carbon less than 2.5 microns)
ratios. Evaporative naphthalene is estimated as a ratio to VOC in MOVES2010a.
Table 3-3
Toxics Ratios for Acrolein and Naphthalene from Gasoline Vehicles
Pollutant
Acrolein
Naphthalene
Naphthalene
Emission Type
Exhaust
Exhaust
Evaporative and
Permeation
Ratio Type
VOC
PM
VOC
Ratio
0.00063
0.0880
0.0004
Transportation and Air Quality, Assessment and Standards Division, Ann Arbor, MI, March 2005; Report
No. EPA-420-R-05-003. Available at http://www.epa.gov/otaa/nmim.htm.
                                        19

-------
4.     Complex Model Algorithms for Carbon Monoxide Fuel
       Effects
       This section describes in detail the calculations used to compute the fuel effects
for carbon monoxide emission in MOVES. These calculations were taken from the EPA
Complex Model developed in the early 1990's.  The algorithms include methodologies
for calculating running and start emissions from gasoline vehicles. The general structure
of this section will be to describe and discuss the calculation algorithms in mathematical
terms, show the underlying equation coefficients and present some limited results in
graphical form for selected cases.

       The Complex Model (CM) algorithm is used to compute the fuel effects for
carbon monoxide. This algorithm was developed in the mid-1990's based on testing of a
sample of mostly TierO vehicles (1993 and earlier model years) over a wide variety of
gasoline fuel compositions.  A complex statistical analysis using multivariate regression
techniques was performed on the primary database and a set of statistical models were
developed.  These statistical models were originally programmed  into the Unconsolidated
Complex Model spreadsheet, and have now been programmed into the MOVES model
and described in this document. The reader who is interested in the original test program
design, the data processing, and/or the details of the statistical analysis must refer to the
original primary documents of the study.

http://www.epa.gov/oms/models.htm

       The algorithm for the Complex Model for CO emission fuel effects is almost
completely analogous to the Complex Model for air toxic emissions.  However, there are
some important differences. The first difference is that the CO fuel effects are not a ratio
to VOC emissions. Instead, the fuel effects for a given target fuel are a ratio to a base
reference fuel whose fuel adjustment is defined as one.  This reference fuel is the fuel(s)
that best represent(s) the test conditions at the Arizona EVI lane where the primary
emission factor data were acquired. See Section 2.1 and Table 2-1 for more details on the
base / reference fuels.   The second difference is the use of the Complex model only to
compute the emission effect contributions from the non-sulfur fuel parameters. For CO,
a separate fuel sulfur model (MOBILE6 Sulfur Model  (M6SM)) is applied to compute
the fuel sulfur effects after the basic complex model calculations are done (see Section
5.0-Fuel Sulfur Model).
3.1    Carbon Monoxide Complex Model Mathematical Algorithm

       The Complex Model algorithm is presented in this section in a series of nine
calculation steps. Each variable is defined in the steps and where possible the actual
calculation coefficients (or at least a sample calculation) are provided.
                                       20

-------
       Because the effects of fuel sulfur are applied in the MOVES Sulfur model, they
have been removed from the CO Complex model by setting the fuel sulfur level to a
dummy value of 30 ppm for all of the target fuels. This will cause the multiplicative
sulfur effects to be unity.  The multiplicative MOVES Sulfur model is applied subsequent
to the CO Complex model.
       The CO algorithm uses the same algorithm as the Complex model air toxics
algorithm for Steps 1 through 6 (i.e., Section 2.2). Different values for the coefficients
are used, and a different set of Excel workbooks are used to contain the coefficients and
other parameters.  These steps are repeated in this section with only minor name changes
for the sake of the reader who only reads the carbon monoxide section. The reader who
has already read Sections 2.2.1 through 2.2.6 may find this section redundant.
3.1.1          Step 1

       This series of steps are used only for the carbon monoxide (CO) pollutant.

       They are performed for each of the MOVES emission processes:

       running,
       start
       extended idle / crankcase

       They are performed for both the Target and Base fuels.  The variable
baseSumCoeff is the sum for the base fuel, and the variable targetSumCoeff is the sum
for the target fuels.
basesumcoeff       =      coeffl + coeff2                                Eq 3-1 a
targetsumcoeff      =      coeffl + coeff2                                Eq 3-lb
       The calculations are performed for each fueltype (i.e., currently gasoline only),
for each fuelFormulation, for each fuel model year combination and for all cmpID
variables (denoted as cmpID in Table 3-1). There are up to 64 different cmpID variables
that represent the effects of individual fuel parameters and cross products of fuel
parameters determined from statistical analysis of the data. Most of the individual CO
fuel models only use a fraction of the 64 different cmpID variables. Currently, the
variable "coeffi" is not used in the calculation. It is reserved for future possible use.
                                       21

-------
Table 3-1 ComplexModelParametersCO
Fields
polProcessID
fuelModellD
cmpID
coeffl
coeffZ
coeffi
Datatype
smallint(6)
smallint(6)
smallint(6)
float
float
float
Description
PK: standard polProcessID variable
PK: basic fuel and technology model index
PK: all fuel property variable
First order coefficients used in calculations
Second order coefficients used in calculations
Third order coefficients (placeholder) used in
calculations.
3.1.2a
Step 2a
       In this Step 2a, the baseSumCoeff variable computed in Step 1 along with two
new variables "baseValue" and "centeringValue" are used to calculate a subsequent
variable BaseProd.  This calculation is performed only for the 'base' fuel (an analogous
calculation is performed for the target fuel). Like Step 1 this calculation is performed for
each pollutant / process / fuel-type / fuelformulationID (base only) / fuelModellD and
cmdID combination.  Mathematically, the equation is given by Eq 3-2a.
Baseprod
       basesumcoeff * (baseValue - centering Value)
Eq3-2a
       The variable "baseValue" is found in Table 2-1 for all of the parameters. The
table MeanFuelParametersCO has the following structure as used in MOVES.

fueltypeid
modelyeargroupid
fuelparameterid
fuelparametername
basevalue
centeringvalue
       The baseValue variable is the particular fuel parameter cmpID for the base fuel
(i.e., sulfurlevel, RVP, E200, etc).  The calculation is performed separately for each fuel
parameter.  The "centering value" is a mean value variable used in the calculations.  It
was determined as part of the original statistical analysis of the data using a mixed model
regression.

       In some cases for the Base fuel, a particular cmdID may include a baseValue fuel
parameter which is a product of two fuel parameters  such as  oxygen and RVP.  This
special case is illustrated in Eq3-2b for a baseprod that is a function of oxygen and RVP.
                                       22

-------
Baseprod    =
       basesumcoeff * (Oxygen - Oxygen centering value) *
              (RVP-RVP centering value)                     Eq3-2b
Table 3-2 MeanFuelParameters
Fields
polProcessID
fuelTypelD
model YearGroupID
fuelParameterlD
baseValue
centeringValue
stdDevValue
Datatype
smallint(6)
smallint(6)
int(ll)
Smallint(6)
float
float
float
Description
PK: standard polProcessID variable
PK: fuelTypelD = 1 only
PK: two new myglDs added
PK: major fuel property variable
Baseline fuel properties. Compared to Arizona
fuel parameters for HC, CO and NOx.
Fuel parameter mean value used in calculations.
Fuel parameter standard deviation used only in
HC and NOx calculations.
3.1.2b
Step 2b
       In this step 2b, the targetSumCoeff variable (calculated in Step 1) along with the
variables "fuelParameter" and "centeringValue" are used to calculate a subsequent
variable TargetProd.  This calculation is performed only for each of the target fuels which
are being evaluated. Like Step 1 this calculation is performed for each pollutant / process
/ fuel-type / fuelformulation (target fuels only) / fuelModellD and all possible cmdID
combinations.  Mathematically, the equation is given by Eq 3-3a. The fuelparameter
variable is the particular fuel parameter for the target fuel (i.e., sulfurlevel, RVP, E200,
etc) that the user wishes to evaluate.  The calculation is performed separately for each
fuel parameter.
Targetprod    =     targetSumCoeff * (fuelparameter - center Value)
                                                           Eq 3-3a
       In some cases for the Target fuel, a particular cmdID may include a fuel
parameter which is a compounding of two fuel parameters such as oxygen and RVP.
This special case is illustrated in Eq3-3bfor a Targetprod that is a function of oxygen and
RVP.
Targetprod   =
             Targetumcoeff * (Oxygen - Oxygen centering value) *
              (RVP-RVP centering value)                     Eq3-3b
                                       23

-------
3.1.3         Step 3


       In this step the variables eBaseProdSum and eTargetProdSum are calculated for each
pollutant, process, fueltypelD, fuelformulationID (base and target), and fuelModellD
combination.  The variables eBaseProdSum and eTargetProdSum are the exponential sum of
each of the individual BaseProd and TargetProd variables. The sums are across all of the
possible cmdID (1 to 64) groups.

eBaseprodSum = EXP  [SUM(Baseprod) where cmdID = 1 to 64 ]               Eq3-4a

eTargetprodSum = EXP [SUM(Targetprod) where cmdID = 1 to 64 ]            Eq3-4b
In cases where the SUM(Baseprod) or SUM(Targetprod) equals zero, the value of
eBaseprodSum or eTargetprodSum is set to zero rather than calculated as unity.
4.1.4         Step 4
       In this step the variable Ratiol is calculated for each pollutant, process,
fueltypelD, fuelModellD, fuelformulationID (the unchanging base fuel and variable
target fuel) and fuelModellD combination using Equation 3-5.  There are ten
fuelModellD's for the carbon monoxide model.

Ratiol =     ( (eTargetprodSum / eBaseprodSum ) - 1.0 )                    Eq3-5
3.1.5         StepS

       In this step, the variable Ratio2 is computed by multiplying Ratiol by a
fuelModelWtFactor variable.  Each of the ten fuel models has an associated weighting
factor which is a function of agelD (31 years) and two broad model year groups (1960-
2000 and post-2000). The structure of the table is shown in Table 3-3, and the names of
the individual fuel models are provided in Table 3-4.  The individual fuelmodellD's
represent fuel metering, EGR, catalyst and air injection technology groupings, and the
weighting factors provide a relative weighting between these groups.  The weighting
factors which are used in MOVES are different from those used in the original Complex
model. They now vary by model year group because of the changing important of
technology groupings (the original does not vary). There is now less emphasis on so
called "High emitters". The original Complex model gave a 55 percent weighting to high
emitters (i.e., fuel model = 10). The fuelmodellD weighting factors now weight the
models according to vehicle age.  With fuelmodel=10 getting a weighting (based on
MOBILE6.2 - MOVES does not contain the concept of "High" and "Normal" emitters)
that ranges for 0.01 percent at age zero to 32.8  percent at 30 years.
                                      24

-------
Table 3-3 FuelModel Weighting Factors
Fields
fuelModellD
modelYearGroupID
agelD
fuelModelWtF actor
Datatype
smallint(6)
int(ll)
smallint(6)
float
Description
PK: basic fuel and technology model index
Only two groups
Age 0 through 30
Weight factor that sums to unity for each
fuelModellD and modelYearGroupID
Table 3-4 FuelModel Names
Fuel Model ID
1
2
3
4
5
6
7
8
9
10
11
Fuel Model Name
PFI & Sway & No Air & EGR
PFI & Sway & No Air & No EGR
TBI & Sway & No Air & EGR
PFI & 3way+Ox & Air & EGR
PFI & Sway & Air & EGR
TBI & Sway & Air & EGR
TBI & 3way+Ox & Air & EGR
TBI & Sway & No Air & No EGR
GARB & 3way+Ox & Air & EGR
All High Emitters
All Tier2
Ratio2
       Ratio 1 * fuelModelWtF actor
Eq3-6
Similar to the air toxics model, model #11 - the All Tier2 model is not part of the
calculation and receives a weighting of zero.
3.1.6
Step 6
       In this step, the Ratio2 variables are summed across the ten fuel models. This
produces the variable CODifferenceFraction. It is a weighted sum of all of the fuel
models and the base and target fuel effects.  In essence, it is the fractional difference in
                                        25

-------
CO emissions between the base and target fuels.  It is calculated for each combination of
pollutant, process, fuel type, age and fuel formulation.

CODifferenceFraction=      SUM (Ratio2)  where i = 1,max(fuelmodels)      Eq3-7

fuelmodel [i]
max (fuelmodels)      =      10     for carbon monoxide
3.1.7          Step?
       This step differs from the air toxics algorithm described in Section 2. The non
sulfur fuel effects on CO emissions are contained in the CODifferenceFraction variable.
By definition, the Base fuel is the fuel formulation that was generally available in the
State of Arizona during the period where EPA primary test data were collected on carbon
monoxide emissions. Since sulfur effects are not included in this model only the low
sulfur (30 ppm sulfur) reference - base fuel A (see Table 2-1) may be used as the Base
fuel.  The target fuel is the fuel which is to be evaluated. The variable
CODifferenceFraction computed in Eq3-8 contains the fractional emission difference
between the target and the base fuels.
COFuelEffectNoSulfur =    1.0 + CODifferenceFraction                      eq3-8

       Eq3-8 calculates the CO fueladjustment factor without sulfur.  This algorithm is
used only for the processes of running and start and for the CO pollutant, and is used only
for gasoline fueled vehicles.  It is computed for each age, fuel formulation and model
year group (MOVES contains two distinct model year groups with different base sulfur
levels).
3.1.8   StepS

       Apply the MOBILE6 Sulfur Model (M6SM) to CORatioNoSulfur to compute the
final CO Fuel Adjustment Factor which is used to correct the CO emission factors in
MOVES for gasoline fuel effects. The details on the M6SM model are shown in Section
5.0 and the process for combining the sulfur and non sulfur fuel models is shown in
Section 6.0.
                                       26

-------
4.0    Predictive Model Algorithms for Total Hydrocarbon and
       Nitrogen Oxide Fuel Effects
4.1    Predictive Model Overview
       The MOVES model uses the Predictive Model to simulate the effect of gasoline
fuel parameters on hydrocarbon (HC) and nitrogen oxide (NOx) emissions. The HC and
NOx Predictive Models utilize two base fuels (30 ppm Sulfur and 90 ppm Sulfur).  The
30 ppm fuel is the base fuel for 2001-and-later model years, and 90 ppm sulfur is the base
fuel for 1975 through 2000 model years.  The remaining fuel parameters are identical
between the two base fuels (see Table 2-1).
       The Predictive Model effects from all fuel parameters are applied to 1960 through
1993 model years. The effects are not applied for model years 2004-and-later (Tier2-
and-later emission standards).  This means that in MOVES2010a, for 2004-model-year-
and-later gasoline vehicles, exhaust HC and NOx emissions are insensitive to all fuel
parameters (vapor pressures, aromatic contents, distillation fractions, etc.), EXCEPT for
the fuel sulfur level parameter.
       The decision to apply no Predictive  Model effects for 2004-and-later model years
in MOVES2010 is based on limited test evidence, and on engineering judgment that the
advanced combustion and emission control technologies used in Tier2-and-later gasoline
vehicles are likely to be insensitive to most fuel parameters (other than sulfur). However,
the EPAct test program and analysis are designed to provide data on this question.
        Since the release ofMOVES2010a, EPAct testing and statistical analysis on
2004 and later model years has shown that virtually all of the pollutants (criteria and air
toxic) are sensitive to a full array of fuel parameters - not just fuel sulfur. Consequently,
the assumption of 'insensitivity' to fuel parameters now used in MOVES2010a will be
substantially changed in future versions of the MOVES model.

4.2    Predictive Model  Step 1

       The overall goal of the  Predictive model algorithm is to develop a multiplicative
fuel correction factor for  a given fuel formulation which can be applied to the base
emission rates (BaseE) as calculated by MOVES.  This is shown mathematically in the
simple Equation 4-1.

Fuel Corrected Emissions   =     Fuel Correction * Sulfur Correction * BaseE   Eq 4-1
       The Fuel Correction shown in Eq 4-1 is the ratio between the Predictive model
output for the Target fuel and the Predictive model output for Base fuel. This is shown in
Equation 4-2

Fuel Correction      =    PredictiveModeljTarget} /PredictiveModeljBase} Eq4-2
                                      27

-------
       Where Target is the emissions using the Target fuel formulation and Base refers
to emissions with the Base fuel formulation.  The same algorithms are used for both the
Target and Base Predictive model calculations. The only difference between the two
calculations is the different fuel properties for the Target fuel and the Base fuel.
4.3    Predictive Model Step 2

       The Predictive model is a series of statistically derived fuel and emission models.
During the analysis process, several models were developed but only three models were
chosen for HC, and six models were chosen for NOx. For a given pollutant, each model
is equally weighted for the computation of the final result.  Table 4.1 lists the individual
models.
Table 4-1
MOVES Fuel Model Naming Conventions
Model Number
1
2
3
4
5
6
HC
107
108
112



NOx
302
303
304
305
306
307
       Each of the fuel models (three for HC and six for NOx) are function of 19 fuel
parameters and combinations (products) of fuel parameters.  All of the models have the same list
of parameters. However, some models have coefficients of zero for certain parameters. The
parameters are listed in Table 4-2 contains a complete list of all of the Predictive model
parameters and equations.  In the MOVES algorithm the "Intercept" and "HI" terms are set to
unity.
                                       28

-------
Table 4-2 Fuel Parameters for the Predictive
Models
Parameter Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Fuel Parameter(s)
Intercept
RVP
T50
T90
AROM
OLEF
OXYGEN
SULFUR
HI
T90T90
T50T50
T90*OXY
SUL*HI
OXY*OXY
T90*ARO
T50*HI
OLE*OLE
T90*OLE
ARO*ARO
       The first step in applying the model is to create a normalized fuel parameter
(N_Parameter) as shown in Equations 4-3a and 4-3b.  It is performed for each of the
parameters listed in Table 4-2, for each of the three HC models or each of the six NOx
models, and for both the Target and Base fuels.
N_Parameter =

Where:
(fuelParmeter -centeringValue) / stdDevValue
Eq 4-3a
fuelParameter - is the particular fuel parameter value of either the target or base fuel
(i.e., RVP(psi)). The values for the Target and Base fuels are taken from the MOVES
Fuel Formulation table.

The related Equation 4-3b is used for a compound term (i.e., T90*OLE)

N Parameter = ((T90-centeringValue)/stdDev Value)* ((Olefin-centeringValue)/stdDev Value)
Eq4-3b

Centering Value and the stdDevValue are fuel parameter constants.  These are shown in Table
4-3.
                                      29

-------
Table 4-3
MOVES Predictive Model CenteringValues and StdDevValues

Term
RVP
T50
T90
AROM
OLEF
OXYGEN
SULFUR

Mean
8.51
205.62
310.65
27.64
6.93
1.49
183.14
Standard
Deviation
0.781459
17.612534
20.869732
6.561886
5.143184
1 .249356
143.055894
4.4    Predictive Model Step 3

       In this step the normalized fuel parameter from Eq 4-3a and Eq 4-3b are
multiplied by the variable "coeffl" for each of the each of the parameters (i) listed in
Table 4-2, for each of the fuel models (i.e., HC has three and NOx has six), and for both
the Target and Base fuels.  This produces a matrix of products for both the Target and
Base fuels. The values for the variable coeffl are shown in the attached spreadsheet
Predi cti veModel C oeffi ci ents .xl s.
Product (i)
N Parameter * coeffl (i)
Eq4-4
4.5    Predictive Model Step 4

       For each of the fuel models, and the Target and Base fuel formulations the values
of Product(i) are summed across the 19 fuel parameters listed in Table 4-2.  The
Exponent of the sums is calculated.
ModelResult(j)      =      EXP  [SUM(Product(i,j))    i = 1, 19 ]

Where (i) is the fuel parameter and (j) is the fuel model.
                                               Eq4-5
4.6    Predictive Model Step 5
       In this step the ModelResult calculated in Step 4 for each fuel model (three HC
models and six NOx models) are weighted (averaged) together. This operation is done
for both the Target and Base fuels to produce the variable "WeightedResult".  For both
HC and NOx, each of the fuel models is equally weighted.
                                       30

-------
WeightedResult[target]=    sum[fuelModelWtFactor * ModelResult(j)]        Eq 4-6

WeightedResult[base]=     sum [fuelModelWtF actor * ModelResult(j)]        Eq 4-7


Where (j) is three for HC and (j) is six for NOx.
4.7    Predictive Model Step 6
       We next calculate the Fuel Correction variable shown in Equation 4-1 and 4-2.
This is the fuel adjustment factor WITHOUT the effects of sulfur. In MOVES it is
computed for each emission process (running and start), each target fuel formulation
which is being evaluated, each model year group / age group and each source type.
Fuel Correction=    WeightedResult[target] / WeightedResult[base]           Eq 4-8a
       By definition, for Tier2-and-later model years the Fuel Correction variable is set
to one for all non sulfur effects.  Since there are no sulfur effects in this algorithm
Equation 4-8b is used.
Fuel Correction      =      1.0                                             Eq 4-8b
4.8    Predictive Model Step 7

       In this step the Sulfur Correction factor is applied to Fuel Correction variable
calculated in Eq 4-8a or Eq 4-8b. The Sulfur Correction factor is computed by applying
the MOBILE6 Sulfur Model (M6SM). The details on the M6SM model are shown in
Section 5.0 and the process for combining the sulfur and non sulfur fuel models is shown
in Section 6.0.
                                        31

-------
5.0   Fuel Sulfur Model
5.0    Fuel Sulfur Model Background

       In the MOVES application of the Complex and Predictive Models, the emission
effects of fuel sulfur were removed and the multiplicative correction factors for sulfur
were set to one. Instead of using the sulfur effects from these two models, the fuel sulfur
effects from the MOBILE6.2 Fuel Sulfur Model (M6Sulf) were applied to both start and
running emissions for HC, CO and NOx emissions for all model years (the exception is
the final sulfur adjustment factor for pre-1975 model years which is unity, ie. no
adjustment). In MOVES2010a, the Tier2 (2004 and later model years) utilize the M6Sulf
model, but this assumption will likely be reviewed once substantial emission / fuel data is
available from the EPAct testing.

       Two issues become apparent when applying the M6Sulf effects in the MOVES
model. These are: (1) The MOBILE6.2 model contains separate effects for High and
Normal emitters, but MOVES does not classify emissions or vehicles according to these
terms.  (2) The MOVES model needs to model Tier2 vehicles using low sulfur fuel (i.e.,
< 30 ppm sulfur). However, the MOBILE6.2 sulfur factors are based entirely on pre-
Tier2 vehicles operating on sulfur levels of 30 ppm or higher.

       In MOBILE6.2, different fuel sulfur effects (coefficients) and equations (i.e., log-
log and log-linear) were used for High and Normal emitters (a  "High " emitter was
defined as a vehicle which emitted twice it applicable HC or NOx standard or three times
its CO standard). MOBILE6.2 computes emission rates for High and Normal emitters
separately and weights them according to a complex algorithm that is built into the
model. However, in general the weighting between High and Normal emitters was
frequently about equal for the typical vehicle on the road (new vehicles had lower rates of
high emitters, though). For the MOVES sulfur effect, a straight arithmetic average
between the High and Normal effects was calculated and used for consistency with the
MOBILE6 approach.

       For some of the late model year vehicles (2001 - 2003), this approach may model
these vehicles to be less sensitive  to sulfur than more recent data from 2004+ model year
- Tier2 vehicles would suggest. This is because (1) the distribution of "high" emitters in
the fleet is smaller (< 50%) than it has been  historically - although, their emission
contribution is probably still very significant, and (2) the "high" emitter corrections are
typically log-linear relationships and less sensitive to sulfur than the log-log "normal"
emitter corrections.

       To model very low sulfur fuels in MOVES, we extrapolated the MOBILE6
emission / fuel sulfur relationship to fuels below 30 ppm sulfur. Since some of these
relationships are log-log in structure, such an extrapolation could lead to meaningless
results. To prevent highly non-linear relationships which are inherent in log-log
mathematics from dominating the MOVES emission results, a fuel adjustment 'floor' of
0.85 was placed in the code.  This 'floor' prevents MOVES from reducing the emission

                                       32

-------
rate by more than 15 percent from its 'base' emission rate as a result of changing base
fuel formulation (30 ppm sulfur) to some other lower level (i.e., 0.01 ppm sulfur).  This
floor should prevent the most asymptotic portion of the log-log relationship from being
utilized.

       As a final thought on this topic, EPA realizes the inherent problems associated
with using these out of date fuel relationships on Tier2 and later vehicles operating on
low sulfur fuels. As such, EPA plans to revise them as soon as sufficient new data are
collected and analyzed as part of the EPAct test program.

5.1    MOBILE6 Sulfur Algorithm in MOVES

       In MOVES, the M6Sulf model data are stored as a set of equation coefficients in a
database table. It has the following table structure:

       pollutantID
       processID
       fuelMYGroupID
       M6emitterID
       SourceTypelD
       SulfurFunctionID
       SulfurCoefficient

       The M6emitterID is designated as either "Normal" or "High" and the
SulfurFunctionID is designed as either "Log-Log" or "Log-Linear". The "Log-Log"
designation refers to a model with a natural log - natural log form and the "log-linear"
refers the natural log-linear model.

       The M6Sulf algorithm as programmed into the MOVES model contains four
separate fuel sulfur models. This document and section details the algorithms in the
MOVES model. It also discusses the minor changes and assumptions that had to be made
to  fit the MOBILE6 algorithms into the MOVES structure. The reader who is interested
in  the primary development of these fuel models needs to refer to the primary
documentation in the EPA report EPA420-R-01-039  "Fuel Sulfur Effects on Exhaust
Emissions - Recommendations for MOBILE6" by Venkatesh Rao. The internet link is:

http://www.epa.gov/otaq/models/mobile6/r01039.pdf
       The individual models which are used in the MOVES model are:

       Short Term Sulfur model
       Long Term Sulfur model
       Sulfur Irreversibility model
       Sulfur GPA model
                                       33

-------
       In addition, space is reserved in the MOVES model to incorporate a new fuel
sulfur model for Tier2 (2004-and-later model year vehicles) if and when it become
available.
5.2    Short Term Sulfur Effects

       The Short Term Sulfur Effects is an adjustment to the base emissions as a
function of the sulfur volume of the gasoline in parts per million.  The calculations begin
by using Equations 5-la and 5-lb. These are used in cases where the log-log relationship
is required (the variable SulfurFunctionID is set to 'log-log'.  This is for TierO and LEV+
vehicles.  The variable "sulfShortTarget" is the correction factor for the sulfur level of the
fuel which is being modeled. SulfShortSO is the correction factor for the basis sulfur
(table SulfurBase. sulfurBasis variable) level.  The sulfur basis is always 30 ppm - even
for model year groups which use 90 ppm Sulfur as the baseline. A 90 ppm sulfur base
fuel will not have a SulfShortAdj  that is equal to 0.0.
sulfShortTarget      =     EXP(sulfurCoeff*LN(sulfur))                    Eq 5-la
sulfShortSO          =     EXP(sulfurCoeff*LN(sulfurBasis))               Eq 5-lb
       The Short Term Sulfur effects for Tierl vehicles (i.e. model years 1994-2000} )
use a log-linear algorithm as in equations 5-2a and 5-2b.  Again the base fuel sulfur is 30
ppm. High emitters may use a different equational form (i.e, log-log) in some cases. The
"Sulfur Model Coefficients.xls" workbook in the variable SulfurFunctionID shows which
form of the equation is used.
sulfShortTarget      =     EXP(sulfurCoeff*(sulfur))                       Eq 5-2a
sulfShortSO          =     EXP(sulfurCoeff*(sulfurBasis))                  Eq 5-2b
The Short term sulfur effect (SulfAdj) for all groups is computed using Eq 5-3.

SulfShortAdj  =     (SulfShortTarget-SulfShortSO)/SulfShortSO           Eq 5-3



5.3    Calculate Long Term Sulfur Effects

       The long term sulfur effects account for sulfur exposure over several thousand
miles of driving.  They values used in MOVES are stored in the sulfurLongCoeff
variable in the MOVES table M6SulfurCoeff.  The values for sulfurLongCoeff are a
function of pollutant. They are:

       HC                        2.50
       CO                        2.36
       NOx                       1.47

                                       34

-------
       These effects are multiplied by the short term sulfur effects calculated above in
Eq 5-3 to produce the variable sulfAdj2.  Sulfur levels of 30 ppm or less are assumed to
have no Long Term effects.
sulfAdj2      =     SulfShortAdj * sulfurLongCoeff                        Eq 5-4
5.4    Calculation of the Sulfur Irreversibility Effects

       In this step, the average emission effects of sulfur which cannot be reversed if
exposure is recorded are computed. They apply to only Tier2 (2004+ model years
vehicles only), and apply only to Target fuel sulfur levels which are greater than 30 ppm
sulfur. Fuel sulfur levels less than or equal to 30 ppm do not cause such effects to occur.
The same effects are applied to all three pollutants (HC, CO and NOx) and processes
(start and running).  The SulfurCap is a function of model year group.

Model Year Group                 SulfurCap

2004 - 2005                       303 ppm sulfur
2006 - 2007                        87 ppm sulfur
2008 +                             80 ppm sulfur
       If the fuel sulfur level is greater than 30 ppm but less than the SulfurCap,
Equation 5-5a is used to compute the SulfTRR effects, and the effect is applied as a
function of model year group.
If sulfur < sulfurCap

SulfTRR             =      EXP (sulfurCoeff * LN(sulfurCap))               Eq 5-5a
       Eq5-5b computes the sulfur irreversibility effects if the selected sulfur level is
greater than the maximum sulfur level. Sulfur levels above the SulfurCap are not
expected in normal use of the MOVES model.

If sulfur > sulfurCap

SulfTRR             =      EXP (sulfurCoeff * LN(sulfur))                   Eq 5-5b
                                       35

-------
5.5    Putting the Short, Long and Irreversibility Sulfur Effects Together

       Equation 5-7 puts all of the sulfur effects together into the final fuel sulfur effects.
It is labeled as sulfAdjS. This variable contains the short term, long term and irreversible
effects together.  Equation Eq5-6 calculates the intermediate sulfMax variable.
sulfMax       =     (SulfIRR-sulfShort30)/sulfShortS0                    Eq 5-6

Where

SulfTRR is from Eq 5-5a or Eq 5-5b.
sulfShortSO is from Eq5-lb or Eq5-2b


sulfAdjS       = 1.0+  [IRFactor* sulfMax  +  (1.0-IRFactor) * sulfAdj2 ]    Eq 5-7

Where

sulfAdj2 is the result of Eq5-4

IRFactor is equal to 0.425.
       To prevent the log-log based sulfur corrections from rapidly approaching zero as
the sulfur levels approach zero, the following limit is placed.

sulfAdjS      =     0.85          Where sulfAdjS <= 0.85                  Eq 5-8
5.6    Sulfur GPA Effects

       During the years 2004-2006, gasoline sulfur levels in the Sulfur "Geographical
Phase-In Area" (Sulfur GPA) were allowed to be higher than elsewhere in the nation.
MOVES accounts for this with the Sulfur GPA Effects. The algorithm applies a
maximum sulfur exposure of 330 ppm sulfur to areas that are GPA areas. The GPA areas
are mostly Rocky Mountain areas and are identified in the county table ofthe
MOVES2010 database.

       In this step (Eq 5-9) the sulfurGPA effects are calculated for both Normal and
High emitter groups.  Sulfur GPA is only applied to model years 2004, 2005 and 2006,
and for fuel sulfur levels which are greater than 30 ppm sulfur.  It is also limited to GPA
areas, and is a sulfur phase-out strategy. In all other cases GPAsulfAdj is equal to the
regular sulfur adjustment, and no additional GPA effect is applied in the MOVES model.
SulfurGPA uses the same basic algorithm as the regular sulfur adjustment, except the
variable SulfurBase. The sulfurGPAMax (330 ppm sulfur) is substituted in the equation
in place ofthe actual  sulfur level ofthe fuel to be evaluated. The value of 330 ppm is a
typical worse case GPA scenario sulfur level.


                                        36

-------
sulfGPAl     =     EXP (sulfurCoeff * LN(sulfurGPAMax))               Eq 5-9

SulfGPA     =     (sulfGPAl - sulfShortSO) / sulfShortSO                 Eq 5-10a


For NOx High emitters the numerator (sulfGPAl and sulfShortSO) is multiplied by 0.60.


sulfGPA     =     SulfGPA * sulfurLongCoeff                           Eq 5-1 Ob




GPASulfadj   =  1.0+  [IRFactor* sulfGPA  + (1.0-IRFactor) *  sulfAdj2 ]   Eq5-lla

Where

sulfAdj2 is the product of Eq5-4

IRFactor is equal to 0.425.
       In non GPA years, or in areas where sulfur < 30 ppm, Eq 5-1 Ib is used.  It is also
used in cases if the sulfur level is greater than sulfurGPAMax (i.e., 330 ppm. As a
general rule, the GPASulfadj can NEVER be smaller than sulfAdj3, and the two are equal
most of the time.
GPASulfadj   =      sulfAdjS                                            Eq5-llb
       The GPASulfadj is applied in the model by weighting it by the fraction of GPA is a
particular county. In the default cases, this fraction is always zero or one. However, GPA
fraction is a user input, so alternative values between zero and one may be entered as
inputs.

Final Sulfur Adj =  SulfAdjS * (1-GPAFraction) + GPASulfadj * GPAFraction  Eq 5-12
5.7   Normal and High Emitter Correction Factor Weighting

      As discussed in general in Section 5.0, the M6Sulf algorithm produces a sulfur
correction for both "Normal" and "High" emitters. Since MOVES does not define vehicles
as "Normal" and "High" emitters in terms of emissions rates, the sets of model coefficients
were regarded as independent models and equaled weighted for consistency with the
MOBILE6 model.  The variable wtHigh is set equal to 0.50.


                                      37

-------
sulfAdj3(Target) = (l-wtHigh)*sulfAdj3(normal) + wtHigh*sulfAdj3(high)      Eq 5-13a
sulfAdj 3 (Base) = (l-wtHigh)*sulfAdj3(normal) + wtHigh*sulfAdj3(high)        Eq 5-13b


       Likewise, a composite of normal and high emitter GPAsulf adjustments are
calculated using the same wtHigh factors.

GPAsulfAdj3(Target) = (1-wtHigh)* GPASulfadj (normal) + wtHigh* GPASulfadj (high)       Eq 5-14a
GPAsulfAdj3(Base) = (1-wtHigh)* GPASulfadj (normal) + wtHigh* GPASulfadj (high)        Eq 5-14b
5.9    Computing the Sulfur Adjustment for the Base Fuels.

       Sections 5-2 through 5-7 must be repeated and applied for the two base fuels of 90
ppm sulfur and 30 ppm sulfur corresponding to the two model year groups (1960-2000 and
2001-2050), respectively. This is because the final sulfur fuel adjustment (see Eq 5-15)
factor is the ratio of the Target fuel adjustment and its Base fuel adjustment. The M6Sulf
model calculations for the Base fuel are exactly the same as those presented in the sections
above and are not repeated here. All other properties of these fuels are held constant for
the sulfur correction.  The Final fuel sulfur correction for the 30 ppm case is one because
the base fuel of 30 ppm sulfur is equal to the 30 ppm sulfur basis. The 30 ppm sulfur level
is called the basis (i.e., sulfShortSO variable) because the entire MOBILE6.2 Fuel Sulfur
algorithm was  developed based on this level. The calculation result is not one for the 90
ppm base sulfur.
FinalSulfAdj  =     sulfAdj3(Target)/sulfAdj3(Base)                       Eq5-15

Where sulfAdj3 is the sulfur adjustment from Sections 5.2 through 5.8 and the
sulfAdj3(base) is the sulfur adjustment for either the 30 ppm or 90 ppm sulfur base (or
reference) cases.
The calculation is similar for the GPA case (Eq 5-16).

FinalGPASulfAdj    =      GPASulfadj (Target)/GPASulfadj(Base)          Eq5-16
                                       38

-------
6.0    Combining the Predictive and Complex Model with the Sulfur
      Model
      This section describes how the fuel adjustment factors generated by the Predictive
and Complex model without sulfur effects are combined with the fuel adjustment factors
generated by the M6Sulf model in Section 5.0. This combination is done only for
pollutants HC, CO and NOx.  The air toxic pollutants do not use the M6Sulf model (i.e.,
the air toxics CM model contains sulfur effects).

      Since all three of these models produce multiplicative correction factors the final
combination of the two factors is multiplicative.  The general equation is shown in Eq 6-
1.

MOVES Fuel Adjustment = Non Sulfur Fuel Adjustment * M6Sulf Adjustment   Eq 6-1
6.1    Overall Fuel Adjustment for Model Year 1960 through 1974 Vehicles.

      In the case of all 1960 through 1974 vehicles, the Overall Fuel Adjustment is set
to one. This was done because little or no data are available for these model years and
most of the vehicles do not contain a catalytic converter or modern electronic fuel /
engine management components which would be sensitive to changes in fuel
composition.

Fuel Adjustment     =      1.0                                           Eq 6-2
6.2   Overall Fuel Adjustment for Model Year 1975 through 2003 Vehicles.


      The Overall Fuel Adjustment for HC, CO and NOx pollutants for 1975 through
2003 model year vehicles is the product of the final fuel adjustments calculated with the
Complex or Predictive models (i.e., COFuelEffectNoSulfur or Fuel Correction (Eq 4-8a),
and the FinalSulfAdj calculated in the MSSulf model.

Fuel Adjustment     =      finalSulfAdj * RatioNoSulfur                    Eq 6-3

Where
      Model Year > = 1975 andModel Year < = 2003 and Pollutant in (HC, NOx) or
      Model Year >= 1975 and Pollutant in (CO)
                                     39

-------
6.3    Overall Fuel Adjustment for Model Year 2004 and Later Vehicles.

      The Overall Fuel Adjustment for HC and NOx pollutants for model year 2004 and
later model year vehicles is the Final Sulfur Adjustment computed by the M6Sulf model.  It
contains only fuel sulfur effects.

Fuel Adjustment     =      final SulfAdj                                  Eq 6-4

Where
      Model Year >=2004 and Pollutant in (HC, NOx)
6.4   Overall fuelAdjustmentGPA for Model Years 2004 Through 2006


      In this step the fuelAdjustmentGPA is set to the final value for use in the MOVES.


fuelAdjustmentGPA  =     FinalGPASulfAdj                              Eq 6-5

Where
      Modelyear in  (2004, 2005, 2006) andfmalGPASulfAdj < 1.0;



fuelAdjustmentGPA  =     fuel Adjustment                               Eq 6-6

      for all other cases
                                     40

-------
7.0   Biodiesel Fuel Effects

       The Draft MOVES2009 version released in April 2009 contains diesel fuel
parameters and emission effects only for diesel fuel sulfur level (ppm sulfur in the fuel).
This parameter is used in a fuel-sulfur mass balance equation to calculate sulfate PM
emission levels, and was expanded in MOVES2010 to include gaseous SO2 emissions.
The fuel sulfur mass balance methodology was taken from EPA PARTS / MOBILE6.2
model and is fully documented in the MOBILE6.2  documentation. See the link:

http!//www. cpa.gov/oms/modcls/mobilc6/m6tcch. htm

       New sulfate (SO4) and sulfur dioxide (SO2) emission factors were also developed
for MOVES2010a for gasoline vehicles. The data were obtained from a set of about 30
vehicles in the EPA Kansas City test program in which both fuel sulfur measurements
and SO4 particulate measurements were made.  See the document "Development of
Gasoline and Diesel Vehicle Sulfate and Sulfur Dioxide Emissions for the MOVES
Model". The sulfate and sulfur dioxide emissions for diesels and pre-1980 gasoline
model years were unchanged from MOBILE6.2

       MOVES2010 contains an additional parameter called BioDieselEsterVolume.  It
represents the percentage of biodiesel ester by volume in a Target fuel.  Currently all
entries in the default fuel supply for MOVES2010 are NULL and the model performs no
biodiesel calculations. However, users could enter local information about biodiesel
fuels.

       Mathematically, the BioDieselEsterVolume variable is  used with the BioDiesel
Fuel Adjustment Factors (presented in Table 7-1).  Together these two parameters
produce an overall diesel Fuel Adjustment factor for biodiesel  fuels.  These fuel
adjustment factors (shown as percentages) give the relative increase or decrease in
emissions as the result of adding biodiesel to standard on-road  diesel fuel. The equation
for calculating the BioDiesel adjustment factor is shown in Eq  7-1. Analogous to the
gasoline fuel adjustment, the BioDiesel factor gets multiplied by the base emission rate of
standard diesel vehicles.
Fuel Adjustment factor = 1.0 +  (BiodieselEsterVolume/100.0) * BioDieselFactor  Eq7-l
       The Biodiesel Fuel Adjustment factors are shown in Table 7-1.
                                       41

-------
Table 7-1
Biodiesel Fuel Adjustment Factors as a Function of Pollutant
Pollutant Name
HC
CO
NOx
PM2.5andPM10.0
Benzene
1,3 Butadiene
Acetaldehyde
Formaldehyde
Naphthalene
Acrolein
BioDiesel Factor
-14.1
-13.8
2.2
-15.6
-14.1
-14.1
-14.1
-14.1
-15.6
-14.1
       The individual pollutant impacts of biodiesel (B20) listed in Table 7-1 are relative
to conventional diesel fuel. For example, the -14.1 factor for HC would reduce the
hydrocarbon emissions from a standard diesel fuel by 14.1 percent.  The biodiesel factors
were based on very limited EPA testing of such fuels.  See the link below.

http://www.arb.ca.gov/fuels/diesel/altdiesel/100519BiodieselWorkshopPresB&W.pdf
       No data were available on air toxic pollutants tested on biodiesel fuels.
Consequently, biodiesel will use the same air toxic / VOC ratios as are used in the
computation of emissions from vehicle operating on ordinary diesel fuels (see Section
8.0).
                                         42

-------
8.0    Toxic Emissions from Diesel Vehicles
       The toxics ratios in MOVES2010 for diesel vehicles were developed well over 10
years ago, and were used in MOBILE6.2.14 These ratios are provided in Table 8-1.
More extensive data are now available, including recent test data on diesel engines
meeting 2007 heavy-duty emission standards.15'  6 EPA intends to update the equations
and the methodology in MOVES using these data, and release the updated version at a
later date. Whereas MOBILE6.2 had separate ratios for light-duty diesel vehicles and
trucks based  on very limited data, MOVES2010a applies the same ratios to all diesel
vehicles, heavy and light-duty.
Table 8-1
Toxirs Ratios for Diesel Vehicles
Pollutant
Benzene
Formaldehyde
Acetaldehyde
Acrolein
Naphthalene
Ratio Type
voc
voc
voc
voc
PM10
Ratio
0.0108
0.0807
0.0298
0.0036
0.001289
       The air toxic pollutants for diesels are computed using equations 8-1 and 8-2.
The equations are completely analogous to the simple equations used for gasoline
vehicles (see Eq 2-1).
AT Emission (g/hr)   =     AT Ratio * VOC Emission (g/hr)

AT Emission (g/hr)   =     AT Ratio * PM10 Emission (g/hr)
Eq8-l

Eq8-2
14 Cook, R., and E. L. Glover. 2002. Technical Description of the Toxics Module for MOBILE6.2 and
Guidance on Its Use for Emission Inventory Preparation. Assessment and Standards Division, Office of
Transportation and Air Quality, Ann Arbor, MI. Report No. EPA420-R-02-011.
http://www.epa. gov/otaq/m6.htm

15 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.

16 Khalek, L, 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.
                                         43

-------
Hydrocarbon Speciation Factors
       Table 8-1 shows a simple relationship between the various hydrocarbon species in
MOVES, and their relationship to the base, total hydrocarbons (THC). As the table
shows the individual hydrocarbon species differ based on the presence or absence of the
hydrocarbon compounds of methane, ethane and aldehydes. The term FID HC refers to
the total hydrocarbons detected by a Flame lonization Detector (FID) instrument.

       All factors used in MOVES were taken from MOBILE6.2 materials and were
originally produced in 1991.  No new speciation factors were developed for
MOVES2010a.
Table 8-1
Hydrocarbon Speciation Types
PollutanflD
1
79
87
86
80
PollutantName
Total
Hydrocarbons
Non Methane
Hydrocarbons
Volatile Organic
Compounds
Total Organic
Gases
Non Methane
Organic Gases
FIDHC
Yes
Yes
Yes
Yes
Yes
Methane
Yes
No
No
Yes
No
Ethane
Yes
Yes
No
Yes
Yes
Aldehydes
No
No
Yes
Yes
Yes
8.1    Non Methane Hydrocarbon Factors

      The MOVES model calculates non-methane hydrocarbons (NMHC).  The
MOVES model computes this hydrocarbon species by first independently calculating
THC emissions and methane emissions.

 (see http://nsdi.epa.gov/otaq/models/ngm/420p05003.pdf).

      As a result of this dependence, NMHC is 'chained' to both THC and methane.
The NMHC calculator subtracts the resulting methane emissions from the THC emissions
(mechanically in MOVES it multiples methane by negative one and adds the result). If a
negative emission value for NMHC is calculated, it is set to zero. The formula for
calculating NMHC is shown in Eq 8-1. This formula is used only for gasoline fuels
      NMHC
THC
Methane
Eq8-l
                                      44

-------
8.2    VOC Speciation Calculation

       The MOVES model calculates Volatile Organic Compounds (VOC).  VOC is
calculated from NMHC using Eq 8-2a through 8-2d and the variables
'speciationConstant' and 'oxySpeciation' coefficients.  The variable
'volToWtPercentOxy' is a standard chemical conversion to converts percent oxygen in
the fuel from a volume basis to a weight basis for variety of oxygenated fuels.
Consequently, all of the standard MOVES fuel inputs for oxygenate (i.e., ETOHVolume,
MTBEVolume, etc) are in terms of a percentage of oxygen by fuel volume.

       For fuels containing ethanol

       VOC   =     NMHC *
       (speciationConstant + oxySpeciation* volToWtPercentOxy*ETOHVolume)          Eq 8-2a


       For fuels containing MTBE

       VOC   =     NMHC *
       (speciationConstant + oxySpeciation* volToWtPercentOxy*MTBEVolume)          Eq 8-2b


       For fuels containing ETBE

       VOC   =     NMHC *
       (speciationConstant + oxySpeciation* volToWtPercentOxy*ETBEVolume)           Eq 8-2c
       For fuels containing TAME

       VOC   =     NMHC *
       (speciationConstant + oxySpeciation* volToWtPercentOxy*TAMEVolume)          Eq 8-2d


       For pure gasoline fuels the second term drops off and only the speciationConstant
is applicable. The speciationConstant is a function of fuel model year group. The
variable volToWtPercentOxy is a function of oxygenate type. The values for each
oxygenate type are:
       Ethanol       0.3488
       MTBE        0.1786
       ETBE         0.1533
       TAME        0.1636


       The general equation for volToWtPercentOxy in MOVES is equation 8-3e.

       volToWtPercentOxy =

       (ETOHVolume * 0.3488 + MTBEVolume * 0.1786 + ETBEVolume * 0.1533 +TAMEVolume * 0.1636)
       /  (ETOHVolume+MTBEVolume+ETBEVolume+TAMEVolume)                      Eq 8-3e
                                        45

-------
8.3    NMOG Speciation Calculation

       The MOVES model calculates Non Methane Organic Gases (NMOG). NMOG
is calculated from NMHC using Eq 8-3a through 8-3d and the variables
'speciationConstant' and 'oxySpeciation' coefficients.  The values for the variable
volToWtPercentOxy are provided in Section 8.2.
       For fuels containing ethanol

       NMOG =      NMHC*
       (speciationConstant + oxySpeciation* volToWtPercentOxy*ETOHVolume)           Eq 8-3a
       For fuels containing MTBE

       NMOG =      NMHC*
       (speciationConstant + oxySpeciation* volToWtPercentOxy*MTBEVolume)           Eq 8-3b
       For fuels containing ETBE

       NMOG =      NMHC*
       (speciationConstant + oxySpeciation* volToWtPercentOxy*ETBEVolume)           Eq 8-3c
       For fuels containing TAME

       NMOG =      NMHC*
       (speciationConstant + oxySpeciation* volToWtPercentOxy*TAMEVolume)           Eq 8-3d
       For pure gasoline fuels the second term drops off and only the speciationConstant
is applicable. The speciationConstant is a function of fuel model year group.
8.4    TOG Speciation Calculation

       The MOVES model calculates Total Organic Gases (TOG).  These are calculated
by calculating both the NMOG emissions and the methane emissions in their respective
calculators. TOG is depend (i.e., chained) to both NMOG and methane. The TOG
calculator in Equation 8-4 simply adds the resulting methane emissions to the NMOG
emissions.

       TOG  =      NMOG + Methane                                      Eq 8-4
                                         46

-------
9.0 Fuel Formulation and Fuel Supply in MOVES2010

9.1    Introduction

     The MOVES2010 model contains about 10,000 default fuel formulations and
default market share values. The large number of fuel formulations reflects a
combination often different fuel parameters, and the need to supply a default fuel
formulation for each combination of U.S. county, month and fuel year.  Each fuel
formulation describes a unique set of fuel properties.

     In MOVES, the fuel models (i.e., complex, predictive, sulfur, etc) are applied to
each fuel formulation selected by the user.  The user selects at least one fuel formulation
for each county, calendar year, month and fuel type (i.e., gas or diesel) combination they
wish to model. Some combinations have two or more fuel formulations which represent
a blend of fuels during that time and in that area. The individual market share values
represent the distribution of fuel formulations within a county, calendar year, month and
fuel type combination, and must sum to unity within the combination.

9.2    Data Sources

       The fuel supply and fuel formulation data in the MOVES2010 model was
assembled from several sources as part of the EPA National Emission Inventory (NEI)
development process.  The process of assembling these data for the MOVES model
consisted of a detailed analysis which is described in the references listed in Section 9.4.
It consisted of a fairly complicated data review and analysis process that utilized several
sources, and will not be repeated here.  The following sources of fuel supply and fuel
property data for MOVES are:

       o  EPA fuel survey data from RFG areas
       o  Vehicle manufacturer fuel surveys
       o  Other commercial fuel surveys
       o  Proprietary Refinery modeling results
       o  State supplied fuel data

       The EPA fuel survey data from RFG areas is an important data source for in-use
fuel property information, and market share data of specific fuels.  It is an annual survey
done in RFG areas. Unfortunately, it is limited to RFG areas, and often is limited to just
a single county in the area. Another shortcoming is that EPA has only analyzed and
incorporated surveys from the 2005 and earlier calendar years. Vehicle manufacturer and
other commercial fuel surveys were also incorporated into the database, or used to
confirm existing data whenever possible. The proprietary refinery modeling results were
a very important source of information and allowed EPA to model areas which did not
have fuel surveys.  This data-source frequently had refinery, pipeline and  other
distribution system volume data on particular fuel formulations and individual fuel
parameters such as the volume of ethanol or the concentration of sulfur in the fuel  flow.
The pipeline fuel flow data could be distributed according to vehicle miles traveled

                                       47

-------
(VMT) through the fuel distribution system so as to produce more complete fuel
information on counties that did not have a fuel survey.  State data on fuel properties and
fuel supplies were also obtained and used where possible.  In most cases, individual states
do not have complete or detailed data on their fuel supplies.

9.3    Analysis Process Overview

       An analysis process was used to assemble the MOVES fuel supply and fuel
formulation tables for use in the model. For calendar years 1990, 1999, 2002 and 2005
historical fuel data were collected and processed, and used in the National Emission
Inventory (NEI) process. These data were generally 'clean' and easily processed into the
required format for MOVES. Analogous tables for calendar years 2000, 2001, 2003 and
2004 were generated by interpolating the primary dataset.

       Calendar years 2005 and 2012 are the key years for fuel property definition.
After 2012, the  fuel properties and supplies become constant. The data for calendar years
2005 are from the 2005 NEI and the data (projections) for 2012 are from refinery
modeling done for the RFS rulemaking.  The 2012 calendar year data are projections
which  start with the 2005 fuels and project the 2012 fuels based on ethanol volume
increases, benzene controls, MTBE phase-out and RVP waiver for ethanol blends. The
impact of future fuel  regulations is a major factor in these projections.  Fuel data for all
intervening years between 2005 and 2012 are interpolations between these two years.
Local fuel regulations and local fuel data, where available, were also incorporated  in the
analysis for each of the counties and calendar years.

       The fuel data is often only by season with winter - summer blends and one or two
shoulder seasons.  As a result, the data are further processed to allow it to be on a
monthly basis.  Frequently, in MOVES2010a, this processing consists of merely copying
it into the monthly database entries. Within a given season, the fuel formulations often
do not vary by month.
9.4    References

9.4.1    1999 NEI report that includes description of how the fuels were developed "DOCUMENTATION FOR THE
       ONROAD NATIONAL EMISSIONS INVENTORY (NEI) FOR BASE YEARS 1970-2002"
       ftp://ftp.epa.gov/EmisInventorv/finalnei99ver3/haps/documentation/onroad/nei onroad ian04.pdf

9.4.2    Updated 2002 report, "DOCUMENTATION FOR THE FINAL 2002 MOBILE NATIONAL EMISSIONS
       INVENTORY, VERSIONS"
       ftp://ftp.epa.gov/EmisInventory/2002fmalnei/documentation/mobile/2002_mobile_nei_version_3_report_09280.pdf

9.4.3    Updated 2005 report, "DOCUMENTATION FOR THE 2005 MOBILE NATIONAL EMISSIONS INVENTORY,
       VERSION 2"
       ftp://ftp.epa.gov/Emislnventory/2005_nei/mobile/2005_mobile_nei_ve rsion_2_report.pdf

9.4.4    EPA RFG fuel survey data:
       http://www.epa.gov/otaq/regs/fuels/rfg/properf/rfgperf.htm

9.4.5    RFS1 Regulatory Impact Assessment document "Draft Regulatory Impact Analysis: Changes to Renewable
       Fuel Standard Program"
       http://www.epa.gov/otaq/renewablefuels/420d09001.pdf
                                          48

-------
10.0  Results

       This section contains some limited emission results based on the algorithms in the
Predictive, Complex and MOBILE6 Sulfur Models.  These results were obtained by
running the MOVES model and charting the emission results.  Results are shown for the
criteria pollutants and the major air toxic compounds. The results are limited in that they
are shown only for areas of the MOVES model that have received substantial changes
from previous versions.  They are also limited because they model the base fuel and vary
only one fuel parameter at a time over its range. For example, in the sulfur effects chart,
fuel sulfur is varied from 4 ppm to 500 ppm and the remaining fuel parameters are kept
constant at the Base Fuel level. The reader who is interested in a complete set of emission
results for all combinations of fuel parameters, particularly the air toxic pollutants, should
consult the primary reference documents for the Complex, Predictive and MOBILE6.2
models.
10.1   Sulfur Effects in MOVES2010a
       The emission effect of fuel sulfur is shown in Figures 10-1 through 10-3 for the
2001+, 1996 and 1988 model years, respectively. The effects are 'net fuel effects'  from
the MOVES model. They were produced by seven separate MOVES runs using a
constant fuel formulation and varying the fuel sulfur level from 4 ppm  sulfur to 500 ppm
sulfur. In each chart, separate curves are shown for THC, CO and NOx.   The oldest
model year represents the fuel effects on TierO vehicles, the middle model year group
represents the Tierl and LEV standards, and the 2001+ model year represents the newer
vehicles.  In all charts the fuel effects are normalized to 90 ppm sulfur for the  1988 and
1996 model year curve and to 30 ppm sulfur for the 2001+ model year curve. In this
context, 'normalization' means the correction factor is set to one.  The  other fuel
parameters  were set at Base Fuel levels (6.9 psi RVP, 0% Ethanol volume, 26.1%
aromatic  content, 5.6% olefm content, 1.0% benzene content, 218F T50 and 329 F  T90.

       Examination of the three figures shows the strongest correlation is for NOx
emissions and for the 2001+ model year group.  This result is not unexpected because it
is this group for which a log-log relationship between emissions and fuel sulfur was
found. This relationship is strongly non-linear and produces a relative  correction of about
0.15 for an  ultra low 2 ppm sulfur fuel and a relative correction of about 3.0 for a 500
ppm high sulfur fuel (actual in-usefuel sulfur levels are not expected to vary that widely).
For the HC and CO pollutants, the sensitivity to fuel sulfur is less for the 2001+ MY
model year group than it is for NOx. The other model year groups mostly utilize log-
linear relationships between emissions and fuel sulfur. This leads to fuel sulfur emission
effects which are less sensitive to sulfur than the Tier2 group.

       In the MOVES2010a model, the 2001+ model year group is the latest model year
group, but is based only on data from pre-tier2 (i.e., 2003 and earlier model years)
vehicles.  Subsequent versions of MOVES (to be released at a later date) will utilize these
curves for the 2001 through 2003 model years, but will apply new curves for 2004  and
later model years based on a new set of vehicle fuel sulfur testing and EPAct testing.
Preliminary results from the analysis of the new data suggests a much stronger

                                       49

-------
relationship at low sulfur levels between fuel sulfur and HC and NOx emissions than is
usedinMOVES2010a.

No air toxic charts are shown because the air toxic / VOC ratio is not a function of fuel
sulfur level in the Complex model.
Figure 10-1   Relative Fuel Sulfur Effect on 2008 Model Year Gasoline Vehicles in
             MOVES
                           Sulfur Fuel Effect
                       2001+ MY - gasoline vehicles
                                                                       •VOC

                                                                       •CO

                                                                       •NOx
                  100
200
300
400
500
600
                       Fuel Sulfur Content (ppm Sulfur)
                                     50

-------
Figure 10-2   Relative Fuel Sulfur Effect on 1996 Model Year Gasoline Vehicles in
             MOVES
         Relative Fuel Sulfur Effect on 1996 Model Year Gasoline Vehicles
                               (90 ppm Sulfur = 1.0)
                  100
200       300       400

    Sulfur Level (ppm)
500
600
                                       51

-------
Figure 10-3   Relative Fuel Sulfur Effect on 1988 Model Year Gasoline Vehicles in
             MOVES
         Relative Fuel Sulfur Effect on 1988 Model Year Gasoline Vehicles
                               (90 ppm Sulfur = 1.0)
                 100
200       300        400

    Sulfur Level (ppm)
500
600
                                      52

-------
10.2   Fuel Ethanol Adjustment Factors in MOVES2010

       The emission effects of fuel ethanol are shown in Figure 10-4 for model year
2003. Charts for other model years are not shown, because the results are similar for the
pre-1994 model years (they differ only slightly for CO) and are identical for the 1995-
through-2002 model years. There are currently no ethanol fuel effects for 2004-and-later
model years in MOVES2010a (only sulfur fuel effects). Other fuel oxygenate types such
as MTBE or ETBE produce the same relative results when the oxygenate volumes are
identical.  The other fuel parameters were set at Base Fuel levels (6.9 psi RVP, 30 ppm
sulfur level, 26.1% aromatic content, 5.6% olefm content, 1.0% benzene content, 218F
T50 and 329 F T90.

       The effects shown in the chart are 'net fuel effects' from the MOVES model.
They were produced with three separate MOVES runs using a constant fuel formulation
and varying the fuel ethanol level from 0 percent to 15 percent by volume. In each chart,
separate curves are shown for THC, CO and NOx.   In all curves the fuel effects are
normalized to 10 percent volume ethanol.  In this context, 'normalization' means the
correction factor is set to unity with the ethanol volume is 10 percent.

       Overall, the results are as expected with only small changes in THC emissions,
moderate reductions in CO emissions, and NOx emission increases from the use of
ethanol.  For HC and NOx, there is about a five percent increase from zero ethanol to 10
percent ethanol.  Carbon monoxide emissions show approximately a 15 decrease from
zero ethanol to 10 percent ethanol.  This is likely due to the extra oxygen introduced by
ethanol into the combustion and exhaust process that oxidizes the carbon monoxide to
carbon dioxide. In MOVES2010a,  there are no fuel effects on particulate matter (PM)
emissions.
Figure 10-4   Relative Fuel Ethanol Effect on Pre-2004 Model Year Gasoline Vehicles
             in MOVES
    ii
   o
0)
3
LL.
0)

ro
                           Ethanol Fuel Effect
                       pre-tier2  - gasoline vehicles
                                                                      •voc

                                                                      •CO

                                                                      •NOx

                                                                      •Benzene

                                                                      •PM
                        4      6      8     10     12

                          Fuel Ethanol Volume (%)
                                                     14
16
                                       53

-------
Figure 10-5   Relative Fuel Ethanol Effect on Pre-2004 Model Year Gasoline Vehicles
             in MOVES
    ii
   o
    0)
    3
    0)
   '•5   0
to

"a.
"5
                           Ethanol  Fuel Effect
                       pre-tier2  - gasoline vehicles
                    •Ethanol

                    •Acetaldehyde
                                  10
15
20
                        Fuel Ethanol Volume  (%)
       Figure 10-5 shows the relationship between fuel ethanol and exhaust ethanol and
acetaldehyde emissions. These are shown on a separate figure because the fuel effects
range from zero to almost a factor of two for these two air toxic pollutants. Both these
curves  show a strong and predictable response to fuel ethanol from 0 to 10 percent
ethanol.

       Figures 10-4 and 10-5 show a maximum value of 15 percent. However, users
should not attempt to model higher ethanol inputs such as 15 percent ethanol (El5) or 85
percent ethanol (E85) with MOVES2010a. For ethanol values greater than 10 percent,
the relationships in MOVES2010a are pure extrapolation.  Future versions of MOVES
will show updated relationships based on recent vehicle testing, rather than mathematical
extrapolations.
10.3   Benzene Adjustment Factors in MOVES2010a

       The emission effects of fuel benzene are shown in Figure 10-6 for model year
2003. Charts for other model years are not shown, because the results are similar. Fuel
benzene levels affect only exhaust and evaporative (not shown) benzene emissions.
Figure 10-6 shows the relative effect of fuel benzene on benzene exhaust emissions with
a relative effect of unity assigned to a fuel benzene level of one percent. The effect is
positive and generally linear over the range of fuel benzene in use.
                                       54

-------
Figure 10-6   Relative Fuel Benzene Effect on Pre-2004 Model Year Gasoline Vehicles
            in MOVES
   0)



   I
   0)
   CO
   £5.  0
      0.2
                        Benzene Fuel Effect
                     pre-tier2 - gasoline vehicles
                                               •Benzene
   0)
   3
0.5
1.5
2.5
                    Fuel Benzene Content (volume %)
                                   55

-------
10.4   Olefm Adjustment Factors in MOVES2010a
       Figure 10-7 shows the relative effect on pre-2004 model year gasoline vehicles to
fuel olefm levels.  The relative effect of unity is assigned to a fuel olefm level of 5.6%
because that is the olefm level of the MOVES2010a reference fuel.  Only pollutants VOC
and 1,3-butadiene are affected by fuel olefm levels.  The VOC response is slightly
negative, but the 1,3-butadiene response is positive and quite significant over the range of
olefm levels in use.  The other fuel parameters were set at Base Fuel levels (6.9 psi RVP,
30 ppm sulfur level, 0% Ethanol volume, 26.1% aromatic content, 1.0% benzene content,
218FT50and329FT90.
Figure 10-7   Relative Fuel Olefm Effect on Pre-2004 Model Year Gasoline Vehicles in
             MOVES
   £

   1
   o
               Olefin Fuel Effect
          pre-tier2  - gasoline vehicles
                                                                   1,3-Butadiene
0      2      4      6      8     10     12

         Fuel Olefin Content (volume %)
                                                          14
                                      56

-------
10.5   E200 / E300 Adjustment Factors in MOVES2010a
       Figure 10-8 shows the relative effect on pre-2004 model year gasoline vehicles to
fuel E200 and E300 levels.  The relative effect of unity is assigned to a fuel e200 level of
41% and a fuel E300 level of 83% (T50 and T90 are 218F and 329F, respectively)
because those are the levels in the MOVES2010a reference fuel. The parameters E200
and E300 are measures of gasoline volatility, and are shown as a pair in Figure 10-8.  In
this example, both move as  a pair either up or down from the 41%/83% reference.  In-
use, this relationship between E200 and E300 does not always occur. Also, some of the
values of E200 and E300 cannot occur in conjunction with the Base Fuel parameters.
The E200 / E300 domain in Figure 10-8 is the entire range of values in the MOVES fuel
database, and the two points at the end of the curves are extreme values. The other fuel
parameters were set at Base Fuel levels (6.9 psi RVP, 30 ppm sulfur level, 0% Ethanol
volume, 26.1% aromatic content, 5.6% olefm content and 1.0% benzene content.

       The pollutants VOC and CO are both affected by E200 / E300 and so are most of
the air toxic pollutants. Most of the pollutants increase as E200 / E300 become larger
(i.e., the fuel contains a larger fraction of heavier constituents).  The exception is 1,3-
butadiene which shows a negative relationship with E200 / E300.
Figure 10-8   Relative Fuel E200 / E300 Effect on Pre-2004 Model Year Gasoline
             Vehicles in MOVES
                        E200 / E300 Fuel Effect
                       pre-tier2  - gasoline vehicles
    0)
    3
              38/78     39.5/81    41/83     50/87

                              E200/E300
64/92
VOC

CO

Benzene

1,3- Butadiene

Formaldehyde

Acetaldehyde
                                      57

-------
Appendix A  Peer Review

       A preliminary draft version of this document was peer reviewed by Professor
Tom Durbin, Ph.D Research Engineer at the University of California, CE-CERT,
Riverside, CA.  Many of his comments were addressed in this final version; some
sections of draft report were removed altogether from the final report. Professor Durbin
made seventeen distinct comments.  These are paraphrased in italics, and addressed
below. The full text of Professor Durbin's comments is available in a separate document.
1.      The document should have a formal reference section.

We have added footnotes to the document and plan a references page. Web links are
inserted throughout the document for important reference documents.

2.      Abbreviations should be better defined in the document.

We agree and have addressed this comment.

3.      It would be interesting if the document could tie in the written text with the
       Appendices with the code.

Between the draft that Prof. Durbin reviewed and the release of MOVES2010, much of
the work that was done outside the  model was incorporated in the MOVES code and the
need for the fuel binner was eliminated.  For more detail on the MOVES code itself, see
the MOVES "Software Design Reference Manual" and the "Programmer's Guide to
MOVES."

4.      Since the values for the fuel formulation table are described in greater detail later
       in the document, it might be useful to indicate that these variables will be
       discussed in the greater detail in the document.

Section 9.0 of the document has been added to provide an overview of the fuel
formulation and fuel supply development process. It describes  the process of developing
these critical data tables and provides a list of references to the  primary documents.

5.      There is not a good logical  development of the thoughts in Section 2.

The draft version reviewed by Professor Durbin has been thoroughly revised, and we
hope it contains a clearer description of the Complex model structure.
6.      The concept of using the 30 ppm fuel for the 2001 and later model year vehicles
       and using the 90 ppm fuel for all 2000 and earlier model years is not very clear.

Unfortunately, the fuel properties of the vehicle dataset used to develop the basic
emission factors were not measured with any real precision. The value of 30 ppm sulfur
was chosen because  is it the standard reference case for most EPA fuel models, and was
the assumed in-use fuel standard during the latter period of testing Arizona. The value of
                                       58

-------
90 ppm sulfur was a typical value during the beginning of the Arizona test period.  The
value of 90 ppm comes with considerable uncertainty.

7.      A slight expansion of the GPA regions would be useful.

This section has been expanded in the final version.
8.      There should be some discussion as to what oxygenate factors are used for
       different cases with ethanol, MTBE, and TAME. It might be interesting to include
       some statistics in the document about what percent of fuels contained these
       various oxygenates.  That would provide the reader a good context for the relative
       importance of as to how fuel formulations have changed over time.

We did not provide the statistics, but they could be developed with effort from the
MOVES Fuel Supply and Fuel Formulation tables. In general, the calendar year 2005
and earlier generally contain modest amounts of MTBE as the primary oxygenate. After
calendar year 2006, MTBE declines dramatically because of ground water contamination
concerns and is replaced almost exclusively by Ethanol.  TAME and ETBE were never
major oxygenate sources and rarely achieve more than a five percent market share.  In
current years (i.e., 2010+) they are all but gone from the fuel mix.
9.      This comment address fuel supplies and the fact that MO VES does not contain
       default fuel supplies past calendar years 2012.

This topic has been removed in the final draft because it pertains to the detailed fuel
supply data and statistics which are now outside the scope of the document. However,
the default MOVES only contains fuel supply information through 2012.  As this date
approaches, this limitation will have to be addressed. For current State Implementation
Plan (SIP) work, users are expected to enter their own fuel supply information for ALL
years.  Once the user supplied data is entered, the model correctly performs all of the fuel
calculations.

10.    A mostly editorial comment regarding clarity on E10, E85 andNMIM.

The topics of E85  and NMEVI have been completely removed from the document.

11.    A comment regarding at 2022 fuel supply database.

This topic has been completely removed from the document.

12.    An editorial comment on the Predictive model.

The Predictive model is now built directly into MOVES and is fully documented in
Section 4.0

13.    A comment on High emitters with the question "What information was used to
       come up with the 20% estimate for high emitters ".
                                       59

-------
This assumption has been changed.  The new MOVES algorithm now uses the High
emitter fraction from MOBILE6.2 that is a function of age (0 through 30).  The
calculation of High emitters is fully document in the MOBILE6.2 documentation.
14.     The assumption that NOx and THC emissions for 1994 and newer vehicles are no
       longer sensitive to the effects of fuel parameters outside of the sulfur is a very
       large simplification of the situation.  Of the fuel parameters,  the addition of
       oxygenates / ethanol is one area that would seemingly be fairly critical given that
       the levels of ethanol in gasoline will be increasing.  He subsequently cites studies
       to boost his argument.

We agree that the 1994 and later assumption was too much of a simplification and
changed the assumption to 2004 and later model years. The 2003 and earlier years are
now fully affected by all of the fuel parameters as appropriate for the pollutant.

75.     The Fuel Binner code is not in Appendix A.

MOVES no longer requires fuel binning, so this code is no longer relevant.

16.     What is the basis for the EPA belief that Tier2 vehicles are less sensitive to fuel
       properties other than sulfur?

At the time of this document development, the EPAct vehicle - fuel  study had not been
completed and its results could not be incorporated into the model.  Based on very limited
knowledge of those vehicles,  we made the assumption that 2004 and later model years
(i.e., Tier2) vehicles are only sensitive to fuel  sulfur in MOVES2010a.

Since the release of MOVES2010a, the EPAct study and the statistical analysis of the
results  have been completed, and these show that Tier2 vehicles are  quite sensitive to an
entire range of fuel properties depending on pollutant.  Subsequent versions of MOVES
will incorporate these new findings.

17.     Professor Durbin includes a set of grammatical and editorial comments.

The document has been restructured and revised to address these comments.
                                       60

-------