Regulatory Impact Analysis:
   Renewable Fuel Standard Program
United States
Environmental Protection
Agency

-------
                         Regulatory Impact Analysis:
                     Renewable Fuel Standard Program
                                  Assessment and Standards Division
                                 Office of Transportation and Air Quality
                                 U.S. Environmental Protection Agency
v>EPA
                  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 an exchange of technical information and to inform the public of
                  technical developments.
United States                                          EPA420-R-07-004
Environmental Protection                                   .  ., „„.,
Agency                                              April 2007

-------
                               Table of Contents


Statement of Need	4

Overview	5

List of Acronyms and Abbreviations	7

Chapter 1: Industry Characterization	11

Chapter 2: Changes to Motor Vehicle Fuel
      Under the Renewable Fuel Standard Program	50

Chapter 3: Impacts on Emissions from Vehicles,
      Nonroad Equipment, and Fuel Production Facilities	118

Chapter 4: National Emission Inventory Impacts	173

ChapterS: Air Quality Impacts	200

Chapter 6: Lifecycle Impacts on Fossil Energy and Greenhouse Gases	218

Chapter 7: Estimated Costs of Renewable Fuels, Gasoline and Diesel	262

ChapterS: Agricultural Sector Impacts	323

Chapter 9: Small-Business Flexibility Analysis	335

Endnotes	342

-------
                                Statement of Need
       The United States currently consumes about 190 billion gallons of gasoline and diesel
fuel annually to meet its transportation fuel needs. Of this volume, about 65 percent, or 124
billion gallons, is derived from foreign sources.  The United States' dependence on imported
petroleum to meet its growing demand for transportation fuel exacts a cost on the nation in terms
of energy security.  In addition, petroleum-based fuel exacts a cost on the nation with respect to
environmental quality.  The Renewable Fuel Standard (RFS) program increases national energy
security by creating a market for renewable fuel as a substitute for petroleum-based fuel. By
incorporating incentives for investing in research and development of renewable fuels, the RFS
program also seeks to accelerate the nation's progress toward energy independence. In addition,
the RFS program helps to reduce the country's greenhouse gas emissions, thereby reducing the
nation's contribution to global climate change and its potential effects on the U.S. economy,
security, and public health.

-------
                                      Overview
       EPA is finalizing standards which would implement a renewable fuel program as
required by the Energy Policy Act of 2005 (the Act). The Act specifies the total volume of
renewable fuel that is required to be used each year, and directs EPA to adjust this amount under
certain circumstances. The resulting standards represent a level of renewable fuel that each
refinery or importer must account for relative to its annual volume of gasoline produced or
imported.  In reality, however, renewable fuel use is forecast to exceed the RFS standards due to
market forces.  The analyses of the impacts associated with this increase in renewable fuel use
are discussed in this Regulatory Impact Analysis (RIA).

Chapter 1:  Industry Characterization
This chapter discusses current gasoline, diesel and renewable fuel production, importation,
marketing and distribution, as well as likely future changes as a result of increased renewable
fuel use.

Chapter 2:  Changes to Motor Vehicle Fuel Under the RFS Program
This chapter discusses our gasoline and renewable fuel consumption predictions (compared to a
2004 base year), and the expected impacts of various ethanol blends on gasoline properties.

Chapter 3:  Impacts on Emissions from Vehicles, Nonroad Equipment and Fuel Production
Facilities
This chapter evaluates the impacts on vehicle and nonroad equipment emissions under various
oxygenate assumptions, specifically increasing ethanol and decreasing MTBE, and different
modeling techniques.  The effect of biodiesel use on diesel-powered vehicle emissions is also
presented. Finally, emissions from ethanol and biodiesel production facilities are discussed.

Chapter 4:  National Emissions Inventory Impacts
This chapter discusses the methods used to develop the national emissions inventories, and
quantifies the impact  of expanded ethanol and biodiesel use on those inventories.

Chapter 5:  Air Quality Impacts
This chapter discusses the impacts of expanded renewable fuel use on ozone and particulate
matter formation.

Chapter 6:  Lifecycle Impacts on Fossil Energy and Greenhouse Gases
This chapter discusses our fuel lifecycle modeling, that is, analysis which accounts for all energy
and emissions of the fuel production process. A description of the model we used, how we used
it, and the results are presented. Impacts on greenhouse gases, including CC>2, fossil fuel use, and
petroleum use are presented.  The effects on  petroleum imports, import expenditures, and
domestic energy security are also discussed.

-------
Chapter 7:  Estimated Costs of Renewable Fuels, Gasoline and Diesel
This chapter contains our analysis of the cost of corn and cellulosic ethanol.  We also discuss
biodiesel and renewable diesel production costs.  Costs associated with distributing the volumes
of ethanol necessary to meet the requirements of the program, and the costs to prepare gasoline
and diesel blendstocks (for blending with renewable fuels) are also presented. Finally, we
present the overall fuel cost impacts of expanded renewable fuel use.

ChapterS:  Agricultural  Sector Impacts
This chapter discusses the likely economic impacts on the agricultural sector that may occur as a
result of the large expansion of renewable fuel production and use expected in the future.  On-
going work using the FASOM model is also described.

Chapter 9:  Small Business Flexibility Analysis
This chapter discusses our Small Business Flexibility Analysis (SBFA) which evaluates the rule
to ensure that concerns regarding small businesses, which would be affected by the rule, are
sufficiently considered.

-------
List of Acronyms and Abbreviations
AAM
ABT
ACE
The Act
ADM
AEO
ANL
AQIRP
ARMS
BO, B5, B20, etc
bbl
BEA
Bgal, bgal, bilgal, billgal,
bg
BGY
BPCD
BPSD
bpd, bbls/day
BTU
BU
Bu/acre
BZ
CA
CAA
CAIR
CARB
CaRFGS
CBG
CBI
CD
CFEIS
CFR
c/gal
CG
CHP
CO
CO2
Co-op
CRC
DDGS
DOE
DRIA
E&C
EO
E10
E85
E200
E300
EIA
Alliance of Automobile Manufacturers
Averaging, Banking, and Trading
American Coalition for Ethanol
Energy Policy Act of 2005 (also the Energy Act)
Archer Daniels Midland
Annual Energy Outlook (an EIA publication)
Argonne National Laboratory
Auto/Oil Air Quality Improvement Research Program
Agricultural Resource Management Survey
Percent of biodiesel, e.g., B5= 5%biodiesel, 95% diesel
Barrel
Bureau of Economic Analysis
Billions of gallons
Billions of gallons per year
Barrels Per calendar day
Barrels per stream day
Barrels Per Day
British Thermal Unit
Bushel
Bushels per acre
Benzene
California
Clean Air Act
Clean Air Interstate Rule
California Air Resources Board
California Phase 3 RFC
Cleaner Burning Gasoline
Caribbean Basin Initiative
Census Division
EPA's Certification and Fuel Economy Information System
Code of Federal Regulations
Cents per gallon
Conventional Gasoline
Combined Heat and Power Technology
Carbon Monoxide
Carbon Dioxide
Cooperative
Coordinating Research Council
Distillers' Dried Grains with Solubles
Department of Energy
Draft Regulatory Impact Analysis
Engineering and Construction
Gasoline Blend which Does Not Contain Ethanol
Gasoline Blend containing a nominal 10 percent ethanol by volume
Gasoline Blend containing 85 percent ethanol by volume
Percent of Fuel Evaporated at 200 Degrees F (ASTM D 86)
Percent of Fuel Evaporated at 300 Degrees F (ASTM D 86)
Energy Information Administration (part of the U.S. Department of Energy)

-------
Energy Act
EO
EPA
EPAct
ETBE
ETOH
exCA
F, °F
FAPRI
FASOM
FBP
FCC
FCCU
FHWA
FOEB
FR
FRM
FRTP
FFV
FTP
GAL
g/Btu
g/day
GDP
GHG
GPA
GREET
GWP
HC
HCO
HDN
HSR
HVGO
IBP
k
kbbl
kwh
Lb
LCD
LEV
LLE
LNS
LP
LSR
mg/m3
MGY, MMgy
MM
MMBTU
MMbbls/cd
MMGal/yr
MOBILE (5, 6, 6.2)
MON
MOVES2006
Energy Policy Act of 2005 (also the Act)
Executive Order
Environmental Protection Agency
Energy Policy Act of 2005 (also 'the Energy Act' or 'the Act')
Ethyl Tertiary Butyl Ether
Ethanol
Excluding California
Fahrenheit
Farm and Agricultural Policy Research Institute
Forestry and Agriculture Sector Optimization Model
Feed Boiling Point (also Final Boiling Point)
Fluidized Catalytic Cracker
Fluidized Catalytic Cracking Unit
Federal Highway Administration
Fuel Oil Equivalent Barrel
Federal Register
Final Rulemaking
Fixed Reduction Trigger Point
Flexible Fuel Vehicle
Federal test procedure
Gallon
Grams per Btu
Grams per day
Gross Domestic Product
Greenhouse Gases
Geographic Phase-in Area
Greenhouse Gas, Regulated Emissions, and Energy Use in Transportation model
Global warming potentials
Hydrocarbon(s)
Heavy Cycle Oil (a refinery stream)
Naphtha Hydrotreater (also Hydro-Denitrogenation Unit)
Heavy Straight Run (a refinery stream)
Heavy Vacuum Gas Oil (a refinery stream)
Initial Boiling Point
Thousand
Thousand barrels
Kilowatt Hour
Pound
Light Cycle Oil (a refinery stream)
Low emission vehicle
Liquid-Liquid Extraction
Light Naphtha Splitter
Linear Programming (a type of refinery model)
Light Straight Run (a refinery stream)
Milligrams per cubic meter
Million Gallons per Year
Million
Million British Thermal Units
Millions of barrels per calendar day
Millions of gallons per year
EPA's Motor Vehicle Emission Inventory Model (versions)
Motor Octane Number
EPA's Next Generation Highway Vehicle Emission Model

-------
MSAT
MSAT1
MSAT2
MTBE
NAAQS
NAICS
NASS
NBB
NCSU
NGL
NMHC
NMIM
NMOG
NONROAD
NONROAD2005
NOX
NPRM
NREL
OMB
OMHCE
ORNL
OTAQ
Oxy-fuel, oxyfuel
PADD
PM
PM10
PM25
PMA
POM
PONA
ppb
ppm
PRTP
PSI
QBtu
Quadrillion
(R+M)/2
RBOB
RFA
RFC
RFS
RIA
RIMS
RIN
RON
RPMG
RSM
RVP
S
SBA
SBAR Panel, or 'the Panel'
SBFA
SBREFA
Mobile Source Air Toxics
200 1 Mobile Source Air Toxics Rule
2006 Proposed Mobile Source Air Toxics Rule
Methyl Tertiary -Butyl Ether
National Ambient Air Quality Standards
North American Industrial Classification System
National Agricultural Statistics Service
National Biodiesel Board
North Carolina State University
Natural gas plant liquids
Non-Methane Hydrocarbons
National Mobile Inventory Model (EPA software tool)
Non-methane organic gases
EPA's Non-road Engine Emission Model
EPA's Non-road Engine Emission Model Released in 2005
Oxides of nitrogen
Notice of Proposed Rulemaking
National Renewable Energy Laboratory
Office of Management and Budget
Organic Material Hydrocarbon Equivalent
Oak Ridge National Laboratory
Office of Transportation and Air Quality
Winter oxygenated fuel program
Petroleum Administration for Defense District
Paniculate Matter
Coarse Particle
Fine Particle
Petroleum Marketing Annual (an EIA publication)
Polycyclic Organic Matter
Paraffin, Olefin, Naphthene, Aromatic
Parts per billion
Parts Per million
Percentage Reduction Trigger Point
Pounds per Square Inch
Quadrillion btu
1015
Octane calculation (RON+MON)/2
Reformulated Blendstock for Oxygenate Blending
Regulatory Flexibility Act
Reformulated Gasoline
Renewable Fuels Standard
Regulatory Impact Analysis
Regional Input-Output Modeling System
Renewable Identification Number
Research octane number
Renewable Products Marketing Group
Response Surface Model
Reid Vapor Pressure
Sulfur
Small Business Administration
Small Business Advocacy Review Panel
Small Business Flexibility Analysis
Small Business Regulatory Enforcement Fairness Act (of 1996)
10

-------
scf
SOA
SOx
SULEV
T50
T90
TAME
ULEV
U.S.C.
USDA
VGO
VMT
voc
vol%
wt%
yr, y
Standard cubic feet
Secondary Organic Aerosol
Oxides of Sulfur
Super ultra low emission vehicle
Temperature at which 50% (by volume) of fuel evaporates (ASTM D 86)
Temperature at which 90% (by volume) of fuel evaporates (ASTM D 86)
Tertiary Amyl Methyl Ether
Ultra low emission vehicle
United States Code
U.S. Department of Agriculture
Vacuum Gas Oil (a refinery stream)
Vehicle Miles Traveled
Volatile Organic Compound
Percent by volume, volume percent
Percent by weight, weight percent
Year
11

-------
Chapter 1:  Industry Characterization
1.1    Transportation Fuel Providers

1.1.1   Petroleum Refiners

       As of the end of 2005, there were 142 crude oil refineries operating in the United States,
representing a total of 16.4 million barrels/day of refining capacity. (These refineries produce
gasoline and other products and are a separate category than "blender refiners" that do not
process crude oil, but make gasoline from blendstocks.) The greatest number of refineries per
PADD is in PADD 3 (the Gulf Coast region) which has 52 operating refineries as of the end of
2005. This PADD also has the greatest refining capacity, at 7.9 million barrels per day.  Table
1.1-1 presents the refineries and their crude oil production capacity, and identifies the PADD
where the refinery is located.

                                     Table 1.1-1.
                       Refining Capacity by Individual Refinery
                              (crude oil processing basis)
Company
Conoco Phillips
Wood River, IL
Belle Chasse, LA
Sweeny, TX
Westlake LA
Linden, NJ
Ponca City OK
Trainer, PA
Borger TX
Wilmington CA
Ferndale WA
Rodeo CA
Billings MT
Valero Energy Corp.
Port Arthur TX
Memphis TN
Lima OH
Texas City TX
Corpus Christ! TX
Houston TX
Sunray TX
Three Rivers TX
Norco LA
Paulsboro NJ
Benecia CA
Wilmington CA
Ardmore OK
Wilmington CA
Capacity
(MMbbls/cd)
2.2
0.31
0.25
0.25
0.24
0.24
0.19
0.19
0.15
0.14
0.10
0.08
0.06
2.0
0.26
0.18
0.15
0.21
0.14
0.08
0.16
0.09
0.19
0.16
0.14
0.01
0.08
0.08
PADD

2
3
3
3
1
2
1
3
5
5
5
4

3
2
2
3
3
3
3
3
3
1
5
5
2
5
                                          12

-------
Company
Krotz Springs LA
Exxon Mobil Corp.
Baytown TX
Baton Rouge LA
Beaumont TX
Joliet IL
Torrance CA
Billings MT
Chalmette, LA
BPPLC
Texas City TX
Whiting IN
Toledo OH
Los Angeles CA
Ferndale WA
Chevron Corp.
Pascagoula MS
ElSegundo CA
Richmond CA
Honolulu HI
Salt Lake City UT
Marathon Oil Corp.
Garyville LA
Cattlettsburg KY
Robinson IL
Detroit Ml
Canton OH
Texas City TX
Saint Paul Park MN
Sunoco, Inc.
Marcus Hook PA
Toledo OH
West/Hie NJ
Tulsa OK
PDV America, Inc.
Citgo; Lake Charles LA
Citgo, Lemont IL
Citgo; Corpus Christi TX
Koch Industries
Corpus Christi TX
Saint Paul MN
Motiva Enterprises LLC
Port Arthur TX
Convent LA
Norco LA
Tesoro Corp.
Anacortes WA
Salt Lake City UT
Capacity
(MMbbls/cd)
0.08
2.0
0.56
0.50
0.34
0.24
0.75
0.06
0.19
1.5
0.44
0.41
0.13
0.26
0.23
0.9
0.33
0.26
0.24
0.05
0.05
1.0
0.25
0.22
0.19
0.10
0.07
0.07
0.07
0.58
0.18
0.16
0.15
0.09
0.81
0.43
0.17
0.16
0.57
0.29
0.28
0.76
0.29
0.24
0.23
0.51
0.12
0.06
PADD
3

3
3
3
2
5
4
3

3
2
2
5
5

3
5
5
5
4

3
2
2
2
2
3
2

2
2
1
2

3
2
3

3
2

3
3
3

5
4
13

-------
Company
Martinez CA
Kapolei HI
Kenai AK
Royal Dutch/Shell Group
Martinez CA
Anacortes WA
Wilmington CA
Saraland AL
Deer Park, TX
Lyondell Chem. Co.
(Houston)
Total SA (Port Arthur, TX)
Sinclair Oil
Tulsa OK
Sinclair WY
Evansville WY
Murphy Oil
Meraux LA
Superior I/I//
Frontier Oil
El Dorado KS
Cheyenne WY
Cenex Harvest States, Inc.
McPherson KS
Laurel MT
Coffeyville Acquisitions
(Coffeyville KS)
Navajo Refining Corp.
Artesia NM
Woods Cross UT
Great Falls MT
Pasadena Refining Systems
(Pasadena TX)
Giant Industries, Inc.
Yorktown VA
Gallup NM
Bloomfield NM
Big West Oil (North Salt
Lake UT)
Capacity
(MMbbls/cd)
0.17
0.09
0.072
0.82
0.16
0.15
0.10
0.08
0.33
0.27
0.23
0.17
0.07
0.07
0.03
0.15
0.12
0.03
0.15
0.11
0.04
0.14
0.08
0.06
0.11
0.11
0.07
0.03
0.01
0.10
0.10
0.06
0.02
0.02
0.10
PADD
5
5
5

5
5
5
3
3
3
3

2
4
4

3
2

2
4

2
4
2

3
4
4
3

1
3
3
4
               Source: Table 5 in Energy Information Administration, Refinery Capacity 2006 found at
       http://www.eia.doe.gov/pub/oil gas/petroleum/data publications/refinery capacity data/current/table5.pdf
       Refining capacity has steadily increased in the U.S. due to increased demand for
petroleum products, with gasoline representing approximately 45 percent of product demand.
Refining capacity (crude oil input) was about 14 million bbls/day in 1973 and 17 million
bbls/day in 2005.  While refining capacity has increased, however, the number of refineries has
                                              14

-------
decreased as less economical refineries have been forced to close.  (Many of these came into
existence for a very short time due to oil price supports in the 1970's.) In the  1970's, the number
of refineries in the U.S. was approximately 270 and has decreased by 47 percent. Figure 1.1-1
shows the number of refineries and total capacity in the U.S. from 1973 through 2004.
                                       Figure 1.1-1.
           Number of Refineries and Total Capacity in the U.S. from 1973-2004
  350
  200 -
E 150 - -
   50 -
                                                                            20.0
                                                                          - - 16.0
                                                                          - - 14.0
                                                                          - - 12.0
                                                                          - - 10.0
                                                                          - - 2.0
                        Source: EIA; Annual Energy Report, 2005 (Table 5.9)
       The increase in capacity combined with the decrease in amount of refineries and the
increased demand for gasoline and diesel fuels, has resulted in an increase in the average
utilization rate of refineries. In the 1970's, the utilization rate ranged from 84 to 94 percent. In
the last ten years, however, the utilization rate has ranged from 91 to 96 percent.  Refineries
therefore have to produce more with less overall capacity. The amount of gasoline and diesel
produced by U.S. refiners has steadily increased.  Since 1973 through 2004, gasoline and diesel
production has increased 27 and 36 percent, respectively. Figure 1.1-2 shows the change in
gasoline and diesel production from 1973 through 2004
                                            15

-------
                                      Figure 1.1-2.
                Amount of Gasoline and Diesel Fuels Produced in the U.S.
     140.0
     120.0
     100.0
      80.0
      40.0
      20.0
0.0 -

 >$

                /A
                oA
              Source: EIA Annual Energy Report, 2005; Table 5.8
1.1.2   Petroleum Imports

       The decrease in U.S. refining capacity discussed in Section 1.1.3, has resulted in
increases in the amount of gasoline and diesel fuels imported into the U.S. As of 2004, 5.4 and
11.5 percent of the total respective volumes of gasoline and diesel consumed in the U.S. were
imported.

Today, the United States imports approximately 70 percent of all petroleum products used, with
two-thirds of these products being used for transportation.  From 1973 to 2004, the amount of
crude oil imported has increased from 1.2 to 3.7 billion barrels per year, a tripling of volume,
representing an average annual increase of about 6 percent. Over the same time period, the
amount of gasoline imported has increased from 2 to 7.4 billion gallons per year, more than three
times the amount of volume. The amount of diesel imported in the same time period decreased
slightly from 6 to 5 percent. Figures 1.1-3 and 1.1-4 show the increase in crude oil and
gasoline/diesel fuel imports, respectively, from 1973 to 2004.
                                           16

-------
                                       Figure 1.1-3.
                 Increase in Crude Oil Imports from 1973-2004
  4.00 -,
  3.50
  3.00
3 2.50
  2.00
  1.50
  1.00
  0.50
  0.00
                                           Years
        (Source: Annual Energy Outlook, 2005; Energy Information Administration)
                                         17

-------
                                      Figure 1.1-4.
           Change in Volumes of Imported Gasoline and Diesel fuels (1973-2004)
    14.00
    12.00
     0.00
                /A
                                           o9>
                                          ^

                                          Years
              Source: Annual Energy Outlook, 2005; Energy Information Administration
       Approximately twenty percent of our trade deficit is from imported petroleum products, a
deficit which reached $782 billion in 2005.  Figure 1.1-5 shows the trade deficits from 1994
through 2004 (earlier data on petroleum imports is not available from the U.S. Census web site at
this time). While the overall contribution of petroleum imports to the total deficit is decreasing
as shown in Figure 1.1-5, this is due to a more rapid growth in the total deficit from other goods
and services.  The portion of the deficit due to petroleum imports by itself is increasing by
approximately 4 percent per year.  Over the last 25 years, the cumulative cost of imported crude
oil has reached $2.0 trillion in 2005 dollars.
                                            18

-------
                                       Figure 1.1-5.
                U.S. Trade Deficit and Portions Due to Petroleum Imports
                                        1994-2004
                        (Millions of dollars, chain weighted to 2000)
    700,000
    600,000
    500,000
  £ 400,000
  o
  T3
  •5
  «
  c
  i 300,000
    200,000
    100,000 --
D Trade deficit
• Deficit due to petroleum
           1994   1995   1996  1997   1998   1999   2000   2001   2002   2003   2004

              Source: U.S. Census Bureau, Foreign Trade Statistics, 2006
       The amount of import facilities in the U.S. has stayed relatively constant since the U.S.
EPA has been requiring such facilities to register. In 1995 there were a total of 39 such facilities
in the U.S. The amount has remained relatively constant, in the 50's since that time and as of
2004 there were 53 such facilities registered with U.S. EPA. The great majority of such facilities
are located in PADD 1; as of 2004, 35 facilities were in PADD 1, and a total of 18 in  the other
four PADDs.
1.2    Renewable Fuel Production

       While the definition of renewable fuel in the Act does not limit compliance with the
standard to any one particular type of renewable fuel, ethanol is currently the most prevalent
renewable fuel blended into motor vehicle fuels today. Biodiesel represents another form of
renewable fuel, which while not as widespread as ethanol use (in terms of volume), has been
increasing in production capacity and use over the last several years.  Ethanol and biodiesel are
                                            19

-------
expected to continue to dominate renewable fuel use in the timeframe when the RFS rule will be
phasing in.

1.2.1  Current U.S. Ethanol Production

1.2.1.1     Overview

       There are currently 110 ethanol production facilities in the United States with a combined
production capacity of 5.2 billion gallons per year1.  This baseline, or starting point, for this
regulatory impact analysis is based on U.S. ethanol production facilities operational as of
October 2006.2ABCDE

       Approximately 92 percent of today's ethanol production capacity is produced exclusively
from corn, mainly from a dry-milling process.  The remainder is derived from corn/grain blends,
cheese whey, and other starches.  The majority of ethanol plants are located in Midwest where
the bulk of corn is produced. PADD 2 accounts for just over 5 billion gallons (or 96 percent) of
the total U.S. ethanol production. Leading the Midwest in ethanol production are Iowa, Illinois,
Nebraska, Minnesota, and South Dakota which together represent 76 percent of the total
domestic product. In addition to the concentration of facilities located in PADD 2, there are also
a sprinkling of ethanol plants situated outside  of the Midwest as far west as California and as far
south as Georgia.

1.2.1.2     Ethanol Feedstocks & Processing Technologies

       All of the ethanol currently produced today comes from grain or starch-based feedstocks
that can easily be broken down into ethanol via traditional fermentation processes.  The primary
feedstock is corn, although grain sorghum (milo), wheat, barley, beverage waste, cheese whey,
and sugars/starches are also fermented to make fuel-grade ethanol.

       The majority of ethanol (almost 92 percent by volume) is produced exclusively from
corn. Most of the corn originates from the Midwest and most of the ethanol is produced in
PADD 2 close to where the corn is grown. However, several corn-ethanol plants are  also
situated outside the traditional "corn belt". In California, Colorado, New Mexico,  and Wyoming
corn is shipped from the Midwest to supplement locally grown grains or in some cases, serve as
the sole feedstock.  As for the remaining ethanol, almost eight percent is produced from a blend
of corn and/or similarly processed grains (milo, wheat, or barley) and less than one percent is
1 This analysis does not consider ethanol plants that may be located in (or planned for) the Virgin Islands or U.S.
territories.

2 The October 2006 ethanol production capacity baseline was generated based on the June 2006 NPRM plant list and
updated on October 18, 2006 based on a variety of data sources including: Renewable Fuels Association (RFA),
Ethanol Producer Magazine (EPM), ICF International, BioFuels Journal, and ethanol producer websites.  The
baseline includes small-scale ethanol production facilities as well as former food-grade ethanol plants that have
since transitioned into the fuel-grade ethanol market.  Where applicable, current ethanol plant production levels have
been used to represent plant capacity, as nameplate capacities are often underestimated.
                                             20

-------
produced from waste beverages, cheese whey, and sugars/starches combined.  A summary of
ethanol production by feedstock is presented in Table 1.2-1.
                                       Table 1.2-1.
                        2006 U.S. Ethanol Production by Feedstock
Plant Feedstock
Cheese Whey
Corn3
Corn, Barley
Corn, Milob
Corn, Wheat
Milo, Wheat
Sugars, Starches
Waste Beverages0
Total
Capacity
MMgy
8
4,780
40
244
90
40
2
16
5,218
%of
Capacity
0.1%
91 .6%
0.8%
4.7%
1 .7%
0.8%
0.0%
0.3%
100.0%
No. of
Plants
2
90
1
8
2
1
1
5
110
%of
Plants
1 .8%
81 .8%
0.9%
7.3%
1 .8%
0.9%
0.9%
4.5%
100.0%
Includes two facilities processing seed corn and another facility
processing corn which intends to transition to corn stalks, switchgrass,
and biomass in the future.
blncludes one facility processing small amounts of molasses in addition
to corn and milo.
clncludes two facilities processing brewery waste.
       There are two primary plant configurations for processing grains into ethanol: dry mill and
wet mill. A summary of the processing technologies used by today's ethanol plants is found below
in Table 1.2-2.

       Dry mill plants simply grind the entire kernel and feed the flour into the fermentation
process to produce ethanol.  At the end, the unfermentable parts are recovered as distillers'
grains along with a soluble liquid containing vitamins, minerals, fat and protein. The distillers'
grains are concentrated with the solubles stream to make a single co-product, referred to as
distillers' grains with solubles (DGS). The co-product is  either sold wet (WDGS) or more
commonly dried (DDGS) to the agricultural market as animal feed. If the feed is going to be
used by local markets,  it's usually sold wet precluding the need for process dryers. However, if
the feed is going to be shipped (usually by train) to more distant locations, the product is usually
dried to facilitate storage and transportation.

       Wet mill plants typically  separate the kernel into four products:  starch, gluten feed, gluten
meal, and oil.  The starch is used in a fermentation process the same as in dry mill plants, while the
gluten, oil, and other co-products are sold into food and agricultural markets. Production of these
multiple streams is more capital-intensive than the dry milling process,  and thus wet mill plants are
generally more expensive to build and tend to be larger in size.
                                            21

-------
                                         Table 1.2-2.
                   2006 U.S. Ethanol Production by Processing Technology
Processing
Technology
Dry Milling
Wet Milling
Other3
Total
Capacity
MMgy
4,057
1,137
25
5,218
%of
Capacity
77.7%
21 .8%
0.5%
100.0%
No. of
Plants
92
10
8
110
%of
Plants
83.6%
9.1%
7.3%
100.0%
aPlants that do not process traditional grain-based crops and thus do not
require milling. This category includes plants processing cheese whey,
sugars & starches, or waste beverages.
       As shown above in Table 1.2-2, dry milling is the most predominant production process
used by today's ethanol plants. Of the 102 facilities processing corn and/or other similarly
processed grains, 92 utilize dry milling technologies and the remaining 10 plants rely on wet
milling processes (refer to Table  1.2-3 below).  The remaining "other" eight plants listed above
process waste beverages, cheese whey, or sugars/starches and operate differently than their
grain-based counterparts. These facilities do not require milling and instead operate a simpler
enzymatic fermentation process.

                                       Table 1.2-3.
                  2006 U.S. Grain Ethanol Production - Wet Mill Plants

Ethanol Plant
Archer Daniels Midland3
Archer Daniels Midland3
Archer Daniels Midland
Archer Daniels Midland3
Archer Daniels Midland
Aventine Renewable Energy
Cargill, Inc.
Cargill, Inc.
Grain Processing Corp
Tate & Lyle
Total

Location
Cedar Rapids, IA
Clinton, IA
Columbus, NE
Decatur, IL
Marshall, MN
Pekin, IL
Eddyville, IA
Blair, NE
Muscatine, IA
Loudon, TN

Capacity
MMgy
300
150
90
250
40
100
35
85
20
67
1,137
Estimated ADM plant capacities
       In addition to grain and starch-to-ethanol production, another method exists for producing
ethanol from a more diverse feedstock base.  This process involves converting cellulosic
materials such as bagasse, wood, straw, switchgrass, and other biomass into ethanol. Cellulose
consists of tightly-linked polymers of starch, and production of ethanol from it requires
additional steps to convert these polymers into fermentable sugars. Scientists are actively
pursuing acid and enzyme hydrolysis as well as gasification to achieve this goal, but the
                                           22

-------
technologies are still not fully developed for large-scale commercial production. As of October
2006, the only known cellulose-to-ethanol plant in North America was logen in Canada, which
produces approximately one million gallons of ethanol per year from wood chips.  Several
companies have announced plans to build cellulose-to-ethanol plants in the U.S., but most are
still in the research and development or pre-construction planning phases. The majority of the
plans involve converting bagasse, rice hulls, wood, switchgrass, corn stalks, and other
agricultural waste or biomass into ethanol.  For more a more detailed discussion on future
cellulosic ethanol plants and production technologies, refer to RIA Sections 1.2.3.6 and 7.1.2,
respectively.

1.2.1.3    Ethanol Plant Energy Sources

       Ethanol production is a relatively resource-intensive process that requires the use of
water, electricity, and steam. Steam needed to heat the process is generally produced onsite or
by other dedicated boilers.  Of today's 110 ethanol production facilities, 101 burn natural gas, 7
burn coal, 1 burns coal and  biomass, and 1 burns syrup from the process to produce steam3. Our
research suggests that 11 plants currently utilize co-generation or combined heat and power
(CHP) technology, although others may exist. CHP is a mechanism for improving overall plant
efficiency.  Whether owned by the ethanol facility, their local utility, or a third party; CHP
facilities produce their own electricity and use the waste heat from power production for process
steam, reducing the energy  intensity of ethanol production.  A summary of the energy sources
and CHP technology utilized by today's ethanol plants is found below in Table 1.2-4.

                                      Table 1.2-4.
                     2006 U.S. Ethanol Production by Energy Source
Plant Energy
Source
Coal
Coal, Biomass
Natural Gasa
Syrup
Total
Capacity
MMgy
1,042
50
4,077
49
5,218
%of
Capacity
20.0%
1 .0%
78.1%
0.9%
100.0%
No. of
Plants
7
1
101
1
110
%of
Plants
6.4%
0.9%
91 .8%
0.9%
100.0%
CHP
Tech.
2
0
9
0
11
Includes three facilities burning natural gas which intend to transition to coal or
biomass in the future.
1.2.1.4    Ethanol Production Locations

       The majority of domestic ethanol is currently produced in the Midwest within PADD 2 -
where most of the corn is grown. Of the 110 U.S. ethanol production facilities, 100 are located
in PADD 2.  As a region, PADD 2 accounts for about 96 percent (or over five billion gallons) of
domestic ethanol production, as shown in Table 1.2-5.
! Facilities were assumed to burn natural gas if the plant fuel type was not mentioned or unavailable.
                                           23

-------
                                      Table 1.2-5.
                         2006 U.S. Ethanol Production by PADD
PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5
Total
Capacity
MMgy
0.4
5,012
30
105
71
5,218
%of
Capacity
0.0%
96.0%
0.6%
2.0%
1 .4%
100.0%
No. of
Plants
1
100
1
4
4
110
%of
Plants
0.9%
90.9%
0.9%
3.6%
3.6%
100.0%
       Leading the Midwest in ethanol production are Iowa, Illinois, Nebraska, Minnesota, and
South Dakota with capacities of 1.62, 0.71, 0.61, 0.55, 0.49 billion gallons, respectively.
Together, these five states' 70 ethanol plants account for 76 percent of the total domestic ethanol
production. However, although the majority of ethanol production comes from PADD 2, there
are a growing number of plants situated outside the traditional corn belt.  In addition to the 15
states comprising PADD 2, ethanol plants are currently located in California, Colorado, Georgia,
New Mexico, and Wyoming.  Some of these facilities ship in feedstocks (namely corn) from the
Midwest, others rely on locally grown/produced feedstocks, while others rely on a combination
of the two.  A summary of ethanol production alphabetically by state is found in Table 1.2-6.
                                           24

-------
                                      Table 1.2-6.
                         2006 U.S. Ethanol Production by State
State
California
Colorado
Georgia
Iowa
Illinois
Indiana
Kansas
Kentucky
Michigan
Minnesota
Missouri
North Dakota
Nebraska
New Mexico
Ohio
Oklahoma
South Dakota
Tennessee
Wisconsin
Wyoming
Total
Capacity
MMgy
71
93
0.4
1,618
706
122
219
38
155
546
155
51
606
30
3
2
493
67
233
12
5,218
%of
Capacity
1 .4%
1 .8%
0.0%
31 .0%
13.5%
2.3%
4.2%
0.7%
3.0%
10.5%
3.0%
1 .0%
1 1 .6%
0.6%
0.1%
0.0%
9.4%
1 .3%
4.5%
0.2%
100.0%
No. of
Plants
4
3
1
25
6
2
8
2
3
16
4
2
12
1
1
1
11
1
6
1
110
%of
Plants
3.6%
2.7%
0.9%
22.7%
5.5%
1 .8%
7.3%
1 .8%
2.7%
14.5%
3.6%
1 .8%
10.9%
0.9%
0.9%
0.9%
10.0%
0.9%
5.5%
0.9%
100.0%
       In addition to the domestic ethanol production described above, the U.S. also receives a
small amount of ethanol from other countries.  A discussion on ethanol imports is found in
Section 1.5

1.2.1.5    Ethanol Producers and Marketers

       The U.S. ethanol industry is currently comprised of a mixture of corporations and farmer-
owned cooperatives (co-ops). More than half (or 60) of the plants are owned by corporations
and the remainder (50 plants) are farmer owned co-ops. On average, a U.S. ethanol production
facility has a mean plant capacity of about 47 million gallons per year.  In general, plants owned
by corporations ("company-owned") are above average in size while farmer-owned co-ops are
below average.  Similarly, company-owned plants tend to have a much broader range in ethanol
production levels than farmer-owned co-ops. A summary of these results is presented in Table
1.2-7.
                                          25

-------
                                      Table 1.2-7.
                    2006 U.S. Ethanol Production by Plant Ownership
Plant
Ownership
Company-Owned3
Farmer-Owned
Total
No. of
Plants
60
50
110
Production Capacity, MMgy
Total
3,315
1,903
5,218
Avg
55
38
47
Min
0.4
3
0.4
Max
300
60
300
Includes ethanol producers with public offerings.
       Based on the dominating number of company-owned plants and their above-average
production size, company-owned plants account for nearly 64 percent of the total domestic
product.  Further, more than 50 percent of today's U.S. ethanol production capacity comes from
plants owned by just 6 different companies.  A list of the top six ethanol producing companies
and their respective plant capacities is found in Table 1.2-8.

                                      Table 1.2-8.
                          2006 Top Six U.S. Ethanol Producers
Company3
Archer Daniels Midland
Broin
VeraSun Energy
Hawkeye Renewables, LLC
Global / MGP Ingredients
Aventine Renewable Energy
Total
Capacity
MMgy
1,070
838
230
200
190
150
2,678
No. of
Plants
7
18
2
2
3
2
34
Includes majority and minority plant ownership.
       Over 80 percent of today's U.S. ethanol production is sold to the gasoline industry by
eight marketing companies4. A list of the top eight ethanol marketers and their respective
marketing capacities based on plant affiliations is found in Table 1.2-9.  The remaining ethanol is
marketed by Kinergy Marketing, The Andersons, Murex International, Noble Americas, and
other small marketing companies.
4 Based on information obtained from ethanol marketer websites, ethanol producer websites, and conversations with
ethanol marketers/producers.
                                           26

-------
                                       Table 1.2-9.
                         2006 Top Eight U.S. Ethanol Marketers

Marketing Company
Archer Daniels Midland
Ethanol Products
Renewable Products Marketing Group
Aventine Renewable Energy
Eco-Energy
Provista (formerly UBE)
Cargill, Inc.
Abengoa Bioenergy
Total
Capacity
MMgya
1,172
991
612
666
325
217
120
110
4,212
No. of
Plants
9
22
15
14
5
5
2
3
75
aVolumes based on marketing agreements and respective ethanol
plant capacities
1.2.2   Forecasted Growth in Ethanol Production

1.2.2.1    Overview

       Over the past 25 years, domestic fuel ethanol production has steadily increased due to
environmental regulation, federal and state tax incentives, and market demand.  More recently,
ethanol production has soared due to the phase out of MTBE, an increasing number of state
ethanol mandates, and elevated crude oil prices.  As shown in Figure 1.2-1, over the past three
years, domestic ethanol production has nearly doubled from 2.1 billion gallons in 2002 to 4.0
billion gallons in 2005. For 2006, the Renewable Fuels Association is anticipating about 4.7
billion gallons of domestic ethanol production5.
5 Based on RFA comments received in response to the proposed rulemaking, 71 FR 55552 (September 22, 2006).
                                           27

-------
                     Figure 1.2-1. U.S. Ethanol Production Over Time
4,500 -

"ra
c"
'o
3
•a
o
~o
as
HI


4,000


























2
2
1


nnnnfl

PI


-



-



-



—











& & <§> <§> <£> O,N <&
\j \j \j \j \j \j \j
PI




r-i





-

















-

















6C



1
0



7"



0



1;




0




81





0

































Year
       Source: Renewable Fuels Association, From Niche to Nation: Ethanol Industry Outlook 2006
       EPA forecasts that domestic ethanol production will continue to grow into the future.  In
addition to the past impacts of federal and state tax incentives, as well as the more recent impacts
of state ethanol mandates and the removal of MTBE from all U.S. gasoline, crude oil prices are
expected to continue to drive up demand for ethanol.  As a result, the nation is on track to exceed
the renewable fuel requirements contained in the Act, as explained below.

1.2.2.2     Expected Increases in Plant Capacity

       Today's ethanol production capacity (5.2 billion gallons) is already exceeding the 2007
renewable fuel requirement (4.7 billion gallons). In addition,  there is another 3.4 billion gallons
of production capacity currently under construct!on.6FGH A summary of the new construction
and plant expansion projects currently underway (as of October 2006) is found in Table 1.2-10.
6 Under construction plant locations, capacities, feedstocks, and energy sources as well as planned/proposed plant
locations and capacities were derived from a variety of data sources including Renewable Fuels Association (RFA),
Ethanol Producer Magazine (EPM), ICF International, BioFuels Journal, and ethanol producer websites.
                                             28

-------
           Table 1.2-10.  Under Construction U.S. Ethanol Production Capacity

PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5
Total
Oct. 2006 Baseline
MMgy
0.4
5,012
30
105
71
5,218
Plants
1
100
1
4
4
110
Under Const.
MMgya
115
2,764
230
50
198
3,357
Plants
1
39
3
1
3
47
Base + Under Const.
MMgya
115
7,776
260
155
269
8,575
Plants
2
139
4
5
7
157
Includes plant expansions
       A select group of builders, technology providers, and construction contractors are
completing the majority of the construction projects described in Table 1.2-10.  As such, the
completion dates of these projects are staggered over approximately 18 months, resulting in the
gradual phase-in of ethanol production shown in Figure 1.2-27.

         Figure 1.2-2. Estimated Phase-In of Under Construction Plant Capacity
                        Oct-06   Dec-06   Mar-07    Jun-07   Sep-07   Dec-07    Mar-08

                                                 Time
                                D Additional Capacity • Total Ethanol Rant Capacity
 Construction timelines based on information obtained from press releases and ethanol producer websites.
                                             29

-------
       As shown in Table 1.2-10 and Figure 1.2-2, once all the construction projects currently
underway are complete (estimated by March 2008), the resulting U.S. ethanol production
capacity would be about 8.6 billion gallons. Without even considering forecasted biodiesel
production (discussed below in 1.2.5), this would be more than enough renewable fuel to satisfy
the 2012 RFS requirements (7.5 billion gallons).  However, ethanol production is expected to
continue to grow. There are more and more ethanol projects being announced each day. These
potential projects are at various stages of planning from conducting feasibility studies to gaining
local approval to applying for permits to financing/fundraising to obtaining contractor
agreements. Together these potential projects could result in an additional 21 billion gallons of
ethanol production capacity (as shown in Table 1.2-11).

                                      Table 1.2-11.
                     Other Potential U.S. Ethanol Production Capacity

PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5
Subtotal
Total"
Base + Under Const.
MMgya
115
7,776
260
155
269
8,575

Plants
2
139
4
5
7
157

Planned
MMgya
548.0
4,633
250
100
232
5,763
14,339
Plants
8
44
4
1
8
65
222
Proposed
MMgya
934
1 1 ,722
876
783
775
15,090
29,428
Plants
21
136
14
14
23
208
430
Includes plant expansions
bTotal including existing plus under construction plants.
       Although there is clearly a great potential for ethanol production growth, it is highly
unlikely that all the announced projects would actually reach completion in a reasonable amount
of time, or at all, considering the large number of projects moving forward. Since there is no
precise way to know exactly which plants will come to fruition in the future, we have chosen to
focus our subsequent discussion on forecasted ethanol production on plants which are likely to
be online by 2012.8 This includes existing plants as well as projects which are under
construction (refer to Table 1.2-10) or in the final planning stages (denoted as "planned" in Table
1.2-11).  The distinction between "planned" versus "proposed" is that as of October 2006
planned projects had completed permitting, fundraising/financing, and had builders assigned
with definitive construction timelines whereas proposed projects did not.

       As shown in Table 1.2-11, once all the under construction and planned projects are
complete, the resulting U.S. ethanol production capacity would be 14.3 billion gallons.  This
volume, expected to be online by 2012, exceeds the EIA AEO 2006 demand estimate (9.6 billion
gallons by 2012, discussed more in RIA Section 2.1.4.1). The forecasted growth would nearly
triple today's production capacity and greatly exceed the 2012 RFS requirement (7.5 billion
8 A more detailed summary of the plants we considered is found in a March 5, 2007 note to the docket titled: RFS
Industry Characterization - Ethanol Production.
                                            30

-------
gallons). While our forecast represents ethanol production capacity (actual production could be
lower), we believe it is still a good indicator or what domestic ethanol production could look like
in the future.  In addition, we predict that domestic ethanol production will continue to be
supplemented by imports in the future. A more detailed discussion on future ethanol imports is
found in Section 1.5.

1.2.2.3    Changes in Feedstocks & Processing Technologies

       Of the 112 forecasted new ethanol plants (47 under construction and 65 planned), 106
would rely on grain-based feedstocks. More specifically, 89 would rely exclusively on corn, 13
would process a blend of corn and/or similarly processed grains (milo or wheat), 3 would process
molasses,  and 1 would process a combination  of molasses and sweet sorghum (milo). Of the
remaining six plants (all in the planned stage), four would process cellulosic biomass feedstocks
and two would start off processing corn and later transition to cellulosic materials. Of the four
dedicated  cellulosic plants, one would process bagasse, one would process a combination of
bagasse and wood, and two would process biomass. Of the two transitional corn/cellulosic
plants, one would ultimately process a combination of bagasse, rice hulls, and wood and the
other would ultimately process wood and other agricultural residues.  In addition to the
forecasted new plants described above, an existing corn ethanol plant plans to expand production
and transition to corn stalks,  switchgrass, and biomass in the future.

       A summary of the resulting overall feedstock usage (including current, under
construction,  and planned projects) is found in Table 1.2-12. A discussion on how the plants
predicted to process cellulosic feedstocks would help the nation meet the Act's cellulosic
biomass ethanol requirement is found in Section 1.2.2.6

-------
           Table 1.2-12.  Forecasted 2012 U.S. Ethanol Production by Feedstock

Plant Feedstock
Bagasse
Bagasse, Wood
Bagasse, Wood, Rice Hulls3
Biomass
Cheese Whey
Cornb
Corn, Barley
Corn, Miloc
Corn, Wheat
Corn Stalks, Switchgrass, Biomass3
Milo, Wheat
Molassesd
Sugars, Starches
Waste Beverages6
Wood Agricultural Residues3
Total
Capacity
MMgy
7
2
108
55
8
12,495
40
1,132
235
40
40
52
2
16
108
14,339
%of
Capacity
0.1%
0.0%
0.8%
0.4%
0.1%
87.1%
0.3%
7.9%
1 .6%
0.3%
0.3%
0.4%
0.0%
0.1%
0.8%
100.0%
No. of
Plants
1
1
1
2
2
178
1
20
3
1
1
4
1
5
1
222
%of
Plants
0.5%
0.5%
0.5%
0.9%
0.9%
80.2%
0.5%
9.0%
1 .4%
0.5%
0.5%
1 .8%
0.5%
2.3%
0.5%
100.0%
facilities plan to start off processing corn.
Includes two facilities processing seed corn.
clncludes one facility processing small amounts of molasses in addition to corn and milo.
Includes one facility planning to process sweet sorghum (milo) in addition to molasses.
Includes two facilities processing brewery waste.
       As shown above, the majority of future plants are predicted to process grains (namely
corn).  Similarly, the vast majority of plants are expected to pursue dry milling technology. Our
analysis does not foresee any new wet mill facilities, with the exception of a new 100 MMgy wet
mill plant that is planned for Fort Dodge, IA and a 37 MMgy plant expansion project that is
underway in Loudon, TN.  Further, we do not predict that there will be any new plants
processing cheese whey, waste beverages, or sugars/starches (which do not require milling).
The forecasted cellulosic feedstock plants (described in more detail in Section 1.2.2.7) will not
require milling. However, these facilities will require complex forms of pretreatment (described
in more detail in Section 7.1.2) to break down the lignocellulosic and hemicellulosic polymers
into fermentable sugars. A summary of the resulting overall feedstock processing technology
utilization is found below in Table 1.2-13.
                                           32

-------
                                      Table 1.2-13.
           Forecasted 2012 U.S. Ethanol Production by Processing Technology
Processing
Technology
Dry Milling
Wet Milling
Other3
Total
Capacity
MMgy
12,668
1,274
397
14,339
%of
Capacity
88.3%
8.9%
2.8%
100.0%
No. of
Plants
192
11
19
222
%of
Plants
86.5%
5.0%
8.6%
100.0%
aPlants that do not process traditional grain-based crops and thus do not require milling.
This category includes plants processing cheese whey, sugars & starches, or waste
beverages as well as plants that plan to process molasses or cellulosic feedstocks.
1.2.2.4    Changes in Plant Energy Sources

       Of the 112 forecasted new plants, 100 would burn some amount of natural gas - at least
initially.  More specifically, 91 plants would rely exclusively on natural gas; two would rely on a
combination of natural gas, bran and biomass; one would burn a combination of natural gas,
distillers' grains and syrup; and six would start off burning natural gas and later transition to
coal. As for the remaining 12 plants, three would burn manure-derived methane (biogas); seven
would rely exclusively on coal; one would burn a combination of coal and biomass; and one
would burn a combination  of coal, tires and biomass. In addition to the new ethanol plants, three
existing plants currently burning natural gas are predicted to transition to alternate boiler fuels in
the future. More specifically, two plants plan to transition to biomass and one plans to start
burning coal.

       Our research suggests that seven of the new plants (mentioned above) would utilize
combined heat and power (CHP) technology, although others may exist. Three of the new CHP
plants would burn natural gas, three would burn coal, and one would  burn a combination of coal,
tires, and biomass.  Among the existing CHP plants, two are predicted to transition from natural
gas to coal or biomass at this time. Overall, the net number of CHP ethanol plants would
increase from 11 to 18. A  summary of the resulting overall plant energy source utilization is
found below in Table 1.2-14. A discussion on how the plants predicted to burn waste materials
could help the nation meet  the Act's cellulosic biomass ethanol requirement is found in Section
1.2.2.6.
                                           33

-------
        Table 1.2-14. Forecasted 2012 U.S. Ethanol Production by Energy Source

Plant Energy Source
Biomass3
Coalb
Coal, Biomass
Coal, Biomass, Tires
Manure Biogasc
Natural Gas
Natural Gas, Bran, Biomass
Natural Gas, Distillers' Grain, Syrup
Syrup
Total
Capacity
MMgy
112
2,095
75
275
144
11,275
264
50
49
14,339
%of
Capacity
0.8%
14.6%
0.5%
1 .9%
1 .0%
78.6%
1 .8%
0.3%
0.3%
100.0%
No. of
Plants
2
21
2
1
3
189
2
1
1
222
%of
Plants
0.9%
9.5%
0.9%
0.5%
1 .4%
85.1%
0.9%
0.5%
0.5%
100.0%
CHP
Tech.
1
6
0
1
0
10
0
0
0
18
aRepresents two existing natural gas-fired plants that plan to transition to biomass.
blncludes two plants planning on burning lignite coal or coal fines. Includes one existing plant currently
burning natural gas that plans to transition to coal. Includes six new plants that will start off burning
natural gas and later transition to coal.
clncludes one facility planning on burning cotton gin in addition to manure biogas.
1.2.2.5    Changes in Ethanol Production Locations

       Once all the forecasted ethanol projects are complete, 87 percent of the domestic
production capacity would originate from PADD 2, followed by PADDs  1, 3, 5, and 4 (all
contributing less than 5 percent). A summary of the findings is found below in Table 1.2-15.

                                      Table 1.2-15.
                   Forecasted 2012 U.S. Ethanol Production by PADD
PADD
PADD1
PADD 2
PADDS
PADD 4
PADDS
Total
Capacity
MMgy
663
12,409
510
255
501
14,339
%of
Capacity
4.6%
86.5%
3.6%
1 .8%
3.5%
100.0%
No. of
Plants
10
183
8
6
15
222
%of
Plants
4.5%
82.4%
3.6%
2.7%
6.8%
100.0%
       While PADD 2 ethanol production is expected to more than double (from 5.0 to 12.4
billion gallons), this represents a decrease in Midwest marketshare (from 96 to 87 percent). This
predicted shift in marketshare is attributed to the growing number of ethanol plants located
outside the cornbelt. Arizona, Florida, Hawaii, Louisiana, New York, Oregon, Pennsylvania and
Texas are scheduled to join the 19 ethanol producing states described in Table 1.2-5. A
summary of future ethanol production by state is found below in Table 1.2-16.
                                          34

-------
                                      Table 1.2-16.
                    Forecasted 2012 U.S. Ethanol Production by State
State
Arizona
California
Colorado
Florida
Georgia
Hawaii
Iowa
Illinois
Indiana
Kansas
Kentucky
Louisiana
Michigan
Minnesota
Missouri
New York
North Dakota
Nebraska
New Mexico
Ohio
Oklahoma
Oregon
Pennsylvania
South Dakota
Tennessee
Texas
Wisconsin
Wyoming
Total
Capacity
MMgy
55
244
243
80.0
150.4
59.2
3,016
1,606
855
569
38
110
212
882
382
325
251
2,543
30
420
112
143
108
953
109
370
463
12
14,339
%of
Capacity
0.4%
1 .7%
1 .7%
0.6%
1 .0%
0.4%
21 .0%
1 1 .2%
6.0%
4.0%
0.3%
0.8%
1 .5%
6.2%
2.7%
2.3%
1 .7%
17.7%
0.2%
2.9%
0.8%
1 .0%
0.8%
6.6%
0.8%
2.6%
3.2%
0.1%
100.0%
No. of
Plants
1
7
5
2
3
5
38
16
11
13
2
2
4
20
6
4
5
31
1
7
3
2
1
16
2
5
9
1
222
%of
Plants
0.5%
3.2%
2.3%
0.9%
1 .4%
2.3%
17.1%
7.2%
5.0%
5.9%
0.9%
0.9%
1 .8%
9.0%
2.7%
1 .8%
2.3%
14.0%
0.5%
3.2%
1 .4%
0.9%
0.5%
7.2%
0.9%
2.3%
4.1%
0.5%
100.0%
1.2.2.6    Meeting the Cellulosic Ethanol Requirement in 2013

       The Energy Policy Act of 2005 (the Energy Act or the Act) requires that 250 million
gallons of the renewable fuel consumed in 2013 and beyond meet the definition of cellulosic
biomass ethanol. The Act defines cellulosic biomass ethanol as ethanol derived from any
lignocellulosic or hemicellulosic matter that is available on a renewable or recurring basis
including dedicated energy crops and trees, wood and wood residues, plants, grasses, agricultural
residues, fibers, animal wastes and other waste materials, and municipal solid waste. The term
also includes any ethanol produced in facilities where animal or other waste materials are
                                           35

-------
digested or otherwise used to displace 90 percent of more of the fossil fuel normally used in the
production of ethanol.

       As discussed above in Section 1.2.2.3, there are seven "planned" ethanol plants planning
on processing cellulosic feedstocks in the future.  A summary of these facilities is found below in
Table 1.2.17.
                    Table 1.2-17.  Potential Cellulosic Feedstock Plants

Ethanol Plant
Worldwide Energy Group3
Celunol Corp.b
GS Agrifuels Corporation0
Xethanol Coastal LLC
Bionol
Xethanol Corporation
BioEnergy International

Location
Kaumakani, HI
Jennings, LA
Memphis, TN
Augusta, GA
Lake Providence, LA
Blairstown, IA
Clearfield County, PA

Plant Feedstock
Bagasse
Bagasse, Wood
Biomass
Biomass
Corn then Bagasse, Rice Hulls, Wood
Corn then Corn Stalks, Switch Grass, Biomass
Corn then Wood, Agricultural Residues
Total Cellulosic Ethanol Potential Based on Plant Feedstocks
Capacity
MMgy
7
2
5
50
108
40
108

Status
Planned
Planned
Planned
Planned
Planned
Planned"
Planned
320
aCompany also/formerly known as Clearfuels Technology
bCompany also/formerly known as BC International
°Project also/formerly known as Mean Green Biofuels
dlncludes 5 Mmqy existing plant capacity plus 35 MMqy planned expansion.
       It is unclear whether the above-mentioned cellulosic feedstock plants would be online
and capable of producing 250 million gallons of ethanol by 2013 to meet the Act's cellulosic
biomass ethanol requirement. However, as described above in Section 1.2.2.4 there are 12
facilities that burn or plan to burn waste materials to power their ethanol plants in the future.
These facilities, summarized below in Table  1.2.18, could also potentially meet the definition of
cellulosic biomass ethanol under the Act.

                       Table 1.2-18.  Potential Waste Energy Plants

Ethanol Plant
Corn LP
E Caruso Ethanol
Archer Daniels Midland
E3 Biofuels, LLC
Harrison Ethanol, LLC
Panda Ethanol
Central Minnesota Ethanol Co-op
Chippewa Valley Ethanol Co.
Ethanex at SEMO Port
Ethanex Southern lllinoisb
Green Plains Renewable Energy0
Corn Plus, LLP

Location
Goldfield, IA
Goodland, KS
Columbus, NE
Mead, NE
Cadiz, OH
Hereford, TX
Little Falls, MN
Benson, MN
Cape Girardeau, MO
Benton, IL
Superior, IA
Winnebago, MN

Plant Energy Source
Coal, Biomass
Coal, Biomass
Coal, Tires, Biomass
Manure Biogas
Manure Biogas
Manure Biogas, Cotton Gin
Natural Gas then Biomass
Natural Gas then Biomass
Natural Gas, Bran, Biomass
Natural Gas, Bran, Biomass
Natural Gas, Distillers Grain, Syrup
Syrup
Total Cellulosic Ethanol Potential Based on Plant Energy Sources
Capacity
MMgy
50
25
275
24
20
100
22
90
132
132
50
49

Status
Existing
Under Construction
Planned
Under Construction
Planned
Under Construction
Existing
Existing3
Planned
Planned
Under Construction
Existing
969
alncludes 45 MMgy existing plant capacity plus 45 MMgy planned expansion.
bJoint venture with Star Ethanol
°Project also/formerly known as Superior Ethanol
                                            36

-------
       Depending on how much fossil fuel is displaced by burning these waste materials (on a
plant-by-plant basis), a portion or all of the above-mentioned 969 MMgy ethanol production
capacity could potentially qualify as "cellulosic biomass ethanol" under the Act. Combined with
the additional 320 MMgy of ethanol production capacity from plants processing cellulosic
feedstocks, the overall cellulosic ethanol potential could be as high as 1.3 billion gallons. Even
if only one fifth of this ethanol were to end up qualifying as cellulosic biomass ethanol or come
to fruition by 2013, it would be more than enough to satisfy the 250 million gallon requirement
specified in the Act.9

1.2.3  Current Biodiesel Production

       Biodiesel is defined in several  sections of the Act, which we have used in formulating our
definition for the regulations, which call for meeting ASTM specifications. Biodiesel is
registered with the EPA for commercial sale and is legal for use at any blend level in both
highway and nonroad diesel engines although most engine manufacturers will only honor the
warranty if biodiesel is used in blends of 2, 5 or, in some limited circumstances, 20 percent.

       Biodiesel can be made from almost any vegetable or animal fat, with most of the world's
production coming from plants oils, notably soy bean and rapeseed (canola) oil.  Biodiesel fuel
production is rapidly increasing in many regions of the world. The choice of the feedstock oil
used to make it is dependent upon the vegetable oils and fat supplies that are economically
available.   For the U.S. market, there  are many potential plant oil feedstocks that can be used to
make biodiesel, including soybean,  peanut, canola, cottonseed and corn oil. Biodiesel can also
be made from animal fats such used restaurant grease (yellow grease) and tallow. Though,
typically for the U.S. market, soybean oil has been the primary major feed stock supply,
followed by use of yellow grease and animal tallow.

       The resulting biodiesel product can be used as a fuel for diesel engines with minor
modifications and is commonly blended with refinery produced diesel fuel. Raw vegetable and
animal oils consist of fatty acids  and glycerine products. Though these oils can directly be used
in engines and give good short term performance, this is highly discouraged as their use can
cause severe engine problems. This is primarily due to the raw oils forming engines deposits,
with coking and  plugging in engine injectors nozzles, piston rings, lubricating oil, etc.  This
happens due to polymerization of the triglycerides in the raw oils as the fuel is combusted.
Therefore, it is necessary to convert the raw oils into a form of esters or biodiesel which prevents
these issues.  The biodiesel production process converts the raw vegetable and animal oils into
esters, though the virgin oils themselves are sometimes (inappropriately) referred to as biodiesel.
The production process called transesterification consists of adding methanol or ethanol to the
virgin vegetable  oil and animal oil, in the presence of a catalyst such as sodium or potassium
hydroxide, resulting in esters or biodiesel and a byproduct glycerol. A subsequent step is usually
9 We anticipate a ramp-up in cellulosic ethanol production in the years to come so that capacity exists to satisfy the
Act's 2013 requirement (250 million gallons of cellulosic biomass ethanol). Therefore, for subsequent analysis
purposes, we have assumed that 250 million gallons of ethanol would come from cellulosic biomass sources by
2012.
                                            37

-------
needed, however, to remove glycerin, catalysts and other compounds, to allow the biodiesel to
meet the required ASTM specifications.

       Biodiesel blends such as B2, B5 and in some cases B20, can be used in existing engines
without modification, and most engines exhibit no performance problems with the use of
biodiesel, though this depends on the blend and the season. However, engine fuel filters may
need to be changed more often, and there may be cold temperature operations due to biodiesel's
higher cloud point. As a result most engine manufacturers will only recognize their warranties if
biodiesel is used in low concentrations. Biodiesel produced from vegetable oil has practically
zero amounts of sulfur and aromatics and a high cetane value, thus making it a good for blending
into 15 ppm highway and offroad diesel fuel, though biodiesel made from yellow grease and
animal fat may contain about 24 ppm of sulfur1.  Biodiesel also has good lubricity qualities and
can be used in concentration (~2 vol%) as a lubricity-enhancing additive for conventional diesel.

1.2.4   Forecasted Biodiesel Production

       Biodiesel production has been increasing rapidly over the past five years and is projected
to continue at a high rate in part because of the Renewable Fuel Standard (RFS)  program. This
expansion has primarily been driven by better economics, due to the recent large increase in
diesel prices associated with the run up in crude prices, along with the Biodiesel Blenders Tax
Credit programs and the Commodity Credit Commission Bio-energy Program, both of which
subsidize producers and offset production costs. The Act extended the Biodiesel Blenders Tax
Credit program to year 2008, which provides about one dollar per gallon in the form of a federal
excise tax credit to biodiesel blenders from virgin vegetable oil feedstocks and 50 cents per
gallon to biodiesel produced from recycled grease and animal fats. This program was started in
2004 under the American Jobs Act. The existing Commodity Credit Commission Bio-energy
Program also pays biodiesel producers grants when the economics to produce biodiesel are poor;
the  program averaged about one dollar per gallon in 2004. Recent payments  through the
Commodity Credit program have  been reduced, however, and the program is expiring in fiscal
year 2006.  Historically, the cost to make biodiesel was an inhibiting factor to production.  The
cost to produce biodiesel was high compared to the price of petroleum derived diesel fuel, even
with consideration of the benefits of subsidies and credits provided by federal and state
programs. Mandates from states and local municipalities that require the use of biodiesel in
transport fuels are another factor which is expanding the use of biodiesel.

       In 2005 approximately 91  million gallons of biodiesel were produced in the U.S. based
on program payments to biodiesel producers under USDA's Bio-energy Program. This volume
represents approximately 0.15 percent of all diesel fuel consumed in the domestic market. EIA
projects the future production volume to expand to 414 million gallons per year in 2007 and then
decrease to about 303 MM gallons per year in 2012, assuming that the biodiesel blender tax
credits program expires in 2007 (see Table 1.2-19).
                                           38

-------
                       Table 1.2-19.  Estimated Biodiesel Production3
Year
2001
2002
2003
2004
2005
2006
2007
2012
Million Gallons per Year
5
15
20
25
91
150
414
303
       a Historical data from 2001-2004 obtained from estimates from John Baize " The Outlook and Impact of
Biodiesel on the Oilseeds Sector" USDA Outlook Conference 06. Year 2005 data from USDA Bioenergy Program.
Year 2006 data from verbal quote based on projection by NBB in June of 06. Production data for years 2007 and
higher are from EIA's AEO 2006.
       With the increase in biodiesel production, there has also been a corresponding rapid
expansion in biodiesel production capacity.  Presently, there are 85 biodiesel plants in operation
with an annual production capacity of 580 million gallons per yearj.  The majority of the current
production capacity was built in 2005 and 2006, and was first available to produce fuel in the
later part of 2005 and in 2006.  Though the capacity has grown, historically the biodiesel
production capacity has far exceeded actual production with only 10-30 percent of this being
utilized to make biodiesel, see Table 1.2-20.

                      Table 1.2-20. U.S. Production Capacity History

Plants
Capacity
(MM gals/yr)
Production,
(MM gals/yr)
Capacity
Utilization for
Biodiesel, %
2001
9
50
5
10

2002
11
54
15
28

2003
16
85
20
24

2004
22
157
25
16

2005
45
290
91
31

2006
85
580
150
26

       Note: Capacity Data based on surveys conducted.
       Excess production capacity is not easily quantified, though since some of these plants
may not run at full rate all of the time and may be "idled" for certain days of the week, seasons,
                                            39

-------
time of day, etc. The capacity can be classified into two types of producers; capacity dedicated
to biodiesel production and capacity available from the ole-chemical industry. The plants that
primarily operate in the ole-chemical industry produce esters for use in the chemical industry.
These plants are swing producers of biodiesel, which means that when the economics are
favorable they can shift their operations and make biodiesel esters instead of products for the ole-
chemical market.10 The capacity from the ole-chemical industry produces mono-alkyl esters
using a similar transesterification process, with the ester products being sold for to make
plasticizers, soaps, paints, solvents and other industrial uses.  Additionally, the biodiesel
production capacity volumes may be optimistic, as this is not officially tracked.  The capacities
listed here are those based on each company's self reported volumes to the National Biodiesel
Board and may have some inaccuracies due to informal reporting procedures.

       We anticipate that future capacity additions will be geared more towards production of
biodiesel for use as transportation fuel, rather than serving primarily the oleochemicals markets.
As of September 2006, there were 65 plants in the construction phase and 13 existing plants that
are expanding their capacity. All of this new capacity when installed would provide about 1.4
billion gallons per year of additional throughput capacity.  Table 1.2-21 presents the data for the
biodiesel plant capacities per the categories discussed.

                         Table 1.2-21.  Biodiesel Plant Capacities

Number of Plants
Total Plant Capacity,
(MM Gallon/year)
Existing Plants
85
580
Construction Phase
78
1,400
       Considering that it takes 12 to 18 months to construct a biodiesel plant (from the time of
project feasibility analysis to startup date), a large portion of the capacity in the construction
phase in late 2006 will be available to produce fuel in 2007.K  Data on biodiesel plant
construction reveals most of the new capacity that is currently being constructed is expected to
be online and producing fuel in 2006 or by end of 2007. Therefore, the existing capacity plus the
capacity in the construction phase totals an aggregate amount of about two billion gallons per
year.  Though there is no volume mandate for biodiesel fuel under the RFS program, the total
capacity available from new and existing plants exceeds EIA's projected biodiesel volume of
414 MM in 2007 and 303 MM in 2012 by a wide margin.

       The plants in the construction phase are larger than existing biodiesel plants, with average
capacity of existing plants at 8.4 MM gallons per year, while plants in construction phase are
10 Oleochemicals are derived from biological fats and oils using hydrolysis or alcoholysis with products of fatty acid
esters and glycerol.
                                            40

-------
averaging 20.9 MM gallons per year, as presented in Table 1.2-22. The distribution of biodiesel
plants by size and number of companies within each size range are presented in Table 1.2-23.
        Table 1.2-22.  Average Plant Capacity by Feedstock (MM gallons per year)
Feedstock
Canola
Multi Feedstock
Other Vegetable
Recycled Cooking Oil
Soybean Oil
Tallow
Existing*

6.0
2.0
0.5
8.8
5.0
Construction*
57.5
16.7

1.0
19.3

              Table 1.2-23. Biodiesel Plant Size versus Number of Companies
Plant Size
(MM gallons per year)a
<1.00
1.0-5.0
5.0-10.0
10.0 to 15 .0
15.0 to 20.0
20.0+
Average Plant Size
Existing Plants
9
28
17
9
2
10
8.4
Construction Phase
5
9
10
7
3
28
20.9
       aTotal capacity of plants in each category; existing plants are 580 MM gal/yr while those in the construction
phase are 1,400 MM gal/yr.
       Because newer plants are likely to be larger than existing plants, have better technology
and may have greater alignment with feedstock and feed sources, some of the older plants may
operate at an economic disadvantage once the new plants come on line. At the moment, it is not
possible to predict actual biodiesel production based on capacity, since in the past the capacity
was used at rates less than maximum.  Thus, how excess production capacity evolves will be
dictated by economics, profitability, and fuel demand.

       The majority of existing biodiesel plant capacity is located in the middle and Midwestern
parts of the country and use soy bean oil as the feedstock. The other plants are scattered with
locations based on the east and west coasts,  with feedstocks based on use of soybean, canola and
other oils as well as yellow grease as the feedstock. The new plants are being built to process a
                                           41

-------
wider variety of feedstocks, with multi-feedstock and recycle grease capability.  The feedstocks
for these plants are listed in Table 1.2-24.
                Table 1.2-24. Feedstock Selection for Biodiesel Producers
Feedstock
Camelia
Canola
Cottonseed
Multi Feedstock
Palm Oil
Recycled Cooking Oil
Soybean oil
Tallow/Poultry Fat
Unknown
Existing


1
29

7
39
2
7
Construction

2

29

O
36

8
1.2.5   Baseline and Projected Biodiesel Volumes for Analysis

       For cost and emission analysis purposes, three biodiesel usage cases were considered: a 2004
base case, a 2012 reference case, and a 2012 control case. The 2004 base case was formed based on
historical biodiesel usage (25 million gallons as summarized in Table  1.2-16). The reference case
was computed by taking the 2004 base case and growing it out to 2012 in a manner consistent with
the growth of diesel fuel (described in Section 2.1.3). The resulting 2012 reference case consisted of
approximately 30 million gallons of biodiesel. Finally, for the 2012 control case, forecasted
biodiesel use was assumed to be 300 million gallons based on EIA's AEO 2006 report (rounded
value from Table 1.2-16).  Unlike forecasted ethanol use (described in 2.1.4), biodiesel use was
assumed to be constant at 300 million gallons under both the statutory and higher projected
renewable fuel consumption scenarios.
1.3   Renewable Fuel Distribution

1.3.1   Current Renewable Fuel Distribution System

       Ethanol and biodiesel blended fuels are not currently shipped by petroleum product
pipeline due to operational issues and additional cost factors.L  The ability to ship by pipeline is
also limited because the sources of ethanol and biodiesel are frequently not in the same locations
as the sources of gasoline and petroleum-based diesel fuel. Hence, a separate distribution system
is needed for ethanol and biodiesel up to the point where they are blended into petroleum-based
fuel as it is loaded into tank trucks for delivery to retail and fleet operators.  Ethanol and
                                           42

-------
biodiesel can either be added by "splash blending" where the renewable is added separately to
the tank truck, or by in-line injection where the renewable is injected into the petroleum fuel
stream as it is being dispensed into the tank truck. Ethanol and biodiesel are sometimes added to
petroleum-based fuels downstream of the terminal, but this accounts for little of the total volume
of used.

       In cases where ethanol and biodiesel are produced within 200 miles of a terminal,
trucking is often the preferred means of distribution. However, most renewable fuel volumes are
produced at greater distances from potential centers of demand. For longer shipping distances,
the preferred method of bringing renewable fuels to terminals is by rail and barge.  Dedicated
pipelines have not been used to distribute renewable fuels to terminals due to the high cost of
installing new pipelines, the relatively large shipping volumes that would be needed to justify
such expenditures, and the fact that renewable fuel production facilities tend to be relatively
numerous and dispersed.

       The relatively low volumes of ethanol used prior to 2002 constrained the ability of the
distribution system to efficiently move ethanol to distant markets. Ethanol  shipments by rail
were typically made on an individual car basis. Under such an approach, small  groups of rail
cars travel to market as part of trains that carry other goods. This approach results in relatively
high transportation costs, longer transit times, and potential delays in delivery.   Substantial
improvements in the efficiency of distributing ethanol by rail are being made due the need to
move large volumes of ethanol over long distances as a consequence of the elimination of MTBE
in California, New York, and Connecticut beginning in  2004.  The use of unit trains, sometimes
referred to as "virtual pipelines" reduces delivery costs, shortens delivery times, and improves
reliability. Unit trains are composed entirely of approximately 100 rail cars containing ethanol.
Ethanol shipped by unit trains is delivered to hub terminals for further distribution to other
terminals by barge and tank truck.

       Substantial volumes of ethanol can potentially be shipped down the Mississippi river by
barge for temporary storage in New Orleans.M  From New Orleans, ethanol can be loaded onto
ocean transport for delivery to the East  and West Coast. There is also potential  to move ethanol
via the Missouri and Ohio as well as other river systems and the Great Lakes. Marine shipments
of ethanol require a relatively large minimum shipment size, determined by the  minimum size of
the marine tank compartment.u  Similar to the case for "unit trains", there are also efficiencies in
dedicating whole barges, barge tows, or marine tankers to ethanol distribution.   The increased
demand for ethanol has made it possible to better benefit from these efficiencies of scale.

       The use of inland barges to transport ethanol from production facilities is in large part
driven by whether there is river access at such facilities. Historically, corn prices tend to be
higher near river systems that serve as arteries for the export of corn than at inland locations
distant from these river systems.  To take  advantage of lower corn prices at inland locations and
to avoid competing for corn with grain elevators that serve the export market, all of the new
ethanol production facilities that have been built since 1999 have been built at inland locations. N
1J River barges typically have a capacity of 10,000 barrels. Ocean barges typically have a capacity of 20,000
barrels. Barges are sometimes subdivided into 2 or 3 compartments.
                                            43

-------
Consequently, the majority of the growth in ethanol freight volumes since 1999 has been in the
rail sector.

1.3.2   Changes to the Renewable Fuel Distribution System Due to Increased Demand

       This section addresses that changes that we expect will take place in the renewable fuel
distribution system in response to the anticipated increase in demand for such fuels through
2012. There may be  some limited opportunity to ship renewable fuels by pipeline in the future
as demand  increases. However, because of the constraints discussed previously (see section
1.3.1), we believe that rail and barge are likely to remain the predominant means of
transportation. The 2002 DOE Study also reached this conclusion.0 While this constraint on the
ability to ship ethanol and biodiesel by pipeline presents logistical challenges that result in
additional transportation costs,  the need to transport these alternative fuels by other means may
work to the overall advantage of the fuel distribution system.  Petroleum product pipelines are
nearing capacity. Thus, it seems likely that the pipeline distribution system will find it
increasingly difficult to keep pace with annual increases in the demand for transportation fuels.
Displacing  some of the volume of transportation fuels from the pipeline distribution system
through the use of ethanol and biodiesel will relieve some of this strain.

       Small volume rail shipments made on a by-car basis are likely to remain an important
feature in supplying markets that demand limited volumes.  However, as the demand for ethanol
increases we anticipate that the expansion of the use of unit trains will continue, and that this will
be a significant means of bringing ethanol to distant markets.  There has been some expansion of
capacity at  existing ethanol plants with river access and some  new plants are projected to be built
with river access. However, we anticipate that most new ethanol capacity will not have river
access.  In addition, at least one new ethanol plant slated for production that does have river
access is planning to  move its ethanol to market via rail. Nevertheless, in cases where rail is  the
means to transporting ethanol to hub terminals, marine transport can play an important role in
further distribution to satellite terminals.

       Substantial improvements to the rail, barge, tank truck, and terminal distribution systems
will be needed to support the transport of the volumes of renewable fuels necessary to meet the
requirements of the RFS program. These improvements include the addition of a significant
number of additional rail cars, and tank trucks. Additional marine barges will also be needed.
To facilitate the increased use of unit trains, new rail spurs will be needed at terminals.
Terminals will also need to add facilities to store and blend ethanol. In addition, those terminals
and retail facilities that had not previously handled ethanol blended fuel will need to make
certain one-time upgrades to ensure the compatibility of their  systems with ethanol. These types
of changes  have been occurring as demand for ethanol and biodiesel has grown rapidly over the
last several years, and there is no reason to suspect that they would not continue as demand
continues to warrant it. The costs associated with these changes are discussed in Chapter 7.3 of
this RIA.

       The most comprehensive study of the infrastructure requirements for an expanded fuel
ethanol  industry was  conducted for the Department of Energy (DOE) in 2002 .p The conclusions
reached in this study  indicate that the changes needed to handle the increased volume of ethanol
                                           44

-------
required under the RFS will not represent a major obstacle to industry.12 While some changes
have taken place since this report was issued (as discussed below), we continue to believe that
the rail and marine transportation industries  can manage the increased growth in an orderly
fashion. This belief is supported by the demonstrated ability of the industry to handle the rapid
increases and redistribution of ethanol use across the country over the last several years as
MTBE was removed.  Given that future growth in ethanol use is expected to take place in an
orderly fashion in response to economic drivers, we anticipate that the distribution system will be
able to respond appropriately.

       The use of unit trains has accelerated beyond that anticipated in the 2002 DOE report,
leading to the more efficient distribution of ethanol by rail.  As a result, rail has taken a relatively
greater role in the transportation of new ethanol volumes as compared to shipment by barge than
was projected in the report. Thus, there is likely to be a relatively greater demand on the rail
distribution system and somewhat less demand on the marine distribution system than was
projected in the DOE study.

       The 2002 DOE study estimated that the increase in the volume of ethanol shipped by rail
needed to  facilitate the use of 10 billion gallons of ethanol annually would represent an increase
in total tank car loadings of 0.33 percent.  The increase in tank car loadings for Class I railroads
was estimated at 4.75 percent.  The DOE report concluded that this increase is relatively modest
by railroad industry standards and could be accommodated given the  available lead time. The
DOE study estimated that the increase in demand on barge movements due to the need to carry
an increased volume of ethanol would equate to a one percent increase in the total tonnage
moved by barge.  Given that on the one hand relatively few new ethanol plants are projected to
be cited with river access, and that on the other hand barge is expected to play an important role
in redistributing ethanol from rail hub terminals, we estimate that the  increase in barge
movements will be 30 percent less than that projected in the 2002 DOE study. This equates to an
increase in total tank car loadings of 0.44 percent rather than the 0.33  percent projected in the
DOE study.  We believe that this relatively modest potential increase  in the demand on the rail
distribution system can be accommodated without major  difficulty given the available lead time.

       Although, the 2002 DOE study generally concluded that the projected one percent
increase in the demand on the river barge industry could be accommodated without major
difficulty, it highlighted two potential concerns.  The report noted that delays are already being
experienced at locks on the Mississippi river.  The question was raised regarding how the
projected increase of one percent in river traffic due to increased ethanol shipments might be
accommodated at these locks. The report also raised concerns regarding the availability of
sufficient marine vessels capable of traveling between two ports in the United States (Jones Act
compliant vessels).  Given that it appears that there will be less demand placed on the river barge
industry to transport ethanol than was projected in the 2002 DOE study, the concerns raised in
the study regarding the capability of the inland waterway system to cope with the increased
traffic associated with shipping the anticipated new volume of ethanol will be less pronounced.
12 See section 7.3 of this RIA regarding the projected costs of the necessary infrastructure improvements.
                                           45

-------
       At the present time, the industry is experiencing a shortage of tractor trailers and drivers
to transport ethanol.  The boom in demand for truck transport is due to a number of factors,
including the precipitous removal of MTBE from gasoline and its replacement by ethanol13
which has taken place when the demand for truck transport was already growing at a rapid place
due to the increased imports. The implementation of EPA's ultra-low sulfur diesel (ULSD)
program this summer may also cause an increase in the demand for tank trucks if more trucks
must be dedicated to ULSD service. Given the gradual increase expected from year to year in
ethanol production, we anticipate that the industry will be able to add sufficient additional tank
truck service in an orderly fashion without undue burden.

       The necessary facility changes at terminals and at retail stations to dispense ethanol
containing fuels have been occurring at a record pace due to the removal  of MTBE from
gasoline.  The use of ethanol has also become more economically attractive due to higher
gasoline prices. Now that MTBE has been removed, a more steady increase in the use of ethanol
is anticipated over time. This will also allow for a smooth transition  for terminals  and retail
operators.

       The volumes of biodiesel that are expected to be used by 2012 to  comply with the RFS
will be relatively modest (approximately 300,000,000 gallons).  Consequently, we anticipate that
biodiesel will continue to be distributed to terminals by tank truck and by individual rail car
shipments.  One hundred percent biodiesel (B100)14 forms wax crystals when the temperature
falls to 35 to 45 degrees Fahrenheit.15 Thus, storage tanks for B100 need to be heated to
maintain flow-ability during the cold seasons. Shipping vessels used to transport B100 such as
barges, rail  cars, and tank truck containers also typically must either be insulated (and sometimes
heated) during the cold season or alternatively facilities can be provided at the terminal to reheat
the vessel prior to delivery.  Biodiesel that is blended with diesel fuel and enhanced with cold
flow additives  (if needed) can have  comparable cold flow performance to petroleum based diesel
fuel.16
13 MTBE is typically blended with gasoline at the refinery. MTBE production plants are often located nearby to
refineries allowing transport to the refinery by dedicated pipeline.  In cases where, the sources of MTBE are more
distant from the refinery, barge and rail are the preferred means of transport and relatively little MTBE is transported
by truck.

14 The concentration of biodiesel in a biodiesel blend is indicated by the number following the "B" designation. For
example, B99.9 indicates a biodiesel blend containing 99.9 percent biodiesel, and B80 indicates a blend containing
80 percent biodiesel. Manufactures of biodiesel sometimes blend in one tenth of one percent diesel fuel into
biodiesel to create B99.9 prior to shipping the fuel to terminals to create more dilute biodiesel blends so that the
producer can claim the biodiesel tax credit (pursuant to Internal Revenue Service requirements).

15 The point at which wax crystals form is referred to as the cloud point. The cloud point of B100 varies depending
on the feed stock used in its production.

16 The relatively low concentration biodiesel blends that are typically used in vehicles (up to 20% biodiesel) can be
formulated to have comparable cold flow performance to petroleum based diesel fuel. Thus, there is no need to heat
such biodiesel blends in vehicle fuel tanks.
                                              46

-------
       As temperatures fall during the cold seasons, some terminals currently avoid the need for
heated B100 tanks and facilities to heat shipping vessels by accepting progressively less
concentrated biodiesel blends (for final blending to produce fuels for use in vehicles).  During
the warm seasons, such terminals typically accept B100 or B99.9. As the weather grows colder,
the terminal might switch to accepting B80 and during the coldest parts of the year might accept
B50 (that contains 50 percent number one diesel fuel).  The need for insulated tank trucks and
tank cars is also sometimes avoided if transit times are brief by shipping warmed biodiesel.  We
believe that as the volume of biodiesel grows, most terminals will opt to receive B100 (or B99.9)
year round for blending into diesel fuel for the consistency in operations which this practice
offers. A number of terminals are already following this practice. These terminals have installed
heated storage tanks for biodiesel and insist that biodiesel be delivered in insulated tank trucks
(or rail cars) so that it may be pumped into the terminal storage tank without concern about the
potential need for reheating. The cost of the necessary heated and/or insulated equipment is not
insignificant.  However, the modest additional volumes that will need to be shipped via rail and
tank truck due to the use of biodiesel do not materially affect the conclusions reached above
regarding the ability of the fuel distribution system to cope with the increased volumes of
renewable fuels.
1.4    Blenders

1.4.1   Ethanol Blending

       Ethanol is miscible with water, and thus can introduce water into the distribution system
causing corrosion and durability problems as well as fuel quality problems. For this reason,
ethanol is blended downstream at terminals or into tank trucks.

       The distribution of ethanol is described in more detail in Section 1.3. Briefly, ethanol
producers provide ethanol either directly to terminals, to marketers or to terminals that are owned
by refiners. In the first case, ethanol is provided to terminals that are owned entities other than
refining companies. They receive ethanol from the ethanol producer,  and gasoline from any
number of refiners.  The blenders then add ethanol to the gasoline at the terminal. For RFG, the
terminals receive the blendstock for RFG, called Reformulated Blendstock for Oxygenate
Blending or RBOB, to which they add the amount of ethanol called for on the Product Transfer
Document that accompanies such shipments. Once the ethanol is added to the RBOB, the
product becomes a finished gasoline (RFG) and is sent via truck to retailers. For conventional
gasoline (CG) ethanol is also added and shipped to retailers.  The tracking mechanism for CG is
not as detailed as it is for RFG, however. The majority of ethanol that is blended into CG has
historically been "splash-blended" although an increasing volume of ethanol is being blended
into special blends of conventional gasoline (e.g. sub-octane), or "match blended". Finally, a
very small amount is blended as E85.

1.4.2   Biodiesel Blending

       Biodiesel generally leaves the production facility in its neat form and is shipped by truck
to locations where it can be blended with conventional diesel fuel. The blending generally
                                           47

-------
occurs at centralized distribution points such as terminals, although it also sometimes occurs
within tank trucks themselves. Biodiesel is only rarely used in its neat (unblended) form.
1.5    Imports/Exports of Renewable Fuel

       Since the early 1980s, the U.S. has maintained a 54 cent per gallon tariff on imported
ethanol, primarily to offset the blending tax subsidy of the same magnitude that had been put in
place to support alternative energy production and domestic agriculture. Legislation and
agreements implemented since then have waived or significantly reduced the tariff on imports
from Canada, Mexico, and about two dozen Central American and Caribbean nations covered by
the Caribbean Basin Initiative (CBI). Under the Caribbean Basin Economic Recovery Act,
which created the CBI, these countries can export ethanol duty free to the U.S. at a rate up to 7%
of the U.S. fuel alcohol market; quantities above this limit have additional stipulations for
feedstocks being grown within the supplying country.

       Historically, the CBI nations have had little ethanol production capacity of their own but
have supplemented it by importing Brazilian ethanol and re-exporting it to the U.S. duty free.
More recently, with the rapid phase-out of MTBE and the high price of ethanol, it has become
economically viable to import significant quantities of ethanol directly from other nations despite
the tariff. Brazil, currently the largest ethanol producing nation in the world, has become the
largest single country supplier to the U.S. market.  As shown in Figure 1.5-1, total imports have
increased more than 30% in 2004-5 over the previous three-year average.
                                           48

-------
             Figure 1.5-1. Historic U.S. Ethanol Import Volumes and Origins3
          300 n
          250 —
                   2001
2002
2003
2004
2005
       a P.O. Licht, "World Ethanol Markets, The Outlook to 2015" (2006).  Gross imports (does not account for
export volumes) including hydrous, dehydrated, and denatured volumes.
       Going forward, as domestic ethanol production capacity increases rapidly, its price is
expected to fall back into the historic range of 30-40 cents per gallon above gasoline (before
blending subsidy).  This is expected to once again make direct imports from Brazil and other
full-tariff producers less attractive, and to decrease total imports.  According to a current report
by P.O. Licht, U.S. net import demand is estimated to be around 300 million gallons per year by
2012, being supplied primarily through the CBI, with some direct imports from Brazil during
times of shortfall or high price.Q

       Changes in the production and trade climate may influence this however. The Caribbean
countries with duty free  status are seeing both internal and foreign investment to increase ethanol
production capacity significantly over the next several years, making more cheap imports
available. It is unclear at this point what volume of ethanol will be supplied through these
channels.

       On the export side, the U.S. has averaged about 100 million gallons per year since 2000,
mostly to Canada, Mexico, and the E.U. Figure 1.5-2 shows historical U.S. exports. There is a
trend over the past five years of exporting larger quantities to fewer countries, with declining
volumes to Asia and increasing volumes to the E.U. and India. The demand for ethanol in all
these areas remains strong, and it appears that Asian imports from Brazil and China are making
up for the decrease in U.S. ethanol moving into the region.
                                           49

-------
             Figure 1.5-2. Historic U.S. Ethanol Export Volumes and Origins'1
       140 n
                                                                 I Canada   100 Mexico

                                                                 I Japan    D EU

                                                                 I India     EB Others
                 2001
2002
2003
2004
2005
       a P.O. Licht, "World Ethanol Markets, The Outlook to 2015" (2006). Gross exports (does not account for
import volumes), includes hydrous, dehydrated, and denatured volumes.
       These numbers are expected to increase modestly as more production comes online, with
more dramatic increases possible during periods of depressed domestic prices or stock surges.
Looking out over the next decade, the E.U. has a biofuels directive in place that will bolster
demand, and Japan and South Korea are expected to increase their use of biofuels steadily as
well. World ethanol production is projected to grow from the current 10 billion gallons per year
to more than 25 in 2015, and the international biofuels markets are just beginning to take shape.
During this period we can expect significant changes in who is supplying and who is demanding
as the players determine their places and forge agreements on subsidies and tariffs.  As of 2005,
the U.S. became largest ethanol producing nation, eclipsing Brazil, and ample foreign markets
will be available if conditions are right.
                                           50

-------
Chapter 2:  Changes to Motor Vehicle Fuel Under the
Renewable  Fuel Standard Program
       In this regulatory impact analysis, we begin by describing the renewable fuel
volume scenarios we used to measure the environmental and economic impacts of
increased renewable fuel blending. From there we narrow our discussion in on ethanol -
the predominant renewable fuel expected to be used in the future. We describe historical
ethanol use, current use and our projections of future ethanol use. The discussion starts
with an in-depth examination of current ethanol use. More specifically, what factors
drive ethanol use and where ethanol blending currently occurs - by state, season, and fuel
type. The discussion then shifts to where ethanol is expected to be used in the future.
We discuss the ongoing trend in increased ethanol use, the anticipated phase-out of
MTBE, and ultimately present our LP modeling results which predict where ethanol will
likely be used in 2012 by PADD, season, fuel type. From there, we describe our
methodology for allocating ethanol usage by state and in some cases, make distinctions
on how we think ethanol would fill urban and rural areas. Once we understand how
ethanol use is expected to change in the future, we measure the anticipated impacts on
gasoline fuel quality (which later feeds into our emissions and air quality analyses). At
the end of this chapter, we also provide a brief estimate on how increased biodiesel
blending will impact diesel fuel properties.
2.1    Renewable Fuel Volume Scenarios

       The Energy Policy Act of 2005 (the Energy Act or the Act) stipulates that the
nationwide volumes of renewable fuel required under the Renewable Fuel Standard
(RFS) program must be at least 4.0 billion gallons in 2006 and increase to 7.5 billion
gallons by 2012.  However, we expect that actual renewable fuel usage will exceed the
RFS requirement by a significant margin. In Annual Energy Outlook 2006 (AEO 2006),
the Energy Information Administration (EIA) projects that total renewable fuel demand
would be 9.9 billion gallons by 2012. More specifically, EIA predicts that 9.6 billion
gallons of ethanol and 303 million gallons of biodiesel would be consumed in 2012. The
projected renewable fuel consumption levels were estimated using EIA's LP  refinery
model which was based on a crude oil price  of $48/bbl. This figure is lower than today's
crude oil price (tracking around $55/bbl at the time of our analysis).17118 Therefore,
current market conditions indicate that renewable fuel production could be even more
                                                      10
favorable and/or prevalent in the future based on economics.   However, EIA's AEO
17 West Texas Intermediate (WTI) crude oil pricing was $59.08/bbl in November, 2006; $61.96/bbl in
December, 2006; and $54.51/bbl in January 2007 according to EIA spot pricing.

18 In AEO 2007, EIA forecasted an even higher ethanol consumption of 11.2 billion gallons by 2012. The
draft report was issued on December 5, 2006, and we were unable to incorporated it into the refinery
modeling used to conduct our analyses.


                                       51

-------
2006 analysis also considers the feasibility of building production facilities to
accommodate for the growing renewable fuel demand.  Accordingly, we interpret EIA's
ethanol and biodiesel projections to be reasonable estimates considering both economics
and the rate at which new plants could feasibly come on-line.  As a result, in assessing
the impacts of expanded renewable fuel use, we evaluated two renewable fuel usage
scenarios (described in more detail below). The first represents the statutorily-required
minimum and the second reflects the higher levels projected by EIA in AEO 2006.
Although the actual renewable fuel volumes produced in 2012 may differ from both the
required and projected volumes, we believe that these two volume scenarios represent a
reasonable range for analysis purposes.

       The Act also requires that at least 250 million gallons of the total renewable fuel
use in 2013 and beyond meet the definition of cellulosic biomass ethanol.  As described
in Chapter 1, there are a number of companies planning to produce ethanol from
cellulosic feedstocks and/or waste-derived energy sources that could potentially meet the
definition of cellulosic biomass ethanol.  Accordingly, we anticipate  a ramp-up in
cellulosic biomass ethanol production in the coming years.  Furthermore, for analysis
purposes, we have assumed that the 250 million gallon requirement would be met by
2012.

       As discussed in more  detail below in  Section 2.2.2, we chose 2004 to represent
current baseline conditions. In 2004, 3.5 billion gallons of ethanol and 25 million gallons
of biodiesel were consumed in motor vehicle fuels.  To compare fuel quality impacts on
emissions and air quality, we created a 2012 reference case that maintained current fuel
quality parameters (with the exception of sulfur) but incorporated forecasted increases in
vehicle miles traveled, changes in fleet demographics, etc. The 2012 fuel reference case
was developed by growing out the 2004 renewable fuel baseline according to EIA's
forecasted energy growth rates. In AEO 2006, EIA predicted that gasoline demand
would grow by 11.2 percent and diesel fuel demand would grow by 20.5 percent from
2004 to 2012. As a result, the 2012 reference case is based  on 3.9 billion gallons of
ethanol use and 30 million gallons of biodiesel use in 2012.

       For our analyses, we created two 2012 control cases representing expanded
renewable use - the "RFS Case" and the "EIA Case". In both cases, cellulosic biomass
ethanol use was assumed to be 250 million gallons (statutory required minimum) and
biodiesel use was assumed to be 303 million  gallons (EIA AEO 2006 estimate).  The RFS
Case was designed to exactly meet the RFS program requirements considering the effects
of higher equivalence values for cellulosic ethanol and biodiesel.  Per § 80.1115, one
gallon of cellulosic ethanol counts 2.5 times towards compliance  and one gallon of
biodiesel counts  1.5  times towards compliance. As  a result, in the RFS Case we predict
that less than 7.5 billion gallons of renewable fuel would actually be  consumed in 2012.
The actual volume of renewable fuel analyzed for the RFS Case was computed to be
approximately 7.0 billion gallons.  The EIA Case represents EIA's projections of
renewable fuel use in 2012. Based on AEO 2006, the actual volume of renewable fuel
                                       52

-------
analyzed for the EIA Case was 9.9 billion gallons.  A summary of the renewable fuel
volume scenarios we evaluated is found below in Table 2.1-1.
              Table 2.1-1 Renewable Fuel Volume Scenarios (MMgal)

Renewable Fuel
Corn Ethanold
Cellulosic Ethanol6
Biodiesel
Total Renewable Volume
Total Compliance Volumef
2004
Base Case3
3,548
0
25
3,573
n/a
2012
Ref Case"
3,947
0
30
3,977
n/a
RFS Case
6,421
250
303
6,974
7,500
EIA Case0
9,388
250
303
9,941
n/a
      aHistorical ethanol usage derived from ElA's June 2006 Monthly Energy Review. Biodiesel
      usage derived from "The Outlook and Impact of Biodiesel on the Oilseeds Sector" presented
      by John Baize at the 2006 USDA Outlook Conference.
      bThe reference case was calculated by applying the 2004-2012 gasoline/diesel energy growth
      rates reported in AEO 2006 to the 2004 Base Case.
      CEIA Case based on ethanol and biodiesel energy contributions reported in AEO 2006.
      Includes ethanol imports.
      eEthanol meeting the definition of cellulosic biomass ethanol in The Act.
      'Based on applying a 2.5 equivalence value to cellulosic biomass ethanol and a 1.5
      equivalence value to biodiesel.
2.2    Current Gasoline Oxygenate Use

2.2.1   Why are oxygenates currently blended into gasoline?

       The blending of oxygenates into gasoline dates back to the 1970's. However,
their use greatly expanded in response to the Clean Air Act (CAA) amendments of 1990.
Areas found to be out of compliance (i.e., in non-attainment) with the National Ambient
Air Quality Standards (NAAQS) for ozone were required to reformulate their gasoline
and use oxygenates year-round. In addition, several states began to use oxygenated fuel
(oxy-fuel) in the wintertime to address carbon monoxide non-attainment. In addition,
oxygenates (namely ethanol) have historically been used as a gasoline volume extender
and more recently, to meet state mandates.  This section summarizes the current driving
forces behind gasoline oxygenate use in the U.S.
2.2.1.1
Federal Reformulated Gasoline Program
       As mentioned above, areas found to be in ozone non-attainment were required to
use reformulated gasoline (RFG) year-round. The federal RFG program contained a
minimum oxygenate requirement as well as other fuel quality standards.19 Adding
19 RFG oxygenate requirement found at 40 CFR 80.41(f).  This requirement was effective for 2004 but has
since been eliminated by the Energy Act Section 1504, promulgated on May 8, 2006 at 71 FR 26691.
                                         53

-------
oxygen to gasoline and reformulating other gasoline properties has helped to reduce the
production of smog-forming pollutants that contribute to unhealthy ground-level ozone.
Besides ozone non-attainment areas, several states/areas also opted into the RFG program
(otherwise known as "opt-in").  In addition, California and Arizona have state programs
that promote the use of oxygenated gasoline.

       A list of the 2004 federal RFG areas and their corresponding oxygenate(s) is
provided in Table 2.2-1.  For the purpose of this analysis, only ethanol (ETOH) and
methyl tertiary-butyl ether (MTBE) have been considered.20
20Other low-usage oxygenates (e.g. ETBE, TAME, etc.) were assumed to be negligible for the purpose of
this analysis.
                                        54

-------
                 Table 2.2-1. 2004 Federal RFG Areas by State
State
California
Connecticut0
Delaware0
District of Columbia0
Illinois
Indiana
Kentucky
Maryland
Massachusetts0
Missouri
New Hampshire
New Jersey0
New York
Pennsylvania
Rhode Island0
Texas
Virginia
Wisconsin
City
Los Angeles
Sacramento
San Diego
San Joaquin Valley
Hartford
Long Island Area
Windham County
Sussex County
Wilmington
Washington DC Area
Chicago Area
Chicago Area
Covington
Louisville
Baltimore
Philadelphia Area
Queen Anne/Kent Counties
Washington DC Area
Boston Area
Springfield
St. Louis
Boston Area
Atlantic City
Long Island Area
Trenton
Warren County
Poughkeepsie
Long Island Area
Philadelphia Area
Providence Area
Dallas/Fort Worth
Houston/Galveston
Norfolk/Virginia Beach
Richmond
Washington DC Area
Milwaukee-Racine
No. of
Counties3
5
6
1
8
6
1
1
1
2
1
8
2
3
3
6
1
2
5
10
4
5
4
2
12
6
1
2
11
5
5
4
8
11
7
10
6
Type of
RFG Area
Req'd
Req'd
Req'd
Req'd
Req'd
Req'd
Opt In
Opt In
Req'd
Opt lnd
Req'd
Req'd
Opt In
Opt In
Req'd
Req'd
Opt In
Opt lnd
Opt In
Opt In
Opt In
Opt In
Opt In
Req'd
Req'd
Opt In
Opt In
Req'd
Req'd
Opt In
Opt In
Req'd
Opt In
Opt In
Opt lnd
Req'd
Primary
Oxygenate13
ETOH
ETOH
ETOH
ETOH
ETOH
ETOH
ETOH
MTBE
MTBE
MTBE
ETOH
ETOH
ETOH
ETOH
MTBE
MTBE
MTBE
MTBE
MTBE
MTBE
ETOH
MTBE
MTBE
Both
MTBE
MTBE
ETOH
ETOH
MTBE
MTBE
MTBE
MTBE
MTBE
MTBE
MTBE
ETOH
Includes partial counties.
bOxygenate determination based on 2004 FHWA gasohol data and EPA fuel survey results.
°Entire state/district operates under the Federal RFG program.
dWas "oot-in" in 2004. now a reauired RFG area.
      As shown above in Table 2.2-1, a little more than half of the Federal RFG areas
(on a county-by-county basis) used MTBE as opposed to ethanol as an oxygenate in
2004.  However, on a volumetric basis, more ethanol was consumed in RFG than MTBE
                                     55

-------
(2.2 billion gallons compared to 1.9 billion gallons as shown in Tables 2.1.5 and 2.1.3,
respectively).
2.2.1.2        State Oxygenated Fuel Programs

       In addition to the RFG program, several states require oxygenated fuel (oxy-fuel)
to be used in the wintertime to address carbon monoxide (CO) non-attainment.  CO is
formed from the incomplete combustion of hydrocarbons (found in all gasoline blends).
Production of the  poisonous gas is more prevalent in oxygen-deficient environments and
more harmful to human health in the wintertime due to temperature inversions.21
Together, the winter oxy-fuel program coupled with improving vehicle emissions control
systems has helped to reduce CO emissions. Many areas have and are continuing to
come into attainment with the CO national ambient air quality standards (NAAQS).
However, many former non-attainment areas continue to use the winter oxy-fuel program
as part of a maintenance plan for remaining in compliance with the CO NAAQS.  A list
of the 2004 oxy-fuel areas is provided in Table 2.2-2.  All oxy-fuel areas were assumed to
use ethanol in 2004 based on information obtained from regional EPA offices.

         Table 2.2-2.  2004 State-Implemented  Winter Oxy-Fuel Programsu
Oxy-Fuel Area Location
State
Alaska
Arizona
California
Colorado
Montana
Nevada
New Mexico
Oregon
Texas
Utah
Washington
City
Anchorage
Tucson
Phoenix
Los Angeles
Denver/Boulder
Longmont
Missoula
Las Vegas
Reno
Albuquerque
Portland
El Paso
Provo/Orem
Spokane
Oxy-Fuel
Period
11/1-2/29
10/1-3/31
11/2-3/15
10/1-2/29
11/1-1/31
11/1-1/31
11/1-2/29
10/1-3/31
10/1-1/31
11/1-2/29
11/1-2/29
10/1-3/31
11/1-2/29
9/1-2/29
Carbon Monoxide Status
Designation
Non-attainment0
Attainment
Non-attainment
Non-attainment
Attainment
Attainment
Non-attainment
Non-attainment
Non-attainment
Attainment
Attainment
Non-attainment
Non-attainment
Non-attainment11
Pursuing RDa
X

X
X


X

X



X
X
Winter Oxy-Fuel Program
Required
X

X
X


X
X
X


X
X
X
Part of MPb

X


X
X



X
X



aCurrently pursuing redesignation to CO attainment.
bArea is in currently in CO attainment but oxy-fuel program remains as part of maintenance plan.
°Area was redesignated to attainment effective 7/23/04.
dArea was redesignated to attainment effective 8/29/05.
2.2.1.3
Other Motivations for Blending Ethanol
21 Temperature inversions in the lower atmosphere are relatively common, especially during winter months
in cold climates. A temperature inversion occurs when cold air close to the ground is trapped by a layer of
warmer air, creating stagnation and trapping pollution close to the ground.
                                        56

-------
       In addition to the RFG and oxy-fuel programs, gasoline refiners have several
other motivations for blending oxygenate (namely ethanol) into gasoline.  First and
foremost, the state they provide gasoline to could be operating under a state ethanol
mandate.  In 2004, Hawaii joined Minnesota in approving a state ethanol requirement.22
Second, blending ethanol into gasoline could help them meet their mobile source air
toxics (MSATI) performance standards as determined by the Complex Model.23 Third,
adding ethanol increases both octane and total fuel volume, thus helping refiners extend
their gasoline production.  Finally, and perhaps most importantly, with record-high crude
oil prices and the growing availability of grain-based ethanol (especially in PADD 2),
ethanol use has become increasingly economical.  The 1.1 billion gallons of ethanol used
in PADD 2 conventional gasoline in 2004 (refer to Table 2.2-5 in Section 2.2.2.4) is a
good indicator of this trend.

       In addition to the increasing  availability of ethanol, consumer demand is also
increasing based on the growing number of ethanol-friendly vehicles on the road.
Conventional vehicles consume the  majority of fuel  ethanol and are limited to gasoline
with 10 volume percent (vol%) ethanol (E10) or less. However, there are currently
around 6 million flexible fuel vehicles (FFVs) on the road today with more being
produced and sold each dayv. FFVs are specifically designed to handle a wide range of
gasoline/ethanol blends up to 85 vol% ethanol (E85).

2.2.2  Development of the Base Case

       As discussed in 2.1, to evaluate the impacts of increased ethanol blending and
decreased MTBE blending on gasoline properties  (and in turn air quality), we had to
create a point of comparison. To do so, we assembled a 2004 Base Case to represent
current baseline conditions, i.e., current gasoline, ethanol, and MTBE use.  The
methodology for assembling the base case, as well as a summary of the results, is
described below.

2.2.2.1        Strategy for Establishing the 2004 Base Case

       For the purpose of this regulatory impact analysis, the 2004 calendar year was
selected to reflect current baseline conditions. This period represented the most current
year for which gasoline and oxygenate data were available and also captured the
California, New York, and Connecticut MTBE bans (effective  1/1/04) while avoiding the
2005 calendar year hurricane upsets.
22 For analysis purposes, both states were assumed to have ethanol mandates which required 100% of the
gasoline to contain 10% ethanol. However, in reality, Hawaii's ethanol mandate only requires that 85% of
the gasoline contain 10% ethanol.

23 This RFS proposal is based on MS ATI conditions. Impacts of the recent MSAT2 rule (72 FR 8428)
which removes individual refinery toxic performance standards (baselines) in exchange for a nationwide
benzene standard are reflected in the analysis for that rulemaking.
                                         57

-------
       The approach for assembling the 2004 base case consisted of obtaining gasoline,
ethanol, and MTBE usage for all 50 states as well as the District of Columbia.  As
mentioned earlier, other low-volume oxygenate use (e.g., ETBE, TAME, etc.) was
assumed to be negligible and thus ignored for this analysis. All ethanol-blended gasoline
was assumed to contain 10 vol% ethanol, with the exception of California "RFG"
(Federal RFG and California Phase 3 RFG (CaRFG3)).24 Current California gasoline
regulations make it very difficult to meet the NOx emissions performance standard with
ethanol content higher than about 6 vol%. For our analysis, all California RFG was
assumed to contain 5.7 vol% ethanol based on discussions with California Air Resources
Board (CARB). This percentage was also applied to California RFG supplied to the
Phoenix metropolitan area in the summertime under Arizona's clean burning gasoline
(CBG) program.25 Finally, all MTBE-blended gasoline was assumed to contain 11 vol%
MTBE.

       Total gasoline consumption was obtained from the 2004 Petroleum Marketing
Annual (PMA) report published by the Energy Information Administration (EIA).W The
reported annual average sales volume for each state was interpreted as total blended
gasoline (including additives, namely oxygenates).  2004 MTBE usage by state was
obtained from EIA.26'X The data received was exclusive to states with RFG programs
(including Arizona's CBG program). Thus, for the purpose of the 2004 base case
analysis, MTBE use was assumed to be limited to RFG areas. 2004 ethanol usage by
state was derived from a compilation of data sources and assumptions.  As a starting
point, total domestic ethanol consumption was acquired from EIA's Monthly Energy
Review published in June 2006Y.  State ethanol contributions originated from the 2004
Federal Highway Administration (FHWA) gasohol report2. However, there was some
ambiguity with the 2004 FHWA data.  First, the total ethanol consumption did  not match
up with EIA's reported value (3.7 billion gallons compared to 3.5 billion gallons).
Second, the gasohol (and thus ethanol) volumes were derived from potentially  imprecise
motor vehicle fuel tax reports.27 And third, not all states using ethanol reported their
24 The small volumes of E85 (85 percent ethanol) gasoline have been ignored for this analysis.

25 For the Base Case analysis, all Arizona CBG was classified as "RFG". In 2004, wintertime Arizona
RFG was assumed to contain 10% ethanol (governed by the Phoenix oxy-fuel program).  Summertime
RFG was assumed to be comprised of 2/3 California RFG (containing 5.7 percent ethanol) and 1/3 PADD 3
RFG (containing either 10 percent ethanol or 11 percent MTBE in 2004).

26 EIA reported 2004 total MTBE usage (in RFG) as 2.0 billion gallons. The reported MTBE usage was
reduced from 2.0 to 1.9 billion gallons under the assumption that CA, NY, and CT implemented their state
MTBE bans on time (by 1/1/04).  (EIA showed small amounts of MTBE use in these states in 2004).
EIA's allocation of MTBE by state was also adjusted based on fuel survey results. Most noteworthy, EIA
reported MTBE usage in Arizona "RFG" as zero. However, the 2004 Phoenix fuel survey results suggest
otherwise. As such, an appropriate amount of MTBE was allocated to Arizona based on the assumption
that 1/3 of all summertime Arizona "RFG" resembles PADD 3 RFG (which contained some level of MTBE
in 2004).

27 The U.S. Department of Treasury requires a distinction between gasohol and gasoline on motor vehicle
fuel tax reports for states with gasohol sales tax exemptions.  These financial records are the source of
FHWA's gasohol/ethanol data. However, since state gasohol tax exemptions have become virtually


                                         58

-------
gasohol usage so FHWA had to model-estimate 19 states' ethanol usage (accounting for
60% of the total ethanol volume). To improve upon the FHWA data, we used a series of
oxygenate verification tools including knowledge of state ethanol mandates, state MTBE
bans, Arizona's CBG program, and fuel survey results.AABB  The state-by-state FHWA
data was adjusted accordingly and allocated by fuel type (RFG, CG, and/or oxy-fuel).
The summarized oxygenate results are presented throughout this section.
2.2.2.2
2004 Gasoline/Oxygenate Consumption by PADD
       In 2004, 3.5 billion gallons of ethanol and 1.9 billion gallons of MTBE were
blended into gasoline to supply the transportation sector with a total of 136 billion gallons
of gasoline.  A breakdown of the 2004 gasoline and oxygenate consumption by PADD is
found below in Table 2.2-3.
          Table 2.2-3. 2004 Gasoline & Oxygenate Consumption by PADD


PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5b
California
Total

Gasoline
MM gal
49,193
38,789
20,615
4,542
7,918
14,836
135,893
Ethanol

MMgal
660
1,616
79
83
209
853
3,500
%of
Gasoline
1 .3%
4.2%
0.4%
1 .8%
2.6%
5.8%
2.6%
%of
Tot ETOH
18.9%
46.2%
2.3%
2.4%
6.0%
24.4%
100.0%
MTBE3

MMgal
1,360
1
498
0
19
0
1,878
%of
Gasoline
2.8%
0.0%
2.4%
0.0%
0.2%
0.0%
1.4%
%of
Tot MTBE
72.4%
0.1%
26.5%
0.0%
1 .0%
0.0%
100.0%
aMTBE blended into RFG
bPADD 5 excluding California
       As shown above, in 2004, almost half (or 46 percent) of the ethanol was
consumed in PADD 2, where the majority of ethanol was produced. The next highest
region of use was the State of California which accounted for nearly a quarter (or 24
percent) of domestic ethanol consumption. This makes sense since California alone
accounts for over 10 percent of the nation's total gasoline consumption. And in 2004,
following their MTBE ban, all fuel (both Federal RFG and CaRFG3) was presumed to
contain 5.7 vol% ethanol. The next highest region of use was PADD 1 (19 percent)
which makes sense considering the high concentration of RFG areas (most of which used
ethanol in  2004 as shown in Table 2.2-1). The remaining 10 percent of ethanol use
occurred collectively in PADDs  3, 4, and 5/
nonexistent over the past several years, gasohol reporting (namely the distinction between gasoline and
gasohol) has suffered.
                                       59

-------
       In 2004, total ethanol use exceeded MTBE use. Ethanol's lead oxygenate role is
relatively new, however the trend has been a progression over the past few years. From
2001 to 2004, ethanol consumption more than doubled (from 1.7 to 3.5 billion gallons),
while MTBE use (in RFG) was virtually cut in half (from 3.7 to 1.9 billion gallons). A
plot of oxygenate use over the past decade is provided below in Figure 2.2-1.
                   Figure 2.2-1.  Oxygenate Consumption vs. Time
                                                                 CC,DD
   4.5 -,
   4.0
        1995     1996     1997     1998     1999     2000     2001     2002     2003     2004
                                          Year
                      -Ethanol Consumption, Total
-MTBE Consumption, RFG
       The nation's transition to ethanol is linked to states' responses to recent
environmental concerns surrounding MTBE groundwater contamination. Traces of
MTBE have been found in both surface and ground water in and around RFG areas. The
MTBE is thought to have made its way into the water from leaking underground storage
tanks, gasoline spills, and engines.  Concerns over drinking water quality prompted
several states to significantly restrict or completely ban MTBE use in gasoline.  At the
time of our analysis, 19 states had adopted MTBE bans. Ten states had bans that
impacted the entire 2004 calendar year, four states had bans that impacted a portion of the
year, and five states had bans that became effective in 2005 and beyond. A list of the
states with MTBE bans (listed in order of phaseout date) is provided below in Table 2.2-
4.
                                        60

-------
               Table 2.2-4. States MTBE Bans by Phaseout Date
                                                             EE
State3
Iowa
Minnesota
Nebraska
South Dakota
Colorado
Michigan
California
Connecticut
New York
Washington
Kansas
Illinois
Indiana
Wisconsin
Ohio
Missouri
Kentucky
Maine
New Hampshire
Phaseout Date
07/01/00
07/02/00; 07/02/05
07/13/00
07/01/01
04/30/02
06/01/03
12/31/03
01/01/04
01/01/04
01/01/04
07/01/04
07/24/04
07/24/04
08/01/04
07/01/05
07/31/05
01/01/06
01/01/07
01/01/07
Type of Banb
Partial
Partial; Complete
Partial
Partial
Complete
Complete
Complete
Complete
Complete
Partial
Partial
Partial
Partial
Partial
Partial
Partial
Partial
Partial
Partial
aArizona is not included because they do not have an official state
MTBE ban. They adopted legislation on 4/28/00 calling for a
complete phaseout of MTBE as soon as feasible but no later than
six months after California's phaseout. The legislation expired on
June 30, 2001 , so it's not official policy. Although the state still
informally encourages the phaseout of MTBE.
bA partial ban refers to no more than 0.5 vol% MTBE except in the
case of MN (1/3%), NE (1 %), and WA (0.6%)
       As explained above in 2.2.2.1, all MTBE consumption was assumed to occur in
reformulated gasoline in 2004. As shown in Table 2.2-3, 99 percent of MTBE use (by
volume) occurred in PADDs 1 and 3.  This reflects the high concentration of RFG areas
in the northeast (PADD 1) and the local production of MTBE in the gulf coast (PADD 3).
PADD 1 receives a large portion of its gasoline from PADD 3 refineries who either
produce the fossil-fuel based oxygenate or are closely affiliated with MTBE-producing
petrochemical facilities in the area.
2.2.2.3
2004 Gasoline/Oxygenate Consumption by Season
       In 2004, according to EIA Petroleum Marketing Annual (PMA), approximately
40 percent of gasoline was consumed in the summertime and 60 percent was consumed in
the wintertime.FF Similarly, according to EIA Monthly Energy Review June 2006, 38
percent of the ethanol was consumed in the summertime and 62 percent was consumed in
the wintertime. 28'GG
                                      61

-------
       Total gasoline use is higher in the wintertime because it's a longer season.  The
RFG regulations define summertime fuel as gasoline produced from May 1st to
September 15th (4.5 months total).29 The remaining 7.5 months are considered to be
wintertime gasoline. Even though on an average per day basis summertime consumption
is higher, more gasoline is still sold and consumed in the wintertime based on the length
of the season.

       Seasonal ethanol use follows the same general trend as gasoline.  However,
besides the associated correlation with seasonal gasoline  consumption, there are
additional reasons why 2004 ethanol use may have been higher in the wintertime.  First,
the oxy-fuel program requires oxygenate to be used in certain areas in the wintertime
only.  These same areas, which do not require oxygenate in the summer, were all
presumed to use ethanol as their oxygenate (as described in 2.2.1.2).  Thus, more areas
use ethanol during the winter months than the summer. Secondly, there is an economic
penalty associated with blending ethanol into summertime RFG.  Refiners supplying
summertime gasoline to RFG areas have to remove butanes and pentanes from their
gasoline in order to  add ethanol and still comply with the 7 psi Reid vapor pressure
(RVP) requirement.

2.2.2.4        2004 Gasoline/Oxygenate Consumption by Fuel Type

       According to fuel survey results, in 2004, approximately 2.2 billion gallons of
ethanol were blended into reformulated gasoline and the remaining 1.3 billion gallons
were used in conventional gasoline (including wintertime oxy-fuel).HH'n A breakdown of
the 2004 ethanol consumption by fuel type and PADD is found in Table 2.2-5.
28 Aforementioned seasonal split for gasoline and ethanol based on RFG production seasons (Summer: May
1 through September 15th; Winter: January 1st through April 30th and September 16th through December
31st).

29 We acknowledge that the aforementioned seasonal split does not exactly match the new summer/winter
seasons defined in the Energy Act (Summer: April 1st through September 30th; Winter: January 1st through
March 31st and November 1st through December 31st).
                                        62

-------
                                   Table 2.2-5.
                2004 Ethanol Consumption by Fuel Type (MMgal)
PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5
California
Total
CG
0
1,07:
31
0
4£
0
1.141
OX?
0
! 0
21
83
89
0
i 19^
RFd
66C
54^
26
0
75
ss:
2.155
Total
66C
1,61(
7£
82
20£
ss:
3.501
aWinter oxy-fuel programs
"Federal RFC plus CA Phase 3 RFC and Arizona CBG
CPADD 5 excluding California
       As mentioned above in Section 2.2.2.1, 100 percent of the 1.9 billion gallons of
MTBE blended into gasoline in 2004, was assumed to be consumed in reformulated
gasoline.
2.2.2.5
2004 Gasoline/Oxygenate Consumption by State
       In 2004, ethanol was blended into gasoline in 34 of the 50 states. No ethanol use
was observed in the remaining 16 states: Maine, New Hampshire, Vermont,
Pennsylvania, Delaware, Georgia, North Carolina, South Carolina, West Virginia,
Tennessee, Oklahoma, Mississippi, Arkansas, Louisiana, Idaho, and West Virginia, nor
was any ethanol used in Washington DC. A summary of 2004 ethanol usage by state is
presented in Table 2.2-6. Note that a state ethanol percentage less than 10 indicates that
only a percentage of the gasoline pool was blended with ethanol, not that ethanol itself
was blended in less than 10 vol% (E10) proportions, except in the case of California
gasoline (E5.7). Figure 2.2-2 shows the percentage of E10 by state.

       The states consuming the highest volumes of ethanol in 2004 were California,
Illinois, New York, Minnesota, and Ohio, respectively. With respect to gasoline use, the
highest percentage of ethanol use occurred in Minnesota, Hawaii, Connecticut, Illinois,
and Iowa. Four out of the five states are not surprising.  The first two states have ethanol
mandates and the last two are located in the "corn belt" where ethanol is produced.
Connecticut's high percentage of ethanol use may come as a surprise at first glance.
However, the entire state operates under the RFG program (refer to Table 2.2-1), and
since they also have a state MTBE ban, ethanol is found in each gallon of gasoline.
                                       63

-------
Table 2.2-6. 2004 Gasoline/Ethanol Consumption by State


State
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida13
Georgia
Hawaii3
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland13
Massachusetts
Michigan
Minnesota
Mississippi
Missouri

Gasoline
MMgal
2,392
302
2,187
1,406
14,836
1,999
1,522
449
119
8,605
4,729
452
632
5,177
3,059
1,635
1,396
2,177
2,287
757
2,480
2,934
4,861
2,684
1,617
3,159
Ethanol

MMgal
31
3
88
0
853
80
152
0
0
0
0
45
0
422
148
117
41
50
0
0
0
18
77
268
0
122
%of
Gasoline
1.3%
1.1%
4.0%
0.0%
5.8%
4.0%
10.0%
0.0%
0.0%
0.0%
0.0%
10.0%
0.0%
8.1%
4.8%
7.1%
2.9%
2.3%
0.0%
0.0%
0.0%
0.6%
1 .6%
10.0%
0.0%
3.9%


State
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island13
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia13
Washington
West Virginia
Wisconsin
Wyoming
Total

Gasoline
MMgal
503
819
857
705
4,235
966
5,626
4,302
350
5,156
2,158
1,500
4,786
490
2,422
434
3,251
11,948
1,097
338
3,920
2,621
772
2,471
311
135,893
Ethanol

MMgal
1
37
23
0
188
8
301
0
11
192
0
31
0
0
0
24
0
39
2
0
0
18
0
109
0
3,500
%of
Gasoline
0.2%
4.5%
2.7%
0.0%
4.4%
0.8%
5.4%
0.0%
3.0%
3.7%
0.0%
2.1%
0.0%
0.1%
0.0%
5.5%
0.0%
0.3%
0.2%
0.0%
0.0%
0.7%
0.0%
4.4%
0.0%
2.6%
aHawaii was assumed to have a 100% E10 mandate in the 2004 Base Case based on RFA's Homegrown for the
Homeland: Ethanol Industry Outlook 2005.
bTrace amounts of ethanol use (<1 MMGal) in FL, MD, Rl and VA.
                          64

-------
             Figure 2.2-2. 2004 Ethanol Consumption, % E10 by State
    2004% E10 by State
       n  o% EIO
       CH  <50% EIO
       O  50-99% EIO
       O  100% EIO
Not Pictured
AK: 13% EIO
HI: 100% EIO
DC: 0% EIO
2.2.3   Development of the 2012 Reference Case
       To establish the 2012 reference case, we started with the 2004 Base Case
(presented in Table 2.2-3) and grew out gasoline/oxygenate use according to the EIA
AEO 2006 motor gasoline energy growth rate from 2004 to 2012.JJ Accordingly, in the
resulting 2012 reference case, ethanol and MTBE use was proportional to 2004 use by
both region and fuel type. A summary of the 2012 ethanol reference case is found in
Table 2.2-7.
                                       65

-------
                                       Table 2.2-7.
           2012 Reference Case - Gasoline & Oxygenate Consumption by PADD
                                    (MMgal)30


PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5b
California
Total

Gasoline
MMgal
54,743
43,166
22,941
5,055
8,812
16,509
151,225
Ethanol

MMgal
735
1,798
88
93
232
949
3,895
%of
Gasoline
1 .3%
4.2%
0.4%
1 .8%
2.6%
5.8%
2.6%
%of
Tot ETOH
18.9%
46.2%
2.3%
2.4%
6.0%
24.4%
100.0%
MTBEa

MMgal
1,513
2
554
0
21
0
2,090
%of
Gasoline
2.8%
0.0%
2.4%
0.0%
0.2%
0.0%
1.4%
%of
Tot MTBE
72.4%
0.1%
26.5%
0.0%
1 .0%
0.0%
100.0%
aMTBE blended into RFG
bPADD 5 excluding California
2.2.4   Development of the 2012 Control Cases

       In Section 2.2.2 we described our methodology behind building the 2004 Base
Case, which was used to produce the 2012 Reference Case (described above). In this
section we will describe how we developed the two 2012 control cases representing
increased ethanol fuel use - the RFS Case and the EIA Case.  Both control cases
incorporate our knowledge of future state ethanol mandates, tax incentives, and
anticipated winter oxy-fuel usage. Our analysis relied on LP modeling (described in
more  detail below) to determine how much ethanol would be used in each PADD, season,
and fuel type.  From there, we conducted post-processing to determine how much
ethanol would be used on a state-by-state basis and in some cases and made predictions
on how ethanol would likely fill urban and rural areas.
2.2.4.1
Forecasting Ethanol Consumption / LP Modeling Results
       As mentioned earlier in Section 2.2.2.2, groundwater contamination concerns
have caused many states to ban the use of MTBE in gasoline. In response to the Energy
Act, all U.S. refiners are expected to eliminate the use of MTBE in gasoline by the end of
2007, and certainly prior to 2012.  Ethanol consumption, on the other hand is expected to
continue to grow in the future. Not only are the Energy Act's RFS requirements
promoting ethanol growth, ethanol is needed to fuel the growing number of ethanol-
friendly vehicles being produced as well as satisfy the growing number of state ethanol
  The total ethanol volume reported in table 2.2-7 (3.895 Bgal) is slightly lower than the reference case
value reported in Table 2.1-1 (3.947 Bgal). The reasonforthe slight discrepancy is because the numbers
presented here were based off the estimated 2004 base case (3.5 Bgal) whereas the numbers presented in
Table 2.1-1 were based off a more precise 2004 ethanol use (3.548 Bgal) reported by EIA in July 2006
Monthly Energy Review.
                                        66

-------
mandates (Washington, Montana, Louisiana, and Missouri recently joined Minnesota and
Hawaii)31'KK'LL
       Based on projections from EIA and others, it's abundantly clear that renewable
fuel use (namely ethanol) is growing much faster than the RFS requirement.  However
quantifying future ethanol use is a difficult task.  The gasoline refining industry and
ethanol industry are currently undergoing a variety of changes/expansions and there is no
precise way to know exactly how things are going to "fall out" in the future.
Accordingly, as explained in Section 2.1, we have considered two different 2012
renewable fuel consumption scenarios to represent a reasonable range of ethanol use. For
the RFS Case we modeled 6.7 billion gallons of ethanol use and for the EIA case we
modeled 9.6 billion gallons (refer to Table 2.1-1). EPA is not concluding that ethanol
consumption could not possibly exceed 9.6 billion gallons by 2012, but rather that this
volume is a reasonable "ceiling" for our analysis.

       To estimate how ethanol use would be allocated in the future, we relied on
Jacob's Consultancy LP refinery modeling.MM For the Base Case and Reference Case,
the LP refinery model was set up to allocate fixed volumes of ethanol/MTBE to regions
consistent with our analysis of current gasoline oxygenate use (described above in
Sections  2.2.2 and 2.2.3). This essentially fixed the total  ethanol and MTBE use in each
PADD.  From there, the oxygenates were further allocated by season and fuel grade to
match the oxygen content for RFG, RBOB and CBOB based on 2004 batch report data.
Any leftover ethanol was allocated to CG.  Based on the resulting fuel allocation, the LP
model generated CG and RFG fuel properties considering the RVP effects and blending
qualities  of ethanol and MTBE (such properties are discussed further and utilized in
Section 2.3).

       For each of the future control cases, MTBE use was assumed to be zero and the
amount of ethanol added to gasoline was varied.  For the  RFS Case, total ethanol use was
fixed at 6.7 billion gallons and for the EIA Case, ethanol  use was fixed at 9.6 billion
gallons.  For  each control case, the LP model used gasoline and ethanol blending
economics (e.g., ethanol distribution costs, seasonal ethanol and gasoline blendstock
prices, etc.) to determine how  much ethanol would be blended into gasoline by PADD,
season, and fuel type. Again, the results were used to generate CG and RFG fuel
properties used in Section 2.3.

       Slight adjustments had to be made to the refinery  modeling outputs to ensure that
sufficient ethanol was supplied in the wintertime to meet  the oxy-fuel requirements in
PADDs 4/5.  In addition, small corrections were  required to ensure that ethanol blending
in a given region/state did not  exceed the maximum blending criteria assumed for the
31 The Montana state mandate requires all gasoline to contain 10 vol% ethanol once plant production ramps
up to 40 MMgal/yr. The Washington state mandate requires 20% of all gasoline to contain 10 vol%
ethanol by 12/1/08. Similarly, the Louisiana state mandate requires 20% of all gasoline to contain 10 vol%
ethanol once plant production ramps up to 50 MMgal/yr. Finally, the Missouri state mandate requires all
gasoline to contain 10 vol% ethanol by 1/1/08. At the time of our analysis, these were the only four new
state ethanol mandates. However, EPA recognizes that as of 7/13/06, several others have new/additional
biofuel standards pending (California, Colorado, Idaho, Illinois, Indiana, Kansas, Minnesota, New Mexico,
Pennsylvania, Virginia, and Wisconsin).
                                         67

-------
analysis - 10 volume percent (vol%) ethanol nationwide, and 5.7 vol% ethanol in
California.  The adjusted LP refinery modeling results for the RFS and EIA control cases
are summarized below in Tables 2.2-8 and 2.2-9, respectively.

                                  Table 2.2-8.
            Adjusted LP Modeling Results for the RFS Case (MMgal)

PADD
PADD1
PADD 2
PADD 3
PADDs 4/5c
California
Total
Summer Ethanol Use
CGa
399
1,667
161
135
0
2,362
RFGb
679
59
47
0
414
1,200
Total
1,078
1,726
208
135
414
3,562
Winter Ethanol Use
CGa
350
1,082
146
138
0
1,717
RFGb
706
288
0
0
398
1,392
Total
1,057
1,370
146
138
398
3,109
Total
Ethanol
2,134
3,096
354
274
813
6,671
Includes Arizona CBG and winter oxy-fuel
bFederal RFG and California Phase 3 RFG
cPADDs 4 and 5 excluding California
                                  Table 2.2-9.
            Adjusted LP Modeling Results for the EIA Case (MMgal)

PADD
PADD 1
PADD 2
PADDS
PADDs 4/5c
California
Total
Summer Ethanol Use
CGa
610
1,735
901
339
0
3,584
RFGb
630
185
47
0
435
1,298
Total
1,240
1,919
949
339
435
4,882
Winter Ethanol Use
CGa
267
1,631
856
154
0
2,908
RFGb
973
366
0
0
470
1,809
Total
1,240
1,998
856
154
470
4,718
Total
Ethanol
2,481
3,917
1,805
492
905
9,600
Includes Arizona CBG and winter oxy-fuel
bFederal RFG and California Phase 3 RFG
cPADDs 4 and 5 excluding California
2.2.4.2
Resulting 2012 Ethanol Consumption by PADD
       Starting with the LP refinery modeling results, we segregated the Rocky
Mountain (PADD 4) and West Coast (PADD 5) ethanol use (represented as an aggregate
above in Tables 2.2-8 and 2.2-9) and examined the resulting ethanol allocation by region.
A summary of the 2012 forecasted ethanol consumption by region (PADDs 1-5 and
California) for each control case is found below in Table 2.2-10.
                                      68

-------
                                  Table 2.2-10.
                 2012 Forecasted Ethanol Consumption by PADD


PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5a
California
Total
6.7 Bgal RFS Case
Gasoline
MM gal
60,468
48,451
24,845
4,869
8,537
16,494
163,664
ETOH
MMgal
2,134
3,096
354
54
220
813
6,671
%of
Gasoline
3.5%
6.4%
1 .4%
1.1%
2.6%
4.9%
4.1%
%of
Tot ETOH
32.0%
46.4%
5.3%
0.8%
3.3%
12.2%
100.0%
9.6 Bgal EIA Case
Gasoline
MMgal
60,468
48,451
25,112
4,928
8,626
16,494
164,078
ETOH
MMgal
2,481
3,917
1,805
151
342
905
9,600
%of
Gasoline
4.1%
8.1%
7.2%
3.1%
4.0%
5.5%
5.9%
%of
Tot ETOH
25.8%
40.8%
18.8%
1 .6%
3.6%
9.4%
100.0%
aPADD 5 excluding California
       As shown above, in 2012 PADD 2 is expected to continue to dominate ethanol
use.  PADD 2 ethanol consumption is expected to double from 1.8 billion gallons (Bgal)
in the Reference Case (refer to Table 2.2-7) to 3.1 Bgal in the RFS Case and 3.9 Bgal in
the EIA Case.  This represents a slight decrease in Midwest marketshare (from 46% in
Reference/RFS Case to 40% in the EIA Case). The predicted shift in marketshare is
attributed to the growing amount of ethanol use outside of the traditional cornbelt.

       The LP modeling suggests that ethanol usage is expected to greatly increase in
PADDs 1 and 3. In PADD 1, ethanol blending is expected to more than triple from 735
million gallons in the Reference Case to 2.1 Bgal in the RFS Case and 2.5 Bgal in the
EIA Case. In PADD 3, ethanol use in expected to sharply increase from 88 million
gallons in the Reference Case to 354 million gallons in the RFS Case and 1.8 billion
gallons in the EIA Case. This projected increase in ethanol blending on the East Coast
and Gulf Coast, reflects the phase out of MTBE (replacement with ethanol) as well as
ethanol blending economics.
2.2.4.3
Resulting 2012 Ethanol Consumption by Season
       Furthermore, we examined the resulting ethanol allocation by season. The LP
refinery modeling assumes equal 182.5-day summer and winter seasons. A summary of
the resulting 2012 forecasted ethanol consumption by season for each of the control cases
is found below in Table 2.2-11.
                                       69

-------
                                  Table 2.2-11.
            2012 Forecasted Ethanol Consumption by Season (MMgal)

PADD
PADD1
PADD 2
PADD 3
PADD 4
PADD 5a
California
Total
6.7 Bgal RFS Case
Summer
1,078
1,726
208
29
106
414
3,562
Winter
1,057
1,370
146
25
113
398
3,109
Total
2,134
3,096
354
54
220
813
6,671
9.6 Bgal EIA Case
Summer
1,240
1,919
949
125
213
435
4,882
Winter
1,240
1,998
856
25
128
470
4,718
Total
2,481
3,917
1,805
151
342
905
9,600
aPADD 5 excluding California
       As shown above, ethanol usage in 2012 is expected to be slightly more prevalent
in the summertime than in the wintertime. This is a shift from our 2004 Base Case (38%
of ethanol use occurred in the summertime and 62% occurred in the wintertime as
explained in Section 2.2.2.3), mainly because we are changing the way we define the
seasons.  In the Base Case we defined the seasons based on the RFG regulations (4.5
months of "summer" and 7.5 months of "winter") whereas in this 2012 forecast we are
defining them based on 6 months of each season.  Since gasoline consumption (gal/day)
is higher in the summertime, more ethanol-blended gasoline could potentially be
consumed during the summer months.  However, since there is an economic penalty
associated with blending ethanol into summertime gasoline (refiners have to remove
butanes and pentanes to comply with the RFG RVP requirements), the result is somewhat
of a seasonal balance in both the RFS Case and the EIA Case.
2.2.4.4
Resulting 2012 Ethanol Consumption by Fuel Type
       In addition to providing a PADD and seasonal breakdown, The LP modeling
determined how much ethanol would be used by fuel type - conventional gasoline (CG)
versus reformulated gasoline (RFG). The first thing we did was allocate a portion of the
CG to the required winter oxy-fuel areas.

Strategy for Allocating Ethanol to Oxy-Fuel Areas

       In the 2004 Base Case, there were 14 state-implemented winter oxy-fuel programs
in 11  states (summarized previously in Table 2.2-2). Of these programs, 9 were required
in response to non-attainment with the CO National Ambient Air Quality Standards
(NAAQS) and 5 were implemented to maintain CO attainment status. However, in the
future 4 of the 9 required oxy-fuel areas are expected to be reclassified from non-
attainment to attainment and discontinue using oxy-fuel in the wintertime32. These areas
are: Anchorage, AK; Las Vegas, NV; Provo/Orem, UT; and Spokane, WA.  In addition,
 : Based on conversations with state officials and regional EPA officials.
                                       70

-------
Colorado is expected to discontinue using winter oxy-fuel in Denver/Boulder and
Longmont to maintain CO attainment status.  The use of oxy-fuel in the above-mentioned
areas is expected to discontinue by 2012 or sooner. With the removal of these 6 state-
implemented programs, that leaves a total of 8 oxyfuel areas in Tucson and Phoenix, AZ;
Los Angeles, CA; Missoula, MT; Reno, NV; Albuquerque, NM; Portland, OR; and El
Paso, TX. We assumed that these areas would continue to blend 10 vol% ethanol into
their gasoline for their entire winter oxy-fuel period (duration varies by area, six month
maximum) in the 2012 control cases.

      Once a portion of the conventional gasoline ethanol was allocated to meet winter
oxy-fuel requirements, this gave use a PADD-by-PADD breakdown of ethanol use by
conventional gasoline, oxy-fuel, and reformulated gasoline as shown below in Table 2.2-
12.
                                  Table 2.2-12.
          2012 Forecasted Ethanol Consumption by Fuel Type (MMgal)

PADD
PADD 1
PADD 2
PADD 3
PADD 4
PADD 5d
California
Total
6.7 Bgal RFS Case
CGa
750
2,749
283
54
106
0
3,942
OXYb
0
0
24
0
113
0
137
RFGC
1,385
347
47
0
0
813
2,592
Total
2,134
3,096
354
54
220
813
6,671
9.6 Bgal EIA Case
CGa
877
3,366
1,733
151
228
0
6,356
OXYb
0
0
24
0
113
0
137
RFGC
1,603
551
47
0
0
905
3,107
Total
2,481
3,917
1,805
151
342
905
9,600
aConventional gasoline including Arizona CBG
bWinter oxy-fuel programs
'Federal RFG plus CA Phase 3 RFG
dPADD 5 excludina California
       However, more post-processing was required to determine how much ethanol
would be used on a state-by-state basis to feed into the emissions and air quality analyses.
We begin the latter part of this discussion by explaining how we allocated the RFG
ethanol to specific RFG areas and how we allocated the CG ethanol to specific
states/regions considering state ethanol mandates and the economic favorability of
ethanol blending

Strategy for Allocating Ethanol Among RFG

       In the 2004 Base Case, there were 18 states/districts with RFG programs covering
a total of 175 counties in  36 areas (summarized previously in Table 2.2-1). For our
analysis of 2012 ethanol use, we assumed that the number of RFG areas would not
change and accordingly, that the RFG fuel contribution to the gasoline pool would remain
the same. However, we considered the amount of ethanol added to RFG to be a variable,
as discussed below.
                                       71

-------
       In the past, all RFG areas were required to use a minimum amount of oxygenate
in their reformulated gasoline year-round, as discussed earlier in 2.2.1.1. However,
effective May 5, 2006, EPA removed the RFG oxygenate requirement in response to the
Act.NN Although the oxygenate requirement has already been eliminated, many refiners
are still operating under contracts with ethanol blenders. As such, refiners true response
to the removal of the oxygenate requirement is relatively unknown at this time.  While it
is difficult to predict exactly how each refinery supplying an RFG area would behave, the
LP modeling has attempted to do so.

       The modeling suggests that some refineries will continue to blend ethanol into
RFG (or even increase blending) in 2012 based on octane, volume, and/or toxic
performance requirements. Some RFG producers may decidedly replace MTBE with
ethanol while others may pare back or discontinue ethanol use all together.  A summary
of the 2012 forecasted RFG ethanol  consumption (by season) for each control case is
found below in Table 2.2-13.
                                  Table 2.2-13.
              2012 Forecasted RFG Ethanol Consumption (MMgal)
33



PADD/State
PADD1
PADD2
PADD 3 / TX
PADD 5 1 CAb
Total

Seasonal
RFG Use
MMgal3
11,380
3,661
2,939
8,247
26,227
6.7 Bgal RFS Case
Summer
ETOH
MMgal
679
59
47
414
1,200
%of
Gasoline
6.0%
1.6%
1.6%
5.0%
4.6%
Winter
ETOH
MMgal
706
288
0
398
1,392
%of
Gasoline
6.2%
7.9%
0.0%
4.8%
5.3%
9.6 Bgal EIA Case
Summer
ETOH
MMgal
630
185
47
435
1,298
%of
Gasoline
5.5%
5.0%
1.6%
5.3%
4.9%
Winter
ETOH
MMgal
973
366
0
470
1,809
%of
Gasoline
8.5%
10.0%
0.0%
5.7%
6.9%

Average
% ETOH
in RFG
6.6%
6.1%
0.8%
5.2%
5.8%
aEqual amounts of reformulated gasoline assumed to be used in the summer and winter seasons.
"includes Federal RFG and CA Phase 3 RFG
       As shown above, the modeling suggests that more ethanol would be consumed in
RFG in the EIA Case in the presence of more ethanol.  The modeling also suggests that
the greatest ethanol marketshare would occur in California RFG (5.2 vol% ethanol on
average across both cases/seasons, or 91% E5.7). The next highest areas of RFG use
would be PADD 1 (6.6 vol% ethanol on average, or 66% E10) followed by PADD 2 (6.1
vol%  ethanol on average, or 61% E10). Little ethanol blending was predicted to occur in
Texas RFG (0.8% ethanol or 8% E10).

       In both control cases, more ethanol was predicted to be blended into wintertime
RFG.  As discussed earlier, this makes sense because in order to meet the RVP
requirements pertaining to summertime RFG (7 psi), refiners have to  remove butanes and
33 Gasoline consumed in the greater Phoenix metropolitan area under the Arizona Clean Burning Gasoline
(CBG) Program, has not been considered "RFG" by the LP refinery modeling and thus discussed in the
conventional gasoline section.
                                       72

-------
pentanes to accommodate for ethanol blending (which increases overall gasoline
volatility).  As such, in the absence of an RVP waiver (which exists exclusively for
summertime CG), refiners are less inclined to blend ethanol into summertime RFG.

       To allocate the RFG ethanol (aggregated by PADD and season in Table 2.2-13)
by state/RFG area, we assumed that each region would behave uniformly with the
exception of PADD 1 (discussed in more detail below).  For example, consider PADD 2
summertime RFG. In the RFS Case, RFG in Chicago, Louisville, Milwaukee, etc. would
all contain 1.6% ethanol on average.  Or more accurately, 16% of all the gasoline
consumed within PADD 2 RFG areas would contain 10% ethanol.

       However, based on our knowledge of the refining industry and distribution
patterns, we did not assume that PADD 1 RFG would be uniform in ethanol content. The
LP modeling assumes that the RFG produced in PADD  1 contains ethanol but the RFG
produced in PADD 3 and shipped to  PADD 1 does not.  RFG from PADD 3 comes up
the Colonial Pipeline and passes through Virginia, Washington DC and Maryland on its
way to Pennsylvania and New York.  With the exception of a small Yorktown refinery,
the southernmost refineries in PADD 1 are located around the Philadelphia area.
However, there is no cheap way to send fuel south. Therefore, the RFG coming from
PADD 3 is likely to completely fulfill the RFG demand in Virginia, Washington DC and
Maryland.  Beyond Maryland, the fuel from PADD 1 refineries is sold along with any
leftover PADD 3 RFG, as distribution costs are roughly the same from Philadelphia
north.  As a result, the Virginia, Washington DC and Maryland RFG areas were assumed
to receive less ethanol (in most cases zero E10) than the other RFG areas located in
PADD 1.  A summary of the resulting RFG ethanol distribution by state is found below
in Table 2.2-14.
                                 Table 2.2-14.
                     2012 RFG Ethanol Distribution by State
PADD/State
PADD1
PADD 2
PADD 3/TX
PADD 5/CA
6.7 Bgal RFS Case
Summer
78% E10 in all states
except DC, MD, VA
(0%E10)
1 6% E10 in all states
16%E10inTX
88% E5.7 in CA
Winter
81 %E10 in all states
areas except DC, MD,
VA(0%E10)
78% E10 in all states
0%E10inTX
85% E5.7 in CA
9.6 Bgal EIA Case
Summer
73% E10 in all states
except DC, MD, VA
(0% E10)
51 %E10 in all states
16%E10inTX
93% E5.7 in CA
Winter
1 00% E10 in all states
except DC, MD, VA
(39%E10)
1 00% E10 in all states
0%E10inTX
100%E5.7inCA
Strategy for Allocating Ethanol Among CG

       The above-mentioned oxy-fuel requirements combined with state ethanol
mandates created a "floor" for conventional gasoline ethanol use within each PADD.
This essentially forced a specific amount of ethanol to be used in wintertime CG in
PADDs 3 and 5 and a specific amount of ethanol to be added year-round in Minnesota,
                                      73

-------
Montana, and Missouri (100% E10 mandates); Hawaii (85% E10 mandate); as well as
Washington and Louisiana (20% E10 mandates).

       To determine how the remaining ethanol would be allocated to the leftover
conventional gasoline, we devised a systematic way to allocate ethanol by state/area.
Since the primary motivation to blend (or not blend) ethanol is expected to be economic,
we devised a way to rank CG areas, on a state-by-state and urban/rural basis, as to the
economic favorability of ethanol blending.  This was done by calculating an ethanol
margin, which is equal to gasoline price minus ethanol delivered price. Ethanol delivered
price is equal to ethanol plant gate price plus transportation  costs minus any additional
state plus other adjustments (explained below).  The greater the ethanol margin, the
greater the economic incentive and the more likely ethanol is to be used in that area.

       At the time the analysis was carried out, ethanol plant gate price was taken from
an older EIA NEMS model.  However, since this price was  assumed to be the same for
all ethanol, the actual value is not important when trying to  estimate relative allocation
preferences between areas.  All ethanol blending was assumed to be done at 10 volume
percent. The gasoline prices for each state were the weighted average rack price of all
conventional grades and all months, taken from EIA Petroleum Marketing Annual
2004. °°

       Ethanol distribution costs were taken from figures given in the documentation for
the EIA NEMS model,  and are based  on a 2002 study by DAI, Inc.pp  For the purpose of
this consumption analysis, all ethanol was assumed to be produced in the Midwest in
census divisions 3 and 4 (corresponding closely to PADD 2).  This is largely consistent
with the production analysis presented in Chapter 1 of the RIA. While the results of the
production analysis do not completely coincide with this assumption (as shown in Table
1.2-15, about 86 percent of the total ethanol plant  capacity is expected to originate from
PADD 2 in 2012 and the rest would originate from other areas throughout the country),
this simplifying assumption is still very reasonable.

       Ethanol consumed within census divisions 3 and 4 was assumed to be transported
by truck, while distribution outside of those areas was via rail, ship, and/or barge. A
single average distribution cost for each destination census division was generated by
weighting together the 2012 freight costs given for each mode in both census divisions 3
and 4 according to their volume share. These cent per gallon figures were first adjusted
upward by 10 percent to reflect higher energy prices, and then additional adjustments
were applied to some individual states based on their position within the census division.
In the cases of Alaska and Hawaii, differences in ethanol delivery prices from the
mainland were inferred from gasoline prices.  Table 2.2-15  shows the gasoline price and
ethanol distribution cost for each state as used in this analysis.
                                        74

-------
                Table 2.2-15.
Gasoline Price & Ethanol Distribution Costs
34
State
Alabama
Alaska
Arizona
Arkansas
Colorado
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Gasoline Rack
Price (c/gal)
123.2
157.0
138.0
123.3
129.5
124.9
125.8
151.7
134.2
125.7
125.6
127.5
124.3
125.9
123.1
125.5
124.8
126.5
127.4
123.0
126.0
130.5
126.0
141.6
125.3
128.4
126.0
124.4
127.7
126.2
123.4
133.8
126.1
124.9
127.8
124.5
122.5
132.3
127.3
123.4
132.1
125.8
125.2
130.4
ETOH Distribution
Cost (c/gal)
7.2
41.5
15.4
7.3
10.4
8.4
11.4
36.5
15.4
4.4
5.4
3.4
4.4
6.2
7.3
13.4
11.4
6.4
4.4
6.2
4.4
13.4
4.4
16.4
12.4
12.4
11.4
11.4
5.4
5.4
8.3
16.5
8.4
11.4
4.4
6.2
10.3
13.4
12.4
11.4
16.5
11.4
4.4
12.4
                     75

-------
       As the final step in the calculation, subsidies and other adjustments were applied.
The federal blending credit of 51 cents per gallon was given to all areas, and five state
retail incentives were included as follows (all cents per gallon of ethanol): Iowa, 29.5;
Illinois, 20.1, South Dakota, 20; Maine: 7.5;  Oklahoma,  1.6.35QQ

       In addition to state subsidies, small penalty adjustments were made for
distributing ethanol into rural areas in several states (as presented in Table 2.2-16). The
reasoning behind this is that when large shipments of ethanol come from the Midwest by
barge,  ship, or rail, they will be unloaded initially at large terminals near metropolitan
areas.  Further storage and handling will be required to allow smaller quantities to be
distributed via truck into rural areas. Several states have gasoline pipelines that traverse
them with connections at various points, helping to reduce distribution burdens, but
ethanol is not expected to be shipped via pipeline. Overall, the largest adjustments were
applied to the Rocky Mountain states since they are generally larger in area and
additional expense is required to transport freight through higher elevations and rugged
terrain. Smaller adjustments were applied to states that are smaller, flatter, or have
navigable water access on one  or more sides. The states that do not appear on this list are
either located in the Midwest (where ethanol is produced and readily available to
virtually all areas at similar costs) or are small northeast states not believed to have
significant differences between rural and urban distribution costs.

                                    Table  2.2-16.
               Adjustment for Ethanol Distribution into Rural Areas
States
OH
AL, AR, FL, GA, KY, LA, ME,
MS, NC, NY, OK, OR, PA, SC,
TN, VA, WA, WV
AK, AZ, CO, ID, NM, NV, UT,
WY, TX
Rural Area
Adjustment (c/gal)
2
4
5
       To determine which in-use areas/counties would receive urban versus rural
ethanol distribution pricing based on the economies of scale described above, we looked
to the U.S. Census Bureau which considers population density and other factors.
  The following states have intentionally been excluded from this CG gasoline/ethanol cost table because
they do not consume any CG (100% RFC): CA, CT, DC, DE, MA, NJ, RI.

35 EPA acknowledges that other states are considering (or may have even approved) retail pump incentives
for gasohol. However, at the time this consumption analysis was completed, these were the only five states
offering retail pump incentives that were likely to be applicable in 2012.
                                         76

-------
       Metropolitan Statistical Areas (MSAs) served as the starting point for determining
the areas within a state that would be considered "urban".  MSAs are geographic entities
defined by the U.S. Office of Management and Budget (OMB) for use by federal
statistical agencies in collecting, tabulating, and publishing federal statistics.  An MSA is
defined as having a core urban area of 50,000 or more people.  Each MSA consists of one
or more counties including the counties contained the core urban area, as well as any
adjacent counties that have a high degree of social and economic integration with the
urban core. For the purposes of this analysis, we only considered MSAs with populations
greater than 1 million people, or other areas having special qualifications. Such
qualifications include MSAs with less than 1 million people that happen to be the largest
MSA in a less-populated state (i.e. Montana and Wyoming), or other MSAs deemed
likely to receive ethanol by rail based on proximity to major rail lines.

       Once the urban counties for each state were determined, county-level  vehicle
miles traveled (VMT) from 2002 were used (as a surrogate for fuel consumption) to
weight the urban counties' approximate fuel demand.  Expressing the urban VMT as a
function of statewide VMT gave us the percentage of ethanol demand that would  be
considered eligible for an urban ethanol distribution cost (values presented in Table 2.2-
15).  The remaining percentage of ethanol demand was considered to be "rural" and
subject to the ethanol blending penalty adjustments found in Table 2.2-16.

       Considering the urban/rural split for each state and the resulting ethanol margin
(ethanol delivered price minus gasoline production cost), we came up with the resulting
ranking system for distributing ethanol into conventional gasoline.  For PADD 1,  refer to
Table 2.2-17; PADD 2, refer to Table 2.2-18, PADD 2, Table 2.2-19; and PADDs 4/5,
Table 2.2-20.  The summer and winter percentages are the same for each urban/rural area
with the exception of states containing winter oxy-fuel areas. For these states, winter
oxy-fuel was deducted from the winter urban fuel since this volume of gasoline was
already accounted for (refer to Table 2.2-12).
                                        77

-------
              Table 2.2-17.
Precedence for Adding ETOH to PADD 1 CG
State
ME
PA
FL
ME
VT
NY
GA
WV
PA
SC
MD
NC
NH
FL
VA
NY
WV
GA
SC
NC
VA
Rural /
Urban
u
u
u
r
-
u
u
u
r
u
-
u
-
r
u
r
r
r
r
r
r
Ethanol
Margin (c/gal)
50.6
48.7
47.5
46.6
45.9
45.6
45.4
45.4
44.7
44.5
44.4
44.0
43.9
43.5
43.0
41.6
41.4
41.4
40.5
40.0
39.0
PADD1
Precedence
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
% of CG Volume
Summer
35.7%
44.7%
59.9%
64.3%
100.0%
67.8%
51.2%
21.1%
55.3%
20.2%
100.0%
14.7%
100.0%
40.1%
63.9%
32.3%
78.9%
48.8%
79.8%
85.3%
36.1%
Winter
35.7%
44.7%
59.9%
65.3%
100.0%
67.8%
51.2%
21.1%
55.3%
20.2%
100.0%
14.7%
100.0%
40.1%
63.9%
32.3%
78.9%
48.8%
79.8%
85.3%
36.1%
                  78

-------
              Table 2.2-18.
Precedence for Adding ETOH to PADD 2 CG
State
IA
SD
IL
ND
NE
OH
Wl
IN
Ml
KS
KY
OH
TN
OK
KY
TN
OK
Rural /
Urban
-
-
-
-
-
u
-
-
-
-
u
r
u
u
r
r
r
Ethanol
Margin (c/gal)
84.6
74.4
72.4
53.3
52.6
51.8
51.8
51.2
51.1
50.9
50.7
49.8
49.3
47.7
46.7
45.3
43.7
PADD 2
Precedence
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
% of CG Volume
Summer
100.0%
100.0%
100.0%
100.0%
100.0%
37.7%
100.0%
100.0%
100.0%
100.0%
10.2%
62.3%
38.4%
29.9%
89.8%
61.6%
70.1%
Winter
100.0%
100.0%
100.0%
100.0%
100.0%
37.7%
100.0%
100.0%
100.0%
100.0%
10.2%
62.3%
38.4%
29.9%
89.8%
61.6%
70.1%
              Table 2.2-19.
Precedence for Adding ETOH to PADD 3 CG
State
MS
NM
AR
AL
LA
MS
TX
AR
AL
LA
NM
TX
Rural /
Urban
u
u
u
u
u
r
u
r
r
r
r
r
Ethanol
Margin (c/gal)
47.8
47.0
47.0
47.0
46.8
43.8
43.2
43.0
43.0
42.8
42.0
38.2
PADD 3
Precedence
1
2
3
4
5
6
7
8
9
10
11
12
% of CG Volume
Summer
6.8%
31.3%
26.4%
31.2%
16.7%
93.2%
61.8%
73.7%
68.8%
63.3%
68.7%
38.2%
Winter
6.8%
16.0%
26.4%
31.2%
16.7%
93.2%
58.2%
73.7%
68.8%
63.3%
68.7%
38.2%
                  79

-------
                                  Table 2.2-20.
                 Precedence for Adding ETOH to PADDs 4/5 CG
State
NV
AZ
NV
CO
UT
ID
WY
AZ
OR
WA
AK
HI
CO
UT
ID
OR
WY
WA
AK
Rural /
Urban
u
u
r
u
u
u
u
r
u
u
u
-
r
r
r
r
r
r
r
Ethanol
Margin (c/gal)
56.2
53.6
51.2
50.1
49.9
49.8
49.0
48.6
48.3
46.6
46.5
46.2
45.1
44.9
44.8
44.3
44.0
42.6
41.5
PADDs 4/5
Precedence
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
% of CG Volume
Summer
57.8%
55.2%
42.2%
48.6%
38.7%
32.3%
1 1 .6%
44.8%
38.2%
43.6%
36.4%
15.0%
51.4%
61.3%
67.7%
61.8%
88.5%
36.4%
63.6%
Winter
49.2%
5.4%
42.2%
48.6%
38.7%
32.3%
1 1 .6%
44.8%
1.4%
43.6%
36.4%
15.0%
51.4%
61.3%
67.7%
61.8%
88.5%
36.4%
63.6%
2.2.4.5
Resulting 2012 Gasoline/Oxygenate Consumption by State
       Applying the CG order of precedence tables to the remaining conventional
gasoline ethanol (less state mandated and winter oxy-fuel volumes) and factoring in the
RFG ethanol distribution (described above in 2.2.4.4), we came up with an ethanol
distribution by state for each control case. The resulting state-by-state ethanol
distribution is summarized below in Table 2.2-21 and a graphical representation for each
control case is provided in Figures 2.2-3 and 2.2-4 below.

                                  Table 2.2-21.
                  2012 Forecasted Ethanol Consumption by State
                             (continued on next page)
                                       80

-------

State
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Total
Abbrv
AL
AK
AZ
AR
CA
CO
CT
DE
DC
FL
GA
HI
ID
IL
IN
IA
KS
KY
LA
ME
MD
MA
Ml
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
Rl
SC
SD
TN
TX
UT
VT
VA
WA
WV
Wl
WY

6.7 Bgal RFS Case
Gasoline
MMgal
2,869
324
2,377
1,686
16,494
2,143
1,827
538
143
10,734
5,899
484
677
6,737
3,820
2,017
1,722
2,742
2,743
944
2,990
3,522
5,995
3,310
1,940
3,986
539
1,010
920
855
5,083
1,152
6,876
5,366
432
6,358
2,661
1,623
5,909
588
3,022
535
4,010
14,454
1,176
422
4,786
2,810
964
3,117
333
163,664
ETON
MMgal
90
0
123
44
813
0
146
43
0
521
0
0
0
454
352
202
86
42
99
34
0
281
346
331
25
343
54
101
6
50
406
35
289
0
43
438
0
34
318
47
0
54
37
62
0
0
0
56
0
268
0
6,671
%of
Gasoline
3.1%
0.0%
5.2%
0.0%
4.9%
0.0%
8.0%
0.0%
0.0%
4.8%
0.0%
0.0%
0.0%
6.7%
9.2%
10.0%
5.0%
1 .5%
0.0%
0.0%
0.0%
8.0%
5.8%
10.0%
0.0%
8.6%
10.0%
10.0%
0.7%
0.0%
8.0%
3.0%
4.2%
0.0%
10.0%
6.9%
0.0%
2.1%
0.0%
8.0%
0.0%
10.0%
0.0%
0.4%
0.0%
0.0%
0.0%
2.0%
0.0%
8.6%
0.0%
4.1%
9.6 Bgal EIA Case
Gasoline
MMgal
2,910
327
2,397
1,710
16,494
2,169
1,827
538
143
10,734
5,899
490
686
6,737
3,820
2,017
1,722
2,742
2,782
944
2,990
3,522
5,995
3,310
1,968
3,986
546
1,010
931
855
5,083
1,167
6,876
5,366
432
6,358
2,661
1,638
5,909
588
3,022
535
4,010
14,574
1,191
422
4,786
2,843
964
3,117
337
164,078
ETON
MMgal
291
0
149
171
905
57
158
46
3
474
12
42
12
570
368
202
172
72
278
64
52
304
599
331
197
372
55
101
52
54
439
109
423
0
43
636
28
34
328
51
0
54
78
759
25
21
52
64
0
291
2
9,600
%of
Gasoline
10.0%
0.0%
6.2%
10.0%
5.5%
2.6%
8.6%
8.6%
2.0%
4.4%
0.2%
8.5%
1 .7%
8.5%
9.6%
10.0%
10.0%
2.6%
10.0%
6.8%
1 .7%
8.6%
10.0%
10.0%
10.0%
9.3%
10.0%
10.0%
5.6%
6.3%
8.6%
9.3%
6.1%
0.0%
10.0%
10.0%
1 .0%
2.1%
5.6%
8.6%
0.0%
10.0%
2.0%
5.2%
2.1%
5.0%
1.1%
2.3%
0.0%
9.3%
0.6%
5.9%
81

-------
      Figure 2.2-3. 2012 Forecasted Ethanol Consumption
               6.7 Bgal RFS Case, % E10 by State
2012% E10 by State
   CH  0% E10
   CH  <50% E10
   d  50-99% E10
   H  100% E10
Not Pictured
AK: 0% E10
HI: 0% E10
DC: 0% E10
      Figure 2.2-4. 2012 Forecasted Ethanol Consumption
               9.6 Bgal EIA Case, % E10 by State
2012%E10byState
   cn  o% EIO
   dl  <50% EIO
   CH  50-99% EIO
   H  100% EIO
Not Pictured
AK: 0% EIO
HI: 85% EIO
DC: 20% EIO
                                82

-------
2.3    Effects of Ethanol and MTBE on Gasoline Fuel Properties

       For the final rulemaking, we estimate the impact of increased ethanol use and
decreased MTBE use on gasoline quality using refinery modeling conducted specifically
for the RFS rulemaking.36  The methods, analyses, and results of the refinery modeling
are discussed in more detail in Chapter 7. In general, adding ethanol to gasoline reduces
the aromatic content of conventional gasoline (CG) and the mid and high distillation
temperatures (e.g.,  T50 and T90). RVP increases except in areas where ethanol blends
are not provided a 1.0 RVP waiver of the applicable RVP standards in the summer.  With
the exception of RVP, adding MTBE direct!onally produces the same impacts.  Thus, the
effect of removing  MTBE results in essentially the opposite impacts.  Neither oxygenate
is expected to affect sulfur levels, as refiners control sulfur independently in order to
meet the Tier 2 sulfur standards.

       The impacts of oxygenate use are smaller with respect to RFG. This is due to the
applicability of VOC and toxics emission performance specifications, which limit the
range of feasible fuel quality values.  Thus, RVP and aromatic and benzene contents are
not consistently affected by oxygenate type or level.

       Table 2.3-1 shows the fuel quality of a typical summertime, non-oxygenated
conventional gasoline and how these qualities change with the additional of 10  volume
percent ethanol (10 vol%).  Similarly, the table shows the fuel quality of a typical MTBE
RFG blend and how fuel quality might change with either ethanol use or simply MTBE
removal. Note that the table does not reflect county-specific fuel properties.
Table 2.3-1. CG and RFG Summer Fuel Quality With and Without Oxygenates

Fuel Parameter
RVP (psi)
T50
T90
E200
E300
Aromatics (vol%)
Olefins (vol%)
Oxygen (wt%)
Benzene (vol%)
Conventional Gasoline
Typical
9 RVP
8.7
218
332
41
82
32
7.7
0
1.0
Ethanol
Blend
9.7
205
329
50
82
27
7.7
3.5
1.0
Reformulated Gasoline a
MTBE
Blend
7.0
179
303
60
89
20
4
2.1
0.74
Ethanol
Blend
7.0
184
335
58
82
20
14
3.5
0.70
Non-Oxygenated
Blend
7.0
175
309
52
88
20
15
0
0.72
a MTBE blend -Reference Case PADD 1 South, Ethanol blend - RFS Case PADD 1 North, Non-oxy
blend - RFS Case PADD 1 South
36 Refinery modeling performed in support of the original RFG rulemaking is also used to help separate the
effects of the two oxygenates.
                                       83

-------
2.3.1   Effect of Ethanol on Conventional Gasoline Fuel Properties
       To estimate effects of ethanol on conventional gasoline, we used the refinery
model output shown below in Table 2.3-2.  These values represent average properties
across the five PADDs and winter and summer seasons.

     Table 2.3-2.  Properties of Conventional Gasoline Per Refinery Modeling
Case
20 12 Reference
2012 RFS
2012EIA
Aromatic
s (vol%)
29.18
29.02
27.97
Olefins
(vol%)
12.39
12.12
12.39
E200
(vol%)
49.96
51.89
53.04
E300
(vol%)
81.64
83.68
82.20
T50
(°F)
190.3
187.4
185.9
T90
(°F)
335.3
326.2
332.8
MTBE
(vol%)
0.65
0.00
0.00
Ethanol
(vol%)
1.66
3.87
5.38
       Using this output, we estimated an average change in each fuel property per vol%
change in ethanol content. To derive these estimations, we first adjusted the Reference
case properties to isolate the effects of ethanol by mathematically removing the effects of
MTBE using factors derived from the RFG RIARR. Then, we calculated the change in
each fuel property per change in ethanol vol% for each combination of cases. That is, we
compared Reference Case to RFS Case, Reference to EIA, and EIA to RFS. Finally, we
averaged the three results from the case-to-case comparisons to derive a useful factor for
adjusting county-level fuel properties for a change in county-level ethanol content. These
ethanol effects are shown below in Table 2.3-3.

  Table 2.3-3. Change in Conventional Gasoline Properties Per Vol% Increase in
                                    Ethanol
Aromatics
(vol%)
-0.46
Olefins (vol%)
0.02
E200 (vol%)
0.91
E300 (vol%)
0.06
T50 (°F)
-1.33
T90 (°F)
-0.28
2.3.2  Effects of MTBE on Conventional Gasoline Fuel Properties

       In support of the final rule implementing the RFG program in 1993, refinery
modeling was performed which estimated the impact of MTBE blending on the various
gasoline properties.ss While this modeling was performed in the context of projecting
the cost of producing RFG, it is applicable to the use of MTBE in CG, as well. The
refinery modeling examined a number of incremental steps involved in the production of
RFG.  Because RFG was mandated to contain oxygen and MTBE was expected to be the
oxygenate of choice, MTBE was added in the first step of the analysis, before the fuel
met the rest of the RFG requirements. Table 2.3-4 shows the results of adding MTBE
based on this refinery modeling.

       This modeling of MTBE effects is somewhat dated (circa 1993).  However, since
removing MTBE does not involve any predictions of its total usage level nor the  location
of its use, economics (such as crude oil price) are not a factor.  It is primarily an issue of
chemical properties and general refinery operation, such as octane management.  Also,
                                       84

-------
MTBE is always match-blended, since gasoline can be shipped with MTBE through
pipelines.  Thus, MTBE is always added at the refinery, allowing the refiner to take full
advantage of its properties.
Table 2.3-4. Effect of MTBE on Gasoline Properties: RFG Final Rule
Fuel Parameter
RVP (psi)
T50a
T90a
E200 (vol%)
E300 (vol%)
Aromatics (vol%)
Olefms (vol%)
Oxygen (wt%)
Benzene (vol%)
Base 9 RVP Gasoline
8.7
218
329
41
83
32.0
13.1
0
1.53
MTBE Blend
8.7
207
321
46.7
84.9
25.5
13.1
2.1
0.95
Difference
0
-11
-8
5.7
1.9
-6.5
0
2.1
-0.58
a Estimated using correlations developed in support of EPA RFG final rule, Docket A-92-12, February
1994.
   T50 = 302 - E200 / 0.49 and T90 = 707 - E300 / 0.22
       As with ethanol blending, MTBE blending reduces aromatic content significantly
as refiners take advantage of MTBE's high octane level. Like ethanol, MTBE also tends
to increase E200 and E200 and decrease T50 and T90.  Unlike ethanol, MTBE does not
increase RVP.

       MTBE blending is shown to modestly reduce sulfur and benzene levels, as well.
This refinery modeling was performed prior to the development of the Tier 2 sulfur
standards for gasoline. With these standards, gasoline must meet a 30 ppm sulfur
standard on average with or without MTBE blending. As refiners can adjust the severity
of their hydrotreating processes to account for various changes in feedstocks and
oxygenate use, we do not expect that the removal of MTBE will result in any increase in
sulfur content. Otherwise, the reversal of the differences shown in Table 2.3-4 are
expected to occur when MTBE is removed from gasoline (when the MTBE content was
11 vol%).

2.3.3 Effects of Ethanol and MTBE on Reformulated Gasoline Fuel Properties

       RFG has historically contained oxygenate due to the applicable 2.0 weight percent
oxygen content requirement.  RFG has contained 11 vol% MTBE or ten vol% ethanol,
except in California, where 6 vol% ethanol blends have been common. As discussed in
Section 2.2, the use of MTBE in RFG has ceased. It has been replaced by either 10 vol%
or ethanol  or high octane hydrocarbon blending components, such as alkylate or
reformate. In either case, as discussed in Section 2.3.2, RFG will continue to have to
meet stringent VOC, NOx, and toxic emission performance standards, though compliance
                                       85

-------
with the NOx standard is essentially assured with compliance with the Tier 2 sulfur
standards applicable to all gasoline37.

       For the NPRM, we assumed that the properties of RFG other than oxygenate
would not be affected by changes in oxygenate use. For the FRM, we are utilizing the
recent refinery modeling to estimate RFG properties by PADD and season for the three
ethanol use scenarios.

       As described above, five refinery models were developed, each representing one
PADD.  These models produced fuel for use in its own PADD, as well as for use in other
PADDs in the several cases of PADD-to-PADD distribution of gasoline. For RFG,
refinery modeling projected that a significant volume of RFG used in PADD 1 would be
produced by PADD 3 refineries.  Because this PADD 3 RFG is shipped to PADD 1 via
pipeline, this fuel tends to be used in the southernmost RFG areas of PADD 1, namely
those in Virginia, the District of Columbia and Maryland. In order to reflect this, we
assumed that the RFG produced in PADD 3 will be used preferentially in the RFG areas
of these three states and the RFG produced in PADD 1 will be used to fulfill the
remaining demand for RFG in PADD 1. For refinery modeling, it was estimated that a
small volume of RFG would be produced in PADD 3 and shipped to PADD 2.  Because
of the small volume involved, we did not assign this volume to a specific RFG area
within PADD 2.

       As part of their work, the refinery modeling contractors calibrated their model to
match EPA's estimate of fuel quality existing in 2004 (i.e., the base case in this analysis).
Therefore, estimates of the properties of RFG for the base case comprise an accurate
estimate of actual 2004 RFG, at least on a PADD-average basis. We also have available
the results of the RFG fuel survey for each RFG area. This survey data sometimes
reflects significant differences in the properties of RFG for specific RFG areas within a
PADD.  Good examples of this would be RFG areas in New York and Connecticut,
which implemented MTBE bans  starting in 2004. We considered using the more precise
RFG survey data to represent RFG fuel quality in the base case, but rejected this
approach for two reasons. One, this would introduce an extraneous difference in RFG
fuel quality between the base case and the RFS/EIA cases. While refinery model base
case projections reasonably match EPA's estimate  of 2004 fuel quality, they do not match
exactly. Comparisons between the base case and the RFS and EIA cases would therefore
include the difference between the RFG survey data and the refinery modeling
contractor's  estimate of this data, plus the effect of additional ethanol use and reduced
MTBE use.  Two, we primarily present the emission impacts of the RFS rule on a
nationwide basis.  On a nationwide basis, reflecting differences between RFG fuel quality
within a PADD would have little impact.  Also, the Ozone RSM can only reflect a single
change in VOC and NOx emissions in non-attainment areas  (e.g., RFG areas). Thus,
differences between  specific RFG areas would be eliminated by the limitation that only
37 Though the MSAT2 final rulemaking (72 FR 8428, February 26, 2007) eliminates these air toxics and
NOx requirements beginning in 2011.
                                       86

-------
the average emission effect can be modeled.  Thus, we used refinery modeling
projections of RFG fuel quality for all three fuel scenarios.
       Tables 2.3-5, 2.3-6, and 2.3-7 present the fuel properties of summertime RFG
under the base, RFS and EIA fuel scenarios, respectively. Under the RFS and EIA cases,
there is no MTBE or TAME in the fuel, so these rows are not shown (i.e., total oxygen
content is the same  as ethanol content in terms of weight percent).

                Table 2.3-5.  RFG Fuel Properties: Base Case - Summer

RVP
Sulfur ppm
Aromatics
Benzene
Olefms
E200
E300
T50
T90
Oxygen (wt%)
MTBE (wt%O)
TAME (wt%O)
Ethanol (wt%O)
PADD 1 South
7.0
6.7
21.0
0.74
4.3
59.9
88.9
179.2
302.7
2.1
1.9
0.2
0.0
PADD 1 North
7.1
22.6
23.9
0.70
13.7
55.0
80.3
189.6
341.7
2.3
0.5
0.0
1.8
PADD 2
7.0
4.5
21.6
0.76
8.0
58.4
86.0
182.5
316.0
2.3
1.3
0.1
0.9
PADD 3
7.0
6.9
20.0
0.70
4.4
59.8
88.9
179.8
302.7
2.1
1.9
0.2
0.1
PADD 5
6.8
10.0
22.0
0.57
5.7
54.6
86.2
190.5
315.0
1.9
0.0
0.0
1.9
All U.S.
6.9
13.0
22.3
0.71
8.5
56.3
84.7
186.9
321.9
2.1
0.7
0.1
1.3
                Table 2.3-6.  RFG Fuel Properties: RFS Case - Summer

RVP
Sulfur ppm
Aromatics
Benzene
Olefms
E200
E300
T50
T90
Ethanol (wt%O)
PADD 1 South
7.0
20.5
20.1
0.72
14.6
52.0
87.5
174.7
308.8
0.0
PADD 1 North
7.0
7.6
20.0
0.70
13.6
57.6
81.9
184.2
334.6
3.7
PADD 2
7.1
21.3
17.9
0.67
17.3
54.1
81.8
185.3
334.8
0.6
PADD 3
7.0
19.8
20.0
0.70
14.1
52.0
87.5
195.7
308.8
0.2
PADD 5
6.8
9.0
22.5
0.57
5.7
54.5
86.2
190.5
315.0
1.8
All U.S.
6.9
13.6
20.5
0.66
11.9
54.5
84.9
185.9
321.0
1.6
                                      87

-------
                Table 2.3-7.  RFG Fuel Properties: EIA Case - Summer

RVP
Sulfur ppm
Aromatics
Benzene
Olefms
E200
E300
T50
T90
Ethanol (wt%O)
PADD 1 South
7.0
23.1
20.2
0.70
18.9
52.0
84.3
179.4
323.6
0.0
PADD 1 North
7.0
10.0
19.7
0.74
10.3
57.7
81.8
184.0
334.7
3.7
PADD 2
7.1
19.3
17.9
0.60
14.7
55.8
80.9
183.4
339.2
1.8
PADD 3
7.0
22.3
20.0
0.67
18.3
52.0
84.3
195.7
323.6
0.2
PADD 5
6.8
8.9
22.6
0.57
5.7
54.6
86.2
190.5
315.0
1.9
All U.S.
6.9
14.9
20.5
0.65
12.1
54.7
83.8
186.4
325.7
1.8
       As shown in Tables 2.3-6 and 2.3-7, summer RFG produced in PADD 1 under the
two increased ethanol use cases contains 10 vol% ethanol (i.e., 3.7 wt% oxygen).
However, RFG produced in the other PADDs contains less than the maximum 10 vol%
ethanol. For other parameters the results generally support the proposed assumption that
they would remain constant, although there were some small changes. The biggest
changes in the RFS and EIA fuel scenarios include higher levels of olefins, T50 and T90
and lower aromatic levels.

       Tables 2.3-8, 2.3-9, and 2.3-10 present the fuel properties of wintertime RFG
under the base, RFS and EIA fuel scenarios, respectively.

                 Table 2.3-8.  RFG Fuel Properties: Base Case - Winter

RVP
Sulfur ppm
Aromatics
Benzene
Olefins
E200
E300
T50
T90
Oxygen (wt%)
MTBE (wt%O)
TAME (wt%O)
Ethanol (wt%O)
PADD 1 South
11.2
28.0
21.1
0.74
12.6
63.6
88.9
179.2
302.7
2.1
2.1
0.0
0.0
PADD 1 North
12.9
25.5
19.0
0.70
16.0
57.4
80.3
184.7
341.7
2.5
0.5
0.0
2.0
PADD 2
12.8
26.2
13.9
0.64
11.6
66.9
79.5
165.4
345.2
3.3
0.0
0.0
3.3
PADD 3
11.2
26.9
20.0
0.70
12.2
63.1
88.9
173.1
302.7
2.1
2.0
0.1
0.1
PADDS
11.5
9.5
20.8
0.47
5.7
59.3
86.2
180.8
315.0
2.2
0.0
0.0
2.2
All U.S.
12.0
21.1
19.3
0.63
11.1
61.0
84.5
178.5
322.7
2.4
0.7
0.0
1.7
                                      88

-------
                 Table 2.3-9.  RFG Fuel Properties: RFS Case - Winter

RVP
Sulfur ppm
Aromatics
Benzene
Olefms
E200
E300
T50
T90
Ethanol (wt%O)
PADD 1 South
11.8
28.0
21.4
0.75
13.3
53.5
87.5
174.7
308.8
0.0
PADD 1 North
13.1
25.0
19.0
0.70
16.0
63.1
81.9
173.0
334.6
3.7
PADD 2
13.0
25.4
17.8
0.89
12.3
63.9
79.5
171.4
345.2
3.0
PADD 3
11.8
26.7
21.2
0.73
12.6
53.4
87.5
192.9
308.8
0.0
PADD 5
11.5
9.4
23.7
0.43
5.7
58.2
86.2
183.0
315.0
1.8
All U.S.
12.3
20.8
20.9
0.65
11.4
59.0
84.5
178.4
322.7
2.0
                 Table 2.3-10. RFG Fuel Properties: EIA Case - Winter

RVP
Sulfur ppm
Aromatics
Benzene
Olefms
E200
E300
T50
T90
Ethanol (wt%O)
PADD 1 South
11.9
27.5
22.2
0.61
14.8
52.5
84.3
179.4
323.6
0.0
PADD 1 North
12.8
25.2
19.0
0.70
16.0
63.6
81.9
172.1
334.3
3.7
PADD 2
12.9
23.8
20.7
0.95
12.3
64.5
79.5
170.1
345.2
3.7
PADD 3
11.9
25.6
21.7
0.58
13.6
52.4
84.3
194.9
323.6
0.0
PADD 5
11.5
8.5
23.7
0.43
5.7
59.4
86.2
180.6
315.0
2.2
All U.S.
12.2
19.9
21.2
0.63
11.9
60.4
83.3
177.5
327.9
2.6
       As shown in Tables 2.3-9 and 2.3-10, winter RFG produced in PADD 1 under the
two ethanol use cases contains 10 vol% ethanol (i..e., 3.7 wt% oxygen). PADD 2 winter
RFG contains 10 vol% ethanol in the EIA case. However, RFG produced in the other
PADDs and cases contains less than the maximum 10 vol% ethanol. On an annual
average basis, RFG produced for use in California contains about 5.7 vol% ethanol (2.1
wt% oxygen).  This is to be expected given the increase in NOx emissions assigned by
CARB's Phase 3 Predictive Model to blends with more than 2.1 wt% oxygen. As for the
summer cases, changes in other fuel parameters were small and mixed.

2.3.4  Estimation of County-Specific Gasoline Properties

       In order to estimate the impact of increased ethanol use and reduced MTBE use
on national emissions and air quality (described in Chapters 4 and 5), we need to estimate
gasoline properties on a county-specific basis throughout the U.S.  In support of previous
analyses of national impacts of various rules, EPA has developed a set of gasoline
                                      89

-------
specifications for each county in the U.S. for various months and calendar years.TT  We
based our analysis on the fuel quality specifications for January and July of 2008, since
2008 is the first year of full implementation of the Tier 2 sulfur standard of 30 ppm.
Some of the EPA county-level gasoline specifications were based on old data, so we
reviewed the estimates and made several modifications before applying the changes
expected due to ethanol addition and MTBE removal.

      First, we adjusted RVP values using more recent information on local RVP
programs and to reflect commingling.  Second, we revised the oxygenate content and
type in each county to match the levels estimated in Section 2.2 to be sold there under
each of the three ethanol use scenarios evaluated. Third, we adjusted the other properties
of gasoline which are affected by the oxygenate use determined in step three. These
modifications are described in more detail below.

2.3.4.1       Adjustments to RVP Levels Prior to Oxygenate Use

      Our review of the NMIM database of county-specific RVP levels for July
indicated that the same RVP level was often applied to all the counties of a specific state.
In many cases, this appeared reasonable, since the same RVP standard applied throughout
the entire state. However, in other cases, for example, Florida, most counties have a 9.0
RVP standard, while those comprising several large urban areas have a 7.8 RVP standard.
The RVP levels in the NMIM database were consistent with the 7.8 RVP control
programs, implying that the 7.8 RVP fuel was sold throughout  the entire state. This was
true for much of the south.

      As mentioned above, the NMIM fuel quality database was based primarily on fuel
survey data from 1999.  Fuel surveys tend to focus on large urban areas, as opposed to
smaller urban or rural areas.  Thus, the only available fuel survey data was likely from the
areas with the tighter local RVP controls. RVP control reduces gasoline supply, since
lighter hydrocarbons must be removed in order to reduce RVP. Some, but not all of these
hydrocarbon components can be moved to higher RVP fuel sold elsewhere.  Obviously
gasoline prices are now much higher than they were in 1999. So  the incentive to increase
supply is greater now than in 1999.  As discussed in Chapter 7, high gasoline prices are
projected for the foreseeable future, at least relative to those existing in 1999. Thus, we
believe that it is reasonable to project that refiners will market gasoline blends with as
high a level of RVP as practical given the applicable standards. For example, in Florida,
two fuels will be marketed: one to meet the  7.8 RVP standard in several urban areas and
another to meet the 9.0 RVP standard applicable elsewhere. There certainly could be
some  spillover of the 7.8 RVP fuel into adjacent 9.0 RVP counties.  However, we lack
data indicating the degree to which this is occurring and might  occur in the future.
Lacking this data, it seems more reasonable to project only that level of RVP control
which is guaranteed by the applicable standards than to assume that refiners will over-
comply with RVP standards and reduce the  volume of gasoline which they can produce.
                                       90

-------
       Past studies have shown that a typical compliance margin for RVP is about 0.3
psi. Thus, for those counties where the standard 9.0 RVP standard applies, we set the
July RVP level to 8.7 psi.

       EPA maintains a list of counties where its 7.8 RVP  standard applies, as well as
any local standards more stringent than 9.0 RVP.UU Using this list, we assigned RVP
values in each county equal to 0.3 psi less than the  standard applicable in July. We also
reduced the RVP levels of two sets of counties which had voluntary local RVP control
programs (and therefore not listed the above Guide).  These two  areas were Seattle and
Tulsa.  Based on a review of annual fuel survey data collected by the Alliance of
Automobile Manufacturers (AAM)W,  the fuel being sold in these areas was very similar
to that for an area with a 7.8 RVP standard.  Thus, we assigned a value of 7.5 psi RVP to
Tulsa County, Oklahoma, and to King, Pierce, and  Snohomish Counties, Washington.

       We then assigned an RVP value of 6.8 psi to counties subject to the Federal RFG
program, again based on an EPA list of the counties subject to this program.ww The EPA
list of RFG counties includes the Baton Rouge, Louisiana, area.  However, litigation has
held up implementation of this program, so these counties were assigned RVP values
consistent with the currently applicable 7.8 RVP standard instead.  The RVP value of 6.8
psi was typical for the RFG areas included in the AAM fuel surveys.

       For the purposes of our analysis, we also assigned the entire State of California an
RVP of 6.8 psi,  since California fuel must meet a similar VOC performance standard to
RFG.  Likewise, RVP in Maricopa and Final counties in Arizona were assigned a level of
6.8 psi.  These two counties are subject to Arizona's unique reformulated gasoline
program. This program basically requires that gasoline sold in these two counties meet
either the California RFG or Federal RFG standards.  Thus, RVP in these two counties
will be the same as in those other two areas, similar to national RFG fuel.

       These RVP levels for  9.0 RVP  and low RVP areas are appropriate when no
ethanol is being blended into  gasoline.  However, most of these areas increase the
applicable standard by 1.0 psi for ethanol blends, which is the typical impact of ethanol
blending. Therefore, these levels need to be adjusted for the expected level of ethanol
use, which is discussed below.

2.3.4.2       County-Specific Oxygenate Type and Content

       The three ethanol use  scenarios developed in Section 2.2  assign ethanol and
MTBE use by state and fuel type (i.e., conventional gasoline, RFG, oxyfuel). In order to
develop county level estimates of ethanol and MTBE use, we simply assume that ethanol
and MTBE use within a state  and fuel type is uniform. For example, if the E10 market
share in conventional gasoline Iowa is  34%, then ethanol use in every county receiving
conventional gasoline in Iowa was assigned an E10 market share of 34%.

       As described above, we nearly always assume that ethanol use is in the form of a
10 vol% blend with gasoline.  The two exceptions are California fuel and Arizona RFG.
                                       91

-------
California fuel containing ethanol is assumed to contain 5.7 vol% ethanol.  Arizona RFG
is assumed to be a mix of 67% California fuel and 33% Federal RFG produced in PADD
3.  Therefore, its ethanol content is a 2/1 mix of the ethanol contents of California RFG
and PADD 3 Federal RFG.

       Similarly, we assume that MTBE is used at an 11 vol% level in RFG, since this
meets the previously mandated oxygen content of 2.1 wt%. MTBE in conventional
gasoline was assumed to be used at a 3 vol% level. This was somewhat arbitrary, but
does not affect the outcome of the analysis.  The effect of MTBE blending on emissions
is very linear. Therefore, whether the fuel pool in a particular area consists of 10% of a
10 vol% MTBE blend or 33%  of a 3 vol% MTBE blend is  immaterial. Though MTBE is
present in the 2012 Reference Case, it is assumed to be completely phased-out in the RFS
and El A ethanol use cases.

       EPA's NMIM model (described in more detail in Chapter 4) will only accept a
single composite fuel for each  county.  Therefore, we could not use the mix of fuels often
projected to be supplied to counties developed in Section 2.2.  In order to produce a
single, composite fuel, we simply multiplied the ethanol  and MTBE contents of each
blend by their market share in that county in order to determine the average ethanol and
MTBE contents of each county's fuel pool, respectively. For example, if the E10 market
share in a specific county was 50%, the ethanol content for that fuel was set to 5 vol%.
We then adjusted the other fuel properties to account for these oxygenates, which is
discussed below.

2.3.4.3       Adjustments to Other Gasoline Properties for Oxygenate Use

       We next adjusted other gasoline properties to account for the level of county-
specific oxygenate use projected to occur under the three ethanol use scenarios.  Our
review of the NMIM fuel database indicated that properties, such as aromatics, reflected
the level of oxygenate use existing in 1999.  Therefore, we used the oxygenate levels in
the NMIM database, which differ from those developed in  Section 2.2 for 2004, as the
basis for our adjustments of the other fuel properties.  For example, if the NMIM
database indicated an ethanol content of 3 vol% for fuel sold in Wayne County,
Michigan,  and the 2004 projection for this county was 5 vol%, we adjusted the NMIM
fuel properties for this county to reflect the addition of 2 vol% ethanol.

       The bases for these adjustments were those developed in Sections 2.3.1 through
2.3.4 above.  As described there, these adjustments apply primarily to conventional
gasoline.  These adjustments are  summarized in Table 2.3-11 below.
                                       92

-------
                                  Table 2.3-11
      Change in Property per 1 Vol% Increase in Ethanol and MTBE Content

E200 (%)
E300 (%)
Aromatics
(Vol%)
Olefins (Vol%)
RVP (psi)
Conventional Gasoline
Ethanol
MTBE
+1.0
+0.52
+0.24
+0.17
-0.5
-0.59
-0.16
0
+0.1
0
Reformulated Gasoline
Ethanol
MTBE
0
0.1
0
0.1
0
0
0
0
0
0
       To calculate new fuel properties for each county, we applied the ethanol and
MTBE factors to the change in county-level ethanol and MTBE content. The overall
adjustment to the fuel property was the addition of the ethanol effect and the MTBE
effect to the baseline fuel property.

       For the impact of ethanol blending on aromatic and olefin contents, we followed a
slightly different approach.  We assumed that the ethanol present in 1999 had been
splash-blended, while that being used in the future will be match-blended.  This
difference doesn't affect the adjustment of RVP, E200, or E300, since we assume that
these parameters are affected in the same way regardless of whether the ethanol is splash-
or match-blended.  However, the change in aromatics does depend on which blending
approach is used. The situation is similar for olefms, though to a lesser extent. Thus, we
employed what can be thought of as a two step  process in adjusting aromatic and olefin
contents for the change in ethanol content between the NMIM estimate and those for the
three ethanol use scenarios developed in Section 2.2.

       The first step is to account for any splash-blended ethanol in the NMIM database.
With splash-blending, aromatic and olefin contents are reduced simply by dilution, since
ethanol contains is neither an aromatic nor an olefin. The following equation shows how
the NMIM level of aromatics was adjusted:
            Intermediate     NMIM
            Aromatic   = Aromatic
             Content       Content
NMIM
Ethanol
        ,,,„,.
 /    /  NMIM \  /    ,   .  \     \
/    /   ,   , \ /  Ethanol  \ .    \
(-(=)(  —  r°°)
       Then, the effect of any ethanol projected to be sold in that county in the three
ethanol use scenarios developed in Section 2.2 was applied using the approach described
above for RVP, E200 and E300 (and for the effect of MTBE on aromatics and olefms).
In this case, the NMIM ethanol content and market share is zero, since we already
adjusted the NMEVI aromatic and olefin contents to represent those existing for a zero
ethanol content.  For example, the equation for the ethanol effect is as follows:
                                      93

-------
       New Fuel       Intermediate
       Property    =  Fuel Property +
        Level          Level
                                                                Fuel Property
 x   RFS    ^ /   RFS   x     s
/   Ethanol   V  Market   \ X  /   C1™e r '   \
I           II          I    I   vol% Ethanol    I
V  C°ntent   A   Shar£  /    V    Increase    /
       We make one final adjustment to RVP to add a commingling effect to account for
areas where vehicles may be fueled by a mix of ethanol-blend gasoline.  Commingling of
ethanol and non-ethanol blends can increase the average RVP of gasoline in vehicle fuel
tanks by 0.1-0.3 psi. Appendix 2-A presents a detailed analysis of the impact of
commingling on the RVP of gasoline in vehicle fuel tanks. Table 2.3-12 presents our
estimate of the net impact of commingling on in-use RVP as a function of the market
share of ethanol blends.
Table
2.3-12. Impact of Ethanol Blends on In-Use RVP (psi)
E10 market share
0%
2%
5%
10%
20%
30%
40%
50%
60%
70%
80%
90%
97%
100%
Commingling Impact
0
0
0.116
0.116
0.202
0.238
0.264
0.273
0.263
0.226
0.172
0.102
0.102
0.000
       EPA's MOBILE6.2 model normally accounts for this effect automatically.
However, when NMEVI is used to run MOBILE6.2, the commingling effect in
MOBILE6.2 is by-passed.  Therefore, any effect of commingling needs to be accounted
for in the average fuel specified to be sold in each county.  To roughly account for this
effect, we increased RVP by 0.1-0.27 psi in all states where the E10 market share was
significant (i.e., more than 5%) but less than 95%. The states which fell into this
category, for CG and RFG, are shown in Table 2.3-13.  The specific RVP increase
depended on the ethanol market share in that county, as indicated in Table 2.3-12.
                                        94

-------
      Table 2.3-13. States Where RVP was Increased Due to Commingling
Fuel Case
                    Conventional Gasoline
Reference
ILLINOIS        SOUTH DAKOTA   INDIANA      NEBRASKA
OHIO           IOWA            KANSAS       NORTH
DAKOTA  WISCONSIN     ALABAMA        MICHIGAN
MISSOURI
  RFS
ALABAMA
COLORADO
LOUISIANA
NEW MEXICO
WASHINGTON
   ARIZONA        ARKANSAS
   FLORIDA         KENTUCKY
  MAINE        MISSISSIPPI       NEVADA
    PENNSYLVANIA   TENNESSEE
  WYOMING
   EIA
COLORADO        FLORIDA        KENTUCKY
LOUISIANA      PENNSYLVANIA   TENNESSEE
WASHINGTON   WYOMING       IDAHO           NEW
HAMPSHIRE NEW JERSEY     NEW YORK       OKLAHOMA
OREGON         TEXAS          UTAH
                              Reformulated Gasoline
Reference
ARIZONA
    MASSACHUSETTS NEW JERSEY
TEXAS
                            PENNSYLVANIA
                            NEW YORK
  RFS
KENTUCKY
NEW JERSEY
MAINE        CONNECTICUT     DELAWARE
INDIANA       MASSACHUSETTS   MISSOURI
ISLAND  VERMONT       WISCONSIN
                     NEW HAMPSHIRE
                     TEXAS
                                    ILLINOIS
                                     RHODE
   EIA
KENTUCKY
NEW JERSEY
MAINE
ILLINOIS
MISSOURI
WISCONSIN
     PENNSYLVANIA   NEW HAMPSHIRE
     NEW YORK       TEXAS
CONNECTICUT      DELAWARE
  INDIANA       MASSACHUSETTS
   RHODE ISLAND  VERMONT
2.4   Effects of Biodiesel on Diesel Fuel Properties

      Our assessment of the effects of biodiesel on diesel fuel properties is found in the
2002 EPA report "A Comprehensive Analysis of Biodiesel Impacts on Exhaust
Emissions"xx. Table 2.4-1 below displays the difference in fuel properties between
biodiesel (B100) and conventional diesel.  Note that by 2010, all highway and nonroad
diesel fuel will meet a 15 ppm cap on sulfur.

      The data in the table below were derived from a wide-range of biodiesels,
primarily plant- and animal-based. The 2002 EPA report did not provide properties for
soy-only based biodiesel.
                                 95

-------
Table 2.4-1. Comparison Between Biodiesel and Conventional Diesel Fuel3

Natural cetane number
Sulfur, ppm
Nitrogen, ppm
Aromatics, vol%
T10, degF
T50, deg F
T90, deg F
Specific gravity
Viscosity, cSt at 40 deg F
Average Biodiesel
55
54
18
0
628
649
666
0.88
6.0
Average Diesel
44
333
114
34
422
505
603
0.85
206
""Conventional diesel fuel sold outside of California.
                                                 96

-------
                           Chapter 2:  Appendix
              Comprehensive Vehicle Refueling Model

       Vehicle refueling patterns affect non-exhaust emissions in a number of ways,
including the distribution of vehicle fuel tank fill levels existing at any given time and the
quality of fuel  in the tank.  Given the interaction between these parameters, we have
developed a single model which represents vehicle refueling patterns.  We then use this
model to estimate the distribution of vehicle fuel tank fill levels and the quality of fuel in
the vehicle fuel tanks.

       Vehicle fuel tank fill levels are primarily a function of the  level at which people
refuel their vehicles and the volume of fuel which they add. In-use, vehicle fuel tanks
will slowly empty until the point of refueling again.  The California Air Resources Board
(CARB) recently conducted a survey of vehicle refueling patterns in three California
cities. We will base our estimates of refueling patterns primarily on these data.

       Most fuel parameters remain unchanged as the fuel is burned. One except is
volatility, particularly RVP, which decreases due to evaporation of the fuel as  the tank
heats up either due to rising ambient temperatures or vehicle operation (e.g., heat transfer
from the exhaust system, engine cooling air flowing under the vehicle, and fuel
recirculation from the engine compartment).

       While ethanol content doesn't change significantly while the vehicle is being
operated, the ethanol content of gasoline in vehicle fuel tanks can  be a function of vehicle
refueling patterns if some of the specific gasolines being marketed in an area contain
ethanol  and some do not. The effect of ethanol on RVP is not linear.  Thus, knowledge
of the distribution of ethanol  content in vehicle fuel tanks is important in estimating the
RVP of gasoline in vehicle fuel tanks and non-exhaust emissions.  We use the vehicle
refueling model to estimate ethanol content, fuel RVP, and average fill level.

       There are four main aspects of the vehicle refueling model. The first two aspects
affect all types of gasoline, ethanol containing or not. The first aspect is a description of
the refueling patterns of vehicle operators. How low is the tank when they refuel?  How
much fuel do they add? Does the volume of fuel added depend on how low the tank was
when they stopped to refuel? The second aspect is the weathering of the fuel as the
vehicle is operated. In general, the degree of weathering, or RVP  reduction, depends on
both the ambient temperature and initial RVP of the fuel.

       The third aspect of the model is the effect of ethanol on RVP. While the ethanol
content  of gasoline tends to be either 5.7 or 10 percent by volume  (vol%) at the service
station, the ethanol content of gasoline in a vehicle's fuel tank can vary from zero to 10
vol%. The fourth aspect of the model is a description of the probability that a vehicle
operator will purchase fuel at the same service station as the last refueling or at another
outlet selling the same brand fuel (i.e., gasoline brand loyalty).  Brand loyalty  is relevant,
because service stations carrying the same brand of gasoline almost always sell either
gasoline with ethanol or gasoline without ethanol, but not both. It is the mixing of
                                       97

-------
gasoline with and without ethanol is vehicles' fuel tanks that can cause the RVP of fuel in
the tank to differ from that dispensed at the service station.  This is referred to as the
commingling effect.

       Each of these four aspects of the vehicle refueling model is described below.

2A.I   Vehicle Refueling Patterns

       During August and September, 2001, the CARB surveyed consumers' refueling
habits at 19 service stations in three local areas (Lake Tahoe, the Bay Area and Los
Angeles).38 Basic refueling information was obtained for 396 vehicle refuelings (i.e.,
initial fuel tank level and volume of fuel added). Fuel samples were also obtained from
254 vehicles, though we are most interested in the volumetric data here.  CARB also
asked those refueling whether they  refueled with the same brand gasoline the last time the
vehicle was refueled.

       We obtained and analyzed the raw volumetric fuel data obtained by CARB. Of
the 396 sets of data, 391 included both initial fuel tank level and volume of fuel added.
One of the two pieces of information was missing for five vehicles, so we discarded these
partial data sets from the analysis.  The tank fill level prior to refueling was recorded in
terms of eighths of a fraction of a full tank, as this is usually how the tank fill level is
indicated on the vehicle dash board. Table 2A-1 shows the probability of a vehicle being
refueled at various fuel tank fill levels.

                     Table 2A-1. Fill Level Prior to Refueling
Fraction of fill level
0.000
0.125
0.250
0.375
0.500
0.625
0.750
0.875
Probability
0.414
0.133
0.253
0.054
0.095
0.020
0.020
0.010
       As can be seen, over 40% of the vehicles surveyed came in with an "empty" tank.

       CARB also recorded whether the vehicle operator "filled up" the tank or not. We
observed that there was a trend towards a greater probability of a "fill up" as the level of
the tank prior to refueling increased. Table 2A-2 shows the probability of a fill-up as a
function of tank fill level prior to refueling.
38  "Draft Assessment of the Real-World Impacts of Commingling California Phase 3 Reformulated
Gasoline, California Environmental Protection Agency, Air Resources Board, August 2003.
                                        98

-------
Table 2A-2.
Probability of Fill-Up as a Function of Initial Tank
Initial Tank Fill Level
0
0
0
0
0
0
0
0
000
125
250
375
500
625
750
875
Likelihood of a
Fill Leve
Fill-up
0.117
0.500
0.586
0.619
0.784
0.875
0.750
0.750
       Overall, when the tank was at least half full, the tank was filled up 79% of the
time.
       In those cases where the fuel tank was not filled to capacity, the volume of fuel
added was recorded in terms of gallons.  Therefore, some processing was required to
estimate the final fill level in terms of fraction of tank capacity. To do so, we had to
estimate the volume of each vehicle's fuel tank.  CARB recorded the basic model type of
each vehicle in the survey.  Based on this model type, we placed each vehicle into one of
six possible categories.  First, each vehicle was identified as either a car or light truck.
Then, we estimated whether it would have a relatively small, medium, or large fuel  tank
for that vehicle class.  The fuel tank sizes assumed for each class are shown in Table 2A-
3 below.

               Table 2A-3. Estimated Fuel Tank Volumes (gallons)
Relative Size
Small
Medium
Large
Car
12
16
20
Light Truck
16
20
24
       Using these tank volumes, we converted the volume of fuel added during partial
fill ups to an equivalent fraction of tank volume and added this to the observed initial fill
level to estimate the final fill level. In five cases (out of the 176 partial fills), the
estimated final fill level exceeded 100%. Either the initial gauge reading was off or
rounded up, or more likely, our estimate of the total tank volume was too small. In these
cases, we reduced the final fill level to 95%.  (Given that this was a partial fill-up, the
final fuel tank level had to be less than 100%.)

       For all of the partial fill-ups, we converted the volume of fuel added from gallons
of fuel to fractional tank volume.  Both the mean and standard deviation of these volumes
were determined as a function of initial fill level.  These figures  are shown in Table 2A-4.
                                       99

-------
      Table 2A-4. Volume of Fuel Added During Partial Fills (% of Fill Level)
Initial Fill Level
0.000
0.125
0.250
0.375
0.500
0.625
0.750
0.875
Volume of Fuel Added
Mean
0.406
0.434
0.382
0.451
0.314
0.325
0.225
0.030
Standard Deviation
0.200
0.157
0.167
0.185
0.120
—
0.04
—
       As can be seen from Table 2A-4, the final fill level from partial fills is very close
to a full tank when the initial fuel tank level was 0.625 or higher. The actual number of
cases when vehicles initiated refueling when the tank fill level was 0.625 or higher was
quite small (20). The number of partial fills surveyed was even smaller (4). Given this
small number and the fact that the fraction of fill-ups for half full tanks exceeded that
found for 0.75 and 0.875 full tanks (from Table 2A-3), we assumed that all tanks which
were at least half full when refueling was initiated were filled-up.  When the initial fill
level was less than 0.5, we assumed that the mean volume of fuel added were those
shown in Table  2A-4.

       As also can be  seen from the figures in Table 2A-4, the estimates of the standard
deviation in the volume of fuel added are substantial relative to the mean volumes of fuel
added. We desired to reflect this variability in the volume of fuel added during partial
fills. Thus, we utilized both the estimates of the mean and standard deviation in the
volume of fuel during refueling.  We accomplished this by multiplying the standard
deviation by a randomly generated standard normal deviate and adding this to the mean
volume of fuel added to estimate the volume of fuel added during each partial refueling.

2A.2  Weathering

       Fuel weathering is the result of the evaporation of the lighter components of
gasoline when the temperature in the fuel tank rises.  This temperature rise can be the
result of diurnal swings in ambient temperature or from vehicle operation.  In the latter
case, the heat can be transferred either convectively from the exhaust system and engine
cooling air flowing under the vehicle or conductively from recirculated fuel from the
engine's fuel system or both. Gasoline is a mixture of many different chemicals. Some
of these chemicals, such as butane, evaporate more quickly than other chemicals with a
higher molecular weight, such as octane.  The loss of lighter chemicals can be sufficient
to reduce the concentration of these lighter chemicals in the liquid gasoline. This reduces
the RVP of the fuel and its tendency to evaporate as the current tank of fuel is consumed.

       We base our estimate of weathering on RVP on the methodology currently in
MOBILE6.2. This estimate was first developed for MOBILE4 and was also used in
                                       100

-------
MOBILES. This methodology first calculates an effective in-use tank temperature (TeVaP)
which drives fuel evaporation. This temperature is a function of the daily minimum
temperature (Tm;n) and maximum temperature (Tmax), as indicated in the following
equation:

     = -1.7474 + 1.029*Tmm + 0.99202*(Tmax - Tmm) - 0.0025173 * Tmm*(Tmax - Tmm)
       The loss in RVP is a function of both TeVaP and dispensed RVP, as indicated by
the following equation:

RVP reduction (psi) due to weathering = -2.4908 + 0.026196 * Tevap
                                        + 0.00076898 * TeVaP * Dispensed Fuel RVP
       This RVP loss is that occurring when the vehicle fuel tank is 54.57% full, which
is the effective in-use average tank level for estimating non-exhaust emissions in
MOBILE6.2. For a typical high ozone day where the ambient temperature might range
from 72 F to 96 F, and for a dispensed RVP of 9 psi, the RVP loss due to weathering is
0.54 psi. In order to estimate weathering at other tank fill levels, we assume that
weathering is linear with tank fill level.

2A.3   Effect of ethanol on RVP as a Function of Ethanol Content

       In general, the chemicals comprising gasoline blend ideally.  That is, the property
of the finished gasoline is the sum of the property of each component weighted by its
molar, volume or mass fraction, whichever is technically appropriate. Each component
of gasoline has its own RVP level. Adding a component to gasoline at the level of 10
volume percent (vol%), which is the typical ethanol concentration would increase the
blend's RVP by 10% of the component's RVP and decrease the blend's RVP by 10% of
the original RVP.  For example, normal butane with an RVP of 42 psi can be added to
gasoline with an RVP of 8 psi. If the butane is added to a final level of 5 vol%, then the
final RVP is 0.05 times 42 plus 0.95 times 8, or 9.7 psi.

       Ethanol blending affects the RVP of the finished gasoline quite differently.
Ethanol is a highly polar compound, due to the presence of the hydroxyl radical. In pure
liquid ethanol, these hydroxyl radicals interact with each other, increasing the degree of
attraction between ethanol molecules and lowering their tendency to evaporate.  This
phenomena is commonly known as hydrogen bonding and is most commonly associated
with water. When added to non-polar hydrocarbons at low concentrations, such as those
comprising gasoline, the evaporative tendency of ethanol increases dramatically. This
increase in vapor pressure is indicated by what is referred to as the activity coefficient.
The activity coefficient is the ratio of a compound's actual vapor pressure in a mixture to
that predicted by ideal blending.  Table 2A-5 shows ethanol's activity coefficient at
various levels of concentration in a typical gasoline.
                                      101

-------
          Table 2A-5.  Activity Coefficient of Ethanol in Gasoline Blends 39
Ethanol Concentration (vol%) *
3%
6%
10%
14%
18%
Activity Coefficient
7.5-8.0
3.8-4.1
2.3-2.5
1.9-2.0
1.6-1.8
       As can be seen, these activity coefficients are substantially greater than 1.0,
indicating a significant increase in the vapor pressure of ethanol beyond that predicted by
ideal blending.

       Adding ethanol to gasoline can also increase the vapor pressure of the
hydrocarbon components.  In general, instead of the hydrocarbons' vapor pressures
decreasing with the addition of another component (e.g., by 10% with the addition of 10
vol% ethanol), they remain constant or even increase.  This could be due to a tendency of
the hydrocarbons in the vapor phase to "bounce" off of the ethanol molecules at the
surface of the liquid phase.

       A number of studies have shown that the full effect of ethanol's impact on RVP is
reached at very low concentrations.  For example,  a study performed by the Energy and
Environmental Research Center at the University of North Dakota indicates that 90% of
the full impact of ethanol on RVP is reached when the ethanol concentration is only 2
vol%.40  Researchers at the University of Delaware found the same relationship.41 Below
2 vol% ethanol, we have assumed that the effect is essentially linear.

       The full effect of ethanol on gasoline RVP is a function of the RVP of the base
hydrocarbon gasoline.  In general, the increase in RVP caused by ethanol blending
increases as the base RVP decreases. The actual RVP of specific commercial ethanol and
non-ethanol gasoline blends are generally known for a specific area being modeled.
Thus, they are not a primary concern here. However, in order to develop realistic
estimates of weathering and commingling across a range of ethanol blend market shares,
39  Harley, Robert A. and Shannon C. Coulter-Burke, "Relating Liquid Fuel and Headspace Vapor
Composition for California Reformulated Gasoline Samples Containing Ethanol," Environmental Science
and Technology, Volume 34, Number 19, 2000.  pp 4088+4094, Figure 3. It should be noted that the
ethanol concentrations shown in the reference are in terms of mole fraction, which are essentially a factor
of 2 higher than volume fraction.

40 Aulich, Ted and John Richter, "Addition of Nonethanol Gasoline to E10 - Effect on Volatility," Energy
and Environmental Research Center at the University of North Dakota, July 15, 1999.

41  Bennett, Alison, Stephan Lamm, Hasan Orbey and Stanley I. Sandier, "Vapor-Liquid Equilibria of
Hydrocarbons and Fuel Oxygenates. 2," Jounal of Chemical Engineering Data, Volume 38, 1993, pp. 263-
269, Figure 7. This reference shows the ethanol  concentration in terms of mass fraction, which is nearly
identical to volume fraction.
                                         102

-------
it would be helpful to use realistic RVP levels for commercial ethanol and non-ethanol
gasoline blends.
       Ethanol blending generally occurs under two types of RVP standards. One type
of standard requires that both ethanol and non-ethanol blends meet the same RVP
standard. This is the case in reformulated gasoline (RFG) areas. In most other areas,
ethanol blends are allowed to meet an RVP standard 1.0 psi higher than that applicable to
non-ethanol blends.

       We will estimate the impact of weathering and commingling for both situations.
For RFG-like situations, we will assume that both ethanol and non-ethanol blends have
the same RVP. In order to estimate the impact of ethanol blending in areas where ethanol
blends are allowed to have a higher RVP level, we evaluated recent fuel quality data
collected by the Alliance of Automobile Manufacturers. We combined the data collected
from 2001-2003 and found six cities where significant numbers of both ethanol and non-
ethanol blends were sampled and analyzed.  These six cities were: Albuquerque,
Cleveland, Denver, Detroit, Minneapolis, and  Seattle.  The average RVP levels of the
non-ethanol gasoline samples in each city ranged from 7.46-9.00 psi, while that of the
ethanol blends ranged from 8.50-9.93 psi. We observed a relationship between the RVP
of the non-ethanol gasolines and the difference between the RVP of the ethanol and non-
ethanol blends. In general, as the RVP of the non-ethanol gasoline increased, the
difference between the RVP of the ethanol and non-ethanol blends decreased. Figure 2A-
1 shows this relationship for the six cities, along with a best-fit line based on least squares
regression.  The r-squared value for the best-fit line was 0.25.

        Figure 2A-1.  Effect of Ethanol Blending on RVP in Six U.S. Cities
Ethanol Blend RVP
Difference
1
1
1
0
0

9
.£.


•
•
^^^^^*
y = -0.1538x+2.3606
R2 = 0.2508

7.0 7.5 8.0 8.5 9.0 9.5
Non-Ethanol Blend RVP
       On average, ethanol blending led to a 1.0 psi RVP higher RVP level when the
RVP level of the non-ethanol gasoline was 9.0 psi. Ethanol's effect increased to 1.2 psi
                                       103

-------
when the RVP level of the non-ethanol gasoline was 7.5 psi. When evaluating the effect
of weathering and commingling, we will evaluate non-ethanol gasoline RVP levels of
6.8, 7.8 and 9.0 psi, as these are common RVP standards in use today. Using the
relationship indicated in Figure 1, the RVP levels of the ethanol blends associated with
these base RVP levels are 8.1, 9.0, and 10.0 psi, respectively.

2A.4  Brand Loyalty

       CARB recently conducted a fairly extensive direct survey of vehicle refueling
patterns.  This study is both more recent and more extensive than those made in the
past.YY During their refueling survey, CARB asked vehicle operators whether the vehicle
was refueled with the same brand of fuel the previous time the vehicle was refueled.  The
resulting responses are summarized in Table 2A-6 below.

            Table 2A-6.  Brand Loyalty in the CARB Refueling Survey
Response to Question: Did you refuel
with the
"Yes"
"No"
"Don't Know"
Los Angeles
62%
31%
7%
Bay Area
59%
38%
3%
Lake Tahoe
31%
67%
2%
Breakdown of Retail Outlets Surveyed
Major Brands (Shell, Chevron,
Texaco, Mobil, ARCO)
Intermediate (Valero)
Local Brands (USA, Fox, United)
4 (100%)
0 (0%)
0 (0%)
5 (84%)
1 (16%)
0 (0%)
3 (33%)
0 (0%)
6 (67%)
       CARB thought that the relatively low level of brand loyalty in Lake Tahoe was
due to a high rate of rental car usage in that city. However, they did not present any data
regarding the fraction of total VMT by rental vehicles to justify their rationale.  Rental
vehicle VMT would have to represent roughly half of all VMT in the Tahoe area
(assuming such use was negligible in the other two areas) for this factor to explain the
large difference seen in Table 2A-6. This seems quite unlikely, despite Lake Tahoe being
a resort area.

       We believe that there is a more likely explanation for the difference. A review of
the service stations surveyed in the three areas shows significant differences in the types
of brands surveyed.  The service stations surveyed in Los Angeles and the Bay Area were
dominated by major, nationally recognized brands.  Those in Lake Tahoe were  dominated
by more local brands.  We believe that brand loyalty could easily be stronger for
nationally known brands which advertise and  which offer their own credit cards.  A few
major brands offer a significant discount when their credit card is used to buy their
gasoline (e.g., Shell, BP).

       The breakdown of service stations into major and local brands in the three areas is
shown in the lower half of Table 2A-6. We have defined major brands to include
                                       104

-------
vertically integrated oil companies which have been in the retail business for several
decades, market in several regions across the U.S. and are known to widely advertise. As
shown in Table 2A-6, all four retail fuel outlets surveyed in Los Angeles fell into this
category, while five out of six stations in the Bay Area reflected major brands.  The sixth
outlet in the Bay Area was a Valero outlet.  Valero is a newcomer in the retail market
relative to Chevron, Shell, etc.  However, it is currently the largest refiner in the U.S. and
offers its own credit card. Thus, Valero appears to fall into an intermediate category
somewhere between the major brands and the local brands.

      The situation is  essentially reversed in Lake Tahoe.  Two-thirds of the retail
outlets surveyed were local brands. Only three out of nine outlets represented major
brands.

      In order to investigate the potential difference between brand loyalty between
major and non-major brands, we assumed that each type of fuel brand had its own level
of loyalty across the three areas. We then estimated these two levels of brand loyalty
identified to best predict the overall brand loyalty in each area. Overall, loyalty levels of
62% for major brands and 15% for non-major brands (including Valero) fit the survey
data reasonably well.

      The U.S. Energy Information Administration (EIA) tracks a broad range of
proprietary physical and financial data from large energy producers through their
Financial Reporting System.  This information includes the volume of gasoline sold
through retail outlets owned by or leased from these companies and  selling fuel with the
company's brand name. Up to  1997, these retail outlets sold about 45% of all gasoline
sold in the U.S. In 1998, EIA expanded the number of companies included in the
Financial Reporting System by 50% (from 22 to 33 companies).  The percentage of
gasoline sold by these firms' retail outlets increased to 62% of all gasoline sold in the
U.S.

      The nature of the firms included in the Financial Reporting System changed with
the  eleven companies added in  1998. Prior to 1998, this system included 22 companies.
A few of these firms were not oil companies, (e.g., Burlington Resources, Enron,
Sonat/El Paso Energy, Union Pacific Resources, USX). Several others  were not major
gasoline retailers, at least under their corporate names (e.g., Anadarko Petroleum, Kerr-
McGee, Occidental Petroleum).

      In 1998, EIA added an additional 11 companies to the Financial Reporting
System. Most of these were gasoline refiners (e.g., Citgo, Clark, Equilon, Lyondell-
Citgo, Motiva, Sunoco, Tesoro, Ultramar Diamond Shamrock (UDS), Valero and
Williams).  At the same time, the volume of gasoline sold by the original 22 companies
decreased to 31% of all gasoline sold in the U.S.  This drop is likely due to the  spin-off of
refineries by companies like Shell and Texaco to partnerships like Motiva and Equilon.
The actual retailing of this 14% of gasoline sales likely didn't change significantly (e.g.,
the  retail brand name continued to be Shell). The net increase of 17% of U.S. gasoline
sales represented the other refiners, such as Tesoro, Valero, Citgo, USD, etc. These latter
                                       105

-------
companies have a much more regional footprint and have not established brand name
familiarity coupled with a perception of higher quality gasoline. With the exception of
the single Valero outlet in the Bay Area, none of the stations surveyed by CARB offered
gasoline from these companies. Thus, we believe that the major brands included in the
CARB survey are more similar to the fuel suppliers included in EIA's Financial
Reporting System prior to 1998 than to those included after 1998.

       Given this, we estimate that 45% of U.S. gasoline sales are sold through stations
carrying a major brand. Weighting the loyalty levels of 62% for major brands by 45%
and the loyalty level of 15% for other brands by 55% yields an overall national average
loyalty level of 36%. For non-loyal consumers, the probability of brand selection is
assumed to be random. Practically, this means that the probability of choosing either a
non-ethanol or ethanol blend depends on each fuel's market share.

       The exact question asked by the CARB surveyors was  whether the vehicle had
been refueled the last time with the same brand of fuel. CARB assumed that this meant
that the vehicle was always refueled with the same brand. However, the question was
limited to only the refueling immediately preceding the current one.  Our primary
estimate of brand loyalty is that directly addressed by the CARB survey: the likelihood
that the previous refueling was with the same brand of gasoline. We also estimate the
sensitivity of our estimate of commingling to the assumption made by CARB below.

2A.5   Procedures for Modeling Vehicle Refueling and Resultant Fuel Quality

       We developed a model to predict the fuel tank level and fuel quality existing in a
typical onroad vehicle through 500 refuelings. The vehicle is  assumed to begin its life
with a full tank of non-ethanol fuel.   The fuel tank level at which the vehicle is refueled
is based on the probabilities shown in Table 2A-1 above.  First, a cumulative distribution
of refueling probabilities was generated by adding the probabilities shown in the second
column of Table 2A-1. Then,  a random number valued between 0.0  and 1.0 is generated.
If the random number is less than the cumulative probability of the tank being empty at
refueling (0.414), the tank is assumed to be empty. If the random number is between this
figure and the cumulative probability of the tank being 1/8 full at refueling (0.547),  the
tank is assumed to be 1/8 full at refueling, etc.

       The RVP level  of this fuel is reduced using the weathering equation shown in
Section 2A.2. The  level of RVP loss is assumed to be proportional to the volume of fuel
used since the last refueling. For example, a vehicle might be  driven from  a full fuel tank
down to a tank which is 20% full. Fuel usage is 80% of a tank. The above RVP
weathering equation represents the RVP drop for a vehicle being driven from a full  tank
down to a 60% full tank, or a fuel usage of 40%. Therefore, the RVP decrease due to
weathering in this case would be twice that indicated by the weathering equation in
Section 2A.2. Fuel composition (i.e., ethanol content) is assumed to be unaffected by
driving.
                                       106

-------
       Only the hydrocarbon portion of the fuel is weathered, since the effect of ethanol
on RVP is essentially independent of its concentration. Thus, the value of RVP used in
the weathering equation is that for the hydrocarbon portion of the fuel, not the total RVP
of the blend. This means that we need to track the RVP of the hydrocarbon portion of the
fuel separately.

       The probability of the fuel tank being completely filled during refueling is
determined from the estimates shown in Table 2A-2. Again an independent random
number is generated with a value between 0.0 and 1.0.  If the value is less than the
probability shown in Table 2A-2 for that initial fill level, the tank is assumed to be filled
up. When a partial fill is indicated, another random number between 0.0 and 1.0 is
generated.  This random number is used in conjunction with the NORMINV function in
Excel to generate a random level of the standard normal deviate, or the number of
standard deviations to add to the mean estimate for the volume of fuel added during a
partial fill-up. The values for the mean and standard deviations for volume of fuel added
are shown in Table 2A-4.  As discussed in  Section 2A.2, whenever the initial fuel tank
level is 0.5  or greater, we assume that the tank is filled up.

       Occasionally, the volume of fuel added during a partial fill will exceed  the
capacity of the tank. This occurs when the random number generator produces a large
positive number of standard deviations to be added to the mean fuel volume typically
added during a partial fill. In these cases, we set the final tank level after refueling to
100%.
       The type of fuel added, ethanol or non-ethanol blend, is determined both by the
level of brand loyalty and the mix of fuels available in the local area.  As discussed in
Section 2A.4, brand loyalty is estimated to be 36%. Again, an independent random
number is generated with a value between 0.0 and 1.0.  If the value is less than 0.36, the
type of fuel added is assumed to be the same as that added during the last refueling.
Otherwise,  the probability of refueling with any particular fuel is assumed to be
independent of the previous fuel used. The probabilities of refueling with a non-ethanol
blend and an ethanol blend are the market shares of the respective  fuels.  This is selection
is made by  choosing a new random number.

       We then determine the quality of the fuel in the tank after the refueling  event.
The ethanol concentration of the tank fuel is simply the ethanol concentration of the fuel
prior to refueling plus that of the fuel added during refueling, each weighted by their
respective volumes. We assume that the fuel tank contains some volume of fuel, even
when it indicates empty.  Consistent with CARB in their assessment of commingling, we
assume that this tank "heel" is 10% of the tank capacity. The "ethanol" portion of the
ethanol blend is assumed to be 95% pure ethanol (i.e., it contains 5% denaturant).  This
denaturant is assumed to have the same RVP as the non-oxygenated gasoline.

       Calculating the RVP of the gasoline blend after refueling is more complicated.
We first calculate the RVP of the hydrocarbon portion of the fuel after refueling in the
same way as described above for the ethanol concentration. The RVP of hydrocarbons
blend linearly or ideally, so we simply weight the RVP of the hydrocarbon portion of the
                                       107

-------
fuel left in the tank just prior to refueling (adjusted downward for weathering as indicated
above) with that of the new fuel by their respective contributions to the total volume of
fuel in the tank after refueling.  We then increase this RVP based on the concentration of
ethanol in the tank. As described in Section 2A.3, we assume that ethanol's impact on
RVP is constant between 2 and 10 vol% and a function of the RVP of the hydrocarbon
blendstock, as discussed above. Between zero and 2 vol%, we assume that its RVP effect
increases linearly up to its full effect.

       At this point, the vehicle has been refueled and we have determined the quality of
the fuel currently in the tank. The next step is to repeat the entire process described
above, starting with a new level at which the tank is refueled once again.

       Once a vehicle has been refueled 500 times, we determine the RVP and ethanol
concentration of the fuel in the tank over a range of fuel tank fill levels. We split the full
range of possible tank fill levels into 10 discrete segments, each representing a 10% range
of tank fill level (e.g., 20-30% full). We then determine which tank fill levels the vehicle
will be driven through prior to the next refueling.  For example, if a vehicle is refueled to
100% of tank capacity and is then driven down to 1/8 full before its next refueling, its
tank moves from 100% full to 90% full to 80% full, etc. until it reaches 12.5% full.
Thus, in this example, the tank was never in the range of 0-10% full.  We assume that the
vehicle spends the same amount of time and accumulates the same amount of VMT at
each tank fill  level between its starting and ending points. (This is  simply equivalent to
assuming that vehicles are driven differently depending on their level of fuel tank fill
level; a safe assumption.) The RVP of the fuel in the tank is adjusted at each fill level as
the vehicle is being driven, including the effect of weathering. The same  is done for
ethanol concentration. For each segment of fuel tank fill level, the RVP and ethanol
concentration occurring between each set of refueling events is averaged.

       The entire process is then  repeated 50 times.  Overall, both RVP and ethanol
concentration versus tank fuel fill level is tracked for 25,000 refuelings (500 refuelings
per model pass-through times 50 model pass-throughs). Overall averages are then
determined and retained for analysis.

       One output of the model which is independent of the RVP levels of the fuels is the
distribution of the fuel tank levels of vehicle on the road at any one time.  This
distribution is shown in Table 2A-7.
                                       108

-------
       Table 2A-7.   Distribution of Fuel Tank Fill Levels for the In-Use Fleet
Range of Fuel Tank Fill Level
Lower Limit
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Upper Limit
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
% of Vehicles
7.6%
9.4%
12.9%
12.5%
12.2%
10.6%
9.4%
8.7%
8.4%
8.2%
       As can be seen, the most frequent onroad fuel tank fill levels are between 20% to
50%.  This distribution will be used to weight the effect of commingling which occurs for
each range of fuel tank fill level.

2A.6   Modeling Results

       We performed the procedure described in  Section 2 A. 5. for a set of base gasoline
RVP levels ranging from roughly 7 RVP to 9 RVP and for the two types of ethanol
blending (matched RVP and increased RVP). An example of the sequence of
calculations is as follows:
    1)     Select the RVP of non-oxygenated gasoline (EO), the RVP of the ethanol
          blend (E10), and the market share of E10,
    2)     Begin with a tank full of non-oxygenated gasoline,
    3)     Choose a random number which is used to probabilistically determine:  a) the
          level at which the tank is being refueled, b) the level to which the tank is
          filled, c) the volume of fuel thus being added,  and d) the type of fuel used to
          fill the tank (EO or E10),
    4)     Determine which tank fill levels the vehicle passed through between the prior
          fill level and the point at which it was refilled  and determine the fuel RVP at
          each 10% increment in tank fill level using the weathering equation,
    5)     Determine the RVP of the hydrocarbon portion of the fuel at the time of refill
          using the weathering equation,
    6)     Determine the concentration of ethanol in the refilled fuel tank using the
          ethanol concentration of the fuel after the prior fill-up, the volume of fuel in
          the tank at the time of refill, the ethanol concentration of the fuel used to refill
          the tank currently and the volume of fuel added during this refill,
    7)     Determine the RVP of the hydrocarbon portion of the fuel after refill by
          weighting the RVP of the hydrocarbon portion of the fuel in the tank at the
          time of refueling and the RVP of hydrocarbon portion of the fuel being added
          by the volume of the hydrocarbon portion of the fuel in the tank at the time of
          refill and the volume of the hydrocarbon portion of the fuel being added
          during refueling, respectively,
                                       109

-------
   8)     Determine the RVP of the total fuel in the tank after refueling from the RVP
          of the hydrocarbon portion of the fuel from step 6 and the effect of ethanol on
          RVP using its concentration from step 5.
   9)     Return to step 2) and proceed through 500 refuelings.

       Once the model has been applied to 500 refuelings (essentially the life of the
vehicle), the results are compiled.  The average RVP level for each interval of tank fill
level is determined. For example, over 500 refuelings, approximately 200  have the tank
being refilled when it was less than one-eighth full, so that the vehicle was driven when
the tank was 10% full.  For these 200 occurrances, the tank RVP level is averaged. This
becomes the average RVP at a fuel tank fill level of 10%. Approximately 260 refueling
involve the vehicle being driven when the fuel tank fill level is 20%.  Fuel  RVP is again
averaged for these 260 situations.  The process is repeated for a 30%  fuel tank fill level,
40%, and so on through 100% full (which occurs every time the tank is completely filled
up).  The RVP predictions of the model for various mixes of 9 RVP non-oxygenated
gasoline and a  10 RVP ethanol blend are shown in Tables 2A-8 through 2A-13. The last
line of each table shows the weighted average RVP level using the distribution of in-use
fuel tank fill levels shown above in Table 2A-7.
                                      110

-------
This page intentionally left blank.
           Ill

-------
 Table 2A-8.  In-Use Fuel Tank RVP Levels - 9 RVP CG with Ethanol Waiver (psi)
Fuel Tank Fill Level
5%
15%
25%
35%
45%
55%
65%
75%
85%
95%
Wtd.Avg.
Ethanol Blend Market Share
0%
8.15
8.23
8.29
8.35
8.40
8.44
8.50
8.58
8.66
8.75
8.42
2%
8.19
8.27
8.33
8.39
8.44
8.48
8.55
8.62
8.71
8.80
8.47
5%
8.26
8.33
8.40
8.45
8.50
8.56
8.62
8.70
8.78
8.87
8.53
10%
8.35
8.42
8.49
8.55
8.60
8.65
8.71
8.79
8.87
8.97
8.63
20%
8.52
8.58
8.65
8.71
8.77
8.82
8.89
8.97
9.05
9.15
8.80
30%
8.67
8.74
8.81
8.87
8.93
8.98
9.05
9.13
9.22
9.31
8.96
40%
8.79
8.87
8.93
8.99
9.05
9.11
9.17
9.25
9.34
9.43
9.08
50%
8.90
8.97
9.04
9.09
9.15
9.21
9.27
9.35
9.44
9.53
9.18
60%
8.98
9.05
9.11
9.17
9.23
9.29
9.35
9.43
9.52
9.61
9.26
70%
9.05
9.12
9.19
9.25
9.31
9.36
9.43
9.50
9.59
9.68
9.34
80%
9.11
9.18
9.24
9.30
9.35
9.40
9.46
9.54
9.63
9.72
9.38
90%
9.15
9.22
9.29
9.35
9.40
9.45
9.51
9.58
9.67
9.76
9.42
95%
9.17
9.24
9.30
9.36
9.41
9.46
9.52
9.59
9.68
9.77
9.44
98%
9.17
9.25
9.31
9.37
9.41
9.46
9.52
9.60
9.68
9.77
9.44
100%
9.17
9.25
9.31
9.37
9.41
9.46
9.52
9.60
9.68
9.77
9.44
Table 2A-9.  In-Use Fuel Tank RVP Levels - 7.8 RVP CG with Ethanol Waiver (psi)
Fuel Tank Fill Level
5%
15%
25%
35%
45%
55%
65%
75%
85%
95%
Wtd.Avg.
Ethanol Blend Market Share
0%
7.08
7.14
7.20
7.24
7.29
7.33
7.38
7.44
7.51
7.59
7.31
2%
7.13
7.19
7.25
7.30
7.34
7.38
7.43
7.49
7.56
7.64
7.36
5%
7.20
7.26
7.31
7.36
7.41
7.45
7.50
7.57
7.64
7.72
7.43
10%
7.30
7.36
7.41
7.46
7.51
7.55
7.61
7.67
7.75
7.82
7.53
20%
7.50
7.56
7.62
7.67
7.71
7.76
7.82
7.88
7.95
8.03
7.74
30%
7.64
7.71
7.76
7.81
7.86
7.92
7.97
8.04
8.12
8.19
7.89
40%
7.79
7.85
7.91
7.96
8.01
8.06
8.12
8.19
8.27
8.35
8.04
50%
7.91
7.97
8.03
8.08
8.13
8.18
8.23
8.30
8.38
8.45
8.15
60%
8.01
8.07
8.12
8.17
8.22
8.27
8.33
8.40
8.47
8.55
8.25
70%
8.10
8.15
8.21
8.26
8.30
8.35
8.40
8.47
8.54
8.61
8.33
80%
8.15
8.21
8.27
8.31
8.35
8.40
8.45
8.52
8.59
8.67
8.38
90%
8.19
8.25
8.30
8.35
8.40
8.44
8.49
8.55
8.62
8.70
8.42
95%
8.21
8.27
8.32
8.37
8.41
8.45
8.51
8.57
8.64
8.72
8.44
98%
8.22
8.28
8.34
8.38
8.42
8.46
8.51
8.58
8.65
8.73
8.45
100%
8.22
8.28
8.34
8.38
8.42
8.46
8.51
8.58
8.65
8.73
8.45
                                   112

-------
  Table 2A-10. In-Use Fuel Tank RVP Levels - 6.8 RVP CG with Ethanol Waiver (psi)
Fuel Tank Fill Level
5%
15%
25%
35%
45%
55%
65%
75%
85%
95%
Wtd.Avg.
Ethanol Blend Market Share
0%
6.19
6.24
6.29
6.33
6.36
6.40
6.44
6.50
6.56
6.63
6.39
2%
6.24
6.29
6.34
6.38
6.42
6.45
6.50
6.55
6.61
6.68
6.44
5%
6.31
6.37
6.41
6.46
6.49
6.53
6.57
6.63
6.69
6.76
6.51
10%
6.43
6.48
6.53
6.57
6.61
6.65
6.69
6.75
6.81
6.88
6.63
20%
6.63
6.68
6.73
6.77
6.81
6.86
6.90
6.96
7.02
7.09
6.84
30%
6.82
6.86
6.91
6.95
7.00
7.04
7.09
7.15
7.21
7.28
7.02
40%
6.96
7.02
7.07
7.11
7.16
7.20
7.25
7.31
7.37
7.44
7.18
50%
7.09
7.13
7.18
7.23
7.27
7.31
7.36
7.42
7.49
7.55
7.30
60%
7.19
7.24
7.29
7.34
7.38
7.42
7.47
7.53
7.59
7.66
7.40
70%
7.29
7.33
7.38
7.42
7.47
7.51
7.56
7.61
7.67
7.74
7.49
80%
7.35
7.40
7.45
7.49
7.53
7.57
7.61
7.67
7.73
7.79
7.55
90%
7.40
7.45
7.50
7.53
7.57
7.61
7.65
7.71
7.77
7.84
7.59
95%
7.41
7.46
7.51
7.55
7.59
7.62
7.67
7.72
7.78
7.85
7.61
98%
7.42
7.47
7.52
7.56
7.59
7.63
7.67
7.72
7.79
7.85
7.61
100%
7.43
7.48
7.53
7.57
7.60
7.64
7.68
7.73
7.79
7.86
7.62
Table 2A-11. In-Use Fuel Tank RVP Levels as - 9 RVP Gasoline with No Ethanol Waiver
Fuel Tank Fill Level
5%
15%
25%
35%
45%
55%
65%
75%
85%
95%
Wtd.Avg.
Ethanol Blend Market Share
0%
8.15
8.23
8.29
8.35
8.39
8.44
8.50
8.58
8.67
8.76
8.43
2%
8.17
8.24
8.31
8.37
8.42
8.47
8.53
8.60
8.69
8.78
8.45
5%
8.20
8.28
8.34
8.40
8.45
8.50
8.56
8.64
8.72
8.81
8.48
10%
8.26
8.33
8.39
8.45
8.50
8.55
8.62
8.69
8.78
8.87
8.53
20%
8.33
8.40
8.47
8.52
8.58
8.63
8.70
8.77
8.86
8.95
8.61
30%
8.38
8.45
8.51
8.58
8.63
8.69
8.75
8.83
8.92
9.01
8.66
40%
8.40
8.47
8.54
8.60
8.65
8.71
8.78
8.86
8.95
9.04
8.69
50%
8.41
8.47
8.54
8.60
8.66
8.72
8.78
8.86
8.95
9.04
8.69
60%
8.40
8.46
8.53
8.59
8.65
8.71
8.77
8.85
8.94
9.03
8.68
70%
8.36
8.43
8.50
8.56
8.61
8.67
8.73
8.81
8.90
8.99
8.64
80%
8.32
8.39
8.45
8.51
8.56
8.62
8.68
8.76
8.85
8.94
8.60
90%
8.26
8.33
8.40
8.45
8.50
8.55
8.62
8.69
8.78
8.87
8.53
95%
8.23
8.30
8.37
8.42
8.47
8.52
8.58
8.66
8.74
8.84
8.50
98%
8.20
8.28
8.35
8.40
8.45
8.50
8.56
8.63
8.72
8.81
8.48
100%
8.19
8.27
8.33
8.39
8.43
8.48
8.54
8.62
8.70
8.79
8.46
                                     113

-------
Table 2A-12. In-Use Fuel Tank RVP Levels - 8 RVP Gasoline with No Ethanol Waiver (psi)
Fuel Tank Fill Level
5%
15%
25%
35%
45%
55%
65%
75%
85%
95%
Wtd.Avg.
Ethanol Blend Market Share
0%
7.26
7.33
7.39
7.43
7.47
7.52
7.57
7.63
7.71
7.79
7.50
2%
7.28
7.35
7.41
7.45
7.50
7.54
7.59
7.66
7.73
7.81
7.52
5%
7.32
7.38
7.44
7.49
7.53
7.57
7.63
7.69
7.77
7.85
7.56
10%
7.36
7.43
7.49
7.54
7.59
7.63
7.69
7.75
7.83
7.91
7.61
20%
7.44
7.50
7.56
7.62
7.67
7.72
7.77
7.84
7.92
8.00
7.69
30%
7.50
7.57
7.62
7.68
7.73
7.78
7.84
7.91
7.98
8.06
7.76
40%
7.53
7.59
7.64
7.70
7.75
7.80
7.86
7.93
8.01
8.09
7.78
50%
7.53
7.60
7.66
7.71
7.76
7.81
7.87
7.94
8.02
8.10
7.79
60%
7.52
7.57
7.64
7.69
7.74
7.79
7.85
7.92
7.99
8.07
7.77
70%
7.48
7.55
7.60
7.66
7.71
7.76
7.81
7.88
7.96
8.04
7.73
80%
7.43
7.49
7.56
7.60
7.65
7.70
7.75
7.82
7.90
7.98
7.68
90%
7.36
7.42
7.48
7.53
7.57
7.62
7.67
7.74
7.82
7.90
7.60
95%
7.33
7.39
7.45
7.50
7.54
7.59
7.64
7.70
7.78
7.86
7.57
98%
7.31
7.37
7.43
7.47
7.52
7.56
7.61
7.67
7.75
7.83
7.54
100%
7.28
7.35
7.41
7.46
7.50
7.54
7.59
7.66
7.73
7.81
7.52
Table 2A-13. In-Use Fuel Tank RVP Levels - 7 RVP Gasoline with No Ethanol Waiver (psi)
Fuel Tank Fill Level
5%
15%
25%
35%
45%
55%
65%
75%
85%
95%
Wtd.Avg.
Ethanol Blend Market Share
0%
6.37
6.43
6.47
6.51
6.55
6.59
6.63
6.69
6.75
6.82
6.57
2%
6.40
6.45
6.50
6.54
6.58
6.61
6.66
6.72
6.78
6.85
6.60
5%
6.43
6.48
6.53
6.58
6.62
6.66
6.70
6.76
6.82
6.89
6.64
10%
6.49
6.54
6.59
6.64
6.68
6.72
6.77
6.82
6.89
6.96
6.70
20%
6.57
6.62
6.68
6.72
6.77
6.81
6.86
6.92
6.98
7.05
6.79
30%
6.63
6.68
6.73
6.78
6.82
6.87
6.92
6.98
7.05
7.11
6.85
40%
6.66
6.71
6.76
6.81
6.85
6.90
6.96
7.02
7.08
7.15
6.88
50%
6.66
6.71
6.76
6.81
6.85
6.90
6.95
7.01
7.08
7.15
6.88
60%
6.65
6.70
6.75
6.79
6.84
6.89
6.94
7.00
7.06
7.13
6.86
70%
6.61
6.66
6.71
6.76
6.80
6.84
6.89
6.95
7.02
7.08
6.82
80%
6.56
6.61
6.66
6.70
6.74
6.78
6.83
6.89
6.95
7.02
6.76
90%
6.48
6.54
6.59
6.63
6.67
6.71
6.75
6.81
6.88
6.94
6.69
95%
6.44
6.49
6.54
6.58
6.62
6.66
6.70
6.76
6.82
6.89
6.64
98%
6.41
6.47
6.51
6.56
6.59
6.63
6.67
6.73
6.79
6.86
6.61
100%
6.40
6.45
6.50
6.54
6.58
6.61
6.65
6.71
6.77
6.84
6.60
                                      114

-------
       The next step is to estimate the impact of commingling on in-use RVP. We use the in-use tank RVP levels shown in the previous six tables
for ethanol blend market shares of zero and 100% to represent situations where no commingling occurs. In the absence of commingling at
intermediate levels of ethanol blend market share, the RVP should vary linearly between those found at zero and 100%.  The impact of commingling
is then the difference between the actual level of RVP estimated by the model and the RVP estimated from the zero and  100% ethanol blend market
share RVP levels. Table 2A-14 shows the impact of commingling for the case where gasoline RVP is 9 psi and ethanol blends are allowed a 1.0 psi
RVP waiver.
                                                           Table 2A-14.
          Commingling as a Function of Fuel Tank Fill Level and Ethanol Blend Market Share - 9 RVP CG with Ethanol Waiver
Fuel Tank
Fill Level
5%
15%
25%
35%
45%
55%
65%
75%
85%
95%
Wtd.Avg.
Ethanol Blend Market Share
2%
0.016
0.017
0.019
0.021
0.023
0.023
0.025
0.025
0.026
0.025
0.022
5%
0.049
0.047
0.049
0.049
0.051
0.052
0.054
0.055
0.055
0.055
0.051
10%
0.101
0.096
0.098
0.097
0.101
0.105
0.110
0.112
0.112
0.112
0.104
20%
0.167
0.164
0.166
0.170
0.175
0.181
0.185
0.187
0.186
0.186
0.176
30%
0.213
0.210
0.210
0.217
0.224
0.232
0.238
0.240
0.242
0.242
0.226
40%
0.235
0.229
0.228
0.234
0.243
0.251
0.259
0.263
0.265
0.265
0.246
50%
0.234
0.228
0.230
0.235
0.243
0.252
0.260
0.263
0.266
0.266
0.246
60%
0.220
0.214
0.217
0.222
0.229
0.240
0.246
0.250
0.251
0.252
0.233
70%
0.183
0.179
0.180
0.184
0.192
0.197
0.201
0.204
0.206
0.206
0.192
80%
0.132
0.129
0.130
0.134
0.138
0.143
0.147
0.150
0.150
0.151
0.140
90%
0.073
0.070
0.072
0.071
0.073
0.076
0.078
0.078
0.078
0.079
0.074
95%
0.036
0.037
0.036
0.037
0.039
0.042
0.043
0.042
0.042
0.043
0.039
98%
0.013
0.015
0.014
0.015
0.013
0.015
0.014
0.013
0.013
0.013
0.014
       As can be seen, the impact of commingling increases slightly moving from low levels of fuel tank fill level to high levels. As found by
previous studies of commingling, the impact of commingling is lowest when either EO or E10 fuels predominate the market and peaks when the mix
of EO and E10 is approximately 50/50.  Again, the weighted average of the commingling impact is determined by applying weighting the
commingling impact at each fuel tank fill level by the distribution of fill levels in-use shown in Table 2A-7.
                                                                115

-------
Table 2A-15 shows the weighted average commingling impacts for the six fuel cases.
             Table 2A-15. Weighted Average Commingling Impact for Various Sets of EO and E10 Fuels (psi)
EO/E10
RVP Level
9/10
7.8/8.9
7/8.2
9/9
8/8
7/7
Ethanol Blend Market Share
2%
0.022
0.029
0.028
0.022
0.021
0.027
5%
0.060
0.065
0.067
0.051
0.055
0.065
10%
0.102
0.110
0.122
0.104
0.110
0.125
20%
0.171
0.201
0.203
0.176
0.189
0.212
30%
0.227
0.239
0.265
0.226
0.249
0.267
40%
0.249
0.273
0.299
0.246
0.269
0.298
50%
0.249
0.274
0.290
0.246
0.278
0.294
60%
0.227
0.255
0.275
0.233
0.253
0.277
70%
0.199
0.218
0.236
0.192
0.218
0.233
80%
0.143
0.160
0.173
0.140
0.160
0.172
90%
0.084
0.082
0.094
0.074
0.080
0.096
95%
0.045
0.043
0.046
0.039
0.045
0.047
98%
0.020
0.019
0.015
0.014
0.019
0.017
                                                       116

-------
This page intentionally left blank.
              117

-------
       Ethanol use in the three RFS rule fuel cases (Reference, RFS, and EIA) occurs
predominately under three situations: 1) 9 RVP CG with an RVP waiver for ethanol, 2) 7.8 RVP
CG with an RVP waiver for ethanol, and 3) 7 RVP RFG.  In order to simplify application of the
impact of commingling to our emission modeling, we averaged the commingling impacts for
these situations from Table 2A-15 and applied that to the entire U.S. as a function of ethanol
blend market share. This average set of commingling impacts is shown in Table 2A-16.
Table2A-16.
Commingling Impact Applied in RFS Rule Emission Modeling
Ethanol Blend Market Share
0%
2%
5%
10%
20%
30%
40%
50%
60%
70%
80%
90%
95%
98%
100%
Commingling Impact (psi)
0
0.026
0.064
0.113
0.194
0.244
0.273
0.272
0.253
0.217
0.159
0.087
0.045
0.019
0.000
                                         118

-------
Chapter 3:  Impacts on Emissions from Vehicles, Nonroad
Equipment, and Fuel Production Facilities
       As described in Chapter 2, there are a large number of potential fuels that qualify as
renewable.  However, only two are expected to be used in significant volumes by 2012: ethanol
and biodiesel. Of these, ethanol use is expected to predominate. In particular, ethanol is
expected to dominate the "growth" in renewable fuel use between now and 2012. Thus, our
primary focus here will be on the impact of the use of ethanol on emissions in spark-ignited
vehicles and equipment. We will more briefly touch on the impact of biodiesel fuel use on
emissions.

       Similarly, we expect that the bulk of the impact of ethanol use on emissions and air
quality will be associated with emissions from spark-ignited vehicles and equipment using low
level ethanol-gasoline blends. We expect the use of high level ethanol-gasoline blends, like E85
to be relatively small in comparison.  Thus, the discussion here will focus on emissions from the
use of low level ethanol blends.  We will more briefly discuss the per vehicle impacts of use of
high level ethanol-gasoline blends relative to gasoline.

       Finally, we present estimates of the emissions related to the production and distribution
of both ethanol and biodiesel. The emissions related to the production and use of ethanol  can be
significant relative to the emission impacts of the use of ethanol blends, due to the significantly
increase in the volume of ethanol expected to be produced in the future.
3.1    Effect of Fuel Quality on Onroad Spark-Ignited Vehicle Emissions

       Ethanol belongs to a group of gasoline additives commonly referred to as oxygenates.
The two most commonly used oxygenates are ethanol and MTBE, though TAME has been used
in significant volumes, as well. All oxygenates have relatively high levels of octane (i.e., greater
than 100 R+M/2). Both ethanol and MTBE have been used historically to meet the gasoline
oxygen requirements for oxyfuel and RFG.  Historically, MTBE was the predominant oxygenate
used in gasoline in the U.S. Over time, MTBE use has decreased in the U.S, while ethanol use
has increased, to the point where ethanol use now predominates. This trend appears to be
accelerating, to the point where it appears that essentially all MTBE use will cease in the U.S
sometime in 2007.

       The impact of oxygenate use on emissions from motor vehicles has been evaluated since
the mid-1980's. Several models of the impact of gasoline quality on motor vehicle emissions
were developed in the early 1990's and updated periodically since that time. We use the most
up-to-date versions of these models here to estimate the impact of changes in oxygenate use on
emissions.  Still, as will be described below,  significant uncertainty exists as to the effect of these
gasoline components on emissions from both motor vehicle and nonroad equipment, particularly
from the latest models equipped with the most advanced emission controls. Assuming adequate
funding, we plan to conduct significant vehicle and equipment testing over the next several years


                                         119

-------
to improve our estimates of the impact of these additives and other gasoline properties on
emissions. The results of this testing are not available for inclusion in this analysis. We hope
that the results from these test programs will be available for reference in the comprehensive
evaluation of the emission and air quality impacts of all the fuel-related requirements of the
Energy Act required by Section 1506.  A draft of this study is required to be completed in 2009.
As we discuss the emission impacts of increased ethanol use below, we identify the areas where
current estimates appear to be the most uncertain and where we hope to obtain additional data
prior to the 2009 study.

3.1.1   Low Level Ethanol and MTBE-Gasoline Blends

       EPA has developed a number of emission models relating the impact of gasoline quality
on emissions from motor vehicles.  In 1993, EPA published the Complex Model, which predicts
the effect of gasoline quality on VOC, NOx and air toxic emissions from 1990 model year light-
duty motor vehicles (i.e., Tier 0 vehicles). This model is used to determine refiners' compliance
with RFG and anti-dumping standards. The Complex Model also contains estimates of the
impact of gasoline RVP on non-exhaust VOC emissions.  These estimates were taken from the
then-current version of the MOBILE emissions model, MOBILES.

       In 2001, in responding to California's request for a waiver of the RFG oxygen mandate,
EPA performed a new analysis of the impact of gasoline quality on exhaust VOC and NOx
emissions from Tier 0 vehicles.  This analysis included essentially  all of the data used to develop
the Complex Model, as well as some additional data developed since 1993. It also used more
advanced statistical tools, such as a mixed statistical model,  which  were not available in 1993.
These VOC and NOx models are referred to here as the EPA Predictive Models. Thus, in terms
of both supporting data and modeling tools, the EPA Predictive Models represent an
improvement over the Complex Model, at least for exhaust VOC and NOx emissions. Because
the criteria for granting California a waiver of the oxygen requirement focused on ozone and PM
impacts, EPA did not develop a similar model for toxics or CO emissions.

       In roughly the same timeframe, EPA developed its latest motor vehicle emission
inventory model, MOBILE6.  Some of the fuel-emission relationships from the Complex Model
were incorporated into MOBILE6. These included the effect of selected gasoline properties on
exhaust VOC and NOx emissions and the fraction of VOC emissions represented by several air
toxics (benzene, formaldehyde,  acetaldehyde, and 1,3-butadiene).  The EPA Predictive Models
were not available in time for their incorporation into MOBILE6. MOBILE6 also contains
estimates of the effect of certain gasoline parameters on CO emissions, namely RVP and oxygen
content. The effect of RVP on non-exhaust VOC emissions contained in MOBILE6.2 represents
an update of the MOBILES and Complex Model estimates.

       We desire in this RFS analysis to utilize the most up to date estimates of the impact of
gasoline quality on emissions currently available. No one model currently contains the most up
to date estimates for all the pollutants of interest. Therefore, we have broken up the remainder of
this sub-section into six parts. The first discusses emissions of VOC and NOx, as the EPA
Predictive Models address these pollutants. The  second discusses CO emissions, as neither the
Complex Model nor the EPA or CARB Predictive Models address  this pollutant. The third
                                          120

-------
section addresses emissions of air toxics, as a combination of models represents the best estimate
of the impact of fuel quality on these emissions. The fourth section addresses non-exhaust VOC
emissions. The fifth section addresses PM emissions. The sixth section addresses emissions of
aromatic hydrocarbons. The seventh and final section presents the impact of ethanol and MTBE
blending on per mile emissions from gasoline-fueled motor vehicles.

3.1.1.1       Exhaust VOC, CO and NOx Emissions

       In this section we evaluate the various models available which predict the impact of
gasoline quality on exhaust VOC, CO and NOx emissions.  Several such models have been
developed over the past 15 years.  We first discuss the EPA Complex Model, the EPA Predictive
Models and the CARB Predictive Models due to the wide range of fuel parameters which they
address and their similar form.  We next describe the fuel effects contained in EPA's
MOBILE6.2 emission inventory model, as this model addresses CO emissions, which the other
models do not. These models best predict emissions from Tier 0 vehicles, as most of the
emission data upon which they were based were from these vehicles. A number of fuel effects
test programs which tested later model year vehicles have been performed since the above
emission models were developed. We summarize these studies below and develop emission
projections based on consistent statistical procedures. Finally, we select which model best
predicts the effect of fuel quality for each pollutant.  Due to the uncertainty involved with
predicting the impact of fuel quality on emissions from Tier 1 and later vehicles, we develop two
alternative approaches to making such predictions for the purpose of this rule.

3.1.1.1.1     EPA Complex Model and CARB and EPA Predictive Models

       In 1993, EPA published the Complex model to investigate the effects of changing
gasoline fuel parameters on the exhaust emissions of Tier 0 and older vehicles.  This model
predicts the effect of gasoline quality on exhaust VOC, toxics and NOx emissions and non-
exhaust emissions of VOC and benzene.  The Complex Model is used to determine compliance
with the emissions performance requirements for federal RFG by comparing the predicted
emissions of a candidate fuel to that of a baseline fuel for common baseline vehicle technology.
The baseline fuel and the baseline vehicle technology represent those fuels and vehicles included
in the 1990 US light duty vehicle fleet (Tier 0 technology).

       In 1999, the state of California petitioned EPA for a waiver of the oxygen requirement in
RFG.  The reasoning behind the waiver request centered on the California Air Resources Board's
(CARB) analysis which showed that reducing the amount of oxygen in RFG would lead to
reduced NOx exhaust emissions. The model that CARB developed to support their claim was
called the Phase 3 predictive model22.  This model differed from the previous version of
CARB's predictive model (the Phase 2 model) in a number of ways. The most significant
difference included a substantially expanded database, mainly for model year 1986 and newer
vehicles, as well as an improved version of the statistical analysis software package used to
develop the model (SAS® PROC MIXED).  According to CARB, the Phase 3 predictive model
displays a  steeper NOx/Oxygen response than the Phase 2 Predictive model as a result of
eliminating the RVP by Oxygen term which the previous model had erroneously included. This
caused an increase in the NOx exhaust emissions predicted, and for many areas this increase
                                          121

-------
would lead to NOx levels exceeding those set by National Ambient Air Quality Standards
(NAAQS).

       According to the 1990 Clean Air Act, EPA can waive the RFG oxygen requirement if it
prevents compliance with a NAAQS. In order to properly perform an environmental impact
analysis in response to this waiver request, EPA considered using both its Complex Model, as
well as CARS's Phase 3 Predictive Model to estimate the impact of gasoline quality on
emissions. The EPA Complex model, while considered statistically robust due to the large
number of vehicles comprising the dataset, was not considered to be adequate for a number of
reasons. First, the Complex Model was based on an emissions database which did not include
several studies that have since been published.  Second, the EPA Complex Model was developed
using a fixed effects statistical modeling approach42.  In contrast, both the CARB Phase 2 and 3
models were mixed models, employing a more sophisticated statistical approach than was
available at the time of development of the Complex model.

       EPA also rejected using CARB's Phase 3 Predictive Model in its analysis of the waiver
request.  While CARB had developed a very detailed protocol for developing the Phase 3 model,
it rejected the results of this protocol because the resulting model differed too substantially from
the Phase 2 model. Thus, EPA decided to create its own "predictive models" for exhaust VOC
and NOx emissions which combined the protocols used to develop the Complex Model with the
expanded database and improved statistical tools which were now  available.  EPA relied on
existing EPA models for evaporative VOC emission effects.  However, these latter estimates
were augmented with recent data indicating that ethanol increased  permeation emissions, as well
as the consideration of several commingling models and associated assumptions about drivers'
refueling behavior.

       One main conclusion drawn by EPA in the California Oxygen waiver analysis was that
insufficient data existed at that time to conclusively determine the  response of Tier 1 and newer
vehicles to fuel parameters other than sulfur.43  Some data indicated that oxygen increased NOx
emissions from Tier 1 and later vehicles, while other data contradicted this.  Due to this
inconsistency, EPA assumed that oxygen did not affect exhaust VOC, NOx or CO emissions
from Tier 1 and later vehicles in its analysis of CARB's request for a waiver of the RFG oxygen
mandate.
42 A "fixed effects" model of this kind makes no attempt to estimate the error introduced by sampling from some
larger population of vehicles or fuels. The model just describes quantitatively the relationships among variables that
are present in the dataset that was analyzed. A "mixed" model, as was used by CARB in both the Phase 2 and Phase
3 predictive models' construction, attempts to go beyond description of the available data to make statistical
inference to some larger population from which the available data were sampled. In this case CARB treated the
vehicle effects as random (assuming that the test vehicles were sampled from some larger fleet) while fuel effects
were treated as "fixed" (assuming that all fuels of interest were represented in the data). Such a modeling approach
makes it possible to estimate the probable error in modeled effects in a way that is not possible with a fixed effects
model. The approach, moreover, improves the accuracy of the significance measures used to decide which terms to
include in the model.

43At the time of that 1999 analysis, sufficient data existed on the emissions effects of Sulfur on Tier 1 vehicles to be
modeled.  However, sulfur levels were not expected to change as a result of the removal of oxygen from RFG in
California, so the effect of sulfur was moot in this situation.
                                            122

-------
       None of the three models discussed above (i.e., the Complex Model, the CARB
Predictive Model, and the EPA Predictive Models) address CO emissions. Historically, this is
because the RFG program did not mandate a specific reduction in CO emissions and the lesser
role of CO emissions in forming ambient ozone. The CARB Predictive Model considers the
impact of fuel oxygen content beyond 2 wt%, but does not address the full range of oxygen
levels on CO emissions, nor the impact of other fuel parameters. The only EPA model which
predicts the impact of fuel quality on CO emissions is MOBILE6.2, which is discussed in the
next section.

3.1.1.1.2     MOBILE6.2

       The exhaust emission effects contained in MOBILE6.2 often  differ for normal and high
emitting vehicles. They can also vary by model year. As it is difficult to determine the fraction
of emissions coming from each model year's vehicles in MOBILE6.2, as well as normal and
high emitters, it is not feasible to predict outside of the model how a specific fuel is going to
affect in-use emissions. In addition, the split between normal and high emitters varies depending
on the presence and type of inspection and maintenance (I/M) program applicable in a particular
local area. Thus, the effect of a specific fuel on emissions can vary to some degree from one
county to another.

       In order to quantify the effect of various fuel parameters on exhaust emissions in
MOBLIE6.2 under the conditions existing in the 2012-2020 timeframe, we compared the
changes in emissions predicted by the NMEVI modeling described in  Chapter 4 of the Draft RIA
with the changes in fuel quality occurring in the ethanol use scenarios. Specifically, we first
determined the percentage change in exhaust VOC, CO and NOx emissions by county for the
base and 7.2 Minimum RFG Use scenarios evaluated for the NPRM.  We then performed a series
of linear regressions of these ratios against the change in fuel RVP, ethanol content and MTBE
content. We did this for the 2012, 2015 and 2020 emission projections separately. For each
combination of county and calendar year, the only property that changed was fuel quality.  All
other parameters relevant to emissions (e.g., the distribution of vehicles by age and class, VMT,
ambient temperature, etc.) were otherwise identical.  The results are summarized in Table 3.1-1.
                                          123

-------
               Table 3.1-1. Fuel-Exhaust Emission Effects in MOBILE6.2

RVP (% / psi)
Ethanol (% / Vol %)
MTBE (% / Vol%)
Adjusted r- Square
20 12 (fleet average)
voc
NOx
CO
7.1%
0.6%
12.7%
-1.1%
0.0%
-0.75%
-0.7%
0.0%
-0.4%
0.83
0.95
0.36
2015 (fleet average)
VOC
NOx
CO
7.0%
0.6%
12.7%
-1.2%
0.0%
-0.73%
-0.7%
0.0%
-0.4%
0.85
0.95
0.36
2020 (fleet average)
VOC
NOx
CO
6.7%
0.6%
12.6%
-1.2%
0.0%
-0.67%
-0.7%
0.0%
-0.4%
0.87
0.95
0.39
       For comparative purposes, the effect of RVP, ethanol and MTBE on exhaust VOC and
NOx emissions from the EPA Predictive Models are shown in Table 3.1-2.  The base fuel is a
typical non-oxygenated, summertime, conventional gasoline, with 8.7 RVP, 30 ppm sulfur, 32
vol% aromatics, 13 vol% olefms, T50 of 218 F, T90 of 329, and no oxygen.
Table 3.1-2.
Fuel-Exhaust Emission Effects per the EPA Predictive Models

VOC
NOx
RVP (% / psi)
1.1%
1.1%
Ethanol (% / Vol %)
-0.16%
0.75%
MTBE (% / Vol%)
-0.17%
0.36%
       As can be seen, the exhaust emission effects contained in the EPA Predictive Models
differ quite dramatically from those in MOBILE6.2. Regarding the effect of RVP, both models
predict that an increase in RVP will increase both exhaust VOC and NOx emissions. However,
MOBILE6.2 predicts that an increase of 1.0 psi will increase exhaust VOC by roughly 7%, while
the EPA Predictive Models predict only a 1% increase.  Regarding NOx emissions, the EPA
Predictive Models predict the larger effect (1%),  while the effect in MOBILE6.2 is smaller
(0.6%). While the ratio of these two effects is significant, the absolute difference (0.4%) is very
small.

       Regarding the addition of ethanol, the two models again predict very different results.
MOBILE6.2 predicts roughly 7 times the exhaust VOC reduction per volume percent of ethanol
added, with no increase in NOx. The EPA Predictive Models project a significant increase in
NOx emissions. The relative differences are similar for the addition of MTBE to gasoline,
though the difference between the two estimates  of exhaust VOC reduction is smaller.

       Regarding CO Emissions from Tier 0 vehicles, MOBILE6.2 projects that a 3.5 wt%
oxygen fuel (i.e., E10) will reduce CO emissions from normal emitters by 11% and those from
                                          124

-------
high emitters by 19%.44 For Tier 1 and later vehicles, MOBILE6.2 projects that a 3.5 wt%
oxygen fuel (i.e., E10) will have no effect on CO emissions from normal emitters and will reduce
CO emissions from high emitters by the same 19% estimated for Tier 0 vehicles. This latter
projection was based on a similar assumption to those made during EPA's review of California's
request for a RFG oxygen waiver due to the same lack of relevant emission data. We estimate
that the fraction of CO emissions coming from high emitters is 64.8% based on the overall effect
of 6.7% in 2020 shown in Table 3.1-1 for 2020, when the fleet is entirely Tier 1 and later
vehicles. This produces a fleet wide CO emission reduction for an E10 blend of 13.8% for Tier 0
vehicles and 6.7% for Tier 1 and later vehicles. MOBILE6.2 does not project any impact of the
other relevant fuel parameters (aromatics, olefins, T50, and T90) on CO emissions for either Tier
0 or Tier 1 vehicles.

       Since that time, the results of several test programs have been published. Given the
dwindling numbers of Tier 1 vehicles on the road, these more recent studies have focused on
vehicles certified to California's low emission vehicle (LEV), ultra low emission vehicle
(ULEV), and super ultra low emission vehicle (SULEV) standards, as well as federal LEV and
Tier 2 standards. The results of these more recent studies, as well as the few available in 2001,
are discussed in the following section.

3.1.1.1.3.      CRCE-67 Study

3.1.1.1.3.1    Overview

       In early 2006, the Coordinating Research Council (CRC) published the results of their E-
67 study investigating the effects of three fuel parameters, ethanol, T50 and  T90, on exhaust
emissions from recent model year vehicles.2 The twelve vehicles tested included both cars and
light trucks, certified to California LEV, ULEV and SULEV standards, with model years ranging
from 2001 to 2003. A matrix of twelve (12) fuels was tested in this program, with varying levels
of ethanol, T50, and T90. Each fuel parameter (ethanol, T50, and T90) was tested at each of
three levels. However, a full factorial matrix of 27 test fuels was deemed unnecessarily large due
to subtle differences between fuels that may not have yielded statistically significant results, or
due to practical considerations regarding the fuels that could be blended using existing refinery
streams.

       The E-67 report presents the results of emission testing for each fuel, as well as a mixed
statistical model created from the emission data. The model indicates that each of the three fuel
parameters always  has a statistically significant effect on both NMHC45 and NOx emissions. In
addition, significant interactions between the three fuel parameters are also often present.

       The first step in our analysis of the CRC E-67 model was to compare the emissions
changes predicted by the CRC mixed model to the actual emissions changes observed for each
44 A normal emitter is generally a vehicle whose emissions are no more than twice its certification emission
standards.  A high emitter is a vehicle whose emissions exceed this level.

45 NMHC is essentially equivalent to VOC for our purposes in this study.


                                           125

-------
fuel in the test program.  We calculated average NMHC and NOx emissions over the Federal
Test Procedure for all twelve vehicles on each fuel. The CRC mixed model predicts the
percentage change in emissions for each fuel relative to another fuel. These changes in predicted
exhaust emissions for each fuel were applied to the measured emissions for fuel "H" in order to
create a set of absolute emission levels for each fuel. We then compared the emissions predicted
by the CRC E-67 model to the measured emission levels to observe how well the model
predicted the effects of each fuel. The fuel properties of the CRC E-67 test fuels are listed in
Table 3.1-3AAA, below, and in greater detail in Table 3A-1 of Appendix 3 A. We selected CRC
fuel "H" as the "base" fuel since its properties are the closest to a national average non-
oxygenated conventional gasoline (0% Ethanol, 215°F T50, 330 °F T90).  (See Table 2.2-6 in
Chapter 2 for data pertaining to gasoline survey results across the U.S.)

                 Table 3.1-3.  CRC E-67 Test Program Fuels Properties"

Fuel
A
B
C
D
E
F
G
H
1
J
K
L
Target Pro
T50 (°F)
195
195
195
195
195
215
215
215
215
235
235
235
parties for Design Variables
T90 (°F)
295
295
330
355
355
295
295
330
355
330
355
355
Ethanol (%)
0
5.7
10
0
10
0
10
0
5.7
5.7
0
10














Actual Values
T50 (°F)
195
191
193
199
198
217
212
216
216
237
236
233
T90 (°F)
294
290
329
355
352
295
291
327
354
329
355
349
Ethanol
0
5.6
10.4
0
10.3
0
10.1
0.1
5.9
5.9
0
10.5
                See Table 3A-1 in Appendix 3 A for detailed properties of all E-67 test fuels
       The comparison between predicted and measured NOx emissions is shown in Figure 3.1-
1.  The fuels are shown to indicate a trend in ethanol content, from the lowest levels on the left to
the highest levels on the right. Within a constant level of ethanol content, the fuels are then
shown in order of their level of T50 (lowest again on the left and highest on the right).  The y-
axis scale in this figure is set to match that for NMHC emissions, which will be presented and
discussed next.
                                           126

-------
                 Figure 3.1-1. CRC E-67 Predicted vs. Actual Emissions
Relative NOx Emissions Change from Base (CRC H: 215 T50, 330 T90, 0% EtOH)
                   0% Ethanol
                                         5.7% Ethanol
                                                            10% Ethanol
     50.0%
     40.0%
     30.0%
  01
  O)
  c
  ro
  O
  5?
20.0%
     10.0%
     0.0%
    -10.0%
                 -CRC E-67 NOx Predicted
                 -Actual CRC Emissions
                            D     K     B     J     I      G
                            CRC Fuels (order of increasing O2, T50, T90)
       As shown in Figure 3.1-1, the CRC model for NOx emissions predicts the general trend
in the emission data, which roughly indicates an increase in NOx emissions with increasing
ethanol content.  However, the model clearly does not reflect many of the fuel to fuel differences
indicated by the actual emissions data. One example of this is the change between fuels G and C
- two 10% ethanol blends with relatively low distillation temperatures.  In changing from fuel G
to fuel C, the CRC E-67 model predicts a 4.3% increase in NOx emissions whereas the actual
test data clearly shows a 9.2% decrease. This likely indicates the existence of interactions
between the fuel parameters which are more complex than those which could be included in the
model. While fuel parameters other than ethanol, T50, and T90 were held as constant as possible
among all the test fuels, the level of specific compounds (such as toluene or the various xylenes)
could not be held constant. It is possible that some of these compounds are affecting NOx
emissions and confounding the ability of the model based just on ethanol, T50, and T90 to
predict the observed changes.

       Figure 3.1-2 repeats this comparison for NMHC.
                                           127

-------
                 Figure 3.1-2. CRC E-67 Predicted vs. Actual Emissions
Relative NMHC Emissions Change from Base (CRC H: 215 T50, 330 T90, 0% EtOH)
                0% Ethanol
                                                               10% Ethanol
   50.0%
   -10.0%
                          CRC Fuels (order of increasing O2, T50, T90)


       First, it is very apparent that NMHC emissions are much more sensitive to fuel quality
than NOx emissions. The largest increase in NMHC emissions relative to Fuel H is three times
that for NOx emissions. Except for Fuels A and F (and of course Fuel H), the CRC model
generally under-predicts the measured NMHC data. Directionally, however, the predicted
emissions changes are very consistent with those observed in the test results. For this dataset at
least, the effect of fuel quality on NMHC emissions are much more predictable than NOx
emissions.

       The fuels studied in this test program were varied independently at low, medium, and
high levels of T50, T90, and Ethanol. If you include all the possible linear, quadratic, and
interactive terms,  there are a total of possible 10 combinations. The CRC E-67 models included
8 out of thelO possible fixed effects for the NOx, NMHC, and CO models. These terms were:
T50, T90, ethanol (EtOH), T50 squared,  T90 squared, EtOH squared, T50 by EtOH, and T90 by
EtOH. The excluded terms were T50 by T90, and T50 by T90 by EtOH, which CRC excluded
from consideration since previous studies had indicated that these terms had little effect on
emissions.

       Also, several of the terms that were included in the CRC model  had p-values greater than
0.1, indicating that those terms are less than marginally significant.46 Specifically, the EtOH by
EtOH term in the NMHC model and the  T90, T90 by T90 and T90 by EtOH terms in the NOx
model all had p-values above 0.10. In developing both the Complex Model and the EPA
46 Using the widely accepted criteria of a 95% confidence interval, p < 0.05 is considered to be statistically
significant, 0.05 < p < 0.10 is marginally significant, and p > 0.10 is not considered statistically significant.
                                           128

-------
Predictive Models, our procedures would normally exclude the least significant term.  A new
regression would then be performed without this term being included in the model.  This process
would be repeated until all the remaining terms were statistically significant.47

       We desired to determine how these statistically insignificant terms might be affecting the
predicted emission changes.  This, plus the discrepancies between the CRC E-67 model and the
actual emissions data, especially for the NOx model, prompted us to create our own NOx and
NMHC models using the CRC E-67 dataset.  Conducting our own modeling also provides us
with the opportunity to apply a wide range of statistical tests in order to better understand the
role of various fuel parameters in affecting emissions from these vehicles. The following
sections provide details pertaining to the verification of the CRC model and the motivation for
constructing a new model from this data.

3.1.1.1.3.2.   Development of a New Mixed Model: The EPA E-67 Model

       Using the E-67 dataset provided by the CRC, EPA first verified the coefficients and p-
values of the CRC E-67 model using the full E-67 dataset (no outliers were removed) with the
same 8 fixed fuel effects that were included by CRC. This was successful and the coefficient
and p-values resulting from this modeling are shown in Table 3.1-4.

                 Table 3.1-4.  CRC E-67 Model P-Values and Coefficients
CRC E-67 NMHC CO NOx
Effect
Intercept
T50
T50*T50
T90
T90*T90
EtOH
EtOH*EtOH
T50*EtOH
T90*EtOH
P-Value
<0001
<0001
<0001
0.0541
0.0035
0.1124
0.2816
0.084
0.0004
Coefficient
-3.2942
0.0063
0.000176
0.001685
0.000058
0.005679
0.000722
0.000195
0.000244
P-Value
0.0001
0.3099
0.0428
0.0051
0.0815
0.0174
0.0005
0.0182
0.0534
Coefficient
-0.7966
0.001227
0.000099
-0.0045
0.000045
-0.01581
0.003118
0.000355
0.000174
P-Value
<.0001
0.8939
0.2182
0.762
0.1163
0.0504
0.0861
0.0414
0.99
Coefficient
-2.6183
-0.00013
-0.00006
0.00024
0.000043
0.00571
0.001622
-0.00032
-1.19E-06
       EPA then created a new model starting with all combinations of T50, T90, and EtOH
along with their squares, cross products, and the interactive terms T50 by T90 by EtOH for a
total of 10 fixed effects. From this "full model", variables were eliminated in order to improve
the fit statistics between the model and the test data until a "final model" was created that
contained 7 fixed fuel effects for NMHC and CO, and 6 fixed fuel effects for NOx. Table 3.1-5,
on the following page, shows the p-values and coefficients for the fixed effect terms of each
model.
47 One exception to this process is that the linear form of a variable, such as ethanol, would always be retained in
the model if a second order term included ethanol (e.g., the ethanol by T90 term).
                                           129

-------
                 Table 3.1-5. EPA E-67 Model P-Values and Coefficients
EPA E-67 NMHC CO NOx
Effect
Intercept
T50
T50*T50
T90
T90*T90
EtOH
T50*EtOH
T90*EtOH
T50*T90
T50*T90*EtOH
P-Value
<.0001
<0001
<.0001
0.0498
0.0039
0.101
0.0987
0.0002


Coefficient
-3.2773
0.006272
0.000168
0.00172
0.000057
0.005892
0.000186
0.00025


P-Value
0.0002
0.5815

0.0059

0.0111
<.0001
0.0299
<.0001
0.0003
Coefficient
-0.7684
0.00066

-0.00437

-0.01726
0.003843
0.000178
0.000126
0.000023
P-Value
<.0001
0.687

0.7761
0.0735
0.062
0.0446
0.0426


Coefficient
-2.6418
-0.00037

0.000224
0.000047
0.005393
0.001854
-0.00031


       As shown in Table 3.1-5, the EPA E-67 Model does not include terms with p-values
greater than 0.10 (except for linear terms included in statistically significant higher order terms).
Statistical tests show that these two models are not significantly different from one another. The
null hypothesis in this case is that EPA E-67 fits the data just as well as the original CRC E-67
model based on a chi-squared test. However, based on several fit statistics (AIC, AICC, and
BIC) the EPA E-67 model provides a slightly better fit to the test data than either the original
CRC E-67 model or the full model with all 10 terms included.  The next step is to compare the
EPA E-67 model predictions  to both the E-67 data and the predictions of the EPA Predictive
Models, which reflect the emission effects for older vehicles.

       Both the EPA E-67 and EPA Predictive models are mixed models that predict the relative
changes in exhaust emissions due to carefully controlled changes in gasoline quality, including
the addition of an oxygenate such as ethanol. The models are not intended to be accurate at
predicting absolute emission levels, but rather the difference in emissions when fuel properties
are varied.  The goal of this analysis is to determine if the EPA Predictive models, which were
developed using data from Tier 0 and earlier vehicles, predict the same relative changes in
emissions as the Tier 1 vehicles used for the EPA E-67 model.

       A key difference between the models is that there are only three fuel parameters used as
inputs for the EPA E-67 model: T50, T90, and ethanol content.  The EPA Predictive Models use
these three properties along with RVP, aromatic content, olefm content, and sulfur content as
fuel parameter inputs to the model.

       We ran the EPA E-67 and EPA predictive models with the 12 fuels used in the CRC test
program, inputting the applicable fuel  properties used in each model. Following the same
procedure as outlined above,  CRC test fuel H was selected as a "base" fuel in order to compare
relative changes between this fuel and others with varying amounts of ethanol, T50, and T90.
The NOx emissions predicted by the EPA E-67 and EPA Predictive models, together with the
actual E-67 study data, are shown graphically in Figure 3.1-3, below.
                                          130

-------
             Figure 3.1-3. EPA E-67 and EPA Predictive vs. Actual Emissions
Relative NOx Emissions Change from Base (CRC H: 215 T50, 330 T90, 0% EtOH)
    50.0%
    40.0%
    30.0%
  O 20.0%
    10.0%
     0.0%
    -10.0%
    -20.0%
                            CRC Test Fuels (in order of increasing O2, T50, T90)
       As shown in Figure 3.1-3, neither model predicts the actual test data with complete
accuracy.  The EPA E-67 shows the same general relationship to the emission data as did the
CRC E-67 NOx model. Thus, removing the statistically insignificant terms had little impact on
the relative fit of the model to the data. The EPA Predictive NOx models, on the other hand,
appear to be primarily sensitive to ethanol, with T50 and T90 playing very limited roles in
affecting NOx emissions. In contrast, the E-67 model shows sensitivities to all three parameters.

       Overall, the E-67 study indicates that NOx emissions from recent model year vehicles
(LEVs, ULEVs and SULEVs) are still sensitive to at least several fuel parameters. As indicated
by the inability of either the EPA E-67 model  or the EPA Predictive Models to accurately predict
all of the changes seen in the E-67 data, this study is very valuable in identifying the continued
sensitivity of LEV and cleaner vehicles to changes in fuel  quality.

       Figure 3.1-4, below, shows the comparison of NMHC emissions predicted by the EPA E-
67 and Predictive models together with the E-67 study data.
                                          131

-------
             Figure 3.1-4.  EPA E-67 and EPA Predictive vs. Actual Emissions
Relative NMHC Emissions Change from Base (CRC H: 215 T50, 330 T90, 0% EtOH)
    50.0%
    40.0%
    -10.0%
    -20.0%
                            CRC Test Fuels (in order of increasing O2, T50, T90)
       From Figure 3.1-4, it is apparent that both models do a better job at predicting changes in
NMHC emissions than was the case for NOx emissions. The EPA E-67 model is clearly the
more accurate of the two models. However, this is to be expected given it was based on the data
being depicted.  The ability of the EPA Predictive Model to predict the general trend of nearly all
the CRC E-67 test fuels is impressive, given it is based on data from Tier 0 vehicles with 5-10
times the NMHC emission levels of the vehicles in the E-67 test program.  Overall, it appears
that NMHC emissions from LEVs and cleaner vehicles are even more sensitive to changes in
fuel quality than NMHC emissions from Tier 0 vehicles.

       The preceding figures illustrate the differences between the models for all 12 fuels
included in the E-67 test program.  Some of these fuels are more practical, or likely to be
commercially produced, than others. Based on the results  of AAM fuel surveys presented in
Chapter 2, summertime E10 blends will generally have levels of T50 and T90 that are about 29
°F and 7 °F lower than non-ethanol blends.  Thus, it could be useful to focus on sets  of fuels in
the CRC E-67 study which reflect these differences.

       The fuel pair which most closely reflects these differences are CRC fuel "C", a 10 vol%
ethanol blend, and CRC fuel "H", a non-oxygenated fuel.  Both fuels have a mid-range level of
T90. A second, more complex set of fuels involve those with higher levels of T90.  The CRC
"E" fuel contains 10 vol% ethanol and has the high level of T90.  However, there is  not a good
match to this fuel which is non-oxygenated. Yet two non-oxygenated fuels ("D" and "K"), when
considered together, represent a reasonable match to fuel "E.  Fuel D reflects no change in T50
relative to fuel E, while fuel K reflects slightly more than a typical drop in T50. Thus, by
averaging the emissions for fuels K and D  and then comparing this to the emissions  with fuel E,
                                          132

-------
we are able to generate a second direct indication of the impact of ethanol blending on emissions
from these low emitting vehicles. The general properties of these five fuels and the emissions
changes predicted by the two models are shown in Table 3.1-6 below.

                                       Table 3.1-6.
                     Predicted NOx and NMHC Emissions Changes
                          for EPA E-67and Predictive Models
                                                             48
Fuel Changes
T50 (°F)
T90 (°F)
Oxygen (vol%)

Change in Emissions
EPA Predictive Model NOx
EPA E-67 NOx
Actual E-67 Data

EPA Predictive Model NMHC
EPA E-67 NMHC
Actual E-67 Data
HtoC
Mid - Low
Mid - Mid
0-10


9.5%
1 1 .0%
3.3%

-3.7%
-3.8%
-6.3%
KtoE
High - Low
High - High
0-10


8.4%
5.8%
1 .6%

-11.1%
-17.3%
-21 .0%
DtoE
Low - Low
High - High
0-10


9.4%
10.4%
-1 .8%

7.3%
8.8%
9.2%
       As shown in Table 3.1-6, the two models agree quite closely on the effect of fuel C
relative to fuel H on both NMHC and NOx emissions. However, that said, both models tend to
overestimate the impact of fuels C and E on NOx emissions and underestimate the impact of
these fuels on NMHC emissions.

       Regarding the comparison of fuel E to fuels K and D, the two models tend to agree on the
effect of fuel E to fuel D, but differ more with respect to the effect of fuel E to fuel K,
particularly for NMHC emissions.  One reason for the difference in the latter comparison is that
the EPA E-67 NMHC model is more sensitive to very high levels of T50 than the EPA
Predictive Model for NMHC.

       Overall, the results of the E-67 study suggest that our assumption that Tier 1 and later
vehicles would not be sensitive to fuel parameters such as ethanol, T50 ad T90 (made in our
consideration of California's request for a waiver of the RFG oxygen requirement) may not be
valid. The observation that NMHC emissions from LEVs, et. al. could actually be more
sensitive than Tier 0 vehicles (on a percentage basis), particularly challenges our assumption.
While the effect of fuel quality on NOx emissions from low emitting vehicles is still not clear
from the recent test data, these emissions do appear to be sensitive to fuel quality.
48 For an additional comparison between the models with an expanded set of fuels to be used later in this analysis,
refer to Table 3 A-2 in Appendix 3 A.
                                           133

-------
3.1.1.1.3.3.   Effect of Oxygen Content Alone on Emissions: CRC E-67 Study

       In order to provide an estimate of fuel-emission effects comparable to those of the other
studies of the effect of fuel quality on emissions from LEV and later vehicles, we selected three
pairs of fuels tested in the CRC E67 study where the only change in fuel properties was oxygen
content.  One pair of fuels compared non-oxygenated fuel versus E6 (fuels A and B). Two pairs
of fuel compared non-oxygenated fuel versus E10 (fuels D  and E and fuels F and G). We
applied mixed univariate statistical models to the logarithm of the emission data from fuels A
and B and to fuels D, E, F and G, with vehicle as a random variable and fuel type as a fixed
variable.  Table 3.1-7 presents the results in terms of the percentage change in emissions between
the non-oxygenated fuel and the ethanol blend in each pair.

                                      Table 3.1-7.
       Matched Fuel Pair Results from the CRC E-67 Study: Effect of Oxygen Alone


6 vol% ethanol vs. no oxygenate
(Fuels A and B)
10 vol% ethanol vs. no oxygenate
(Fuels D, E, F and G)
NMHC
% Change
-4.5%*
4.7%
CO
% Change
-7.5%
-18.1%
NOx
% Change
2.6%
9.5%
               * Bold type indicates the difference was significant at a 90% confidence level.
       As can be seen, the addition of 6 vol% ethanol increased NOx emissions, while
decreasing NMHC and CO emissions. None of the effects were statistically significant at a 90%
confidence level.  The addition of 10 vol% ethanol decreased CO emissions, while increasing
NMHC and NOx emissions.  All three pollutant effects were statistically significant at a 90%
confidence level.
3.1.1.1.4.
AAM-AIAM Sulfur and Oxygenate Study
       AAM, together with AIAM and Honda, performed a test program in 2001 to evaluate the
effect of fuel sulfur and oxygen on emissions of CARB Tech 5 vehicles (i.e., LEV and later
vehicles)888. This program was performed at the request of the CARB in conjunction with the
MTBE ban and new Phase 3 Cleaner Burning Gasoline regulations. The first of the program's
two distinct goals was to evaluate the emissions of very low sulfur fuels at 1, 30, and 100 ppm
sulfur, with other fuel parameters held constant. Part two was to compare the emissions effects
of MTBE and ethanol blended fuels to non-oxygenated fuel, again with other parameters held as
constant as possible.  The fuel specifications for the oxygenated fuel matrix are shown in Table
3.1-8. (The sulfur-related testing is not relevant here, so the fuels which  only reflect a change in
sulfur are not shown.)
                                          134

-------
                                      Table 3.1-8.
                     AAM Fuel Properties for Oxygenated Test Fuels
Test Fuel
RVP (psi)
Arom (vol.%)
Olefins (vol.%)
T50 (°F)
T90 (°F)
MTBE/EtOH (vol.%)
Sulfur (ppm)
Non-Oxy
7.1
24.6
4.6
210
297
0
35
MTBE
6.9
22
3.4
202
294
11.3
31
E10
7.1
21.6
4.1
205
291
11.5
28
       The sulfur effects portion of this study tested 13 TLEV, LEV, and ULEV vehicles while
the oxygenate phase of the study used 8 and 2 of phase 1's LEV and ULEV vehicles,
respectively. These vehicles were a mixture of light and medium duty passenger cars and trucks
of undisclosed make and model year.  The FTP 75 was selected as the drive cycle, with regulated
data collected both at the tailpipe and engine out for a subset of vehicles to evaluate catalyst
efficiency.

       The sulfur related testing found a clear relationship between the level  of sulfur in the fuel
and the natural log of emissions (CO, NOx and NMHC). Bag weighted emissions of NOx, CO,
and HC emissions were reduced by 16%, 12%, and 11% (respectively) when fuel sulfur levels
were reduced from 30 to 1  ppm.  The effects found in the oxygenate portion of the study were
less clear, since some results were not statistically significant.  Average bag weighted emissions
of HC and CO tended to decrease with increasing oxygen  content.  Both oxygenated fuels
showed a decrease in NOx emissions, with MTBE having slightly lower emissions than ethanol.
The details of the statistical procedures applied to the data were not described in the
documentation.

       In order to provide  a consistent basis for comparing the results of this study to the other
studies of the effect of oxygenate on LEV and later vehicle emissions, we applied a mixed
univariate statistical model to the logarithm of the emission data from the AAM-AIAM
oxygenate study, with vehicle as a random variable and fuel as a fixed variable.  Table 3.1-9
presents the results in terms of the percentage change in emissions between the 11 vol% MTBE
blend, the 11 vol% ethanol blend and the non-oxygenated  fuel.

                                       Table 3.1-9.
         Fuel Effects from the AAM-AIAM Oxygenate Study: EPA Mixed Model


1 1 vol% MTBE vs. no oxygenate21
1 1 vol% ethanol vs. no oxygenate
1 1 vol% ethanol vs. 1 1 vol% MTBE
NMHC
% Change
-15.3%b
-12.6%
-1.4%
CO
% Change
-23.7%
-24.6%
-5.7%
NOx
% Change
-12.7%
-6.6%
25.2%
 Fewer vehicles were tested on the MTBE blend.
' Bold type indicates the difference was significant at a 90% confidence level.
                                          135

-------
       As can be seen in Table 3.1-9, the addition of 11 vol% MTBE was found to significantly
reduce the emissions of all three pollutants by 10% or more relative to the non-oxygenated
California RFG. The addition of 11 vol% ethanol was also found to significantly reduce the
emissions of all three pollutants relative to the non-oxygenated California RFG.  However, the
reductions were smaller than those of the MTBE blend for all three pollutants. Also, the
reduction in NOx emissions was not statistically significant at a 90% confidence level.  When we
compared the emissions with the ethanol blend to those with the MTBE blend, we found that the
increases were all statistically  significant at a 90% confidence level.

       Based solely on this  single study of seven vehicles and three fuels, it appears that adding
ethanol to a severely reformulated gasoline while holding other properties constant tends to
reduce NMHC, CO and NOx emissions.  However, replacing MTBE with the same volume of
ethanol tends to increase these emissions. Focusing on just this one study, the effect of increased
ethanol use could differ in RFG areas, where MTBE has traditionally been used, and
conventional fuel areas, where no oxygenate has traditionally been used. However, this one
study is not a sufficient basis to draw a firm conclusion regarding the effect of ethanol blending
on exhaust emissions.
3.1.1.1.5.
ExxonMobil Study of Oxygenate Type and Content
       In the fall of 1999, ExxonMobil (Mobil Oil at that time) conducted a test program to
investigate the emissions effects of MTBE and ethanol on LEV and ULEV vehicles. The
information which follows was made publicly available on the CARB websiteccc and is taken
directly from a presentation posted on that webpage. The vehicles tested are listed in Table 3.1-
10.
             Table 3.1-10. Vehicles Tested in 1999 ExxonMobil Study

Make/Model
1999 Dodge Stratus
1 999 Chevrolet Malibu
1999 Mazda Protege
1999 Ford Crown Victoria
1998 Honda Accord
Emissions
Calibration
LEV
LEV
ULEV
LEV
ULEV

Enqine
2.4L
3.1L
1.6L
4.6L
2.3L
Fuel
System
MFI
PFI
MPF
SFI
MFI
       Four test fuels were developed with varying MTBE and ethanol content, with other
parameters controlled as tightly as possible. The fuel specifications are listed below in Table
3.1-11.
                                          136

-------
                    Table 3.1-11. ExxonMobil Test Fuel Specifications
Test Fuel
RVP, psi
T50 (°F)
T90 (°F)
MTBE (vol%)
Ethanol (vol%)
Aromatics (vol%)
Olefins (vol%)
Oxygen (wt%)
M1
6.5
192
279
0
0
21
<1
0
M2
6.4
191
272
10
0
21
<1
1.82
M3
7.3
198
274
0
7
21
<1
2.436
M4
7.2
203
282
0
10
21
<1
3.48
       Vehicles were tested in duplicate over the FTP 75 drive cycle, and composite weighted
emissions for regulated pollutants reported.  The average results for all vehicles, based on least
squares means from the analysis of variance, is shown in Table 3.1-12., below.  The oxygenate
effect is the percent change in emissions between a given fuel and the base fuel Ml.

       Table 3.1-12. Average Exhaust Emissions: ExxonMobil Study (all vehicles)

Fuel
M1
M2
M3
M4

Oxvqenate
None
10% MTBE
7% EtOH
10%EtOH
Exhaust Emissions, q/mi
HC
0.058
0.058
0.059
0.061
CO
0.70
0.72
0.66
0.66
NOx
0.187
0.198
0.213
0.239

HC

0%
1%
5%
Oxygenate
Effect (a'
CO

3%
-6%
-6%

NOx

6%
14%
28%
       (a) Bold font represents statistically significant (a=0.1)
       The bold numbers in the above table represent statistically significant differences in
emissions at the 90% confidence level.  Only the effect of ethanol on NOx emissions was found
to be statistically significant while the effects on other criteria pollutants were not. As was the
case for the AAM-AIAM study, the presentation to CARB did not describe the statistical
analysis applied to the detail in sufficient detail to replicate the results. Thus, we applied a
mixed univariate statistical model to the logarithm of the emission data from the ExxonMobil
oxygenate study, with vehicle as a random variable and fuel type as a fixed variable. Because of
the greater number of fuels in this study, we applied two different types of mixed models.  One
set of models compared emissions between various pairs of fuels (e.g., Ml and M2). We applied
five models of this type. A sixth model used oxygen content as the fuel variable and considered
the significance of the square  of oxygen content, as well. The upper half of Table 3.1-13
presents the results in terms of the percentage change in emissions between the various fuel
pairs.  The bottom half of Table 3.1-13 presents the estimated emission effects for 2.1 wt% and
3.5 wt% oxygen fuels using the results of the fuel oxygen content based model.
                                           137

-------
       Table 3.1-13. Fuel Effects from the ExxonMobil Oxygenate Study


THC
% Change
CO
% Change
NOx
% Change
Two Fuel Comparisons
10 vol% MTBE vs. no oxygenate
7 vol% ethanol vs. no oxygenate
10 vol% ethanol vs. no oxygenate
10 vol% ethanol vs. 10 vol% MTBE
-2.5%a
-0.9%
3.6%
6.3%
-5.1%
-11.3%
-11.6%
-7.6%
-0.4%
9.3%
20.3%
21.1%
Oxygen Content model
2.1 vol% Oxygen
3.5 vol% Oxygen
-2.0%
3.5%
-9.1%
-15.2%
2.9%
17.3%
        Bold type indicates the difference was significant at a 90% confidence level.
       As can be seen, the results of our statistical analysis differ from those performed by
ExxonMobil. This may have been due to our focus on the logarithm of emissions or the use of a
mixed model.  As shown in Table 3.1-13, the addition of 10 vol% MTBE was found to reduce
the emissions of all three pollutants relative to the non-oxygenated California RFG. While the
differences tended to be substantial in magnitude on average, they were not statistically
significant at a 90% confidence level. The addition of 7 vol% ethanol to the non-oxygenated
California RFG (with an increase of roughly 1.0 psi RVP) was also found to reduce the
emissions of all three pollutants. Again, the reductions in NMHC and NOx emissions were not
statistically significant. While the  reduction in CO emissions was slightly smaller on average
than that for MTBE, the effect was more consistent across vehicles and statistically significant at
a 90% confidence level.  The addition of 10 vol% ethanol to the non-oxygenated California RFG
(again with an increase of roughly  1.0 psi RVP) was also found to reduce emissions of NMHC
and CO, but increased NOx emissions slightly. Like that for the 7 vol% ethanol blend, only the
CO emission effect was statistically significant at a 90% confidence level. Finally, the
substitution of 10 vol% ethanol for 10 vol% MTBE was found to increase emissions of THC and
NOx substantially, while reducing  CO emissions slightly. None of the effects were significant at
a 90% confidence level, though the NOx increase was nearly so (e.g., 89% confidence).

       With respect to the model using oxygen content to describe the four fuels in the study, the
effect of oxygen on emissions was found to be statistically significant for all three pollutants. In
addition, the square of the oxygen  content was statistically significant at the 90% confidence
level for THC and NOx emissions.  The predictions shown in Table 3.1-13 utilize the square of
oxygen content for these two pollutants, but not for CO emissions. As can be seen, a 2.1 wt%
oxygen fuel (e.g., an 11 vol% MTBE or 6% ethanol blend) is predicted to decrease THC and CO
emissions, but increase NOx emissions. A 3.5 wt% oxygen fuel (e.g., a 10% ethanol blend) is
predicted to decrease CO emissions, but increase THC and NOx emissions.

       The two modeling approaches produce markedly different predictions for a 10 vol%
ethanol blend,  especially for THC and NOx. The direct comparison of the clear fuel and the El 0
blend shows the E10 blend to reduce THC and NOx slightly. In contrast, the oxygen content
approach predicts that the E10 blend will increase both THC and NOx emissions. This
                                           138

-------
underlines the need for more data to accurately predict the emission effect of various types and
levels of oxygenate on modern vehicles.
3.1.1.1.6.
Toyota Study of MTBE and Ethanol Blends
       Toyota Motor Company of Japan performed a small study, presented in 2000, evaluating
the effect of MTBE and ethanol blended gasoline on exhaust emissions of nine LEV, TLEV,
ULEV vehicles.DDD  The model years for these vehicles were not presented.. Three test fuels
were evaluated in this study: Phase 2 California RFG containing MTBE, a matched RVP E10
and a higher RVP E10.  However, exhaust emission testing was only performed on the MTBE
fuel and the higher RVP ethanol fuel. The fuel properties for these test fuels are listed in Table
3.1-14, below:

              Table 3.1-14. Fuels Tested in the Toyota Oxygenate Test Program
Fuel
Parameter
RVP (psi)
Arom (vol.%)
Olefins (vol.%)
T50 (°F)
T90 (°F)
MTBE (vol.%)
EtOH (vol.%)
Sulfur (ppm)
MTBE Blend
6.8
24
5
156
290
11.1
0
30
Matched RVP Ethanol
Blend
7
23
5
208
294
0
11.2
29
Higher RVP Ethanol Blend
7.6
24.1
3.4
212
297
0
8.9
30
       As can be seen in the above table, aromatics, olefins, and sulfur were held relatively
constant while other parameters varied. The emission test cycle used was not stated. We assume
it was the FTP 75 test. Only regulated emissions results are provided.

       The study found that, on average across all vehicles and tests, NOx emissions increased
by 5% for E10-B relative to MTBE. Correspondingly, CO emissions were reduced by 6% and
NMHC emissions were decreased by 0.6% for E10-B relative to MTBE.

       Again for comparison purposes, we applied a mixed univariate statistical model to the
logarithm of the emission data from the Toyota oxygenate study, with vehicle as a random
variable and fuel type as a fixed variable. Table 3.1-15 presents the results in terms of the
percentage change in emissions between the 11 vol% MTBE and the 9 vol% ethanol blends.

             Table 3.1-15.  Fuel Effects from the Toyota  Oxygenate Study


9 vol% ethanol vs. 1 1 vol% MTBE
NMHC
% Change
-0.8%a
CO
% Change
-6.1%
NOx
% Change
3.8%
        Bold type indicates the difference was significant at a 90% confidence level.
                                          139

-------
       As can be seen, the substitution of 9 vol% ethanol for 11 vol% MTBE in a RFG-type
blend was found to reduce the emissions of NMHC and CO, but increase NOx emissions.  None
of the emission effects was found to be statistically significant at a 90% confidence level.

       Based solely on this single study of nine vehicles and three fuels, it appears that replacing
MTBE with roughly the same volume of ethanol in reformulated gasoline while holding other
properties constant (except RVP) tends to reduce NMHC and CO, and increases NOx emissions.
It is not possible to separate the effect of ethanol from RVP in this study.
3.1.1.1.7.
Mexican Petroleum Institute Fuel-Emission Effects Study
       In 2006, the Institute Mexicano del Patroleo (hereafter referred to as Mexican Petroleum
Institute) published a paper on a recently conducted research program investigating the effects of
gasoline properties on exhaust and evaporative emissions on 30 light duty cars and trucks,
ranging in model year from 1993 to 2002.EEE  The fuel quality parameters investigated include
RVP, oxygen, Sulfur, olefms, aromatics, and distillation parameters.  The results of this study
were used to develop a statistical model for predicting emissions based on fuel quality for use in
guiding national air quality improvement program policy.  In their analysis, a comparison was
made between the "predictive model," developed based in this test data, and EPA's Complex
model. The end result is a general qualitative agreement with the EPA Complex model, with
some quantitative differences pertaining to the vehicles and fuels used in each model's
development. Selected properties of the test fuels are shown below in Table 3.1-16.

      Table 3.1-16.  Fuel Tested in the Mexican Petroleum Institute Fuel Effects Study
Fuel
1
2
3
4
5
6
7
8
9
10
11
12
L-S
Ref
ZM
Aromatics
(vol%)
19.1
19.3
18.8
20.6
19
18.7
40.2
20.7
35.8
19.8
20
19.8
40.3
28
24.1
Olefins
(vol%)
6.6
6.6
6.5
6.7
6.6
6.9
6.9
15
15.5
5
7.4
6.1
4.8
13.5
9
Oxygen a
(wt%)
0
0.98
2.03
2.1
1
1.03
0.98
1.07
1.06
1.03
0.98
1.05
1.14
0.34
1.21
Benzene
(vol%)
0.86
1.09
0.9
1.13
1.15
0.98
2.26
1.31
1.25
0.6
0.8
0.75
1.06
1.14
1
RVP
(psi)
6.8
6.8
6.9
7
8.6
10.7
6.6
10.9
10.8
6.6
6.7
6.6
8.1
8.9
7.7
T50
(°F)
224
222
216
227
213
199
233
197
209
220
223
222
232
207
213
T90
(°F)
325
324
324
326
323
325
324
322
326
335
321
321
324
332
326
Sulfur
(ppm)
411
412
406
386
423
387
402
415
403
89
209
817
34
724
403
        Oxygenate is MTBE, except fuel 4 (Ethanol)
                                          140

-------
       The base fuel from which other fuel parameters were varied to form additional fuels is
indicated as fuel 2 in the above table, and is a low-level MTBE fuel blended to represent a
composite average of all brands of commercially available Mexican gasoline.  The reference fuel
for comparative fuel effects purposes, fuel "Ref", is a fuel blended to have refinery average
levels of sulfur, benzene, RVP, aromatics, olefins, and distillation properties for the year 2000.

       The technology groups investigated were referred to as either "Tier 0"  and "Tier 1"
vehicles, which is a bit of a misnomer as these vehicles were not equipped with on-board
diagnostic equipment (OBD), nor were they subject to emissions durability standards49. Rather,
each technology class acts as a surrogate for emission control technologies with the certification
standards shown in Table 3.1-17.
Table 3.1-17. Certification Standards of Test Vehicles by Technology Class
Technology class
"Tier 0"
"Tier 1 "
# of Vehicles
12
14
Model Years
1993-1998
1 999-2002
CO (g/km)
2.1
2.1
NMHC (g/km)
0.25 (THC)
0.156
NOx (g/km)
0.62
0.25
       Vehicles were tested on a chassis dynamometer over the FTP-75 test. Regulated
emissions, as well as speciated hydrocarbons and carbonyls, were collected for each test.  The
procedures and statistical methods employed to develop the predictive model for this test
program were similar to those used to construct the complex models for exhaust VOC and NOx
emissions.

       The emissions test results were separated into "Tier 0" and "Tier 1" categories, and
reported as mean emissions rates for all vehicles of that type on each test fuel. The natural
logarithm of emissions was then regressed to develop a predictive model, using a statistical
approach that is "similar to the techniques used to construct the complex model for exhaust VOC
and NOx emissions" (Schifter et al, 1275). The report then goes into details on the experimental
vs. model predicted results, as well as a validation of their model which will not be discussed
here.  The model predicted percent changes in emissions (for both vehicle fleets) for the Mexican
Petroleum Institute predictive model and EPA Complex model are shown in Figures 3.1-5 and
3.1-6, below.
49 The Mexican Environmental Agency currently mandates emissions durability certificates to be issued on new
vehicles
                                           141

-------
       Figure 3.1-5.  Emissions Effects Predicted by EPA Complex Model
                       3 Nth,
      Onk-    L-nlbc   H-ntbc   BOH   M-RVP   H-RVP   Hiarom  H-olefRVP H-afct-orom  L-s

                                            Fuels
                                                                     M-s     H-s
            Figure 3.1-6.  Emissions Effects Predicted by Mexican Petroleum Institute
                                    "Predictive" Model
                                      II-RVI-   l|;a«m  M-olePKV? I I-ntf-amm
                                            lutls
       The values above represent average emissions across both vehicle technology fleets
considered together, and with the industry average fuel ("Ref') as a baseline for predicted
changes in emissions. As you can see, their model predicts an even greater reduction in exhaust
THC and NOx emissions with ethanol (and MTBE) than the complex model, along with
directionally inconsistent results for fuels with high aromatics, RVP, and olefms. This is partly
due to differences in the fuels and vehicles used to develop each model (along with the properties
of the base fuel selected) and speaks to the fact that there is a high degree of uncertainty and
sensitivity to consider when extrapolating the results  of a fuel effects model to the larger vehicle
and fuel population.

       As with the other studies, we applied a mixed univariate statistical model to the logarithm
of the emission data from the Mexico fuel study, with vehicle as a random variable  and fuel type
as a fixed variable.  We restricted the vehicles to those which had NOx emissions of 0.20 g/mi or
                                            142

-------
less on the low sulfur fuel (30 ppm) included in the test matrix. This resulted in the inclusion of
seven vehicles (numbers 9, 15, 25, 27, 28, 29, and 31).  Table 3.1-18 presents the results in terms
of the percentage change in emissions between the non-oxygenated base fuel, the 5.5 vol% and
11 vol% MTBE blends and the 6 vol% ethanol blend.

                 Table 3.1-18.  Fuel Effects from the Mexico Fuels Study


5.5 vol% MTBE vs. no oxygenate
1 1 vol% MTBE vs. no oxygenate
6 vol% ethanol vs. no oxygenate
6 vol% ethanol vs. 6 vol% MTBE
6 vol% ethanol vs. 1 1 vol% MTBE
NMHC
% Change
11.1% a
-11.1%
3.9%a
-5.8%
19.5%
CO
% Change
-5.5%
-9.7%
-6.3%
-0.9%
4.0%
NOx
% Change
-3.1%
10.0%
27.2%
32.8%
14.2%
        Bold type indicates the difference was significant at a 90% confidence level.
       As can be seen, the addition of 5.5 vol% MTBE to a low-RVP type fuel at constant RVP
was found to increase the emissions of NMHC and CO, and decrease NOx emissions.  The
addition of 11 vol% MTBE to a low-RVP type fuel at constant RVP was found to increase the
emissions of NOx and CO, and decrease NMHC emissions. Only the NMHC reduction was
found to be statistically significant at a 90% confidence level. The addition of 6 vol% ethanol to
a low-RVP type fuel at constant RVP was found to reduce the CO emissions, and increase
NMHC and NOx emissions.  Only the NOx emission increase was statistically significant at a
90% confidence level.

       The substitution of 6 vol% ethanol for 11 vol% MTBE at constant RVP and oxygen
content was found to decrease NMHC emissions and increase NOx emissions.  CO emissions
were essentially unchanged.  Only the NOx emission increase was statistically  significant at a
90% confidence level. The substitution of 10 vol% ethanol for 11 vol% MTBE was found to
increase emissions of all three pollutants. Only the NMHC emission increase was statistically
significant at a 90% confidence level.

       Based solely on this single study of six vehicles and three fuels, it appears that either
MTBE or ethanol blends with roughly 2 wt% oxygen increase NOx emissions, while the effect
on NMHC and CO emissions are inconsistent.  Substituting ethanol for MTBE at the same
oxygen content appears to increase NMHC, CO and NOx emissions.
3.1.1.1.8.
Overview of LEV and Later Vehicle Studies
       The differences in the details of the various studies prevent a simple quantitative
comparison of their results.  However, we have performed a qualitative comparison by simply
determining whether the study found an increase or a decrease in emissions of 2% or more and
whether the effect was statistically significant at 90% confidence or not. The results of this
determination are shown in Table 3.1-19 below for a number of fuel pairs.
                                          143

-------
     Table 3.1-19. Summary of Oxygenate Emission Effects: LEV and Later Vehicles


10/11
vol%
MTBE
E6/E7
E10/E11
E10/11
vs.
MTBE
THC/NHMC
Lower
AAM-AIAM
Mexico
ExxonMobil*
Mexico
CRC E67
AAM-AIAM
Toyota
AAM-AIAM
Higher


CRC E67
ExxonMobil
ExxonMobil
CO
Lower
AAM-AIAM
Mexico
ExxonMobil
Mexico
ExxonMobil
CRC E67
AAM-AIAM
ExxonMobil
CRC E67
ExxonMobil
AAM-AIAM
Higher




NOx
Lower
AAM-
AIAM
Mexico

AAM-
AIAM
Toyota
Higher

Mexico
ExxonMobil
CRC E67
ExxonMobil
CRC E67
AAM-AIAM
ExxonMobil
       Starting with MTBE, none of the three studies which tested both non-oxygenated and
MTBE fuels found MTBE to increase the emissions of any of the pollutants. One to two studies
found statistically significant reductions in THC/NMHC, CO and NOx emissions with the use of
MTBE.

       Three studies tested a non-oxygenated fuel and a 6-7 vol% ethanol blend. Ethanol
blending at this level likely reduces CO emissions and increases NOx emissions. Two of the
three studies showed an increase in NMHC emissions, but neither result was statistically
significant at a 90% confidence level. Thus, the effect of E6 on NMHC emissions is particularly
unclear.

       Three studies tested a non-oxygenated fuel and a 10-11 vol% ethanol blend. Ethanol
blending at this level likely reduces CO emissions, as all three studies showed a statistically
significant reduction. Two of three studies found an increase in NOx emissions, while one found
a decrease,  all statistically significant at a 90% confidence level. Again, the effect of ethanol
blending on exhaust NMHC emissions is not clear. Two of three studies found an increase in
THC/NMHC emissions, while one found an increase. The results of one of the two studies
finding an increase and those of the study finding a decrease were statistically significant at a
90% confidence level.

       Finally, three studies tested both MTBE and 10-11 vol% ethanol blends. Ethanol
blending at this level appears to reduce CO emissions relative to MTBE.  The effect on the other
two pollutants is less clear.  Two of three studies found a decrease in THC/NMHC emissions,
though the one study finding an increase was the only one where the result was statistically
significant at a 90% confidence level. Two of three studies found an increase in NOx emissions
and both of these were statistically significant at a 90%  confidence level.
3.1.1.1.9
Selection of Models for Exhaust NMHC, CO and NOx Emissions
                                           144

-------
3.1.1.1.9.1     Exhaust NMHC and NOx Emissions

       For Tier 0 vehicles, the EPA Predictive Models are based on more data and the most
advanced statistical tools. Therefore, we will use these models here to project the effect of
increased ethanol use and decreased MTBE use on exhaust VOC and NOx emissions.

       For Tier 1 and later vehicles, the choice is much less clear.  In our analysis of California's
request for an RFG oxygen waiver, we assumed that there was too little data upon which to
project the  effect of fuel quality on the emissions of these vehicles.  Substantially more data exist
on the effect of oxygenates on LEV and later vehicle exhaust emissions than did in 2001.
However, as indicated by the summary of these results shown in Table 3.1-19, there does not
appear to be sufficiently consistency to confidently predict the impact of oxygenate type and
content on exhaust NMHC and NOx emissions.

       For the NPRM, we developed two separate sets of predictions: a primary analysis
assuming no effect of oxygen on NMHC and NOx emissions from Tier 1  and later vehicles, and
a sensitivity analysis which applied the Predictive Model effects to Tier 1 and later vehicles.  The
qualitative  summary shown in Table 3.1-19 supports a continuation of this approach. The effect
of ethanol blending on NMHC emissions is unclear in Table 3.1-19.  This is reasonable
bracketed by the primary and analysis, which assumes no effect and the sensitivity analysis
which assumes a reduction.  Table 3.1-19 also indicates that five out of six studies found that 6-
10 vol% ethanol blends increased NOx emissions from LEV and later vehicles.  This is also
reasonable bracketed by the primary and analysis, which assumes no effect and the sensitivity
analysis which assumes an increase. Given the uncertainty in the fuel-emissions effects for Tier
1 and later vehicles, there is also some value in maintaining consistency with our analysis
conducted in response to California's request for an RFG oxygen waiver. The primary analysis
does this.

       The varied results across these studies indicate the need for additional test data. It may
also be possible in the future to combine the emission data from  all such studies (as was done for
the Complex and Predictive Models) in order to develop a more robust estimate of the impacts of
oxygenate blending on emission from late model year vehicles.

3.1.1.1.9.1     Exhaust CO Emissions

       For Tier 0 vehicles, MOBILE6.2 is the standard modeling tool for estimating the effect of
fuel quality on CO emissions. Therefore, we will use that model here to project the effect of
increased ethanol use and decreased MTBE use on CO emissions.

       Regarding later vehicles, the five studies of LEV and later vehicles all found that
increasing oxygen content in terms of MTBE or ethanol reduces CO emissions. This is
consistent with both the primary and sensitivity analyses, as both approaches include a reduction
in CO emissions.
                                           145

-------
       The five studies of LEV and later vehicles all tested normally emitting vehicles.
MOBILE6.2 estimates that an E10 blend will reduce CO emissions from Tier 0 vehicles by 11%.
The quantitative results of the five studies generally support this degree of reduction (Toyota,
ExxonMobil, Mexican Petroleum Institute) or perhaps a larger degree of reduction (CRC E67,
AAM-AIAM). Due to the absence of an EPA Predictive Model for CO emissions, for the
NPRM, we did not develop a sensitivity case for CO emissions.  However, given the possibility
that the CO emission reduction is larger than that currently estimated by MOBILE6.2 for these
vehicles, it appears reasonable to include a sensitivity analysis for CO emissions, as well as
NMHC and NOx. An approach analogous to that taken for NMHC and NOx emissions appears
reasonable. That is, for the primary analysis, we will continue to use MOBILE6.2 to project the
effect of fuel properties on CO emissions. This means essentially a 6.7% reduction in CO
emissions from Tier 1 and later vehicles for an E10 blend. For the sensitivity analysis, we will
apply the MOBILE6.2 CO emission reduction of 13.8% for Tier 0 vehicles to Tier 1 and later
vehicles.

       As discussed above, the five studies  of LEV and later vehicles are not sufficient for use in
quantitatively projecting the impact of fuel quality on emissions from these vehicles.  Additional
data must still be collected over a broader set of vehicles, fuel changes, and conditions.

3.1.1.2       Exhaust Toxic Emissions

       Two EPA models project the impact of fuel quality on exhaust toxic emissions: the
Complex Model and MOBILE6.2. The Complex Model projects the impact of fuel quality on
toxic emissions directly. That is, any impact of fuel quality on total exhaust VOC emissions
(which includes the air toxics) is implicitly included in the model's predictions. MOBILE6.2
separates the process into  two steps. Total exhaust VOC emissions are projected first, in part
based on fuel quality. Then, the fraction of VOC represented by each air toxic is estimated, in
part based on fuel quality.

       The effect of fuel quality on exhaust VOC emissions in MOBILE6.2 was already
discussed above. The effect of fuel quality on the fraction of exhaust VOC emissions
represented by each air toxic in MOBILE6.2 is based on the projections contained in the
Complex Model. These Complex Model's effects  of fuel quality on exhaust toxic emissions
were used with the effect of fuel quality on exhaust VOC emissions backed out. Thus, with
respect to the effect of fuel quality on the fraction of exhaust VOC emissions represented by each
air toxic, the Complex Model is the basis of both the Complex Model and MOBILE6.2
predictions.

       With respect to exhaust VOC emissions, we already decided above that the EPA
Predictive Models represent the best estimate for Tier 0 vehicles. For Tier 1 and later vehicles,
we assume in our primary analysis that these vehicles' exhaust VOC emissions are unaffected by
fuel quality.  As a sensitivity analysis, we decided to extend the impacts indicated by the EPA
Predictive Models to all vehicles.

       We follow the two-step process taken in MOBILE6.2 here in modeling the impact of fuel
quality on exhaust toxic emissions. We will use the EPA Predictive Models to project the impact
                                          146

-------
of fuel quality on exhaust VOC emissions, as discussed above. We will use the effects of fuel
quality on the toxic fractions of exhaust VOC emissions contained in MOBILE6.2.

3.1.1.3       Non-Exhaust Emissions

       Both the Complex Model and MOBILE6.2 evaluate the effect of gasoline quality on non-
exhaust VOC emissions. However, the effects in the Complex Model were taken from an older
version of MOBILE, as was mentioned above.  Therefore, MOBILE6.2 represents the better of
the two estimates of the effect of gasoline quality on non-exhaust VOC emissions.  The EPA
Predictive Models do not address non-exhaust emissions, so they are not applicable here.

       In EPA's second analysis of California's request for a waiver of the RFG oxygen
requirement, we enhanced the estimate of non-exhaust emissions in MOBILE6.2 by adding
additional permeation emissions related to the use of ethanol.FFF Recent testing at that time
indicated that ethanol increases the rate of permeation of hydrocarbons through plastic fuel tanks
and elastomers used in fuel line connections, as well as permeating ethanol itself.  Subsequent
testing has confirmed this effect. Therefore, we have added the effect of ethanol on permeation
emissions to MOBILE6.2's estimate of non-exhaust VOC emissions in assessing the impact of
gasoline quality on emissions here.

       Air Improvement Resource, Inc. for the American Petroleum Institute, recently
summarized the available test data on the effect of ethanol on permeation emissions and
developed a methodology for estimating in-use permeation emissions in several U.S. cities.GGG
This study provides a useful starting point for incorporating these emissions into this RFS
analysis.

       Before examining this study, it is useful to point out that the non-exhaust emission
estimates in MOBILE6.2 include permeation emissions for non-oxygenated gasoline. Typical
extended diurnal emission tests (e.g., those lasting 2-3 days) automatically include any emissions
permeating through plastic and elastomeric fuel system components. However, since the
emission tests used as the basis of the MOBILE6.2 estimates of non-exhaust emissions primarily
were performed with non-oxygenated gasoline.  Those tests that did include ethanol blends only
exposed the vehicle to this fuel for a few days.  The CRC study of ethanol-related permeation
indicates that it takes at least a week or two for the effect of ethanol to fully develop.  Therefore,
it is very unlikely that the tests performed by EPA and others to assess the impact of ethanol and
other fuel components on non-exhaust emissions included the effect of ethanol on permeation
emissions.  In those cases where a vehicle may have been exposed to an ethanol blend for some
time prior to testing, the increased permeation emissions likely were still present when the
vehicle was tested on a non-oxygenated gasoline, still masking the effect.  Therefore, our task
here is to develop an estimate of the incremental impact of ethanol use on permeation emissions,
and not an estimate of total permeation emissions with and without ethanol.

       The primary source of ethanol permeation emission data is the CRC E-65 study.HHH This
study tested 10 vehicles, 6 cars and 4 light trucks, ranging in model year from 1989 to 2001.
Permeation emissions were measured using two fuels, a non-oxygenated gasoline and a 6 vol%
ethanol blend like that commonly sold in California. AIR placed these vehicles into three
                                          147

-------
groups, based on a combination of model year and applicable evaporative emission standards.
The vehicles in the test program were certified to two distinct evaporative emission
requirements.  The older vehicles were certified to EPA's or California's 2 gram hot soak plus
diurnal emission standard based on an accelerated one-hour diurnal test. The three newest
vehicles were certified to the enhanced evaporative emission requirements first implemented in
the 1996 model year, which included an extended two or three day diurnal test. In addition, the
data indicated that the three pre-1990 model year vehicles had much larger incremental ethanol
permeation emissions than the later pre-enhanced evaporative emission vehicles.  Therefore, AIR
split the pre-enhanced evaporative emission category into two groups, pre-1990 model year
vehicles and 1990 and later model year vehicles. We believe that this is appropriate and apply
this split here, as well.

       Since the earliest calendar year during which emissions are assessed in the RFS analysis
is 2012 and MOBILE6.2 only considers vehicles which are 24 years old or newer, at most only
two model years of pre-1990 vehicles are present in our analysis.  Due to accumulated
scrappage, these vehicles comprise a very small percentage of the on-road fleet in 2012 and
disappear from our analysis by 2015.  Therefore, we decided to ignore the pre-1990 model year
data here.

       AIR estimated the average incremental ethanol permeation emission rates for the  1990
and later model year pre-enhanced evaporative emission vehicles to be 0.86 gram per day
(g/day), while that enhanced evaporative emission vehicles was 0.80 g/day. Given the small
number of vehicles tested and the variability in the rates measured for individual vehicles, for the
purposes of this analysis, we consider these two levels to be generally equivalent.  Therefore, we
use an average incremental ethanol permeation emission rate of 0.8 g/day for all vehicles. A
follow-on study performed by CRC indicates that the permeation emissions associated with a 10
vol%  ethanol blend could not be distinguished statistically from those of the 6 vol% blend.111
Therefore, we use the 0.8 g/day incremental permeation emission rate for both 6 vol% and 10
vol%  ethanol blends here.

       Beginning with the 2004 model year, EPA and California implemented further
enhancements to their evaporative emission standards. The EPA "Tier 2" requirements include
accumulating mileage on durability data vehicles with an ethanol blend. However, actual
emission testing is still performed using non-oxygenated gasoline. We believed that this
combination of requirements would incorporate any effects of ethanol on emissions, including
potential permeation effects. Because of these and other aspects of the 2004 and later standards,
AIR estimated that the permeation emissions due to ethanol would be reduced to 0.43 g/day for
these vehicles.

       We believe that it is likely that permeation emissions for non-oxygenated blends will be
lower for these vehicles, due to the fact that the diurnal emission standard was reduced from 2
g/day to 0.95 g/day and lower in some cases.  However, as mentioned above, the effect of
ethanol on permeation emissions takes about 2 weeks to fully develop and to fully disappear.
Therefore, it is possible to  accumulate mileage on a certification vehicle using an ethanol blend,
change the fuel to the emission test fuel, wait two weeks and then test the vehicle. In this case,
the effect of ethanol on permeation will have disappeared during the certification testing.  Thus,
                                           148

-------
until EPA requires certification emission testing with ethanol blends, we have no assurance that
manufacturers will modify their vehicle designs to address the effect of ethanol on permeation.
Thus, for the purposes of this analysis, we maintain the estimate of 0.8 g/day for Tier 2, and
earlier vehicles.

       Permeation emissions vary significantly with ambient temperature, with emissions
increasing with increases in temperature. The 0.8 g/day emission estimate applies at an average
temperature of 95 F. The literature indicates that permeation varies exponentially with
temperature. The CRC testing indicates that permeation emissions double with every increase in
temperature of 18 F. Vice versa, permeation emissions drop 50% with every decrease in
temperature of 18 F. We apply this relationship in Chapter 4 in developing incremental ethanol
permeation emissions for each hour of the day in each county in the U.S.

       We  plan to update our projections of the effect of gasoline quality on non-exhaust VOC
emissions from Tier 1 and later model year vehicles based on additional testing which is
expected to begin in 2007. Additional testing of permeation emissions is already underway with
the CRC E-77 test program.  These updated projections will be used in the comprehensive
assessment of the impact of the fuel-related provisions of the Energy Act which due in 2009.

       Non-exhaust emissions are a function of ambient temperature and temperatures vary
across the nation. Therefore, it is not as simple to determine the effect of RVP and other fuel
qualities on non-exhaust emissions on a per vehicle basis as it is for exhaust emissions.
Therefore, we performed a regression of the non-exhaust VOC and benzene emissions  developed
in Chapter 4 as a function of fuel properties in order to estimate these effects on a per vehicle
basis. Specifically, we regressed the ratio of non-exhaust VOC and benzene emissions in each
county in July between two fuel scenarios (the RFS case and the base case) against the change in
RVP, ethanol content and MTBE content. The results are summarized in Table 3.1-20.

         Table 3.1-20. Fuel-Non-Exhaust Emission Effects in MOBILE6.2: 2012

VOC
Benzene
RVP (%/psi)
11.8%
2.0%
Ethanol (%/Vol %)
0.3%
1.0%
MTBE (%/Vol%)
0.0%
-0.5%
Adjusted r-Square
0.53
0.50
       Non-exhaust emissions of benzene are estimated in MOBILE6.2.  MOBILE6.2 adjusts
these emissions for RVP, benzene fuel content, MTBE fuel content, temperature and the total
non-exhaust VOC emissions produced by the vehicle fleet in question. We will use MOBILE6.2
here to project the impact of decreased MTBE use and increased ethanol use on non-exhaust
benzene emissions.

       Benzene is also emitted via permeation. Just as MOBILE6.2 does not include the effect
of ethanol on VOC emissions via permeation, it does not include the effect of ethanol on benzene
emissions via permeation. Thus, we will add this effect outside of the MOBILE6.2 model in the
same way as was just described for VOC emissions via permeation.
                                          149

-------
       Both the CRC E-65 study and the E-65 Phase 3 follow-on study referenced above
measured the benzene content of permeation emissions during some of the tests performed.
Table 3.1-21 presents the available data by fuel type for both studies.

             Table 3.1-21. Benzene Permeation Emissions: CRC E-65 Studies
E-65 Phase
No. of
Vehicles
EO
E6 (Average
Aromatic s)
E6 (High
Aromatics)
E10
MTBE
Blend
Fuel Benzene Content (vol%)
Phase 1
Phase 3
—
—
0.73
0.41
0.72
0.55
—
0.43
___
0.51
0.53
___
Benzene Emissions (% of total permeation emissions)
Phase 1
Phase 3
10
4-5
2.5-2.6%
1.4%
2.2%
1.4%
—
1.4%
___
1.7%
2.2%
___
Benzene Emissions Adjusted to 0.88 vol% Fuel Benzene (% of total permeation emissions)
Phase 1
Phase 3
10
4-5
3.0-3.1%
3.0%
2.7%
2.2%
—
2.8%
___
2.9%
3.6%
—
       As can be seen from Table 3.1-21, the benzene content of permeation emissions is
slightly higher for the Phase 1 study than Phase 3.  This is consistent with the higher benzene
contents of the fuels tested in Phase 1. The Phase 1 fuels have particularly low benzene contents
compared to levels typical across the U.S.  Therefore, assuming a linear relationship between
benzene fuel content and benzene permeation emissions, we adjusted the benzene permeation
emissions to those for a fuel benzene content of 0.87 vol%, which is the average of the benzene
fuel content of summertime gasoline produced nationwide in the base, RFS and EIA cases from
the recent refinery modeling described in Chapter 2. We focused on summer benzene content
since permeation emissions are a strong function of temperature. These figures are shown in the
bottom third of Table 3.1-21. As can be seen, the benzene content of permeation emissions is
much more consistent across the two Phases of E-65 after being adjusted to consistent fuel
benzene content than before this adjustment.

       There appears to be no definite trend in the benzene content of permeation emissions with
increasing ethanol fuel content. Both CRC studies found that the addition of ethanol does not
simply increase permeation emissions via increased emissions of ethanol, but also through
increased emissions of other fuel components. Thus, we will  assume for this analysis that the
benzene fraction of permeation emissions is independent of ethanol fuel content (i.e., benzene
emissions increase at the same rate as total permeation emissions).  This is also consistent with
our conclusion in Chapter 2 that ethanol blending does not affect benzene fuel  content.  Given
this, we determined the average benzene fraction of permeation emissions by averaging the
benzene fraction across all the fuels and vehicles tested, excluding the MTBE blend.  This
average was 3%.  This figure will be used here to estimate the benzene portion of the increase in
permeation emissions resulting from ethanol blending.

       Given this, we determined the average benzene fraction of permeation emissions by
averaging the benzene  fraction across vehicles. Thus, the average benzene fraction of
                                          150

-------
permeation emissions is the average of the figures in the rightmost column of Table 3.1-21, or
2%.

3.1.1.4        PM Emissions

       The amount of data evaluating the impact of ethanol and MTBE blending on direct
emissions of PM from gasoline-fueled vehicles is extremely limited.  Three fairly limited studies
have evaluated the impact of ethanol blending on PM emissions from gasoline vehicles. These
studies are summarized below.

       The Colorado Department of Public Health and Environment tested 24 vehicles on two
winter grade commercial fuels at 35°F in Denver (i.e., at high altitude)/" Both fuels were
obtained from a local refinery.  One fuel was non-oxygenated and represented fuel sold outside
of the Denver area. The other contained  10 vol% ethanol and represented fuel sold in the Denver
area, which has an oxygenated fuel mandate.  As would be expected, the fuels differed in other
qualities besides ethanol content. The ethanol blend had a 2 vol% lower aromatic content, which
is somewhat less than expected. However, it also had a 53  F lower T50 level, which is a much
greater difference than is typical. The two fuels used during this testing appear to have been
used in random order (i.e., sometimes the non-oxygenated fuel was tested first, other times the
E10 fuel was tested first).

       Half of the 24 vehicles were certified to Tier 0 emission standards, while the other half
were certified to Tier 1  standards.  Each group of 12 vehicles included 8 cars and 4 light trucks.

       The study found that PM emissions for the 24 vehicles over the FTP decreased from
about 9 mg/mi to about 6 mg/mi with the ethanol blend, for a reduction of 36%. In addition, the
vehicles with the highest base PM emission rates showed by far the largest reductions, both in
absolute terms and in terms  of percentage. PM emissions from Tier 1 vehicles decreased from
roughly 5.5 mg/mi to 4 mg/mi with the ethanol blend, for a reduction of 27%. Essentially all of
the emission reduction occurred during Bag 1 of the test (i.e., related to the  cold start).

       PM emissions were also measured over a warmed up California Unified Cycle (i.e., no
cold start). PM emissions for the 24 vehicles over this cycle for the two fuels were not
statistically different. The ethanol blend  increased PM emissions from the Tier 0 vehicles
slightly and decreased those from the Tier 1 vehicles slightly.

       Finally, PM emissions were also measured over an EPA REP05 Cycle, again with no
cold start. PM emissions over this cycle were 4-5 times those over the California Unified Cycle,
indicating the impact of high speed, aggressive driving on PM emissions. However, despite this
general increase in PM emissions, for the 24 vehicles PM emissions over the REP05 cycle were
again very similar for the two fuels. This time, however, the ethanol blend decreased PM
emissions from the Tier 0 vehicles slightly and increased those from the Tier 1 vehicles slightly.

       Overall, this testing indicates that the effect of ethanol (together with lower aromatics and
T50 levels) may reduce PM emissions due to cold starting at 35 F under high altitude conditions.
                                           151

-------
However, PM emissions during warmed up driving are very low and an effect of fuel quality was
indiscernible.

       The State of Alaska, in conjunction with General Motors Corp. and EPA, measured PM
emissions from ten vehicles ranging in model year from 1977 to 1994 using two fuels.KKK  The
non-oxygenated fuel was a commercial wintertime fuel from the Fairbanks areas.  The ethanol
blend in this study was created from the non-oxygenated fuel via splash blending. Testing was
performed in Alaska using a portable dynamometer. Three of the vehicles were also tested at
EPA's laboratory in Research Triangle Park, N.C, ranging in model year from 1987 to 1994.
The testing in Alaska was performed at -20°F, 0°F, and 20°F.  The EPA testing was performed at
these same temperatures plus 75 F. Both sets of testing began with the non-oxygenated fuel,
followed by testing with the E10 fuel. This could introduce a bias into the results, but the degree
of this is unknown.

       The cold conditions led to difficulties in measuring PM emissions in Alaska. Therefore,
few acceptable measurements of PM were made and the results were not presented in the paper.
The fact that the EPA testing was conducted in a laboratory made vehicle conditioning and
operation and particulate collection more feasible. The PM emissions  from the three vehicles
tested by EPA on the two fuels are presented in the paper.

       Only one measurement of PM emissions was made for each combination of vehicle, fuel
and temperature. Thus, no direct measurement of test to test variability is available. We
calculated the percentage difference in PM emissions between the E10 and EO fuel for each of
the eleven combinations tested. PM emissions with the ethanol blend ranged from 81% lower to
84% higher than those with the EO fuel.  Thus, there appears to be considerable variability  in the
test results.  Taken together, the average of the percentage changes for each condition showed
the ethanol blend reducing PM emissions by 21%. However,  this decrease was not statistically
significant at the 90% confidence level.  The ethanol blend more consistently decreased PM
emissions at -20 F and OF, but not at 20 F or 75 F. The paper states that PM emissions at the
higher two temperatures were very low and the differences tended to be within measurement
accuracy. It is important to note, however, that the lower end of this range is 20 F. Only a small
percentage of driving in the U.S. occurs below this temperature.

       The third and final study was performed by EPA's laboratory in Research Triangle Park,
N.C.LLL  This study was conducted in three phases; the last two of which are relevant here. In
Phase II, PM emissions from two 1993-1995 model year vehicles were tested at -20 F, 0 F, and
20 F. In Phase III, PM emissions from an additional five 1987-2001 model year vehicles were
tested at -20 F, 0 F, 20 F and  40 F. Both phases utilized two fuels, one a wintertime non-
oxygenated fuel and the other a 10 vol% ethanol fuel created from the  non-oxygenated fuel via
splash blending.  Both phases measured PM emissions over the FTP and over a series of four
back-to-back EVI240 tests.50 It is not clear whether the fuels were always tested in the same
order or tested randomly. Some testing was performed with various malfunctions induced on the
50 The IM240 test is a warmed up test consisting of a portion of the FTP driving cycle. It was designed as a short
transient test cycle for use in vehicle inspection and maintenance programs.


                                          152

-------
vehicles, like disconnecting the oxygen sensor. We focus on the emissions from the properly
operating vehicles here.

       Of the 26 combinations of vehicles and temperatures tested, valid PM measurements over
the FTP were successfully obtained for both fuels in 21 of them. The average percent change in
PM emissions due to ethanol blending was +1%, in other words a very slightly increase.  In
contrast to the results of the two test programs discussed previously, the ethanol blend did not
show a benefit at -20 F, and showed only a very slight 1% reduction in PM emissions at 0 F.
The data show some tendency for the ethanol blend to produce a greater PM emission reduction
for the highest emitting vehicles. However, this trend is not as clear as in the Colorado study.
Thus, this study indicates no clear effect of ethanol on PM emissions.

       The IM240 testing showed much lower PM emission levels due to the warmed up nature
of the test.  There was also no clear trend in the effect of ethanol on PM emissions in this testing.

       The available data indicate that ethanol blending might reduce exhaust PM emissions
under very cold weather conditions (i.e., 0 F or less), particularly at high altitude. There is no
indication of PM emission reductions at higher temperatures and under warmed up conditions.
The data are certainly too limited to support a quantitative estimate of the effect of ethanol on
PM emissions.

       Fine particles can also be formed through a series of chemical reactions in the atmosphere
from gasses such as sulfate (SC>42"), nitrate  (N(V), ammonia (NHa), and volatile organic
compounds (VOC) emitted from motor vehicles.  This aerosol formed secondarily in the
atmosphere through these gas to particle conversions will be discussed in greater detail in
Chapter 5 of this document. Emerging science is indicating that gaseous aromatic compounds
are likely among the most important VOCs which  are precursors of carbonaceous PM which is
formed in the atmosphere.  Therefore, we discuss the effect of fuel quality on aromatic
hydrocarbon emissions in the next section.

3.1.1.5       Aromatic Emissions

       The Auto/Oil Air Quality Improvement Research Program tested over 100 vehicles from
model years 1983 - 1989 on a fuel matrix of over 80 fuel blends to determine the exhaust
emission effects of varying fuel parameters - including ethanol and aromatics.MMM Phase 1 of
this study tested two fleets of vehicles: twenty (20) 1989 model year vehicles, and fourteen (14)
1983-1985 model year vehicles. A matrix of 16 fuels (Matrix A) was developed in the first
portion of the study with half the fuels containing 20% aromatics by volume and half with 45%
aromatics by volume. This data was used to investigate the impact of changing aromatic levels
in the fuel on the aromatics emitted in the exhaust  as a function of total hydrocarbon emissions.
Linear regression of the test data indicates that there is a linear relationship between  the level of
aromatics by volume in the fuel and the mass of aromatics emitted in the exhaust.  Based on the
results of this regression, aromatics have a tendency to be emitted less than proportionally to
their percent volume in the fuel, as shown in the following equation:

                Aromatics Exhaust(wt%) = 0.64x Aromatics Fuel(vol%)


                                          153

-------
       The coefficient in the above equation was statistically significant at the 90% confidence
level (0.64 ± 0.02). However, when we considered the presence of an intercept, it was not
statistically significant (-0.35 ± 2.26). Therefore, we forced the regression line through zero and
repeated the regression.

       The Auto/Oil program also produced data which allows the effect of ethanol on aromatic
hydrocarbon emissions to be assessed.  As discussed in Section 2.2, ethanol blending tends to
reduce the aromatic content of gasoline. Of interest here is whether ethanol has any other effect
on aromatic hydrocarbon emissions beyond that associated with reducing the aromatic content of
gasoline.

       The Auto/Oil data contained a subset of fuels designed specifically for this analysis. A
total of 4 ethanol blends were produced by splash blending ethanol into four non-oxygenated
fuels.  Two of the non-oxygenated fuels came from the original group of sixteen tested during
Phase 1 of the research study. Base fuel A (industry average fuel) was a 9 RVP fuel with 32%
aromatics.  Base fuel F was also a 9 RVP gasoline with 20% aromatics. Two new non-
oxygenated fuels were created from base fuels A and F. In both cases, and butane was removed
to lower the RVP level by 1 psi resulting in fuels V and S, respectively. To each of these 4 non-
oxygenated fuels,  10% ethanol was splash blended resulting in the final 8 fuel test matrix. A
summary of the differences between the expected and actual aromatic content in the fuel and in
the exhaust as a function of THC is shown in Table 3.1-22.NNN

                                      Table 3.1-22.
              Expected vs. Predicted Non-Oxy and E10 Fuel Properties and
           Exhaust Aromatics Reductions for Auto/Oil AQIRP "Fuel Matrix B'

Non-
Oxy
Fuel
A
V
F
S


Fuel
Aromatics
32
33.5
20
21.2


E10
Fuel
X
w
u
T

Fuel
Aromatics
Expected3
28.8
30.15
18
19.08

Measured
Fuel
Aromatics
27.2
29.0
19.1
18.1
% Reduction in
Exhaust
Aromatics
Expected
2.05
2.14
1.28
1.36
% Reduction
in Exhaust
Aromatics
Actual
2.39
1.54
0.93
0.63
       aBased on dilution as a result of splash blending 10% ethanol
       The aromatic contents of the ethanol blends, as listed in Table 3.1-22, do not reflect the
10% reduction from the aromatic contents of their non-oxygenated base fuels which would be
expected from splash blending with 10 vol% ethanol. The discrepancies between the measured
and estimated fuel aromatic contents are small, on the order of 1-2 vol% aromatics.  However,
when the total  difference is on the order of 3 vol% ethanol, these discrepancies are significant.
The discrepancies are likely the result of measurement uncertainty of both base and ethanol
fuels.

       Using the relationship between fuel aromatic content and aromatic hydrocarbon
emissions developed above, we can predict the reduction in exhaust aromatic emissions
                                          154

-------
associated with the differences in the aromatic contents of the non-oxygenated fuels and their
ethanol containing counterparts. We believe that it is most accurate to use the expected aromatic
contents for the ethanol blends rather than the measured levels, since it is likely that the volume
of ethanol added was very close to 10 vol%.

       The expected reduction in the percentage of VOC emissions represented by aromatic
hydrocarbons based on the expected reduction in fuel aromatic content is shown in the second to
the last column in Table 3.1-22.  The measured reduction in the percentage of VOC emissions
represented by aromatic hydrocarbons is shown in the last column. In three out of four cases (all
but fuels A and X), the actual reduction in aromatic emissions is less than the predicted reduction
based on dilution. Had we used the measured aromatic contents for the ethanol blends, the
outcome would have been the same: In three out of four cases (in this case all but fuels F and U),
the actual reduction in aromatic emissions is less than the predicted reduction based on dilution.

       Qualitatively, this indicates that there does not appear to be any additional benefit in
reducing aromatic hydrocarbon emissions associated with the use of ethanol beyond that
expected from the reduction in the aromatic content of gasoline portion associated with ethanol
blending.

       Based on our analysis in Section 2.2, increased ethanol blending will  significantly reduce
gasoline aromatic content. This could cause a corresponding reduction in the aromatic fraction
of exhaust VOC emissions relative to non-oxygenated conventional gasoline. In addition,
ethanol also reduces total exhaust VOC emissions from older vehicles and may do so from newer
vehicles, based on the CRC E-67  study. This would further reduce emissions of aromatic
hydrocarbons. As will be discussed further in Chapter 5, this reduction in aromatic hydrocarbon
emissions could reduce ambient levels of secondary organic PM.

3.1.1.6        Emission Effects Associated with Specific Fuel Blends

3.1.1.6.1      Conventional Gasoline Analysis

       In Section 2.2 of Chapter 2,  we estimated the effect of blending ethanol and MTBE on
the quality of conventional gasoline (see Tables 2.2-3 and 2.2-4). Here, we present the effect of
these changes in fuel quality on emissions from motor vehicles in percentage terms, relative to
those of a typical non-oxygenated U.S. gasoline blend.  Because of the Tier 2 sulfur standards,
sulfur is held constant at 30 ppm.  Also, due to the MSAT standards, we assume that benzene
levels are not affected, as well. Table 3.1-23 presents the gasoline  qualities of a typical 9 RVP
CG,  as well as MTBE and ethanol blends which reflect the effect of adding these two oxygenates
to gasoline.
                                           155

-------
              Table 3.1-23. CG Fuel Quality With and Without Oxygenates"
Fuel Parameter
RVP (psi)
T50
T90
Aromatics (vol%)
Olefins (vol%)
Oxygen (wt%)
Sulfur (ppm)
Benzene (vol%)
Typical 9 RVP CG
8.7
218
332
32
7.7
0
30
1.0
MTBECG Blend
8.7
207
321
25.5
7.7
2
30
1.0
Ethanol CG Blend
9.7
205
329
27.4
7.5
3.5
30
1.0
       Assumes summer (July) conditions
       Table 3.1-24 presents the differences in emissions of the MTBE and ethanol blends
relative to that of non-oxygenated conventional gasoline.

         Table 3.1-24. Effect of Oxygenates on Conventional Gasoline Emissions"
Pollutant
Exhaust VOC
NOx
C0b
Exhaust Benzene
Formaldehyde
Acetaldehyde
1,3-Butadiene
Non-Exhaust VOC
Non-Exhaust
Benzene
Source
EPA Predictive
Models
MOBILE6.2
EPA Predictive
and Complex
Models
MOBILE6.2
MOBILE6.2 &
Complex Models
1 1 Vol% MTBE
-9.2%
2.6%
-6% 7-11%
-22.8%
+21.3%
+0.8%
-3.7%
Zero
-9.5%
10Vol% Ethanol
-7.4%
7.7%
-11% 7-19%
-24.9%
+6.7%
+156.8%
-13.2%
+30%
+ 15.8%
9.7 RVP
CG
+1.1%
+1.1%
+12.7%
-2.6%
-3.7%
-2.0%
-2.6%
+30%
+15.8%
 Assumes summer (July) conditions
 The first figure shown applies to normal emitters; the second applies to high emitters.
       The two oxygenated blends both reduce exhaust VOC and CO emissions, but increase
NOx emissions. The MTBE blend does not increase non-exhaust VOC emissions due to the fact
that non-oxygenated and MTBE blends have to meet the RVP standard. Ethanol blending
increases non-exhaust VOC emissions in two ways. First, ethanol blends are allowed 1.0 psi
higher RVP levels in most areas with CG.  Second, ethanol increases permeation emissions. The
most notable effect on toxic emissions in percentage terms is the increase in acetaldehyde with
the use of ethanol.  Acetaldehyde emissions more than double. However, as will be seen below,
base acetaldehyde emissions are low relative to the other toxics. Thus, the absolute increase in
emissions is relatively low.
                                          156

-------
3.1.1.6.2     Reformulated Gasoline (RFG) Analysis

       The previous section discussed the relative emission changes to expect when adding
ethanol to the conventional non-oxygenated gasoline pool. A second scenario to consider is the
case where RFG areas change from MTBE, a commonly used oxygenate in RFG areas, to either
ethanol RFG or a non-oxygenated RFG.

       Table 3.1-25 presents the gasoline qualities of three types of RFG: non-oxygenated, a
typical MTBE RFG as has been marketed in the Gulf Coast and a typical ethanol RFG which has
been marketed in the Midwest. The fuel specifications shown are based on specific RFGs
predicted by the refinery modeling discussed in Chapter 2 for the cases and PADDs shown.
These specific fuels were selected as they represented PADD-wide RFGs which contained
primarily one oxygenate at the desired volumetric concentration.

         Table 3.1-25. Summer RFG Fuel Quality With and Without Oxygenates"
Fuel Parameter
Case
OXYGEN (wt%)
SULFUR (ppm)
RVP (psi)
E200 (%)
E300 (%)
T50 (F)
T90 (F)
AROMATICS (vol%)
OLEFINS (vol%)
BENZENE (vol%)
Non-Oxygenated
RFG
PADD 3 Reference
0.0
30.0
7.0
52.0
87.5
184
335
20.1
14.6
0.7
MTBE RFG
PADD 1 Reference
2.1
30.0
7.0
59.9
88.9
190
342
21.0
4.3
0.7
Ethanol RFG
PADD 2 RFS
3.7
30.0
7.0
57.6
81.9
185
335
20.0
13.6
0.7
       Assumes summer (July) conditions
       Table 3.1-26 presents the emission impacts of these three types of RFG relative to the 9
RVP CG described in Table 3.1-23.
                                         157

-------
                                     Table 3.1-26.
                Effect of RFG on Per Mile Emissions from Tier 0 Vehicles
                      Relative to a Typical  Conventional Gasoline"
Pollutant Source
Non-Oxy
RFG
11 Volume
Percent MTBE
10 Volume
Percent Ethanol
Exhaust Emissions
voc
NOx
CO
Exhaust Benzene
Formaldehyde
Acetaldehyde
1,3-Butadiene
EPA Predictive
Models
MOBILE6.2
EPA Predictive
and Complex
Models
-13.4%
-2.4%
-22%
-21.2%
-5.9%
-0.2%
20.9%
-15.3%
-1.7%
-31%
-29.7%
19.4%
-9.5%
-29.2%
-9.7%
7.3%
-36%
-38.9%
2.3%
173.7%
6.1%
Non-Exhaust Emissions
VOC MOBILE6.2 &
CRC E-65
Benzene MOBILE6.2 &
Complex Models
-30%
-40%
-30%
-43%
-18%
-32%
       Assumes summer (July) conditions
       As can be seen, the oxygenated RFG blends are predicted to produce a greater reduction
in CO emissions, but increase NOx emissions. Exhaust VOC emission effects are mixed. Non-
exhaust VOC emissions with the exception of permeation are roughly the same due to the fact
that the RVP level of the three blends is the same.  However, the increased permeation emissions
associated with ethanol reduces the overall effectiveness of ethanol RFG. The most notable
effect on toxic emissions in percentage terms is the increase in acetaldehyde with the use of
ethanol. Acetaldehyde emissions more than double.  However, as will be seen below, base
acetaldehyde emissions are low relative to the other toxics.  Thus, the absolute increase in
emissions is relatively low. The ethanol RFG also produces a greater reduction in exhaust
benzene emissions and somewhat lower reduction in non-exhaust benzene emissions.

       The exhaust emission effects shown for VOC and NOx emissions only apply to Tier 0
vehicles in our primary analysis. In the sensitivity analysis, these effects are extended to Tier 1
and later vehicles. The effect of RVP on non-exhaust VOC emissions is temperature dependent.
The figures shown above represent the distribution of temperatures occurring across the U.S.
under summer conditions (average July fuel specifications).

3.1.2   High-Level Ethanol Blends

       The vast majority of ethanol blended into gasoline as a result of the RFS is expected to be
used in a 10 vol% ethanol blend (E10) rather than an 85 vol% ethanol blend (E85), as discussed
in Chapter 1.  At the same time, some ethanol is likely to be used as E85, and its use is growing.
Current estimates indicate that roughly 6 million FFVs  are on the road today, with US
automakers projecting an additional 3 to 5 million FFVs produced annually over the next several
years.  The analysis to follow relies upon the limited  amount of data available on both older and
                                          158

-------
current technology (Tier 2) FFVs, the later of which will dominate the FFV fleet in future years.
Based on this data of Tier 2 vehicles, we believe that with the increased use of E85 (which is
inherently a low-sulfur fuel) emissions should be neutral or better than operation on EO or E10
fuel blends for CO and NOx. NMOG emissions may be higher primarily due to emissions of
unburned ethanol at cold starts, while running NMOG emissions are lower with E85 based on
certification data. The sections to follow examine these issues in greater detail.

3.1.2.1        Exhaust emissions

3.1.2.1.1      Regulated Gaseous Pollutants

       Relatively little data is available for investigating the effects of high level ethanol blends
on exhaust emissions.  Part of the 1993 Auto/Oil Air Quality Improvement Research Program
(AQIRP) investigated the emissions associated with the use of E85 blends. Emissions over the
Federal Test Procedure (FTP) were measured from three Tier 0 and Tier 1 certified flexible-fuel
vehicles with  three test fuels. A source of emission data for Tier 2 FFVs is EPA's Certification
and Fuel Economy Information System (CFEIS) database, which contains certification data for
                                    ..                       ooo
five model year 2006 FFVs certified to Tier 2 standards (bins 5-8).    However, certification
data, composed of regulated emissions while operating on E85, represents very limited operating
conditions. It  does not include aggressive driving or cooler ambient temperature starts or
operation.

       The Auto/Oil Study found that E85 reduced FTP composite NOx emissions by 49%
compared to conventional gasoline with  1988 industry average fuel properties.  This is likely the
result of improved catalyst efficiency due to the low sulfur concentration in E85 (only 5 ppm vs.
339 ppm in the industry average fuel). The 2006 CFEIS data from Tier 2 FFVs, on the other
hand, shows only a 3% decrease in NOx emissions with E85 from a cold start test but a
significant 45% decrease in hot running NOx emissions. CO emissions are reduced at least 33%
on a cold start according to CFEIS, while the Auto/Oil study did not find statistically significant
changes in CO emissions. Emissions of Non-Methane Organic Gases (NMOG) increased 10%
for the Tier 2  CFEIS vehicles. The Auto/Oil data showed NMOG increased by 26%, but this
change was not statistically significant (p-value of 0.28). However, CFEIS data indicated a 50%
reduction in CO and HC emissions during hot operation. The measurement used to determine
NMOG in both CFEIS and Auto/Oil data include the mass of oxygen in all measured organic
species  except methane.

       While the emissions of NMOG appear to increase with E85 compared to EO for Tier 2
certified vehicles, the majority (-55%) of E85 NMOG emissions are direct emissions of ethanol,
which has a relatively low reactivity compared to other NMOG species.  Thus there may still be
a slight NMOG benefit based on ozone reactivity despite a potential net increase in total NMOG
emissions.  An important point worth noting is that the cold start emissions with E85 represent a
greater %  of bag weighted emissions than with EO.  This manifests itself primarily in the form of
unburned ethanol emissions during cold  start, before the combustion chamber has reached a high
enough temperature to promote complete ethanol vaporization. Thus NMOG emissions with
E85 at colder temperatures could be much greater (2 to 3 times higher) than with EO due to
prolonged periods at low temperature. Because of this unique start behavior and the lower
                                          159

-------
emissions observed during hot operation, emissions from E85 may be better quantified if
separated between starts and hot driving operation.
3.1.2.1.2
Air Toxics
       With increasing use of E85, some air toxics may increase while others decrease relative
to EO.  Emissions of Benzene and 1,3-Butadiene decrease while acetaldehyde, formaldehyde, and
emissions of ethanol increase. The net result is an increase in total air toxics, but this is largely
driven by increase ethanol and acetaldehyde emissions. Table 3.1-27, below, shows the percent
change in FTP composite g/mile emissions of several air toxics for the three FFVs tested on three
fuels as part of the 1993 Auto/Oil study.  The fuels tested were AQIRP gasoline with!988
Industry average qualities (CG), a 1996 California phase 2 reformulated gasoline (RFG blended
with MTBE), and an E85 blend with identical gasoline specs as RFG.PPP

                                      Table 3.1-27.
                Percent Difference in Toxic Emissions Between EO and E85



Formaldehyde
Benzene
1,3-Butadiene
Acetaldehyde
Total Toxics
% Difference Between Fuels
RFG vs CG

-2
-55
-31
-18
-42
E85 vs CG

93
-87
-85
2620
108
E85 vs RFG

97
-72
-79
3220
255
       The increase in acetaldehyde emissions is substantial, on the order of 20 to 30 times that
of EO.  This is substantially higher than the 15 to 20 fold increase shown with Tier 2 FFVs in the
CFEIS data. Emissions of benzene, 1,3-butadiene, ethylbenzene, hexane, styrene, toluene, m-
xylene, p-xylene, o-xylene, and naphthalene are all expected to decrease significantly (50-80%)
with the use of E85 vs. EO according to CFEIS, which is consistent with the Auto/Oil results
presented in Table 3.1-27. Regardless of vehicle technology, the increased emissions of
acetaldehyde  could be a potential concern due to its strong odor, as well as its respiratory system
irritating and  potentially carcinogenic properties.
3.1.2.1.3
Particulate Matter
       Even less data exists to draw firm conclusions on direct particulate matter emissions due
to increased E85 use. Theoretically, E85 use has the potential to increase direct emissions of PM
under modes of rich engine operation. This is especially important at cold start, before the
catalyst has reached its operating temperature and when an E85 fueled vehicle runs substantially
richer than if it were fueled with EO. In this situation, the low temperatures in the combustion
chamber, compounded by the evaporative cooling effect of ethanol, makes fuel vaporization
difficult and may increases exhaust emissions of raw fuel  and PM at cold start. Sustained
periods of high load may also have increased emissions of PM with E85 than with EO due to
richer operation with E85.  Results from a 2003 SAE paper showed a negligible increase in
                                           160

-------
direct PM emissions from E85 vs. EO fueled vehicles over the European Test Cycle (Directive
70/220/EEC and its amendments).QQQ Tests conducted at 23°C and 16°C showed an increase in
PM emissions with decreasing temperature for both EO and E85, with slightly higher PM
emissions at cold temperature with E85. This study only used one E85 blend and one model year
2002 FFV, however, so these results cannot be considered entirely representative of the on-road
FFV vehicle fleet.  Again, since the projected use of ethanol as E85 is very small compared to
its use as E10, the emissions impacts associated with E85 will be also be quite small on an
absolute scale.

       As discussed above, emerging science is beginning to identify gaseous aromatics as an
important precursor to secondary organic aerosol. Exhaust aromatic emissions should be
reduced with E85 since the fuel aromatics content of E85 is much lower than that for EO or E10
blends. This reduction in exhaust aromatics should reduce the formation of secondary organic
aerosol. However, as mentioned earlier, no specifications currently exist for the 15% gasoline
portion of E85.  Thus, the degree to which the aromatic content of E85 will be lower than
gasoline is not known with any confidence. Lack of data regarding the speciation of VOC
emissions also prevents any quantitative estimate of any benefit in this area.

3.1.2.2       Non-Exhaust Emissions

       We currently have very little data regarding non-exhaust emissions from E85 vehicles.
Theoretically, evaporative emissions of E85 fueled vehicles have the potential to be lower than
with EO or E10.  This is because ethanol blended with  a given gasoline at the 85% level is likely
to be less volatile than EO or E10 (with the same gasoline fuel quality). This is not entirely
certain, however, since there is no fuel specification for the hydrocarbon composition of the 15%
of E85 that is gasoline. Thus, the RVP of the final E85 blend could be closer to that of EO or
E10 fuels than commonly thought to be the case. Moreover, since the volatility of ethanol blends
peaks between 6 and 30 vol% ethanol, the fuel in the tank of drivers of flex-fuel vehicles who
alternate between E85 and gasoline will experience a wide range of ethanol concentrations in the
fuel at any given time, and therefore a wide variation in the corresponding evaporative
emissions.

       Similarly, we have very little data with which quantitative predictions of the impact of
E85 use on non-exhaust emissions of air toxics (e.g., benzene) can be drawn. The Auto/Oil
study  mentioned in the previous section tested  the same three Tier 0 and Tier 1 vehicles for hot-
soak evaporative emissions. They  found no statistically significant change in NMOG or
OMHCE51 evaporative emissions, yet found a  statistically significant 60% reduction in benzene
emissions. Directionally, you would expect both hydrocarbon and air toxic evaporative
emissions to decrease due to the dilution of the hydrocarbon portion of the fuel with ethanol.
However, again, it is highly dependent on the volatility of the gasoline component of the specific
E85 used and its benzene content, neither of which is regulated.
  Organic Material Hydrocarbon Equivalent


                                          161

-------
3.2    Effect of Fuel Quality on Spark-Ignited Nonroad Equipment Emissions
       We use EPA's NONROAD emission model to estimate the effect of gasoline quality on
emissions from nonroad equipment.  We use the 2005 version of this model, NONROAD2005,
which includes the effect of ethanol on permeation emissions from several types of equipment:
all small spark-ignition equipment (including handheld and non-handheld equipment less than or
equal to 25 hp), all spark-ignition recreational marine watercraft (includes all outboard, stern-
drive inboard, and personal watercraft). Note that these categories do not include recreational
vehicles (motorcycles, ATVs, and snowmobiles) or large spark-ignition equipment.

       Only a limited number of fuel parameters affect emissions in NONROAD. Exhaust
VOC, CO and NOx emissions are a function of sulfur and oxygen. Here, only the latter fuel
parameter is of interest.  Emissions of all three pollutants are assumed to change proportionally
with fuel  oxygen content.  Table 3.2-1 shows the effect of moving to a 10 volume percent
ethanol blend (3.5 wt% oxygen) on these emissions, either from a non-oxygenated fuel or from
an 11 volume percent MTBE blend (2.0 wt% oxygen).RRR

                                     Table 3.2-1.
       Effect of a 10 Volume Percent Ethanol Fuel on Nonroad Exhaust Emissions

Base Fuel
VOC
CO
NOx
4-Stroke Engines
Non-
Oxygenated
-16%
-22%
+40%
11 Volume
Percent MTBE
-7%
-9%
+17%
2-Stroke Engines
Non-
Oxygenated
-2%
-23%
+65%
11 Volume
Percent MTBE
-1%
-10%
+28%
       As can be seen, the higher oxygen content of ethanol blends reduces exhaust VOC and
CO emissions. However, it also increases NOx emissions quite substantially, especially from 4-
stroke engines. However, it should be noted that NOx emissions from these engines tend to be
fairly low to start with, given the fact that these engines run richer than stoichiometric. Thus, a
large percentage increase of a relative low base value  can be a relatively small increase in
absolute terms. This will be seen below in Chapter 4, when we evaluate the impact of increased
ethanol use on the local and national emission inventories.

       Non-exhaust VOC emissions (other than permeation) are a function of gasoline RVP and
ethanol content in NONROAD2005. Ethanol content only affects permeation emissions. Both
of these emissions are temperature dependent, so the effect of ethanol and RVP is also
temperature dependent. Based on the results of modeling national emissions in July, a 10
volume percent ethanol blend increases non-exhaust VOC emissions by 15 percent. This
assumes a 1.0 psi increase in RVP.

       Hose permeation emissions in the public version of NONROAD2005 are independent of
fuel quality. In support of the development of new emission standards for small nonroad
engines, EPA has been testing small nonroad engines  for hose permeation emissions using fuels
                                         162

-------
with and without ethanol. Based on this testing, as well as discussions with nonroad equipment
manufacturers, we developed new hose permeation emission rates for NONROAD2005 for both
gasoline and E10 fuels. Roughly, these revised permeation rates indicate that emissions with
E10 are 2-3 times higher than those for gasoline.  This increase is similar to that found for the
three oldest onroad vehicles in the CRC E-65 study, discussed in section 3.1.1.1.4 above.52 The
NONROAD2005  hose permeation factors888 were adjusted as follows:

       1)  Permeation emissions in the public version of NONROAD2005 were 450 grams per
       meter-squared per day (g/m2/day) for both small spark-ignition engines and for the supply
       hoses on portable fuel tanks in recreational marine watercraft, applicable to all fuels. For
       both types of equipment, the permeation emission rates were changed to 122 g/m2/day for
       gasoline and 222 g/m2/day for El0.

       2)  Permeation emissions in the public version of NONROAD2005 were 100 g/m2/day
       for supply hoses on outboard recreational marine watercraft (> 25 hp), 300 g/m2/day for
       supply hoses on personal watercraft (PWC), and 110 g/m2/day for fill neck hoses on both
       outboards  and PWC. These three permeation emission rates were changed to 42
       g/m2/day for gasoline and 125 g/m2/day for E10.

       3)  Permeation emissions from sterndrive/inboard recreational marine watercraft in the
       public version of NONROAD2005 were 100 g/m2/day for supply hoses and 110 g/m2/day
       for fill neck hoses.  Both of these permeation emission rates were changed to 22 g/m2/day
       for gasoline and 40 g/m2/day for El0.

       4)  Permeation emissions in the public version of NONROAD2005 were 0 g/m2/day for
       vent hoses on all recreation marine watercraft. This permeation emission rate was
                        rt                          rt
       changed to 2.5 g/m /day for gasoline and 4.9 g/m /day for El0.

       5)  One final adjustment was to double the vent hose length for all gasoline-fueled
       outboards, personal watercraft, and sterndrive/inboard watercraft.

       The NONROAD emissions model does not estimate emissions of toxic air pollutants
from nonroad equipment. However, the National Mobile Inventory Model (NMIM) does make
such estimates.  NMIM utilizes the MOBILE and NONROAD models to develop national
emission estimates for motor vehicles and nonroad equipment. For the most part, NMIM
provides the relevant inputs to MOBILE6.2 and NONROAD and processes the results.
However,  with respect to nonroad toxic emissions, NMEVI takes exhaust and non-exhaust VOC
emission estimates from NONROAD and applies a set of toxic fractions of VOC emissions
based on fuel quality.TTT NMIM contains estimates of the toxic fractions of VOC emissions for
three fuels: a non-oxygenated gasoline, an MTBE blend and an ethanol blend. NMIM applies
the fraction of VOC emissions represented by each of the air toxics to either the exhaust or non-
52 Permeation emissions from nonroad equipment are not regulated. Thus, the elastomers used in the fuel systems of
nonroad equipment are likely to be more similar to those of older onroad vehicles than those of later onroad vehicles
which were subject to stringent non-exhaust VOC emission standards.


                                           163

-------
exhaust VOC emissions estimated by NONROAD. The toxic fractions of VOC were derived
from motor vehicle testing. Thus, they are considered approximate.

       We hope to update our projections of the effect of gasoline quality on exhaust and non-
exhaust emissions from nonroad equipment based on additional testing to be conducted over the
next several years if funding allows. These updated projections could be used in the
comprehensive assessment of the impact of the fuel-related provisions of the Energy Act which
is due in 2009.
3.3    Effect of Fuel Quality on Compression-Ignited Vehicle and Equipment
Emissions - Biodiesel

       Biodiesel is expected to be one of two renewable fuels to be used in significant volumes
through 2020.  While ethanol will dominate the market, biodiesel use is likely to grow
considerably reaching 300 million gallons by 2012, according to EIA estimates.  It is produced
domestically from vegetable oils, animal fats and recycled cooking oils, with the majority of this
product coming from soybean oil. It is typically used in 2%, 5% and 20% blends with diesel fuel
which have been assigned B2, B5 and B20 designations, respectively.

       In 2002, EPA issued a report entitled "A Comprehensive Analysis of Biodiesel Impacts
on Exhaust Emissions" based on existing data from various test programs. This report included a
technical analysis of biodiesel effects on regulated and unregulated pollutants from diesel
powered vehicles and concluded that biodiesel fuels improved PM, HC and CO emissions of
diesel engines while slightly increasing their NOx emissions.

       While the conclusions reached in the 2002 EPA report relative to biodiesel effects on
VOC, CO  and PM emissions  have been generally accepted, the magnitude of the B20 effect on
NOx remains controversial due to conflicting results from different studies.  Significant new
testing is being planned with broad stakeholder participation and support in order to better
estimate the impact of biodiesel on NOx and other exhaust emissions from the in-use fleet of
diesel engines. We hope to incorporate the data from such additional testing into the analyses for
other studies required by the Energy Act in 2008 and 2009, and into a  subsequent rule to set the
RFS program standard for 2013 and later.
3.4    Emissions from Fuel Production Facilities

3.4.1   Ethanol

       The primary impact of renewable fuel production and distribution regards ethanol, since
it is expected to be the predominant renewable fuel used in the foreseeable future. For the
NPRM, we estimated the impact of increased ethanol production, including corn farming, on
emissions based on DOE's GREET model, version 1.6. This estimate also included emissions
related to distributing the ethanol and take credit for reduced emissions related to distributing
displaced gasoline. Since the time  of the NPRM analysis, DOE has published the next version of
                                          164

-------
GREET, version 1.7. The emission estimates related to ethanol production and distribution in
GREET1.7 differ significantly from those in GREET1.6. In addition, through EPA's regional
offices, we contacted a number of States to obtain the latest emission estimates for ethanol plants
currently in production. These plant-specific estimates provide a useful comparison to the
inherently generic emission factors used by a nationwide-average model, such as GREET.

       In Section 3.4.1.1, we describe and compare the emission estimates from the GREET
model, both versions 1.6 and 1.7. In Section 3.4.1.2, we describe the data obtained from the
States and consolidate it into two sets of emission factors; one for wet mills and one for  dry
mills. Finally, in Section 3.4.1.3, we describe how we will use both the GREET and State
estimates in estimating national emissions from new ethanol plants in Chapter 4.

3.4.1.1        GREET Emission Estimates

       The emissions related to producing and distributing ethanol for use in gasoline blends
from both GREET1.6 and 1.7 are summarized in Table 3.4-1.  GREET presents emission factors
in a variety of units, such as per bushel of corn harvested, gallons of ethanol produced, etc.  All
the emission factors shown in this table have been converted to a per gallon of ethanol produced
or distributed basis using the default conversion factors contained in GREET.  One of GREET's
default assumptions is that 80% of ethanol plants are associated with dry corn milling, while the
other 20% are associated with wet milling. Nearly all future ethanol plants are planned to be dry
mill facilities.  Therefore, we only show the emission factors for dry mill ethanol plants  below.
                                          165

-------
                                      Table 3.4-1.
       Well-to-Pump Emissions for Producing and Distributing Ethanol from Corn
                       Dry Mill Facility (grams per gallon ethanol)
Pollutant
Corn Farming
and
Transportation
Ethanol
Production
Co-
Product
Credits
Ethanol
Trans-
portation
Gasoline
Transportation
Credit
Total
Emissions
Dry Mills -GREET 1.6
VOC
CO
NOx
PM10
SOx
0.8
4.3
11.3
8.1
1.2
6.8
2.9
4.6
0.4
6.1
-4.1
-3.3
-6.9
-2.5
-0.9
0.5
0.2
1.5
0.0
0.2
-0.9
-0.1
-0.4
0.0
-0.1
3.1
4.1
10.1
6.1
6.5
Dry Mills -GREET 1.7
VOC
CO
NOx
PM10
SOx
1.6
4.0
10.7
1.1
5.0
2.6
1.7
3.9
4.2
3.7
-2.5
-1.7
-4.1
-0.5
-2.4
1.6
0.2
1.3
0.0
0.2
-1.5
-0.1
-0.4
0.0
-0.1
1.8
4.1
11.4
4.9
6.4
Wet Mills -GREET 1.6
VOC
CO
NOx
PM10
SOx
0.8
4.3
11.3
8.1
1.2
6.8
3.3
6.2
0.5
9.2
-1.8
-2.1
-4.8
-2.3
-0.6
0.5
0.2
1.5
0.0
0.2
-0.9
-0.1
-0.4
0.0
-0.1
5.5
5.7
13.8
6.4
9.9
Wet Mills -GREET 1.7
VOC
CO
NOx
PM10
SOx
1.6
4.0
10.7
1.1
5.0
2.6
2.3
4.9
5.9
4.1
-3.2
-2.1
-4.6
-0.6
-2.7
1.6
0.2
1.3
0.0
0.2
-1.5
-0.1
-0.4
0.0
-0.1
1.1
4.3
11.9
6.5
6.5
       As can be seen, the emission estimates from the two versions of GREET differ
significantly. In particular, VOC emissions in GREET 1.7 are about 60% as large as those
estimated in GREET 1.6 for dry milss and even lower for wet mills.  The other differences are in
the +20% range, and differ in direction depending on pollutant.

       The default mix of dry and wet mill ethanol plants is 80/20 in both versions of GREET.
This is a reasonable estimate for current production.  However, the vast majority of new plants
are expected to be of the dry mill variety. Therefore, we will use the above emission factors for
a 80/20 mix of dry  and wet mill plants to estimate the emissions from current ethanol  plants and
the dry mill emission factors to estimate the emissions from future plants.

       While emissions related to ethanol production and distribution will increase,,areas with
refineries might experience reduced emissions, not necessarily relative to current emission
                                          166

-------
levels, but relative to those which would have occurred in the future had ethanol use not risen.
However, to the degree that increased ethanol use reduces imports of gasoline, as opposed to the
domestic production of gasoline, these reduced refinery emissions will occur overseas and not in
the U.S.  Therefore, we will not take any credit for reduced refinery emissions here.

       Similarly, areas with MTBE production facilities might experience reduced emissions
from these plants as they cease producing MTBE. However, some of these plants are likely to be
converted to produce other gasoline blendstocks, such as iso-octane or alkylate. In this  case,
their emissions are not likely to change substantially.

       The emission factors shown in Table 3.4-1 do include a credit for reduced emissions
related to a reduction in the volume of non-oxygenated gasoline being distributed. These are
taken directly from the GREET emission estimates for conventional gasoline production and
distribution. We assumed that ethanol use will reduce gasoline use one for one on an energy
basis.
3.4.1.2
Ethanol Production Emissions Received from States
       The emissions from most of the steps involved in ethanol production and distributions are
very diffuse (e.g., tractors plowing corn fields). However, the emissions from ethanol production
plants are point sources whose emissions are often measured and tracked by local or state
governments. We contacted over a dozen States in an attempt to improve our estimate of
emissions from ethanol production. The results of this process are summarized below.uuu

       We received emission  estimates from 13 States for current ethanol plants with a
combined capacity of 3 billion gallons per year. The emission data involved annual emission
estimates for one or more years between 2001-2005.  Overall, these plants represent roughly
three-fourths of current ethanol capacity.  The emission data cover a very wide range of plant
capacities, ranging from 0.4 to 274 million gallons of ethanol per year.

       The capacity-weighted average emissions of wet and dry mill ethanol plants are
summarized in Table 3.4-2.

          Table 3.4-2. Emissions From Ethanol Plants: State Data (g/gal ethanol)

voc
CO
NOx
PM10
SOx
Wet Mills
17.5
15.0
18.3
8.8
24.6
Dry Mills
4.0
1.9
5.5
2.2
7.0
       As can be seen, the emissions from wet mills are much higher than those from dry mills.
In general, the wet mill plants are older than the dry mills.  They also involved a different set of
processes and produce a different set of by-products.
                                           167

-------
below.
       The ethanol plant emissions from GREET1.6 and GREET1.7 are shown in Table 3.4-3
           Table 3.4-3. Emissions From Ethanol Plants: GREET (g/gal ethanol)


voc
CO
NOx
PM10
Sox
Wet Mills
GREET 1.6
6.8
3.3
6.2
0.5
9.2
GREET 1.7
2.6
2.3
4.9
5.9
4.1
Dry Mills
GREET 1.6
6.8
2.9
4.6
0.4
6.1
GREET 1.7
2.6
1.7
3.9
4.2
3.7
       As can be seen from comparing the emission estimates in Tables 3.4-2 and 3.4-3, the
State data indicate that the emission from wet mills are much higher than those estimated in
either version of GREET. In contrast, the emission data obtained from the States for dry mills is
generally consistent with the estimates in GREET1.6 and higher than those in GREET1.7. An
exception to the latter are emissions of PM10, which the state data indicate are lower than the
estimate in GREET 1.7 and higher than that in GREET 1.6.

       The reasons for the differences in the State data and the estimates in GREET 1.6 and
GREET1.7 are not known.  It is possible, particularly for wet mills, that the State estimates
include the emissions from an entire geographically-defined facility which may include more
operations than just corn milling.  It is also possible that the estimates in GREET represent
emissions from plants which would be designed today or in the future and are less representative
of plants which were built over 20 years  ago. This deserves further investigation.  At this time,
we will use the average of the State emission data as a second estimate of ethanol plant
emissions, along with GREET1.7, in order to better indicate the range of possible emissions from
these plants.
3.4.1.3
Selection of Ethanol Production and Distribution Emission Estimates
       We have available three estimates of the emissions from ethanol plants and two estimates
of the emissions from the other steps in the process of growing corn through ethanol distribution.
The estimates contained in GREET1.7 represent an update of those in GREET1.6. Therefore, we
will use the emission factors from GREET1.7 in lieu of those in GREET 1.6 in Chapter 4 where
we estimate national emissions from ethanol production and distribution.

       In addition, we develop a second estimate of these emissions by substituting the average
emission factors based on the State data for the dry mill ethanol plant emissions contained in
GREET1.7.  While the State data represents emissions from current plants and our primary focus
is future plants, it is not certain that the emissions of the two sets of ethanol plants will differ.
Therefore, the use of the  State data will provide a useful indication of the potential uncertainty in
the GREET 1.7 estimates. We chose not to use the state data for wet mills, as these emission
factors are often a factor  of 10 higher than those from GREET 1.6 or GREET 1.7. More
understanding of the processes producing these emissions is needed before they can be all
                                          168

-------
assigned to ethanol production.  Thus, we use the GREET 1.7 emission factors for wet mill
ethanol production, while using the state data for dry mill ethanol production.

       The two sets of estimates are shown in Table 3.4-4. In both sets of estimates, the
emission factors for current ethanol plants assume an 80/20 mix of dry and wet mills, while those
for future ethanol plants assume 100% dry mills.

                                      Table 3.4-4.
            Selected Emission Factors for Ethanol Production and Distribution
                                      (g/gal ethanol)


voc
CO
NOx
PM10
Sox
GREET 1.7
Current Plants
1.8
4.0
11.4
4.9
6.4
Future Plants
1.8
4.1
11.4
4.9
6.4
GREET 1.7 + State Data
Current Plants
3.6
4.4
10.8
6.1
7.2
Future Plants
3.2
4.3
13.0
2.8
9.7
3.4.2   Biodiesel

       Like ethanol, we base our emission factors for biodiesel production distribution on the
estimates contained in the GREET model, version 1.7.  Table 3.4-6 shows the emission factors
associated with soybean farming, soy oil production and esterification, and biodiesel distribution.
We also include emissions related to distributing the biodiesel and take credit for reduced
emissions related to distributing displaced diesel fuel.

                                       Table 3.4-5.
    Well-to-Pump Emissions for Producing and Distributing Biodiesel from Soybeans
                               (grams per gallon biodiesel)



Pollutant
VOC
CO
NOx
PM10
SOx
Total:
Soybean
Farming and
Transportation
2.7
10.6
19.3
2.6
17.6


Biodiesel
Production
34.8
2.1
5.5
2.2
4.1


Biodiesel
Transportation
0.2
0.2
0.9
0.0
0.1

Diesel Fuel
Transportation
Credit
-0.2
-0.1
-0.6
0.0
-0.1


Total
Emissions
37.6
12.7
25.1
4.8
21.8
       At the same time, areas with refineries might experience reduced emissions, not
necessarily relative to current emission levels, but relative to those which would have occurred in
the future had biodiesel use not risen. However, to the degree that increased biodiesel use
reduces imports of diesel fuel, as opposed to the domestic production of diesel fuel, these
reduced refinery emissions will occur overseas and not in the U.S.
                                           169

-------
                    Chapter 3: Appendix
Fuel Property Tables and Summary of Predicted Emissions Changes
                            170

-------
Table 3A-1. CRC E-67 Study Test Fuel Properties
Inspection
API Gravity
Relalive Density
DVPE
Oxygenales-04815
MTBE
ETBE
BOH
02
Sulfur Conlenl
D 86 Disiillalion
IBP
5% Evaporated
10% Evaporated
20% Evaporated
30% Evaporated
40% Evaporated
50% Evaporated
60% Evaporated
70% Evaporated
80% Evaporated
90% Evaporated
95% Evaporated
EP
Recovery
Residue
Loss
Dnveability Index
E200
E300
Units
"API
60/60 °F
psi

vol %
vol %
vol %
wt %
ppm

T
•F
°F
°F
8F
°F
'F
°F
5F
°F
op
°F
°F
vol %
vol %
vol %

vol %
vol %
Fuel A
62.1
0.7310
7.74

0.03
0.02
0.02
0.02
18.8

94.2
126.3
136.0
148.6
163.6
179.8
194.7
209.0
224.2
243.4
294.3
327.4
351.2
97.0
1.8
1.2
1082,4
53.6
90.9
FuelB
59.9
0.7393
7.84

0.03
0.02
5.62
2.10
16.7

107.6
127.2
133.2
140.8
154.1
176.1
190.9
203.2
219.3
240.9
289,8
325.9
352.0
97.9
1.1
1.0
1075.8
57.6
91.5
FueIC
57.6
0.7482
7.70

0.13
0.01
10.37
3,84
19.0

104.3
124.6
130.5
138.8
146.6
153.7
192,7
223.5
245.7
281.5
329.2
343.4
374.0
97.7
1.2
1.1
1128,0
52.1
84
FuelD
61.4
0.7337
7.65

0.03
0.01
0.00
0,01
18.2

88.8
123.2
133.3
147.6
164.1
182.3
199,5
216.9
237.9
274.3
355.0
367.3
392.0
97.9
0.8
1.3
1153,3
50.6
83.6
FuelE
56.7
0.7519
7.80

0.11
0.01
10.26
3,78
17,2

106.3
124.3
130.5
139.5
147.2
153.8
197,7
226.2
259.2
299.7
351,7
364.9
385.4
97.4
1.4
1.2
1165.1
50,6
80.0
FuelF
60.1
0.7387
7.62

0.08
0.08
0.00
0,03
18.1

94.2
121.6
135.0
154.7
177.0
200.2
216.8
227.6
238,2
254.7
295.0
324.0
361.2
97.2
1.7
1.1
1148,0
40.0
90.9
FueIG
57.1
0.7502
7.78

0.13
0.04
10.15
3.76
17,5

103.7
125.3
133.2
143.7
152.9
163.4
212.2
226.7
237,0
251.7
290.7
327.8
365.4
96.7
1.5
1.8
1151,2
47,4
79.5
FuelH
60.6
0.7366
7.85

0.09
0.01
0.05
0.04
18,6

94.2
122.7
134.0
151.6
173.3
197.0
216,3
230.4
245.9
273.7
326,9
343.7
374.4
98.0
1.0
1.0
1176,8
41.7
85.2
Fuel I
57.2
0.7498
7.68

0.16
0.01
5.94
2,22
16.8

100.7
124.0
130.2
139.0
150.8
191.0
215,9
235.9
260.9
311.3
3542
366.6
391.8
97.9
1.1
1.0
1211.5
431
77.8
FueiJ
56.6
0.7525
757

013
001
5.90
219
191

102.6
1260
1344
1466
1755
220.5
2366
251.5
2719
305.2
3292
3387
3658
97.6
1.2
1.3
12549
35,2
78.4
FuelK
59.3
07416
771

016
002
0.00
003
219

93.9
1179
1297
1484
1744
208.5
2361
255.2
2796
319.1
3555
3686
3903
981
1.0
09
12582
376
75. 2
FuelL
54.4
0.761 1
7.69

016
001
1049
383
206

1061
1294
1400
1524
1588
202.1
2327
248.7
2735
3077
3491
3674
389.6
973
1.0
1 7
12823
394
780
                     171

-------
Table 3A-2. Summary of EPA E-67 vs. EPA Predictive Model Effects of E10 and MTBE Use Relative to CG and RFG
Base Fuel: EO: AAM CG EO: AAM CG E10: AAM E10:AAM E10:AAM 11% MTBE: Phase 2 RFG Class C RFG E10:
AAM Summer RVP -1 psi RVP-2psi Summer Avg Summer avg, Summer avg, Fuel props RFG: Non- MTBE: 1993 1993 region 2
avg. non-oxy delta delta T50 limited to T50 T90 02 are deltas oxy, from region 2 data, data (L),T50
L1 95°F only for EPA from AAM CG 1 993 region 2 low RVP & T90 delta
model class C Data, from AAM
(T50=195°F) low RVP data, low RVP
Fuel Parameters







RVP (psi)
T50 (°F)
T90 (°F)
Aromatics (vol %)
Olefins (vol %)
Oxygen (wt%)
Sulfur (ppm)

8.7
218
332
32
7.7
0
30
7.8
218
332
32
7.7
0
30
6.8
218
332
32
7.7
0
30
9.7
186
325
27
6.1
3.5
30
9.7
195
325
27
6.1
3.5
30
8.7
195
325
32
7.7
3.5
30
8.7
206
324
25.5
7.7
2.1
30
6.7
214
325
25.48
13.1
0
30
6.7
212
321
25.48
13.1
2.1
30
6.7
194
322
25.48
13.1
3.5
30

Predicted Emissions Changes
EPA Predictive Models (% change)


NOx
NMHC
EPA E-67 Model (% change)


NOx
NMHC


0.0
0.0

0.0
0.0

-1.0
-1.0

0.0
0.0

-2.1
-2.0

0.0
0.0

7.7
-7.4

9.2
-5.7

7.3
-7.0

6.7
-6.5

7.7
-7.5

6.7
-6.5

2.6
-9.2

-1.9
-4.7

-1.7
-7.7

-0.7
-3.2

2.4
-11.1

-2.2
-1.9

6.3
-12.9

8.8
-7.2
172

-------
This page intentionally left blank.
              173

-------
Chapter 4:  National Emission Inventory Impacts
       This chapter describes the methods used to develop national emissions inventories under
the Renewable Fuel Standard (RFS) program. These inventories account for impacts from
ethanol use, the removal of MTBE, and the resulting changes to gasoline.  These inventories also
account for the impacts of ethanol and biodiesel production and distribution.  This chapter also
presents and discusses these inventories.
4.1    Impact of Ethanol Use

       This section describes the methods used to develop national emissions inventories with
respect to ethanol consumption. This section also presents and discusses these inventories.
These inventories reflect only emissions from vehicles and equipment operating on ethanol-
blend gasoline, from both onroad and off-road sources.  The off-road sources do not include
nonroad diesel, locomotive, or marine applications.

4.1.1   Overview of Cases

       As described in Section 2.1, we consider three cases for the future use of ethanol-blend
gasoline: a Reference Case, an RFS Case, and an EIA Case. The main difference between the
cases is our assumption about how much ethanol will be used and where it will go. The
Reference case represents our estimate of fuel quality by county which existed in 2004 when
approximately 3.5 billion gallons of ethanol were consumed nationwide. In terms of 2012 fuel
consumption, about 4.0 billion gallons of ethanol is consumed nationwide in the Reference case.
The RFS case assumes 6.7 billion gallons of ethanol consumption in 2012, in accordance with
the requirements of the RFS mandate. The EIA case assumes 9.6 billion gallons of ethanol is
used nationwide in 2012, based on projections made in the Energy Information Agency's 2006
Annual Energy Outlook. We evaluate each case by predicting fuel quality in each county of the
U.S. in 2012. This 2012 fuel matrix is then used for all inventory and air quality assessments.

       While Chapter 2 discusses our methods for determining how much ethanol will go to
each state in  each case and how fuel properties will be affected, this section of the RIA uses
those distributions to derive estimates of the impact on national emissions inventories.

4.1.2   National Emissions Inventory Estimation Procedure

       Having approximated the effects of adding ethanol and removing MTBE on fuel
properties (see Chapter 2), the next step was to use the EPA's National Mobile Inventory Model
(NMIM)VVV  to calculate emissions inventories for gasoline fueled motor vehicles and nonroad
equipment in years 2012, 2015, and 2020. For all three years, we ran NMIM for January and
July, assuming that each was representative of winter and summer conditions, respectively.  We
estimate annual emission inventories by summing the two monthly inventories and multiplying
by six.  This  was done in order to reduce the amount of time needed to actually run the model.
                                          174

-------
       One additional simplification was made to shorten the time required to run NMIM for the
three years, three fuel cases, two months and roughly 3100 counties in the U.S.  Counties within
a state with identical fuels and inspection-maintenance programs and similar temperatures were
grouped together and run through NMIM as a single geographical area.  The temperatures used
for this area were those of the county with the highest VMT in the group.  As the specific
counties within a state with identical fuels sometimes changed across the three fuel cases, the
groupings of counties sometimes changed across the NMIM runs of the three fuel cases. This
occasionally introduced a change in the temperatures estimated for a county between fuel cases.
This in turn produced a change in emissions independent of changes in fuel quality.

       We evaluated the potential for this simplification to bias the projected emission impacts
of the various fuel cases. Counties where RFG is sold were always modeled consistently across
all three fuel cases and so are unaffected by this simplification.  Counties with low RVP and 9
RVP fuel were sometimes affected. On average, the changes in emissions occurring due to a
change in temperature appear to be unbiased (i.e., emissions increase as often as they decrease).
Also,  many of the emission impacts of changing fuel quality (e.g., exhaust VOC and NOx
impact) were applied outside of the NMIM model and so are unaffected by this simplification.
Since we do not present or use the emission impacts for individual counties, we believe that this
simplification does not significantly impact the emission impacts presented below.

       We chose 2012 as the first projection year, because it is the year of full RFS program
implementation. We also chose 2015 and 2020 to illustrate how the emissions will change over
time as the fleet changes. We increased ethanol consumption beyond 2012 only by volumes
required to maintain the same proportion to gasoline that existed in 2012, and not by growth
predicted in EIA estimates. By restricting ethanol growth in this way, the same fuel quality that
existed in 2012 would apply to 2015 and 2020, which would better highlight the effects of fleet
turnover.

       NMEVI's estimates of both onroad and nonroad emissions were "post-processed" to
reflect factors not yet included in the model.  For onroad emissions, the effect of fuel quality on
exhaust VOC and NOx emissions contained in the model (i.e., those in MOBILE6.2) were
replaced with those from the EPA Predictive Model. We further adjusted the NMIM estimates
of exhaust VOC, CO and NOx emissions from onroad vehicles in a "sensitivity" analysis in
order  to reflect the significant degree of uncertainty which currently  exists with respect to these
effects. Air toxic emissions were adjusted in order to reflect changes in total exhaust VOC
emissions,. Finally, the effect of ethanol on permeation VOC and benzene emissions also were
added to the onroad emission estimates.  This series of post-processing steps are further
described in the sections below.

       For nonroad emissions, the only adjustment  to the NMIM estimates was to adjust air
toxic emissions in the two control cases to reflect the change in the toxic fraction of VOC
emissions to that estimated for onroad vehicles, as opposed to that estimated for nonroad
equipment. These steps for calculating emissions inventories  are described in the following
sections.  A summary of the models used and fundamental post-processing steps are shown in
Table 4.1-1 below.
                                          175

-------
                Table 4.1-1.  Estimation of National Emissions Inventories:
                   Models Used and Fundamental Post-Processing Steps
                      Exhaust Emissions
                                                  Non-Exhaust Emissions
Onroad
Model: NMIM which runs MOBILE6.2.

Post-processing:

1. Replace VOC and NOx fuel effects for Tier 0
vehicles from MOBILE6.2 with fuel effects from
EPA Predictive Model;

2. Conduct sensitivity analysis by applying fuel
effects for Tier 0 vehicles to all vehicles.

3. Adjust exhaust air toxics emissions to reflect
adjustment to exhaust VOC emissions.
                                                   Model: NMDVI which runs MOBILE6.2.

                                                   Post-processing:

                                                   1. Add effect of ethanol on permeation emissions
                                                   of VOC and benzene.
Nonroad
Model: NMIM which runs NONROAD2005
(modified to account for hose permeation).
Post-processing:

1. Changes in toxic fraction of VOC emissions in
two fuel control cases based on onroad estimates
instead of nonroad estimates.
Model: NMIM which runs NONROAD2005
(modified to account for hose permeation).
Post-processing:

1. Changes in toxic fraction of VOC emissions in
two fuel control cases based on onroad estimates
instead of nonroad estimates.
4.1.2.1
   Onroad Emission Estimation Procedures
       We ran NMIM to estimate county-specific emissions from gasoline motor vehicles for
January and July in years 2012, 2015, and 2020. For each month and year combination, we ran
the three onroad cases (Reference, RFS, and EIA).  The NMIM model utilizes the
MOBILE6.2WWW model to estimate motor vehicle emissions, as well as the effect of fuel quality
on emissions. As discussed in Chapter 3, the EPA Predictive Model contains more recent
estimates of the impact of fuel quality on exhaust VOC and NOx emissions.  Therefore, we
removed the impact of fuel quality on exhaust VOC and NOx emissions as estimated by
MOBILE6.2 and replaced these impacts with those of the EPA Predictive Model. As also
discussed in Chapter 3, MOBILE6.2 does not include the impact of ethanol on permeation
emissions.  Therefore, we added these emissions to those estimated by NMIM. Finally, we
arrived at annual emissions estimates by summing the January and July results, then multiplying
by six. The procedures for making these changes are discussed below.
4.1.2.1.1
   Onroad Exhaust Emissions
       MOBILE6.2 performs most of its emission estimation procedures for a non-oxygenated
8.7 RVP gasoline.  The effect of differing fuel quality is represented by a set of adjustment
factors, which can vary by vehicle type, model year, and whether the vehicle is properly
operating or not (i.e., is a low or high emitter). Because the mix of vehicle types, model years,
and low and high emitters varies by county and calendar year, it is infeasible to estimate the net
impact of each fuel parameter on emissions outside of the model. In  Section 3.1.1.1.2 of Chapter
3, we describe a process whereby we performed linear regressions on the exhaust emissions
                                           176

-------
estimated by NMIM in order to determine the average effect of RVP, ethanol content and MTBE
content on exhaust VOC and NOx emissions. Also in Section 3.1.1.1.2, we describe these same
impacts using the EPA Predictive Model. We combined these fuel-emission effects with the fuel
quality expected to exist in each county under each ethanol use case to estimate the adjustment
which NMIM had applied to exhaust VOC and NOx emissions. This NMIM adjustment for fuel
quality was removed and replaced by one based on the EPA Predictive Models. In our primary
analysis,  the fuel-emission effects from the EPA Predictive Models were only applied to the
fraction of exhaust VOC and NOx emissions which are emitted by Tier 0 vehicles. In our
sensitivity analysis, the fuel-emission effects from the EPA Predictive Models were applied to all
exhaust VOC and NOx emissions.

       Table 4.1-2 shows the values for "Tier 0 Fraction"; i.e., the fraction of VOC and NOx
emissions from vehicles with Tier 0 emissions characteristics. Note that the fraction drops as
time progress, reflecting the attrition of such vehicles in the national fleet. In the sensitivity
analysis,  the Tier 0 vehicle emission fraction is 1.0 for all years and pollutants.

            Table 4.1-2.  Fraction of In-Use Exhaust Emissions Attributable to
                     Vehicles with Tier 0 Emissions Characteristics
Calendar Year
2012
2015
2020
VOC
0.339
0.183
0
NOx
0.162
0.065
0
       After adjusting exhaust VOC and NOx according to the methods described above, we
adjusted the four exhaust toxic emissions: benzene, 1,3 -butadiene, formaldehyde, and
acetaldehyde. MOBILE6.2 estimates exhaust toxic emissions by first estimating the fraction of
exhaust VOC emissions represented by each toxic based on fuel quality. The model then applies
this fraction to exhaust VOC emissions to estimate absolute emissions of air toxics.  Since we
adjusted exhaust VOC emissions, it was necessary to adjust exhaust toxic emissions, as well, by
the ratio of the change in exhaust VOC emissions.

       As described in Section 3.1.1.1.2 of Chapter 3, carbon monoxide emissions were also
adjusted. The following equation illustrates the CO adjustment:
CO  =   C    X  (1+(Etoh Vol% x Etoh ^ Shr + MTBE Vol% x MTBE Mkt Shr x 0.5454) x CO Adj. Factor )
4.1.2.1.2     Onroad Non-Exhaust Emissions

       The only adjustment to the non-exhaust emission estimates from NMIM was to add
county-specific  estimates of the increase in permeation emissions due to ethanol use. In Section
3.1.1.3 of Chapter 3, we determined that a 10 vol% ethanol blend increased permeation
emissions by 0.8 grams per day at 95 F.  We also concluded there that permeation emissions
double with every increase in temperature of 18 °F. Because of this temperature relationship,
                                          177

-------
permeation effects were only accounted for in the July emission estimate since emissions during
the winter months could be at least four times lower, and thus negligible.
       Permeation emissions occur whether a vehicle is being used or is parked.  Therefore, the
average hourly emission factor in each county in July is determined by adjusting the 0.8 gram
per day emission rate for the average fuel tank temperature occurring in that hour of the day in
each county in July and multiplying by the market share of E10 fuel in that county. Total
monthly emissions in each county were determined by summing across hours of the day,
multiplying by 31 days and multiplying by the number of vehicles estimated to reside in that
county.

       The average fuel tank temperature is a function of the average ambient temperature at
that hour of the day, adjusted to account for the increase in fuel tank temperature for those
vehicles which are operating or which are still cooling down from operating. We obtain
estimates of these latter two factors from EPA's Draft MOVES2006 model.xxx These are
shown in Table 4.1-3.  The fuel tank temperature of vehicles which have been parked some time
tend to lag the ambient temperature both when the latter is rising and falling. We assume here
that the fuel tank temperature of these parked vehicles is equal to the ambient temperature, which
is true on average for the day.

           Table 4.1-3. Increase in Fuel Tank Temperature Relative to Ambient
Hour of the Day
Midnight
1:00 AM
2:00 AM
3:00 AM
4:00 AM
5:00 AM
6:00 AM
7:00 AM
8:00 AM
9:00 AM
10:00 AM
11:00 AM
Noon
1:00 PM
2:00 PM
3:OOPM
4:00 PM
5:00 PM
6:00 PM
7:00 PM
8:00 PM
9:00 PM
10:00 PM
11:OOPM
Vehicles Operating or in Hot Soak
2.6%
2.8%
1.2%
0.9%
0.8%
2.6%
6.6%
12.3%
14.0%
10.0%
11.1%
12.5%
15.6%
16.0%
17.2%
21.0%
23.7%
28.5%
30.0%
25.7%
18.7%
13.5%
10.6%
7.8%
Average Tank Temperature Rise (F)
10.0
6.9
6.1
4.9
3.1
3.0
3.7
4.6
3.5
3.8
3.8
4.9
4.8
5.5
6.6
7.7
8.6
8.3
8.8
9.2
8.3
7.6
8.0
8.4
                                          178

-------
       The total number of gasoline vehicles in the U.S. in 2004 is estimated to be 228
million.YYY We increased this figure by 1.9% per year53 to derive estimates of the gasoline
vehicle fleet in 2012, 2015  and 2020. This produced estimates for the fleet of gasoline vehicles
in the U.S. of 265, 281 and 308 million vehicles in 2012, 2015 and 2020, respectively. These
vehicles were allocated to each county based on the county-specific distribution of national VMT
by gasoline vehicles contained in NMIM.

       As described in Section 3.1.1.3 of Chapter 3, we estimate that benzene represents 3% of
the increased VOC permeation emissions due to ethanol use.  Thus, we added this 3% to the non-
exhaust emissions of benzene estimated by NMEVI.

4.1.2.2       Nonroad Emissions

       NMEVI is capable of utilizing any one of a series of EPA's NONROAD emission models.
We chose to use the NONROAD2005ZZZ model to estimate emissions from nonroad equipment
here, as it reflects EPA's latest estimates of emission factors for nonroad equipment.  EPA has
also recently developed a set of emission factor inputs for the NONROAD model which include
the effect of ethanol on permeation emissions from a number of types of nonroad equipment (see
Chapter 3).

       For the proposed rule inventories, the NONROAD  model was not able to select ethanol
related emission factors based on the fuel quality  inputs to the model. It was therefore necessary
to run NMIM for two extreme ethanol use cases (no ethanol use and 100% ethanol use) and use
those results to estimate emissions for the five ethanol use  cases which were the focus of the
proposed rule.

       For the final rule, NONROAD model capabilities were updated to account for oxygenate
effects.  Therefore, we were able to run NMIM (which runs NONROAD) using the same fuel
property inputs that were used for onroad emissions inventories. This eliminated the need to
interpolate between the "No Oxygen" and "All Oxygen" NONROAD runs that were needed for
the proposal.

       For nonroad toxic exhaust emissions, the toxic emissions factors for nonroad equipment
are based on very limited data. In EPA's recent final rule which implemented new Mobile
Source Air Toxic (MS AT)  standards, we adjusted the fraction of nonroad VOC emissions
represented by the various air toxics contained in  NMIM for a reduction in fuel benzene content
with those estimated for the same fuel change by MOBILE6.2 for onroad motor vehicles.  This
was done because of the very limited amount of nonroad emission test data which both varied
fuel quality and measured toxics emissions. We take the same approach here. We begin with
the estimate of nonroad toxic emissions from NMIM for the Reference Case.  Then, any change
in the toxics fraction of nonroad VOC emissions due to a change in fuel  quality predicted by
NMIM is replaced by the change in the toxics fraction of onroad VOC emissions due to the same
53 Annual growth rate in gasoline consumption on an energy basis per EIA Annual Energy Outlook, 2006 (therefore
it applies regardless of future ethanol use scenario). Assumes constant annual mileage per vehicle over this
timeframe.
                                          179

-------
change in fuel quality predicted by MOBILE6.2. This adjustment is illustrated in the following
equation:
Adjusted
Nonroad
Toxic
Emissions
NMIM
Toxic
Emissions
NMIM VOC
Emissions (RFS
or EIA case)
NMIM VOC
Emissions
(Reference case)
MOBILE6.2 Toxic Emissions (RFS or EIA Case)
MOBILE6.2 VOC Emissions (RFS or EIA Case

MOBILE6.2 Toxic Emissions (Reference Case)
MOBILE6.2 VOC Emissions (Reference Case)
4.1.3   National Emissions Inventory Projections
4.1.3.1
Emission Inventories: Primary Analysis
       This section provides the national emissions inventories for the primary case analyses.
Criteria pollutant inventories are included, along with a brief discussion of the trends.  A short
discussion of air toxics inventories is also included. See Tables 4A-1 through 4A-7 in the
Chapter 4 Appendix for complete primary-case inventories on air toxics and criteria pollutants,
as well as the percent changes in inventories from the Reference case.

       Table 4.1-4 shows ethanol impacts on VOC inventories for each of the three cases of
renewable fuel use in years 2012, 2015, and 2020.  In any given year, the data suggest that total
VOC emissions will increase as ethanol use increases. The largest increase is seen in the EIA
case, where the increase is about 1% of the Reference case inventory.

       Our analysis indicates that this increase is a result of VOC non-exhaust emissions, such
as those from evaporation or permeation. While VOC exhaust emissions decrease, they do not
decrease enough to counteract the increase from non-exhaust emissions.

                                      Table 4.1-4.
            National VOC Emissions from Gasoline Vehicles and  Equipment:
    Reference Case Inventory and Change in Inventory for Control Cases (Tons/Year)
Primary Case
Total
Reference
RFS Case (Change)
EIA Case (Change)
On-Road
Reference
RFS Case (Change)
EIA Case (Change)
Non-Road
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
5,882,000
18,000
43,000
2012
3,417,000
10,000
32,000
2012
2,465,000
8,000
11,000
2015
5,569,000
25,000
49,000
2015
3,269,000
16,000
36,000
2015
2,300,000
9,000
13,000
2020
5,356,000
34,000
58,000
2020
3,244,000
23,000
42,000
2020
2,112,000
11,000
16,000
                                           180

-------
       Table 4.1-5 shows ethanol impacts on CO inventories for each of the three cases of
renewable fuel use in years 2012, 2015, and 2020. In any given year, data suggest that total CO
emissions will decrease as ethanol use increases.  The largest reduction is seen in the EIA case;
this decrease is still less than 3% of the Reference inventory.

                                      Table 4.1-5.
             National CO Emissions from Gasoline Vehicles and Equipment:
    Reference Case Inventory and Change in Inventory for Control Cases (Tons/Year)
Primary Case
Total
Reference
RFS Case (Change)
EIA Case (Change)
On-Road
Reference
RFS Case (Change)
EIA Case (Change)
Non-Road
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
55,022,000
-483,000
-1,366,000
2012
37,656,000
-45,000
-359,000
2012
17,366,000
-438,000
-1,007,000
2015
53,702,000
-473,000
-1,329,000
2015
36,171,000
-39,000
-321,000
2015
17,531,000
-434,000
-1,008,000
2020
53,949,000
-460,000
-1,286,000
2020
35,723,000
-19,000
-252,000
2020
18,226,000
-441,000
-1,034,000
       Table 4.1-6 shows ethanol impacts on NOx inventories for each of the three cases of
renewable fuel use in years 2012, 2015, and 2020. In any given year, the data suggest that total
NOx emissions will increase as ethanol use increases. The largest increase is seen in the EIA
case, which is around 2% of the Reference inventory.

       Our analysis also indicates that nonroad NOx emissions increase much greater than
onroad emissions. While onroad inventories increase less than one percent in control cases,
nonroad inventories increase up to 11% in the EIA case.
                                          181

-------
                                     Table 4.1-6.
            National NOx Emissions from Gasoline Vehicles and Equipment:
    Reference Case Inventory and Change in Inventory for Control Cases (Tons/Year)
Primary Case
Total
Reference
RFS Case (Change)
EIA Case (Change)
On-Road
Reference
RFS Case (Change)
EIA Case (Change)
Non-Road
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
2,487,000
23,000
40,000
2012
2,240,000
9,000
13,000
2012
247,000
14,000
27,000
2015
2,059,000
18,000
33,000
2015
1,797,000
3,000
4,000
2015
262,000
15,000
29,000
2020
1,695,000
17,000
32,000
2020
1,407,000
0
0
2020
288,000
17,000
32,000
       Table 4.1-7 shows ethanol impacts on air toxic emissions for each of the three cases of
renewable fuel use in 2012.

       For all air toxics shown, the most extreme changes occur in the EIA case. The data
suggest that, in 2012, total benzene emissions will decrease by about 4% due to decreases in both
onroad and nonroad emissions. Total 1,3-butadiene emissions decrease  by less than 2% due to
decreases in both onroad and nonroad emissions. Total formaldehyde emissions decrease by up
to 1.5%. Total acetaldehyde emissions increase by as much as 36% due to increases in both
onroad and nonroad emissions.

       Generally, the trends in 2015 and 2020 parallel those of 2012 and are shown in the
appendix to this chapter.  Benzene maintains a drop of up to about 6% with increased ethanol
use.  Formaldehyde remains fairly flat, ranging from a 0.5% increase to  a 1.2% decrease.
Acetaldehyde maintains an increase of as much as 36.5%.  Finally, 1,3-butadiene remains fairly
flat, ranging from no change to a 0.5% increase.

       Again, we  emphasize that the toxics inventories are based on very limited data, especially
when it comes to emissions from nonroad equipment.
                                         182

-------
                                      Table 4.1-7.
        National Toxic Emissions from Gasoline Vehicles and Equipment in 2012:
       Reference Case Inventory and Change in Inventory for Control (Tons/Year)
Primary Case
Benzene
1,3-Butadiene
Formaldehyde
Acet aldehyde
Total
Reference
RFS Case (Change)
EIA Case (Change)
178,000
-3,200
-7,200
18,900
-200
-300
40,400
-600
-200
19,900
3,400
7,100
Onroad
Reference
RFS Case (Change)
EIA Case (Change)
124,100
-2,300
-5,400
12,000
-200
-200
29,900
-600
-300
15,500
2,400
5,400
Nonroad
Reference
RFS Case (Change)
EIA Case (Change)
53,900
-900
-1,800
6,900
0
-100
10,500
0
100
4,400
1,000
1,700
4.1.3.2
Emission Inventories: Sensitivity Analyses
       This section provides the national emissions inventories for the sensitivity case analyses.
Criteria pollutant inventories are included, along with a brief discussion of the trends.  See
Tables 4A-1 through 4A-7 in the Chapter 4 Appendix for complete sensitivity-case inventories
on air toxics and criteria pollutants, as well as the percent changes in inventories from the
reference case.

       Table 4.1-8 shows ethanol impacts on VOC inventories for each of the three cases of
renewable fuel use in years 2012, 2015, and 2020. Where the primary analysis showed total
VOC emissions increasing with ethanol use in all cases, the sensitivity analysis shows that total
VOC emissions decrease. Onroad  emissions decrease in all cases, while nonroad emissions
increase to the same extent as under the primary analysis.
                                          183

-------
                                     Table 4.1-8.
            National VOC Emissions from Gasoline Vehicles and Equipment:
    Reference Case Inventory and Change in Inventory for Control Cases (Tons/Year)
Sensitivity Case
Total
Reference
RFS Case (Change)
EIA Case (Change)
On-Road
Reference
RFS Case (Change)
EIA Case (Change)
Non-Road
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
5,834,000
-20,000
-4,000
2012
3,369,000
-28,000
-15,000
2012
2,465,000
8,000
11,000
2015
5,510,000
-23,000
-10,000
2015
3,210,000
-32,000
-23,000
2015
2,300,000
9,000
13,000
2020
5,281,000
-27,000
-17,000
2020
3,169,000
-38,000
-33,000
2020
2,112,000
11,000
16,000
       Table 4.1-9 shows ethanol impacts on CO inventories for each of the three cases of
renewable fuel use in years 2012, 2015, and 2020. In any given year, the data suggest that total
CO emissions will decrease as ethanol use increases.  The onroad vehicle CO emission
reductions increase by roughly a factor of three compared to the primary analysis. This increases
the overall CO emissions reduction from  about 3% in the primary case to 4% in the sensitivity
case.
                                     Table 4.1-9.
             National CO Emissions from Gasoline Vehicles and Equipment:
    Reference Case Inventory and Change in Inventory for Control Cases (Tons/Year)
Sensitivity Case
Total
Reference
RFS Case (Change)
EIA Case (Change)
On-Road
Reference
RFS Case (Change)
EIA Case (Change)
Non-Road
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
54,315,000
-692,000
-1,975,000
2012
36,949,000
-254,000
-968,000
2012
17,366,000
-438,000
-1,007,000
2015
52,998,000
-676,000
-1,929,000
2015
35,467,000
-242,000
-921,000
2015
17,531,000
-434,000
-1,008,000
2020
53,183,000
-676,000
-1,937,000
2020
34,957,000
-235,000
-903,000
2020
18,226,000
-441,000
-1,034,000
       Table 4.1-10 shows ethanol impacts on NOx inventories for each of the three cases of
renewable fuel use in years 2012, 2015, and 2020. In any given year, the data suggest that total
NOx emissions will increase as ethanol use increases. The largest increase is seen in the EIA
case, where the increase in total emissions is as high as 4.6% of the reference inventory. As in
                                         184

-------
the primary analysis, nonroad NOx emissions increase much greater than onroad emissions.
While onroad inventories increase up to 3.5%, nonroad inventories increase upwards of 11.1% in
the El A case.
                                    Table 4.1-10.
            National NOx Emissions from Gasoline Vehicles and Equipment:
    Reference Case Inventory and Change in Inventory for Control Cases (Tons/Year)
Sensitivity Case
Total
Reference
RFS Case (Change)
EIA Case (Change)
On-Road
Reference
RFS Case (Change)
EIA Case (Change)
Non-Road
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
2,519,000
68,000
106,000
2012
2,272,000
54,000
79,000
2012
247,000
14,000
27,000
2015
2,087,000
57,000
91,000
2015
1,825,000
42,000
62,000
2015
262,000
15,000
29,000
2020
1,717,000
48,000
79,000
2020
1,429,000
31,000
47,000
2020
288,000
17,000
32,000
       Table 4.1-11 shows ethanol impacts on air toxic emissions for each of the five cases of
renewable fuel use in 2012.  The impacts in 2015 and 2020 are shown in the Appendix to this
chapter.

                                    Table 4.1-11.
        National Toxic Emissions from Gasoline Vehicles and Equipment in 2012:
       Reference Case Inventory and Change in Inventory for Control (Tons/Year)
Sensitivity Case
Benzene
1,3-Butadiene
Formaldehyde
Acet aldehyde
Total
Reference Case
RFS Case (Change)
EIA Case (Change)
175,700
-5,000
-9,400
18,600
-400
-600
39,600
-1,100
-700
19,500
3,000
6,600
Onroad
Reference Case
RFS Case (Change)
EIA Case (Change)
121,800
-4,100
-7,600
11,700
-400
-500
29,100
-1,100
-800
15,100
2,000
4,900
Nonroad
Reference Case
RFS Case (Change)
EIA Case (Change)
53,900
-900
-1,800
6,900
0
-100
10,500
0
100
4,400
1,000
1,700
      As in the primary analysis, the most extreme changes in the sensitivity analysis tend to
occur in the EIA case.
                                         185

-------
       The data suggest that, in 2012, total benzene emissions will decrease by as much as 5.4%
due to decreases in both onroad and nonroad emissions. Total formaldehyde emissions decrease
by up to 2.8%. Nonroad formaldehyde emissions tend to remain relatively flat, while onroad
emissions decrease.  Total acetaldehyde emissions increase by as much as 34% due to increases
in both onroad and nonroad emissions. Total 1,3-butadiene emissions decrease by about 3%.
4.1.3.3
Local and Regional VOC and NOx Emissions (Summer 2015)
       We also estimate the percentage change in VOC, NOx, and CO emissions from gasoline
fueled motor vehicles and equipment in those areas which actually experienced a significant
change in ethanol use.  Specifically, we focused on areas where the market share of ethanol
blends was projected to change by 50 percent or more. We also focused on summertime
emissions, as these are most relevant to ozone formation as discussed in Chapter 5. We modeled
2015 because the ozone Response Surface Model (RSM) used for air quality modeling (also
discussed in Chapter 5) is based upon a 2015 emissions inventory, though we would expect
similar results in 2012.  Finally, we developed separately estimates for:  1) RFG areas, including
the state of California and the portions of Arizona where their CBG fuel programs apply, 2) low
RVP areas (i.e., RVP standards less than 9.0 RVP, and 3) areas  with a 9.0 RVP standard.  This
set of groupings helps to highlight the emissions impact of increased ethanol use in those areas
where emission control is most important.

       Table 4.1-12 presents our primary analysis estimates of the percentage change in VOC,
NOx, and CO emission inventories for these three types of areas when compared to the 2015
reference case. Note that the analyses here is very similar to that described in  Section 5.1, with
the exception that Table 4.1-12 below reflects 50-state emissions (instead of 37 eastern states)
and excludes diesel emissions.
                                     Table 4.1-12.
 Change in July 2015 Emissions from Gasoline Vehicles and Equipment in Counties Where
                 Ethanol Use Changed Significantly - Primary Analysis
Ethanol Use
RFS Case
EIA Case
RFG Areas
Ethanol Use
VOC
NOx
CO
Down
0.8%
-3.4%
6.1%
Up
2.3%
1.6%
-2.6%
Low RVP Areas
Ethanol Use
VOC
NOx
CO
Up
4.2%
6.2%
-12.5%
Up
4.6%
5.7%
-13.7%
Other Areas (9.0 RVP)
Ethanol Use
VOC
NOx
CO
Up
3.6%
7.3%
-6.4%
Up
4.6%
7.0%
-6.0%
                                          186

-------
       As expected, increased ethanol use tends to increase NOx emissions. The increase in low
RVP and other areas is greater than in RFG areas, since the RFG in the RFG areas included in
this analysis all contained MTBE.  Also, increased ethanol use tends to increase VOC emissions,
indicating that the increase in non-exhaust VOC emissions exceeds the reduction in exhaust
VOC emissions. This effect is muted with RFG due to the absence of an RVP waiver for ethanol
blends.  See Chapter 2 for a discussion of how ethanol levels will change at the state-level.

       Table 4.1-13 presents the percentage change in VOC, NOx, and CO emission inventories
under our sensitivity analysis (i.e., when we apply the emission effects of the EPA Predictive
Models to all motor vehicles).

                                      Table 4.1-13.
 Change in July 2015 Emissions from Gasoline Vehicles and Equipment in Counties Where
                 Ethanol Use Changed Significantly - Sensitivity Analysis
Ethanol Use
RFS Case
EIA Case
RFG Areas
Ethanol Use
VOC
NOx
CO
Down
-1.0%
-0.9%
7.3%
Up
1.0%
5.6%
-3.0%
Low RVP Areas
Ethanol Use
VOC
NOx
CO
Up
3.4%
10.4%
-15.0%
Up
3.7%
10.8%
-16.4%
Other Areas (9.0 RVP)
Ethanol Use
VOC
NOx
CO
Up
3.0%
10.8%
-9.0%
Up
3.9%
11.0%
-8.9%
       Directionally, the changes in VOC and NOx emissions in the various areas are consistent
with those from our primary analysis. The main difference is that the increases in VOC
emissions are smaller, due to more vehicles experiencing a reduction in exhaust VOC emissions,
and the increases in NOx emissions are larger.
4.2    Impact of Biodiesel Use

       As discussed in Chapter 1, biodiesel use totaled 25 million gallons in 2004 and is
projected to increase to 300 million gallons in 2012. Total diesel fuel use in onroad diesels in
2004 was roughly 39.4 billion gallons and is expected to grow to 47.5 billion gallons per year by
2012.54 The volumes of biodiesel produced thus represent 0.06% and 0.6% of onroad diesel fuel
54  Based on linear interpolation between estimate for 2001 from Table 7.1.2-1 and that for 2014 from Table 7.1.3-4,
both from the 2010 Nonroad FRM Final RIA, EPA420-R-04-007, May 2004, available in EPA Docket OAR-2003-
0012.
                                           187

-------
consumption in 2004 and 2012, respectively.  Given the very small contribution of biodiesel to
the pool of diesel fuel, the nationwide emission impacts of biodiesel are expected to be similarly
small for the foreseeable future.  As a result, we have not included biodiesel emission impacts in
our emission inventory estimates for this rule.

       We do intend to investigate these impacts in the future, however. As stated in Chapter 3,
the 2002 EPA report entitled "A Comprehensive Analysis of Biodiesel Impacts on Exhaust
Emissions" concluded that biodiesel fuels improved PM, CO and HC emissions of diesel engines
while slightly increasing their NOx emissions. Nevertheless, these conclusions remain
controversial due to conflicting results from different studies. As a result, preparations are being
made to launch a test program with stakeholder participation to further investigate the emission
impacts of biodiesel.
4.3    Impact of Renewable Fuel Production and Distribution

4.3.1   Ethanol

       In Chapter 2, we estimated that 3.5 billion gallons of ethanol was produced for use in
motor fuel in 2004, which comprises our estimate of fuel quality for the base case .  Maintaining
fuel quality, but increasing fuel volume to that expected in 2012,55 ethanol use would increase to
3.9 billion gallons. The increases in emissions associated with ethanol production and
distribution under the RFS and EIA cases are, thus, determined relative to the emissions
associated with producing and distributing 3.9 billion gallons of ethanol.

       We describe the emissions associated with producing and distributing ethanol on a per
gallon basis  in Chapter 3.4.1.  There, we compare emissions factors from DOE's GREET model,
versions  1.6 and 1.7, as well as estimates of ethanol plant emissions obtained from the States.
We decided  there to use two emission estimates here, one from GREET 1.7, and the other from
GREET1.7 augmented by the State  estimates for ethanol plant emissions. Here, we simply
multiply  those emission factors by the volume of ethanol being used in each scenario. Table 4.3-
1 shows estimates of annual emissions expected to occur nationwide due to increased production
of ethanol.   It should be noted that emissions in the base case assume a 80/20 mix of dry mill
and wet mill facilities. New plants (and thus, the emission increases) assume 100% dry mil
facilities.
55 EIA projects gasoline demand of 16.93 and 18.84 quadrillion Btu in 2004 and 2012, respectively. This represents
overall growth between these two years of 11.3%.
                                          188

-------
                                      Table 4.3-1.
    Annual Emissions Nationwide from Ethanol Production and Transportation: 2012
                                     (tons per year)



voc
NOx
CO
PM10
SOx
GREET 1.7
Base Case
Emissions
8,000
17,000
49,000
21,000
27,000
RFS Case
El A Case
Increase in Emissions
5,000
13,000
35,000
15,000
20,000
11,000
26,000
72,000
30,000
41,000
GREET 1.7 + State Data
Base Case
Emissions
14,000
18,000
56,000
12,000
42,000
RFS Case
EIA Case
Increase in Emissions
10,000
14,000
40,000
9,000
30,000
20,000
27,000
81,000
18,000
61,000
       As can be seen, the potential increases in VOC and NOx emissions from ethanol
production and transportation are of the same order of magnitude as those from ethanol use.
Generally, ethanol plants are not located in ozone non-attainment areas, so the ozone impact of
the increased VOC and NOx emissions should be minimal.

       According to our estimates, almost 120 counties throughout the nation are constructing
new ethanol plants, expanding existing plants, or planning construction for future plants.  The
increases in ethanol production across these counties range from as low as 2 million gallons per
year for modest expansions, to over 270 million gallons per year due to the construction of
entirely new facilities. To estimate the potential increase in VOC and NOx emissions associated
with these plants, whether construction is planned or underway, we apply the ethanol production
emission factors (EFs) derived from state data as well as those found in GREET 1.7. See
Chapter 3.4 for a discussion of the emission factors related to ethanol production and plant
emissions.

       The ethanol production emission factors are applied to the increase in the volume of
ethanol production expected in each of the counties. Figures 4.3-1  and 4.3-2 illustrate potential
increases in future monthly VOC and NOx emissions, respectively, in counties that can expect a
growth in ethanol production. The emissions reflect plants operating for one month at 90%
capacity. In each figure, the distribution of counties is presented in order from the lowest-to-
highest increase in ethanol production volume.  The figures show results based upon both  state-
based emission factors and GREET 1.7 emission factors.
                                           189

-------
                 Figure 4.3-1.
 Monthly VOC Emissions in Counties Expecting an
         Increase in Ethanol Production
                              > VOC (State data EFs) I
          20     40      60
                                    100     120
               Distribution of Counties
         (in order of increasing ethanol production)
                Figure 4.3-2.
Monthly NOx Emissions in Counties Expecting an
        Increase in Ethanol Production
             Distribution of Counties
        (in order of increasing ethanol production)
       As the figures indicate, most counties will see an increase of less than 40 tons/month
VOC and less than 60 tons/month NOx, according to the distribution based upon the state data
emission factors. The average emissions are about 26 tons/month VOC and 35 tons/month NOx
using state data, and about 17 tons/month VOC and 25 tons/month NOx using GREET 1.7
emission factors. However, average VOC and NOx emissions increase to about 61 tons/month
and 83 tons/month, respectively, in the 10% of counties expecting largest increases in ethanol
production.  The average emissions for the remaining 90% of counties is about 21 tons/month
VOC and 29 tons/month NOx.  For both VOC and NOx, emissions estimates are about 35% less
when using the GREET 1.7 emission factors.
4.3.2  Biodiesel

       In Chapter 1, we estimated that 25 million gallons of biodiesel were produced for use in
motor fuel in 2004. Based on growth in overall diesel fuel demand between 2004 and 2012,56
this would represent the equivalent of 30 million gallons of biodiesel in 2012 for our reference
case. Here, we estimate the increase in emissions which will occur with an increase in biodiesel
production and distribution from 30 million gallons to 300 million gallons per year.

       We describe the emissions associated with producing and distributing biodiesel on a per
gallon basis in Chapter 3.  Here, we simply multiply those emission factors by the volume of
biodiesel being used in each scenario.  Table  4.3-2 shows estimates of annual emissions expected
to occur nationwide due to increased production of biodiesel.
56 EIA projects gasoline demand of 16.93 and 18.84 quadrillion Btu in 2004 and 2012, respectively. This represents
overall growth between these two years of 11.3%. Source: Annual Energy Outlook 2006, DOE/EIA-0383(2006),
Reference Case Table 2, available in docket EPA-HQ-OAR-2005-0161.
                                            190

-------
                                     Table 4.3-2.
    Annual Emissions Nationwide from Biodiesel Production and Transportation: 2012
                                    (tons per year)

voc
NOx
CO
PM10
SOx
Reference Inventory:
28 mill gal biodiesel per year
1,400
1,500
800
50
250
Increase in Emissions:
300 mill gal biodiesel per year
14,000
15,000
8,000
500
2,500
       As can be seen, the potential increases in emissions from biodiesel production and
transportation are of the same order of magnitude as those from biodiesel use, with the exception
of CO emissions. Generally, biodiesel plants are not located in ozone non-attainment areas, so
the ozone impact of the increased VOC and NOx emissions should be minimal.
4.4    Total Emission Impacts of Renewable Fuel Production and Use

       Tables 4.4-1 and 4.4-2 combine the VOC, CO and NOx emission impacts for ethanol use
from Section 4.1 and renewable fuel production and distribution from Section 4.3.  Table 4.4-1
includes the emission impacts from gasoline vehicles and equipment under our primary analysis
and renewable fuel production and distribution from GREET 1.7. Table 4.4-2 includes the
emission impacts from gasoline vehicles and equipment under our sensitivity analysis and
renewable fuel production and distribution from GREET 1.7 augmented with the State data for
ethanol production plants. Emissions from renewable fuel production and distribution in 2012
were increased by 1.9% per year to account for growth in gasoline and diesel fuel demand.
                                         191

-------
                                Table 4.4-1.
  National Emissions from Gasoline Vehicles and
      Production and Distribution: Primary Case
Equipment and Renewable Fuel
and GREET1.7 (Tons/Year)

VOC Emissions
Reference
RFS Case (Change)
EIA Case (Change)
CO Emissions
Reference
RFS Case (Change)
EIA Case (Change)
NOx Emissions
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
5,891,000
33,000
63,000
2012
3,467,000
50,000
108,000
2012
2,483,000
33,000
38,000
2015
5,578,513
41,969
71,311
2015
3,321,850
58,337
116,446
2015
2,319,026
36,482
42,596
2020
5,366,368
51,584
83,496
2020
3,301,600
69,232
130,856
2020
2,132,736
39,952
48,256
                                Table 4.4-2.
  National Emissions from Gasoline Vehicles and
Production and Distribution: Sensitivity Case and
Equipment and Renewable Fuel
GREET1.7/State Data (Tons/Year)

VOC Emissions
Reference
RFS Case (Change)
EIA Case (Change)
CO Emissions
Reference
RFS Case (Change)
EIA Case (Change)
NOx Emissions
Reference
RFS Case (Change)
EIA Case (Change)
Tons/Year
2012
5,849,000
-1,000
25,000
2012
3,426,000
16,000
70,000
2012
2,484,000
34,000
50,000
2015
5,525,855
-746
22,824
2015
3,270,249
15,622
67,959
2015
2,320,083
37,539
55,280
2020
5,298,280
-3,656
18,864
2020
3,234,664
13,992
66,224
2020
2,133,888
41,104
62,080
                                    192

-------
Chapter 4: Appendix
        193

-------
Table 4A-1. VOC Emission Inventories under Various Ethanol Use Cases
Primary
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
5,882,000
5,900,000
5,925,000
2012
3,417,000
3,427,000
3,449,000
2012
2,465,000
2,473,000
2,476,000
2015
5,569,000
5,594,000
5,618,000
2015
3,269,000
3,285,000
3,305,000
2015
2,300,000
2,309,000
2,313,000
2020
5,356,000
5,390,000
5,414,000
2020
3,244,000
3,267,000
3,286,000
2020
2,112,000
2,123,000
2,128,000
Change from Reference (tons)
2012
—
18,000
43,000
2012
—
10,000
32,000
2012
—
8,000
11,000
2015
—
25,000
49,000
2015
—
16,000
36,000
2015
—
9,000
13,000
2020
—
34,000
58,000
2020
—
23,000
42,000
2020
—
11,000
16,000
% Change from Reference
2012
—
0.3%
0.7%
2012
—
0.3%
0.9%
2012
—
0.3%
0.4%
2015
—
0.4%
0.9%
2015
—
0.5%
1.1%
2015
—
0.4%
0.6%
2020
—
0.6%
1.1%
2020
—
0.7%
1.3%
2020
—
0.5%
0.8%
Sensitivity
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
5,834,000
5,814,000
5,830,000
2012
3,369,000
3,341,000
3,354,000
2012
2,465,000
2,473,000
2,476,000
2015
5,510,000
5,487,000
5,500,000
2015
3,210,000
3,178,000
3,187,000
2015
2,300,000
2,309,000
2,313,000
2020
5,281,000
5,254,000
5,264,000
2020
3,169,000
3,131,000
3,136,000
2020
2,112,000
2,123,000
2,128,000

2012
__
-20,000
-4,000
2012
__
-28,000
-15,000
2012
__
8,000
11,000

2015
__
-23,000
-10,000
2015
__
-32,000
-23,000
2015
__
9,000
13,000

2020
__
-27,000
-17,000
2020
__
-38,000
-33,000
2020
__
11,000
16,000

2012
__
-0.3%
-0.1%
2012
__
-0.8%
-0.4%
2012
__
0.3%
0.4%

2015
__
-0.4%
-0.2%
2015
__
-1.0%
-0.7%
2015
__
0.4%
0.6%

2020
__
-0.5%
-0.3%
2020
__
-1.2%
-1.0%
2020
__
0.5%
0.8%
                               194

-------
Table 4A-2. CO Emission Inventories under Various Ethanol Use Cases
Primary
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
55,022,000
54,539,000
53,656,000
2012
37,656,000
37,611,000
37,297,000
2012
17,366,000
16,928,000
16,359,000
2015
53,702,000
53,229,000
52,373,000
2015
36,171,000
36,132,000
35,850,000
2015
17,531,000
17,097,000
16,523,000
2020
53,949,000
53,489,000
52,663,000
2020
35,723,000
35,704,000
35,471,000
2020
18,226,000
17,785,000
17,192,000
Change from Reference (tons)
2012
—
-483,000
-1,366,000
2012
—
-45,000
-359,000
2012
—
-438,000
-1,007,000
2015
—
-473,000
-1,329,000
2015
—
-39,000
-321,000
2015
—
-434,000
-1,008,000
2020
—
-460,000
-1,286,000
2020
—
-19,000
-252,000
2020
—
-441,000
-1,034,000
% Change from Reference
2012
—
-0.9%
-2.5%
2012
—
-0.1%
-1.0%
2012
—
-2.5%
-5.8%
2015
—
-0.9%
-2.5%
2015
—
-0.1%
-0.9%
2015
—
-2.5%
-5.7%
2020
—
-0.9%
-2.4%
2020
—
-0.1%
-0.7%
2020
—
-2.4%
-5.7%
Sensitivity
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
54,315,000
53,623,000
52,340,000
2012
36,949,000
36,695,000
35,981,000
2012
17,366,000
16,928,000
16,359,000
2015
52,998,000
52,322,000
51,069,000
2015
35,467,000
35,225,000
34,546,000
2015
17,531,000
17,097,000
16,523,000
2020
53,183,000
52,507,000
51,246,000
2020
34,957,000
34,722,000
34,054,000
2020
18,226,000
17,785,000
17,192,000

2012
__
-692,000
-1,975,000
2012
__
-254,000
-968,000
2012
__
-438,000
-1,007,000

2015
__
-676,000
-1,929,000
2015
__
-242,000
-921,000
2015
__
-434,000
-1,008,000

2020
__
-676,000
-1,937,000
2020
__
-235,000
-903,000
2020
__
-441,000
-1,034,000

2012
__
-1.3%
-3.6%
2012
__
-0.7%
-2.6%
2012
__
-2.5%
-5.8%

2015
__
-1.3%
-3.6%
2015
__
-0.7%
-2.6%
2015
__
-2.5%
-5.7%

2020
__
-1.3%
-3.6%
2020
__
-0.7%
-2.6%
2020
__
-2.4%
-5.7%
                               195

-------
Table 4A-3. NOx Emission Inventories under Various Ethanol Use Cases
Primary
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
2,487,000
2,510,000
2,527,000
2012
2,240,000
2,249,000
2,253,000
2012
247,000
261,000
274,000
2015
2,059,000
2,077,000
2,092,000
2015
1,797,000
1,800,000
1,801,000
2015
262,000
277,000
291,000
2020
1,695,000
1,712,000
1,727,000
2020
1,407,000
1,407,000
1,407,000
2020
288,000
305,000
320,000
Change from Reference (tons)
2012
—
23,000
40,000
2012
—
9,000
13,000
2012
—
14,000
27,000
2015
—
18,000
33,000
2015
—
3,000
4,000
2015
—
15,000
29,000
2020
—
17,000
32,000
2020
—
0
0
2020
—
17,000
32,000
% Change from Reference
2012
—
0.9%
1.6%
2012
—
0.4%
0.6%
2012
—
5.7%
10.9%
2015
—
0.9%
1.6%
2015
—
0.2%
0.2%
2015
—
5.7%
11.1%
2020
—
1.0%
1.9%
2020
—
0.0%
0.0%
2020
—
5.9%
11.1%
Sensitivity
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
2,519,000
2,587,000
2,625,000
2012
2,272,000
2,326,000
2,351,000
2012
247,000
261,000
274,000
2015
2,087,000
2,144,000
2,178,000
2015
1,825,000
1,867,000
1,887,000
2015
262,000
277,000
291,000
2020
1,717,000
1,765,000
1,796,000
2020
1,429,000
1,460,000
1,476,000
2020
288,000
305,000
320,000
Change from Reference (tons)
2012
__
68,000
106,000
2012
__
54,000
79,000
2012
__
14,000
27,000
2015
__
57,000
91,000
2015
__
42,000
62,000
2015
__
15,000
29,000
2020
__
48,000
79,000
2020
__
31,000
47,000
2020
__
17,000
32,000
% Change from Reference
2012
__
2.7%
4.2%
2012
__
2.4%
3.5%
2012
__
5.7%
10.9%
2015
__
2.7%
4.4%
2015
__
2.3%
3.4%
2015
__
5.7%
11.1%
2020
__
2.8%
4.6%
2020
__
2.2%
3.3%
2020
__
5.9%
11.1%
                               196

-------
Table 4A-4. Benzene Emission Inventories under Various Ethanol Use Cases
Primary
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
178,000
174,800
170,800
2012
124,100
121,800
118,700
2012
53,900
53,000
52,100
2015
175,400
164,800
169,100
2015
124,200
122,600
119,400
2015
51,200
42,200
49,700
2020
179,900
178,200
174,200
2020
130,600
129,400
126,100
2020
49,300
48,800
48,100
Change from Reference (tons)
2012
—
-3,200
-7,200
2012
—
-2,300
-5,400
2012
—
-900
-1,800
2015
—
-10,600
-6,300
2015
—
-1,600
-4,800
2015
—
-9,000
-1,500
2020
—
-1,700
-5,700
2020
—
-1,200
-4,500
2020
—
-500
-1,200
% Change from Reference
2012
—
-1.8%
-4.0%
2012
—
-1.9%
-4.4%
2012
—
-1.7%
-3.3%
2015
—
-6.0%
-3.6%
2015
—
-1.3%
-3.9%
2015
—
-17.6%
-2.9%
2020
—
-0.9%
-3.2%
2020
—
-0.9%
-3.4%
2020
—
-1.0%
-2.4%
Sensitivity
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
175,700
170,700
166,300
2012
121,800
117,700
114,200
2012
53,900
53,000
52,100
2015
172,700
168,000
163,600
2015
121,500
117,500
113,900
2015
51,200
50,500
49,700
2020
176,500
171,900
167,200
2020
127,200
123,100
119,100
2020
49,300
48,800
48,100

2012
__
-5,000
-9,400
2012
__
-4,100
-7,600
2012
__
-900
-1,800

2015
__
-4,700
-9,100
2015
__
-4,000
-7,600
2015
__
-700
-1,500

2020
__
-4,600
-9,300
2020
__
-4,100
-8,100
2020
__
-500
-1,200

2012
__
-2.8%
-5.4%
2012
__
-3.4%
-6.2%
2012
__
-1.7%
-3.3%

2015
__
-2.7%
-5.3%
2015
__
-3.3%
-6.3%
2015
__
-1.4%
-2.9%

2020
__
-2.6%
-5.3%
2020
__
-3.2%
-6.4%
2020
__
-1.0%
-2.4%
                                 197

-------
Table 4A-5. Acetaldehyde Emission Inventories under Various Ethanol Use Cases
Primary
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
19,900
23,300
27,000
2012
15,500
17,900
20,900
2012
4,400
5,400
6,100
2015
20,000
23,400
27,300
2015
15,800
18,300
21,500
2015
4,200
5,100
5,800
2020
21,100
24,700
28,800
2020
17,000
19,800
23,300
2020
4,100
4,900
5,500
Change from Reference (tons)
2012
—
3,400
7,100
2012
—
2,400
5,400
2012
—
1,000
1,700
2015
—
3,400
7,300
2015
—
2,500
5,700
2015
—
900
1,600
2020
—
3,600
7,700
2020
—
2,800
6,300
2020
—
800
1,400
% Change from Reference
2012
—
17.1%
35.7%
2012
—
15.5%
34.8%
2012
—
22.7%
38.6%
2015
—
17.0%
36.5%
2015
—
15.8%
36.1%
2015
—
21.4%
38.1%
2020
—
17.1%
36.5%
2020
—
16.5%
37.1%
2020
—
19.5%
34.1%
Sensitivity
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
19,500
22,500
26,100
2012
15,100
17,100
20,000
2012
4,400
5,400
6,100
2015
19,500
22,400
26,100
2015
15,300
17,300
20,300
2015
4,200
5,100
5,800
2020
20,400
23,400
27,200
2020
16,300
18,500
21,700
2020
4,100
4,900
5,500

2012
__
3,000
6,600
2012
__
2,000
4,900
2012
__
1,000
1,700

2015
__
2,900
6,600
2015
__
2,000
5,000
2015
__
900
1,600

2020
__
3,000
6,800
2020
__
2,200
5,400
2020
__
800
1,400

2012
__
15.4%
33.8%
2012
__
13.2%
32.5%
2012
__
22.7%
38.6%

2015
__
14.9%
33.8%
2015
__
13.1%
32.7%
2015
__
21.4%
38.1%

2020
__
14.7%
33.3%
2020
__
13.5%
33.1%
2020
__
19.5%
34.1%
198

-------
Table 4A-6. Formaldehyde Emission Inventories under Various Ethanol Use Cases
Primary
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
40,400
39,800
40,200
2012
29,900
29,300
29,600
2012
10,500
10,500
10,600
2015
40,100
39,600
40,100
2015
30,100
29,700
30,100
2015
10,000
9,900
10,000
2020
41,400
41,200
41,600
2020
32,000
31,800
32,200
2020
9,400
9,400
9,400
Change from Reference (tons)
2012
—
-600
-200
2012
—
-600
-300
2012
—
0
100
2015
—
-500
0
2015
—
-400
0
2015
—
-100
0
2020
—
-200
200
2020
—
-200
200
2020
—
0
0
% Change from Reference
2012
—
-1.5%
-0.5%
2012
—
-2.0%
-1.0%
2012
—
0.0%
1.0%
2015
—
-1.2%
0.0%
2015
—
-1.3%
0.0%
2015
—
-1.0%
0.0%
2020
—
-0.5%
0.5%
2020
—
-0.6%
0.6%
2020
—
0.0%
0.0%
Sensitivity
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
39,600
38,500
38,900
2012
29,100
28,000
28,300
2012
10,500
10,500
10,600
2015
39,200
38,100
38,400
2015
29,200
28,200
28,400
2015
10,000
9,900
10,000
2020
40,300
39,200
39,500
2020
30,900
29,800
30,100
2020
9,400
9,400
9,400

2012
__
-1,100
-700
2012
__
-1,100
-800
2012
__
0
100

2015
__
-1,100
-800
2015
__
-1,000
-800
2015
__
-100
0

2020
__
-1,100
-800
2020
__
-1,100
-800
2020
__
0
0

2012
__
-2.8%
-1.8%
2012
__
-3.8%
-2.7%
2012
__
0.0%
1.0%

2015
__
-2.8%
-2.0%
2015
__
-3.4%
-2.7%
2015
__
-1.0%
0.0%

2020
__
-2.7%
-2.0%
2020
__
-3.6%
-2.6%
2020
__
0.0%
0.0%
199

-------
Table 4A-7. 1,3-Butadiene Emission Inventories under Various Ethanol Use Cases
Primary
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
18,900
18,700
18,600
2012
12,000
11,800
11,800
2012
6,900
6,900
6,800
2015
18,500
18,600
18,500
2015
12,000
12,000
12,000
2015
6,500
6,600
6,500
2020
19,100
19,100
19,100
2020
12,800
12,800
12,800
2020
6,300
6,300
6,300
Change from Reference (tons)
2012
—
-200
-300
2012
—
-200
-200
2012
—
0
-100
2015
—
100
0
2015
—
0
0
2015
—
100
0
2020
—
0
0
2020
—
0
0
2020
—
0
0
% Change from Reference
2012
—
-1.1%
-1.6%
2012
—
-1.7%
-1.7%
2012
—
0.0%
-1.4%
2015
—
0.5%
0.0%
2015
—
0.0%
0.0%
2015
—
1.5%
0.0%
2020
—
0.0%
0.0%
2020
—
0.0%
0.0%
2020
—
0.0%
0.0%
Sensitivity
Total
Reference
RFS Case
EIA Case
On-Road
Reference
RFS Case
EIA Case
Non-Road
Reference
RFS Case
EIA Case
Tons/Year
2012
18,600
18,200
18,000
2012
11,700
11,300
11,200
2012
6,900
6,900
6,800
2015
18,200
18,000
17,800
2015
11,700
11,400
11,300
2015
6,500
6,600
6,500
2020
18,700
18,300
18,200
2020
12,400
12,000
11,900
2020
6,300
6,300
6,300

2012
__
-400
-600
2012
__
-400
-500
2012
__
0
-100

2015
__
-200
-400
2015
__
-300
-400
2015
__
100
0

2020
__
-400
-500
2020
__
-400
-500
2020
__
0
0

2012
__
-2.2%
-3.2%
2012
__
-3.4%
-4.3%
2012
__
0.0%
-1.4%

2015
__
-1.1%
-2.2%
2015
__
-2.6%
-3.4%
2015
__
1.5%
0.0%

2020
__
-2.1%
-2.7%
2020
__
-3.2%
-4.0%
2020
__
0.0%
0.0%
200

-------
Chapter 5:  Air Quality Impacts
5.1    Ozone

       We performed ozone air quality modeling simulations for the eastern United States using
the ozone Response Surface Model (RSM) to estimate the effects of the projected changes in
emissions from gasoline vehicles and equipment.  The ozone RSM is a screening-level air
quality modeling tool that allows users to quickly assess the estimated air quality changes over
the modeling domain. The ozone RSM is a model of a full-scale air quality model and is based
on statistical relationships between model inputs and outputs obtained from the full-scale air
quality model.  In other words, the ozone RSM uses statistical techniques to relate a response
variable to a set of factors that are of interest, e.g., emissions of precursor pollutants from
particular sources and locations.  The following section describes the modeling methodology,
including the development of the multi-dimensional experimental design for control strategies
and implementation and verification of the RSM technique. Additional detail is available in the
Air Quality Modeling Technical Support Document (AQMTSD) that was drafted for the Mobile
Source Air Toxics Rule Proposal (published March 29, 2006).AAAA

       The foundation for the ozone response surface metamodeling analyses was the CAMx
modeling done in support of the final Clean Air Interstate Rule (CAIR). The CAIR modeling is
fully described in the CAIR Air Quality Modeling Technical Support Document, but a brief
description is provided below.BBBB  The modeling procedures used in the CAIR analysis (e.g.,
domain, episodes, meteorology) have been used for several EPA rulemaking analyses over the
past five years  and are well-established at this point.

       The ozone RSM uses the 2015 controlled CAIR emissions inventory as its baseline,
assuming future fuel quality remains unchanged from pre-Act levels, which serves as the
baseline for the analysis of the final RFS standards.cccc  We then compare these baseline
emissions to the emissions which would have occurred in the future if fuel quality had remained
unchanged from pre-Act levels to those which will occur with fuel quality reflecting the
increased renewable fuel use projected in the future. This approach differs from that
traditionally taken in EPA regulatory impact analyses. Traditionally, we would have compared
future emissions with and without the requirement of the Act. However, as described in Chapter
1, we expect that total renewable fuel use in the U.S. in 2012 to exceed 7.5 billion gallons even
in the absence of the Renewable Fuel Standard (RFS). Thus, a traditional regulatory impact
analysis would have shown no impact on emissions or air quality.

       The modeling simulations that comprised the metamodeling were conducted using
CAMx version 3.10.  It  should be noted that because the ozone RSM is built from CAMx air
quality model runs, it therefore has the same strengths and  limitations of the underlying model
and its inputs.  CAMx is a non-proprietary computer model that  simulates the formation and fate
of photochemical oxidants including ozone for given input sets of meteorological conditions and
emissions. The gridded meteorological data for three historical episodes were developed using
the Regional Atmospheric Modeling System (RAMS), version 3b.DDDD   In all, 30 episode days
were modeled using frequently-occurring, ozone-conducive,  meteorological conditions from the
                                          201

-------
summer of 1995. Emissions estimates were developed for the evaluation year (1995) as well as a
future year (2015).

       The CAMx model applications were performed for a domain covering all, or portions of,
37 States (and the District of Columbia) in the Eastern U.S., as shown in Figure 5.1-1. The
domain has nested horizontal grids of 36 km and 12 km. However, the output data from the
metamodeling is provided at a 12 km resolution (i.e., cells from the outer 36 km cells populate
the nine finer scale cells, as appropriate).  Although the domain of the ozone RSM is the 37
Eastern states, the expanded use of ethanol in fuel is expected to occur nationwide.  Chapter 4
describes the nationwide inventory impacts associated with the F standards.
       Figure 5.1-1. Map of the CAMx Domain Used for RFS Ozone Metamodeling
       The ozone RSM used for assessing the air quality impacts of expanded ethanol use in fuel
was developed broadly to look at various control strategies with respect to attaining the 8-hour
ozone NAAQS. The experimental design for the ozone RSM covered three key areas: type of
precursor emission (NOx or VOC), emission source type (i.e., onroad vehicles, nonroad vehicles,
area sources, electrical generating utility (EGU) sources,  and non-utility point sources), and
location in or out of a 2015 model-projected residual ozone nonattainment area. This resulted in
a set of 14 emissions factors.

       The 14 emission factors were randomly varied and used as inputs to CAMx. The
experimental design for these 14 factors was developed using a Maximin Latin Hypercube
method. Based on a rule of thumb of 10 runs per factor, we developed an overall design with
                                          202

-------
140 runs (a base case plus 139 control runs).  The range of emissions reductions considered
within the metamodel ranged from 0 to  120 percent of the 2015 CAIR emissions. This
experimental design resulted in a set of CAMx simulations that serve as the inputs to the ozone
response surface metamodel.

       To develop a response surface approximation to CAMx, we used a multidimensional
kriging approach, implemented through the MIXED procedure in SAS. We modeled the
predicted changes in ozone in each CAMx grid cell as a function of the weighted average of the
modeled responses in the experimental design. A response-surface was then fit for several ozone
metrics, namely the ozone design value, the 1-hour maximum value, the 24-hour average value
and the average ozone level between 9 am and 5 pm. The effect of changes in VOC and NOx
emissions on ozone was estimated in each grid cell covered by the model for each ozone metric
except the ozone design value. The ozone design value is the mathematically determined
pollutant concentration at a particular monitoring  site that must be reduced to, or maintained at or
below the National Ambient Air Quality Standard to assume  attainment.  The 8-hour ozone
design value is the 3-year average of the fourth-highest daily  maximum 8-hour average ozone
concentrations measured at each monitor within an area over each year, which must not exceed
0.08 ppm (85 ppb, considering round-off). Thus,  ozone design values only exist for grid cells
which contain ozone monitoring stations and  where ozone attainment has been an issue. Ozone
design values have been developed for 525 of the 2696 counties in the 37 state region,  of the 31
these  The specific ozone design values used in this analysis  are those for 2001, which represent
the average of the ozone design values determined for three, three-year periods (1999-2001,
2000-2002, and 2001-2003).  Validation was  performed and is summarized in the Mobile Source
Air Toxics rule Air Quality Modeling Technical Support Document.  The validation exercises
indicated that the ozone RSM replicates CAMx response to emissions changes very well for
most emissions combinations and in most locations.

       The ozone RSM limits the number of geographically distinct changes in VOC and NOx
emissions which can be simulated. Emissions from motor vehicles and nonroad equipment can
be varied separately. Distinct percentage changes in either the motor vehicle or nonroad
inventories can also be applied in ozone nonattainment and attainment areas. However, distinct
emission impacts cannot be simulated in various ozone nonattainment areas (e.g., Chicago and
Houston or New York and Kansas City).  This limits our ability to simulate the impact of
increased ethanol use in a couple of ways. First, ethanol use is not geographically uniform
across the U.S., either currently or in the future. Thus, the emission impacts resulting from
changes in ethanol use also varies geographically.  Second, the emission impacts of ethanol use
are not uniform. Ethanol use in RFG and other areas which do not grant ethanol blends a 1.0 psi
RVP waiver will not experience as much of an increase in VOC emissions with increased
ethanol use as areas which grant ethanol blends an RVP waiver.  Third, the impacts of new
ethanol plants will be even more geographically focused.  The Ozone RSM cannot generally be
applied to model the emission impacts from such local sources for a couple of reasons.  One, the
location of new ethanol plants is difficult to predict in many cases. Two, the impact of these
plants on local emissions can be very large in percentage terms given the absence of a lot of
other  industrial activity. The Ozone RSM was designed to represent the ozone impact of the
same  change in VOC or NOx emissions across a broad region (e.g., all attainment areas).
Therefore, it cannot be used to model the impact of a large change in one county's emissions
                                          203

-------
without also assuming the same change in the upwind county's emissions. As not every county
will contain a new ethanol plant, the assumptions inherent in the Ozone RSM do not match the
situation of a new individual point source, such as an ethanol plant.

       We developed a methodology which would best approximate the impact of changes in
local emissions on the ozone level in each local area, while maintaining as much of the impact of
ozone transport from other areas as possible given the above mentioned limitations. We do this
by running the ozone RSM twice for each scenario and drawing the resultant ozone impact from
the run which best matched the emission impact expected in a particular local area, considering
both the change in emissions modeled for that particular local  area, as well as that occurring in
upwind areas.

       First, as mentioned above, ethanol use is expected to change dramatically  in some areas,
but not at all in others. Averaging the emission impacts across these two types of areas and
estimating the associated ozone impact would be very misleading.  No area would be likely to
experience the ozone impact predicted.  Some areas would experience a much greater impact,
while others would experience no impact. Therefore, the first step in using the Ozone RSM to
predict the ozone impacts related to the RFS is to estimate the change in VOC and NOx
emissions in those areas ethanol blend market share changed significantly. As was done in the
analysis of local emission impacts presented in  Section 4.1.3.3 above, we defined a significant
change in ethanol blend market share as a change of 50% or more. This focuses the change in
emissions in those areas where the change is likely to occur.

       As discussed in Chapters 3 and 4, the effect of ethanol  use on emissions differs
depending on the baseline fuel quality and the applicable RVP standards. In particular, ethanol
use has significantly different impacts on emissions in RFG, low RVP and 9 RVP areas.
Therefore, in order to better predict the ozone impact likely  to occur in specific areas, we
estimate the change in VOC and NOx emissions separately for RFG, low RVP  and 9 RVP areas
(per above, only for those areas in each case where ethanol blend market share  changed by 50%
or more).

       The Ozone RSM only covers the 37 easternmost states in the U.S.  Therefore, we limited
the calculation of VOC and NOx emission impacts to only those states. The Ozone RSM was
developed with the year 2015 as the default year.  Since we  develop most of our impacts of the
RFS for the year 2012 and 2015, we chose to run the Ozone RSM for, 2015.  The Ozone RSM is
designed to accept emission changes in terms of total onroad and total nonroad sources,
respectively, and both emission categories include diesels. Therefore, we included estimates of
VOC and NOx emissions from diesel vehicles and equipment  in 2015 in our calculation of the
emission impacts.  These diesel emissions do not change between the various RFS scenarios.
However, they do reduce the effective percentage change in VOC and NOx emissions which is
projected to occur.  Overall, these analyses are very similar to  those described in Section 4.1.3.3
above, with the exceptions of the limitation to 37 states and  the inclusion of diesel emissions.
The results of these calculations are shown in Table 5.1-1.
                                          204

-------
                                     Table 5.1-1.
   Emission Inputs to Ozone Modeling: Change in Total Mobile Sources Emissions in 37
   Eastern States where Ethanol Use Changes Significantly, July 2015 (percent change)

VOC
On Road
Attain.
(9 RVP)
Non-
Attain.
Non Road
Attain.
(9 RVP)
Non-
Attain.
NOx
On Road
Attain.
(9 RVP)
Non-
Attain.
Non Road
Attain.
(9 RVP)
Non-
Attain.
Primary Analysis
RFS
EIA
RFC
LRVP
RFC
LRVP
7.5%
7.5%
8.2%
8.2%
-1 .2%
8.9%
1 .7%
9.3%
1 .5%
1 .5%
2.3%
2.3%
1 .8%
1 .7%
2.4%
2.0%
0.2%
0.2%
0.2%
0.2%
0.1%
0.2%
0.1%
0.2%
3.0%
3.0%
3.2%
3.2%
-1 .9%
3.3%
0.7%
3.3%
Sensitivity Analysis
RFS
EIA
RFC
LRVP
RFC
LRVP
6.0%
6.0%
6.6%
6.6%
-5.3%
7.0%
-1 .2%
7.4%
1 .5%
1 .5%
2.3%
2.3%
1 .8%
1 .7%
2.4%
2.0%
2.9%
2.9%
3.0%
3.0%
1 .7%
3.5%
2.2%
3.6%
3.0%
3.0%
3.2%
3.2%
-1 .9%
3.3%
0.7%
3.3%
       Our category of 9 RVP areas is very similar to the set of attainment areas in the Ozone
RSM. Therefore, the application of the emission impacts expected in 9 RVP areas in the Ozone
RSM was straightforward. However, both RFG and low RVP areas together generally comprise
the set of nonattainment areas in the Ozone RSM.  As seen in Table 5.1-1, the expected emission
impacts of the various RFS scenarios differ significantly depending on whether the area has RFG
or low RVP fuel. Both sets of emission impacts could not be run in the Ozone RSM at the same
time.  Therefore, we ran the Ozone RSM twice.  The first run applied the emission impacts
estimated for RFG areas to the ozone nonattainment areas in the Ozone RSM and applied the
emission impacts for 9.0 RVP areas to the ozone attainment areas in the Ozone RSM.  This run
should produce satisfactory projections of ozone impacts for all areas except those areas with
low RVP, as well as those areas where ethanol use is not expected to change.

       The second run applied the emission impacts  estimated for Low RVP areas to the ozone
nonattainment areas in the Ozone RSM and applied the emission impacts for 9.0 RVP areas to
the ozone attainment areas in the Ozone RSM.  This run should produce satisfactory projections
of ozone impacts for all areas  except those areas with RFG, as well as those areas where ethanol
use is not expected to change.

       For both runs of the Ozone RSM, we set the predicted change in ozone to zero in those
counties not expected to experience a significant change in ethanol use.  This ignores any impact
from ozone transport from other areas where ethanol use did change. However, we believe that
the ozone impacts due to transport are much smaller than those associated with changes in local
emissions.  This is particularly true in this case, where the percentage change in emissions would
be the same in both the local and upwind areas.
                                         205

-------
       We merged the results of the two runs by attributing ozone impacts from each county
according to their nonattainment designation and fuel type. For non-attainment areas, this choice
was obvious. Non-attainment counties with RFG programs were assigned the ozone impacts
from the first run (i.e., the run where the changes in VOC and NOx emissions were the average
of those observed for RFG areas). Non-attainment counties with Low RVP programs were
assigned the ozone impacts from the second run (i.e., the run where the changes in VOC and
NOx emissions were the average of those observed for Low RVP areas). For attainment areas
(i.e., 9 RVP areas), the results of either run could be used,  as both runs of the Ozone RSM
applied the same emission  changes to attainment areas.  Thus, the local emission impacts would
be identical in the two Ozone RSM runs.  Ozone transport is also likely identical for the vast
majority of these counties,  given that they are likely downwind from other attainment area
counties.  The only difference occurs if an attainment area is downwind of a RFG or Low RVP
area.  For a nationwide analysis such as this one, we were not able to determine for each
attainment area whether a potential upwind area was more likely to be an RFG or Low RVP area.
Therefore, we chose to use the ozone impacts results from the first Ozone RSM run of the model
(i.e., where the emission impacts for RFG areas were applied to ozone nonattainment areas) for
all attainment areas. We chose this run because RFG areas tend to have the higher ozone levels
than Low RVP areas and thus, would be more likely to affect areas downwind.  We present the
ozone impacts of increased ethanol use resulting from this methodology in the following section.

5.1.1   Ozone Response Surface Metamodel Results

       This  section summarizes the results of our modeling of ozone air quality impacts in the
future with and without the expanded use of ethanol in fuel.  The impact of increased ethanol use
on the 8-hour ozone design values in 2015 are presented in Table 5.1-2.  The changes presented
in Table 5.1-2 are for those counties with 2001  modeled design values.57 The Chapter 5
Appendix presents the impacts of increased ethanol use on a number of alternative measures of
ambient ozone concentration.
57 2001 design values were calculated as an average of the 1999-2001, 2000-2002 and 2001-2003 design values at
each monitoring site. Please see the Air Quality Modeling Technical Support Document for the final Clean Air
Interstate Rule for additional information.
                                          206

-------
                                      Table 5.1-2.
      Impact of Increased Ethanol Use on 8-hour Ozone Design Values in 2015 (ppb)

RFS Case
EIA Case
Primary Analysis
Minimum Change
Maximum Change
Average Change Across 37 States
Population-Weighted Change Across
Average Change Where Ethanol Use
Population-Weighted Change Where
States
37 States
Changed Significantly States
Ethanol Use Changed Significantly
-0.015
0.329
0.057
0.052
0.153
0.154
0.000
0.337
0.079
0.056
0.181
0.183
Sensitivity Analysis
Minimum Change
Maximum Change
Average Change Across 37 States
Population-Weighted Change Across
Average Change Where Ethanol Use
Population-Weighted Change Where
States
37 States
Changed Significantly States
Ethanol Use Changed Significantly
-0.115
0.624
0.111
0.092
0.300
0.272
0.000
0.549
0.142
0.096
0.325
0.315
       As can be seen, ozone levels generally increase with increased ethanol use.  This is likely
due to the projected increases in both VOC and NOx emissions. Some areas do see a small
decrease in ozone levels.  In our primary analysis, where exhaust emissions from Tier 1 and later
onroad vehicles are assumed to be unaffected by ethanol use, the population-weighted increase in
ambient ozone levels is 0.052-0.056 ppb.  Since the 8-hour ambient ozone standard is 0.08 ppm
(85 ppb), this increase represents about 0.06 percent of the standard, a very small percentage58.
While small, this figure includes essentially zero changes in ozone in areas where ethanol use did
not change. When we focus just on those  areas where the market share of ethanol blends
changed by 50 percent or more, the population-weighted increase in ambient ozone levels rises
to 0.154-0.183 ppb. This increase represents about 0.2 percent of the standard.

       In our sensitivity analysis, where exhaust emissions from Tier 1 and later onroad vehicles
are assumed to respond to ethanol like Tier 0 vehicles, the population-weighted increase in
ambient ozone levels across the entire 37 state  area is slightly less than twice as high, or 0.092-
0.096 ppb.  This increase represents about 0.11 percent of the standard.  When we focus just on
those areas where the market share of ethanol blends changed by 50 percent or more, the
population-weighted increase in ambient ozone levels rises to 0.272-0.315 ppb.  This increase
represents about 0.35 percent of the standard.

       For the primary analysis, we also present the counties with the largest increases in the
ozone design value. Table 5.1-3 presents the county level ozone design value impacts of the
RFS case, while Table 5.1-4 presents the same information for the EIA case. It is important to
note that the results of this ozone response surface metamodeling exercise is meant for
screening-level purposes only and does not represent the results that would be obtained from
 : Appendix I of 40 CFR Part 50.
                                          207

-------
full-scale photochemical ozone modeling.  It is also important to note that the ozone RSM results
indicate that the counties which are projected to experience the greatest increase in ozone design
values are generally counties that are projected to have ambient concentrations well below the
0.08 ppm ozone standard in the 2015 baseline.

                            Table 5.1-3. RFS Case, Primary Analysis:
          2015 Ozone Response Surface Metamodeling Results3 for Counties with
    Largest Increases in Ozone 8hr Design Value (ppb) Due to Increased Use of Ethanol
State Name
Arkansas
Ohio
Ohio
Indiana
Ohio
Ohio
Maine
Ohio
Ohio
Louisiana
Louisiana
Illinois
Indiana
Ohio
Alabama
Louisiana
County Name
Crittenden Co
Geauga Co
Mahoning Co
Lake Co
Medina Co
Summit Co
York Co
Stark Co
Clinton Co
West Baton
Rouge Parish
Livingston
Parish
Cook Co
Shelby Co
Knox Co
Mobile Co
Jefferson Parish
2015 Baseline
(Post-CAIR)b
78
82.5
74.7
80.7
72
77.4
77.6
71.7
75.7
77.4
76.6
81.1
76.2
71.4
68
77.1
2015 RFS Case
78.3289
82.7972
74.9943
80.9929
72.2909
77.6901
77.8825
71.9707
75.9705
77.6685
76.8656
81.3605
76.4587
71.6541
68.2514
77.351
Effect of Expanded
Ethanol Use (ppb)
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
2015
Population
53,852
108,600
248,545
490,796
187,686
557,892
210,006
384,672
50,635
23,202
141,807
5,362,932
47,904
62,138
430,341
512,963
 a The Ozone RSM is meant for screening-level purposes only and does not represent results that would be
 obtained from full-scale photochemical ozone modeling.  In particular, the model does not account for changes in
 CO emissions or VOC reactivity, both of which should decrease with increased ethanol use and directionally
 reduce ozone, in areas where ozone formation is VOC-limited.
 b The Clean Air Interstate Rule (CAIR) modeling is fully described in the CAIR Air Quality Modeling Technical
 Support Document (Docket EPA-HQ-OAR-2005-0036).
                                            208

-------
                       Table 5.1-4.  EIA Case, Primary Analysis:
          2015 Ozone Response Surface Metamodeling Results for Counties with
   Largest Increases in Ozone 8hr Design Value (ppb) Due to Increased Use of Ethanol
State Name
Ohio
Ohio
Ohio
Arkansas
Ohio
Mississippi
Ohio
Indiana
Maine
New York
Texas
Ohio
Maine
Maine
Louisiana
Louisiana
Louisiana
Mississippi
Michigan
Ohio
Ohio
Louisiana
Florida
Ohio
Ohio
Florida
Indiana
Indiana
Massachusett
s
Michigan
Pennsylvania
New York
Mississippi
County Name
Geauga Co
Clinton Co
Mahoning Co
Crittenden Co
Summit Co
Adams Co
Stark Co
Shelby Co
York Co
Wayne Co
Travis Co
Medina Co
Hancock Co
KennebecCo
Livingston Parish
West Baton
Rouge Parish
Lafourche Parish
Warren Co
Huron Co
Franklin Co
Trumbull Co
Jefferson Parish
Pinellas Co
Delaware Co
Knox Co
Duval Co
Marion Co
Madison Co
Middlesex Co
Oakland Co
Beaver Co
Monroe Co
Harrison Co
2015 Baseline
(Post-CAIR)c
82.5
75.7
74.7
78
77.4
67.2
71.7
76.2
77.6
71.6
69.4
72
76.8
64.9
76.6
77.4
72.7
56.2
71.9
77
80
77.1
62.3
72.1
71.4
50.6
74.6
72.9
75.8
79.2
70.5
74.3
69.3
2015 EIA Case
82.8369
76.0218
75.0213
78.3204
77.7175
67.5164
72.0153
76.5115
77.902
71.8926
69.6912
72.2909
77.0904
65.1903
76.8883
77.6869
72.984
56.4827
72.1766
77.2716
80.2713
77.3707
62.5639
72.3606
71.6579
50.8568
74.8565
73.1564
76.0564
79.4542
70.7528
74.5521
69.5517
Effect of Expanded
Ethanol Use (ppb)
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
2015
Population
108,600
50,635
248,545
53,852
557,892
33,495
384,672
47,904
210,006
103,846
1,022,772
187,686
55,606
122,363
141,807
23,202
95,881
52,075
37,530
1,181,578
227,546
512,963
998,864
149,341
62,138
895,678
889,645
140,575
1,498,849
1,355,671
184,649
754,405
216,926
       There are a number of important caveats concerning our estimated ozone impacts using
the Ozone RSM. The Ozone RSM does not account for changes in CO emissions. As shown in
Chapter 4, ethanol use should reduce CO emissions significantly, directionally reducing ambient
ozone levels in areas where ozone formation is VOC-limited.  Accounting for the reduction in
CO emissions in NOx-limited areas, however, may have little impact on the ozone impact of
ethanol use.
                                        209

-------
       The Ozone RSM also does not account for changes in VOC reactivity. With additional
ethanol use, the ethanol content of VOC should increase. Ethanol is less reactive than the
average VOC. Therefore, this change should also reduce ambient ozone levels in a way not
addressed by the Ozone RSM. Again, like the impact of reduced CO emissions, this effect
applies to those areas where ozone formation is VOC-limited.  Another limitation is the RSM's
inability to simulate the spatial distribution of emission impacts associated with the standard.
Instead, we are forced to make simplifying assumptions about the geographic uniformity of RFS
emissions impacts, explained above. The caveats and limitations associated with the RSM
highlight the fact that it should only be used as a screening-level tool to characterize broad trends
associated with changes in different source categories of ozone precursors.

       Finally,  our application of the Ozone RSM here does not include the impact of emissions
from new ethanol plants. Direct!onally, this will increase ozone levels in the vicinity of the new
plant.  As discussed in Chapter 4, the overall VOC and NOx emission impacts of new ethanol
plants are only slightly lower than the emission impacts  resulting from increased use of ethanol
in vehicles and equipment. Given the concentrated nature of these impacts, the ozone impacts of
these new plants should be a focus of further study in the future.

       Keeping these limitations in mind, the expanded use of ethanol will impact the national
emissions inventory of precursors to ozone, such  as VOCs and NOx, as described in Chapter 4.
Exposure to ozone has been linked to a variety of respiratory effects including premature
mortality, hospital admissions and illnesses resulting in school absences. Ozone can also
adversely affect the agricultural and forestry sectors by decreasing yields of crops and forests.

       The Wisconsin Department of Natural Resources (DNR) recently performed a similar
study of the impact of increased  ethanol use on ozone.EEEE They estimated that the conversion of
gasoline outside of RFG areas in Wisconsin to E10 blends would increase ozone in these areas
on the  order of 1 ppb to as much as 2 ppb.  (RFG areas in Wisconsin already  contain 10 vol%
ethanol.) This ozone increase was due to the predicted increase in NOx emissions associated
with ethanol use, since the non-RFG areas in Wisconsin are generally NOx limited for ozone
formation.

       The Wisconsin DNR estimated the ozone  impact for calendar year 2003 and assumed that
all vehicles experience the increase in NOx emissions.  Thus, their results are more comparable
to our sensitivity analysis, than our primary analysis. For the two increased ethanol use
scenarios, our sensitivity analysis projects increased ozone levels for several Wisconsin counties
of 0.35-40 ppb. Because the Wisconsin DNR analyzed calendar year 2003 emissions and air
quality, their base emission levels are much higher than  those estimated here for the year 2015.
Emission standards applicable to new vehicles and equipment are continually reducing emissions
over time. Per the emission models used here and by the State of Wisconsin  (NONROAD and
MOBILE6), the effect of fuel quality is generally estimated in terms of a percentage change in
the base emission level.  As  emissions from vehicles and equipment decrease over time, the
absolute impact of fuel quality changes decreases at the  same rate. Thus, the absolute emission
changes predicted here for 2015  could easily be a factor of two lower than those predicted by
                                          210

-------
Wisconsin for 2003. This is likely the primary cause of the difference in the two sets of
projected ozone impacts.

       We received a comment from the Missouri Department of Natural Resources on the
ozone impact analysis performed for the NPRM which expressed a concern that our approach of
zeroing out the ozone impact in areas which did not experience a significant change in ethanol
use had the effect of ignoring the impact of ozone transport due to increased ethanol use in
upwind areas. This comment is correct. In a national analysis such as this one, it is not practical
to go through over 3100 counties to determine which counties might have not experienced a
change in ethanol use in a particular ethanol use case, but is downwind of an area which did.
Still, the issue is a potentially relevant one and of reasonable interest particularly to those tasked
with air quality management.

       In an attempt to approximate the impact of ozone transport from areas which did
experience a change in ethanol use on ozone in areas which did not, we performed one additional
run of the Ozone RSM. This additional run applied the changes in VOC and NOx emissions
estimated above for attainment areas from our sensitivity analysis for the EIA case to emissions
in attainment areas, and applied no change in emissions in non-attainment areas.  We then
compared the resulting ozone levels to those from the base case, focusing on the difference in
ozone levels in non-attainment areas. Emissions in non-attainment areas were the same in both
cases (no change from the base case).  Thus, the difference in ozone levels in non-attainment
areas should only be due to changes in emissions and ozone levels in upwind attainment areas.

       The results of this comparison indicated that, in terms of the 8-hour ozone design value,
ozone levels in non-attainment areas (i.e., RFG or Low RVP areas) decreased by 0.03 ppb.
Thus, the average impact due to ozone transport is a reduction in ozone in downwind areas.
However, the standard deviation in the ozone impact was 0.05 ppb, indicating that a significant
number of areas experienced an increase,  though most experienced a decrease. This is not
surprising given that ozone in some attainments areas is VOC limited and may be decreasing in
this fuel case, while others are NOx limited and may be increasing. The maximum ozone
reduction was 0.17 ppb, while the maximum increase was 0.12. More precise local atmospheric
dispersion modeling will be needed in order to estimate this type of impact for specific non-
attainment areas.

       In summary, we estimate that the measurable changes in VOC and NOx which are a
result of increased ethanol use will, on average, result in small increases in ambient ozone
formation. As we discussed above, the ozone modeling results in a net increase in the average
population weighted ozone design value metric measured within the modeled domain (37
Eastern states and the District of Columbia). In Appendix A, we also present the impacts  of
increased ethanol use on a number of alternative measures of ambient ozone concentration. We
acknowledge, however, that to the extent it occurs, increased future levels of ambient
concentrations of ozone related to the increased use of ethanol may result in  detrimental health
and welfare effects due to ozone.
                                          211

-------
5.2    Particulate Matter

5.2.1   Impact of Changes in Direct PM Emissions

       The amount of data evaluating the impact of ethanol and MTBE blending on direct
emissions of PM from gasoline-fueled vehicles is extremely limited, as discussed in Chapter 3.
Most studies do not test PM emissions from vehicles fueled with unleaded gasoline, because the
level of PM emissions from properly operating vehicles is usually very low, less than 0.1 g/mi.

       Two studies indicate that the addition of ethanol might reduce direct PM emissions from
gasoline vehiclesFFFF 'GGGG.  However, both studies were performed under wintertime conditions
and one at high altitude.  One of the studies only consisted of three vehicles.  The available data
indicate that ethanol blending might reduce exhaust PM emissions under very cold weather
conditions (i.e., 0 F or less), particularly at high altitude.  There is no indication of PM emission
reductions at higher temperatures or under warmed up conditions.  Thus, the data are certainly
too limited to support a quantitative estimate of the effect of ethanol on PM emissions.

5.2.2   Potential Impact of Changes in Secondary PM Formation

       In addition to being emitted directly from a combustion source, fine particles can be
formed through a series of chemical reactions in the atmosphere when SC>2, NOx, and VOC
oxidize or otherwise react to form a wide variety of secondary PM.  For example, SC>2 oxidizes
to SO3 and sulfuric acid and NOx oxidizes to NO3 and nitric acid which, in turn, react with
ammonia in the atmosphere to form ammonium sulfate and ammonium nitrate. Particles
generated through this gas to particle conversion are referred to as secondary aerosols (SA) and
represent a significant portion of ambient fine particulate matter. Studies have shown that as
much as 70% of the total organic  carbon in urban parti culate matter can be attributed to
secondary organic aerosol (SOA) formation although the amount can also be less.HF£HF£
Secondary PM tends to form more in the summer with higher temperatures and more intense
sunlight.

       Source-receptor modeling studies conducted in the Los Angeles area is 1993 by Schauer
et alira indicate that as much  as 67% of the fine parti culate matter collected could not be
attributed to primary sources. The authors concluded that much of this unidentifiable organic
matter is secondary organic aerosol formed in the atmosphere. This is consistent with previous
studies conducted by Turpin and Huntzicker in 1991 who concluded that 70% of the total
organic carbon in urban PM measurements made in southern California can be attributed to
SOA.

       Gas phase VOCs are  oxidized by OH, NO2, peroxyacetylnitrate (PAN), and ozone in the
atmosphere, but their propensity to  condense in the particle phase is a function of two factors:
volatility and reactivity.  To accumulate as an aerosol, a reaction product must first be formed in
the gas phase at a concentration equal to its saturation concentration. This requirement will not
be met if the relevant gas-phase reactions of the VOC are too slow or if the vapor pressure of the
reaction product is higher than the initial concentration of its VOC precursor.JJJJ  Limited data
for reaction rate constants determined both experimentally  and estimated by structural
                                          212

-------
relationships are available in the published literature. However, the atmospheric chemistry
behind SOA reaction rates and the estimated aerosol yield is highly complex and carries with it a
great deal of uncertainty. Research in this area is ongoing and thus the capacity to quantitatively
model SOA formation is not yet a straightforward process.

       In general, all reactive VOC are oxidized by OH or other compounds. Additionally,
alkenes, cycloalkenes,  and other olefinic compounds can react with ozone and NO2 to form
secondary aerosols.  In fact, ozone is responsible for nearly all the SOA formation from olefms,
while OH plays little or no role at all (Grosjean and Seinfeld, 1989; Izumi and Fukuyama, 1990).
Many VOC, however,  will never form secondary organic aerosol under atmospheric conditions
regardless of their reactivity. This is because the products of reactions of these compounds have
vapor pressures that are too high to form aerosols at atmospheric temperatures and pressures.
These include all alkanes and alkenes with up to 6 carbon  atoms, benzene and many low-
molecular weight carbonyls, chlorinated compounds and oxygenated solvents (Grosjean, 1992).

       The VOC that have the greatest propensity to form SOA include aromatic hydrocarbons
(such as toluene but even including benzene), higher molecular weight olefms and cyclic olefms,
and higher molecular weight paraffins. Kleindienst et al suggest that a high fraction of SOA is
due to aromatic hydrocarbon precursors. Furthermore, "aromatic products having a single alkyl
group on the aromatic  ring were found to represent a 'high-yield' family (e.g., toluene,
ethylbenzene); compounds having multiple methyl groups (e.g., m-xylene, 1,2,4-
trimethylbenzene) were found to represent a 'low-yield' family" (Kleindienst, 269). All of the
above mentioned VOC precursors are important either because there are large amounts of these
particular VOC emitted per day, or because a large fraction of the VOC reacts, or a combination
of the two.  Based on VOC emissions inventory data collected in the Los Angeles area, the most
important aerosol precursors (in the LA area using 1982 VOC emissions inventories) are listed in
Table 5.2-1 below:

                                      Table 5.2-1.
                           Predicted In Situ SOA Formation
                         During a Smog Episode in Los Angeles
VOC Functional Group kg emitted daily*
Aromatics
Olefins



Paraffins




Alkenes
Cyclic Olefins
Terpenes

Alkanes
Cycloalkanes
223985

31163
3220
6000

140493
37996
Secondary PM %
Produced (kg)* yield
3061

608
144
626

368
96
1.37

1.95
4.47
10.43

0.26
0.25
•Source: Grosjean et al, 1992
       These predictions are a function of input data collected in the Los Angeles area, and
assume ambient levels of [ozone] = 100 ppb, [OH]=1.0xl06 molecules/cm3, and [NO3]=0 with 6
hours of reaction time. Aromatics  are the largest functional group in terms of the absolute
                                          213

-------
quantities of VOC emitted daily, and thus they eventually form the most SOA.  Likewise, many
high molecular weight paraffins (alkanes) form SOA on a significant scale simply because their
emissions are high. However, the relative fraction of paraffins that react is less than that of
aromatics in smog chamber experiments simulating SOA formation in the atmosphere. For
olefins, the alkenes exhibit a combination of both relatively high emissions, and a high fraction
of VOC reacted to form SOA. Cyclic olefins, in contrast, are emitted in relatively low levels, but
a high fraction of these VOC react and the end result is a proportionally higher SOA yield than
with the alkenes. Lastly, there are several "miscellaneous" compounds and terpenes that are
emitted on a relatively small scale (in southern California), but that produce a substantial amount
of secondary organic aerosol.

       Researchers at EPA recently completed a field study in the Raleigh/Durham area of
North Carolina that investigated the contribution of various sources to ambient PM 2.5
concentrations.KKKK In the study they identified toluene as an SOA precursor.  They estimate
that mobile sources contribute nearly 90% of the total toluene emissions in that region based on a
chemical mass balance approach. At the same time, however, SOA attributable to non-fuel-
related VOC (i.e., biogenic emissions) was found to be an even larger contributor to SOA (i.e.,
toluene was not likely the dominant source of SOA in this area).  This study is currently
undergoing peer review and will be published shortly.  Qualitatively, however, this information
is still quite useful since the study identifies a contributing source of SOA that is attributable
almost entirely due to mobile sources.

       VOC reaction rates increase with increasing ambient temperature and sunlight intensity,
so the level of SOA formed is much higher in summer than in winter.  Even in the more
temperate coastal climates of southern CA, studies have found the summertime concentration of
SOA calculated through Chemical Mass Balance models show SOA formation to be anywhere
from 2-5 times higher in summer than winter. In a study conducted at both urban  and rural
locations in the southeastern United Sates, the concentration of SOA in the summer and early fall
was roughly 2-3 times that of colder monthsLLLL.

       As mentioned in Chapters 2 and 3, the addition of ethanol should reduce aromatics in
gasoline, which will in turn reduce the aromatics emitted in the exhaust. However, quantifying
the emission reduction is not possible at this time due to a lack of speciated exhaust data for
newer vehicles running on ethanol blends. In addition, increased NOx emissions resulting from
the increased use of ethanol could increase the formation of nitrate PM.

       Based on the following, we believe that it is likely that the decrease in secondary PM
from organic aromatic hydrocarbons is likely to exceed the increase in secondary nitrate PM.  In
1999, NOx emissions from gasoline-fueled vehicles and equipment comprised about 20% of
national NOx emissions from all sourcesMMMM. In areas where ethanol use increases
dramatically, NOx emissions from gasoline-fueled vehicles and equipment increases roughly 5-
10%. This is roughly equivalent to a 1-2% increase in NOx emissions nationwide.

       In contrast, gasoline-fueled vehicles and equipment comprised over 60% of all national
gaseous aromatic VOC emissions59. In areas where ethanol use increases dramatically, fuel
                                          214

-------
aromatic content decreases by about 4 vol% in the summertime, averaged across conventional
gasoline and RFG.  This represents about a 15% reduction from a base level of around 27 vol%.
Assuming a proportional relationship between fuel aromatics and aromatic emissions, this
represents about a 24% reduction in aromatic emissions nationwide.

       In most urban areas, ambient levels of excess summer carbonaceous PM (a reasonable
estimate of secondary organic PM) tend to exceed those of secondary nitrate PM. Thus,
directionally, it  appears likely that a net reduction in ambient PM levels will result from
increased ethanol use.  However, this should be considered a rough comparison at this time. A
more precise comparison will have to await the incorporation of secondary organic aerosol
formation into models, such as CMAx.

       The research to facilitate this incorporation is currently underway. EPA ORD scientists
are currently carrying out a wide variety of laboratory studies to refine the SO A chemistry
mechanisms for use in the next version of the  CMAQ model, which is expected to be completed
in 2007 and submitted for peer review. This information should be available in time for the
comprehensive  study of the Act's fuel requirements which is due in 2009.60
59 Based on internal analyses of emissions inventories.

60 Subject to funding.
                                          215

-------
Chapter 5: Appendix
        216

-------
Table 5A-1. 2015 Ozone Response Surface Metamodeling Summary Statistics for the RFS
Rule"'; Primary Scenario
Shour Design Value (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.015
0.329
0.057
0.086
0.052
EIA Scenario
0.000
0.337
0.079
0.100
0.056
24hr Average (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.168
0.197
0.008
0.018
0.014
EIA Scenario
-0.162
0.074
0.013
0.021
0.013
Ihr Maximum (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.104
0.213
0.015
0.033
0.029
EIA Scenario
-0.094
0.180
0.024
0.040
0.033
A verage 9-to-5 (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.107
0.203
0.012
0.027
0.022
EIA Scenario
-0.097
0.141
0.019
0.031
0.024
Average 10-to-3 (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.108
0.207
0.012
0.027
0.023
EIA Scenario
-0.107
0.149
0.019
0.032
0.024
   a Note that the statistics presented here represent ethanol use changes across the entire 37-state
   ozone RSM domain.
                                          217

-------
Table 5A-2.  2015 Ozone Response Surface Metamodeling Summary Statistics for the RFS
Rule"; Sensitivity Scenario
Shour Design Value (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.115
0.624
0.111
0.158
0.092
EIA Scenario
0.000
0.549
0.142
0.170
0.096
24hr Average (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.181
0.184
0.015
0.034
0.025
EIA Scenario
-0.173
0.142
0.024
0.041
0.028
Ihr Maximum (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.150
0.498
0.027
0.060
0.052
EIA Scenario
-0.133
0.346
0.043
0.072
0.062
A verage 9-to-5 (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.177
0.400
0.022
0.049
0.040
EIA Scenario
-0.163
0.260
0.034
0.057
0.046
Average 10-to-3 (ppb)
Statistic
Minimum Change
Maximum Change
Average Change
Standard Deviation
Population-Weighted Change
RFS Scenario
-0.182
0.431
0.022
0.050
0.041
EIA Scenario
-0.167
0.273
0.035
0.058
0.047
    a Note that the statistics presented
    entire eastern U.S. 37-state ozone
here reflect the impact of ethanol use changes across the
RSM domain.
                                           218

-------
Chapter 6: Lifecycle Impacts on Fossil Energy and Greenhouse
Gases
6.1    Lifecycle Modeling

       Lifecycle modeling accounts for the energy and emissions from a production process. It
incorporates the material aspects, input and output, of each step in a product system. This
method helps to identify key processes and emission sources and facilitates comparisons between
processes, consumption of natural resources, pollutant generation and environmental burden. It
is important to note that lifecycle modeling typically provides only general comparisons, based
on industry-wide estimates and assumptions; it does not reflect general  equilibrium impacts, such
as effects on input markets.  The results of this type of analysis are highly dependent upon the
input data used, the variables considered, and the assumptions made. Nevertheless, within these
limitations, it can be an extremely useful tool for evaluating some of the environmental impacts
of products and processes.

       For transportation fuels, lifecycle modeling considers all steps in the production of the
fuel. This includes production of the fuel feedstock, transportation of the fuel feedstock to a
processing facility, fuel processing, and distribution of the fuel to the retail outlet.  If the analysis
considers only the finished product, it is sometimes called a 'well-to-pump' analysis; if the fuel
combustion emissions are included, it can be called a 'well-to-wheel' analysis. While both
approaches have advantages, in this work we have considered 'well-to-wheel' impacts.
However,  we are not addressing the issues of vehicle technology and energy efficiency, since we
are making the assumption that the vehicle issues will not be affected by the presence  of
renewable fuels (i.e., efficiency of combusting one Btu of renewable fuel is equal to the
efficiency of combusting one Btu of conventional fuel).

       To put this type of analysis into perspective, consider the example of gasoline. The fuel
feedstock is crude oil.  The lifecycle analysis accounts for the energy used to extract the oil from
the ground and any associated emissions, such as the natural gas that is flared at the well head.
Next you evaluate transportation of the crude oil to the refinery.  If it is domestic crude oil, it
may be delivered by pipeline and/or barge. The analysis takes into account national trends for
domestic oil transportation, and apportions energy used and emissions generated to each type of
transportation. For foreign crude oil, the energy and emissions from ocean tankers is included,
with an estimate of the average distance traveled by these tankers. Next is an estimation of the
energy use and emissions from the refinery.  Because gasoline is not the only product  produced
at the refinery, only a portion of the energy and emissions is allocated to gasoline production.
There are different methods for making this allocation, based on the value of the co-products or
an engineering assessment of the energy use and emissions from the various units in the refinery.
You then evaluate the  energy use and emissions from transporting the gasoline to market, via
pipeline and truck, based on national average distances. Finally, vehicle energy use and
emissions  are estimated.  Figure 6.1-1 illustrates this process.
                                          219

-------
         Figure 6.1-1: Lifecycle Production Process, 'Well-to-Wheel', for Gasoline
                                                          Refueling
          Grade  Crude
             ticpranspo
              via pipeline
                   Grade Refining
   Fuel
 Transport
via pipeline
                                                             Fuel
                                                             Use
       Lifecycle modeling has been a useful tool in evaluating the environmental benefits of
various alternative transportation fuels. It allows the replacement fuel to be fairly compared
against the conventional transportation fuels - gasoline and diesel fuel. There have been several
significant lifecycle analyses of transportation fuels done in the last decade. The lifecycle
analysis done for this Renewable Fuel Standard (RFS) program uses a model developed by the
Department of Energy (DOE) Argonne National Laboratory (ANL) called the Greenhouse
Gases, Regulated Emissions, and Energy Use in Transportation (GREET) model. EPA has
reviewed and modified GREET somewhat to reflect the data and assumptions appropriate for the
RFS.  These modifications are discussed further in section 6.1.2.

6.1.1   Scope of the Lifecycle Analysis

       An important step in conducting a lifecycle analysis is to define the scope of the study.
Varying results can be obtained depending on the scope identified. The scope of the analysis
includes (1) the goal (2) the system boundaries (3) what flows are considered (4) temporal
considerations and (5) modeling tools used. Each of these components is examined in the
following sections.
6.1.1.1
Goal
       The goal of this analysis is to determine the GHG emission and fossil fuel impact of the
increased use of renewable fuels.  This analysis is based on comparing future scenarios
representing an increased percentage of the overall transportation sector fuel pool coming from
renewable fuels compared to a reference case with the percentage of renewable fuels use at
current levels. This implies that our future scenarios assume renewable fuels are displacing their
petroleum based counterparts and causing less to be used. This RIA reflects increases in ethanol
production of 85% and 150% respectively from the baseline. As this analysis is compared to a
reference case we are only interested in the savings of the new or marginal renewable fuels used.
                                          220

-------
       We have evaluated the absolute savings (e.g., tons of GHG emissions) as well as
determining what percentage these absolute savings are in terms of overall transportation sector
and economy wide emissions and energy use.

6.1.1.2        System Boundaries

       The lifecycle analysis for the relevant activities identified in the GREET model is
conducted without  any regard to the geographic attributes of where emissions or energy use
occurs.  While the primary emphasis of a rulemaking analysis is typically to examine the
domestic implications of a rulemaking, the lifecycle analysis of this final rule represent global
reductions in  GHG emissions and energy use, not just those occurring in the U.S. For example,
under a full lifecycle assessment approach, the savings associated with reducing overseas crude
oil extraction and refining are included here, as are the international emissions associated with
producing imported ethanol.  This assumes that for every gallon of gasoline that's not imported
into the US, the corresponding quantity of crude oil is not extracted or processed to make this
gasoline regardless where the extraction or production takes place. This type of modeling does
not allow for  behavioral changes that may be occur, called "rebounding effect," discussed later.

       There are two important caveats to this analysis, both dealing with secondary impacts that
may result internationally due to the expanded use of renewable fuels within  the United States.
The first caveat is the emissions  associated with international land use change. Due to
decreasing corn exports some changes to international land use may occur, for example, as more
crops are planted in other regions to compensate for the decrease in crop exports from the U.S.
While the emissions associated with domestic land use change are well understood  and are
included in our lifecycle analysis, we did not include the potential impact on  international land
use and any emissions that might directly result.  Our currently modeling  capability does not
allow us to assess what international land use changes would occur or how these changes would
affect greenhouse gas emissions.  For example, we would need to know how international
cropping patterns would change  as well as farming inputs and practices that might affect
emissions assessment. The second caveat results from the assumption of reduced petroleum
imports. It is commonly presumed in economic analyses that demand for a normal  good (i.e.,
oil) will increase as price decreases. A world wide reduction of oil price that could result from
reduced U.S.  imports can reduce the cost of producing transportation fuel which in  turn would
tend to reduce the price consumers would have to pay for this fuel. To the extent fuel prices are
decreased, demand and consumption would tend to increase; this impact of reduced cost of
driving is sometimes referred to  as a "rebound effect." Such a greater consumption would
presumably result in an increase in greenhouse gas emissions as consumers would drive more.
These increased emissions would in part offset the emission benefits  otherwise accounted for this
rule61. It is important to note that GREET does not model behavioral changes that may affect
prices of relevant commodities and goods which through various feedback loops ultimately
energy use. The model does not include a general equilibrium approach that examines how a
shock (whether economic, technical or legal) affects not only the sector of interest but also other
sectors and the economy as a whole.62  While such impacts of U.S. actions are important to
  The extent to which this offset would occur would depend on sensitivity of demand to price.
                                          221

-------
understand, we have not have fully considered and quantified the rebound effects of this
renewable fuel  standard.  Nevertheless, such impacts remain an important consideration for
future analysis.

       The system boundaries for this study encompass both the renewable fuels lifecycle stages
as well as their petroleum based counterparts. Table 6.1-1 shows the lifecycle stages considered
for each fuel.

                    Table 6.1-1. Lifecycle Stages Included in Analysis
Corn Ethanol
Corn Farming
Corn Transport

Ethanol Production
Ethanol T&D
Ethanol Tailpipe
Emissions
Cellulosic Ethanol
Biomass Farming
Biomass Transport

Ethanol Production
Ethanol T&D
Ethanol Tailpipe
Emissions
Biodiesel
Soybean Farming
Soybean Transport
Soybean Crushing
Biodiesel
Production
Biodiesel T&D
Biodiesel Tailpipe
Emissions
Petroleum-Based
Gasoline
Crude Oil
Extraction
Crude Oil Transport

Refining
Gasoline T&D
Gasoline Tailpipe
Emissions
Petroleum-Based
Diesel Fuel
Crude Oil
Extraction
Crude Oil Transport

Refining
Diesel Fuel T&D
Diesel Fuel
Tailpipe Emissions
       The boundaries around each lifecycle stage include the emissions and energy use
associated with that operation as well as upstream components that feed into it. For example, the
corn farming stage includes emissions from fuel used in tractors as well as from producing and
transporting the fertilizer used in the field. Electricity production emissions are included in
almost all of the stages shown.  These components typically have the biggest  impact on the
results.  We did not include for  example, energy and emissions associated with producing the
steel and concrete used to construct the ethanol plants or petroleum refineries.

       As other lifecycle studies of renewable fuels have included an expanded set of system
boundaries, a sensitivity analysis was performed that includes the energy use  and the emissions
associated with producing farm equipment, and is described in section 6.1.2.7.

       A potentially important  system boundary affect, however, could be changes in land use.
This is particularly the case for  GHGs if new land (e.g., rainforest land) must first be cleared in
order to grow the biofuel feedstocks. This lifecycle analysis is conducted without any regard to
the geographic attributes of where emissions or energy use occurs. The benefits of this final rule
represent global reductions in GHG emissions and energy use, not just those occurring in the
U.S. For example, the savings associated with reducing overseas crude oil extraction and
refining are included here, as are the international emissions associated with producing imported
ethanol.  One exception to this is the emissions associated with international land use change.
Due to decreasing corn exports  and modest decreases in soybean exports, there may be some
62 Since GREET is not a behavioral model, it cannot assess any economic efficiency implications associated with
increased ethanol production.  Analyzing these implications would be important for future ethanol rulemakings.
                                           222

-------
additional corn and soybean acres planted internationally to meet world demand. The emissions
associated with domestic land use change are included in our lifecycle analysis but international
land use change was not as it was outside the scope of our agriculture sector analysis. However,
if emissions from international land use change were included it would lower the overall benefits
of this rule. This is an area we will continue to examine for future analysis.

6.1.1.3        Environmental Flows Considered

       One issue that has come to the forefront in the assessment of the environmental impacts
of transportation fuels relates to the effect that  the use of such fuels could have on emissions of
greenhouse gases (GHGs). The combustion of fossil fuels has been identified as a major
contributor to the increase in concentrations of atmospheric carbon dioxide (CO2) since the
beginning of the  industrialized era, as well as the build-up of trace GHGs such as methane (CH4)
and nitrous oxide (N2O).  This lifecycle analysis evaluates the impacts of increased renewable
fuel use on greenhouse gas emissions.

       The relative global warming contribution of emissions of various greenhouse gases is
dependant on their radiative forcing, atmospheric lifetime, and other considerations. For
example, on a mass basis, the radiative forcing of CH4 is much higher than that of CO2, but its
effective atmospheric residence time is much lower. The relative warming impacts of various
greenhouse gases, taking into account factors such as atmospheric lifetime and direct warming
effects, are reported on a 'CO2-equivalent' basis as global warming potentials (GWPs).  The
GWPs used in this analysis were developed by the UN Intergovernmental Panel on Climate
Change (IPCC) as listed in their Third Assessment Report63, and are shown in Table 6.1-2.

                                      Table 6.1-2.
                    Global Warming Potentials for  Greenhouse Gases
Greenhouse Gas
CO2
CH4
N2O
GWP
1
23
296
       Greenhouse gases are measured in terms of CO2-equivalent emissions, which result from
multiplying the GWP for each of the three pollutants shown in the above table by the mass of
emissions for each pollutant. The sum of impacts for CH4, N2O, and CO2, yields the total
effective GHG impact.

       The impact increased volumes of renewable fuels use has on GHG emissions (in terms of
CO2-eq.) as well as for only CO2 emissions which represent a subset of the overall GHG
emissions, is considered in this analysis. The impact increased volumes of renewable fuels use
has on fossil energy (in terms of Btus) is  also considered.  Fossil energy use includes energy
63 IPCC "Climate Change 2001: The Scientific Basis", Chapter 6; Intergovernmental Panel on Climate Change; J.T.
Houghton, Y. Ding, DJ. Griggs, M. Noguer, P.J. van der Linden, X. Dai, C.A. Johnson, and K. Maskell, eds.;
Cambridge University Press. Cambridge, U.K. 2001. http://www.grida.no/climate/ipcc_tar/wgl/index.htm


                                          223

-------
associated with coal, natural gas, and petroleum products.  Fossil energy use is strongly linked
with CC>2 and GHG emissions and is an important consideration when looking at overall
sustainability.

       Petroleum energy use is a subset of fossil energy use and is the major contributor to
overall transportation sector energy use.  Petroleum energy use is also linked to CC>2 and GHG
emissions but also has impacts on national energy concerns such as dependence on foreign
sources of petroleum. Therefore, petroleum energy was also considered separately in this
analysis and examined in terms of overall energy use, as well as in terms of petroleum imports
avoided through the increased use of renewable fuels.
6.1.1.4
Time Frame and Volumes Considered
       The results presented in this analysis represent a snapshot in time.  They represent annual
GHG and fossil fuel savings in the year considered, in this case 2012.

       Consistent with the renewable fuel volume scenarios described in Chapter 2, our analysis
of the GHG and fossil fuel consumption impacts of renewable fuel use was conducted using
three volume scenarios. The first scenario was a reference case representing 2004 renewable
fuel production levels, projected to 2012.  This scenario provided the point of comparison for the
other two scenarios. The other two renewable fuel scenarios for 2012 represented the RFS
program requirements and the volume projected by EIA.

       In both the RFS and EIA scenarios, we assumed that the biodiesel production volume
would be 0.303 billion gallons based on an EIA projection. Furthermore, the Energy Act
requires that 250 million gallons of cellulosic ethanol be produced starting in the year 2013, for
both scenarios we assume that 250 million gallons of ethanol that qualify for cellulosic credit
will be produced in 2012.  The remaining renewable fuel volumes in each scenario would be
ethanol made from corn and imports. The import volume is based on EIA's projections for the
percent of total ethanol volume supplied by imports in 2012. The total volumes for all three
scenarios are shown in Table 6.1-3.

                  Table  6.1-3. Volume Scenarios in 2012 (billion gallons)

Corn-ethanol
Cellulosic ethanol
Biodiesel
Ethanol imports
Total volume
Reference
Case
3.947
0.0
0.030
0.0
3.977
RFS Case
5.985
0.25
0.303
0.436
6.974
EIA Case
8.758
0.25
0.303
0.630
9.941
       As we are comparing against a reference case, we are only interested in the emissions and
energy savings associated with new or marginal renewable fuels production that comes on-line
after 2004 (the baseline assumed for the reference case).
                                          224

-------
6.1.1.5       Model Used

       The lifecycle model used in the evaluation of the impacts of the RFS program is the fuel-
cycle model developed by DOE's Argonne National Laboratory. For this work, EPA used the
most recent version of this model, GREET 1.7 (November 10, 2006 release). GREET, a multi-
dimensional spreadsheet model, is one of the most widely used model of this type for
transportation fuels.  It has been reviewed, used, and referenced by a wide variety of analysts,
including General Motors, National Corn Growers Association, several fuel industry
organizations, and a wide  variety of academic institutions.  It is the most comprehensive and
user-friendly model of its type. It has been under development for over 10 years, with input
from EPA, USD A, DOE laboratories, and industry representatives.  The model addresses the full
lifecycle for an exhaustive number of alternative transportation fuels and automotive
technologies.  For these reasons, EPA felt it was the best tool for evaluating the energy and
emission impacts of the RFS program.

       The GREET model has been developed to calculate per-mile energy use and emission
rates of various  combinations of vehicle technologies and fuels for both fuel cycles and total
energy cycles.  The model actually consists of three components:  GREET 1.x, which calculates
fuel cycle energy use and emissions, GREET 2.x, which calculates light-duty vehicle cycle
energy use and emissions, and GREET 3.x, which calculates heavy-duty vehicle cycle energy
use and emissions.  All discussion here refers to GREET 1.7, the most recent version of the fuel
component of GREET.

       To estimate fuel cycle energy use and emissions, GREET first estimates energy use and
emissions for a given upstream stage.  The model then combines the energy use and emissions
from all upstream stages for a fuel cycle, to estimate total upstream fuel cycle energy use and
emissions. Inputs are national-average energy usage rates, efficiencies and emission factors for
each stage.  The model calculates total energy use, fossil energy use, and emission rates for the
regulated pollutants and greenhouse gases, reported as grams per mile or grams per million Btu.
These results allow comparison of transportation fuels, based on energy use and/or emissions.

       One of the main comments we received on our lifecycle approach was that our sole
reliance on the GREET model should be avoided, given other models are available. There are
several other models that have been developed for conducting renewable fuels lifecycle analysis.
For example, researchers at the Energy and Resources Group (ERG) of the University of
California Berkeley have developed the ERG Biofuel Analysis  Meta-Model (EBAMM) and
Mark Delucchi at the Institute of Transportation Studies of the University of California Davis has
developed the Lifecycle Emissions Model (LEM).  There are also other non-fuel specific
lifecycle modeling tools that can be used to perform renewable fuel lifecycle analysis. The main
differences in these models are with input assumptions used as  described below.

       Several studies have been released recently making use  of these other models and
showing different results than we find in the analysis done  for this rule.  For example, whereas
GREET estimates a net GHG reduction of about 22% for corn ethanol compared to gasoline, the
previously cited works by Farrell et al. utilizing the EBAMM show around a 13% reduction.
While there may be small  differences in the models in terms of emissions and energy uses
                                          225

-------
associated with ancillaries (e.g., emissions to produce fertilizer, electricity, etc.) the main
difference in results is not due to model used but assumptions on scope and input data used.

       For example, most studies focus on average or current ethanol production which uses a
current mix of wet and dry mill ethanol production and use of coal and natural gas as process
energy. In contrast, we consider new or marginal ethanol production which implies a higher
portion of more efficient dry mill production and mix of process fuels.  Other studies also
typically base ethanol  and farm energy use on historic data while we are assuming a state of the
art dry milling plant and most current farming energy use data.  Assumptions concerning land
use change CC>2 emissions and agriculture related GHG emissions could also have an impact on
overall results. Other  studies also differ in the environmental flows considered.  For example,
DelucchiNNNN uses different types of greenhouse gases and GWPs compared to those used in this
analysis as shown in Table 6.1-2 to determine GHG emissions.

       Other researchers  have performed lifecycle analysis of renewable fuels not specifically
focused on GHG emissions.  One result that has been debated recently is the net  energy balance
of corn-based ethanol  fuel.  Some analysts have suggested that there is actually a negative energy
balance for corn ethanol, meaning that it takes more fossil energy to produce the ethanol than is
contained in the resulting fuel, making it an unattractive transportation fuel.  While we do not
believe this is an appropriate metric to use when examining renewable fuels, as discussed in
Section 6.2.3, it is still useful in examining the range of lifecycle results.  Two studies Pimental
(2003)64 and Patzek (2005), 65 concluded that the energy balance is negative.  . Many other
researchers, however,  have criticized that work as being based on out-dated farming and ethanol
production data, including data not normally considered in lifecycle analysis for  fuels, and not
following the standard methodology for lifecycle analysis in terms of valuing co-products.
Furthermore, several recent surveys have concluded that the energy balance is positive, although
they differ in their numerical estimates.66'67'68 Authors of the GREET model have also
concluded that the lifecycle amount of fossil energy used to produce ethanol is less than the
amount of energy in the ethanol itself. Based on our review of all the available information, and
the results of our own  analysis, we also believe that the energy balance is positive.
64 Pimentel, David "Ethanol Fuel: Energy Balance, Economics, and Environmental Impacts are Negative", Vol. 12,
No 2, 2003 International Association for Mathematical Geology, Natural Resources Research

65 Pimentel, D.; Patzek, T. "Ethanol production using corn, switchgrass, and wood; biodiesel production using
soybean and sunflower." Nat. Resour. Res. 2005, 14 (1), 65-76.

66 Hammerschlag, R. "Ethanol's Energy Return on Investment: A Survey of the Literature 1990 - Present." Environ.
Sci. Technol. 2006, 40, 1744 -1750.

67 Farrell, A., Pelvin, R., Turner, B., Joenes, A., O'Hare, M, Kammen, D., "Ethanol Can Contribute to Energy and
Environmental Goals", Science, 1/27/2006, Vol 311, 506-508.

68 Hill, I, Nelson, E., Tilman, D., Polasky, S., Tiffany, D., "Environmental, economic, and energetic costs and
benefits of biodiesel and ethanol biofuels", Proceedings of the National Academy of Sciences, 7/25/2006, Vol. 103,
No. 30, 11206-11210.


                                            226

-------
       The differences found by different studies and models used emphasize the importance of
the input data and methodology when using lifecycle analysis. It also shows how dependent this
type of analysis is on the assumptions made throughout the model. Based on differences in
scopes and input data considered between these other studies and what we defined in this
analysis, we believe the differences in results that are seen are reasonable and the values we are
obtaining from our use of the GREET model are acceptable for this analysis.

6.1.2   Modifications to GREET

       EPA chose to use GREET 1.7 to evaluate the lifecycle impacts of the RFS program.
GREET 1.7 is the most recently released version of the GREET model. However, this version of
the model  does not reflect the potential impacts on transportation fuel industries as a result of the
RFS program. In addition, for this regulation our intent was to evaluate the impact of
incremental renewable fuel production resulting from the RFS program and not a current
industry average. Therefore, EPA has modified some of the input variables and assumptions
made in the GREET model.  The renewable fuels considered in this analysis were modeled as
being produced from the following feedstocks and processes:

   -   CornEthanol:
          o  Wet Milling
                •   Mix of coal and natural gas as process fuel
          o  Dry Milling
                •   Natural gas as process fuel
                •   Coal as process fuel
                •   Biomass as process fuel

   -   CellulosicEthanol:
          o  Hybrid Poplar Feedstock
                •   Fermentation route
          o  Switchgrass Feedstock
                •   Fermentation route
          o  Corn Stover Feedstock
                •   Fermentation route
          o  Forest Waste  Feedstock
                •   Gasification route

   -   Biodiesel:
          o  Soybean Oil Feedstock
                •   Transesterification route
          o  Yellow Grease Feedstock
                •   Transesterification route

       These feedstocks and processes were primarily based on what was available in the
GREET model with some minor modifications as described below. However, there are other
pathways for producing renewable fuels not covered here, for example different feedstocks for
cellulosic ethanol production (e.g., MSW) as well as different process for the feedstocks
                                          227

-------
considered, like gasification of switchgrass and production of soybean oil diesel fuel through
hydrotreating.

       Furthermore, the lifecycle analysis used for this rulemaking is based on averages of the
different renewable fuels modeled. For example, the GHG emission and fossil energy savings
associated with increased use of corn ethanol are calculated based on a mix of process fuels,
assuming a certain projected mix of each process fuel as outlined below. While this method may
not exactly represent the reductions associated with a given gallon of renewable fuel, it is
reasonable for the purpose of this analysis which is to determine the impact of the total increased
volume of renewable fuels used.

       We recognize that different feedstocks and processes will each have unique
characteristics when it comes to lifecycle GHG emissions and energy use.  However, we
understand that other feedstocks and processes as well as differences in other parts of the
renewable fuel lifecycle will impact the savings associated with their use and this is the focus of
ongoing work at the agency.

       GREET is subject to periodic updates by ANL, each of which results in some changes to
the inputs and assumptions that form the basis for the lifecycle estimates of emissions generated
and energy consumed. These updates generally focus on those  input values for those fuels or
vehicle technologies that are the focus of ANL at the time. As a result there  are a variety of
other inputs related to ethanol and biodiesel that may not have been updated  in some time. In the
context of the analysis of the RFS and EIA scenarios, we determined that some of the GREET
input values that were either based on outdated information or did not appropriately reflect
market conditions under a renewable fuels mandate should be examined more closely, and
updated if necessary.

       Since the analysis done for the NPRM, several changes  have been made to the GREET
model, some as part of periodic updates ANL had planned and some as part of an interagency
agreement between ANL and EPA to investigate a variety of GREET input values.  A summary
of the changes is as follows:

   -  Included CC>2 emissions from corn farming lime use
   -  Updated the corn farming fertilizer use inputs
   -  Added cellulosic ethanol production from corn stover and forest waste
   -  Modeled biomass as a process fuel source in corn ethanol dry milling

       In addition to the changes above we also examined and updated other GREET input
assumptions for corn ethanol and biodiesel production.  A summary of the GREET input values
we investigated and modified is given below.  We also examined several other GREET input
values, but determined that the default GREET values should not be changed for a variety of
reasons as discussed in the following sections. These included  corn and ethanol transport
distances and modes and byproduct allocation methods.  Our investigation of these other GREET
input values are discussed more fully below.  The current GREET default factors for these other
inputs were included in the analysis for this final rule.
                                          228

-------
       We did not investigate the input values associated with the production of petroleum-
based gasoline or diesel fuel in the GREET model for this final rule.  However, the refinery
modeling discussed in Chapter 7 will provide some additional information on the process energy
requirements associated with the production of gasoline and diesel under a renewable fuels
mandate. We will use information from this refinery modeling in future analysis to determine if
any GREET input values should be changed.

       A summary of the GREET corn ethanol input values we investigated for this final rule is
given below.

6.1.2.1       Wet Mill versus Dry Mill Ethanol Plants

       As described in Chapter 1, the two basic methods for producing ethanol from corn are dry
milling and wet milling.  In the dry milling  process, the entire corn kernel is ground and
fermented to produce ethanol. The remaining components of the corn are then dried for animal
feed (dried distillers grains with solubles, or DDGS). In the wet milling process, the corn is
soaked to separate the starch, used to make  ethanol, from the other components of the corn
kernel. Wet milling is more complicated and expensive than dry milling, but it produces more
valuable products (ethanol plus corn syrup,  corn oil, and corn gluten meal and feeds). The
majority of ethanol plants in the United States are dry mill plants, which produce ethanol more
simply and efficiently.

       While other lifecycle models often base the mix of wet and dry milling on existing plants,
for this analysis, we are only interested in marginal ethanol production. We expect most new
ethanol plants will be dry mill operations. That has been the trend in the last few years as the
demand for ethanol has grown, and our analysis of ethanol plants under construction and planned
for the near future has verified this. Our analysis of production plans, as outlined in  Chapter 1,
indicates that essentially all new ethanol production will be from dry mill plants (99%).

6.1.2.2       Coal versus Natural Gas in Ethanol Plants

       The type of fuel used within the ethanol plant for process energy to power the various
components that are used in ethanol production (dryers, grinders, heating, etc.) can vary among
ethanol plants. The type of fuel used has an impact on the energy usage, efficiency,  and
emissions of the plant, and is primarily determined by economics. Most new dry mill plants built
in the last few years have used natural gas.  However, some new plants are using coal. For these
cases, EPA is promoting the use of combined heat and power, or cogeneration, in ethanol plants
to improve plant energy-efficiency and to reduce air emissions.  This technology, in the face of
increasing natural gas prices, may make coal a more attractive energy source for new ethanol
plants.

       GREET default factors represent the average percentage of fuel use for the entire
industry, and may not reflect the recent growth in the industry.  Therefore,  we based our fuel mix
assumptions on the review of plants under construction and those planned for the near future
outlined in Chapter 1. Our analysis indicates that coal will be used as process fuel for
approximately 14% of the new dry mill under construction and planned ethanol production
                                          229

-------
volume capacity. This is the value we used in GREET for our analysis of dry milling ethanol
production fuel mix.

       As opposed to typical dry mill plants, corn wet mill ethanol plants can use a mix of
process fuel sources at the same plant. For the 1% of additional ethanol production from wet
mills, the GREET model defaults of 40% coal and 60% natural gas process fuel was used in this
analysis.

       As described below, the ethanol production stage of the lifecycle typically represents the
stage where the largest amount of fossil fuel energy is consumed and where the impact on
lifecycle emissions is the greatest.  Therefore, the type of process fuel used in ethanol production
will have a significant impact on the fuel's lifecycle GHG  results.  For example, our analysis
indicates that ethanol produced in a coal fired dry mill plant would not have any GHG benefits as
compared to petroleum gasoline. Given that the relative prices of natural gas and coal could
change over time, and thus change  the percentage of each used in ethanol production, our
analysis of fuels used in plants under construction and those planned for the future would need to
be reevaluated for future work.
6.1.2.3
Ethanol Plant Process Efficiency
       For the corn-to-ethanol fuel cycle, the largest amount of fossil fuel energy consumed
occurs at the ethanol production plant. The energy use at a dry mill plant using natural gas was
based on the model developed by USD A which was documented in a peer-reviewed journal
paper on cost modeling of the dry-grind corn ethanol process.0000  This model was modified by
EPA for use in the cost analysis of this rulemaking described in Chapter 7.  GREET inputs are
total energy use per gallon of ethanol produced.  The USDA model predicts the annual thermal
(natural gas) and electricity demand shown in Table 6.1-4.

                                      Table 6.1-4.
                     Annual Energy Use at Dry Mill Ethanol Plant
Energy Input
Purchased Electricity (MWh/yr.)
Natural Gas (mmBtu/yr.)
Output
Ethanol (mmgal/yr.)
Value
41,308
1,617,094

50
       Electricity energy use was converted from MWh to Btu based on a conversion of 3,410
btu/kWh.  The primary energy used to produce electricity is accounted for in the GREET model.
Table 6.1-5 shows the GREET input used for natural gas process fuel dry milling plants in this
analysis.
                                          230

-------
                                      Table 6.1-5.
                            GREET Inputs for Corn Ethanol
                            Natural Gas Dry Mill Energy Use
Total Energy Use (mmBtu/gal.)
% electricity
35,159
8.0%
       Energy requirements for a coal fired ethanol plant are different from a natural gas fired
plant. Typically coal boilers are slightly less efficient than natural gas boilers. Furthermore
additional electricity is required for coal storage and handling as compared to natural gas.
Additionally a large portion of the energy at an ethanol plant is due to drying the DDGS. A
natural gas plant utilizes natural gas driers for this process while a coal fired plant would use
steam dryers, the efficiency loss of converting  coal to steam represents additional thermal energy
required at a coal fired plant vs. a natural gas one.

       Most other lifecycle models assume the same energy efficiency for both coal and natural
gas ethanol plants, however, for this analysis, it was assumed that a coal plant would require
15%69 more electricity demand due to coal handling and have a 13% increase in thermal demand
for steam dryers as compared to the natural gas fueled plant.  The increase in thermal demand
was based on breaking out the drying energy in the USDA process model and assuming the same
amount of energy would be produced by 78% efficient coal boilers. Table 6.1-6 shows the
GREET input used for coal process fuel dry milling plants in this analysis.

                                      Table 6.1-6.
                            GREET Inputs for Corn Ethanol
                                Coal Dry Mill Energy Use
Total Energy Use (mmBtu/gal.)
% electricity
40,079
8.1%
       The Energy Act also allows ethanol made from non-cellulosic feedstocks to receive
cellulosic ethanol production volume credit if 90 percent of the process energy used to operate
the facility is derived from a renewable source.  In the context of our cost analysis, we have
assumed that 250 million gallons of corn ethanol will be produced using 90 percent or more
biomass energy and receive the cellulosic ethanol volume credit.  Further discussion of this issue
can be found in Chapter 1.

       For the lifecycle analysis we considered the case where a corn ethanol dry mill plant
utilized biomass as a fuel source. For this case the same amount of fuel and purchased electricity
energy per gallon as a coal powered plant was assumed.  This assumption is based on the
biomass plant having more fuel handling than a natural gas plant and producing steam for DDGS
drying.
69 Baseline Energy Consumption Estimates for Natural Gas and Coal-based Ethanol Plants - The Potential Impact of
Combined Heat and Power (CHP), Prepared for: U.S. Environmental Protection Agency Combined Heat & Power
Partnership, Prepared by: Energy and Environmental Analysis, Inc., July 2006.


                                          231

-------
       As discussed in section 6.2.3, CC>2 emissions from combustion of biomass are not
assumed to increase net atmospheric CC>2 levels. Therefore, CC>2 emissions from biomass
combustion as a process fuel source are not included in the lifecycle GHG inventory of the
ethanol plant.  The fossil energy use and GHG emissions from producing the electricity used at
the plant are included.

       For the 1% of corn ethanol produced from wet milling, the GREET process energy use
default of 49,950 Btu/gallon of ethanol produced by the wet milling process was used in the
analysis.

6.1.2.4       Corn Transport Distances

       Corn transport distances selected for use in this analysis are 100 miles round trip. Corn
used in the ethanol production process is assumed to travel from corn fields to ethanol
production facilities in a two-step process; first, corn is transported from outlaying farms to
centrally-located collection facilities, such as county elevators. Second, this corn is transported
from the collection facilities to the ethanol production facilities.  The first leg of the corn
transport process is assumed to be a 20-mile round trip and the second leg is assumed to be an
80-mile round trip.  These assumptions coincide with those used in GREETPPPP Version 1.7 and
GREET Version 1.5.

       Corn transport data is limited, however; Graboski70 found that the average one-way
hauling distance for corn from fields to county elevators was 7.5 miles and from county elevators
to ethanol processing facilities was 49.7 miles for an effective average round-trip corn transport
distance of 74.6 miles.  Similarly, Gervais and BaumelQQQQ found that average one-way corn
transport distances for the 1994-1995 Iowa growing season was 37.2 miles for semi-trucks
(35.8%), 4.9 miles for wagons (33.3%), and 9.1 miles  for single and tandem axel vehicles
(30.9%). Several Minnesota corn mills indicated that the maximum radius of supply for their
mills was 65 to 80 miles (values apparently cited in the same study).

       The available data on corn transport distances does not provide a clear indication that the
default values in GREET are unreasonable. Therefore, we retained the GREET default values
for our analysis. This assumes that the land use pattern (where corn is planted) and the plant
location decisions by ethanol plants will not change significantly.  We believe this is reasonable
for the fuel volumes  considered. This is an area we will continue to examine for future analysis.
70 The authors assume that the corn payload weight is equal to the transport vehicle weight, that the vehicle returns
empty, and the effective average round-trip vehicle distance can be estimated as being one and a half times the one-
way travel distance (1.5 times 49.7 miles = 74.6 miles); Graboski, 2002, Fossil Energy Use in the Manufacture of
Com Ethanol, Colorado School of Mines, (Prepared for the National Corn Growers Association).


                                           232

-------
6.1.2.5
Ethanol Transportation Distances and Modes
       The default values in GREET for ethanol transportation and modes are shown in Table
6.1-7.  These values correspond to numbers in a USDA study on the energy balance of corn
ethanol.RRRR

                Table 6.1-7.  GREET Ethanol Transportation Input Data
Mode

Rail
Barge
Truck
Plant to Terminal
%
40%
40%
20%
Distance (miles)
800
520
80
Terminal to Station
%
0%
0%
100%
Distance (miles)


30
       The GREET default values are consistent with the analysis we performed on ethanol
distribution infrastructure.  Chapter 1  of this document discusses current ethanol transportation
and distribution and indicates that if ethanol facilities are located within 100-200 miles of a
terminal, trucking is preferred.  Rail and barge are used for longer distances. Pipelines are not
currently used to transport ethanol and are not projected to play a role in ethanol transport in the
future time frame considered.

       We also discuss in Chapter 1 future ethanol transportation and distribution needs based
on the increased amounts of renewable fuels used as a result of this rule.  We concluded that
most new ethanol capacity will not have river access. In addition, at least one new ethanol plant
slated for production that does have river access is planning to move its ethanol to market via rail
so most new ethanol freight volumes will be handled by rail and that ethanol transport by inland
waterway will remain constant.

       A recent USDA Cost of Ethanol Production report also provides information on ethanol
distribution distances and modes.ssss  The report includes 2002 data from a survey of 21 dry mill
ethanol plants.  The survey collected data on modes and distances traveled for ethanol transport
from the facilities.  The report concluded that 46 percent of the ethanol produced at the surveyed
plants in 2002 was shipped by truck an average one way distance of 93 miles, with a range of 30
to 250 miles.  The remaining 54 percent of ethanol produced was shipped by rail an average one
way distance  of 1,163 miles, with a range of 800 to 2,500 miles. However, this data is for a
subset of existing plants where, for example, there is no barge transportation listed, and also does
not take into account the increased demand for ethanol projected by this rule.
                                          233

-------
       Comparing the GREET default values to these other sources indicates that the GREET
defaults values for percent of ethanol transported by rail may be low. However, due to lack of
precise data on future ethanol transportation by mode, we concluded that the current GREET
default values for percent of ethanol transported by mode are appropriate for the RFS analysis.

       The GREET default values for miles shipped by mode fall within the range of values
listed in the USDA survey data of existing plants. The USDA survey data indicate higher than
average transportation distances; however the data is not comprehensive enough, only
representing a small fraction of total and projected ethanol production capacity, thus not
warranting a change to the default GREET values.  Therefore, the default values shown in Table
6.1-7 were used in this analysis. This is an area we will continue to examine for future analysis.

6.1.2.6       Biodiesel Transportation Distances and Modes

       The default values in GREET for biodiesel transportation and modes are shown in Table
6.1-8.
Table 6.1-8. GREET Biodiesel Trans
Mode

Barge
Pipeline
Rail
Truck
Plant to Terminal
%
8%
63%
29%
0%
Distance (miles)
520
400
800

portation Input Data
Terminal to Station
%
0%
0%
0%
100%
Distance (miles)



30
       The GREET default assumptions for mode of biodiesel transportation are not consistent
with the analysis we performed on biodiesel distribution infrastructure. The distribution
infrastructure discussion in Chapter 1 of this document indicates pipelines are not currently used
to transport biodiesel and are not projected to play a role in biodiesel transport in the future time
frame considered.

       Therefore, GREET default factors for biodiesel transportation from plant to terminal were
modified to remove pipeline transport.  The percent of biodiesel shipped by barge and rail were
increased in the same proportion as the current percentage split. The result was 22% of biodiesel
shipped by barge and 78% shipped by rail.  The GREET default distances for biodiesel rail and
                                          234

-------
barge transport as well as terminal to station assumptions are consistent with ethanol
transportation and distribution assumptions and were used in this analysis.
6.1.2.7
Corn Yield and Related Inputs
       GREET includes a collection of energy use and material inputs to corn farming per
bushel (bu) of corn produced.  Several corn farming input data parameters and default values
were updated from the version of GREET used for the NPRM to the version used in the FRM
analysis. The current GREET corn farming input data default values are shown in Table 6.1-9.

                     Table 6.1-9. GREET Corn Farming Input Data
Input Parameter
Energy Use for Corn Farming
- Energy use from diesel fuel
- Energy use from gasoline
- Energy use from natural gas
- Energy use from LPG
- Energy use from purchased electricity

Nitrogen Fertilizer (as N)
Phosphate Fertilizer (as P2Os)
Potash Fertilizer (as K2O)
Lime (as CaCOs)

Herbicide Use:
Insecticide Use:
Default Value
22,500 Btu/bu
38.3%
12.3%
21.5%
18.8%
9.0%

420 g/bu
149 g/bu
174 g/bu
1,202 g/bu

8.1 g/bu
0.68 g/bu
       The default GREET input values for corn farming shown in Table 6.1-9 are based in part
on farm energy use and material inputs per acre divided by an assumed corn yield in bu/acre.
Therefore, while corn yield is not a direct input in GREET, it is a critical part of the calculation
of corn energy and material input requirements.  Although corn yields have been generally rising
over time, see Figure 6.1-2, the annual variation  is volatile.
                                          235

-------
                        Figure 6.1-2. U.S. Average Corn Yield
                                                               TTTT
                    180 n
                     60
                       1990
1995
2000
2005
2010
       We examined data on farm energy use, material input, and yield data to determine if the
GREET default values needed to be updated. The lifecycle modeling conducted for the RFS
program is based on future predictions. Unfortunately, no good projections of future energy use
associated with corn farming are available. USD A does list projections for corn yield. The 2012
projected U.S. average corn yield is 158.5 bu/acre.uuuu  Historic data on corn farming energy
use is available from the following USDA information sources.

    •   The USDA Agricultural Resource Management Survey (ARMS) provides data from
       selected States on fuel, electricity, natural gas, and seed corn used per acre on the farm
       and activities of moving farm products to initial storage facilities.
    •   The USDA National Agricultural Statistics Service (NASS) produces annual data on crop
       production including yields per acre and total production of corn by state.

       USDA NASS data on corn yields and production values are provided annually.
However, the three most recent years of the ARMS data and specifically the costs-of-production
portion of the survey dedicated to corn are 1991, 1996, and 200171.  Table 6.1-10 lists corn
farming energy input data for the three years of the ARMS study.
71 Use of historic farming energy use may not be representative of current practice. Higher energy prices relative to
the years considered here could lead to farmers adopting practices that lower overall energy use.
                                            236

-------
                      Table 6.1-10. Farm Energy Use Data per Acre
Input

Seed
Energy:
- Diesel
- Gasoline
-LPG
- Electricity
- Natural Gas
Total Energy Use
Units

bu/acre

Gallons/acre
Gallons/acre
Gallons/acre
kWh/acre
Cubic ft/acre
mmBtu/acre
9-State Weighted Average Values
1991
1.51

7.81
3.42
3.86
32.72
284.73
2.12
1996a
1.50

9.80
3.07
7.25
79.38
208.12
2.71
2001
1.69

6.40
1.65
5.10
38.22
207.09
1.78
3 Yr. Avg.
1.57

8.01
2.71
5.41
50.11
233.31
2.20
"High energy use in the 1996 survey is due to increased corn drying requirements. See the discussion below.
       Although USDA corn data is available for every state that produces corn, the data
documented in Table 6.1-10 is for nine major corn producing States: Illinois, Indiana, Iowa,
Minnesota, Nebraska, Ohio, Michigan, South Dakota, and Wisconsin. In 2005, these nine States
accounted for 80 percent of U.S. corn production. In 2001 these nine States represented 92
percent of U.S. ethanol production, and based on our analysis outlined in Chapter 1 are projected
to represent 82 percent of ethanol production in 2012.  The data in Table 6.1-10 are weighted
based on corn production data for each of the nine States from the NASS. The total energy use
values listed in Table 6.1-10 were calculated by converting fuel use to Btu based on the lower
heating values of the fuels as listed in the GREET model. These estimates may be biased
downward if the corn production attributable to the incremental increase in ethanol production
will occur on less productive land than was used in the 1991-2001 period, when corn prices were
lower than they are projected to be in this analysis.  Also, as corn production expands due to
expanded ethanol production, it may increasingly take place in dryer climates that may increase
irrigation demand and result in different yields. This is an area we will continue to examine for
future analysis.

       The ARMS surveys include information on energy use and also on dollars spent by
farmers on custom work.  This custom work includes farmers contracting outside services  for
corn drying, planting, fertilizing and harvesting.  The cost of custom work includes machine
overhead, fuel charges, and labor costs.  Therefore, there is some energy use associated with the
dollars spent on fuel used in custom work. It was assumed that 10% of custom work cost was
spent on  fuelwvv.  This fuel cost was assumed to be split between LPG and diesel fuel in the
same percentage as reported energy use for each state. Cost was converted to gallons based on
price paid by farmers for LPG and diesel fuel in each of the survey yearswwww.  Custom work
energy use is included in Table 6.1-10.

       It can be seen from Table 6.1-10 that there is substantial variation in the three years of
energy use survey data. Several factors can influence corn farming energy use. For example, it
was reported that 1991 was a dry year, lowering the moisture content of the  corn crop and  thus
requiring less energy to dry the corn, whereas the 1996 crop  was reported to have a higher
moisture content and thus require more energy  to dry resulting in the high energy use values for
                                          237

-------
1996. Farm diesel use is also dependent on tillage type and soil conditions, wetter soil requiring
more diesel use, and decreased tillage requirements (e.g., no till) reducing diesel use.xxxx

       To project corn farming fuel use in 2012, the average energy use from the three years of
survey data were taken, in terms of energy per acre.  As energy use is somewhat weather related
and it is impossible to confidently predict future conditions, it was felt that the three years of
historic data represented a good mix of high and low energy use years.  The average energy use
in terms of Btu/acre was divided by the projected corn yield in 2012 of 156.9 bu/acre. This is the
USDA projected corn yield adjusted to account for seed corn energy use as shown in Table 6.1-
10.  The seed use shown in Table 6.1-10 accounts for seed corn energy use.  We assumed that
growing seed corn requires 4.7 times the energy and material inputs to grow than cornYYYY'zzzz.
The result was 14,036 Btu of energy needed to produce a bushel of corn, which was used in
GREET for this analysis.

       The GREET default values for corn farming material inputs were updated from the
values in the NPRM version. GREET defaults were based on historic data provided from the
following USDA sources.

   •   The USDA National Agricultural Statistics Service (NASS) produces annual reports
       listing quantities of fertilizers and chemicals used per acre of corn.
   •   The USDA Economic Research Service (ERS) produces an Agricultural Resources and
       Environmental Indicators report that has data on lime used per acre of corn.

       The USDA sources provide average material use data per harvested acre of corn.  The
GREET defaults are based on the assumption that material input use per acre will be flat from
2005 into the future.  The 2005 values are based on a three year average of 2003 through 2005
data. Data on inputs per acre are divided by projected corn yields to get GREET defaults in
terms of g/bushel of corn. While these values are felt to be reasonable to be used in this analysis,
the agency cautions that these  estimates are based on the historical record while the incremental
corn production attributable to expanded ethanol production may occur on less productive land
than was used historically.  As a result, these estimates may be biased downward, resulting in
over-estimates of ethanol displacement indices.

       Another potential input to corn farming is the energy and emissions associated with
producing farm equipment.  As described in Section 6.1.1.2, this input is considered outside the
system boundaries of our lifecycle analysis. However, the latest version of GREET has an
option to include energy use and emissions associated with producing farm equipment in the
corn ethanol lifecycle results.  We performed a sensitivity analysis on expanding the corn
production system to include farm  equipment production to determine the impact it has on the
overall results of our analysis.

       It was found that including farm equipment production energy use and emissions
increases ethanol lifecycle energy use and GHG emissions and decreases the corn ethanol
displacement index by approximately 1 percent. Furthermore, to be consistent in the modeling if
system boundaries are expanded to include production of farming equipment they should also be
expanded to include producing other material inputs to both the ethanol and petroleum lifecycles.
                                          238

-------
For example, this expansion of system boundaries would include the energy use and emissions
associated with producing concrete and steel used in the petroleum refinery72. The net effect of
this would be a slight increase in both the ethanol and petroleum fuel lifecycle results and a
smaller or negligible effect on the comparison of the two.

       The corn farming material and energy use used in the lifecycle analysis is based on
producing and average bushel of corn. There are differences associated with variations in corn
yield, inputs required for existing land vs. land converted to crops, etc.  Furthermore, there are
ripple effects associated with increased corn used for ethanol that could have GHG emission
implications, ranging from changes in manure management to the acres of rice grown. One such
effect is CO2 associated with land use change which is examined in the following section. Other
effects and variations in corn farming will be examined as part of future analysis.

6.1.2.8        CO2 from Land Use Change

       Farming practices could potentially release carbon stored in soil as CO2 emissions. If
non-cropland (e.g., pastureland, Conservation Reserve Program (CRP)  land) is converted to crop
production, carbon sequestered in the soil and existing cover could be released.  The agricultural
sector modeling work done for this rulemaking examined the issue of land use change due to
increases in renewable fuel production and use.  The agricultural sector modeling results indicate
that, compared to the 2012 Reference Case, approximately two and a half million acres will
come out of CRP land as a result of increased renewable fuel production.  Not all of these two
million acres will go directly into corn production used to produce ethanol.  However, the entire
amount of CO2 emissions from the CRP land use change is attributable to the increased amount
of ethanol produced, as without the increased demand for corn there would be no change in CRP
land.  The agricultural modeling results also indicated a reduction in U.S. corn exports and a
modest decline in U.S. soybean exports which could impact crop production in other countries.
However, we did not consider impacts on non-U.S. land use that might result from decrease in
U.S. exports of corn and soybeans.

       The GREET model has a default factor for CO2 from land use change that was included
in the NPRM analysis.  This factor was updated based  on the results of the agricultural sector
modeling mentioned above and included in the final rulemaking lifecycle analysis. The CO2
emissions from land use change used in the final rulemaking represent approximately 1% of total
corn ethanol lifecycle GHG emissions. However, this value could be more significant if
increased amounts  of renewable fuels are used.

       The issue of CO2 emissions from land use change associated with converting forest or
CRP land into crop production for use in producing renewable fuels is an important factor to
consider when determining the overall sustainability of renewable fuel use.  While the analysis
described above  is  indicating that this rulemaking will  not cause a significant change in land use,
this is an area we will continue to research for any future analysis.
72 The expansion of system boundaries would apply to existing refineries as ethanol is assumed to replace gasoline
from existing production.


                                          239

-------
6.1.2.9        Ethanol Production Yield

       Modern ethanol plants are now able to produce more than 2.7 gallons of ethanol per
bushel of corn compared with less than 2.4 gallons of ethanol per bushel of corn in 1980.  The
development of new enzymes continues to increase the potential ethanol yield. We used a value
of 2.7173 gal/bu in our analysis, which may underestimate actual future yields.  However, this
value is consistent with the ethanol  model developed by USDA described in Section 6.1.2.2 and
was used in the cost modeling of corn ethanol discussed in Chapter 7.

6.1.2.10       Byproduct Allocation

       There are a number of by-products made during the production of ethanol. In lifecycle
analyses, the energy consumed and emissions generated by an ethanol plant must be allocated
not only to ethanol, but also to each of the by-products. There are a number of methods that can
be used to estimate by-product allocations.  These include methods based on the economic value
of each by-product, or on energy usage, based on engineering analysis of the actual processes
related to each product. The method preferred by EPA is called the displacement method. This
method most accurately accounts for these by-products by  calculating the lifecycle emissions of
the products that will be displaced by them.  In this method the lifecycle emissions of the
displaced product are calculated and subtracted from the ethanol lifecycle.  The ethanol receives
a credit for the lifecycle emissions of whatever product is displaced, since a quantity of that
product is no longer needed and is displaced by the ethanol by-products.

       For example, the DDGS  produced by an ethanol dry mill plant is a replacement for corn
and soybean animal feed. We based the amount of DDGS  produced by an ethanol dry mill plant
on the USDA model used in the cost analysis work of this rulemaking. That model predicted
6.21  dry Ib. of DDGS per gallon of ethanol produced. As per the agricultural sector modeling
done for this rulemaking,  we assumed that this DDGS displaces 50% corn and 50% soybean
meal on a mass basis.  So the lifecycle emissions of producing 3.1 Ib. of corn and 3.1  Ib. of
soybean meal were calculated and subtracted from the lifecycle emissions associated  with
producing a  gallon of ethanol.

       By-products from the ethanol wet milling process include corn gluten meal and corn
gluten feed that are assumed to displace corn production, as well as corn oil that is assumed to
displace soybean oil.  Ethanol produced from cellulosic feedstock through the fermentation route
is assumed to produce excess electricity as a by-product, from onsite combustion of lignin.  This
excess electricity is assumed to displace electricity from the grid. The fermentation process used
to produce ethanol in corn wet and dry milling and cellulosic ethanol production also produces
CC>2  as a by-product.  This CC>2 could be sold to an organization that specializes in cleaning and
pressurizing it for use in the food industry for example to carbonate beverages, to manufacture
dry ice, and to flash freeze meat.  While CC>2 could potentially displace other sources of CC>2
production, this was not considered in our analysis and no value was associated with this CO2
co-product.
 1 All yield values presented represent pure ethanol production (i.e. no denaturant).

                                          240

-------
       The displacement method for by-product allocation is the default for the GREET model.
EPA supports that approach and continues to use that method in this analysis. However, other
researchers have used different allocation methods in their ethanol fuel cycle studies.  We
evaluated one of these other methods used by USD A in a recent ethanol energy balance
reportAAAAA to determine the impact this assumption has on the overall results of the analysis.
The method used by USDA was to split the energy use and emissions of corn agriculture and
ethanol production between the ethanol and co-products. The lifecycle analysis results were then
based on only the ethanol portion. A process simulation was used to allocate the energy used in
the ethanol plant to ethanol and by-products. Using this approach they determined that on
average 59 and 64 percent of the energy used in dry and wet mills respectively is used to produce
ethanol.  The remaining energy is used for the production of by-products. Therefore,  for dry mill
ethanol production only 59 percent of the plant energy use and associated emissions were
allocated to the ethanol lifecycle. Corn production energy use and emissions were allocated
based on the starch content of the corn, assumed to be 66 percent of corn kernel weight. So, only
66 percent of the energy and emissions used to produce corn were allocated to the ethanol
lifecycle.

       Use of the process energy based  allocation method reduces ethanol lifecycle energy use
and GHG emissions by approximately 30 percent compared to the displacement allocation
approach. This indicates that ethanol lifecycle analysis  results are extremely sensitive to the
choice  of allocation method used. However, as mentioned above, EPA feels that the
displacement allocation method is the most reasonable and is the preferred method to use. This
decision is supported by international lifecycle assessment standards which indicate that
whenever possible  the product system should be expanded to include the additional functions
related to the co-productsBBBBB.

6.1.2.11       Biodiesel Production

       Two scenarios for biodiesel production were considered, one utilizing soybean oil as a
feedstock and one using yellow grease.

       For the soybean oil scenario, the energy use and inputs for the biodiesel production
process were based on a model developed by NREL and used by EPA in the cost modeling of
soybean oil biodiesel, as discussed in Chapter 7.

       The GREET model does not have a specific case of biodiesel production from yellow
grease. Therefore, as a surrogate we used the soybean oil based model with several adjustments.
For the yellow grease case, no soybean agriculture emissions or energy use was included.
Soybean crushing was still included as a surrogate for yellow grease processing (purification,
water removal, etc.).  Also, due to additional processing requirements, the energy use associated
with producing biodiesel from yellow grease is higher than for soybean oil biodiesel production.
As per the cost modeling of yellow grease biodiesel  discussed in Chapter 7, the energy use for
yellow grease biodiesel production was assumed to be 1.72 times the energy used for soybean oil
biodiesel.
                                          241

-------
       The biodiesel lifecycle results were based on a 50% / 50% split between soybean oil and
yellow grease biodiesel production based on EIA's AEO 2006 projections for biodiesel produced
from the different feedstocks.
6.2    Methodology

       As outlined in the scoping discussion, the goals of this analysis are to both examine the
total GHG and fossil fuel reductions of increased renewable fuel use in absolute tons and gallons
and to compare these reductions to the U.S. transportation sector and nationwide GHG emissions
and fossil fuel use.  The output of the GREET model can be used directly to calculate tons of
GHG and gallons of petroleum reduced.  However, these results are not entirely consistent with
transportation sector and nationwide emissions inventories which are based on slightly different
assumptions concerning fuel heating values and carbon content. As a result we could not use
GREET directly to estimate the nationwide impacts of replacing some gasoline and diesel with
renewable fuels.

       To be consistent between our modeling of savings and overall sector inventories, we used
GREET instead to generate comparisons between renewable fuels and the petroleum-based fuels
that they displace.  These comparisons allowed us to develop displacement indexes which
represent the percent of lifecycle GHGs or fossil fuel reduced when a Btu of renewable fuel
replaces a Btu of gasoline or diesel. In this way GREET was used to generate percent reductions
and not absolute values.  These percent reductions or displacement values were then applied to
the same gasoline and diesel fuel inventories used to generate transportation sector and
nationwide inventories. This ensured that savings and sector wide inventories in terms of
absolute values were calculated in a consistent manner.

       In order to estimate the impacts of increased use of renewable fuels on fossil energy and
greenhouse gases, we first determined how much gasoline and diesel would be replaced as  a
result of this rule. We then combined lifecycle percent reductions from GREET with lifecycle
inventories and petroleum consumption values for gasoline and diesel fuel use to get the amounts
of fossil energy and greenhouse gases reduced. For example, to estimate the impact of corn-
ethanol use on GHGs, these factors were combined in the following way:
                                          242

-------
                     GHG,corn ethanol
                                     corn ethanol X
                                                            GHG,corn ethanol
where:
SoHG,c
      orn ethanol
J^corn ethanol


J-^ '-'gasoline
DI
   GHG,corn ethanol
=    Lifecycle GHG emission reduction over the reference case associated with
     use of corn ethanol (million metric tons of GHG)

=    Amount of gasoline replaced by corn ethanol on an energy basis (Btu)

=    Lifecycle emissions associated with gasoline use (million metric tons of
     GHG per Btu of gasoline)

=    Displacement Index for GHGs and corn ethanol, representing the percent
     reduction in gasoline lifecycle GHG emissions which occurs when a Btu
     of gasoline is replaced by a Btu of corn ethanol
       Variations of the above equation were also generated for impacts on all four endpoints of
interest (fossil fuel consumption, petroleum consumption emissions of CC>2, and emissions of
GHGs) as well as all three renewable fuels examined (corn-ethanol, cellulosic ethanol, and
biodiesel). These values are then compared to the total U.S. transportation sector and nationwide
inventories of fossil energy and greenhouse gases to get the overall impacts of the rule.

       In this regard, the impact on overall transportation sector GHG emissions due to the
increased use of renewable fuels can be described mathematically as follows:

TSeCtOro/0;GHG   =    SGHG,corn ethanol +  SoHG,cell ethanol + SoHG,biodiesel
                                   TSectorGHG
where:
TSectoro/0jGHG
 ^GHG,corn ethanol
SoHG,cell ethanol    ~
 ^GHG,biodiesel
TSectorGHG
                     Percent reduction in overall transportation sector GHG emissions resulting
                     from the use of renewable fuels (%)
                     Lifecycle GHG emission reduction over the reference case associated with
                     use of corn ethanol (million metric tons of GHG)

                     Lifecycle GHG emission reduction over the reference case associated with
                     use of cellulosic ethanol (million metric tons of GHG)

                     Lifecycle GHG emission reduction over the reference case associated with
                     use of biodiesel (million metric tons of GHG)

                     Overall transportation sector GHG emissions in 2012 (million metric tons
                     of GHG)
                                            243

-------
       We used the same approach to estimate fossil energy, petroleum energy, and CC>2
reductions in the transportation sector. We also used the same approach to estimate nationwide
reductions.

       Section 6.2.1 describes how we estimated the amount of gasoline and diesel fuel replaced
as modeled for this rule.  Section 6.2.2 describes the lifecycle emissions and energy associated
with gasoline and diesel fuel use.  In Section 6.2.3 below, we outline how we generated
displacement indexes using GREET.  Section 6.2.4 outlines how we developed the overall
transportation sector and nationwide fossil energy and greenhouse gas emissions.

6.2.1  Modeling Scenarios

       In general, the volume fraction (R) in the equation above represents the amount of
conventional fuel no longer consumed - that is, displaced - as a result of the use of the
replacement renewable fuel.  Thus R represents the incremental amount of renewable fuel used
under each of our renewable fuel volume  scenarios, in units of Btu. We make  the assumption
that vehicle energy efficiency will not be  affected by the presence of renewable fuels (i.e.,
efficiency of combusting one Btu of ethanol is  equal to the efficiency of combusting one Btu of
gasoline).

       As described in Section 6.1.1.4, our analysis of the GHG and fossil fuel consumption
impacts of renewable fuel use was  conducted using three volume scenarios. The total  volumes
for all three scenarios are shown in Table 6.1-3. For the purposes of calculating the R values, we
assumed the ethanol volumes shown in Table 6.1-3 are 5% denatured, and the  ethanol  volumes
were adjusted down to represent pure (100%) ethanol.  The adjusted volumes were then
converted to total Btu using the appropriate volumetric energy content values (76,000  Btu/gal for
ethanol, and 118,000 Btu/gal for biodiesel).

       Since the impacts of increased renewable  fuel use were measured relative to the 2012
reference case, the value of R actually represented the incremental amount of renewable  fuel
between the  reference case and each of the two other scenarios.  The results are shown in Table
6.2-1.  The results shown in Table 6.2-1 are direct reductions in fuel use and do not represent
lifecycle savings.
                                          244

-------
                                      Table 6.2-1.
               Direct Conventional Fuel Replaced in 2012 (quadrillion Btu)

Gasoline Replaced by Corn Ethanol
Gasoline Replaced by Cellulosic Ethanol
Diesel Fuel Replaced by Biodiesel
Gasoline Replaced by Ethanol Imports
Total Energy
RFS Case
0.147
0.018
0.032
0.031
0.229
EIA Case
0.347
0.018
0.032
0.045
0.443
6.2.2   Lifecycle Impacts of Conventional Fuel Use

       In order to determine the lifecycle impact that increased renewable fuel volumes may
have on any particular endpoint (fossil fuel consumption or emissions of GHGs), we also needed
to know the conventional fuel inventory on a lifecycle basis. Since available sources of GHG
emissions are provided on a direct rather than a lifecycle basis, we converted these direct
emission and  energy estimates into their lifecycle counterparts.

       To do this, we used GREET to develop multiplicative factors for converting direct
(vehicle-based) emissions of GHGs, or direct (vehicle-based) consumption of petroleum, into
full lifecycle factors.  GREET output was used to generate the conversion factors shown in Table
6.2-2.
                                      Table 6.2-2.
            Direct (wheel only) Conversion Factors to Well-to-Wheel (lifecycle)
                                Emissions or Energy Use

Petroleum
Fossil fuel
GHG
C02
Gasoline
1.11
1.22
1.26
1.23
Diesel
1.10
1.21
1.25
1.21
                                          245

-------
       The factors in Table 6.2-2 were applied to gasoline and diesel fuel inventories of
emissions or energy consumption at the consumer level (i.e. direct emissions or energy) to
convert them into alternative inventories representing full lifecycle contributions.

       The direct petroleum energy for gasoline and diesel fuel is just the energy content of the
fuels used. Consistent with U.S. EPA National Inventory calculations00000, we converted
energy use values for gasoline and diesel fuel to direct CC>2 emissions by multiplying by a carbon
content coefficient, a carbon oxidation factor, and converting the resulting carbon emissions into
CC>2.  The CC>2 emissions were then scaled up by assuming a fraction increase to the CC>2
emissions to account for non-CC>2 GHGs (CH4 and N2O). The fraction increase was based on the
U.S. EPA National Inventory 2004 values for both CO2 and total GHG emissions. Table 6.2-3
shows the total lifecycle petroleum and GHG emissions associated with direct use of a Btu value
of gasoline or diesel fuel. These values represent factor LC in the equation described above.

                                      Table  6.2-3.
                       Lifecycle Emissions and Energy (LC Values)

Petroleum (Btu/Btu)
Fossil fuel (Btu/Btu)
GHG (Tg-CO2-eq/QBtu)
C02 (Tg-CCVQBtu)
Gasoline
1.11
1.22
99.4
94.2
Diesel
1.10
1.21
94.5
91.9
6.2.3   Displacement Indexes

       In order to permit a quantitative evaluation of the degree to which a renewable fuel
reduces lifecycle fossil fuel consumption or GHG emissions, several metrics have been
developed.  Three of the most prominent metrics are shown in Table 6.2-4.

       Table 6.2-4. Metrics Used to Measure Lifecycle Impacts of Renewable Fuels
Metric
Net energy balance
Energy efficiency
Displacement index
Calculation
Renewable energy out - fossil energy in
Fossil energy in + renewable energy out
(or alternatively renewable energy out + fossil
% reduction in emissions or energy compared
that it replaces
energy in)
to the fuel
       Of these metrics, we believe the displacement index is the most appropriate to use as it
compares the renewable fuel to the petroleum fuel it is displacing.  The net energy balance and
energy efficiency approaches only consider the renewable fuel itself and do not account for the
                                          246

-------
fact that the use of renewable fuels result in decreased use of petroleum fuels and thus provide
misleading results.

       As an example, if 81,000 Btu of fossil fuels were required to make, transport, and store
one gallon of ethanol, then the energy efficiency would be calculated as follows:

                 Energy efficiency = 81,000 Btu/gal - 76,000 Btu/gal = 1.07
       This result would imply that ethanol cannot be labeled "renewable," since one gallon of
ethanol contains less energy than was required to make that one gallon. However, the use of
ethanol may still reduce overall lifecycle fossil fuel use even in this case.  If, for example, 18,000
Btu of fossil fuels were required to make one ethanol-equivalent gallon of gasoline (i.e. 76,000
Btu of gasoline), then a total of 94,000 Btu of fossil fuel energy would be consumed whenever
76,000 Btu of gasoline energy was combusted in a conventional vehicle.  Since 81,000 Btu is
less than 94,000 Btu, the use of ethanol would result in less fossil fuel consumption than the use
of gasoline, even though the energy efficiency is greater than 1.0.  The 81,000 Btu of fossil
energy required to produce the ethanol includes lifecycle energy.  The energy content of the
ethanol (76,000 Btu) itself is not considered fossil energy and therefore not included in the
comparison with gasoline calculation above. Thus, even in cases where the net energy balance
of a renewable  fuel is negative or has energy efficiency less than  l.O74, there may still be an
overall reduction in lifecycle fossil fuel use (and associated GHG emissions) due to decreased
petroleum fuel use.

       Therefore, studies that rely on the energy balance metric and conclude for example that
the net energy balance of corn ethanol is negative, or the energy efficiency is less than 1.0,
making it an unattractive transportation fuel, are not capturing the full implications of the use of
the fuel and are providing misleading results.

       Because of this potential for the net energy balance and energy efficiency metrics to
provide misleading information, for our analysis of this rule we have chosen to use the
displacement index.  The displacement index provides the most direct measure of the impacts of
replacing conventional gasoline or diesel with a renewable fuel, and is also better suited to
describing impacts of renewable fuel use on fossil fuel consumption and GHGs.

       The displacement index (DI) represents the percent reduction in GHG emissions or fossil
fuel energy brought about by the use of a renewable fuel in comparison to the conventional
gasoline  or diesel that the renewable fuel replaces.  The formula for calculating the displacement
index depends on which fuel is being displaced  (i.e. gasoline or diesel), and which endpoint is of
interest (e.g. petroleum energy,  GHG).  For instance, when investigating the CC>2 impacts of
ethanol used in gasoline, the displacement index is calculated as follows:
74 A net energy balance of zero, or an energy efficiency of 1.0, would indicate that the full lifecycle fossil fuels used
in the production and transportation of ethanol are exactly equal to the energy in the ethanol itself.
                                           247

-------
       DIco2 = 1  -   lifecycle CO; emitted for ethanol in g/Btu
                    lifecycle CC>2 emitted for gasoline in g/Btu
       The units of g/Btu ensure that the comparison between the renewable fuel and the
conventional fuel is made on a common basis, and that differences in the volumetric energy
content of the fuels is taken into account. The denominator includes the CCh emitted through
combustion of the gasoline itself in addition to all the CO2 emitted during its manufacturer and
distribution. The numerator, in contrast, includes only the CC>2 emitted during the manufacturer
and distribution of ethanol, not the CC>2 emitted during combustion of the ethanol.

       The combustion of biomass-based fuels, such as ethanol from corn and woody crops,
generates CC>2.  However, in the long run the CC>2 emitted from biomass-based fuels combustion
does not increase atmospheric CC>2 concentrations, assuming the biogenic carbon emitted is
offset by the uptake of CC>2 resulting from the growth of new biomassDDDDD. As a result, CC>2
emissions from biomass-based fuels combustion are not included in their lifecycle emissions
results and are not used in the CC>2 displacement index calculations shown above. Net carbon
fluxes from changes in biogenic carbon reservoirs in wooded or crop lands are accounted for
separately in the GREET model.

       When calculating the GHG displacement index, however, the CH4 and N2O emitted
during biomass-based fuels combustion are included in the numerator. Unlike CC>2 emissions,
the combustion of biomass-based fuels does result in net additions of CH4 and N2O to the
atmosphere. We assume that combustion CH4 and N2O emissions are not offset by carbon
uptake of renewable biomass production. As  shown in Table 6.1-2, CH4 and N2O emissions
contribute to the total GHG impact. Therefore, combustion CH4 and N2O emissions are included
in the lifecycle GHG emissions results for biomass-based fuels and are used in the GHG
displacement index calculations.

       Using GREET, we calculated the lifecycle values for energy consumed and GHGs
produced for corn-ethanol, cellulosic ethanol, and soybean-based biodiesel, as well as the
gasoline and diesel fuel that would be displaced. For both renewable and conventional fuels, we
summed the lifecycle results for both the feedstock and the fuel.  The results are shown in Table
6.2-5.
                                         248

-------
             Table 6.2-5. Output from GREET Used to Develop Displacement Indexes

Units
Gasoline"
Corn
ethanol
Corn ethanol
(biomass fuel)
Cellulosic
ethanol
LS
Diesel
Biodiesel
Well-to-Pump
Fossil
energy
Petroleum
energy
CO2
CO2-eq
Btu/mmBtu
Btu/mmBtu
g/mmBtu
g/mmBtu
224,133
107,298
17,893
20,435
742,411
90,771
56,275
75,219
290,324
88,896
26,089
43,043
88,973
91,977
-71
6,427
207,008
98,656
16,629
19,134
464,594
96,539
28,468
31,193
End point combustion
Fossil energy
Petroleum
energy
C02
combustion13
Fossil CO2
combustion
CO2-eq
combustion0
Btu/mmBtu
Btu/mmBtu
g/mmBtu

g/mmBtu
1,000,000
1,000,000
76,419
76,419
79,015
0
0
74,755
0
2,596
0
0
74,755
0
2,596
0
0
74,755
0
2,596
1,000,000
1,000,000
77,570
77,570
77,669
0
0
79,388
0
99
 Volume-weighted average of conventional gasoline (65%), RFG blendstock (25%), and CaRFG blendstock (10%).
 Based on carbon content of the fuel.
0 Includes Fossil CO2, CH4, and N2O tailpipe emissions. CH4 and N2O emissions based on assuming an increase over CO2
emissions, the percent increase is from the U.S. EPA National Inventory for CO2 and GHG emissions from on-road sources.
           We used the values from the table above to calculate the displacement indexes. The
    results are shown in Table 6.2-6.
                                                 249

-------
                Table 6.2-6.  Displacement Indexes Derived from GREET

Dlpossil Fuel
DIpetroleum
DIoHG
DIco2
Corn ethanol
39.3%
91.8%
21.8%
40.3%
Corn ethanol
(biomass fuel)
76.3%
92.0%
54.1%
72.3%
Cellulosic ethanol
92.7%
91.7%
90.9%
100.1%
Imported
ethanol
69.0%
92.0%
56.0%
71.0%
Biodiesel
61.5%
91.2%
67.7%
69.8%
       The displacement indexes in this table represent the impact of replacing a Btu of gasoline
or diesel with a Btu of renewable fuel.  Thus, for instance, for every Btu of gasoline which is
replaced by corn ethanol, the total lifecycle GHG emissions that would have been produced from
that Btu of gasoline would be reduced by 21.8 percent. For every Btu of diesel which is replaced
by biodiesel, the total lifecycle petroleum energy that would have been consumed as a result of
burning that Btu of diesel fuel would be reduced by 91.2 percent.

       Consistent with the cost modeling done for this rule, for the 2012 cases we assume the
"cellulosic" ethanol volume is actually produced from corn utilizing a biomass fuel source at the
ethanol production plant. The displacement index for that fuel as shown in Table 6.2-6, is used
in the calculation of reductions.

       The displacement index for imported ethanol in all cases is based on an average of corn
and cellulosic ethanol. While not exclusively, we anticipate much imported ethanol to be
primarily sugarcane based ethanol. There currently is no sugarcane ethanol lifecycle values
included in GREET. The GHG emissions when producing sugarcane ethanol differs from corn
ethanol in that the GHG  emissions from growing sugarcane is likely different than for growing a
equivalent amount of corn to make a gallon of ethanol, the process of turning sugar into ethanol
is easier and therefore less energy intensive (which typically translates into lower GHG) and,
importantly, we understand that at least some of the ethanol produced in Brazil uses the bagasse
from the sugarcane itself as a process fuel source.  We know from our analysis that using  a
biomass source for process energy greatly improves the GHG benefit of the renewable fuel.
These factors would result in sugarcane ethanol having a greater GHG benefit per gallon than
corn ethanol, certainly where natural gas or coal is the typical process fuel source used.
Conversely, sugarcane ethanol production does not result in a co-product such as distillers grain
as in the case of corn ethanol. In our analyses, accounting for co-products significantly
improved the GHG displacement index for corn ethanol.  Furthermore, there would be additional
transportation emissions associated with transporting the imported ethanol to the U.S.  as
compared to domestically produced ethanol.  Developing a technically rigorous lifecycle
estimate for energy needs and GHG impacts for sugarcane ethanol is not a simple task and was
not available in the timeframe of this rulemaking.  Considering all of the differences between
imported and domestic ethanol, for this rulemaking, we assumed imported ethanol would be
predominately from sugarcane and have estimated DFs approximately mid-way between the
DFs for corn ethanol and DFs for cellulosic ethanol. We are continuing to develop  a better
                                          250

-------
understanding of the lifecycle energy and GHG impacts of producing ethanol from sugarcane
and other likely feedstocks of imported ethanol for any future analysis.

6.2.4  Transportation Sector and Nationwide Inventories

       For our analysis described above, we need estimates of transportation sector and
nationwide fossil energy and GHG emissions to determine the percent reduction impacts of the
program (e.g., TSectorGHG factor in the equation above).  These inventories are direct not
lifecycle and are needed for 2012 to compare to the projected renewable fuel savings in 2012.
6.2.4.1
Fossil Fuel Inventory
       The transportation sector and nationwide fossil fuel inventory is just the energy content
of the fuels used. Fossil fuel use in the transportation sector includes gasoline and diesel as well
as other petroleum fuels,  such as residual oil and LPG. It also includes other fossil energy use in
the form of natural gas and the fossil portion of electricity used.  Inherent with the assumptions
on the amounts of renewable fuels use projected to 2012, there are also assumed values for
gasoline and diesel fuel use.  Values for energy use of the different transportation fuels other
than gasoline and diesel (e.g., jet fuel, natural gas, etc.) were taken directly from the 2006
Annual Energy Outlook.

       The nationwide fossil fuel inventory includes petroleum, natural gas, and coal energy use.
The direct fossil fuel inventory values are shown in Table 6.2-7.

                    Table 6.2-7.  Direct Fossil Fuel Inventories (QBtu)

Nationwide
Transportation Sector
2012
94.53
31.41
6.2.4.2
Petroleum Inventory
       As with fossil energy, the transportation sector and nationwide petroleum inventory is
just the energy content of the fuels used. The transportation sector petroleum inventory includes
gasoline and diesel as well as other petroleum fuels, such as residual oil and LPG.

       The nationwide petroleum inventory includes petroleum use in the transportation sector
as well as other sectors. The direct petroleum inventory values are shown in Table 6.2-8.
                                           251

-------
                    Table 6.2-8. Direct Petroleum Inventories (QBtu)

Nationwide
Transportation Sector
2012
43.87
30.47
6.2.4.3
COi Inventories
       We calculated direct CC>2 emissions for the transportation sector in 2012 by applying
carbon emissions factors to the projected amount of fuels used in those years.

       Direct CC>2 emissions from the transportation sector as a whole are calculated in the same
way as direct gasoline and diesel emissions are calculated as described in Section 6.2.2. We
converted energy use values for transportation sector fuels to direct CC>2 emissions by
multiplying by a carbon content coefficient, a carbon oxidation factor, and converting the
resulting carbon emissions into CC>2.  Emissions from electricity use in the transportation sector
(rail) are calculated based on the U.S. average mix of fossil fuels used to generate electricity.

       Consistent with the EPA inventory report we made an adjustment to diesel fuel, jet fuel
and residual oil use to subtract out the emissions associated with bunker fuel.  The AEO values
include the energy use of bunker fuels, but the emissions of these fuels are not considered part of
the U.S. transportation sector emissions. This adjustment was done by decreasing emissions of
diesel fuel, jet fuel, and residual oil by the portion of emissions associated with bunker fuels as
determined in the EPA inventory report.

       Direct nationwide CC>2 emissions are also calculated in the same way applying factors for
all fossil fuels used as reported by the 2006 Annual Energy Outlook.  This type of analysis
results in a small understatement of total Nationwide CC>2 emissions as it does not capture other
industrial sources of CC>2 emissions for example CC>2 emissions from calcinations of limestone in
the cement industry.  However, there are no projections of these other emissions sources for
2012, and they are a relatively small part of total Nationwide CC>2 emissions, representing only
6% of total CC>2 emissions in 2004 according to the EPA National Inventory values.  Therefore,
while impacts of increased renewable fuel use as a percent of nationwide CC>2 emissions may be
slightly overestimated the impacts on results are not thought to be significant.  The results of
direct CC>2 emission calculations are shown in Table 6.2-9.
                                           252

-------
                      Table 6.2-9. CO2 Direct Inventories (Tg CO2)

Nationwide
Transportation Sector
2012
6,406
2,108
6.2.4.4
GHG Inventories
       Projections for direct GHG emissions can not be calculated directly from the energy
projections as was done for CO2.  The approach to estimating CO2 emissions from mobile
combustion sources varies significantly from the approach to estimating non-CO2 GHG
emissions (CH4 and N2O emissions). While CO2 can be reasonably estimated by applying an
appropriate carbon content and fraction of carbon oxidized factor to the fuel quantity consumed,
CH4 and N2O emissions depend largely on the emissions control equipment used (e.g., type of
catalytic converter) and vehicle miles traveled.  Emissions of these gases also vary with the
efficiency and vintage of the combustion technology, as well as maintenance and operational
practices. Due to this complexity, a much higher level of uncertainty exists in the estimation of
CH4 and N2O emissions from mobile combustion sources, compared to the estimation of CO2
emissions.

       Projections for direct transportation sector and nationwide GHG emission are done by
assuming a fraction increase to the CO2 emissions to account for non-CO2 GHGs.  The fraction
increase was based on the U.S. EPA National Inventory 2004EEEEE values for both CO2 and total
GHG emissions. This same increase is applied to 2012 CO2 values. Table 6.2-10 shows the
fraction increase values for GHGs over CO2 emissions calculated from the U.S. EPA National
Inventory report.

          Table 6.2-10. U.S. National Inventory 2004 CO2 and GHG Inventories

Nationwide
Transportation
Sector
C02 (Tg-C02)
5,988
1,860
GHG (Tg-CO2-eq.)
7,074
1,960
Fraction Increase
1.1807
1.0538
       The results of direct GHG emission calculations are shown in Table 6.2-11.
                                          253

-------
                  Table 6.2-11. GHG Direct Inventories (Tg CO2-eq.)

Nationwide
Transportation Sector
2012
7,564
2,222
6.3    Impacts of Increased Renewable Fuel Use

       We used the methodology described above to estimate impacts of increased use of
renewable fuels on consumption of petroleum and fossil fuels and also emissions of CC>2 and
GHGs. This section describes our results.
6.3.1   Fossil Fuels and Petroleum

       We used the S equation in Section 6.2 to estimate the reduction associated with the
increased use of renewable fuels on lifecycle fossil fuel and petroleum consumption. These
values are then compared to the total U.S. transportation sector and nationwide inventories to get
a percent reduction. The estimates are presented in Tables 6.3-1 and 6.3-2.

                                     Table 6.3-1.
       Estimated Fossil Fuel Impacts of Increased Use of Renewable Fuels in 2012,
                         In Comparison to the Reference Case

Reduction (quadrillion Btu)
Percent reduction in
Transportation Sector Energy Use
Percent reduction in Nationwide
Energy Use
RFS Case
0.15
0.48 %
0.16%
EIA Case
0.27
0.85 %
0.28 %
                                         254

-------
                                     Table 6.3-2.
       Estimated Petroleum Impacts of Increased Use of Renewable Fuels in 2012,
                         In Comparison to the Reference Case

Reduction (billion gal)
Percent reduction in
Transportation Sector Energy Use
Percent reduction in Nationwide
Energy Use
RFS Case
2.0
0.82 %
0.57 %
EIA Case
3.9
1.60%
1.11 %
6.3.2   Greenhouse Gases and Carbon Dioxide

       We used the S equation in Section 6.2 to estimate the reduction associated with the
increased use of renewable fuels on lifecycle emissions of CC>2. These values are then compared
to the total U.S. transportation sector and nationwide emissions to get a percent reduction. The
estimates are presented in Table 6.3-3.

                                     Table 6.3-3.
      Estimated COi Emission Impacts of Increased Use of Renewable Fuels in 2012,
                         In Comparison to the Reference Case

Reduction (million metric tons CC^)
Percent reduction in Transportation
Sector Emissions
Percent reduction in Nationwide
Emissions
RFS Case
11.0
0.52 %
0.17%
EIA Case
19.5
0.93 %
0.30 %
       Carbon dioxide is a subset of GHGs, along with CH4 and N2O as discussed above.  It can
be seen from Table 6.2-6 that the displacement index of CC>2 is greater than for GHGs for each
                                         255

-------
renewable fuel. This indicates that lifecycle emissions of CH4 and N2O are higher for renewable
fuels that for the conventional fuels replaced. Therefore, reductions associated with the
increased use of renewable fuels on lifecycle emissions of GHGs are lower than the values for
CO2. The estimates for GHGs are presented in Table 6.3-4.

                                     Table 6.3-4.
     Estimated GHG Emission Impacts of Increased Use of Renewable Fuels in 2012,
                         In Comparison to the Reference Case

Reduction
(million metric tons CO2-eq.)
Percent reduction in Transportation
Sector Emissions
Percent reduction in Nationwide
Emissions
RFS Case
8.0

0.36 %
0.11 %
EIA Case
13.1

0.59 %
0.17%
6.4    Implications of Reduced Imports of Petroleum Products

6.4.1   Impacts on Imports of Petroleum Products

       To assess the impact of the RFS program on petroleum imports, the fraction of domestic
consumption derived from foreign sources was estimated using results from the AEO 2006. We
describe in this section how fuel producers might change their levels and mix of imports in
response to a decrease in fuel demand.

       We compared the levels and mix of imports in the AEO reference case with the AEO low
macroeconomic growth case and AEO high oil price case. The latter two cases reflect different
assumptions by EIA regarding economic growth and world oil prices, respectively. The net
effect for both cases is a reduction in domestic petroleum consumption compared to the AEO
reference case. The changes in the level and mix of imports were examined, given a reduction in
petroleum consumption similar to the amount estimated in the RFS for 2012 (0.25 to 0.49
Quads). Note that the EIA has conducted three separate analyses of Congressional bills which
included earlier forms of the renewable fuel standard. These separate analyses however were
based on earlier AEO versions and, in some instances, considered numerous provisions in
addition to an RFS which collectively affected world oil prices and domestic oil consumption.
                                         256

-------
Thus, we did not directly use these earlier analyses, rather opting to use only the results in the
AEO 2006 cases, as discussed above, to assess the RFS impacts on imports.

       Comparison of the AEO 2006 reference case against the low macroeconomic growth case
allowed us to evaluate how a decrease in domestic petroleum demand might affect the mix of
imported finished products, imported crude oil, and domestic production. Note that the world
price of crude oil remains the same between the AEO low macroeconomic growth and reference
cases. Comparison  of the two cases show that with an initial decrease in petroleum consumption
(approximately 300,000 barrels per day or 0.61 Quads, higher than 2012 values), net imports will
account for approximately 95% of the reductions on an energy basis.FFFFF These net imports
include imports of crude oil or petroleum products minus exports of crude oil or petroleum
products.GGGGG Both reduced domestic crude production and natural gas plant liquids account for
most of the remainder. Note that for all levels of reduced petroleum demand, domestic crude
production appears to account for less than 5% of the change. In addition, the reductions shown
here do not reflect any rebound effect that may occur. Out of the initial reductions in net
petroleum imports, imported finished products account for almost all the reductions.  As domestic
petroleum demand  is reduced even further (over 860,000 barrels per day), approximately 50% of
the reductions come from imported finished products, 44% from imported crude oil,  and the
remainder from reduced domestic, natural gas plant liquid (NGL) production, and exports.

       Under the low macroeconomic growth case assumptions, imported finished products are
initially reduced presumably because they represent the higher marginal cost source for refineries
versus imported crude oil. Refineries may prefer to refine crude oil as opposed to importing
finished products because of the higher margins involved with the former and the potentially
more optimum use  of refining capacity. Crude oil,  as an international commodity, will be
purchased at the market price by refineries. Thus,  while crude oil from abroad may be produced
more cheaply than domestic production sources, refineries that purchase from either  source will
pay the international market price for that specific grade of crude oil based on specific gravity
and sulfur content plus the cost of transport to the U.S.

       Note that there is uncertainty in quantifying how refineries will change their mix of
sources with a decrease in petroleum demand, particularly at the levels estimated for the RFS.
Changes in world oil price from the reference case could also significantly alter the mix of
sources from which refineries choose. For example, a comparison between the AEO  high price
case and the reference case (under a decrease in petroleum consumption of 0.64 Quads) shows
that 80% of the reductions (on an  energy basis) come from reductions in net petroleum imports,
while the remaining 20% comes from reductions in domestic production. As petroleum
consumption is reduced even further, reductions in net petroleum imports make up an even
greater percentage.  For the reductions in  net petroleum imports, imports of finished products are
observed to actually increase while imports of crude oil decrease even more.

       We believe that the actual refinery response might range between these two AEO cases,
so that net import reductions could compose 80-95% of the reductions in petroleum demand for
2012. The split between the changes in imports of finished products versus crude oil  are more
uncertain. Discussions with EIA suggest the split could be close to 50-50. Thus, we believe the
range could be between these two estimates (nearly all to 50% finished product). For the
                                          257

-------
purposes of this RIA, we show values for the case where net import reductions come entirely
from imports of petroleum products, with an example shown below.

       By using the petroleum reduction levels as discussed in 6.3.2 of the RIA and comparing
these to the AEO 2006 results, we estimate that 95% of the lifecycle petroleum reductions will
be met through reductions in net petroleum imports. Table 6.4-1 shows the reductions in net
petroleum imports estimated for the RFS program. We expect that these import reductions will
be met almost exclusively from finished petroleum  products rather than from crude oil, for the
reasons given above and consistent with the results  of the AEO 2006 low macroeconomic growth
case. As an example calculation, we apportioned 95% of the total reductions in gasoline and
diesel to displaced finished product imports. By 2012, imports of finished products are estimated
to be reduced by 123,000 and 240,000 barrels per day, respectively, for the RFS and EIA cases.
We compare these reductions in imports against the AEO projected levels of net petroleum
imports. The range of reductions in net petroleum imports are estimated to be between 0.9 to
1.7%.
                      Table 6.4-1. Net reductions in Imports in 2012

Reduction in finished products21
(barrels per day)
Percent reduction15
RFS Case
123,000
0.89%
EIA Case
240,000
1.73%
              1 Net reductions relative to 2012 reference case
               Compared to AEO2006 projections for 2012 reference case
6.4.2   Impacts on Import Expenditures

       The reductions in petroleum imports were discussed in Section IX.D of the preamble. As
noted in the preamble, we calculate the change in expenditures on petroleum imports and ethanol
imports assuming this would not result in any other changes in consumer behavior that would be
reflected in fuel use.  95% of all reductions in petroleum imports were calculated to be from
finished petroleum products rather than crude oil, as discussed in the prior section. The economic
savings in petroleum product imports was calculated by multiplying the reductions in gasoline
and diesel imports by their corresponding price. According to the EIA, the price of imported
finished products is the market price minus domestic local transportation from refineries and
minus taxes.HHHHH An estimate was made by using the AEO 2006 wholesale gasoline, distillate,
and ethanol price forecasts for the specific analysis years. The current ethanol import tariff of
$0.54/gallon placed on countries outside the Caribbean and Central America is not included in
the import expenditures, since the tariff revenue collected would remain  in the U.S.

       As an example calculation, the RFS case is expected to yield a reduction of 2.0 billion
gallons of gasoline in the year 2012.  95% of these reductions, or 1.9 billion gallons, are
expected to come from imports of finished gasoline. Thus, the domestic refining sector would
avoid purchases of 1.9 billion gallons of gasoline and diesel at the wholesale price. According to
the AEO 2006, the end-user prices of gasoline and diesel are forecasted to be $2.01 per gallon
and $1.98  per gallon respectively. Minus federal taxes, state taxes, and distribution costs, the
                                          258

-------
wholesale prices of gasoline and diesel forecasted in the AEO 2006 are $1.376 and $1.382 per
gallon, respectively (2004$). Note that the AEO wholesale prices were used for this calculation,
as opposed to the gasoline and diesel production costs in Chapter 7 of the RIA, to stay consistent
with the other AEO results used herein. The avoided petroleum payments abroad thus total $2.6
billion in 2012 as shown in Table 6.4-2. The additional ethanol import expenditures, using the
same approach, is estimated to be $0.7 billion in 2012. The net avoided expenditures in imports
is thus the difference, or $1.9 billion in 2012 as shown in Table 6.4-2.

       We compare these avoided petroleum import expenditures against the projected value of
total U.S. net exports of all goods and  services economy-wide. Net exports is a measure of the
difference between the value of exports of goods and services by the U.S. and the value of U.S.
imports of goods and services from the rest of the world. For example, according to the AEO
2006, the value of total import expenditures of goods and services exceeds the value of U.S.
exports of goods and services to the rest of the world by $695 billion for 2006 (for a net export
level of minus $695 billion) and by $383 billion for 2012 (for a net export level of minus $383
billion).75 In Table 6.4-2, we compare  the avoided expenditures in imports versus the total value
of U.S. net exports of goods and services for the whole economy for 2012. Note that changes to
corn exports, discussed in Chapter 8 of the RIA, are also included in the calculation of net
exports. Relative to the 2012 projection, the avoided import expenditures due to the RFS would
represent 0.4 to 0.7% of economy-wide net exports.

                                       Table 6.4-2.
                      Avoided Import Expenditures ($2004 billion)
Cases
RFS Case
EIA Case
AEO Total
Net Exports
-$383
(year 20 12)
Expenditures
on Petroleum
Imports
-$2.6
-$5.1
Expenditures
on Ethanol
Imports
+ $0.7
+ $1.0
Decreased
Corn Exports
+ $0.6
+ $1.3
Net
Expenditure
s on Imports
-$1.4
-$2.8
Percent of
Total Net
Exports
0.4%
0.7%
6.5    Energy Security Implications of RFS

6.5.1   Background

       One of the effects of increased use of renewable fuels in the U.S. from the RFS is that it
diversifies the energy sources in making transportation fuel. A potential disruption in supply
reflected in the price volatility of a particular energy source carries with it both financial as well
as strategic risks. These risks can be reduced to the extent that diverse sources of fuel energy
reduce the dependence on any one source. This reduction in risks is a measure of improved
energy security.
75 For reference, the U.S. Bureau of Economic Analysis (BEA) reports that the 2005 import expenditures on energy-
related petroleum products totaled $235.5 billion (2004$) while petroleum exports totaled $13.6 billion - for a net of
$221.9 billion in expenditures. Net petroleum expenditures made up a significant fraction of the $591.3 billion
current account deficit in goods and services for 2005 (2004$). (http://www.bea.gov/)
                                           259

-------
       At the time of the proposal, EPA stated that an analysis would be completed and
estimates provided in support of this rule. In order to understand the energy security implications
of the RFS, EPA has worked with Oak Ridge National Laboratory (ORNL), which has
developed approaches for evaluating the social costs and energy security implications of oil use.
In a new study produced for the RFS, entitled  "The Energy Security Benefits of Reduced Oil Use,
2006-2015 " ORNL has updated and applied the method used in the 1997 report "OilImports:
An Assessment of Benefits and Costs", by Leiby, Jones, Curlee and Lee.76,77  While the 1997
report including a description of methodology and results at that time has been used or cited on a
number of occasions, this updated analysis and results have not been available for full public
consideration. Since energy security will be a key consideration in future actions aimed at
reducing our dependence on oil, it is important to assure estimates of energy security impacts
have been thoroughly examined in a full and open public forum. Since the updated analysis was
only recently available, such a thorough analysis has not been possible. Therefore, EPA has
decided not to rely on the results of this report for the purposes of this rulemaking. Rather, we
are including it as part of the record of this rulemaking and are inviting further public analysis
and consideration of both this particular report and other perspectives on how to best quantify
energy security benefits.  To facilitate that additional consideration, we highlight below some of
the key aspects of this particular analysis.

       The approach developed by ORNL estimates the incremental benefits to society, in
dollars per barrel, of reducing U.S. oil imports, called "oil  premium."  Since the 1997 publication
of this report, changes in  oil market conditions, both current and projected, suggest that the
magnitude of the oil premium has changed.  Significant  driving factors that have been revised
include: oil prices, current and anticipated levels of OPEC production, U.S. import levels, the
estimated responsiveness of regional oil supplies and demands to price, and the likelihood of oil
supply disruptions. For this analysis, oil prices from the EIA's AEO 2006 were used.  Using the
"oil premium" approach,  estimates of benefits of improved energy security from reduced U.S. oil
imports from increased use  of renewable fuels are calculated.

       In conducting this analysis, ORNL considered the full economic cost of importing
petroleum into the U.S. The full economic cost of importing petroleum into the U.S. is defined
for this analysis to include two components in addition to the purchase price of petroleum itself.
These are: (1) the higher costs for oil imports resulting from the effect of U.S. import demand on
the world oil price and OPEC market power (i.e., the so  called "demand" or "monoposony"
costs); and (2) the risk of reductions in U.S. economic output and disruption of the U.S. economy
caused by sudden disruptions  in the supply of imported oil to the U.S. (i.e., macroeconomic
disruption/adjustment costs).
76 Leiby, Paul N., Donald W. Jones, T. Randall Curlee, and Russell Lee, Oil Imports: An Assessment of Benefits and
Costs, ORNL-6851, Oak Ridge National Laboratory, November, 1997.

77 The 1997 ORNL paper was cited and its results used in DOT/NHTSA's rules establishing CAFE standards for
2008 through 2011 model year light trucks. See DOT/NHTSA, Final Regulatory Impacts Analysis: Corporate
Average Fuel Economy and CAFE Reform MY 2008-2011, March 2006.


                                           260

-------
       1.      Effect of Oil Use on Long-Run Oil Price, U.S. Import Costs, and Economic
              Output

       The first component of the full economic costs of importing petroleum into the U.S.
follows from the effect of U.S. import demand on the world oil price over the long-run. Because
the U.S. is a sufficiently large purchaser of foreign oil supplies, its purchases can affect the world
oil price. This monopsony power means that increases in U.S. petroleum demand can cause the
world price of crude oil to rise, and conversely, that reduced U.S. petroleum demand can reduce
the world price of crude oil.  Thus, one consequence of decreasing U.S. oil purchases due to
increased use of renewable fuel is the potential decrease in the crude oil price paid for all crude
oil purchased.

       2.      Short-Run Disruption Premium From Expected Costs of Sudden Supply
              Disruptions

       The second component of the  external economic costs resulting from U.S. oil imports
arises from the vulnerability of the U.S. economy to oil shocks.  The cost of shocks depends on
their likelihood, size, and length, the capabilities of the market and U.S. Strategic Petroleum
Reserve (SPR), the largest stockpile of government-owned emergency crude oil in the world, to
respond, and the sensitivity  of the U.S. economy to sudden price increases. While the total
vulnerability of the U.S. economy to oil price shocks depends on the levels of both U.S.
petroleum consumption and imports, variation in import levels or demand flexibility  can affect
the magnitude of potential increases in oil price due to  supply disruptions.  Disruptions are
uncertain events, so the costs of alternative possible disruptions are weighted by disruption
probabilities. The probabilities used by the ORNL study are based on a 2005 Energy Modeling
Forum78 synthesis of expert judgment and are used to determine an expected value of disruption
costs, and the change in those expected costs given reduced U.S. oil imports.

       3.      Costs of Existing U.S.  Energy Security Policies

       The last often-identified  component of the full economic costs of U.S. oil imports is the
costs to the U.S. taxpayers of existing U.S. energy security policies. The two primary examples
are maintaining a military presence to help secure stable oil supply from potentially vulnerable
regions of the world and maintaining  the SPR to provide buffer supplies and help protect the
U.S. economy from the consequences of global oil supply disruptions.

       U.S. military costs are excluded from the analysis performed by ORNL because their
attribution to particular missions or activities is difficult. Most military forces serve a broad
range of security and foreign policy objectives. Attempts to attribute some share of U.S. military
costs to oil imports are further challenged by the need to estimate how those costs might vary
with incremental variations  in U.S. oil imports. Similarly, while the costs for building and
maintaining the SPR are more clearly related to U.S. oil use and imports, historically these costs
78 Stanford Energy Modeling Forum, Phillip C. Beccue and Hillard G. Huntington, "An Assessment of Oil Market
Disruption Risks," Final Report, EMF SR 8, October, 2005.


                                           261

-------
have not varied in response to changes in U.S. oil import levels. Thus, while SPR is factored into
the ORNL analysis, the cost of maintaining the SPR is excluded.

       As stated earlier, we have placed the report in the docket of this rulemaking for the
purposes of inviting further consideration. However, the results of that report have not been used
in quantifying the impacts of this rule.
                                          262

-------
Chapter 7:  Estimated Costs of Renewable Fuels, Gasoline and
Diesel

       This section describes our methodology for estimating the cost impacts of increased
production, distribution, and blending of renewable fuels, including corn and cellulosic ethanol
and biodiesel.  Detailed information is given on expected changes to the nation's fuel distribution
system, as well as changes in refining processes that will likely occur as larger volumes of
ethanol are blended into gasoline. The impact of subsidies is also addressed.

7.1    Ethanol

       This subsection provides a description of the analysis we conducted for estimating the
cost of corn and cellulosic ethanol. Our analysis indicates that corn ethanol will cost $1.32 per
gallon to produce (2004  dollars) in 2012. We also estimated that using cellulosic feedstock, the
production costs for ethanol would be approximately $1.65  per gallon (2004 dollars).  By 2012
this cellulosic cost may decline with breakthroughs and advances in technology. Based on
reports from a variety of sources and discussions we held with members of academia as well as
those directly involved in the industry, we believe several hurdles still have to be overcome in
the production of large volumes of cellulosic derived ethanol. However, it appears that good
progress continues to be made and we remain optimistic that cellulosic ethanol will become
increasingly important in the future.

7.1.1   Corn Ethanol

       Of the new ethanol production capacity expected to  be built, according to Section 1.2.2 of
this RIA, less than three percent combined is expected to be produced from cellulosic feedstocks
or in plants that differ significantly from dry mill corn ethanol plants. Several plants will be able
to utilize other starchy feedstocks besides corn, such as milo, barley, wheat, and sorghum.
However, corn is the primary feedstock, and therefore, the following analyses will focus on dry
mill starch ethanol production.

7.1.1.1        Engineering and Construction Requirements for Corn Ethanol Plants

       To meet the RFS Case goal of 6.7 billion gallons per year (Bgal/y) of ethanol in 2012
from the October 2006 capacity of 5.2 Bgal/yr, 1.5 Bgal/yr  of additional capacity will have to be
constructed.79  If we consider that it is likely that at least 9.6 Bgal/yr of actual ethanol capacity
will come on-line by 2012 (EIA Case), the annual capacity  increase is 4.4 Bgal/yr. Our industry
characterization work considering plants that are either under construction or are planned to be
constructed in the next 2-3 years suggests the average new plant size will be 81 million gallons
per year (MMgal/yr) (including a small number of expansions).
79 For details on current and expected ethanol capacity, refer to Section 1.2 of this RIA. Note that volumes
considered cellulosic are also included here, since we believe that virtually all near-term cellulosic ethanol
production will be from starch-based feedstocks that meet the alternative definition in the Act (discussed further in
Section 7.1.1.2).


                                           263

-------
       Based on conversations with representatives from design-build firms working in this
field, as well as material from public sources, each new plant requires design engineering work
lasting about six months followed by construction lasting 12-14 months before plant startup is
possible, resulting in a total project timeline of 18-20 months. The design phase for a basic 50
MMgal/yr plant is expected to require the attention of about 12 engineers full time, and the
construction phase will employ an average of about 125-150 workers each day. To correlate
these figures with requirements for an 81 MMgal/yr plant, the number of construction personnel
(150) were scaled proportionally, while the number of engineering personnel were assumed to be
constant.

       These figures provide a basis for estimating the personnel requirements of the total
volume needed to meet the expected volumes. Over the six-year build-up period, a maximum of
970 construction workers and 30 engineers would be required on a monthly basis for the RFS
case, while for the EIA case, these numbers increase to 2,328 and 75, respectively.

       These figures simply estimate the number of workers required at the final assembly stage
of the plant, and do not capture many more personnel hours that will go into designing and
constructing vessels, pipe fittings, control systems, and other pieces of equipment that will be
installed and brought online by the plant construction crews.  A report produced by one
consultant suggested that expansion of the ethanol industry was responsible for more than 65,000
construction jobs in  2005.nra

7.1.1.2        Corn Ethanol Production Costs

       Corn ethanol costs for our work were estimated using a model developed by USDA that
was documented in a peer-reviewed journal paper on cost modeling of the dry-grind corn ethanol
process.JJJJJ It produces results that compare well with cost information found in  surveys of
existing plants.KKKKK

       The USDA model is for a 40 MMgal/yr corn plant producing ethanol with a primary co-
product of distillers  dried grains with solubles (DDGS). The ethanol yield used in the model is
2.87 gallons per bushel with 5.0% gasoline denaturant. The model is based on work done in
chemical process simulation software to generate equipment sizes,  stream flowrates, and material
and energy balances. These results were then put together with feedstock, energy, and
equipment  cost information in a spreadsheet format to arrive at a final per-gallon  cost estimate.
Although the model is current in terms of technology, yields, and capital estimates, we made
some modifications  to allow estimation of costs for ethanol plants of different sizes and
operating under different energy and feedstock prices.  We believe that these updates, in
combination with the industry and supplier surveys done by USDA in developing the model,
result a reasonable estimate for projected ethanol production costs.

       We estimate an average corn ethanol production cost of $1.26 per gallon in 2012 (2004
dollars) for the RFS case and $1.32 per gallon for the EIA case. The cost of ethanol production
is most sensitive to the prices of corn and the primary co-product, DDGS. Utilities, capital, and
labor expenses also have an impact, although  to a lesser extent. Corn feedstock minus DDGS
                                          264

-------
sale credit represents about 48% of the final per-gallon cost, while utilities, capital and labor
comprise about 19%, 9%, and 6%, respectively. For this work, we used corn and DDGS price
projections generated by the Forestry and Agricultural Sector Optimization Model, which is
described in Chapter 8.1.1  of this RIA. Corn and DDGS prices are given there in Table 8.1-1.
Figure 7.1-1 shows the cost breakdown for production of a gallon of ethanol. Note that this
production model does not account for the cost to pelletize or ship the DDGS.  Those costs are
external and are expected to increase the price of DDGS an end user located far from the plant.
More details are given in Section 8.1.1 where the FASOM model is discussed.

                                      Figure 7.1-1.
                  Cost Breakdown of Corn Ethanol Production (2004$).
 I .T-U
 i .£.\j
 A f\f\
Cpl .1)1)
tf»n on
c $0.80
0
— \J.£.\J
- 0
II J
** 1
o ">
(~) ro

/



V.



















-vpU.'HJ


D DDGS Sales Credit
• Denaturant
D Other Raw Materials
• Corn Feedstock Hauling
D Labor, Supplies, Overheads
D Depreciation
• Utilities
D Shelled Corn
Assumes 81 MMgal/yr plant size,
5% denaturant, $2.50/bu corn,
$83.35/ton DDGS.

       The ability to address plant scaling in the model was accomplished by applying an
engineering scaling factor to all plant equipment.  In past rulemakings involving modifications to
refineries we have used a material scaling factor of 0.65. This factor is applied as an exponent to
the ratio of the new size to the original size, the result of which is then multiplied by the original
capital cost. However, there is information suggesting that a general factor may be considerably
higher for ethanol plants. Based on a recent journal publication, a factor of 0.84 was used in this
work.LLLLL With this factor, the model indicates that the change in per-gallon production cost
due to economies  of scale is very small over the range of typical plant sizes, on the order of
$0.02 between 40 and 100  MMgal/yr. In this analysis we used an average new plant size of 81
MMgal/yr, derived from our industry characterization work in Chapter 1.
                                          265

-------
       We also added functions to estimate the per-gallon cost impact of coal combustion as a
process energy source rather than natural gas.  Our industry characterization work suggests that
about 14% of ethanol production from new plants being constructed will use coal for the process
energy source, so the effect on average costs is relatively small. Capital cost used for an 81
MMgal/yr gas-fired plant was $99 million (2004$).  For the coal system versus natural gas,
additional requirements were estimated at $45 million in capital for the same size plant, as well
as one additional operator per shift and 10% additional electric utility use.  These figures should
be considered conservative estimates, and were based on information from press releases as well
as a conversation with staff of a company that designs and builds ethanol plants. Additionally,
we adjusted the thermal efficiency of coal combustion processes downward by 13% relative to
natural gas, and electricity consumption upward by 10% to reflect operational differences in the
processes.MMMMM  Using this information in the model, the cost savings is about $0.04 per gallon
of ethanol for a coal-fired plant compared to natural gas firing. The  results presented here are a
weighted average of coal and gas production costs (using  14% coal).

       Under the Energy Act, starch ethanol can be counted as cellulosic if at least 90% of the
process energy is derived from animal wastes or other waste materials.NNNNN It is expected that
the vast majority of the 250 million gallons per year of cellulosic ethanol production required by
2013 will be made using this  provision.  While we have been unable to develop a detailed
production cost estimate for ethanol from corn which meets cellulosic criteria, we  assume that
the costs will not be  significantly different from conventionally produced corn ethanol.  We
believe this is reasonable because to the extent that these processes are utilized, we expect them
to be in locations where the very low or zero cost of the feedstock or biogas itself will likely
offset the costs of hauling the material and the additional capital for  handling and combusting it.
In addition, because  the quantity of ethanol produced using these processes is still  expected to be
a relatively small fraction of the total ethanol demand, the  sensitivity of the overall analysis to
this assumption is also very small.

       In  general,  energy prices used in the model were taken from historical EIA data for 2004
and scaled according the ratios of 2004-2012 price forecasts published in the Annual Energy
Outlook 2006. O0000'ppppp The prices used in the modeling are shown in Table 7.1-2.  Several
sensitivity cases were run using the model, and the results  are shown in Table 7.1-3. Input
values in this table were chosen to give a significant margin around current and anticipated
future prices.

                                        Table 7.1-2.
	Energy Prices Used for Ethanol Cost Modeling for 2012 (2004$)	^_
      Natural Gasa               Coal3                Electricity3          Natural Gasolineb
	$/MMBtu	$/MMBtu	$/kWh	$/gal	
	6.16	1.94	0.044	L36	
"Historical data based on averages for Iowa, Illinois, Minnesota, and Nebraska
b Natural gasoline (or natural gas liquids) is the typical denaturant used in ethanol  production, since it is cheaper than
finished gasoline. The price used was based on its value being 20 cents per gallon below wholesale gasoline.
                                           266

-------
                                      Table 7.1-3.
                     Energy and Feedstock Price Sensitivities (2004$)
Natural Gas = $6.00/MMBtu
Corn
$/bu
$2.00


$2.50


$3.00


$3.50


$4.00


$5.00


$6.00


DDGS
$/ton
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
Ethanol
$/gal
$1.18
$1.02
$0.86
$1.36
$1.20
$1.04
$1.53
$1.37
$1.21
$1.71
$1.55
$1.39
$1.88
$1.72
$1.56
$2.23
$2.07
$1.91
$2.58
$2.42
$2.26
Natural Gas = $12.00/MMBtu
Corn
$/bu
$2.00


$2.50


$3.00


$3.50


$4.00


$5.00


$6.00


DDGS
$/ton
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
$50.00
$100.00
$150.00
Ethanol
$/gal
$1.35
$1.19
$1.03
$1.52
$1.36
$1.20
$1.70
$1.54
$1.38
$1.87
$1.71
$1.55
$2.04
$1.88
$1.72
$2.39
$2.23
$2.07
$2.74
$2.58
$2.42
7.1.2   Cellulosic Ethanol

7.1.2.1    How Ethanol Is Made from Cellulosic Feedstocks

       It is not clear when the first processes to produce ethanol from cellulosic biomass were
discovered. While ethanol produced from starch can be traced historically to ancient times,
cellulosic derived ethanol appears to have been investigated in the 1800's. Until recently, the
demand for fuel ethanol has been somewhat limited, and not sufficient to support a cost-
competitive, commercial process to convert cellulose into ethanol.

       With the increasing demand for fuel ethanol during the past few years, good progress has
been made toward producing ethanol from cellulosic feedstocks.  Interest in ethanol has
continued to grow, initially fostered in part from EPA's reformulated gasoline (RFG) regulations
that required such gasoline to contain a minimum of 2 percent oxygen by weight in the fuel.
This minimum oxygen requirement has recently been revoked by EPA in response to the Energy
Act, which revised the Clean Air Act requirement for oxygen in RFG. The Renewable Fuel
Standard (RFS) continues to create a demand for ethanol. Likewise, there is an additional
incentive to produce cellulosic ethanol because the Energy Act mandates that, starting in 2013,
renewable fuels used in gasoline must include.
                                          267

-------
       There is a wide variety of government and renewable fuels industry research and
development programs dedicated to improving our ability to produce renewable fuels from
cellulosic feedstocks. There are at least three completely different approaches to producing
ethanol from cellulosic biomass, sometimes referred to as "platforms". The first is based on
what NREL refers to as the "sugar platform,"80 which refers to pretreating the biomass, then
hydrolyzing the cellulosic and hemicellulosic components into sugars, and then fermenting the
sugars into ethanol. Corn grain is a nearly ideal feedstock for producing ethanol by
fermentation, especially when compared with cellulosic biomass feedstocks.  Corn grain is easily
ground into small particles, following which the exposed starch which has a-linked saccharide
polymers is easily hydrolyzed into simple, single component sugar which can then be easily
fermented into ethanol.  By comparison, the biomass lignin  structure must be either mechanically
or chemically broken down to permit hydrolyzing chemicals and enzymes access to the
saccharide  polymers. The central problem is that the cellulose/hemicellulose saccharide
polymers are p-linked which makes hydrolysis much more difficult.  Simple microbial
fermentation used in corn sugar fermentation is also not possible, since the cellulose and
hemicellulose (6 & 5 carbon molecules, respectively) have not been able to be fermented by the
same microbe. We discuss various pretreatment, hydrolysis and fermentation technologies,
below. The second and third approaches have nothing to do with pretreatment, acids, enzymes,
or fermentation.  The second is sometimes referred to as  the "syngas" or "gas-to-liquid"
approach; we will call it the "Syngas Platform." Briefly, the cellulosic biomass feedstock is
steam-reformed to produce syngas which is then converted to ethanol over a Fischer-Tropsch
catalyst.  The third approach uses plasma technology.

       Technologies that are currently being developed may solve some of the problems
associated with producing cellulosic ethanol.  Specifically, one problem, mentioned previously,
with cellulosic feedstocks is that the hydrolysis reactions produce both glucose, a six-carbon
sugar, and xylose, a five-carbon sugar (pentose sugar, CsHioOs; sometimes called "wood sugar").
Early  conversion technology required different microbes to  ferment each sugar. Recent research
has developed better cellulose hydrolysis enzymes and ethanol-fermenting organisms.81  Now,
glucose and xylose can be co-fermented—hence, the terminology, weak-acid enzymatic
hydrolysis  and co-fermentation.  In addition, at least one group is researching the use of recently
developed genome modifying technology to produce a variety of new or modified enzymes and
microbes that show promise for use in a process known as weak-acid, enzymatic-prehydrolysis 82
  Enzyme Sugar Platform (ESP), Project Next Steps National Renewable Energy, Dan Schell, FY03 Review
Meeting; Laboratory Operated for the U.S. Department of Energy by Midwest Research Institute • B NREL,
Golden, Colorado, May 1-2, 2003; U.S. Department of Energy by Midwest Research Institute • Battelle • Bechtel

81 "Purdue yeast makes ethanol from agricultural waste more effectively." Purdue News, June 28, 2004; Writer:
Emil Venere, (765) 494-4709, venere@purdue.edu; source: Dr. Nancy Ho, (765) 494-7046,
nwyho@ecn.purdue.edu.

82 DOE Genomics: GTL Roadmap, Systems Biology for Energy and Environment,  U.S. Department of Energy,
Office of Science, Office of Biological and Environmental Research, Office of Advanced Scientific Computing
Research, Germantown, MD 20874-1290, August 2005; DOEGenomesToLife.org/roadmap: downloadable as whole
or in sections.
                                           268

-------
7.1.2.2        Difficulties in Estimating Capital and Operating Costs for New or Pioneer
              Process Plants

       Many years ago the petroleum and chemical process industries learned that it can be
financially problematic to scale-up a bench or laboratory scale process to a full commercial sized
operation. There are simply too many process variables that act one way at a batch rate of one-
or two-gallons per day, or even 100-gallons per day, but then act a completely different way in a
continuous, 70,000 gallon per day operation.  Under these, admittedly somewhat extreme
expectations, there is also absolutely no reasonable way to optimize a process.  We expect that at
least pilot or demonstration size projects will be necessary before a fully commercial sized,
reasonably optimized plant can be constructed.

       The petroleum and chemical process industries have also learned that if a different
feedstock, with similar, but at the same time sufficiently different characteristics, becomes
available, it is nearly always necessary to  make several pilot plant runs before the feedstock in
introduced into the process. There are a wide variety of potential cellulosic feedstocks, such as
switch grass,  forest thinnings, municipal waste, wood chips, and corn stover (corn stalks).  The
physical characteristics of these materials, such as size, composition, and density vary widely.
As a result, there could be  significant differences  in the process configurations, as well as
differences in the enzyme "cocktails" required to  hydrolyze and convert each of them into
ethanol.  Compositional and density variations may require different reactor residence times for
each feedstock, which will impact throughput. Many of the process streams will actually be
slurries of the feedstock. It is also quite likely that each slurry stream will have its own flow and
compositional characteristics. The flow characteristics of any slurry, under real operating
conditions, must be well understood in order to properly design  an optimum system.
Additionally, valve and pump types, sizes, and materials of construction, as well as line sizes and
configurations, may vary.  Apart from the various process issues, questions also remain
regarding which of the feedstocks is actually the best in terms of ethanol yield per dollar.

       Consequently, we believe a good deal more process data is necessary before a reasonably
accurate cost to design, engineer, and build a commercial scale cellulosic based ethanol plant can
be expected.  At the present time, there is  only one cellulosic ethanol plant in North America
(IogenQQQQQ a privately held company, based in Ottawa, Ontario, Canada).  On February 28,
2007, however, the Department of Energy (DOE) announced that it will provide grants of up to
$385 billion for six biorefmery projects over the next four years. These facilities are expected to
produce more than 130 million gallons of celluloic ethanol per year.  As additional information
on these future facilities are made available, EPA will have more information on process design
from which we will better be able to project production costs for cellulosic ethanol.

       Although the industry seems to be moving down several different pathways, one of the
more mature process being tested and improved uses dilute acid enzymatic prehydrolysis with
simultaneous saccharification (enzymatic) and co-fermentation.   Because there is more publicly
available information about this process, the model we used incorporates this type of process to
estimate the cost of producing ethanol from corn stover. We chose corn stover because it is
ubiquitous and because of the likelihood it will eventually be used as a feedstock.
                                           269

-------
       In 1999, the National Renewable Energy Laboratory (NREL) published a report outlining
its work with the USDA to design a computer model of a plant to produce ethanol from hard-
wood chips.RRRRR  Although the cellulosic model was originally prepared for hardwood chips, it
was meant to serve as a modifiable-platform for ongoing research using cellulosic biomass as
feedstock to produce ethanol.  Their long-term plan was that various indices, costs, technologies,
and other factors would be regularly updated.

       NREL modified the model in order to compare the cost of using corn-grain with the cost
of using corn stover to produce ethanol. We used the corn stover model from the second
NREL/USDA study for this analysis.  Because there are no operating plants that could
potentially provide real world process design, construction, and operating data for processing
cellulosic ethanol, NREL had originally considered modeling the plant based on assumptions
associated with a pioneer plant. Such assumptions would likely result in costs significantly
higher than corn ethanol plants due to the higher level of uncertainty in both the design and
engineering as well as the final construction and operating costs.  The literature indicates that
such models often underestimate actual costs since the high performance assumed for pioneer
process plants is generally unrealistic.

       The NREL analysis assumed that the corn stover plant was an Nth generation plant, built
after the industry had been well enough established to provide verified costs. The corn stover
plant was normalized to the corn kernel plant, e.g., placed on a similar basis. Additional costs
for risk financing,  longer start-ups, and other costs associated with first-of-a-kind or pioneer
plants were not included in the study.sssss  It is also reasonable to expect the cost of cellulosic
ethanol will be higher than corn ethanol because of the  complexity of the cellulose conversion
process. During the recent past, process improvements and other advancements in corn
production have considerably reduced the cost of producing corn ethanol. We also believe it is
realistic to assume that cellulose-derived ethanol process improvements will be made and that
one can likewise reasonably expect that as the industry  matures, the cost of producing ethanol
from cellulose will also decrease.

7.1.2.3        Methods, Data Sources, and Assumptions

       For our analysis, we used the spreadsheet model that NREL developed for its comparison
of the costs of producing ethanol from corn grain and corn stover.83  We believe NREL's
approach was reasonable and well thought out.  They also had an outside engineering firm
validate their work, to the extent possible. The Delta-T Corporation (Delta-T) assisted in
preparing, reviewing, and estimating costs for the process design. Delta-T worked with NREL
process engineers to review all the process design and equipment costs (with the exclusion of
wastewater treatment and the burner-boiler system, which were reviewed by Merrick
Engineering and Reaction Engineering, Inc., respectively).  For the plant areas that are actively
83 The first, woodchip-plant study was designed to produce 52.2 million gallons of ethanol per year from about
2,200 tons per day (350 operating days per year; 15 days for downtime, including turn-around) of woodchips. The
second study normalized the original woodchip plant into the corn stover plant to produce 25 million gallons of
ethanol per year (about 1,235 wet tons per day), in 1999 dollars.  The adjustments included feedrate and feedstock
volume and cost adjustments; equipment sizes with adjustments to capital and installation costs, and the cost of
capital, labor, and process chemicals, including denaturant.


                                           270

-------
being investigated under DOE programs (e.g., prehydrolysis, cellulase enzyme production, and
simultaneous saccharification and co-fermentation), Delta-T used the results of the DOE
sponsored research to identify process design criteria and equipment requirements. These were
used as a basis for sizing and costing major equipment components in the facility. The results of
Merrick Engineering's work on wastewater treatment and REI's work on the burner/boiler were
also included.TTTTT The NREL model used the Aspen Plus™ process simulator to calculate the
flows and the heat and material balances for the process. We decided to use the NREL
spreadsheet corn stover model, as is, since we did not have access to the Aspen Plus™ model nor
to all the input.  Rather, we left the feedrate, yields, and streams flows as they  were, but adjusted
equipment capital and installation costs, and utility, chemical, and labor costs to 2004 dollars.
We used the same indices used by NREL to update their corn stover study; however, we used
actual costs and indices for 2004 where possible. For example, in their 2000 calculations, NREL
had extrapolated the Chemical Engineering Plant Cost Index and the Chemical Cost Index
through 2012.  However, we used actual 2004 data rather than the extrapolated data.

       We did not change the corn stover cost.  Several issues remain to be settled regarding the
amount of stover that should be left in place and how it should be gathered, baled, and shipped.
We found cost ranges of from $25 per dry ton to $45 per dry ton.  For purposes of this analysis
we used the $35 per dry ton that NREL assumed in its analysis.

       For the analysis, we  calculated the annual production cost in dollars per gallon of fuel
ethanol.  The annual production cost includes equipment straight-line depreciation for the life of
the plant (10 years), and variable costs, labor, supplies and  overhead, minus any by-product
credits.  Gasoline for denaturant and diesel for bulldozers to move the stover were projected into
2012 prices using lEA's AEO 2006UUUUU report. The market selling price minus the  annual
production cost is the before-tax profit. We calculated variable operating costs using NREL's
best estimate of quantities of chemicals and additives based on their laboratory work.  NREL
calculated fixed costs using  industry standards for percentages of direct labor (indirect labor was
40% of direct labor and overhead was 60% of total labor); other operating supplies, insurance,
etc. totaled 3.25% of total installed cost.  According to the analysis three major cost categories
made up the majority of the  total production cost:  feeds stock- 31.2%; fixed costs - 23.8%; and
depreciation (reflects installed capital cost of equipment) - 33.8%.

       As previously stated, several feedstock issues remain to be settled, not least being which
of the many available feedstocks will be the best or most efficient. We chose an average cost of
$35 per dry ton; we don't believe the cost will rise and as a result of the research that is currently
under way, there is reason to expect it to come down a little. On the other hand, several
researchers have indicated switch grass may be better than corn stover;  others  point to forest
wastes, etc. In the end, the best feedstock will likely be the one that is readily available  and close
to the plant; gathering, baling, and hauling continue to be important issues that will definitely
impact the viability of a feedstock. Equipment cost reductions may have a significant impact on
future costs. For example, there appear to be reasons to expect  significant savings from
purchasing enzymes rather than growing them onsite. Another issue that remains to be
investigated is whether a particular kind of feedstock can be processed using one type of
technology while a different kind may require the use of a completely different technology.
                                          271

-------
7.1.2.4        Results and Discussion

       Given the limitations we've already discussed, and perhaps others, we determined that it
would have cost approximately $1.65 in 2012 (2004 dollars) to produce a gallon of ethanol using
corn stover as a cellulosic feedstock.

       The provisions offering grants and shared financing included in Title XV of the Energy
Actvvvvv will likely encourage process development work to generate the necessary
construction and operating cost estimates. We assume the results produced by the above
referenced NREL study are accurate and reasonable given the state of our current knowledge.
7.2    Biodiesel and Renewable Diesel Production Costs

7.2.1 Overview of Analysis

       We based our estimate for the cost to produce biodiesel on the use of USDA's, NREL's
and EIA's biodiesel computer models, along with estimates from engineering vendors that
design biodiesel plants. Biodiesel fuel can be made from a wide variety of virgin vegetable oils
such as canola, corn oil, cottonseed, etc. though, the operating costs (minus the costs of the
feedstock oils) for these virgin vegetable oils are similar to the costs based on using soy oil as a
feedstock, according to an analysis by NREL84. Biodiesel costs are therefore determined based
on the use of soy oil, since this is the most commonly used virgin vegetable feedstock oil, and
the use of recycled cooking oil (yellow grease) as a feedstock. Production costs are based on the
process of continuous transesterification, which converts these feedstock oils to esters, along
with the ester finishing processes and glycerol recovery. The models and vendors data are used
to estimate the capital, fixed and operating costs associated with the production of biodiesel fuel,
considering utility, labor,  land and any other process and operating requirements, along with the
prices for feedstock oils, methanol, chemicals and the byproduct glycerol.

       The USD A, NREL and EIA models are based on a medium sized biodiesel plant that
was designed to process raw degummed virgin soy  oil as the feedstock.  Additionally, the EIA
model also contains a representation to estimate the biodiesel production cost for a plant that
uses yellow grease as a feedstock.  In the USDA model, the equipment needs and operating
requirements for their biodiesel plant were estimated through the use of process simulation
software. This software determines the biodiesel process requirements based on the use of
established engineering relationships, process operating conditions and reagent needs. To
substantiate the validity and accuracy of their model, USDA solicited feedback from major
biodiesel producers. Based on responses, they then made adjustments to their model and updated
their input prices to year 2005. The NREL model is also based on process simulation software,
though the results are adjusted to reflect NREL's modeling methods, using prices based on year
2002. The origin of the EIA model is not known, though it is based on 2004 prices. The output
for all of these models was provided in spreadsheet format. We also use engineering vendor
84 NREL Presentation "U.S. Biodiesel Feedstock Supply" June 2004.


                                          272

-------
estimates as another source to generate soy oil and yellow grease biodiesel production costs.
These firms are primarily engaged in the business of designing biodiesel plants.
       The production costs are based on a 10 million gallon per year biodiesel plant located in
the Midwest using feedstock oils and methanol, which are catalyzed into esters and glycerol by
use of sodium hydroxide. Because local feedstock costs, distribution costs, and biodiesel plant
type introduce some variability into cost estimates, we believe that using an average plant to
estimate production costs provides a reasonable approach. Therefore, we simplified our analysis
and used costs based  on an average plant and average feedstock prices since the total biodiesel
volumes forecasted are not large and represent a small fraction of the total projected renewable
volumes.

       The models and vendor estimates are further modified to use input prices for feedstocks,
byproducts and energy that reflect the effects of the fuels provisions in the Energy Act. In order
to capture a range of  production costs,  we generated cost projections from all of the models and
vendors. We present  the details on these estimates later in this section.

       For soy oil biodiesel production, based on the USDA model, we estimate a production
cost  of $2.06 per gallon in 2004 and  $1.89 per gal in 2012 (in 2004 dollars). With the NREL
model, we estimate soy oil biodiesel production costs of $2.28 and $2.11 per gallon in 2004 and
2012, respectively, which is slightly higher than the USDA results. The EIA model generated
soy oil based costs of $2.33 and $ 2.15/gal, while the engineering vendor's costs averaged $2.27
and $2.09/gal, in years 2004 and 2012, respectively.

       For yellow grease derived biodiesel, we used the EIA and vendor estimates and generated
a range of costs, as discussed later. The total production costs ranged from $1.24 to $1.60/gal in
2004, and from $1.11 to $1.56 for year 2012.

                                      Table 7.2-1.
        Summary of Production Costs for Biodiesel made from Soy Oil, per Gallon
                                      (2004 cents)

USDA
NREL
EIA
PSI-Lurgi
Superior
Process
Technologies
Total
Production
Cost
189
211
215
220
224
Subsidized
Production
Cost
89
111
115
120
124
Feed
156
165
161
174
175
Capital
11.3
17.0
14.4
18.8
11.7
Reagent
and
Chemicals
12.7
17.0
NA
12.6
16.5
Labor
5.0
6.0
NA
8.2
7.6
Energy/Utilities
4.8
7.4
16.0
5.5
5.0
                                           273

-------
                                       Table 7.2-2.
     Summary of Production Costs for Biodiesel made from Yellow Grease, per Gallon
                                       (2004 cents)

EIA
Superior
Process
Technologies
Total
Production
Cost
138
167
Subsidized
Production
Cost
88
117
Feed
80
114
Capital
14.4
14.7
Reagent
and
Chemicals
NA
18.3
Labor
NA
8.7
Energy/Utilities
16.9
9.0
        With the current Biodiesel Blender Tax Credit Program, producers using virgin vegetable
 oil stocks receive a one dollar per gallon tax subsidy while yellow grease producers receive 50
 cents per gallon, reducing the net production cost to a range of 89 to 115 c/gal for soy oil and 61
 to 106 c/gal for yellow greased derived biodiesel fuel in 2012. This compares favorably to the
 projected wholesale diesel fuel prices of 138 cents per gallon in 2012, signifying that the
 economics for biodiesel are positive under the effects of the blender credit program, though the
 tax credit program will expire in 2008 if it is not extended. Congress may later elect to extend
 the blender credit program, though, following the precedence used for extending the ethanol
 blending subsidies.  Additionally, the Small Biodiesel Blenders Tax credit program and state tax
 and credit programs offer some additional subsidies and credits, though the benefits are modest
 in comparison to the Blender's Tax credit.

 7.2.2   Inputs to and Results of USDA's Model

        We used USDA's biodiesel model as a source to generate an estimate for the cost to
produce biodiesel fuel. The model is in spreadsheet format with inputs in 2005 dollars, and
contains all of the capital and operating costs for a plant to produce 10 million gallons per year of
biodiesel fuel.
 7.2.2.1
Feedstock Costs
        Feedstock prices are the largest component in generating production costs for biodiesel
 fuel. For soy oil prices, we used prices based on USDA's 2006 Outlook, which has forecasted
 soy oil prices considering production of biodiesel under EPAct 2005.  USDA's Outlook is a
 national forecasting analysis that models the effects of demand for farm products and farm
 product prices for soy beans, soy bean oil, corn and other farm commodities. The 2006 Outlook
 estimated soy oil prices considering the demand of soy oil derived biodiesel fuel at
 approximately 160  MM gallons per year in 2006 and 312 MM gallons a year in 200785. This is
 in close proximity to EIA's soy oil derived biodiesel volume projection of 135 MM gals and 265
 MM gals in 2006 and 2007, respectively. We therefore used the soy oil prices from USDA's
 Outlook to determine biodiesel production costs. The USDA does not forecast yellow grease
 prices, so we assumed that yellow grease feedstocks costs would maintain the same relative
 historical pricing differential to virgin soy oil.  In the past, some analysis has shown that yellow
  ' Per USDA phone discussion 6/22/06
                                           274

-------
grease has sold for about half the price of soy oil86.  The resulting feedstock costs to make a
gallon of biodiesel under projected volumes for RFS are in Table 7.2-3.

                                       Table7.2-3.
                  Projected Prices of Feedstock (2004 Dollars per Gallon)
Marketing Year
2004
2012
Soy Oil3
1.71
1.56
Yellow Grease
0.86
0.78
""Production of Biodiesel assumed to consume 7.42 Ibs of soy oil per gallon. USDA prices in 2012 are adjusted to
2004 dollars to account for inflation, using GDP index of 109.7 in year 2004 and 130.8 in year 2012.
7.2.2.2        Capital Costs

For capital costs we used USDA's total installed capital cost of $10.66 MM for a 10 MM gallon
per year plant.  This estimate was determined by the USDA, using a detailed analysis to
generate costs for equipment needs, installation, land, engineering and construction work,
buildings, utility needs, contingencies, startup costs etc. The USDA model is based on 2005
dollars, so we adjusted the numbers to 2004 values using the GDP index. Per the USDA method,
the total installed capital costs on a per gallon basis was amortized on a 10 year straight line
depreciation rate using a facility dependent cost of 10 percent times the capital costs.
Maintenance charges, insurances and facility supply costs were also calculated as percentages of
the capital. The total of all of these are equal to 16 cents per gallon.

7.2.2.3        Operating Costs

       The total operating expenses were 20 and 18 c/gal for a soy based biodiesel plant in 2004
and 2012, respectively. The operating cost included a 4 cent per gallon offset from sale of the
glycerol product at a price of 5 cents/lb. The operating costs include values for utilities, feed
reagents, manpower and were based on the USDA's model.  The components of the operating
costs are discussed below.

7.2.2.4        Utility and Labor Costs

       We estimated utility costs using energy requirements from USDA's model and adjusted
the inputs to match the energy and electricity prices for the Midwest, using prices from EIA's
AEO.  The cost for steam was estimated using the price of natural gas. Each pound of steam
was produced from heating water, which required 810 British Thermal Units (BTUs) per pound
of steam.  Additionally, the steam costs are estimated assuming that the BTUs to make steam are
increased  by a factor of two, to account for steam distribution efficiency losses, treatment of
boiler water to prevent fouling, maintenance and other miscellaneous costs. The utility
requirements per gallon of biodiesel and energy prices are presented in Tables 7.2-4 and 7.2-5
86Energy Information Administration NEMS Petroleum Marketing Model Documentation page J-2


                                           275

-------
                                      Table 7.2-4.
                       Utility Requirements per Gallon Biodiesel3
Medium Pressure Steam, Ibs
Electricity, kWh
Cooling Tower Water, Ibs
4.0
0.10
96.1
"Utilities per USDA model from the production of biodiesel from soy oil.
                                      Table 7.2-5.
                       Midwest Energy Prices per Year (in 2004 $)
Year
Electricity, $/kWh
Natural gas, $/MM BTU
2004
0.046
7.16
2012
0.044
6.16
       Labor costs include the salaries and benefits for personnel to operate a biodiesel plant.
This was estimated in the USDA model, though the labor costs were in 2005 dollars, which we
adjusted to 2004 dollars using the GDP price index.  The resulting labor costs are 5 cents per
gallon of produced biodiesel fuel.
7.2.2.5
Chemical Reagents
       Another operating expense, the production of biodiesel also requires the use of chemicals
and chemical reagents, as these act as a catalyst in the transesterification process.  Additionally,
methanol is required as it is the feedstock that is chemically combined with soy oil and yellow
grease during the transesterification process, yielding the biodiesel product.  The amount of
chemicals and methanol required to make a gallon of biodiesel are listed in Table 7.2-6.

                                      Table 7.2-6.
                                 Reagent Requirements
Reagent
Water
Hydrochloric acid
Methanol
Sodium Methoxide
Sodium Hydroxide
Annual Requirement,
Ibs per gallon
0.0323
0.0185
0.8006
0.0231
0.0031
       For the prices of chemical reagents, we used prices that were supplied in USDA's 2005
model and adjusted them to 2004 dollars. Additionally, since we have no forecasting
mechanism we assumed that the chemical reagent prices remained unchanged in 2012.
                                          276

-------
However, we estimated methanol prices, as the cost for this feedstock is a significant component
of the total operating costs. For our analysis, we generated values by use of a correlation that
calculates methanol's price as a function of the price of natural gas87.  In 2004 and 2012, using
Midwest natural gas prices, we estimated methanol prices of 13.1 and 11.6 cents per pound,
respectively.  All other chemical prices, we assumed were constant over time and are in Table
7.2-7.
                                       Table 7.2-7.
                                Reagent Prices (in 2004 $)
Reagent
Hydrochloric acid
Sodium Methoxide
Sodium Hydroxide
Prices, $/lb
0.167
1.358
0.273
7.2.2.6
Glycerol Byproduct
       The feedstock cost credit for the glycerin by product in our modeling work was 5 cents
per pound, based on recent pricing trends, assuming that additional glycerol generated from
expansion of biodiesel production will continue to keep prices low. The model, like many
biodiesel plants, produces a crude 80% glycerin stream, which is usually sold to glycerin refiners
for purification.  In the past, crude glycerin has sold for around $0.15 / pound.  Because of the
increase in biodiesel production around the world, however, the crude glycerin market has
become saturated and the price is now around $0.05 / pound. As more biodiesel capacity comes
on line, this price may very well drop further, though other markets for the use of glycerol are
likely to develop because glycerol is a platform chemical used throughout industry. We assumed
that the current glycerin pricing environment will continue in the future.  For our cost estimation,
the byproduct glycerin was sold at 5 cents per pound, reflecting current saturated market and low
pricing conditions. The income from sale of the byproduct glycerin lowered biodiesel production
costs by 2 percent and 4 percent for soy oil and yellow grease derived biodiesel fuel,
respectively.

       The total biodiesel production costs derived using the USDA's model are summarized in
Table 7.2-8
87 Per EIA paper "MTBE Production Economics" Tancred C. M. Lidderdale, methanol price cents per gallon:
15.79 + 0.099 *natural gas price ($ per million BTU)
                                           277

-------
                                       Table7.2-8.
             Projected Production Costs for Biodiesel by Feedstock per Gallon
                                      (2004 Dollars)
Marketing Year
2004
2012
Soy Oil
2.06
1.89
 7.2.3   Inputs to and Results of NREL's Biodiesel Model

        We used NREL's biodiesel model as another source to generate an estimate for the cost
to produce biodiesel fuel.  Similar to the USDA's model, the NREL biodiesel model also
represents a continuous transesterification process that uses sodium hydroxide and methanol to
convert soy oil to biodiesel and which has the finishing processes for biodiesel and glycerol. The
model is in spreadsheet format, and contains all of the capital and operating costs for a plant to
produce 10 million gallons per year of biodiesel fuel. We again simplified our analysis, and used
the NREL model to estimate production costs for an average biodiesel plant that makes 10 MM
gallon per year.  To make the results directly comparable to USDA's model, we used energy
costs in the Midwest, and based the analysis on production of soy oil derived biodiesel.

        Based on the results of the NREL model, we estimate that the total production costs to
make soy oil derived biodiesel fuel are $2.28 and $2.11 per gallon for years 2004 and 2012,
respectively.  This is 22 cents more per gallon than the estimate derived from USDA's model.
The components that make up our NREL estimate are discussed in the sections that follow.
 7.2.3.1
Feedstock Costs
        The feedstock costs increase because the NREL model assumes 7.87 pounds of soy oil
are required to make a gallon of biodiesel fuel.  This is slightly higher than the pounds required
by the USDA model, though the difference may be due to each model being based on soy oils
with differing chemical structures, i.e. more esters, differing densities.  The higher amount of soy
oil required by the NREL model raises the production costs for biodiesel by about 10 cents per
gallon for feedstock costs alone, versus the USDA model.  The feedstock costs are summarized in
Table7.2-9.

                                       Table7.2-9.
                  Projected Prices of Feedstock (2004 Dollars per Gallon)
Marketing Year
2004
2012
Soy Oil
181.0
165.2
                                           278

-------
7.2.3.2
Capital Costs
       The total capital cost in the NREL model account for all of the costs for building a plant,
including but not limited to the expenses for equipment, tanks, installation costs, engineering,
tanks, construction, land and site development, start up and permitting charges. These costs do
not account for expenses incurred from maintenance, insurance and taxes, however.  The total
capital coats for a plant are $14.8 million in 2002 dollars, which we adjusted to 2004 dollars
using the GDP price index.  The capital costs were amortized assuming a seven percent return  on
investment, resulting in a cost of 17 cents per gallon.  All of the economic factors used for
amortizing the capital costs are summarized in Table 7.2-10.

                                      Table 7.2-10.
        Economic Factors Used in Deriving the Capital Cost Amortization Factor
Amortization
Scheme


Societal Cost
Depreciation
Life


10 Years
Economic and
Project Life


15 Years
Federal and
State Tax Rate


0%
Return on
Investment
(ROI)

7%
Resulting
Capital
Amortization
Factor
0.11
7.2.3.3
Operating Costs
       The total operating costs are 31 and 30 cents per gallon for years 2004 and 2012,
respectively.  These costs are not directly comparable to those from the USDA model, as fixed
operating cost are included in the operating costs for the NREL model, while the USDA model
accounts for fixed costs in the capital estimate. The operating cost for the NREL analysis
includes items for utilities, reagents,  manpower, insurance, taxes, general administration and
maintenance costs, though do not account for capital costs. Additionally, the sale of the glycerol
byproduct (80% strength) generated income of 4 cents per gallon of produced biodiesel, using
glycerol price of 5 cents per pound. The cost associated with insurance, taxes, general
administration and supplies incur a cost of  2.4 cents per gallon of biodiesel.  The remaining
components of operating costs for the NREL modeling analysis are discussed below.
7.2.3.4
Utility and Labor
         The utility costs were estimated using the energy requirements in the NREL model
along with the same prices for energy, steam and electricity, as those used in our USDA analysis.
The utility requirements per gallon of biodiesel fuel are listed in Table 7.2-11
                                          279

-------
                                     Table 7.2-11.
                      Utility Requirements per Gallon of Biodiesel
Natural Gas, SCF
Medium Pressure Steam, Ibs
Electricity, kWh
Cooling Tower Water, Ibs
2.0
3.2
0.1
8.3
       The NREL model accounts for the salaries of 4 employees per shift to run and maintain
the plant. In addition to salaries for these personnel, the labor expenses also accounted for
employee fringe benefits and the cost for a plant supervisor. The resulting labor costs are 6 cents
for each gallon of biodiesel.
7.2.3.5
Chemical Reagents
       The NREL model also requires the use of the same chemicals and chemical reagents that
are used in the USD A model. The amount of chemical reagents in the NREL model, however,
reflect the use of diluted hydrochloric acid (HC1) and sodium methoxide for the biodiesel
production process.  Hydrochloric acid is listed as being at 33 percent strength, which we
assumed also applied to the strength of sodium methoxide, since the amount of HC1 in the model
is reflective of about one third the value of the USDA's model. For the chemical and reagents
prices, we used the same pricing values as those in our USDA modeling analysis.  The resulting
total chemical and reagent costs on a per gallon basis are about 17 cents for each gallon of
biodiesel fuel produced. All of the required chemicals and reagents for the production of
biodiesel are presented on an undiluted basis in Table 7.2-12.

                                     Table 7.2-12.
                                 Reagent Requirements
Reagent
Water
HCf
Methanol
NAOCH3a
Sodium Hydroxide
Annual
Requirement, Ibs per gallon
3.4646
0.0098
0.6037
0.0338
0.1901
aHCl is Hydrochloric acid, NAOCH3 is sodium methoxide.
   The total biodiesel production costs derived from the NREL model are summarized in Table
7.2-13.
                                          280

-------
                                       Table 7.2-13.
             Projected Production Costs for Biodiesel by Feedstock per Gallon
                                      (2004 dollars)
Marketing Year
2004
2012
Soy Oil3
2.28
2.11
    ""Production consumes 7.87 Ibs of soy oil per gallon of biodiesel. USDA prices in 2012 are adjusted to 2004
    dollars to account for inflation, using GDP index of 109.7 in year 2004 and 130.8 in year 2012.
7.2.4  EIA NEMS Model for Biodiesel

       We also estimated production costs using the biodiesel plant representation in EIA's
NEMS model. This biodiesel model is in spreadsheet format and has the aggregated cost
components for an average soy oil and yellow grease based biodiesel plant. We could not locate
written documentation that describes the basis for these models, though we will assume for our
analysis that it represents an average biodiesel plant.

       EIA's model requires 7.65 Ibs/gal of soy oil and yellow grease feedstock to produce a
gallon of biodiesel fuel. Using the oil feedstock costs as discussed in section 7.2.2.1, feed stock
costs for soy oil are $1.76/gal and $1.61/gal for 2004 and year 2012, respectively, while yellow
grease feedstock costs are $0.88/gal and $0.80/gal for years 2004 and 2012, respectively.

       The EIA model does not provided specific individual cost components for biodiesel
production, though it does have an estimate for total energy, operating and capital costs for  both
plant types. Capital costs are estimated at 14.4 cents/gal in 2004 for both plant types, which we
assumed contains all of costs associated with building a plant, along with the depreciation and
capital payoff costs.   The energy  costs are provided in the model on an aggregated basis and do
not contain the individual amounts of natural gas, electricity and steam used by a plant.  The
model, though, has the total energy needs in year 2004, which are 13.7 c/gal for soy oil and 14.5
cents/gal for yellow grease.  For 2012, we determined the energy costs by adjusting the 2004
aggregate energy cost by the EIA projected price change of natural gas in the Midwest from
2004 to 2012, resulting in an energy cost of 16.0 and  16.9 cents/gal for soy oil and yellow grease,
respectively. All of the other operating costs are represented by an aggregate number, which in
year 2004 is 32.6 and 34.5 cent/gal for soy oil and yellow grease, respectively.  This cost
represents all of the operating costs not associated with energy and capital requirements.  For our
analysis, we assume that this cost does not change in  2012. We used a glycerin price of 5 c/lb,
which generates income and offsets operating cost by 4 c/gal.

       The net production cost for yellow greases (minus feedstock costs) is about 3 c/gal more
than the net production cost for soy based biodiesel, indicating the extra cost incurred for the
yellow grease process. The resulting total production cost are presented in Table 7.2-14
                                           281

-------
                                        Table7.2-14.
              Projected Production Costs for Biodiesel by Feedstock per Gallon
                                       (2004 Dollars)
Marketing Year
2004
2012
Soy Oil3
2.33
2.15
Yellow Grease
1.47
1.38
 "Production consumes 7.65 Ibs of soy oil per gallon of biodiesel. USDA prices in 2012 are adjusted to 2004
 dollars to account for inflation, using GDP index of 109.7 in year 2004 and 130.8 in year 2012.
 7.2.5   Vendor Production Estimates for Biodiesel

        We used engineering vendor estimates as another source to generate the cost to produce
biodiesel fuel. For this, we used engineering details from two firms, Superior Technologies and
PSI-Lurgi Engineering Inc.  These engineering vendors are engaged in the business of designing,
constructing and building biodiesel plants. The biodiesel production processes provided by these
firms are also based on the continuous transesterification process, using sodium hydroxide and
methanol to convert soy oil and yellow grease to biodiesel, along with the finishing processes for
biodiesel and glycerol, similar to the other models.  The vendors generated estimates of the total
cost to build and operate a biodiesel plant, providing the requirements for the equipment, energy,
capital and operating.  We adjusted these estimates to a 2004 year costs basis for comparative
purposes, and used energy costs in the Midwest.

        The vendor estimates we used for PSI-Lurgi are those listed in the report "Economic
Feasibility of Producing Biodiesel in Tennessee"  88.   The biodiesel plant in this analysis was
sized for soy  oil feedstock, based on a 13 MM gallon per year plant, which we assumed is directly
comparable to the 10 MM gallon plant used in the USDA, NREL and EIA models.  In making
this comparison, we relied on a  report89 from Superior Process Technologies to generate the
production costs to make biodiesel from soy oil and yellow grease.  This report has the various
cost components for a 10 MM gallon per year plant.

        The total soy oil based biodiesel production cost estimate for year 2004 is $2.20/gal and
$2.34/gal for PSI-Lurgi and Superior Technology, respectively with an average cost of $2.27/gal.
For 2012, we project the soy oil production cost from both vendors would average approximately
$2.09 gal.  The Superior Technology's yellow grease biodiesel production cost is $1.66/gal in
2004.
 88 "Economic Feasibility of Producing Bio-diesel in Tennessee" AIM-AG Agri-Industry Modeling & Analysis
 Group, 2002

 89 Superior Process Technologies, " Biodiesel Plant Economics and Process Description", 8/18/06


                                            282

-------
 7.2.5.1        PSI-Lurgi estimate of Biodiesel Costs

        The feedstock costs assume that 7.569 pounds of soy oil are required to make a gallon of
biodiesel fuel.  The resulting feedstock cost is 174.1 cents/gal for year 2004 . We adjusted all
other cost items in the PSI estimate from a 2002 year to 2004 basis, scaling by the relative change
of GDP, though the costs for reagent needs and utilities are adjusted using the methods as
discussed in the following sections.

 7.2.5.2        Capital Costs

        The total capital cost provided in the estimate account for all of the costs for building a
 plant, though excluding maintenance costs, similar to the capital requirements in the NREL
 model. The total capital coats for a plant are $19.7 million in 2004 dollars, which we adjusted
 from 2002 dollars.  The capital costs were amortized assuming a seven percent return on
 investment, resulting in an annualized cost of 16.7 cents per gallon.  The economic factors used
 for amortizing the capital costs are the same as those listed in Table 7.2-8.

 7.2.5.3        Operating Costs

        The total operating costs are 29 cents per gallon for year 2004, excluding capital charges.
 The cost associated with insurance, taxes and general administrations is 7.3 cents per gallon,
 while the cost for maintenance is 2 cents/gal.  The sale of the glycerol byproduct at 80%
 strength generates incomes of 4 cents per gallon of produced biodiesel, assuming a glycerol price
 of 5 cents per pound. The remaining components of operating costs are discussed below.

 7.2.5.4        Utility and Labor

          The utility costs were estimated using the energy requirements presented in Table 3.1
 of the report, along with the same prices for energy, steam and electricity, as those used in our
 2004 model analysis. The  total utility requirements are 5.5 cents per gallon. The utility
 requirements per gallon of biodiesel fuel are listed in Table 7.2-15.

                                       Table 7.2-15.
                        Utility Requirements per Gallon of Biodiesel
Natural Gas, SCF
Medium Pressure Steam, Ibs
Electricity, kWh
Cooling Tower Water, Ibs
0
3.92
0.093
200.7
        The PSI-Lurgi estimate accounts for the salaries of 4 employees per shift to run and
 maintain the plant. In addition to salaries for these personnel, the labor expenses also accounted
 for employee fringe benefits.  The resulting labor costs are 6.3 cents for each gallon of biodiesel.
 In addition to these costs, the SG&A expenses are estimated at 6.3 c/gal.
                                            283

-------
 7.2.5.5
Chemical Reagents
        The PSI-Lurgi estimate also requires the same chemicals and chemical reagents in the
 USDA/NREL models. We assumed that the hydrochloric acid and the sodium methoxide used
 in the PSI estimate is 33% strength, as the prices listed in the study are reflective of being on a
 diluted basis. We also assumed that the price for the amount of caustic soda required is on an
 undiluted basis. We adjusted the prices for the chemical and reagents using the 2004 year
 pricing values used in our USDA/NREL modeling analysis, though the price of phosphoric acid
 was adjusted using the GDP index.  The resulting total chemical and reagent costs on a per
 gallon basis are about 12.6 cents for each gallon of biodiesel fuel produced. All of the required
 chemicals and reagents for the production of biodiesel are presented on an undiluted basis in
 Table 7.2-16.

                                       Table 7.2-16.
                                  Reagent  Requirements
Reagent
Water
HCla
Methanol
NAOCH3a
Caustic Soda / Sodium Hydroxide
Phosphoric Acid
Annual Requirement,
Ibs per gallon
1.666
0.026
0.700
0.038
0.04
0.013
 aHCl is Hydrochloric acid, NAOCH3 is sodium methoxide.
        The total soy oil based biodiesel production costs derived from the PSI-Lurgi are $2.20
 dollars in year 2004.

 7.2.6   Superior Process Technologies Estimate

        The Superior feedstock costs assume that 7.60 pounds of soy oil and yellow grease are
required to make a gallon of biodiesel fuel, which results in a feedstock cost of 175 cents/gal for
year 2004 for soy oil and 87.5 c/gal for yellow grease, using the feedstock costs in section 7.2.2.1.
We adjusted all other cost items in the Superior estimate from a 2006 year to 2004 basis,
adjusting the cost by the relative change of GDP, though the costs for reagent needs and utilities
are adjusted using the methods as discussed below.
 7.2.6.1
Capital Costs
        The total capital coats for a soy oil based plant are $10.7 million, while the costs for a
 yellow grease plant are $13.4 million, in 2004 dollars. These costs are inclusive of the amount
 needed for a new plant, which is similar to the other biodiesel estimates.  The capital costs were
 amortized assuming a seven percent return on investment, resulting in a cost of 11.7 cents per
 gallon and 14.7 cents per gallon for a soy oil and yellow grease based plant, respectively.
 7.2.6.2
Operating Costs
                                           284

-------
       The total operating costs, excluding capital charges are 47 and 64 cents per gallon for soy
oil and yellow grease plants in year 2004, respectively. Insurance, taxes, rent and local taxes
incur a cost of about 6 and 7 cents per gallon, for soy oil and yellow grease, respectively. The
costs for maintenance, plant overhead costs and supplies account for about 16.0 and 21.5
cents/gal, for soy oil and yellow grease.  The sale of the glycerol byproduct at 80% strength
generates incomes of 4 cents per gallon of produced biodiesel, assuming glycerol price of 5 cents
per pound.  The remaining components of operating costs are discussed below.

7.2.6.3        Utility and Labor

       The overall utility costs were provided, though the specific amounts of natural gas,
electricity were not provided for the 10 MM gallon plant.  We adjusted the Superior utility cost
estimate from a 2006 year to a 2004 year basis, using the relative price change of natural gas and
electricity, assuming that natural gas supplies 90 percent of the energy,  and electricity supplies
the remaining 10 percent. The resulting energy requirement is 5.0 and 8.6 cents per gallon, for
soy oil and yellow grease, respectively.

       The estimate accounts for personnel cost to run and maintain the plant, including
laboratory, plant supervisory and administration  costs. The overall labor costs on a 2004 year
basis are 7.6 and 12.8 cents for each gallon of biodiesel, for soy oil and yellow grease,
respectively.

7.2.6.4        Chemical Reagents

       The Superior vendor estimate also requires the same chemicals and chemical reagents as
used in the USDA/NREL models and Tennessee study. The total chemical reagent cost from
Superior on a 2006 year reagent pricing basis is 18.1 and 20.1 c/gal, respectively for soy and
yellow grease plants. Superior provided the prices for each of the chemicals, though the specific
amounts of each chemical were not provided for the soy oil based estimate.  We therefore,
adjusted the total chemical reagent cost to a 2004 year basis, assuming the demands for reagent
as documented in the PSI-Lurgi estimate.  The resulting reagent costs on a 2004 year basis are
16.5 and 18.3 cents/gal, for soy oil and yellow grease.  The Superior prices for the required
chemicals and reagents are presented in Table 12-11.

                                      Table 7.2-17.
                                Superiors Reagent Prices
Reagent
NAOCH3 a (25% solution)
HC1 a (32%)
Methanol
Phosphoric Acid (75% )
Caustic Soda / Sodium Hydroxide (50%)
Dollar /lb
0.50
0.091
0.146
0.42
0.14
aHCl is Hydrochloric acid, NAOCH3 is sodium methoxide.
       The total resulting biodiesel production costs derived from Superior Process description
and engineering estimate is $2.34 and $ 1.66 per gallon for soy oil and yellow grease derived
                                           285

-------
biodiesel respectively, in year 2004.  Yellow greases net production cost (minus feedstock costs)
are about 20 c/gal more than the net production cost for soy based biodiesel.
7.2.7   Yellow Grease Production Costs

       Yellow grease's production cost is higher than soy oil produced biodiesel fuel, due to the
extra capital and operating costs required to remove contaminants in the grease feedstock.  In the
prior sections, the EIA and Superior analysis indicated that the yellow grease production cost for
biodiesel is higher than the production cost based on use of soy oil, excluding the feedstock
costs. The EIA analysis showed that yellow grease's production cost is 3 c/gal higher, while the
Superior results showed that the yellow grease's production cost is about 20 c/gal higher than
soy oil production costs. Both of these provide a measurement of the extra production costs
(excluding feedstock) associated with making biodiesel from yellow grease versus soy oil.

       In this section, we use the EIA and Superior results to generate yellow grease costs as
inputs to the models, adjusting the soy oil production costs to reflect the extra cost for producing
yellow grease. We assume the same feedstock costs for yellow grease as those listed in section
7.2.2.1, and that it takes 7.6 Ibs of yellow grease to produce a gallon biodiesel fuel.  Table 7.2-
18, contains the resulting yellow grease production costs based on the EIA and Superior
analyses.

                                      Table 7.2-18.
                 Yellow Grease Costs Based on EIA and Superior Results

USD A, c/gal
NREL, c/gal
EIA, c/gal
Vendor avg, c/gal
Average
EIA 2004
124
138
143
142
136.8
Superior 2004
141
155
160
159
153.8
EIA 2012
111
131
132
139
128.3
Superior 2012
128
148
149
156
145.2
       We averaged all of the yellow grease results, and generated an average production cost of
$1.45/gal in 2004 and $1.37/gal in 2012 for yellow grease derived biodiesel.

7.2.8   Biodiesel Blending Credit Programs

       There are numerous credit and incentive programs that encourage the blending of
biodiesel.  These programs reimburse blenders and producers for adding biodiesel to transport
diesel fuel, which acts to lower the production costs and makes the production of biodiesel more
economically competitive with petroleum derived diesel fuel. There are several
federal/nationwide biodiesel credit programs that offer subsidies for blending or use of biodiesel
as a transport diesel fuel which are discussed below.

       The Commodity Credit Commission Bio-energy Program is an existing program that
expires at the end of fiscal year 2006, though  due to a funding shortfall the program will
terminate on July 31, 2006. This program was administered by the USDA and pays biodiesel
producers grants when the economics to produce biodiesel are poor.  The stipend is determined
                                           286

-------
based on available funding and the volume of renewable fuel that can receive the credit. For
historical purposes, the payments in 2004 and 2005 averaged about 107 and 50 c/gal of fuel
produced, respectively. For the first half of 2006, the credit on a per gallon basis is reduced
further, as the payment is diluted by increased production volume of fuels available to receive
the credit.

       The Energy Act extended the Biodiesel Blenders Tax Credit program to the end of year
2008. This program was created under the American Jobs Creation Act of 2004 which created an
excise tax credit that can be claimed by anyone who blends biodiesel into transport diesel fuel.
Under this program, blenders may claim a credit against the applicable federal motor fuels excise
tax for blends containing biodiesel.  According to IRS guidelines, the credit may be claimed by
anyone who adds biodiesel into diesel fuel at a level greater than 0.1 percent in the final blend.
The full credit for biodiesel made from virgin vegetable oils and animal fats is $1.0 per gallon,
while biodiesel  derived from recycled grease receives 50 cents/gallon. A blender with more
excise tax credits than taxes owed can receive a refund from the IRS. Additionally, under the
current program, imported biodiesel and fuel made from imported feedstocks can also receive the
credit.

       The Income Tax Credit Alternative is a program that is also available. This program does
not require any blending of biodiesel, though it does offers allow a similar excise tax credit as in
the blenders tax credit program. The excise tax can only be taken against actual income,
however, which makes the program less economically attractive than the blenders' credit
program.

       The Energy Act also created the Small Biodiesel Blenders Tax credit program. Under this
program, a credit of 10 c/gal is available to small producers who make biodiesel fuel from virgin
vegetable oils. This stipend is limited to companies with annual production volumes less than 60
MM gallons per year, using the aggregated capacity from all production sites for an individual
company. The maximum payment per company is capped at $15 MM per year and the program
is set to expire at the end of year 2008.

       In addition to the federal programs, there are state and local programs that offer state fuel
tax exemptions, tax credits, and incentives that are more modest.
7.3    Distribution Costs

7.3.1.  Ethanol Distribution Costs

       There are two components to the costs associated with distributing the volumes of
ethanol necessary to meet the requirements of the Renewable Fuels Standard (RFS): 1) the
capital cost of making the necessary upgrades to the fuel distribution infrastructure system, and
2) the ongoing additional freight costs associated with shipping ethanol to terminals. The most
comprehensive study of the infrastructure requirements for an expanded fuel ethanol industry
was conducted for the Department of Energy (DOE) in 2002  wwwww  This study provides the
foundation for our estimates of the capital costs associated with upgrading the distribution
                                          287

-------
infrastructure system as well as the freight costs to handle the increased volume of ethanol
needed in 2012.  Distribution costs are evaluated here for the RFS and EIA cases.

7.3.1.1        Capital Costs to Upgrade the Ethanol Distribution System

       The 2002 DOE study examined two cases regarding the use of renewable fuels.  The first
case assumed that 5.1 Bgal/yr of ethanol would be used in 2010, and the second case assumed
that 10 Bgal/yr of ethanol would be used in 2015.  We interpolated between these two cases to
provide the foundation for our estimate of the capital costs to support the use of 6.67 Bgal/yr of
ethanol in 2012 (the RFS case). The 10 Bgal/yr case from the DOE study was used as the
foundation to estimate the  capital costs under the EIA case. For both the 6.67 Bgal/yr and 9.64
Bgal/yr cases, we adjusted the  results from the DOE study to reflect a 3.9 Bgal/yr 2012 ethanol
use baseline.  Table 7.3-1 contains our estimates of the infrastructure changes and associated
capital costs for the two ethanol use scenarios examined in today's rule.90
90 These capital costs will be incurred incrementally during the period of 2007-2012 as ethanol volumes increase.
For the purpose of this analysis, we assumed that all capital costs will be incurred in 2007.
                                           288

-------
                                          Table 7.3-1.
                        Ethanol Distribution Infrastructure Capital Costs
                     Relative to a 3.9 Billion Gallon per Year Reference Case

New Terminal Blending Systems for Ethanol
Number of terminals
Capital cost
New Ethanol Storage Tanks at Terminals
Number of tanks
Capacity
Capital cost
Terminal Storage Tanks Converted to Ethanol
Number of tanks
Capacity
Capital cost
Terminals Using Ethanol for the First Time3
Number of terminals
Capital cost
New Rail Delivery Facilities at Terminals
Number of terminals
Capital cost
Retail Facilities Using Ethanol for First Time3
Number of retail facilities
Capital cost
New Tractor Trailer Transport Trucks
Number of Trucks
Capital Costs
New Barges
Number of new barges
Capital cost
New Rail Cars
Number of new rail cares
Capital cost
Total Capital Costs
Capital Costs Attributed to Terminal and Retail (i.e. fixed)
Facilities
Capital Costs Attributed to Mobile Facilities (tank trucks,
rail cars, & barges)
RFS Case
(6.67 Bgal/yr)

243
$73,044,000

168
1,526,000 barrels
$21,939,000

44
3 19,000 barrels
$931,000

212
$4,238,000

42
$14,869,000

33,600
$19,824,000

209
$24,027,000

11
$21,475,000

2,024
$172,012,000
$352,361,000
$134,847,000
$217,514,000
EIA Case
(9.64 Bgal/yr)

515
$154,530,000

370
3, 4 15,000 barrels
$48,803,000

83
592,000 barrels
$1,739,000

453
$9,065,000

76
$27,127,000

74,820
$44,146,000

435
$50,075,000

23
$43,204,000

3,491
$296,729,000
$675,418,000
$285,410,000
$390,008,000
3 Terminal and retail facilities using ethanol for the first time will need to make various modifications to ensure the
compatibility of their systems with ethanol.
       Our estimated capital costs in this final rule differ from those in the proposal for several
reasons.  First, the volume for the RFS case was updated to reflect the fuel rule provisions.
Second, we adjusted our estimate of capital costs from those in the proposal to reflect an increase
in the cost of rail tank cars and barges since the DOE study was conducted.  Third, we are
assuming a 30 percent increase in the reliance on rail versus marine transport over that projected
in the DOE study.  The 2002 DOE study estimated that 53 percent of the increase in ethanol
volume shipped between PADDs would be carried by barge and 47 percent by rail.  For the
purposes of this analysis, we assumed that 30 percent of the increased volume in ethanol
                                           289

-------
shipments that were projected to be carried by barge in the DOE study would instead be carried
by rail.  This equates to 37 percent of the increase in ethanol shipments being carried by barge
and 63 percent by rail. To provide a conservatively high estimate of the potential economic
impact, we assumed that this shift translates into a 30 percent increase in rail infrastructure costs.
This incorporates the increased cost to prepare additional terminals to receive ethanol by rail, and
to provide a sufficient number of additional rail  tank cars for ethanol transport. The actual
increase in rail infrastructure costs may be somewhat lower given improvements in the efficiency
of ethanol transport by rail.

       Amortized over 15 years at a 7 percent cost of capital, the total capital costs (of
$352,361,000 under the RFS case and 675,418,000 under the EIA case) equates to an annual cost
of approximately $38,687,000 under the RFS case and $74,157,000 under the EIA case. This
translates to approximately 1.4  cents per gallon of new ethanol volume under the RFS case and
1.2 cents per gallon under the EIA case. Under both cases, approximately 0.5  cents per gallon is
attributed to mobile facilities and the remainder to fixed facilities.
7.3.1.2
Ethanol Freight Costs
       The 2002 DOE study contains estimated ethanol freight costs for each of the 5 PADDs.
These estimated costs are summarized in the following Table 7.3-2,xxxxx  A map of the PADDs
is contained in Figure 7.3-1.

                                      Table 7.3-2.
               Estimated Ethanol Freight Costs from the 2002 DOE Study
PADD
1
2
3
4
5
National Average
5.1 billion gallons per year
(cents per gallon)
11.1
4.3
6.6
4.7
12.7
7.7
10.0 billion gallons per year
(cents per gallon)
7.2
2.4
5.8
7.4
10.7
5.7
                                          290

-------
                                      Figure 7.3-1.
                                PAD District Definitions.
  Petroleum Administration for Defense {PAD} Districts
       The Energy Information Administration (EIA) translated the cost estimates from the 2002
DOE study to a census division basis.YYYYY  A summary of the resulting (EIA) ethanol
distribution cost estimates are contained in the following Table 7.3-3.  A map of the census
divisions is contained in Figure 7.3-2.
                                          291

-------
           Table 7.3-3.
EIA Estimated Ethanol Freight Costs
 (derived from the 2002 DOE Study)
Census Division
From
East North Central
East North Central
East North Central
East North Central
East North Central
East North Central
West North Central
West North Central
West North Central
West North Central
West North Central
West North Central
West North Central
West North Central
West North Central
To
New England
Middle Atlantic
East North Central
South Atlantic
East South Central
Pacific
New England
Middle Atlantic
East North Central
West North Central
South Atlantic
East South Central
West South Central
Mountain
Pacific
Freight Cost
(cents per gallon)
9.8
9.8
4
9.8
4.7
14.0
11.4
11.4
4
4
11.4
4.7
4.7
4.5
13.0
            292

-------
                                          Figure 7.3-2.
                                        Census Divisions
                                                         Enst
                                     3*u1H Cttrtnri  SGUtti Cwitral

                                                        SOUTH

Allan*
       We took the EIA projections and translated them into State-by-State ethanol freight costs.
For the purposes of this analysis, all ethanol was assumed to be produced in the East and West
North Central Census Divisions (corresponding closely to PADD 2). We believe that this is a
reasonable approach because the cost of shipping corn feedstock from PADD 2 to ethanol plants
located outside of PADD 2 will typically negate any potential reduction in freight cost from
reduced shipping distances for ethanol or dried distiller grains.  The vast majority of ethanol
plants planned for outside of PADD 2 are projected to begin operation using corn supplied from
PADD 2.91  Many have stated plans to transition to local feedstocks.  However, we believe that
such a transition will typically not be accomplished within the timeframe considered by this
91 Hawaii is a special case because plants potentially located there will use local feedstocks from their initial start up
date.
                                           293

-------
analysis (i.e. 2012). Other local considerations may provide a unique cost advantage to locating
a plant outside of PADD 2. One such consideration might be that if by locating a plant outside
of PADD 2 the ethanol producer could avoid the need to dry the distiller grains it produces and
sell the wet distiller grains to a local market. Although this might result in a significant cost
savings, it is unclear the extent to which this will be possible given the short shelf life of wet
distiller grain (~3 days). Also, any potential cost savings might be offset by the relatively lower
price that can be negotiated for wet versus dry distiller grains.  In any event, there is insufficient
data at this time to evaluate the extent to which such local conditions may result in an advantage
in lower freight costs for ethanol plants outside of PADD 2. Further, our projection of where
new ethanol production plants might be located indicates that only 10 percent of production
capacity could be located outside of PADD 2. Thus, any potential freight cost advantage that
might be enjoyed by such  plants would not likely have a significant impact on our national
analysis.  Furthermore, to  the extent that the location of ethanol plants outside of PADD 2
imparts a savings in ethanol distribution costs, this would suggest that our estimates of ethanol
freight costs in this rule are conservatively high.

       Ethanol consumed within census divisions belonging to PADD 2 was assumed to  be
transported by truck, while distribution outside of these areas was assumed to be by rail, ship,
and/or barge. A single average distribution cost for each destination census division was
generated by weighting together the 2012 freight costs given for each mode in both source
census divisions according to their volume share. These cents per gallon figures were first
adjusted upward by 10 percent to reflect the increased cost of transportation fuels used to ship
ethanol since the 2002 DOE study, and then additional adjustments were applied to some
individual states based on  their position within the census division.  In the case of Alaska and
Hawaii, differences in ethanol delivery prices from the mainland were inferred from gasoline
prices.

       For some states, different freight costs for ethanol supplied to large hub terminals  versus
small satellite terminals was estimated. The reasoning behind this is that large shipments of
ethanol shipped from the Midwest by barge, ship, and/or unit train will often be initially
unloaded at hub terminals  for further distribution to satellite terminals. In cases where
redistribution from a hub to a satellite terminal doesn't take place, the volume of ethanol shipped
directly from the producer to a lesser volume ("satellite") terminal will also incur a higher freight
rate than ethanol shipped to a larger-volume "hub" terminal. The largest adjustment was  applied
to the Rocky Mountain States since they are generally large in area and additional expense is
required to transport freight through higher elevations and rugged terrain.  Smaller adjustments
were applied to states  that are smaller, flatter, or have access by navigable waterways. The states
to which an adjustment was not applied were generally in the Midwest or were so small as to not
warrant different distribution costs.  Given the large number of ethanol plants in the Midwest, we
do not believe that there are substantial differences  in the cost of distributing ethanol with that
area.

       We made several adjustments in our estimates of ethanol freight costs from those  in the
proposal. First, the differential cost of shipping ethanol to satellite terminals versus hub
terminals was increased to better reflect the additional costs incurred in either redistributing the
ethanol from a hub to  a satellite terminal, or of shipping  ethanol directly from the producer to the
                                           294

-------
satellite terminal in a lesser volume. The estimated additional freight cost of shipping ethanol to
satellite terminals versus hub terminals is contained in the following Table 7.3-4.

                                       Table 7.3-4.
             Additional Freight Cost to Deliver Ethanol to a Satellite Terminal
                              Compared to a Hub Terminal
States
OH
AL, AR, FL, GA, KY, LA, MD, ME, MS, NC, NH, NY,
OK, OR, PA, SC, TN, VA, VT, WA, WV
AK, AZ, CO, ID, MM, NV, TX, UT, WY
cents per gallon
2
4

5
       Another change that we made from the proposal was with respect the volume of ethanol
we estimated would be delivered to hub versus satellite terminals. The proposal assumed a 50/50
spit.  For this final rule, we project that all of the ethanol volume blended into reformulated
gasoline would be used in urban areas served by hub terminals. The percentage of ethanol
blended into conventional gasoline that is used in an urban area (and hence delivered to a hub
terminal) versus that used in a rural area (and hence delivered to a satellite terminal) was based
on our analysis of the percentage of vehicle miles traveled in urban versus rural areas.92

       The final change from the proposal pertains to our consideration of the cost of shipping
ethanol from the production plant to the rail head / marine terminal either for large volume
shipment by unit train or marine shipment to hub terminals, or for shipment at single car rates via
multiple-product trains directly to satellite terminals. Our review of current ethanol  freight rates
conducted in response to a comment on the proposed rule indicates that we did not adequately
account for this added cost in the proposal. Chicago is a primary ethanol gathering point from
producers for further distribution. A 4 cent per gallon conveyance fee is charged to account for
delivery of ethanol from the production plant gate to the Chicago Board of Trade delivery point
for taking ethanol. This includes train shipments, loading costs, and other miscellaneous fees.
Based on this information, we have added 4 cents per gallon to our ethanol freight estimates.

       Our estimates of the State-by-State ethanol freight costs under the RFS and EIA cases are
contained in Tables 7.3-5 and 7.3-6.  National and PADD average freight  costs under both the
RFS and EIA cases are contained in Table 7.3-7. We are assuming that these freight costs do not
include the costs associated with the recovery of capital for the distribution facility changes that
are necessary to accommodate the increased volume of ethanol.  This may tend to overstate
distribution costs to some extent because some capital recovery may be incorporated into the 4
cent  per gallon conveyance fee. The inclusion of rail tank car lease fees also suggests that these
estimated freight costs may be conservatively high given that rail car lease fees incorporate a
capital recovery and profit margin.
92 See Chapter 2 of this RIA for additional discussion of our estimate of the percentage of ethanol that will be used
in urban versus a rural areas
                                           295

-------
                                    Table 7.3-5.
                       State-by-State Ethanol Freight Costs
State
Connecticut
Maine
Massachusetts
New Hampshire
Rhode Island
Vermont
New Jersey
New York
Pennsylvania
Delaware
District of Columbia
Florida
Georgia
Maryland
North Carolina
South Carolina
Virginia
West Virginia
Illinois
Indiana
Michigan
Ohio
Wisconsin
Iowa
Kansas
Minnesota
Missouri
Nebraska
North Dakota
South Dakota
Kentucky
Tennessee
Oklahoma
PADD


















2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Ethanol Freight Cost*: (cents per gallon)
Hub
Terminal
15.4
17.4
15.4
16.4
15.4
16.4
15.4
15.4
12.4
15.4
15.4
12.4
15.4
15.4
15.4
15.4
15.4
15.4
4.4
5.4
6.4
5.4
4.4
3.4
4.4
4.4
4.4
4.4
5.4
4.4
6.2
6.2
8.3
Satellite
Terminal
15.4
21.4
15.4
16.4
15.4
16.4
15.4
19.4
16.4
15.4
15.4
16.4
19.4
15.4
19.4
19.4
19.4
19.4
4.4
5.4
6.4
7.4
4.4
3.4
4.4
4.4
4.4
4.4
5.4
4.4
10.2
10.2
12.3
Average Freight Cost"
RFS Case
15.4
20.0
15.4
16.4
15.4
NA***
15.4
15.4
13.8
15.4
NA*"
14.0
NA***
NA***
NA***
NA***
NA***
NA***
4.4
5.4
6.4
6.6
4.4
3.4
4.4
4.4
4.4
4.4
5.4
4.4
7.1
8.7
NA***
EIA Case
15.4
20.0
15.4
16.4
15.4
16.4
15.4
15.7
13.7
15.4
15.4
14.0
17.4
15.4
NA***
NA***
15.4
NA***
4.4
5.4
6.4
6.6
4.4
3.4
4.4
4.4
4.4
4.4
5.4
4.4
7.3
8.7
11.1
* Freight rates from PADD 2 production facilities
** Hub and satellite freight rates were volume weighted to arrive at
*** No significant ethanol use.  See Chapter 2 of this RIA regardinj
used.
an average freight rate.
; our estimates of where ethanol will
be
                                         296

-------
                                        Table 7.3-5.
                     State-by-State Ethanol Freight Costs (continued)
State
Alabama
Mississippi
Arkansas
Louisiana
Texas
New Mexico
Colorado
Idaho
Montana
Utah
Wyoming
Arizona
Nevada
Alaska
Hawaii
Oregon
Washington
California
PADD
3
3
3
3
3
3
4
4
4
4
4
5
5
5
5
5
5
5
Ethanol Freight Cost*: (cents per gallon)
Hub
Terminal
11.2
10.2
11.3
11.3
14.3
16.4
14.4
19.4
17.4
17.4
16.4
19.4
20.4
45.5
40.5
20.5
20.5
20.5
Satellite
Terminal
15.2
14.2
16.3
15.3
19.3
21.4
19.4
24.4
22.4
22.4
21.4
24.4
25.4
50.5
40.5
24.5
24.5
20.5
Average Freight Cost"
RFS Case
14.0
13.9
15.0
13.8
14.3
19.0
NA*"
NA*"
21.8
NA*"
NA*"
20.3
21.1
NA*"
NA*"
20.5
21.9
20.5
EIA Case
14.0
13.9
15.0
13.8
16.0
19.6
17.0
22.8
21.8
20.5
20.8
20.5
22.3
NA*"
40.5
20.5
21.9
20.5
* Freight rates from PADD 2 production facilities.
** Hub and satellite freight rates were volume weighted to arrive at the average freight rate.
*** No significant ethanol use.  See Chapter 2 of this RIA regarding our estimates of where ethanol will be used.
                                        Table 7.3-6.
                    National and PADD Average Ethanol Freight Costs

National Average
PADD1
PADD 2
PADD 3
PADD 4
PADD 5 excluding AK & HI
PADD 5 including AK & HI
Ethanol Freight Cost* (cents per gallon)
RFS Case
11.3
14.9
5.1
14.6
21.8
20.6
20.6
EIA Case
11.9
15.1
5.3
15.2
19.8
20.7
22.1
        * Freight rates from PADD 2 production facilities.
       The national average ethanol freight cost of 11.3 cents per gallon under the RFS case and
11.9 cents per gallon under the EIA case translates to an annual freight cost for the additional
volume of ethanol used in 2012 of $313,123,000 and $678,300,000 respectively. Adding in the
annualized capital costs, results in a total annual ethanol distribution cost of 351,810,000 or 12.7
cents per gallon under the RFS case and $752,457,000 or  13.1 cents per gallon under the EIA
case.93
                                            297

-------
7.3.2  Biodiesel Distribution Costs

       The volume of biodiesel used by 2012 under the RFS is estimated at 300 million gallons
per year.  The 2012 reference case against which we are estimating the cost of distributing the
additional volume of biodiesel needed to meet the requirements of the RFS is 30 million
gallons.94

       The capital costs associated with distribution of biodiesel are higher per gallon than those
associated with the distribution of ethanol due to the need for storage tanks, blending systems,
barges, tanker trucks and rail cars to be insulated and in many cases heated during the winter
months.95 In the proposal, we estimated that that these capital costs would be approximately
$50,000,000. We adjusted our estimate of these capital costs for this final rule based on
additional information regarding the cost to install the necessary storage and blending equipment
at terminals and the need for additional rail tank cars for biodiesel.96  We now estimate that
handling  the increased biodiesel volume will require a total capital cost investment of
$145,500,000 which equates to about 6 cents per gallon of new biodiesel volume.97

       In the proposal, we estimated that the freight costs for ethanol adequately reflect those for
biodiesel  as well. In response to comments, we sought additional information regarding the
freight costs for biodiesel.  This information indicates that freight costs for biodiesel are typically
30 percent higher than those for ethanol which translates into an estimate of 15.5 cents per gallon
for biodiesel freight costs.98

       Including the cost of capital recovery for the necessary distribution facility changes, we
estimate the cost of distributing biodiesel to be 21.5 cents per gallon.
93 All capital costs were assumed to be incurred in 2007 and were amortized over 15 years at a 7 percent cost of
capital.

94 See Chapter 1 of this RIA regarding the 2012 reference case.

95 See Chapter 1.3 of the Regulatory Impact Analysis associated with today' s rule for a discussion of the special
handling requirements for biodiesel under cold conditions.

96 Information on biodiesel facility costs was obtained from a number of biodiesel blenders on the condition that the
specific source of such information would not be identified.  Biodiesel rail tank cars typically have a capacity of
25,500 gallons as opposed to 30,000 gallons for an ethanol tank car. Thus, additional tank cars are need to transport
a given volume of biodiesel relative to the same volume of ethanol.

97 Capital costs will be incurred incrementally over the period of 2007-2012 as biodiesel volumes increase. For the
purpose of this analysis, all capital costs were assumed to be incurred in 2007 and were amortized over 15 years at a
7 percent cost of capital.

98 This is based on our review of publicly available biodiesel and ethanol freight rates from CSX and BNSF rail at
www.csx.com and www.bnsf.com, on information regarding the lease rates for biodiesel versus ethanol freight cars
considering the smaller size of biodiesel tank cars, and on discussions with biodiesel distributors. The estimated
ethanol freight costs were increased by 30 percent to arrive at the estimate of biodiesel freight costs.


                                              298

-------
7.4    Gasoline and Diesel Blendstock Costs

       The previous sections of this chapter have presented estimates of the cost of producing
and distributing ethanol and biodiesel. In this section, we summarize the results of refinery
modeling conducted by Jacobs Consultancy under contract to EPA. Jacobs's used the Haverly
Linear Programming (LP) model to conduct the analysis. This model is widely used by the
refining industry, consultants, engineering firms and government agencies to analyze refinery
economics, refinery operations, fuel quality changes, refinery capital investments,  environmental
changes and demand changes. The Haverly model uses Jacobs's Refining Process Technology
Database to represent refining operations

       The modeling was conducted to analyze the effect of the increased renewable fuel use on
the production costs and composition of the nation's gasoline and diesel fuel.  The refinery
modeling output described in this section includes the changes in volumes and capital
investments as well as the resulting capital  and fixed operating costs, the variable costs, and the
total of all these costs. The costs are expressed in 2006 dollars and capital costs are amortized at
7% before tax return on investment (ROI).  The costs for the RFS and EIA cases are expressed
incremental to the reference case. We first report the results of the RFS case, followed by the
results of the EIA case.

7.4.1   Description of Refinery Modeling Cases Modeled

       The modeling cases were set up to analyze the RFS and EIA cases described in Chapter
2.  The primary renewable fuel modeled was  ethanol in gasoline, while considering a fixed
production amount of biodiesel as projected by EIA in 2012.  Along with the increased use of
renewable fuels, the analyses for the RFS and EIA cases both include the elimination of the RFG
oxygen content standard and the resulting removal of MTBE from the U.S. gasoline market.
These scenarios both assume the current Mobile Source Air Toxics standard (MSAT1) is in
place. The effects of the MSAT2 standard are modeled in that rulemaking which has just recently
been made final.

       Jacobs conducted a Linear Programming (LP) modeling analysis of the refining industry
for the various RFS scenarios using a model developed by Haverly's LP technology. The
modeling was set up to analyze the extent to which ethanol will be used in CG versus RFG by
region and the resulting effects on gasoline composition. The refining industry was modeled
based on five aggregate complex refining regions, representing PADD's 1, 2, 3,  4 & 5 together
minus California and  California separately. All of the PADDS were modeled simultaneously
together in the LP model, in order to balance  and meet the national gasoline and fuel demands.

7.4.1.1        RVP

       The analysis modeled summer and winter  seasons, with all gasoline types including
California RFG, Federal RFG, 7.0, 7.8 RVP controlled areas and 9.0 CG. The control cases
consisted of the minimum renewable fuel volume as specified by EPAct and discussed in
Chapter 2 and the 2006 AEO projection of 9.6 billion gallons of ethanol per year in 2012.
                                          299

-------
Winter gasoline RVP levels were adjusted higher than EPA data, to account for refiner RVP
reporting inaccuracies from use of the complex model in the winter season.  (Some refiners
reported lower RVP levels than actually produced, as the complex model has a fixed upper
reporting limit of 8.8 in the winter season.)
7.4.1.2
Base case (2004)
       The base case was established by modeling fuel volumes for 2004. Information was
based on process capacities from Oil and Gas Journal, EIA data and gasoline emissions and
property data from EPA. Fuel property data for this base case was built off of 2004 refinery
batch reports provided to EPA; however, the base case assumed sulfur standards based on
gasoline data in 2004, not with fully phased in of Tier 2 gasoline standards at the 30 ppm level.
In addition we assumed the phase-in of 15 ppm sulfur standards for highway, nonroad,
locomotive and marine diesel fuel.  The supply/demand balance for the U.S. was based on
gasoline volumes from EIA and the California Air Resources Board (CARB). Our decision to
use 2004 rather than 2005 as the baseline year was because of the refinery upset conditions
associated with the Gulf Coast hurricanes in 2005.
7.4.1.3
MSAT1 Provisions for Refinery Cases in 2012
       For CG and RFG, gasoline qualities were modeled to assure Complex model Phase 2
calculations  seasonal and annual compliance, taking into account the elimination of the oxygen
requirement for RFG; (by PADD and California), and under MSAT1 gasoline standards.
Incremental gasoline volumes above the 2004 base case volumes for each PADD were allowed
to conform to less stringent toxics performance standards as allowed by the MSAT1 provisions.
For this, the MS AT 1  PADD constraints were calculated using gasoline data from 1998-2000
EPA batch reports, considering that new incremental volumes of gasoline above the 1998-2000
annual average would comply with MSAT1 provisions, as predicated by EPact 2005. The
following tables show the resulting conventional and RFG gasoline MS ATI baseline constraints,
which was applied to  gasoline produced for the cases modeled in year 2012.
                                     Table 7.4-1.
                   Conventional Gasoline MSAT 1
                                     2012 Baseline Data

PADD 1
PADD 2
PADD 3
PADD 4 & 5, excluding
California
Exhaust Toxics mg/mi*
88.33
92.79
88.79
99.85
NOx mg/mi*
1,440.84
1,432.57
1,438.76
1,414.00
*mg/mi is milligram per mile.
                                         300

-------
                                     Table 7.4-2.
                      RFG Gasoline MSAT 1 2012 Baseline Data

PADD 1
PADD2
PADD 3
PADD 4 & 5, excluding
California
Total Toxics mg/mi*
75.11
80.11
74.74
NA
Toxics Percent Reduction
27.39
22.56
27.75
NA
*mg/mi is milligram per mile.
7.4.1.4       Reference case (2012)

       The reference case was based on modeling the base case, using 2012 fuel prices, and
scaling the 2004 fuel volumes to 2012 based on growth in fuel demand. In addition, we scaled
MTBE and ethanol upward, in proportion to gasoline growth, and assumed the RFS program
would not be in effect.  For example, if the PADD 1 gasoline pool MTBE oxygen was 0.5 wt%
in 2004, the reference case assumed it should remain at 0.5 wt%. Finally, we assumed the
MSAT 1 standards would remain in place as would the RFG oxygen mandate.   We assumed the
crude slate quality in 2012 is the same as the baseline case.

7.4.1.5       Control cases (2012)

       Two control cases were run for 2012. The assumptions for the control cases are
summarized below:

          •  Control Case  1 (RFS case): 6.7 billion gallons/yr (BGY) of ethanol in gasoline; it
             reflects the renewable fuel mandate. In addition, it is assumed that no MTBE is in
             gasoline, MSAT1 is in place, the 1 psi waiver for CG containing 10 volume
             percent ethanol  remains in effect for all  states where it currently applies, the RFS
             is in effect, and there is no RFG oxygenate mandate.

          •  Control Case 2 (EIA case): Same as Control Case 1, except that the ethanol
             volume in gasoline is 9.6 BGY.

7.4.2   Assumptions made for Refinery Modeling

7.4.2.2       Fuels Production and Demand

       The production of and demand for gasoline and other refinery fuels in the reference and
control cases were based on EIA's AEO 2006 projections for year 2012 . The modeling also was
set up to meet demand based on terminals' sales in each refining area, using EIA fuel sales data.
The LP modeling accounted  for inter-PADD transfers of finished products and gasoline
blendstocks from refiners, to meet demand at terminals, based on historical transfer data from
EIA, including CBOB and RBOB.  Both the RFS and EIA control cases did not model any
production of biodiesel fuel in  fulfilling transportation  diesel fuel demand.
                                         301

-------
7.4.2.3       Ethanol

        The control cases were based on fixed national ethanol volumes as specified in Chapter
2.  For the control cases, however, the LP modeling analysis used ethanol blending economics
and ethanol distribution costs to allocate ethanol to each PADD, and to allocate ethanol use in
CG and RFG grades of gasoline. Additionally, the modeling assumed that all ethanol added to
gasoline is match-blended for octane by refiners in the reference and control cases, while splash
blending of ethanol was assumed as appropriate for the base case using EPA gasoline data.

        The price of ethanol was based on the 2004 yearly average price spread between
regular conventional gasoline sold on spot market in Houston and ethanol sold on spot market on
Chicago Board of Trade (CBOT). This was used to determine a Midwest ethanol production
price. To derive ethanol prices for all other PADDs outside the Midwest, the Midwest ethanol
production price was then adjusted for transportation costs to deliver ethanol from the Midwest
to end use terminals (see section 7.3 for additional details). The price of ethanol was also
adjusted to account for the 51 cent/gal rebate from the Federal subsidies, but did not account for
the impact of state subsides.

       The reference and  control cases where modeled assuming that ethanol  CG blends are
entitled to the 1.0  psi RVP waiver during the summer (i.e., for all 9.0 RVP and low RVP control
programs) so as to assess the impact on summertime butane removal.

7.4.2.4       Processes and Capital

       All changes in refining capital was assessed at a 15%  Return on Investment (ROI) after
taxes, which was adjusted to 7 % ROI before taxes.  Crude and other input prices were based on
Jacobs' projection of refinery margins and crude prices in 2012 cases, which was also based on
the historical price spreads of fuels between PADDS, using information from EIA's 2004 price
information tables, Platts,  and AEO 2006, see the Jacob's report for the petroleum fuel prices
used in the modeling analysis.

7.4.3   Results of Refinery Modeling

7.4.3.1       Summary  of Changes in Refinery Inputs and Outputs to the RFS Case

       There are a number of changes in individual and overall volume for specific gasoline
blendstocks between the RFS case and reference case based on the refinery model  results. The
changes include the increased blending of ethanol, the removal of MTBE, and the increased
volumes of isooctane, isooctene and alkylate from the reuse of isobutylene formerly used to
produce MTBE.  The isooctane and isooctene are produced by merchant MTBE plants that
formerly produced MTBE from mixed butanes, ethylene crackers, and propylene oxide plants as
determined by a survey of how those plants are being converted to produce other gasoline
blendstocks. The  alkylate is produced from the isobutylene previously used to produce MTBE at
captive (refinery-based) MTBE plants.  The total volume of these gasoline blend stocks is
                                          302

-------
summarized in Table 7.4-3 for both the reference case and RFS case adjusting the volume of
ethanol and MTBE to reflect their gasoline energy-equivalent volumes.
                                     Table 7.4-3.
       Comparison of Ethanol, MTBE, Isooctane, Isooctene, and Alkylate Volumes
            by PADD for the RFS Case and Reference Case (barrels per day)
Case
Reference
Case
RFS Case
Gasoline Blendstock
Ethanol
MTBE
Isooctane/lsooctene
Gasoline-equivalent
Volume
Ethanol
MTBE
Isooctane/lsooctene
Alkylate from MTBE
Uasolme-equivalent
Volume
uridnye in udbunrm
Equivalent Volume
PADD1
57,620
54,887
200
82,687
139,224
0
11,042
102,930
20,243
PADD 2
114,900
0
200
76,034
201,989
0
200
133,513
57,479
PADD 3
5,242
122,474
200
102,864
23,091
0
212,177
227,417
124,553
PADD 4/5
20,676
0
200
13,846
17,853
0
200
1 1 ,983
-1 ,863
CA
58,934
0
200
39,096
53,004
0
21 ,484
56,466
17,370
USA
257,372
177,360
1,000
314,527
435,160
0
245,103
532,308
217,781
       As the bottom row in Table 7.4-3 shows, the gasoline-equivalent volumes for the
aggregated volume of these gasoline blendstocks are expected to increase as we compare the
RFS case to the reference case.  It is this net increase in gasoline blendstock volume that is
expected to result in a net reduction in petroleum consumption.

       The addition of ethanol to wintertime gasoline, and to summertime RFG, will cause an
increase of approximately  1 psi in RVP that needs to be offset to maintain constant RVP levels.
An obvious means that refiners could choose to offset the increase in RVP is to reduce the
butane levels in their gasoline. To some extent, the modeling results showed some occurrences
of that, but it also did not report an overall increase in butane sales as a result of the increased
use of ethanol.

       To convert the captive MTBE over to alkylate, after the rejection of methanol, refiners
will need to combine one molecule of refinery produced isobutane with the isobutylene that was
the feedstock for MTBE. The use of the isobutane will reduce the RVP of the gasoline pool
from which it comes,  helping to offset the RVP impacts of ethanol.  Also, the increased
production of alkylate provides a low RVP  gasoline blendstock that offsets a portion of the
cracked stocks produced by the fluidized catalytic cracker unit.  Other  means that the refinery
model used to offset the high blending RVP of ethanol includes purchasing gasoline components
with lower RVP, producing more poly gasoline which has low RVP and selling more high-RVP
naphtha to petrochemical sales.

       In Table 7.4-4, we  summarize the inputs into and  gasoline outputs from the refinery
model separate from the ethanol and converted MTBE blendstocks summarized above. The
summary shows that crude oil and vacuum  gas oil and residual fuel purchases are expected to
decrease about 1  percent averaged over all the PADDs. The refinery model also estimates that
the volume of purchased gasoline components will increase in most PADDs. These gasoline
components include renewable blendstocks for ethanol blending (RBOB), which is a very low
                                         303

-------
RVP gasoline blendstock. A likely reason for the increased gasoline blendstock volumes over
purchased crude oil purchases is that the refinery model is seeking to purchase low RVP
blendstocks to offset the volatility impacts of ethanol, as opposed to having to crack crude oil
which produces more volatile four carbon compounds. Table 7.4-4 also shows the volume of
gasoline projected to be produced for both the RFS case and reference case. We adjusted the
gasoline volume for the RFS case to reflect the same energy density of the gasoline reported for
the reference case.  While the national energy-adjusted gasoline production levels for RFS case
are about the same as the reference case, the energy adjusted gasoline production levels vary
significantly by PADD. The refineries in  PADDs 1 and 2 are projected to produce more gasoline
in the RFS case compared to the reference case, while the refineries in PADD 3 are projected to
produce less gasoline in the RFS case.

                                      Table 7.4-4.
               Summary of Refinery Model Input and Output Volumes by
              PADD for the RFS Case  and Reference Case (barrels per day)
Case
Reference
Case
RFS Case
Crude Oil and Gasoline
Crude Oil
VGO and Residual Fuel
Gasoline Component Input:
Gasoline Volume
Gasoline Energy Content
Crude Oil
VGO and Residual Fuel
Change in Crude oil and
VGO/resid
Inputs
Change in Gasoline
Component Inputs
Gasoline
Gasoline Energy Content
Total Gasoline at Constant
Energy
Volume at Constant
Energy
PADD1
1 ,823,008
152,467
144,293
1,378,811
5.012
1,762,018
75,044
-138,413
200,272
55,979
1 ,483,535
4.951
1 ,465,385
86,573
PADD 2
3,650,044
59,552
69,233
2,398,179
4.997
3,579,232
59,552
-70,812
69,233
0
2,584,977
4.957
2,564,330
166,151
PADDS
9,071,056
680,329
144,782
4,004,675
5.093
9,071,056
680,329
0
1 78,080
33,298
3,753,849
5.073
3,739,352
-265,323
PADD 4/5
1 ,529,442
0
49,247
778,262
5.019
1,520,709
0
-8,733
49,247
0
765,880
5.015
765,181
-13,081
CA
1,952,560
27,400
51 ,475
1,184,533
5.024
1,952,560
40,707
13,307
67,146
15,672
1,184,533
5.046
1,189,801
5,268
USA
18,026,111
919,748
459,030
9,744,461
5.044
17,885,576
855,631
-204,651
563,979
104,949
9,772,775
5.016
9,719,422
-25,039
       The addition of ethanol, the phase out of MTBE and the reuse of former MTBE
feedstocks to make other gasoline blendstocks is expected to change the capital investments that
would otherwise occur if these changes were not made. Table 7.4-5 summarizes the change in
refinery unit throughputs by PADD comparing the RFS case to the reference case.
                                          304

-------
                                      Table 7.4-5.
  Change in the Refinery Unit Capacities by PADD between the RFS Case and Reference
                            Case (thousand barrels per day)
Unit or Category
Crude Tower
Vacuum Tower
Sats Gas Plant
Unsats Gas Plant
FCC DeC5 Tower
FCC
FCC Splitter
Hydrocracker
Delayed Coker
Visbreaker
Thermal Naphtha Splitter
CRU Reformer
SRU Reformer
BTX Reformer
C4 Isomerization
C5/C6 Isomerization
HFAIkylation
H2SO4 Alkylation
Dimersol
Cat Poly
Isooctane
DHT - Total
DHT 2nd RCT - Total
DHT Arom Saturation
NHT- Total Fd
NHT-lsom/ThermalFd
NHT- Reformer Fd
CGH - Generic
CGH - Olefin Sat'n
FCCU Fd HOT
LSR Splitter
LSR Bz Saturator
Reformate Saturator
SDA
MTBE
TAME
Hydrogen Plant - Total BSCF
Lube Unit
Sulfur Plant
Fuel System - Fuel Oil
Fuel System - CO2 (BLb/Day)
Utilities - Steam (Bibs)
Utilities - Steam Vent (Bibs)
Utilities - Power (Mwh)
Utilities - Cooling H2O (Bgal)
PADD1
-61
-22
-13
-38
-8
-130
-97
0
-13
0
-2
0
0
0
-13
0
0
-22
0
1
0
0
0
0
-9
-9
0
-81
0
-39
-12
-6
-2
0
-5
0
-11
0
-276
0
-18
-22
0
-715
-213
PADD 2
-71
-32
-12
-19
-3
-61
-5
0
-16
0
-2
8
0
0
0
3
-1
0
0
0
0
1
3
0
3
-5
8
-34
0
0
-32
-15
-24
0
0
0
1
0
-298
0
-9
-12
0
-226
-93
PADD 3
0
0
6
0
56
0
116
0
0
0
0
48
0
0
1
-57
1
59
0
18
0
22
2
0
-13
-61
48
-42
0
0
0
0
0
0
-114
0
-75
0
138
0
-5
-17
0
749
-156
PADD 4/5
-9
-2
-2
-4
-4
-16
-8
0
0
0
0
0
0
0
0
15
0
-1
-2
0
0
5
5
0
-1
0
0
-4
0
0
0
0
0
0
0
0
3
0
-13
0
-1
0
0
1
-14
CA
0
0
0
3
-9
2
3
-6
3
-3
0
0
-4
0
-2
-26
0
-3
0
0
0
-1
0
-4
-4
0
-4
0
28
0
-6
-2
-4
0
0
0
-27
0
81
0
-2
-6
0
-133
-27
USA
-141
-56
-21
-59
33
-205
9
-6
-27
-3
-3
56
-4
0
-14
-65
0
33
-2
19
0
28
10
-4
-23
-75
52
-161
28
-39
-50
-22
-30
0
-119
0
-109
0
-368
0
-35
-58
0
-324
-502
       Most of the capacity throughput changes are negative, reflecting the decreased processing
of crude oil and vacuum gas oil and decreased downstream refining units as projected by the
refinery model. Of the negative throughput changes, the large reduced volume of the fluidized
catalytic cracker is important.  As discussed above, the refinery model likely chose to decrease
the fluidized catalytic cracker throughput to crack less heavy hydrocarbons to light
hydrocarbons,  producing less four-carbon compounds to offset the volatility impacts of ethanol.
There are several units which show throughput capacity increases, primarily in PADD 3. PADD
3 refineries will have a substantial loss in octane because of the removal of a substantial volume
                                          305

-------
of MTBE, but the refinery model did not choose to blend much ethanol in PADD 3.  Instead, the
refinery model chose to make additional alkylate from the captive MTBE plants formerly
operating in PADD 3, and blend in isooctene from the conversion of merchant MTBE plants.
The refinery model also added some reformer capacity to make up the balance of octane loss.
The refinery model added depentanizer capacity mostly in PADD 3 to enable the blending of
ethanol into RFG.

      Refiners can also control the gasoline production and quality by adjustments they can
make to several of their refinery conversion units. Refiners can adjust the conversion of their
FCC and hydrocracker units, and change the severity of their reformers.  Table 7.4-6 contains the
percent conversions and severities of these units.

                                      Table 7.4-6.
             Comparison of Key Refinery Unit Operations by PADD between
                       the RFS Case and Reference Case (percent)
Case
Reference
Case
RFS Case
Refinery Unit Operations
FCCU Conversion
continuous Kerormer
Severity
Semi-Regen Severity
Hydrocracker
Conversion
FCCU Conversion
Continuous Reformer
Severity
Semi-Regen Severity
Hydrocracker
Conversion
PADD 1
73
99
0
80
72
100
0
80
PADD 2
74
99
0
80
74
96
0
80
PADD 3
74
97
0
85
74
97
0
85
PADD 4/5
71
0
94
85
71
0
93
85
CA
75
0
95
85
75
0
96
85
       The refinery model maintains the same FCC unit conversion percentage for the RFS case
compared to the reference case, except for PADD 1 which showed a small decrease in FCC unit
conversion.  For all PADDs, hydrocracker conversion percentage remains the same.  Continuous
reformer severity is projected to increase slightly in PADD 1 likely because of the octane loss
caused by the removal of MTBE from the RFG pool which is not completely made up by the
increased ethanol volume there.  In PADD 2 where a lot of ethanol is being blended, reformer
severity decreases significantly from 99 RON to 96 RON.  Reformer severity remains the same
in PADD 3.  Reformer severity is projected to increase slightly in California due to an
anticipated small decrease in ethanol. Finally, reformer severity is projected to decrease slightly
in PADDs 4 and 5 despite the small decrease in ethanol there.

       These changes in refinery unit throughputs are associated with changes in capital
investments.  Table 7.4-7 summarizes the projected change in capital investments between  the
reference case and the RFS control case. Table 7.4-7 shows that incremental to the reference
case, refiners are expected to reduce their capital investments by $5.8 billion compared to
business as usual. Most of the reduction occurs in PADDs 1  and 2 where large volumes of
ethanol, and other gasoline blendstocks, are expected to enter the gasoline pool. Of course, this
capital cost decrease is countered by the $2.3 billion in capital costs being incurred to build new
ethanol plants and put into place the distribution system required to distribute the new ethanol.
                                          306

-------
                     Table 7.4-7.
     Comparison of Capital Expenditures by PADD
between the RFS Case and Reference Case (million dollars)

Unit
Crude Tower
Vacuum Tower
Sats Gas Plant
Unsats Gas Plant
FCC DeC5 Tower
FCC
FCC Splitter
Hydrocracker
H-Oil Unit
Delayed Coker
Visbreaker
Thermal Naphtha Splitter
CRU Reformer
SRU Reformer
BTX Reformer
C4 Isomerization
C5/C6 Isomerization
HFAIkylation
H2SO4 Alkylation
Dimersol
Cat Poly
Isooctane
DHT - Total
DHT2nd RCT - Total
DHT Arom Saturation
NHT- Total Fd
CGH - Generic
CGH - Olefin Sat'n
FCCU Fd HOT
LSR Splitter
LSR Bz Saturator
Reformate Saturator
Reform ate Splitter
SDA
MTBE
TAME
Hydrogen Plant
Lube Unit
Sulfur Plant
Merox Jet
Merox Diesel
BTX Reformer - Tower feed
BTX Reformer - Extract feed
Total Capital Costs $MM
PADD 1
CAPEX vs
Reference
Case
-228.6
-141.4
-101.2
-280.1
17.4
-1426.9
-144.2
0.0
0.0
0.0
-0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-698.3
0.0
29.0
0.0
1.6
0.0
0.0
-39.7
-472.8
0.0
-525.0
0.0
-44.7
-8.2
-4.4
0.0
0.0
0.0
-109.6
0.0
-1.9
0.0
0.0
0.0
0.0
-4,179
PADD 2
CAPEX vs
Reference
Case
0.0
0.0
-13.6
-225.5
-52.0
-1160.4
-37.0
0.0
0.0
0.0
-0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
11.0
0.0
0.0
-154.5
0.0
0.0
-47.2
-151.7
-272.4
-142.7
0.0
0.0
0.0
2.2
0.0
-2.8
0.0
0.0
0.0
0.0
-2,247
PADD 3
CAPEX vs
Reference
Case
0.0
0.0
-1.8
-2.5
54.3
0.0
49.5
0.0
0.0
0.4
0.0
0.0
0.0
0.0
2.4
28.9
0.0
0.0
607.5
0.0
100.3
0.0
217.8
6.1
0.0
0.0
-139.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-175.4
0.0
-196.1
0.0
0.0
0.0
0.0
0.0
0.0
552
PADD 4/5
CAPEX vs
Reference
Case
0.0
2.6
2.6
-29.5
-16.6
-103.8
-6.6
0.0
0.0
0.0
0.0
0.0
0.0
23.9
0.0
0.0
153.7
0.0
0.0
-23.3
0.0
0.0
32.0
21.1
0.0
69.3
-70.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-7.4
0.0
-0.1
0.0
0.0
0.0
0.0
48
CA
CAPEX vs
Reference
Case
0.0
0.0
-22.7
-1.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-60.6
0.0
0.0
0.0
-25.0
0.0
0.0
11.2
0.0
0.0
0.0
0.0
160.5
0.0
0.0
0.4
-47.3
-8.8
0.0
0.0
0.0
-58.3
0.0
0.1
0.0
0.0
0.0
0.0
-51
U.S. Total
CAPEX vs
Reference
Case
-228.6
-138.7
-136.8
-538.6
3.1
-2691.0
-138.3
0.0
0.0
0.4
-0.2
0.0
0.0
23.9
2.4
-31.7
153.7
0.0
-90.9
-48.3
129.3
0.0
262.6
38.2
0.0
29.6
-837.2
160.5
-525.0
-47.2
-196.0
-328.0
-155.9
0.0
-175.4
0.0
-369.2
0.0
-4.7
0.0
0.0
0.0
0.0
-5,878
                        307

-------
7.4.3.2
Summary of Changes in Refinery Inputs and Outputs to the EIA Case
       The EIA case has some similarities to the RFS case. The MTBE is still estimated to no
longer be blended into gasoline, and the former MTBE feedstocks are converted over to other
low-RVP gasoline blendstocks.  The annual volume of ethanol blended into gasoline, however, is
almost 3 billion gallons higher.  This increased volume of ethanol is expected to be spread over
all the PADDs, although PADD 3 is projected to absorb the most.  The much increased volume
of very high octane ethanol is expected to slightly reduce the consumption of the gasoline
blendstocks produced from former MTBE feedstocks.  The net gasoline-equivalent volume
increase by ethanol and other gasoline blendstock changes is expected to be over 100 thousand
barrels per day.  Table 7.4-8 contains the volumes of these gasoline blendstocks by PADD.

                                     Table 7.4-8.
       Comparison  of Ethanol, MTBE, Isooctane, Isooctene, and Alkylate Volumes
            by PADD for the EIA Case and  Reference Case (barrels per day)
Case
Reference
Case
EIA Case
Gasoline Blendstock
Ethanol
MTBE
Isooctane/lsooctene
Gasoline-equivalent
Volume
Ethanol
MTBE
Isooctane/lsooctene
Alkylate from MTBE
Gasoline-equivalent
Volume
Change in Gasoline
Equivalent Volume
PADD1
57,620
54,887
200
82,687
161,821
0
1 1 ,042
117,844
35,157
PADD 2
1 1 4,900
0
200
76,034
255,512
0
200
168,838
92,804
PADDS
5,242
122,474
200
102,864
117,722
0
200,119
277,816
1 74,952
PADD 4/5
20,676
0
200
1 3,846
32,113
0
200
21,395
7,548
CA
58,934
0
200
39,096
59,055
0
17,010
55,986
16,889
USA
257,372
177,360
1,000
314,527
626,223
0
228,571
641 ,878
327,351
       Table 7.4-9 summarizes the inputs into and gasoline outputs from the refinery model
separate from the ethanol and converted MTBE blendstocks summarized in Table 7.4-12 above.
Crude oil and vacuum gas oil and residual fuel purchases are expected to decrease about 1.7
percent averaged over all the PADDs. The refinery model also estimates that the volume of
purchased gasoline components will increase incrementally over the RFS case. It seems that a
likely reason for the increased gasoline blendstock volumes over purchased crude oil purchases
is that the refinery model is seeking to purchase low RVP blendstocks to offset the volatility
impacts of ethanol, as opposed to having to crack crude oil which produces more volatile four
carbon compounds. Table 7.4-15 also shows the energy-adjusted volume of gasoline projected
to be produced for both the RFS case and reference case.  The national energy-adjusted gasoline
production levels for EIA case is somewhat lower than the reference case which suggests that the
crude oil savings described above are somewhat overstated. The refineries in PADDs 1 and 2
are projected to produce much more gasoline in the EIA case compared to the reference case,
while the refineries in PADD 3 are projected to produce much less gasoline in the EIA case.
                                         308

-------
                                     Table 7.4-9.
            Summary of Refinery Model Input and Output Volumes by PADD
                 for the EIA Case and Reference Case (barrels per day)
Case
Reference
Case
EIA Case
Crude Oil and Gasoline
Crude Oil
VGO and Residual Fuel
Gasoline Component Inputs
Gasoline Volume
Gasoline Energy Content
Crude Oil
VGO and Residual Fuel
Change in Crude oil and
VGO/resid
Gasoline Component
Inputs
Change in Gasoline
Component Inputs
Gasoline
Gasoline Energy Content
Total Gasoline at Constant
Energy
Change in Total Gasoline
Volume at Constant
Energy
PADD1
1 ,823,008
152,467
144,293
1,378,811
5.012
1 ,667,893
97,621
-209,962
208,809
64,516
1,602,258
4.924
1 ,574,007
195,196
PADD 2
3,650,044
59,552
69,233
2,398,179
4.997
3,539,369
59,552
-110,675
69,233
0
2,584,977
4.922
2,546,302
148,123
PADD 3
9,071,056
680,329
144,782
4,004,675
5.093
9,058,059
675,959
-17,368
166,023
21,240
3,657,519
5.042
3,620,530
-384,145
PADD 4/5
1,529,442
0
49,247
778,262
5.019
1,528,255
0
-1,188
49,247
0
775,512
5.000
772,501
-5,761
CA
1,952,560
27,400
51,475
1,184,533
5.024
1,952,560
41,194
13,794
62,913
11,438
1,184,533
5.042
1,188,742
4,208
USA
18,026,111
919,748
459,030
9,744,461
5.044
17,746,134
874,325
-325,398
556,224
97,194
9,804,799
4.988
9,695,751
-48,710
       The addition of ethanol, the phase out of MTBE and the reuse of former MTBE
feedstocks to make other gasoline blendstocks is expected to change the capital investments that
would otherwise occur if these changes were not made.

       Table 7.4-10 summarizes the change in refinery unit throughputs by PADD comparing
the EIA case to the reference case.
                                        309

-------
                                     Table 7.4-10.
  Change in the Refinery Unit Capacities by PADD between the EIA Case and Reference
                            Case (thousand barrels per day)
Unit or Category
Crude Tower
Vacuum Tower
Sats Gas Plant
Unsats Gas Plant
FCC DeC5 Tower
FCC
FCC Splitter
Hydrocracker
Delayed Coker
Visbreaker
Thermal Naphtha Splitter
CRU Reformer
SRU Reformer
BTX Reformer
C4 Isomerization
C5/C6 Isomerization
HF Alkylation
H2S04 Alkylation
Dimersol
Cat Poly
Isooctane
DHT - Total
DHT 2nd RCT - Total
DHT Arom Saturation
NHT - Total Fd
NHT - Isom/Thermal Fd
NHT- Reformer Fd
CGH - Generic
CGH - Olefin Sat'n
FCCU Fd HOT
LSR Splitter
LSR Bz Saturator
Reformate Saturator
SDA
MTBE
TAME
Hydrogen Plant - Total BSCF
Lube Unit
Sulfur Plant
Fuel System - Fuel Oil
Fuel System - CO2 (BLb/Day
Utilities - Steam (Bibs)
Utilities - Steam Vent (Bibs)
Utilities - Power (Mwh)
Utilities - Cooling H2O (Bgal)
PADD1
-155
-63
-17
-37
-7
-130
-99
0
-10
3
-1
-15
0
0
-13
0
0
-22
0
0
0
-24
0
0
-32
-17
-15
-81
0
-39
-12
-6
-3
-3
-5
0
-34
0
-340
0
-20
-25
0
-1,029
-246
PADD 2
-111
-50
-20
-28
0
-90
0
0
-25
3
-3
-19
0
0
3
-14
-3
0
0
0
0
-3
-15
0
-27
-7
-19
-53
0
0
-32
-15
-24
0
0
0
7
23
-551
0
-13
-12
0
-582
-128
PADD 3
-13
12
-20
-6
40
-18
82
0
-7
1
-1
-109
0
0
0
-93
1
49
0
18
0
-66
17
0
-145
-36
-109
26
0
0
0
0
0
0
-114
0
4
0
-261
0
-10
-28
0
-98
-256
PADD 4/5
-1
4
0
-7
-6
-24
-12
0
0
0
0
0
-1
0
0
-3
0
-1
-2
0
0
16
15
0
-1
0
-1
-6
0
0
0
1
0
0
0
0
4
0
-4
0
-2
-3
0
40
-24
CA
0
0
1
3
-9
3
3
-7
3
-3
0
0
-3
0
-2
-30
0
-5
3
0
0
0
0
-4
-3
0
-3
0
29
0
-6
-2
-5
0
0
0
-30
0
86
0
-2
-7
0
-168
-28
USA
-280
-97
-56
-74
17
-260
-27
-7
-39
3
-5
-143
-4
0
-12
-140
-2
22
1
19
0
-78
17
-4
-208
-60
-147
-114
29
-39
-50
-21
-32
-3
-119
0
-49
23
-1,070
0
-47
-75
0
-1,837
-682
       Most of the capacity throughput changes are negative, reflecting the decreased processing
of crude oil and vacuum gas oil and decreased downstream refining units as projected by the
refinery model. Of the negative throughput changes, the reduced volume of the fluidized
catalytic cracker is important. As discussed above, the refinery model likely chose to decrease
the fluidized catalytic cracker throughput to crack less heavy hydrocarbons to light
                                          HO

-------
hydrocarbons, producing less four-carbon compounds to offset the volatility impacts of ethanol.
The reduction in FCC unit throughput is relatively less for the EIA case than it was for the RFS
case.  There are several units which show throughput capacity increases, primarily in PADD 3.
PADD 3 refineries will have a substantial loss in octane because of the removal of a substantial
volume of MTBE. Unlike the RFS case, however, much of the ethanol is projected to be blended
into PADD 3's gasoline pool making up for the octane loss. This can be seen in Table 7.4-15 as
reformer and alkylation throughputs volumes are projected to be lower for the EIA case. The
refinery model added depentanizer capacity mostly in PADD 3 to enable the blending of ethanol
into RFG.

       Refiners can also control the gasoline production and quality by adjustments they can
make to several of their refinery conversion units. Refiners can adjust the conversion of their
FCC and hydrocracker units, and change the severity of their reformers. Table 7.4-11 contains
the percent conversions and severities of these units.

                                     Table 7.4-11.
             Comparison of Key Refinery Unit Operations by PADD between
                       The EIA Case and Reference Case (percent)
Case
Reference
Case
EIA Case
Refinery Unit Operations
FCCU Conversion
continuous Ketormer
Severity
Semi-Regen Severity
Hydrocracker
Conversion
FCCU Conversion
Continuous Reformer
Severity
Semi-Regen Severity
Hydrocracker
Conversion
PADD 1
73
99
0
80
72
100
0
80
PADD 2
74
99
0
80
74
94
0
80
PADD 3
74
97
0
85
74
96
0
85
PADD 4/5
71
0
94
85
71
0
93
85
CA
75
0
95
85
75
0
97
85
       The refinery model maintains the same FCC unit conversion percentage for the RFS case
compared to the reference case, except for PADD 1 which showed a small decrease in FCC unit
conversion.  For all PADDs, hydrocracker conversion percentage remains the same.  Continuous
reformer severity is projected to increase slightly in PADD 1 despite the ethanol blended into
that PADD's gasoline. In PADD 2 where a lot of ethanol is being blended, reformer severity
decreases significantly from 99 RON to 94 RON.  Reformer severity is projected to decrease
slightly in PADD 3 due to the large volume of ethanol being blended into the gasoline in that
PADD. Reformer severity is projected to decrease slightly in California due to an anticipated
small increase in ethanol.  Finally, reformer severity is projected to decrease slightly in PADDs 4
and 5 due to the increase in ethanol there.

       These changes in refinery unit throughputs are associated with changes in capital
investments. Table 7.4-12 summarizes the projected change in capital investments between the
reference case and the EIA control case.  Table 7.4-12 shows that incremental to the reference
case, refiners are expected to reduce their capital investments by $7.3 billion compared to
business as usual.  Most of the reduction occurs in PADDs 1 and 2 where large volumes of
                                          ill

-------
ethanol, and other gasoline blendstocks are expected to enter the gasoline pool. Of course, this
capital cost decrease is countered by the estimated $6.5 billion in capital costs incurred to build
new ethanol plants and put into place the distribution system that the new ethanol requires.
                                             112

-------
                   Table 7.4-12.
Comparison of Capital Expenditures by PADD between
  the EIA Case and Reference Case (million dollars)

Unit
Crude Tower
Vacuum Tower
Sats Gas Plant
Unsats Gas Plant
FCC DeC5 Tower
FCC
FCC Splitter
Hydrocracker
H-Oil Unit
Delayed Coker
Visbreaker
Thermal Naphtha Splitter
CRU Reformer
SRU Reformer
BTX Reformer
C4 Isomerization
C5/C6 Isomerization
HF Alkylation
H2S04 Alkylation
Dimersol
Cat Poly
Isooctane
DHT - Total
DHT 2nd RCT - Total
DHT Arom Saturation
NHT- Total Fd
CGH - Generic
CGH - Olefin Sat'n
FCCU Fd HOT
LSR Splitter
LSR Bz Saturator
Reformate Saturator
Reformate Splitter
SDA
MTBE
TAME
Hydrogen Plant
Lube Unit
Sulfur Plant
Merox Jet
Merox Diesel
BTX Reformer - Tower feed
BTX Reformer - Extract feed
Total
PADD 1
CAP EX vs
Reference
Case
-453.8
-295.0
-115.9
-275.6
17.7
-1426.9
-147.0
0.0
0.0
0.0
7.2
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-715.4
0.0
4.3
0.0
-169.8
0.0
0.0
-39.7
-471.6
0.0
-525.0
0.0
-44.7
-19.2
-20.1
0.0
0.0
0.0
-188.5
0.0
-2.5
0.0
0.0
0.0
0.0
-4,882
PADD 2
CAPEX vs
Reference
Case
0.0
0.0
-13.6
-261.5
-58.9
-1160.4
-48.7
0.0
0.0
0.0
6.2
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-165.3
0.0
0.0
-179.5
0.0
0.0
-47.2
-151.7
-272.4
-142.7
0.0
0.0
0.0
22.9
0.0
-2.8
0.0
0.0
0.0
0.0
-2,476
PADD 3
CAPEX vs
Reference
Case
0.0
103.7
-55.8
-20.1
50.9
-68.1
46.6
0.0
0.0
-185.9
0.0
-0.3
0.0
0.0
1.8
0.0
0.0
0.0
497.6
0.0
114.9
0.0
-219.5
105.6
0.0
0.0
102.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-175.4
0.0
6.6
0.0
-0.2
0.0
0.0
0.0
-0.1
304
PADD 4/5
CAPEX vs
Reference
Case
0.0
8.9
0.1
-49.6
-18.5
-331.7
-9.6
0.0
0.0
0.0
0.0
0.0
0.0
-2.6
0.0
0.0
-56.6
0.0
0.0
-17.9
0.0
0.0
93.2
138.0
0.0
-1.9
-77.1
0.0
0.0
0.0
21.9
0.0
0.0
0.0
0.0
0.0
54.2
0.0
0.0
0.0
0.0
0.0

-249
CA
CAPEX vs
Reference
Case
0.0
0.0
-23.4
2.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-60.6
0.0
0.0
0.0
7.8
0.0
0.0
21.6
0.0
0.0
0.0
0.0
161.0
0.0
0.0
0.4
-49.9
-10.1
0.0
0.0
0.0
-58.3
0.0
0.1
0.0
0.0
0.0

-9
U.S. Total
CAPEX vs
Reference
Case
-453.8
-182.4
-208.7
-604.4
-8.8
-2987.1
-158.7
0.0
0.0
-185.9
13.4
-0.3
0.0
-2.6
1.8
-60.6
-56.6
0.0
-217.8
-10.1
119.2
0.0
-274.6
78.3
0.0
-41.7
-626.1
161.0
-525.0
-47.2
-174.0
-341.6
-172.9
0.0
-175.4
0.0
-163.2
0.0
-5.4
0.0
0.0
0.0

-7,311

-------
7.4.3.3
Adjustments to the LP Refinery Model's Cost Estimate
       We made several adjustments to the costs directly estimated by the LP refinery cost
model for the RFS and EIA cases which are included in the costs reported below. One
adjustment made was to adjust the costs based on the ethanol prices used in the LP cost model to
reflect the ethanol production costs estimated and reported above in section 7.1.1. This
adjustment resulted in much lower ethanol costs to refiners because Jacobs largely based its
ethanol prices on ethanol's octane costs instead of its historical price relationship to gasoline,
which is much lower. We also adjusted the ethanol distribution costs from those used in the LP
refinery cost study, which roughly corresponded to those used for the proposed rule cost
analysis, to those estimated for the final rule as discussed above in section 7.3.1. In Table 7.4-13
we summarize the ethanol production and distribution costs used in the LP refinery cost model
and those we estimated for the final rule.

                                      Table 7.4-13
        Ethanol Price and Distribution Costs used in the LP Refinery Model versus
              Those used for the Final Rule Cost Analysis (cents per gallon)

Prices
used in
LP
Refinery
Cost
Model
Costs
used in
Final
Cost
Analysis

Ethanol
Price in
Midwest
Ethanol
Distribution
Cost
Ethanol
Price in
PADD
Ethanol
Production
Cost
Ethanol
Distribution
Cost
Ethanol Cost
in PADD
Case
RFS and
EIA
Case
RFS and
EIA
Case
RFS and
EIA
Case
RFS
Case
EIA
Case
RFS and
EIA
Case
RFS
Case
EIA
Case
PADD 1
158
12
170
126
131
16
142
147
PADD 2
158
0
158
126
131
6.5
132.5
137.5
PADD 3
158
10
168
126
131
16
142
147
PADD
4/5 ex CA
158
17
175
126
131
23
149
154
CA
158
18
176
126
131
22
148
153
                                           114

-------
       Another adjustment we made to the costs directly estimated by the LP refinery cost
model was to add a cost for distributing gasoline.  The refinery cost model did not include
distribution costs for gasoline for moving the gasoline from the refinery to the terminal. We
assigned gasoline distribution costs to be 4 cents per gallon applied as a cost savings to the
gasoline-equivalent volume of ethanol blended into each PADD's gasoline, since this roughly
corresponded to the volume of gasoline displaced by the ethanol.

7.4.3.4        Estimated Costs

7.4.3.4.1      Estimated Costs for the RFS Case

       Table 7.4-14  summarizes the costs for the RFS case excluding federal and state ethanol
consumption subsidies.  The costs are reported by different cost component as well as aggregated
total and the per-gallon costs." This estimate of costs reflects the changes in gasoline that are
occurring with the expanded use of ethanol, including the corresponding removal of MTBE and
reuse of MTBE feedstocks. The operating costs include the labor, utility and other operating
costs and are a direct output from the refinery model. These costs are adjusted to reflect
ethanol's production cost plus distribution costs instead of the ethanol prices used in the refinery
cost model. The fixed costs are 3 percent of the capital costs.  The  costs associated with lower
energy density gasoline  are accounted for using the fractional change in energy density shown in
Table 7.4-4, multiplied times the wholesale price of gasoline.  By excluding the federal and state
ethanol consumption subsidies in the table, we avoid the transfer payments caused by these
subsidies that would  hide a portion of the program's costs.

                                       Table 7.4-14.
           Summary of RFS Case Costs without Ethanol Consumption Subsidies
  (million  dollars per year and c/gal, except as noted; 2004 dollars, 7% ROI before taxes )

Capital Costs ($MM)
Amortized Capital Costs ($MM/yr)
Fixed Operating Cost ($MM/yr)
Variable Operating Cost ($MM/yr)
Lower Energy Density Gasoline ($MM/yr)
Total Cost ($MM/yr)
Capital Costs (c/gal)
Fixed Operating Cost (c/gal)
Variable Operating Cost (c/gal)
Lower Energy Density Gasoline (c/gal)
Total Cost Excluding Subsidies (c/gal)
RFS Case
6.7 Billion Gals
Incremental to Reference Case
-5,878
-647
-178
-201
1,848
823
-0.40
-0.11
-0.12
1.13
0.50
99 EPA typically assesses social benefits and costs of a rulemaking. However, this analysis is more limited in its
scope by examining the average cost of production of ethanol and gasoline without accounting for the effects of
farm subsidies that tend to distort the market price of agricultural commodities.


                                           315

-------
       Our analysis shows that when considering all the costs associated with these fuel changes
resulting from the expanded use of ethanol that these various possible gasoline use scenarios will
cost the U.S. $820 million in the year 2012.  Expressed as per-gallon costs, these fuel changes
would cost the U.S. 0.50 cent per gallon of gasoline.

       Table 7.4-15 expresses the total and per-gallon gasoline costs for the RFS case with the
federal and state ethanol subsidies included.  The federal tax subsidy is 51 cents per gallon for
each gallon of new ethanol blended into gasoline.  The state tax subsidies apply in 5 states and
range from 1.6 to 29 cents per gallon. The cost reduction to the fuel industry  and consumers are
estimated by multiplying the subsidy times the volume of new ethanol estimated to be used in the
state.

                                      Table 7.4-15.
           Estimated RFS Case Cost including Ethanol Consumption Subsidies
     (million dollars per year and cents per gallon; 2004 dollars, 7% ROI before taxes)

Total Cost ($MM/yr)
Federal Subsidy ($MM/yr)
State Subsidies ($MM/yr)
Revised Total Cost ($MM/yr)
Per-Gallon Cost Excluding Subsidies (c/gal)
Federal Subsidy (c/gal)
State Subsidies (c/gal)
Total Cost Including Subsidies (c/gal)
RFS Case
6.7 Billion Gals
Incremental to Reference Case
823
-1376
-5
-558
0.50
-0.84
-0.003
-0.34
       The cost including subsidies better represents gasoline's production cost as might be
reflected to the fuel industry as a whole and to consumers "at the pump" because the federal and
state subsidies tends to hide a portion of the actual costs.  Our analysis estimates that the fuel
industry and consumers will see a 0.34 cent per gallon decrease in the apparent cost of producing
gasoline for the RFS case.

7.4.3.4.2      Estimated Costs for the EIA Case

       Table 7.4-16 summarizes the costs for the EIA case. The costs in this table exclude
federal and state ethanol consumption subsidies.  The costs are  reported by different cost
components as well as the aggregated total and the per-gallon costs.  This estimate of costs
reflects the changes in gasoline that are occurring with the much expanded use of ethanol,
including the removal of MTBE and reuse of MTBE feedstocks. The operating costs include the
labor, utility and other operating costs and are a direct output from the refinery model, adjusted
for ethanol's production cost at this higher volume including ethanol distribution costs. The
fixed costs are 3 percent of the capital costs.  The costs associated with lower energy density
gasoline, as shown in Table 7.4-9, are estimated by the fractional change in energy content times
the wholesale price of gasoline.  The increment of the EIA case to the RFS case indicates the
economic impact of the additional volume of ethanol between the two cases.
                                           116

-------
                                      Table 7.4-16.
           Summary of EIA Case Costs without Ethanol Consumption Subsidies
  (million dollars per year and c/gal, except as noted; 2004 dollars, 7% ROI before taxes)

Capital Costs ($MM)
Amortized Capital Costs ($MM/yr)
Fixed Operating Cost ($MM/yr)
Variable Operating Cost ($MM/yr)
Lower Energy Density Gasoline
($MM/yr)
Total Cost ($MM/yr)
Capital Costs (c/gal)
Fixed Operating Cost (c/gal)
Variable Operating Cost (c/gal)
Lower Energy Density Gasoline (c/gal)
Total Cost Excluding Subsidies (c/gal)
EIA Case
9.6 Billion Gals
Incremental to Reference Case
-7,311
-804
-222
-491
3,255
1739
-0.49
-0.14
-0.30
1.98
1.06
EIA Case
9.6 Billion Gals
Incremental to RFS Case
-1,433
-158
-43
-290
1407
915
-0.10
-0.03
-0.18
0.86
0.56
       Our analysis shows that when considering all the costs associated with these fuel changes
resulting from the expanded use of subsidized ethanol that these various possible gasoline use
scenarios will cost the U.S. $1,740 million in the year 2012 for the EIA case. Expressed as per-
gallon costs, these fuel changes would cost the U.S. about 1.1 cents per gallon of gasoline. The
incremental volume of ethanol added between the RFS and EIA cases is expected to cost $915
million in the year 2012, resulting in a 0.56 cent per gallon cost.

       Table 7.4-17 expresses the total and per-gallon gasoline costs for the EIA case with the
federal and state ethanol subsidies included. The federal tax subsidy is 51 cents per gallon for
each gallon of new ethanol blended into gasoline. The state tax subsidies apply in 5 states and
range from 1.6 to 29 cents per gallon.  The cost reduction to the fuel industry and  consumers are
estimated by multiplying the subsidy times the volume of new ethanol estimated to be used in the
state.
                                           117

-------
                                       Table 7.4-17.
            Estimated EIA Case Cost including Ethanol Consumption Subsidies
     (million dollars per year and cents per gallon; 2004 dollars, 7% ROI before taxes)

Total Cost ($MM/yr)
Federal Subsidy ($MM/yr)
State Subsidies ($MM/yr)
Revised Total Cost ($MM/yr)
Per-Gallon Cost Excluding Subsidies
(c/gal)
Federal Subsidy (c/gal)
State Subsidies (c/gal)
Total Cost Including Subsidies
(c/gal)
EIA Case
9.6 Billion Gals
Incremental to Reference Case
1739
-2865
-31
-1158
1.06
-1.74
-0.02
-0.71
EIA Case
9.6 Billion Gals
Incremental to RFS Case
915
-1489
-26
-600
0.56
-0.90
-0.02
-0.37
       The cost including subsidies better represents gasoline's production cost as might be
reflected to the fuel industry as a whole and to consumers "at the pump" because the federal and
state subsidies tends to hide a portion of the actual costs. Our analysis estimates that the fuel
industry and consumers will see a 0.71 cent per gallon decrease in the apparent cost of producing
gasoline for the EIA case. Incremental to the RFS case, the consumer would be expected to see a
0.37 cent per gallon price decrease "at the pump."
7.4.3.4.3
Sensitivity Cost Analyses for the RFS and EIA Cases
       In Table 7.1-5 above, we presented various corn-ethanol production cost estimates based
on varying corn and dried distillers grain prices. We entered a range of low and high production
ethanol cost estimates from that table into our cost spreadsheet created from the output from the
LP refinery cost modeling. The range of ethanol production costs that we chose represents a
reasonable bound around the possible range of future ethanol production costs.  This allowed us
to estimate the cost of using ethanol at these other possible ethanol production costs at the
ethanol volumes analyzed for the RFS and EIA cases. We present these costs in Table 7.4-18.
We did not conduct sensitivity analyses around higher or lower crude oil prices. 10°
100 This sensitivity analysis conducted at lower and higher ethanol production costs can also be used as a surrogate
for a sensitivity analysis of higher and lower crude oil prices.  Analyzing a lower ethanol cost is similar to analyzing
a higher crude oil price with ethanol production costs at the levels we analyzed them at which was 126 and 131 cents
per gallon, and vice versa for our sensitivity analysis at the higher ethanol production cost.
                                            118

-------
                                      Table 7.4-18.
 Summary of the Sensitivity Cost Analysis at Higher and Lower Ethanol Production Costs
                    (Costs in 2012, 2004 dollars, 7% ROI before taxes)
Ethanol Production
Cost
0.86
2.04

Cost without
Subsidies
Cost with
Subsidies
Cost without
Subsidies
Cost with
Subsidies
Units
$MM/yr
c/gal
$MM/yr
c/gal
$MM/yr
c/gal
$MM/yr
c/gal
Costs
RFS Case
-260
-0.16
-1640
-1.00
2930
1.79
1546
0.95
Costs
EIA Case
-846
-0.52
-3740
-2.28
5784
3.53
2890
1.76
7.4.4   Impact on Diesel Prices

       Biodiesel fuel is added to highway and nonroad diesel fuel, which increases the volume
and therefore the supply of diesel fuel and thereby reduces the demand for refinery-produced
diesel fuel. In this section, we estimate the overall cost impact, considering how much refinery
based diesel fuel is displaced by the forecasted production volume of biodiesel fuel.  The cost
impacts are evaluated considering the production cost of biodiesel with and without the subsidy
from the Biodiesel Blenders Tax credit program.  Additionally, the diesel cost impacts are
quantified with refinery diesel prices as forecasted by Jacobs's which are based on EIA's AEO
2006.

       We estimate the net effect that biodiesel production has on overall cost for diesel fuel in
year 2012 using total production costs for biodiesel and diesel fuel. The costs are evaluated
based on how much refinery based  diesel fuel is displaced by the biodiesel volumes as forecasted
by EIA, accounting for energy density differences between the fuels.  The cost impact is
estimated from a 2012 year basis, by multiplying the production costs of each fuel by the
respective changes in volumes for biodiesel and estimated displaced diesel fuel.  We further
assume that all of the forecasted bio-diesel  fuel volume is used as transport fuel, neglecting
minor uses in the heating oil market.

       For this analysis, the production costs for biodiesel fuel are based on the costs generated
using the USD A, NREL, EIA and the design vendors estimates in the preceding sections. We
average these results to developed costs for soy oil and yellow grease feedstocks. Additionally,
the production costs are based on EIA's projection in 2012 that half of the total biodiesel volume
will be made from soy oil feedstock with the remaining volume being produced from yellow
grease.  To these estimates, we add distribution costs of 21.5 c/gal to the biodiesel production
costs, reflecting the distribution estimates derived in  section 7.3.2.  For the refinery diesel
production costs in 2012, we used the projected wholesale national average diesel price of 160
c/gal projected by the Jacobs' pricing forecast.  Distribution costs of 4 c/gal were added to the
Jacobs's wholesale diesel price projection,  to account for the additional costs to move diesel fuel
from the wholesale market to  end use terminals.
                                           119

-------
       Our estimate for the reduction in refinery produced diesel fuel is based on EIA's forecast
for approximately 300 MM gallons of biodiesel in 2012, along with the 2012 Reference Case
year biodiesel production volume of 28 MM gallons. With this and accounting for differences in
energy density between biodiesel and diesel fuel, we estimate in 2012 that the additional
biodiesel production reduces the need for 250 MM gallons of refinery produced diesel fuel.
Table 7.4-19 contains the energy densities used in this analysis.

                                      Table 7.4-19.
	Energy Content of Fuels per Gallon	
 Fuel                                                Lower Heating Value (BTU/gallon)

 Biodiesel                                                       117,093

 Refinery Produced Diesel                                          128,700
       For all RFS case scenarios, the net effect of biodiesel production on diesel fuel costs,
including the biodiesel blenders' subsidy, is a reduction in the cost of transport diesel fuel costs
by $114 MM per year, which equates to a fuel cost reduction of about 0.20 c/gal101. Without the
subsidy, the transport diesel fuel costs are increased by $91 MM per year, or an increase of 0.16
c/gal.
7.5    Other Potential Economic Impacts

       Ideally, we would prefer to assess all economic and environmental impacts of increased
ethanol use and decreased fossil fuel use in a holistic manner. Such an analysis is beyond the
scope of this RIA. However, we can approximate some of the impacts of increased ethanol
production and use,  and we can discuss other impacts qualitatively. The preceding discussion
quantifies the impact of expanded use of renewable fuels on the cost of gasoline and diesel fuel.
It does so by quantifying the direct costs of ethanol production, as well as the direct costs of state
and federal tax subsidies for the renewable fuels, which are financed through tax payments.
There are many other economic impacts associated with the use of renewable fuels and the fossil
fuels they replace which go well beyond the scope of the analysis conducted for the RIA. We
have not attempted to quantify all of them here. For example, increased renewable fuel
production and use may have adverse impacts on surface and ground water quality and soil
erosion, while decreased fossil fuel, distribution and use may have positive impacts.  To quantify
the economic impact associated with this would require extensive analysis of the likely responses
of farmers to the increased demand for renewable fuels, the cost of actions taken to remedy the
impacts, and the cost of any resulting health and welfare impacts.
101 Based on EIA's AEO 2006, the total volume of highway and off-road diesel fuel consumed in 2012 was
estimated at 58.9 billion gallons.
                                           320

-------
       Furthermore, the renewable fuel production costs assumed in our analysis do not reflect
the entire cost to society associated with the production of the corn and soybean feedstocks used
in their production due other state and federal agricultural policies. Direct payments, counter-
cyclical payments, marketing loans, and subsidized crop insurance are all examples of policies
outside of this rulemaking that impact the price of corn and soybeans that are not reflected in the
production cost for ethanol and biodiesel, but do impact costs borne by consumers indirectly
through taxes. Quantifying the incremental impacts of this rulemaking on the effects of these
pre-existing programs would represent a significant challenge. However, the challenge is
complicated even more by the direct and indirect economic support provided for the production,
supply, and distribution of the fossil fuels which would be replaced by these renewable  fuels.
Again, any assessment of the  overall costs to society for increase renewable use would have to
look at the economic support  provided across the entire fuel  supply. Such an analysis is well
beyond the scope of this RIA.

       Despite our inability to fully capture all the potential impacts on the cost to society of
increased renewable fuel use, two potential impacts were touched on briefly in our analysis, and
these are discussed in this subsection.

Economic Impacts of Emission Changes

       As discussed in Chapters 4.1 and 5.1, we estimate that there may be an increase  in
emissions and a corresponding small increase in ozone resulting from the expanded use of
renewable fuels. Our vehicle and equipment emission estimates are highly uncertain, however,
given the lack of data in particular on vehicles and engines complying with the latest standards.
However, to the extent that there are emission and ozone increases resulting from the expanded
use of renewable fuels, there can be a cost associated with them. In some cases, areas that see an
increase in emissions resulting from renewable fuel use may be forced to take other actions to
offset these emission increases. In other cases, particularly in attainment areas, the impact, while
not affecting attainment, may adversely impact air quality and human health. It is difficult to
provide any quantitative estimate of what the mitigation costs might be to offset emission
increases, or to quantify the health impacts resulting from the air quality impacts. Not only are
the emission and air quality impacts highly uncertain, but they are also very location dependent.
While we have made projections on where the ethanol use may rise or fall for the purposes of
estimating nationwide fuel cost impacts and potential emissions impacts, these projections are
much less reliable when trying to predict specific local air quality impacts.

       Despite all of the above caveats, we have attempted to provide a rough estimate  of the
potential national-level cost impacts; As a surrogate for additional emission control costs in
nonattainment areas and potential health impacts in attainment areas, we looked at the potential
health costs associated with the secondary nitrate PM resulting from the decreases in NOx
emissions estimated in previous EPA rules.  We note again that we actually expect an overall
decrease in ambient PM2.5 formation due to the increased use of ethanol in fuel (See Chapter
5.2).102 Thus, we expect most areas to have lower health impact costs and certainly lower
abatement costs related to PM control.
102 Overall, we expect that the decrease in secondary organic PM is likely to exceed the increase in secondary nitrate
PM. In 2006, NOX emissions from gasoline-fueled vehicles and equipment comprise about 37% of national NOX
                                           321

-------
       In recent rulemakings we monetized PM emission impacts, including those resulting from
changes in secondarily formed PM2.5 due to NOx emission changes.  Using this information as a
guide, we provide a screening-level estimate of the monetized PM-related health impacts
associated with an increase in NOx emissions associated with the final rule. For this analysis, we
derived a dollar-per-ton value based on recent benefits modeling conducted for the Clean Air
Nonroad Diesel Rule (CAND).ZZZZZ  This value ($8,000 in PM-related monetized health
impacts per ton of NOx reduced) is based on air quality modeling  conducted in 2004 for the
CAND rule. This benefits transfer method is consistent with approaches used in other recent
mobile and  stationary source rules.103  We refer the reader to the final CAND RIA for more
details on this benefits transfer approach.  The dollar-per-ton value represents monetized health
impacts in 2015  (in year 2000 dollars).

       We combined the dollar-per-ton estimate of monetized health effects with the projected
2015 emission changes presented in Table 4.4-1, which includes emissions from gasoline
vehicles and equipment and renewable fuel production and distribution. We estimate that the
potential PM2.5-related monetized impact associated with NOx emissions from increased use of
ethanol to be up to $290 million for the RFS  control scenario, and up to $340 million for the EIA
control scenario.  Note that this impact is based on monetized changes in health effects,
including changes in mortality risk, chronic bronchitis, nonfatal  heart attacks, respiratory hospital
admissions, asthma attacks, and other minor  health endpoints. It is also important to point out
that this value does not represent the cumulative monetized health impacts associated with the
potential PM changes associated with the future use of ethanol described above.

       This estimate  is subject to a number of additional caveats.  The dollar-per-ton values
reflect specific geographic patterns of emissions reductions and  specific air quality and benefits
modeling assumptions which are derived from previous analyses and will not match those
associated with increased ethanol use in fuel  for two reasons. One, the geographical distribution
of the emission sources affected by increased ethanol use differs from that addressed in the
CAND rulemaking.  Two, the CAND rule was national in scope and the emission  reductions
were spread out across the entire nation. Increased ethanol use will be very geographically
focused. Many major population centers will not experience an  increase in ethanol use as their
fuel already contains  ethanol. Care should be taken when applying these estimates to emission
reductions that occur in any specific location, since the dollars-per-ton for emission reductions in
specific locations may be very different than the national average.  Given these caveats and the
emissions from mobile sources. In contrast, gasoline-fueled vehicles and equipment comprise almost 90% of
national gaseous aromatic VOC mobile source emissions. The percentage increase in national NOX emissions due to
increased ethanol use should be smaller than the percentage decrease in national emissions of gaseous aromatics.
Finally, in most urban areas, ambient levels of secondary organic PM exceed those of secondary nitrate PM. Thus,
directionally, we expect a net reduction in ambient PM levels due to increased ethanol use. However, we are unable
to quantify this reduction at this time.

103 See: Clean Air Nonroad Diesel final rule (69 FR 38958, June 29, 2004); Nonroad Large Spark-Ignition Engines
and Recreational Engines standards (67 FR 68241, November 8, 2002); Final Industrial Boilers and Process Heaters
NESHAP (69 FR 55217, September 13, 2004); Final Reciprocating Internal Combustion Engines NESHAP (69 FR
33473, June 15, 2004); Final Clean Air Visibility Rule (EPA-452/R-05-004, June 15, 2005); Ozone Implementation
Rule (70  FR 71611, November 29, 2005).


                                            322

-------
potential decrease in ambient PM2.5 due to the decrease in aromatic fuel content, we can not say
for certain in which direction the total monetized PM-related health impact will be.  In reality
there may be an overall reduction in PM-related health costs, despite the increase due to
increased NOx emissions.
                                           323

-------
Chapter 8:  Agricultural Sector Impacts
       As described in the Preamble, we used the Forest and Agricultural Sector Optimization
Model (FASOM) to estimate the U.S. agricultural impacts of increasing renewable fuel volumes
to 7.5 billion gallons per year (BGY) by 2012, as required by the Renewable Fuels Standard
(RFS) and to 9.9 BGY, the volume of renewable fuel the Energy Information Administration
(EIA) predicts for the year 2012 in the Annual Energy Outlook 2006.104  Although these
renewable fuel volumes are lower than current market predictions, these assumptions were
established during the NPRM and are used throughout this FRM.

       FASOM is a long term economic model  of the U.S. agriculture sector that maximizes
total revenues for producers while meeting the demands of consumers. Using a number of
inputs, FASOM determines which crops, livestock, and processed agricultural products will be
produced in the U.S. In each model simulation,  crops compete for price sensitive inputs such as
land and labor at the regional level.  The cost of these and other inputs are used to determine the
price and level of production of primary commodities (e.g., field crops, livestock, and biofuel
products). FASOM also estimates prices using costs associated with the processing of primary
commodities into secondary products (e.g., converting livestock to meat and dairy, crushing
soybeans to soybean meal and oil). FASOM does not capture short-term fluctuations (i.e.,
month-to-month, annual) in prices and production, however, as it is designed to identify long
term trends.105

       FASOM uses supply and demand curves for the 11 major U.S.  domestic regions,106
which are calibrated to historic price and production data. FASOM also includes  detailed supply
and demand data for corn, wheat, soybeans, rice and sorghum across 37 foreign regions.107
FASOM maintains transportation costs to all regions and then uses all  of this information to
determine U.S. exports to the point where prices are then equated in all markets.108
104 We analyzed the U.S. agricultural impacts of producing renewable fuels domestically after adjusting for
equivalence values of cellulosic ethanol and biodiesel and projected U.S. imports. For the RFS Case, we assumed
440 million gallons of corn based ethanol will be imported, while we assumed 630 million gallons of corn based
ethanol will be imported for the EIA Case.  For both cases, we assume 250 million gallons of cellulosic ethanol will
be produced (with a 2.5 equivalence value), and 300 million gallons of biodiesel will be produced (with a 1.5
equivalence value).

105 FASOM calculates output in five year increments. For this analysis, 2010 and 2015 data were interpolated to
estimate 2012 values.

106 U.S. regions consist of the Pacific Northwest (West and East), Pacific Southwest, Rocky Mountains, Great
Plains, Southwest, South Central, Corn Belt, Lake States, Southeast, and the Northeast.

107 FASOM Foreign Regions include:  the European Economic Community, North Central Europe, Southwest
Europe, Eastern Europe, Adriatic, Eastern Mediterranean, Former Soviet Union, North Africa, East Africa, West
Africa, South Africa, Red Sea, Iran, India, Taiwan, Japan, South Korea, North Korea, China, Bangladesh, Indonesia,
Myanmar, Pakistan, Philippines, Thailand, Vietnam, West Asia, Southeast Asia, Australia, Caribbean, Eastern
Mexico, Eastern South America, Western South America, Argentina, Brazil, Canada,  Other.
                                            324

-------
8.1    Commodity Prices

8.1.1  Corn and DDGS Prices

       FASOM predicts that as renewable fuel volumes increase, agricultural prices over a range
of products (not just corn and soybean renewable fuel feedstocks) will increase as well. Since the
principal feedstock for ethanol is corn, corn prices are anticipated to rise. For consistency, all of
the dollar estimates are presented in 2004 dollars.  In the RFS Case, corn prices increase to
$2.50/bushel by 2012 (compared to a Reference Case price of $2.32/bushel in 2012). With the
higher renewable fuels volumes in the EIA Case, corn prices rise to $2.71/bushel (2004$) by
2012. (See Table 8.1-1)  To place this difference in perspective, in 2012, corn prices are about 8
percent higher in the RFS Case and 17 percent higher in the EIA Case relative to the Reference
Case.109
                        Table 8.1-1. Corn and DDGS Prices in 2012

Corn Price
Distillers Dried Grains with
Solubles (DDGS) Price
Reference Case
$2.32/bushel
$85.55/ton
RFS Case
$2.50/bushel
$83.35/ton
EIA Case
$2.7 I/bushel
$86.15/ton
       The cost of producing ethanol is dependent upon, among other factors, the price of corn
and the price of related byproducts.  As part of the analytical approach described in the NPRM,
we used FASOM to estimate the future prices of the major ethanol production byproduct:
distillers dried grains with solubles (DDGS). FASOM estimates that the price of DDGS will
remain relatively constant with the renewable volume scenarios that we are examining in this
rulemaking.  An increase in DDGS supply is anticipated to be offset by an increase in DDGS
demand as technology improves to pelletize and distribute DDGS to a wider market.  DDGS
prices in the U.S. in 2012 are predicted to be $83.35/ton in the RFS Case and $86.15/ton in the
EIA Case.  (See Table 8.1-1) Hence, the overall price of DDGS remains within 3 percent of the
DDGS Reference Case price.

       Note that the DDGS price given here is the price an ethanol producer would expect to
receive at the plant gate. FASOM predicts a higher value for the DDGS at the place of end use,

108 For additional details on the FASOM model, see the report by Professor Bruce McCarl, Texas A&M University,
"The Impacts of the Renewable Fuel Standard Program on the U.S. Agricultural Sector," February 2007, included in
the docket.

109 The current price of corn in the U.S. is approximately $3.50 per bushel (2004$), which is considerably higher
than the FASOM prediction and is a likely a result of the fact that recent demand for corn for ethanol is higher than
the currently available stocks.  The model results for 2012 reflects medium-term spatial equilibrium prices, where
rising demand for corn is  met by rising supply ~ due to increased acres planted to corn and to increased corn yields
per acre. Note that while the model assumes that markets for corn and related agricultural commodities will settle at
a price of $2.50 per bushel (in the RFS case) by 2012, that this may be a conservative estimate to the extent that the
agricultural sector is able  to adjust to the increased use of corn in ethanol production by 2012.
                                            325

-------
based on its nutritional ability to be substituted as half soy meal and half corn in animal feed.
The difference between its feed value and the ethanol plant gate price is made up of the cost of
handling and shipping, which may include pelletizing or other measures required to support a
national commodity market for DDGS.

8.1.2   Soybean and Soybean Byproducts

       FASOM predicts relatively modest changes in soybean prices as a result of the increases
in the renewable fuel volumes examined in this rulemaking. In the RFS Case, soybean prices rise
to $5.44/bushel (2004$) by 2012 (compared to a Reference Case price of $5.26/bushel in 2012).
In the EIA Case, soybean prices rise to $5.47/bushel  (2004$) by 2012.  (See Table 8.1-2)
Soybeans prices are expected to increase by about 3 percent (RFS Case) and 4 percent (EIA
Case) relative to the Reference Case by 2012.  The slightly higher prices of soybeans reflect the
consequences of the higher demand for soybeans for renewable fuels as well as the slightly
higher input costs (e.g., land prices). It is also expected that in medium-term the acres planted to
soybeans will fall, due to increased corn plantings, which will also increase soybean prices.

                  Table 8.1-2.  Soybean and Soybean Meal Prices in 2012

Soybean Price
Soybean Meal Price
Reference Case
$5.26/bushel
$176.70/ton
RFS Case
$5.44/bushel
$171.73/ton
EIA Case
$5.47/bushel
$170.05/ton
       Soybean meal is produced when crushing soybeans and extracting soybean oil, the
primary feedstock of biodiesel in the U.S. Under the RFS scenario, FASOM estimates the price
of soybean meal will decrease by about 3 percent in 2012, relative to the Reference Case.110
(See Table 8.1-2)  This decrease is slightly larger under the EIA scenario, with the price of
soybean meal dropping by about 4 percent. Several factors influence the small change in
soybean meal prices. First, more acres of soybeans are being planted to rotate with increased
planting of corn, and this leads to increased soybean supplies.  Second, increased DDGS supplies
can substitute for soybean meal as  a feed ingredient by reducing the soybean meal needed in feed
rations using higher levels of DDGS. Third, the size of the livestock herd is smaller due to higher
meat prices, reducing the demand for animal feeds overall.
  The current price of soybeans in the U.S. is considerably higher than the FASOM prediction and is a likely a
result of the fact that the market expects acres planted to soybeans in the short term are likely to decline due to
increased corn plantings. As with the corn results,.the model reflects medium-term spatial equilibrium prices, where
rising demand for corn is met by rising supply ~ due to increased acres planted to corn and to increased corn yields
per acre by 2012. Similarly, over time, farmers will begin to plant more soybeans in response to relatively higher
short-term prices. The model expects soybean prices to reach an equilibrium price of $5.44 per bushel (in the RFS
case) by 2012.
                                            326

-------
8.2    Impact on U.S. Farm Income
       The increase in renewable fuel production provides a significant increase in farm income
to the U.S. agricultural sector. FASOM predicts that in 2012, U.S. farm income from the sale of
agricultural commodities will increase by $2.65 billion dollars in the RFS Case and $5.41 billion
in the EIA Case. (See Figure 8.2-1) The RFS and EIA farm income changes represent roughly a
5 and 10 percent increase, respectively, in U.S. farm income from the sale of farm commodities
over the Reference Case of roughly $53 billion111. Most of the increase in net income is likely to
be concentrated in rural areas, and may contribute to rural wealth creation.

       Figure 8.2-1. Change in Net Farm Income Relative to Reference Case in  2012
      6.00 -—
      5.00
!2  4.00
to
&
°  3.00
c
o
ffi  2.00
      1.00
      0.00
                                   $2.65
                                                       $5.41
                                              2012
                                 D FASOM RFS • FASOM EIA
8.3    Impact on Employment

       Agricultural employment was not directly modeled but is likely to be very small since
modern farm practices are not labor intensive and increases in production as modeled here will
have negligible impact on direct farm employment. Some additional employment will result
111 While U.S. government farm payments are currently part of the U.S. farm income, what programs will be in
place in 2012 and their impact on farm income is unclear. For our modeling, we assumed the support programs
were in place in 2010 but none were in place in 2015; interpolation between 2010 and 2015 provided the assumed
impact in 2012.
                                          327

-------
from ethanol plant construction and operation.  30 to 50 people per ethanol production facility
seems typical.
8.4    Commodity Use Change

8.4.1   Corn and Ethanol Byproducts

       For this analysis, U.S. corn uses are broken down into four categories: domestic (i.e.,
household) consumption, ethanol production, livestock feed, and U.S. corn exports. (See Figure
8. 4-1) As the demand for corn increases to produce more renewable fuel, U.S. corn utilization
patterns are expected to be altered. In 2005, approximately 13 percent of all corn produced in the
U.S. was used for ethanol production. With the two renewable fuel volumes that we are
examining, the percentage of corn feedstock used for renewable fuels increases significantly. By
2012, in the RFS Case, 20 percent of all corn produced in the U.S. is used to produce ethanol. In
2012, in the EIA Case, 26 percent of all corn is used to produce fuel ethanol. These estimates are
similar to the percentages included in the NPRM.

       The increasing use of corn for ethanol raises the price of corn which has a direct impact
on the other uses of corn.  FASOM predicts higher U.S. corn prices leads to lower U.S. exports
of corn. U.S. corn exports drop from about 2 billion bushels in the Reference Case to 1.6 billion
bushels in the RFS  Case and 1.3 billion bushels in the EIA Case by 2012. In value terms, U.S.
exports of corn fall by $573 million  in the RFS Case and by $1.29 billion in the EIA Case in
2012.

                            Figure 8.4-1. Corn Uses in 2012
   
-------
       Higher U.S. prices for corn due to increased demand for ethanol production results in
decreased use of corn for U. S. livestock feed. The amount of corn used for livestock feeding
decreases by about 320 million bushels in the RFS Case and by about 690 million bushels in the
EIA Case relative to the Reference Case. Substitutes are available for corn as a feedstock, and
this market is highly price sensitive.  One alternate feedstock is DDGS because feed ration using
increased levels of DDGS would need less corn.  The relatively flat prices for DDGS predicted
across all ethanol volume scenarios results from the significant increase in the demand for DDGS
as a feed ingredient parallels the increase in supply of DDGS. FASOM estimates that DDGS use
for livestock feeding for the RFS Case will almost double by 2012, increasing from 8.5 million
tons to 15.2 million tons.  Under the EIA Case, FASOM predicts that DDGS will increase to
22.2 million tons by 2012. (See Figure 8.4-2) Domestic (i.e., household) consumption of corn
for food use declines  slightly with the different renewable fuel volumes analyzed in this
rulemaking.

                      Figure 8.4-2. Livestock Feed Sources in 2012
    600
    500
    400
    300
    200
    100
            FASOM Reference Case
FASOM RFS
FASOM EIA
    DCorn • DDGS D Hay • Silage D Soybean Meal D Wheat • Sorghum D Barley DOats • Gluten Meal
                                           329

-------
8.4.2   Soybean and Soybean Byproducts

       As described previously, only a modest increase in the demand for soybeans are expected
to be used to produce biodiesel in the renewable fuel scenarios analyzed in this rulemaking.
Although changes in soybean uses in this analysis are limited, U.S. exports are expected to drop
by 41.8 million bushels (RFS Case) and 35.6 million bushels (EIA Case).  In terms of export
earnings, U.S. exports of soybeans fall by $220 million in the RFS Case and by $194 million in
the EIA Case in 2012.
8.5    U.S. Land Use Patterns and Land Prices

8.5.1   Corn Acreage

       FASOM predicts that total production of corn in the U.S. in 2012 will be 11.9 billion
bushels under the RFS Scenario and 12.1 billion bushels under the EIA Scenario (compared to
11.7 billion bushels in the Reference Case)112. (See Table 8.5-1) With higher renewable fuel
volumes, more corn will be produced in the U.S.  Increased U.S. corn production can result from
two sources: greater productivity on existing acres of land devoted to corn or from "new" acres
that are brought into the corn production. Much of the high quality, suitable land in the U.S. is
already being used to produce corn. Improvement in the productivity of growing corn on existing
U.S. land is projected to grow by roughly 1 percent annually through 2012. As a result, most of
the increased demand for corn from increased use of renewable fuels will be met from increased
productivity on existing acres of corn relative to the 2005 baseline year. However,  corn
production from new acres plays an important role in corn supply. FASOM estimates an
increase in land devoted to corn production of 1.6 million acres (RFS Case) and 2.6 million acres
(EIA Case) in 2012 compared to the Reference Case.

                                      Table 8.5-1.
    U.S. Corn Acres Harvested, Corn Production, and Agricultural Land Prices in 2012

Corn Acres Harvested
(million acres)
Total Corn Production
(billion bushels)
Land Prices (percent
increase relative to
Reference Case)
Reference Case
78.5
11.7
N/A
RFS Case
80.1
11.9
8.4%
EIA Case
81.1
12.1
16.8%
  ; FASOM includes corn equivalent feeding of by products.in the estimate for total corn production.
                                          330

-------
       Higher renewable fuel volumes will have a direct impact on the value of U.S. agricultural
land.  As demand for corn and other farm products increases, the price of U.S. farm land will
also increase. The FASOM analysis shows that in 2012, higher renewable fuel volumes increase
average agricultural land prices in the U.S. by about 8 percent (RFS Case) and  17 percent (EIA
Case).

8.5.2   Soybean Acreage

       Increasing use of biodiesel fuel in the renewable fuel scenarios does not cause a
significant change in U.S. soybean production.  Soybean production stays relatively flat at 3.3
billion bushels in all three scenarios analyzed. (See Table 8.5-2) Soybean acreage increases
modestly as well in the renewable fuel scenarios examined. In the RFS Case, total soybean acres
are 74.6 million. For the EIA Case, total soybean acres are 74.4 million acres, compared to 73.4
million acres of soybeans in the Reference Case.

       Table 8.5-2. U.S. Soybean Acres Harvested and Soybean Production in 2012

Soybean Acres Harvested
(million acres)
Total Soybean Production
(billion bushels)
Reference Case
73.4
o o
J.J
RFS Case
74.6
3.3
EIA Case
74.4
3.3
8.5.3   CRT Acreage

       Current lands in the Conservation Reserve Program (CRP) total approximately 40 million
acres. To qualify for inclusion in the CRP, the acres must have been at one time in active
agricultural use. Farmers are paid to take these lands out of production and place them in CRP to
provide environmental benefits, including limiting erosion and providing wildlife habitat.
Farmers put land into the CRP voluntarily, considering among other factors the value of the land
if it were to remain in agricultural production versus the amount paid under the CRP contract.
The amount of government payments can change over time.

       For this analysis, we have assumed current per-acre payment levels to landowners are
maintained through 2012. However higher commodity prices and higher land rents associated
with higher renewable fuel volumes would likely require higher CRP payments to maintain the
same level of CRP enrollments.  The RFS and EIA renewable fuel volumes are estimated to
result in CRP withdrawals of 2.3 million and 2.5 million acres, respectively, relative to the
Reference Case. Most of the CRP lands are not likely to go into corn or soybean production
since much of the CRP lands tend to be marginal lands due to their location and productivity.
For example, only a relatively small portion of CRP lands are in the Corn Belt. Instead,
additional corn or soybeans acres will probably be planted on lands that were previously used for
other crops or pasture,  for example, wheat, grain, sorghum or planted forage crops.  It is
                                          331

-------
expected that some of the land removed from CRP will be used for these other agricultural
purposes.  Table 8.5-3 depicts the estimate of CRP impacts.
             Table 8.5-3.  CRP Acreage Changes Relative to Reference Case

Reduction in CRP Acreage (million acres)
RFS Case
2.3
EIA Case
2.5
8.6    Fertilizer Use

       Under the RFS scenario, the total amount of nitrogen applied on all farms increases by
1.2 percent, or 480,000 pounds, relative to the Reference Case in 2012. Under the EIA scenario,
the total amount of nitrogen applied on all farms increases by 2 percent, or 790,000 pounds,
relative to the Reference Case in 2012.  (See Table 8.6-1) We note that this percent increase in
fertilizer is largely accounted for by the 2 percent increase in land used for corn production and 1
percent increase in land for soybean production.  The fact that the amount of nitrogen used
increases at a smaller percent than the amount of land increase for corn production suggest that
much of the corn production land is already in agricultural use (with fertilizer applied) and is not
likely to be land newly released from CRP.

                   Table 8.6-1.  Nitrogen and Phosphorous Use in 2012

Total Nitrogen Applied
(million pounds)
Total Phosphorous Applied
(million pounds)
Reference Case
40.28
4.24
RFS Case
40.76
4.27
EIA Case
41.07
4.29
       Under the RFS scenario, the total amount of phosphorous applied on all farms increases
by 0.7 percent, or 30,000 pounds, relative to the Reference Case in 2012. Under the EIA
scenario, the total amount of phosphorous applied on all farms increases by 1.2 percent, or
50,000 pounds, relative to the Reference Case in 2012.  See Table 8.6-1.
8.7    Environmental Analysis

       Although this analysis does not include a comprehensive and integrated environmental
assessment of the impacts in the agricultural sector of higher renewable fuel volumes from this
rulemaking, we looked at two factors directly impacted by the production of agricultural crops
                                          332

-------
that may relate to environmental impacts. FASOM does estimate the amount of fertilizer used
and changes in CRP land, two indicators that could be associated with water pollution.
113
       Marathon commented that it believes that EPA's assessment of environmental impacts
does not consider all environmental impacts and is therefore incomplete, especially with respect
to water quality impacts.  As described above, our analysis predicts a modest increase in
fertilizer use and modest withdrawals of CRP lands due to the higher renewable fuel volumes.
While increased agricultural development would likely increase pressure on environmentally
sensitive areas such as wetlands and prairie lands and rural ecosystems in general, FASOM does
not represent this level of land detail in the national model and therefore cannot quantify any
potential impacts on these subsets of land types. To the extent that CRP withdrawals are
managed in an environmentally sustainable way, however, water pollution impacts would be
minimized.

       Increasing worldwide demand for biofuels and decreasing U.S. exports of feedstocks
used in producing renewable fuels will likely lead to increased prices, production, and different
trade patterns for renewable fuel feedstocks (i.e., corn and soybeans) in parts of the world outside
of the U.S. FASOM includes the export effect as it contains supply curves for rest of world
production of key agricultural products, but it does not contain a mechanism for appraising world
environmental implications since FASOM is a  domestic model of the U.S. agricultural sector.
Therefore, this analysis focuses only on impacts of the higher renewable fuels volumes in the
U.S.
8.8    U.S. Food Prices

       Despite the wider use of U.S. agricultural feedstocks, principally corn, for renewable
fuels, FASOM estimates only a modest increase in U.S. household food costs.  Annual wholesale
U.S. food costs are estimated to increase by approximately $7 per person with the RFS
renewable volumes and by about $12 per person annually with the EIA renewable volumes by
2012. (See Figure 8.8-1) Agricultural costs are only a portion of ultimate household food costs
so significant increases in corn prices and, to a lesser degree, soybean prices results in a much
smaller relative increase in household food costs.
113 The FASOM model can describe the proportion of fertilizer that potentially will affect groundwater quality and
surface water quality.  FASOM also details the extent to which shifts in agricultural production may affect soil
erosion and carbon sequestration.  In the short timeframe available, we were not able to devote significant efforts to
this type of analysis, but this area of inquiry could be investigated more extensively in the future. We do note that
we capture the sequestration impacts in the GREET analysis.


                                           333

-------
                 Figure 8.8-1. Increase in Annual Food Costs Per Person
                            Relative to Reference Case in 2012
      14
      12
      10
       0
   _re   8
   "o
   a
   I   *
   cs
                                  $7.0
$11.9
                                             2012
                                  DFASOM RFS • FASOM EIA
       FASOM estimates a relatively modest increase in U.S. prices for meat and agricultural
products associated with the higher renewable fuel volumes.  When evaluating changes in overall
U.S. food prices, FASOM uses the All Farm Products Price Index, which is a weighted average
of prices received by farmers at the "farm gate" for crop and livestock products relative to the
Reference Case.114  FASOM estimates a 4 percent increase in the RFS Scenario and a 7 percent
increase in the EIA Scenario in the weighted price of all farm products. (See Table 8.8-1)

       To evaluate changes in U.S. meat prices, FASOM uses the All Meat Products Price Index
which is a weighted average of the prices that farmers receive for meat products at the farm gate.
This index is based upon changes in the weighted average of beef, pork, chicken, and turkey
prices. U.S. meat prices that farmers receive in 2012 are estimated to increase by 0.3% in the
RFS Case and by 1.3% in the EIA Case compared to the Reference Case.
114 The All Farm Products Price Index includes: cotton, corn, soybeans, wheat, sorghum, rice, oats, barley, silage,
hay, sugarcane, sugar beet, potatoes, tomatoes, oranges, grapefruit, switch grass, hybrid poplar, willow, beef, cows,
milk, pigs, lamb, wool, horses and mules, eggs, chicken, and turkey.
                                           334

-------
   Table 8.8-1. Increase in the All Farm Products Price Index and the All Meat Products
                   Price Index Relative to the Reference Case in 2012

All Farm Products Price
All Meat Products Price
RFS Case
3.8% increase
0.3% increase
EIA Case
6.9% increase
1.3% increase
       Because corn is a major component of the All Crop Price Index, a significant change in
corn prices will result in a pronounced change in this index. The impact of corn price changes
on the Meat Price Index will be less pronounced for two reasons.  First, as corn prices rise, meat
producers will modify  feed rations and production systems to reduce their corn usage.  Second,
there will also be substitution among meats leading to higher consumption of meat from animals
using less of the higher priced corn (e.g., increased production of poultry products relative to
beef products).
                                          335

-------
Chapter 9: Small-Business Flexibility Analysis
       This chapter presents our Small Business Flexibility Analysis (SBFA) which evaluates
the potential impacts of the new standards on small entities. The Regulatory Flexibility Act, as
amended by the Small Business Regulatory Enforcement Fairness Act of 1996 (SBREFA),
generally requires an agency to prepare a regulatory flexibility analysis of any rule subject to
notice and comment rulemaking requirements under the Administrative Procedure Act or any
other statute unless the agency certifies that the rule will not have a significant economic impact
on a substantial number of small entities. Prior to issuing a proposal for this rulemaking, we
analyzed the potential impacts of these regulations on those entities that we believe are small
entities (see section 9.3, below). As a part  of this analysis, we conducted outreach with those
entities to gather information and recommendations from these entities on how to reduce the
impact of the rule on small businesses.
9.1    Requirements of the Regulatory Flexibility Act

       We are generally required under the Regulatory Flexibility Act (RFA) to conduct a
regulatory flexibility analysis unless we certify that the requirements of a regulation will not
cause a significant impact on a substantial number of small entities. The key elements of the
RFA include:

•      a description of and, where feasible, an estimate of the number of small entities to which
       the proposed rule will apply;
       the projected reporting, record keeping, and other compliance  requirements of the
       proposed rule, including an estimate of the classes of small entities which will be subject
       to the requirements and the type of professional skills necessary for preparation of the
       report or record;
       an identification to the extent practicable, of all other relevant  Federal rules which may
       duplicate, overlap, or conflict with the proposed rule; and,
•      any significant alternatives to the proposed rule which accomplish the stated objectives of
       applicable statutes and which minimize any significant economic impact of the proposed
       rule on small entities.

       The RFA was amended by SBREFA to ensure that concerns regarding small entities are
adequately considered during the development of new regulations that affect them. Although we
are not required by the Clean Air Act to provide special treatment to small businesses, the RFA
requires us to carefully consider the economic impacts that our proposed rules will have on small
entities. Specifically, the RFA requires  us to determine, to the extent feasible, our rule's
economic impact on small entities, explore regulatory options for reducing any significant
economic impact on a substantial number of such entities, and explain our ultimate choice of
regulatory approach.

       We have concluded that the final RFS  rule will not have  a significant impact on a
substantial number of small entities.  We based this conclusion on several criteria.  First, the
                                          336

-------
industry is expected to be overcomplying by a wide margin independent of the standard, thus
causing compliance costs to be minimal. Second, the Energy Policy Act of 2005 (Energy Policy
Act) already provides relief from the renewable fuels standards until 2011 for the majority of the
small entities; and lastly, we are extending this relief to the remaining small entities.  This is
discussed further below.
9.2    Need for the Rulemaking and Rulemaking Objectives

       A detailed discussion on the need for and objectives of this rule are in the preamble to the
final rule. As previously stated, EPA is required to promulgate regulations implementing a
renewable fuel program under Section 1501 of the Energy Policy Act, which amended the Clean
Air Act by adding Section 21 l(o). The Energy Policy Act requires EPA to establish a program
to ensure that U.S. gasoline contains specific volumes of renewable fuel for each calendar year
beginning in 2006, to increase the amount of renewable fuel used in vehicles and engines in the
U.S.
9.3    Description of Affected Entities

9.3.1  Definition of Small Entities

       Small entities include small businesses, small organizations, and small governmental
jurisdictions. For the purposes of assessing the impacts of the rule on small entities, a small
entity is defined as: (1) a small business that meets the definition for business based on the Small
Business Administration's (SBA) size standards (see Table 9-1); (2) a small governmental
jurisdiction that is a government of a city, county, town, school district or special district with a
population of less than 50,000; and (3) a small organization that is any not-for-profit enterprise
which is independently owned and operated and is not dominant in its field. Table 9.3.1-1
provides an overview of the primary SBA small business categories potentially affected by this
regulation.

                         Table 9.3.1-1. Small Business Definitions
Industry
Gasoline refiners
Defined as small entity by SBA if:
<1,500 employees115
NAICS Codes a
324110
     North American Industrial Classification System
9.3.2  Summary of Small Entities to Which the Rulemaking Will Apply

       The refiners that are potentially affected by this proposed rule are those that produce
gasoline. For our recent rulemaking "Control of Hazardous Air Pollutants from Mobile Sources"
115 In the Draft RIA, we also referred to a 125,000 barrels of crude per day (bpcd) crude capacity limit.  This
criterion was inadvertently used and is not applicable for this program (as it only applies in cases of government
procurement).  We note that the number of small entities remains the same whether this criterion is used or not.


                                           337

-------
(72 FR 8428, February 26, 2007), we performed an industry characterization to determine the
universe of potentially affected gasoline refiners. Information about the characteristics of
refiners comes from sources including the Energy Information Administration within the U.S.
Department of Energy, and from Hoover's (a division of Dun and Bradstreet). The refining
industry is located primarily in NAICS code 324110.

       The industry characterization was then used to determine which refiners met the SBA
definition of a  small refiner. From the industry characterization, and further analysis following
the Notice of Proposed Rulemaking (71 FR 55552,  September 22, 2006), we determined that
there were 15 gasoline refiners (owning 16 refineries) that met the definition of a small refiner.
It should be noted that because of the dynamics in the refining industry (e.g., mergers and
acquisitions), the actual number of refiners that ultimately qualify for small refiner status could
be different from this estimate.

       Title  XV the Energy Policy Act provides, at Section 1501(a)(2) [42 U.S.C.
7545(o)(9)(A)-(D)], special provisions for "small refineries", which includes a temporary
exemption from the standards until calendar year 2011.  Further, the Energy Policy Act states
that EPA must use the definition of "small refinery" and apply the special provisions provided
for small refineries in the RFS program. The Energy Policy Act defines the term "small
refinery" as "... a refinery for which the average aggregate daily crude oil throughput for a
calendar year.. .does not exceed 75,000 barrels."

       A small refinery (as defined by the Energy Policy Act) is very different from a small
refiner (as defined in SBA's regulations at 13 CFR  121.201). Per 13 CFR 121.201, and as stated
above in Table 9-1, a small refiner is a small business that employs less than or equal to 1,500
employees. A  small refinery, per the Energy Policy Act, is a small-capacity refinery and could
be owned by a larger refiner that exceeds the criterion specified in SBA's small entity definition;
whereas small  refiners generally only own a few (and more often than not, only one) refineries.

       In our analysis of the potentially affected small refiners,  we found that 42 refineries met
the Energy Policy Act's definition of a small refinery. Of these, we determined that 13 of these
refineries were owned by small refiners. Therefore, 12 of the 15 small refiners owned refineries
that also met the Energy Policy Act's definition of a small refinery. As a result, we believe that
all but three small refiners would automatically be granted relief by implementing the provisions
specified in the Energy Policy Act.
9.4    Issues Raised By Public Comments

       During the public comment period we received numerous comments regarding various
aspects of the proposed rule, including our proposed small refiner provisions.  The following
section provides a summary of the comments that we received on our proposed provisions. More
information on these comments can be found in the Final Summary and Analysis of Comments,
which is a part of the rulemaking record.
                                          338

-------
9.4.1   Extension of the Small Refinery Exemption to Small Refiners

       Commenters that supported the provision extending the small refinery exemption to small
refiners generally stated that they believe that a small refiner exemption is necessary as those
entities that would qualify as small refiners are generally at an economic disadvantage due to
their company size - whereas the Energy Policy Act only recognizes facilities, based on the size
of each location. These commenters also stated that they have concerns with the cost and the
availability of credits under the RFS program, and believe that provisions for small refiners are
necessary to help mitigate any significant adverse economic impact on these entities.

       Commenters that opposed the provision commented that they believe that EPA exceeded
its discretionary authority, that there appears to be no basis on which the Agency can legitimately
expand this statutory exemption to add small refiners, and that Congress "clearly did not intend
that the exemption be broadened to also include small refiners."  One commenter also stated that
it does not believe that small  refiner provisions are necessary  because the RFS program does not
require costly capital investments like previous fuel regulations.

9.4.2   Application Deadline

       We proposed that refiners would need to apply for the small refinery exemption, and that
the exemption would be effective 60 days after receipt of the  application by EPA (unless EPA
notifies the applicant that the application was not approved or that additional documentation is
required). We received comments on this provision in which commenters stated that requiring
small refinery applications was inconsistent with the language set out in the Energy Policy Act.
The commenters stated that the Energy Policy Act intended that small refineries would
automatically receive the small refinery exemption upon the effective date of the standard, and
that these parties should not be considered obligated parties in 2007 even if they do not submit a
small refinery application.

9.4.3   Provisions for Foreign Small Refineries and Refiners

       For consistency with prior gasoline-related fuel programs, we also proposed to extend the
RFS small refinery (and small refiner) exemption to foreign refiners, and we requested comment
on this provision. We received some comments in which commenters stated that they believe
that there is no reason to extend the small refinery exemption to these refiners.  One commenter
even stated that it believed that such an allowance would be unlawful.

9.4.4   Other

       We received some comments which stated that EPA needed to clarify whether or not
exempt small refineries (and  small refiners) could separate a RIN simply by owning a batch of
fuel. We also received  a comment which stated that it was not clear in the proposed rule whether
or not small refineries (and small refiners) blending ethanol at a terminal or any location without
formally  opting into the program could separate RINs.
                                          339

-------
9.5    Related Federal Rules

       Other current and proposed Federal rules that are related to this rule are: the Mobile
Source Air Toxics (MSAT2) rule (72 FR 8428, February 26, 2007), the Tier 2 Vehicle/Gasoline
Sulfur rulemaking (65 FR 6698, February 10, 2000), and the fuel sulfur rules for highway diesel
(66 FR 5002, January 18, 2001) and nonroad diesel (69 FR 38958, June 29, 2004).
9.6    Projected Reporting, Recordkeeping, and Other Compliance
Requirements

       For any fuel control program, EPA must have the assurance that refiners meet the
applicable standards.  Thus, requirements are imposed to ensure that compliance obligations are
met.

       The recordkeeping, reporting and compliance provisions of this program are fairly
consistent with those currently in place for our other 40 CFR part 80 fuel programs, including the
highway and nonroad diesel and MSAT regulations. These provisions include:

       Registration (the registration numbers will also be used in the RINs)
       Submission of annual reports summarizing a refiner's annual gasoline production and a
       demonstration of its compliance with the renewable fuels standard and submission of
       annual reports detailing and tracking a refiner's RINs; EPA's Central Data Exchange will
       be used for report submissions
       Recordkeeping will consist of the retention of all compliance documents (such as Product
       Transfer Documents and all reports submitted to EPA) for at least five years

       For a more detailed discussion of these provisions, please see section IV of the preamble
to this final rule.
9.7    Steps to Minimize Significant Economic Impact on Small Entities

       As stated above, we conducted outreach to a number of stakeholders that met the
definition of a small entity to gain feedback and advice on the needs of small businesses and
potential challenges that these entities may face.  The feedback that we received from these
entities as a result of these meetings was used during the development of the proposed rule for
developing regulatory alternatives to mitigate the impacts of the rulemaking on small businesses.
General concerns raised by these entities were the potential difficulty  and costs of compliance
with the upcoming standards given the other fuel compliance requirements that the fuel refining
industry is subject to.  Below we discuss the regulatory flexibility alternatives and provisions
which are being finalized in this action.

       While we do not believe that the RFS program with just the statutorily-prescribed
temporary relief for small refineries would have a significant economic impact on a substantial
number of small entities, we continue to believe that some refiners, due to their size, generally
                                         340

-------
face greater challenges compared to larger refiners. These refiners generally have greater
difficulty in raising and securing capital for investing in capital improvements and in competing
for engineering resources and projects.  This rulemaking does not require that refiners make
capital improvements, however there are still costs associated with meeting the standard.  Thus,
we find it appropriate to extend the small refinery temporary exemption, as set out in the Energy
Policy Act, to small refiners.  Under this exemption, any gasoline produced at a refinery owned
by a small refiner will not be counted in determining the renewable fuel obligation of a refiner
until January 1, 2011; further, the small refiner may exclude gasoline produced at its refineries
from its compliance calculations. Beginning in 2011, refineries owned by small refiners will be
required to meet the same renewable fuel obligation as all other refineries.

       Past fuels rulemakings have included a provision that, for the purposes of the regulatory
flexibility provisions for small entities,  a refiner must also have an average crude capacity of no
more than 155,000 barrels of crude per  day (bpcd). To be consistent with these previous rules,
we are finalizing in this rule that refiners that meet this criterion (in addition to having no more
than 1,500 total corporate employees) will be considered small  refiners for the purposes of the
regulatory flexibility provisions for RFS program. Further, the refiner must have produced
gasoline at its refineries by processing crude oil through refinery processing units. We are also
finalizing that eligibility will be based on 2004 data.

       We agree with statements from commenters that the Energy Policy Act did in fact intend
to provide the small refinery exemption without the need for the submission of small refinery
applications,  and that these parties should receive the exemption upon the effective date of the
rule. We also believe that this should be the case for small refiners as well. Therefore, we are
finalizing that small refiners will also receive the exemption immediately upon the effective date
of the rule. However, to ensure that only those refiners who meet the criteria above receive this
exemption, we believe that it is necessary for refiners to verify that they do in fact meet the
criteria. Therefore, these refiners will also be required to submit a verification letter showing
that they meet the criteria for qualification as a small refiner for the regulatory flexibility
provisions. This letter will be similar to the small refiner status applications required under other
EPA fuel programs (and must contain all the required elements specified  at §80.1142 of the
regulations),  except the letter will not be due prior to the program. Small refiner status
verification letters for this rule that are later found to contain false or inaccurate information will
be void as of the effective date of this rule.  Small refiners who subsequently do not meet all of
the RFS program's regulatory flexibility qualification criteria (i.e., cease producing  gasoline by
processing crude oil, employ more than 1,500 people, or exceed the 155,000 bpcd crude oil
capacity limit) as a result of a merger with or acquisition of or by another entity, are disqualified
as small refiners, except in the case of a merger between two previously approved small refiners.
As in other EPA programs, where such  disqualification occurs, the refiner must notify EPA in
writing no later than 20 days following  the disqualifying event.

       We are finalizing the proposed provision allowing foreign refiners to apply for a small
refinery or small refiner exemption under the RFS program.  The Energy Policy Act definition of
"small refinery" is not limited to domestic  facilities, and we believe that we have the discretion
to apply the definition of small refinery, and the similar relief that we are providing  to small
refiners, to foreign producers.  We believe that this provision is necessary for consistency with
                                           341

-------
prior fuel programs (anti-dumping, MSAT, and the fuel sulfur rules) which allowed foreign
refiners to receive such exemptions. Under this provision, gasoline produced at approved
foreign small refineries, and by approved foreign small refiners, will be exempt from the RFS
standard such that obligated parties (importers or blenders) would not count these volumes
towards their renewable volume obligations.

       We are also finalizing the proposed provision that the automatic five year exemption, and
any small refinery extended exemptions (extensions of the small refiners exemption will only be
available to small refineries), may be waived upon notification to EPA. Gasoline produced by a
small refiner who waives its exemption will be included in the RFS program and will be included
in the gasoline used to determine the refiner's renewable fuel obligation. If a refiner waives the
exemption, the refiner will be able to separate and transfer RINs like any other obligated party.
However, exempt small refiners cannot separate a RIN simply by owning a batch, a RIN can
only be separated by these parties once the volume of renewable fuel is blended with gasoline or
diesel to produce a motor vehicle fuel (as stated in the regulations at §80.1129).  If a small
refiner does not waive its small refiner exemption, it can still separate and transfer RINs, but
only for the renewable fuel that the refiner itself blends into gasoline (i.e., the refinery operates
as an oxygenate blender facility).  Lastly, exempt small refiners who blend ethanol can separate
RINs from batches without formally opting in to the program.
9.8    Conclusions

       After considering the economic impacts of today's proposed rule on small entities, we do
not believe that this action will have a significant economic impact on a substantial number of
small entities.  While the Energy Policy Act provided for a temporary exemption for small
refineries from the requirements of today's proposed rule, these parties will have to comply with
the requirements following the exemption period. Therefore, we had to take into account the
economic effects of the program on small entities when they would need to comply with the
standard. As described in section VI of the preamble to this final rule, the annual projections of
ethanol production are greater than the annual renewable fuel volumes required by the Energy
Policy  Act. For example, in 2011, when the Energy Policy Act's small refinery exemption ends,
over one billion gallons in excess RINs are projected to be available.  Further, excess RINs are
anticipated for each year of the program. Due to this projected excess supply in comparison to
the standard, the cost of RINs should be very low—near the level of the transaction costs.

       Due to the low cost to affected small entities, and the projected RIN availability, as well
as the temporary relief provided to small refineries and small refiners, we do not believe that this
program will impose a significant economic impact on a substantial number of small entities.
                                          342

-------
                                    Endnotes
A Renewable Fuels Association (RFA), Ethanol Biorefmery Locations (Updated October 16,
2006).

B Ethanol Producer Magazine (EPM), Plant List (Downloaded October 18, 2006) monthly
magazine publications (June 2006 through October 2006).

c Ethanol Producer Magazine (EPM), monthly magazine publications (June 2006, July 2006,
August 2006, September 2006 and October 2006).

D ICF International, Ethanol Industry Profile (September 30, 2006).

E BioFuels Journal, News & Information for the Ethanol and BioFuels Industries (breaking news
posted June 16, 2006 through October 18, 2006).

F Renewable Fuels Association (RFA), Ethanol Biorefmery Locations (Updated June 19, 2006).

G Ethanol Producer Magazine (EPM), U.S. & Canada Fuel  Ethanol Plant Map (Spring 2006).

H International Fuel Quality Center (IFQC), Special Biofuels Report #75 (April 11, 2006).

1 "Biodiesel Performance, Costs, and Use", Anthony Radich, EIA page 6.

JNBB Survey September 13, 2006 "U.S. Biodiesel Production Capacity".

K From Independent Biodiesel Feasibility Group Presentation.

L "Pipeline Considerations for Ethanol", Department of Agricultural Economics, Kansas State
University, August 2002.  http://www.agmrc.org/NR/rdonlyres/4EEOE81C-C607-4C3F-BBCF-
B75B7395C881/0/ksupipelineethl.pdf, and "Shipping Ethanol Through Pipelines", American
Petroleum Institute, April 6, 2006,  http://api-
ep.api.org/printerformat.cfm?ContentID=54FDlAll-95FA-4B7C-ACElD7C6F121FBlC.

M Infrastructure Requirements for an Expanded Fuel Ethanol Industry, Downstream Alternatives
Inc., January 15,2002.

N "Ethanol Industry's Interest in River Sites Grows", River Transport News, Vol. 15, No. 10,
May 22, 2006.
                                         343

-------
0 Infrastructure Requirements for an Expanded Fuel Ethanol Industry, Downstream Alternatives
Inc., January 15,2002.

P Infrastructure Requirements for an Expanded Fuel Ethanol Industry, Downstream Alternatives
Inc., January 15,2002.

Q F.O. Licht, "World Ethanol Markets, The Outlook to 2015" (2006).

R EIA AEO 2006 Table 12: Petroleum Product Prices

s EIA Crude Oil Spot Pricing (http://tonto.eia.doe.gov/dnav/pet/pet_pri_spt_sl_d.htm)

T Sources: U.S. EPA Office of Transportation & Air Quality, List of Federal Reformulated
Gasoline Areas (Updated February 23, 2004) and U.S. EPA Office of Transportation & Air
Quality, 2004 RFG Fuel Survey Results
(http ://www. epa. gov/otaq/regs/fuel s/rfg/properf/rfgperf.htm).

u Source: U.S. EPA Office of Transportation and Air Quality, State Winter Oxygenated Fuel
Program Requirements for Attainment or Maintenance of CO NAAQS (November 2005).

v National Ethanol Vehicle Coalition (http://www.e85fuel.com/e85101/faqs/number  ffvs.php).

w EIA 2004 Petroleum Marketing Annually (Table 48: Prime Supplier Sales Volumes of Motor
Gasoline by Grade, Formulation, PAD District, and State).

x EIA Historical RFG MTBE Usage (file received from EIA representative on March 9, 2006).

Y EIA Monthly Energy Review, June 2006 (Table 10.1: Renewable Energy Consumption by
Source,  Appendix A:  Thermal Conversion Factors).

z FHWA Highway Statistics 2004: Estimated Use of Gasohol (April 2006).
AA
   EPA Reformulated Gasoline (RFG) Survey
(http ://www. epa. gov/otaq/regs/fuel s/rfg/properf/rf gperf.htm).

BB Alliance of Automobile Manufacturers (AAM) North American Fuel Survey 2004. The
report and data can be purchased through the AAM website,
http://autoalliance.org/fuel/fuel_surveys. php, or by contacting the Alliance at 1401 Eye Street,
N.W., Suite 900, Washington, DC 20005, Tel: (202) 326-5500.

cc EIA Monthly Energy Review, June 2006 (Table 10.1: Renewable Energy Consumption by
Source, Appendix A: Thermal Conversion Factors).

DD EIA Historical RFG MTBE Usage (file received from EIA representative on March 9, 2006).
                                          344

-------
EE U.S. EPA, State Actions Banning MTBE (Statewide), June 2004.

FF EIA 2004 Petroleum Marketing Annually (Table 48: Prime Supplier Sales Volumes of Motor
Gasoline by Grade, Formulation, PADD, and State).

GG EIA Monthly Energy Review June 2006 (Table 10.1: Renewable Energy Consumption by
Source, Appendix A: Thermal Conversion Factors).

HH EPA Reformulated Gasoline (RFG) Survey
(http ://www. epa. gov/otaq/regs/fuel s/rfg/properf/rfgperf.htm)

11 AAM North American Fuel Survey 2004 (http://autoalliance.org/fuel/fuel_surveys.php).

JJ EIA Annual Energy Outlook 2006 (Table 2: Energy Consumption by Sector and Source).

KK International Fuel Quality Center (IFQC), Special Report: United States - State Renewable
Content Standards (June 6, 2006).

LL Renewable Fuel News and World Refining & News Today (January 2006).

MM Modeling Impacts of Energy Bill RFS, Jacobs Consultancy December 2006.

NN The Energy Act Section 1504, promulgated on May 8, 2006 at 71 FR 26691.

°°EIA 2004 Petroleum Marketing Annually (Table 32: Conventional Motor Gasoline Prices by
Grade, Sales Type, PAD District, and State).
pp
  EIANEMS Petroleum Market Model Documentation, Appendix I, Table 14.
QQ American Coalition for Ethanol, STATUS: State by State Ethanol Handbook 2006
(supplemented by information obtained from state websites and conversations with state
government officials).

RR Final Regulatory Impact Analysis for Reformulated Gasoline, Table VI-A6, December 13,
1993.

ss U.S. EPA, Final Regulatory Impact Analysis for Reformulated Gasoline, Table VI-A6,
EPA420-R-93-017, December 13, 1993. Available from RFG Final Rule, Docket No. A-92-12,
February 1994.

TT U.S. EPA, EPA 's National Mobile Inventory Model (NMIM), A Consolidated Emissions
Modeling System for MOBILE6 and NONROAD, EPA420-R-05-024, December 2005, version
NMEVI20060310 with county database NCD20060201 modified according to discussion in
Chapter 2.2.
                                         345

-------
uu U. S. EPA, Guide on Federal and State Summer RVP Standards for Conventional Gasoline
Only, EPA420-B-05-012, November 2005.

vv Alliance of Automobile Manufacturers North American Fuel Survey 2005. The report and
data can be purchased through the AAM website, http://autoalliance.org/fuel/fuel_surveys.php,
or by contacting the Alliance at 1401 Eye Street, N.W., Suite 900, Washington, DC 20005, Tel:
(202) 326-5500. Information about the data and methods is contained in the report Survey
Sampling and Testing Details (May 2005), which may be downloaded from the AAM website at
no charge.
ww
XX
 U.S. EPA website www.epa.gov/otaq/rfg/whereyoulive.htm, as of April 11, 2006.

U.S. EPA, A Comprehensive Analysis of Biodiesel Impacts on Exhaust Emissions, Draft
Technical Report, EPA420-P-02-001, October 2002.

YY Caffrey, Peter I, and Paul A. Machiele, U.S. EPA, "In-Use Volatility Impact of
Commingling Ethanol and Non-Ethanol Fuels," SAE 940765, 1994.

zz "Analysis of California's Request for Waiver of the Reformulated Gasoline Oxygen Content
Requirement for California Covered Areas." U.S. EPA, EPA420-R-01-016, June 2001.

AAADurbin, T., Miller, J., Younglove, T., Huai, T., and Cocker, K. "Effects of Ethanol and
Volatility Parameters on Exhaust Emissions". CRC Project No. E-67. Coordinating Research
Council, Inc. 3650 Mansell Road,  Suite 140. Alpharetta, GA 30022. January, 2006.

BBB "AAM and AIAM Study: Sulfur Oxygen Vehicle Emissions Test Program".  Available on
the California Air Resources Board Website:
http://www.arb.ca.gov/fuels/gasoline/carfg3/carfg3.htm.

ccc "ExxonMobil: LEV/ULEV Gasoline Oxygenate Study". Available on the California Air
Resources Board's Website:  http://www.arb.ca.gov/fuels/gasoline/carfg3/carfg3.htm.

ODD "jOyOta study: Effects of Ethanol on Emissions of Gasoline LDVs".  Available on the
California Air Resources Board Website:
http://www.arb.ca.gov/fuels/gasoline/carfg3/carfg3.htm.

EEE Schifter, I, Diaz, L., and Lopez-Salinas, E. "A Predictive Model to Correlate Fuel
Specifications with On-Road Vehicles Emissions in Mexico". Environmental Science and
Technology. 2006, 40, pp. 1270-1279.

FFF "California Waiver Decision Document," US EPA, EPA420-S-05-005, June 2005.

GGG "Final Report, Effects of Gasoline Ethanol Blends on Permeation Emissions Contribution to
VOC Inventory from On-Road and Off-Road Sources," Air Improvement Resource, Inc., for the
American Petroleum Institute, March 3, 2005.
                                         346

-------
HHH Haskew, Harold M., Thomas F. Liberty and Dennis McClement, "Fuel Permeation from
Automotive Systems," Final Report, for the Coordinating Research Council and the California
Air Resources Board, CRC Project E-65, September 2004.

111  Haskew, Harold M., Thomas F. Liberty and Dennis McClement, "Fuel Permeation from
Automotive Systems: EO, E6, E10 and E85," Prepared for the Coordinating Research Council,
Interim Report CRC Project No. E-65-3, August 2006.

JJJ Ragazzi, Ron and Ken Nelson, "The Impact of a 10% Ethanol Blended Fuel on the Exhaust
Emissions of Tier 0 and Tier 1 Light-Duty Gasoline Vehicles at 35 F," Colorado Department of
Public Health and Environment, March 26, 1999. For additional information or copies, contact
Gerald L. Gallagher or Margie M. Perkins.

KKK Mulawa, Patricia. A. et. al.  "Effect of Ambient Temperature and E-10 Fuel on Primary
Exhaust Particulate Mater Emissions from Light-Duty Vehicles," Environ. Sci. Technol. 1997,
31, 1302-1307.

LLL Stump, F., et. al., Characterization of Emission from Malfunctioning Vehicles Fueled with
Oxygenated Gasoline - E10 Fuel, Part II and Part III, EPA RTF.

MMM "jnj^aj Mass Exhaust Emissions from Reformulated Gasolines". Technical Bulletin No. 1.
Auto/Oil Air Quality Improvement Research Program.  Coordinating Research Council Inc.
Atlanta, Georgia. December 1990.

NNN Hochhauser, A.M., et al. "Effects of Oxygenated Fuels and RVP on Automotive Emissions
- Auto/Oil Air Quality Improvement Research Program." Society of Automotive Engineers
publication No. 920326.

ooo gpA Certification and Fuel Economy Information System.
http://www.epa.gov/otaq/cfeis.htm.

ppp Benson, J., Knepper, J., Reuter, R., Koehl,W., Leppard, W., Rippon, B., Burns, V., Painter,
L., Rutherford, J., Hochhauser,  A., and Rapp, L. "Emissions with E85 and Gasolines in
Flexible/Variable Fuel Vehicles-The Auto/Oil Air Quality Improvement Research Program."
SAE paper no. 952508. Warrendale, PA:  SAE. October 1995.

QQQ Aakko, P. and Nylund, N.-O. Particle Emissions At Moderate and Cold Temperatures Using
Different Fuels.  SAE Technical Paper 2003-01-3285.

RRR "Exhaust Emission Effects of Fuel Sulfur and Oxygen on Gasoline Nonroad Engines," U.S.
EPA, December 2005, EPA420-R-05-016, NR-003c.
                                         347

-------
sss "Nonroad Evaporative Emission Rates," Table 2, U.S. Environmental Protection Agency,
EPA420-R-05-020, NR-012c, December 2005. This document is available in docket EPA-HQ-
OAR-2004-0008-0362.

TTT "EPA's National Mobile Inventory Model (NMEVI), A Consolidated Emissions Modeling
System for MOBILE6 and NONROAD," U.S. EPA, EPA420-R-05-024, December 2005.

uuu Jackson, Cleophas, "Ethanol Plant Emissions as Reported to the States," EPA Memorandum
to the Docket, February 16, 2007.


vvv U.S. EPA, EPA 's National Mobile Inventory Model (NMIM), A Consolidated Emissions
Modeling System for MOBILE6 and NONROAD, EPA420-R-05-024, December 2005, version
NMIM20061128 with county database NCD20061227RfsFinal modified according to discussion
in Chapter 2.2.
www n § EpA User ,s Guide to MOBILE6 j andMOBlLE6.2 Mobile Source Emission Factor
Model, EPA420-R-03-010, August 2003, model used version M6ChcOxFix.
XXX
    Gururaja ,Prashanth, U.S. EPA, "Fuel Tank Temperature generation for evaporative
emissions modeling in MOVES," EPA Memorandum, August 31, 2006.


YYY Federal Highway Administration, Highway Statistics 2004, Section V, Table VM-1 "Annual
Vehicle Distance Traveled in Miles and Related Data", October 2005.


zzz U.S. EPA, User's Guide for the Final NONROAD 2 005 Model, EPA420-R-05-013, December
2005, model used version NR05c-Bond Base.


AAAA y g EPA. 2006. Technical Support Document for the Proposed Mobile Source Air Toxics
Rule - Air Quality Modeling. This document is available in Docket EPA-HQ-OAR-2005-0036.

BBBB U.S. EPA. 2005. Technical Support Document for the Final Clean Air Interstate Rule - Air
Quality Modeling. Available from: http://www.epa.gov/cair/technical.html. This document is
available in Docket EPA-HQ-OAR-2003-0053.

cccc jj g  EPA 2005. Clean Air Interstate Rule Emissions Inventory Technical Support
Document.  Available from: http://www.epa.gov/cair/technical.html. This document is available
in Docket EPA-HQ-OAR-2003-0053.
                                        348

-------
DDDD Pielke, R.A., W.R. Cotton, R.L. Walko, et al. 1992. "A Comprehensive Meteorological
Modeling System - RAMS." Meteor. Atmos. Phys., Vol. 49, pp. 69-91.

EEEE Ozone Air Quality Effect of a 10% Ethanol Blended Gasoline in Wisconsin, Wisconsin
Department of Natural Resources, September 6, 2005.

FFFF Ragazzi, R., Nelson, K. "The Impact of a 10% Ethanol Blended Fuel on the Exhaust
Emissions of Tier 0 and Tier 1 Light Duty Gasoline Vehicles at 35F." Colorado Department of
Public Health and Environment. March 26, 1999. For additional information or copies, contact
Gerald L. Gallagher or Margie M. Perkins.

GGGG Mulawa, P. A. "Effect of Ambient Temperature and E-10 on Primary Exhaust Particulate
Matter Emissions from Light-Duty Vehicles." Environmental Science and Technology 31, pp
1302-1307. 1997.

HHHH Qrosjean^ j)  "jn sj^ Organic Aerosol Formation During a Smog Episode: Estimated
Production and Chemical Functionality." Atmospheric Environment. Vol. 26A. No. 6. pp. 953-
963: 1992.

1111 Schauer, J.J., Fraser, M.P., Cass, G.R., Simoneit, B.R.T., "Source Reconciliation of
Atmospheric Gas-Phase and Particle-Phase Pollutants during a Severe Photochemical Smog
Episode." Environmental Science and Technology 36, pp 3806-3814. 2002.

JJJJ  Pun, Betty K. and Christian Seigneur, "Investigative Modeling of New Pathways for
Secondary Organic Aerosol Formation," Final Report,  CRC Project A-59, Coordinating
Research Council, March 2006.

KKKK j^jngy Rjp pjy[ 25 field study.  Currently undergoing peer review and will be published
shortly.

LLLL Zheng, M., Cass, G.R., Schauer, J.J., Edgerton, E.S. "Source apportionment of PM2.5 in the
Southeast United States using solvent-extractable organic compounds as tracers." Environmental
Science and Technology 36, pp. 2361-2371. 2002.

MMMM y s EpA "Nationai Emissions Inventory (NEI) Air Pollutant Emissions Trends Data,
1970 - 2002 Average annual emissions."  Posted August 2005 on
http://www.epa.gov/ttn/chief/trends/, filename "NOX07182005.xls." (Accessed August 14,
2006). July 2005.

NNNN Delucchi, Mark A. (2003) A Lifecycle Emissions Model (LEM): APPENDIX D: CO2
Equivalency Factors. Institute of Transportation Studies, University of California,  Davis,
Research Report UCD-ITS-RR-03-17D.
                                         349

-------
oooo Kwaitkowski, j R ^ McAloon, A., Taylor, F., Johnston, D.B., Industrial Crops and
Products 23 (2006) 288-296.  A copy of the current USDA model can be obtained by contacting
the corresponding author.

PPPP Wang, Saricks, and Wu, 1997, Fuel-Cycle Fossil Energy Use and Greenhouse Gas
Emissions of Fuel Ethanol Produced from U.S. Midwest Corn, pp. 13; cited in ANL/ESD-38, pp.
65.

QQQQ Qervajs an(j Baumel, 19XX, The Iowa Grain Flow Survey: Where and How Iowa Grain
Producers Ship Corn and Soybeans, CTRE, Iowa State University.

RRRR The Energy Balance of Corn Ethanol: An Update, Shapouri, H, J.A. Duffield, and M.
Wang, 2002 AER-813, Washington DC: USDA Office of the Chief Economist.

ssss uSDA's 2002 Ethanol Cost-of-Production Survey, Shapouri, H, P  Gallagher, Agricultural
Economic Report Number 841, July 2005.

TTTT rpjie USDA National Agricultural Statistics Service (NASS) website, national statistics on
field corn: http://www.nass.usda.gov:8080/QuickStats/Create_Federal_All.jsp.

uuuu USDA Agricultural Baseline Projections to 2015, USDA Office of the Chief Economist,
World Agricultural Outlook Board, Baseline Report OCE-2006-1, Feb  2006.

vvvv Fossil Energy Use in the Manufacture of Corn Ethanol, Graboski, M., Report Prepared for
the National Corn Growers Association, August 2002.

wwww
           an(j ^jgggj fuej prices from USDA Agricultural Statistics report, Prices Paid by
Farmers, multiple years.

xxxxlbid.

YYYY Shapouri, H., Duffield, J., Mcaloon, A.J. the 2001 Net Energy Balance of Corn-Ethanol.
2004. Proceedings of the Conference on Agriculture As a Producer and Consumer of Energy,
Arlington, VA., June 24-25.

zzzz Fossil Energy Use in the Manufacture of Corn Ethanol, Graboski, M., Report Prepared for
the National Corn Growers Association, August 2002.

AAAAA
             j^ H  Duffield, J., Mcaloon, A.J. the 2001 Net Energy Balance of Corn-Ethanol.
2004. Proceedings of the Conference on Agriculture As a Producer and Consumer of Energy,
Arlington, VA., June 24-25.

BBBBB ISO i4044;2006(E), "Environmental Management - Life Cycle Assessment -
Requirements and Guidelines", International Organization for Standardization (ISO), First
edition, 2006-07-01, Switzerland.
                                         350

-------
ccccc Inventory of U.S. Greenhouse Gas Emissions and Sinks:  1990-2004, EPA 430-R-06-002,
April 2006.

DDDDD Inventory of 'U.S. Greenhouse Gas Emissions and Sinks:  1 990-2004, EPA 430-R-06-002,
April 2006.

EEEEE Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1 990-2004, EPA430-R-06-002,
April 2006.

FFFFF -j^et jmp0rts of petroleum include imports of crude oil, petroleum products, unfinished oils,
alcohols, ethers, and blending components minus exports of the same.

GGGGG petroieum products, according to the Annual Energy Outlook 2006, includes imports of
finished petroleum products, unfinished oils, alcohols, ethers, and blending components.

HHHHH mA (September 1997); "Petroleum 1996: Issues and Trends",  Office of Oil and Gas,
DOE/EIA-0615, p. 71. (http://tonto.eia.doe.gov/FTPROOT/petroleum/061596.pdf).

11111 Report entitled "Contribution of the ethanol industry to the economy of the United States,"
prepared by LECG, LLC February 21, 2006.  The impact figures in the report were generated
using the U.S. Bureau of Economic Analysis Regional Input-Output Modeling System.

JJJJJ Kwaitkowski, J.R., McAloon, A., Taylor, F., Johnston, D.B., Industrial Crops and Products
23 (2006) 288-296. A copy of the current USDA model can be obtained by contacting the
corresponding  author.

KKKKK Shapouri, H., Gallagher, P., USDA's 2002 Ethanol Cost-of-Production Survey (published
July  2005).

LLLLL
               p w., Brubaker, H., and Shapouri, H. "Plant Size: Capital Cost Relationships in
the Dry Mill Ethanol Industry," Biomass andBioenergy Vol. 28. 2005. Pp. 565-7.

MMMMM Baseline Energy Consumption Estimates for Natural Gas and Coal-based Ethanol Plants
- The Potential Impact of Combined Heat and Power (CHP), Prepared by Energy and
Environmental Analysis, Inc., July 2006.

NNNNN Energy Policy Act of 2005, Section 1501(a)(2).

ooooo Historical data at
http://tonto.eia.doe.gov/dnav/pet/pet_pri_allmg_d_nus_PTA_cpgal_m.htm (gasoline),
http://tonto.eia. doe.gov/dnav/ng/ng_pri_sum_dcu_nus_m.htm (natural gas),
http://www.eia. doe. gov/cneaf/electricity/page/sales_revenue.xls (electricity),
http://www.eia.doe.gov/cneaf/coal/page/acr/table28.html (coal).
                                          351

-------
ppppp EIA Annual Energy Outlook 2006, Tables 8, 12, 13, 15.

QQQQQ http://www.iogen.ca/company/about/index.html.

RRRRR Lignoceifaiosic Biomass to Ethanol Process Design and Economics Utilizing Co-Current
Dilute Acid Prehydrolysis and Enzymatic Hydrolysis Current and Futuristic Scenarios, Robert
Wooley, Mark Ruth, John Sheehan, and Kelly Ibsen, Biotechnology Center for Fuels and
Chemicals Henry Majdeski and Adrian Galvez, Delta-T Corporation; National Renewable
Energy Laboratory, Golden, CO,  July 1999, NREL/TP-580-26157.

sssss Determining the Cost of Producing Ethanol from Corn Starch and Lignocellulosic
Feedstocks; A Joint Study Sponsored by: USD A and USDOE, October 2000 • NREL/TP-580-
28893 • Andrew McAloon, Frank Taylor, Winnie Yee, USD A, Eastern Regional Research
Center Agricultural Research Service; Kelly Ibsen, Robert Wooley, National Renewable Energy
Laboratory, Biotechnology Center for Fuels and Chemicals, 1617 Cole Boulevard, Golden, CO,
80401-3393; NREL is a USDOE  Operated by Midwest Research Institute • Battelle • Bechtel;
Contract No. DE-AC36-99-GO10337.

TTTTT Lignoceiiuiosic Biomass to Ethanol Process Design and Economics Utilizing Co-Current
Dilute Acid Prehydrolysis and Enzymatic Hydrolysis Current and Futuristic Scenarios, Robert
Wooley, Mark Ruth, John Sheehan, and Kelly Ibsen, Biotechnology Center for Fuels and
Chemicals Henry Majdeski and Adrian Galvez, Delta-T Corporation; National Renewable
Energy Laboratory, Golden, CO,  July 1999, NREL/TP-580-26157.

uuuuu        information Administration, U.S. Department of Energy. See publications,
Annual Energy Outlook, 2006.

vvvvv Energy Policy Act of 2005: TITLE XV— ETHANOL AND MOTOR FUELS, Subtitle
A— General Provisions, SEC. 1511, 1512, 1514.

wwwww
                    Requirements for an Expanded Fuel Ethanol Industry, Downstream
Alternatives Inc., January 15, 2002.

xxxxx jabies ES-9 ancj £§ \Q^ Infrastructure Requirements for an Expanded Fuel Ethanol
Industry, Downstream Alternatives Inc, January 15, 2002.

YYYYY petroieum Market Model of the National Energy Modeling System, Part 2, March 2006,
DOE/EIA-059 (2006), http://tonto.eia.doe.gov/FTPROOT/modeldoc/m059(2006)-2.pdf .

zzzzz
     U.S. Environmental Protection Agency.  May 2004.  Final Regulatory Analysis: Control of
Emissions from Nonroad Diesel Engines. Prepared by: Office of Air and Radiation. EPA420-R-
04-007  Available athttP://www-ePa-gov/nonroad-diesel/2004fr.htm#documents .  Accessed
August 1, 2006.
                                         352

-------