The Life Cycle Assessment of Cellulosic Ethanol Production in The

Home , Cellulosic ethanol

The Life Cycle Assessment of Cellulosic Ethanol Production in the Wisconsin and Michigan Agricultural Contexts: The Influence of LCA Methods and Spatial Variability on Environmental Impact Assessment

Julie C. Sinistore

A dissertation submitted in partial fulfillment of

the requirements for the degree of

Doctor of Philosophy

(Biological Systems Engineering)

at the

UNIVERSITY OF WISCONSIN-MADISON

2012

Date of final oral examination: 07/13/2012

The dissertation is approved by the following members of the Final Oral Committee:

Douglas J. Reinemann, Professor, Biological Systems Engineering

Robert P. Anex, Professor, Biological Systems Engineering

Carol L. Barford, Associate Scientist, Sustainability and Global Environment

Randall D. Jackson, Associate Professor, Agronomy

Christopher J. Kucharik, Associate Professor, Agronomy

Paul J. Meier, Associate Scientist, Engineering Physics, Energy Institute

I. DEDICATIONS

To Aunt Joan for teaching me so many things like how to ride a bike and how to work hard.

To Mr. Berger for all of the fatherly advice and for encouraging me to join a choir.

To my dear Anthony for putting up with me all these years.

To the dearly departed Dr. Joan Ehrenfeld of Cook College, Rutgers University and Dr. Josh Posner of UW-Madison. You are greatly missed, but your contributions to this world live on.

To Hillary Colton for teaching me to make new mistakes.

To the Kazmans for teaching me how to be a performer.

To Mrs. McGinn and Mr. Smith for teaching me how to think critically and show, not tell, in writing.

To Bonnie Berenger for introducing me to rigorous scientific research.

To Dr. Greg Graffin for reminding me to raise my voice.

To Francis Moore Lappé for inspiring me to care about the environment back in high school and reinvigorating my will to protect it.

And last, but certainly not least, to Dr. George Clark for the Tabasco sauce bottles and because you rule. ii

II. SPECIAL THANKS

Committee Members: Doug Reinemann, Rob Anex, Carol Barford, Randy Jackson, Chris Kucharik and Paul Meier

UW Faculty, Staff and TAs: Debby Sumwalt, Chris Elholm, Bill Bland, Karyn Biasca, Emily Stanley, Pat Eagan, Pat Walsh, Beverly Taylor, Bill Provencher, Luciana Moscoso Boedo, Mike Bell, Jonathan Foley, John Norman, Judith Harackiewicz, Stephen Powers, Chris Solomon, Teri Balser, William Beam, Birl Lowery, Philip Barak, Rick Wayne, I.M. Isaacs, Charles Burnette, Xuejun Pan, Gloria Mari-Beffa, Mike Lock, Ravikumar Prathivadi, Colin Hackbarth, Sigurd Angenent, Anna Saputura, Nicola Ferrier, Richard Shaten, Debra Deppeler, Leslie Watkins, Scott Sanders, Leslie Smith, David Dynerman, Marty Kanarek, Mario Trujilo, Franklin Miller, Ajit Naik and Gunasekaran Sundaram.

Rutgers Faculty, Staff and TAs: Kristen Goodrich, Zane Helsel, Priscilla Hayes, Tom Gianfagna, Doc Hamilton, Doc Locandro, Jay Kelly, David Ehrenfeld, Daniel Gimenez and Tim Casey.

Family and Friends: Grandma, Mom, Aunt Pam, Uncle Mark, Edgar, Jenny, Erin, Paul and Judy, Deborah Paulsen, Kristen Radecsky, Naveen VK, John Foster, Kat Farley, Aaliyah Green, Stacey Kennealy, Erica Foote, Anthony D’Agostino, Regen Borchardt, Holly Lahd, David Duncan, Natalie Hunt, Hannah Gaines, Horacio Aguirre-Villegas, Patty Yanez, Simone Kraatz, Thais Passos Fonseca, Sonia Ares Gomez, Alex Lyon, Laurel Gutenberg, Mary Saunders, Kiera Mulvey, Andrew Barrett, Rachel Mallinger, Martha Rideout, Katelin Holm, Jennifer Blazek, Herika Kummel, Nate Bard, Chris and Leslie Anderson, Carl Wahl, Jeff Sensei, Maria Sensei, Barb Sensei, Jane Sensei, John Sensei, Joanna, Sarah, Jared, Kian and Elena, all the folks at East Madison Karate and Aikido of Madison and anyone else who has ever been a friend to me!

Collaborators: The GLBRC, César Izaurralde, David Manowitz, Xuesong Zhang, Mac Post, Shujiang Kang, Pragnya Eranki, Bryan Bals, Tim Meehan, Sarynna Lopez, Bruce Dale, John Greenler, Sara Krauskopf, Steve Slater, David Pluymers and all the staff at the GLBRC.

Other Influences and Essential Partners: Andreas Kapsalis, Enya, the sounds of nature, farmer’s markets, the number eight, if/then and concatenate statements and keyboard shortcuts.

I could not have done this without all of you! iii

III. TABLE OF CONTENTS I. Dedications ...... i II. Special Thanks ...... ii III. Table of Contents ...... iii IV. List of Figures ...... vi V. List of Tables ...... ix VI. Dissertation Abstract ...... xi CHAPTER 1. Introduction ...... 1 1.1. Introduction ...... 1 1.2. Introduction to Life Cycle Assessment ...... 3 1.3. Introduction to Cellulosic Ethanol Production ...... 7 1.4. Challenges in Cellulosic Ethanol Production ...... 12 1.5. Dissertation Objectives and Format ...... 14 1.6. References ...... 17 CHAPTER 2. Applying Life Cycle Assessment to low carbon fuel standards – How allocation choices influence carbon intensity for renewable transportation fuels .....21 2.1. Publication and Contribution Details ...... 21 2.2. Abstract ...... 22 2.3. Introduction and Background ...... 23 2.4. Methods...... 28 2.4.1. System description and boundaries ...... 30 2.4.2. Process steps and emissions estimation ...... 33 2.4.3. Allocation approaches ...... 37 2.5. Results ...... 43 2.6. Discussion ...... 56 2.7. References ...... 60 CHAPTER 3. Methodological review of recent ethanol production studies and recommendations to facilitate comparison ...... 66 3.1. Publication and Contribution Details ...... 66 3.2. Abstract ...... 67 3.3. Introduction and Background ...... 68 iv

3.4. Methods...... 72 3.5. Results and Discussion ...... 84 3.5.1. Functional unit ...... 84 3.5.2. System boundary ...... 88 3.5.3. Allocation ...... 95 3.5.4. Impact category metrics ...... 101 3.6. Conclusions ...... 105 3.7. References ...... 110 CHAPTER 4. Life Cycle Assessment of switchgrass and corn stover ethanol production in the Wisconsin and Michigan agricultural contexts...... 117 PART 1. LCA of Switchgrass Cellulosic Ethanol Production: Testing LCA results for sensitivity to spatial variability ...... 117 4.1. Publication and Contribution Details ...... 117 4.2. Abstract ...... 120 4.3. Introduction ...... 122 4.4. Methods...... 127 4.4.1. Functional unit ...... 128 4.4.2. Analysis, geographic and temporal system boundaries ...... 129 4.4.3. Data sources, assumptions and sensitivity analysis ...... 131 4.4.4. Allocation ...... 137 4.5. Results and Discussion ...... 137 4.5.1. Cropping system results and spatial variation ...... 138 4.5.2. Ethanol production LCA results ...... 142 4.5.3. Sensitivity analysis...... 147 4.5.3.1. GWP ...... 148 4.5.3.2. NER...... 152 4.5.3.3. AP and EP ...... 154 4.6. Conclusions ...... 156 4.7. References ...... 161 PART 2. LCA Corn Stover Cellulosic Ethanol Production: The interplay of spatial- variability and allocation decisions ...... 170 v

4.8. Additional Objectives ...... 170 4.9. Additional Methods ...... 170 4.9.1. Functional unit ...... 173 4.9.2. Analysis, geographic and temporal system boundaries ...... 173 4.9.3. Data sources, assumptions and sensitivity analysis ...... 175 4.9.4. Allocation ...... 176 4.10. Results and Discussion ...... 177 4.10.1. Spatial variation and cropping system results ...... 177 4.10.2. Ethanol production results ...... 189 4.10.3. Sensitivity analysis...... 204 4.10.3.1. GWP ...... 209 4.10.3.2. NER...... 212 4.10.3.3. AP and EP ...... 213 4.11. Conclusions ...... 214 4.12. References ...... 219 CHAPTER 5. Conclusions and Reflections ...... 221 Works Cited ...... 228 Works Referenced ...... 240 Appendices ...... 246

I. Full cropping system variable table ...... 246 II. Full ethanol production system variable table ...... 250 III. Abbreviations and acronyms...... 256 IV. Electricity grid data ...... 257 V. Nitrogen application rates for all scenarios ...... 258 VI. Full size maps of all crop production scenarios and all results...... 259 VII. Full size graphs of all corn stover ethanol LCA results ...... 284 VIII. Life Cycle Inventory data input and source table ...... 287 IX. WI and MI Cation Exchange Capacity and soil pH maps ...... 290 X. Yates computation results for SG and CS ethanol production LCA ...... 293

IV. LIST OF FIGURES

Figure 1.1. Iterative connections between the four phases of LCA...... 4

Figure 2.1. Process flow for system producing corn for animal feed, conventional ethanol from refining corn grain, and cellulosic ethanol from refining the corn stover...... 32

Figure 2.2.a. Energy-content-based allocation...... 46

Figure 2.2.b. Market-value-based allocation...... 46

Figure 2.3.a. System expansion with energy-allocated feedstock...... 48

Figure 2.3.b. System expansion with market value-allocated feedstock...... 48

Figure 2.4. Carbon intensity as a percentage of gasoline...... 55

Figure 3.1. Sample system boundary diagram for ethanol production LCA...... 89

Figure 3.2. Ethanol production examples of allocated inputs and outputs...... 96

Figure 4.1. An example of ecological and political boundaries for two regions in Wisconsin and Michigan, after to be referred to as the study areas of this research...... 124

Figure 4.2. Process flow and system boundary diagram for feedstock production and cellulosic ethanol production...... 130

Figure 4.3. Resulting metrics by watershed from the switchgrass production portion of the LCA...... 140

Figure 4.4. NER versus GWP Intensity for all SG ethanol production scenarios...... 143

Figure 4.5. AP versus EP for all SG ethanol production scenarios...... 144

Figure 4.6. GWP Intensity with relative contributions of each GHG type and net GWP with

error introduced by spatial variability in CO2 and N2O emissions from soils and yield...... 150

Figure 4.7. Relative contributions to energy use and production in the ethanol production and NER including energy in ethanol...... 153 vii

Figure 4.8. AP with relative contributions from each stage of ethanol production and net AP with error introduced by spatial variability in biomass yield...... 155

Figure 4.9. EP with relative contributions from each stage of ethanol production and net EP with error introduced by spatial variability in N and P movement in soils...... 156

Figure 4.10. Process flow and system boundary diagram for feedstock production and cellulosic ethanol production...... 174

Figure 4.11. Corn stover yields across all scenarios...... 178

Figure 4.12. Corn stover production NER results across all scenarios...... 179

Figure 4.13. Corn stover production GWP results across all scenarios...... 180

Figure 4.14. Corn stover production AP results across all scenarios...... 181

Figure 4.15. Corn stover production EP results across all scenarios...... 182

Figure 4.16. NER versus GWP Intensity for the Wisconsin continuous corn scenarios with allocation methods...... 193

Figure 4.17. EP versus AP for the Wisconsin continuous corn scenarios with allocation methods...... 193

Figure 4.18. NER versus GWP Intensity for the Michigan continuous corn scenarios with allocation methods...... 194

Figure 4.19. AP versus EP for the Michigan continuous corn scenarios with allocation methods...... 194

Figure 4.20. NER versus GWP Intensity for the Wisconsin corn soybean scenarios with allocation methods...... 195

Figure 4.21. AP versus EP for the Wisconsin corn soybean scenarios with allocation methods...... 195 viii

Figure 4.22. NER versus GWP Intensity for the Michigan corn soybean scenarios with allocation methods...... 196

Figure 4.23. AP versus EP for the Michigan corn soybean scenarios with allocation methods...... 196

Figure 4.24. Corn stover ethanol LCA NER across all scenarios and allocation methods with error introduced by spatial variability...... 205

Figure 4.25. Corn stover ethanol LCA GWP Intensity across all scenarios and allocation methods with error introduced by spatial variability...... 205

Figure 4.26. Corn stover ethanol LCA AP across all scenarios and allocation methods with error introduced by spatial variability...... 206

Figure 4.27. Corn stover ethanol LCA EP across all scenarios and allocation methods with error introduced by spatial variability...... 206

V. LIST OF TABLES

Table 2.1. System intermediate and final products...... 33

Table 2.2. Annual GHG emission for combined corn and stover ethanol system...... 34

Table 2.3. Allocation ratios for intermediate and final products...... 40

Table 2.4. Annual DDGS co-product displaced GHG emissions...... 41

Table 2.5. Annual electricity co-product displaced GHG emissions...... 42

Table 2.6. Carbon intensity for ethanol products and % GHG emissions relative to gasoline using multiple LCA allocation approaches...... 44

Table 2.7. Annual system GHG emission for corn-ethanol refining absent corn feed and stover-ethanol production...... 52

Table 2.8. Annual system GHG emissions with and without corn-stover refining ...... 53

Table 3.1. Articles reviewed in this study and the feedstocks considered...... 73

Table 3.2. Functional units, system boundaries, allocation methods, and impact category metrics in reviewed studies...... 82

Table 3.3. Example of a table to be included with the abstract of a biofuel production LCA...... 109

Supplemental Table 3.1. Location of LCA component definitions in seven studies...... 114

Table 4.1. SG ethanol production LCA summary table...... 128

Table 4.2. SG ethanol production LCA scenario variable combination abbreviations...... 128

Table 4.3. Average, range and StDev of SG production metrics across the RIMA watersheds for SG production only...... 139

Table 4.4. Results for impact category metrics across all SG ethanol production scenarios...... 143 x

Table 4.5. Percent reduction in GHG emissions compared to gasoline for each SG ethanol production scenario and the new percent reduction value if the emissions from soils and yield modeled in each watershed increases or decreases by one StDev...... 151

Supplemental Table 4.1. Life Cycle Inventory (LCI) data inputs and sources...... 167

Table 4.6. CS ethanol production LCA summary table...... 172

Table 4.7. CS ethanol production scenario variables...... 172

Table 4.8. CS production nitrogen application rates...... 173

Table 4.9. CS feedstock production LCA variables and abbreviations...... 177

Table 4.10. Average, range and StDev of CS production metrics across the RIMA watersheds for CS production only...... 183

Table 4.11. CS ethanol production LCA variable and abbreviation table...... 189

Table 4.12. Results for impact category metrics across all CS ethanol production scenarios...... 190

Table 4.13. Percent reduction in GHG emissions for CS ethanol compared to gasoline for all CS ethanol production scenarios and the new percent reduction value if the emissions from soils and yield modeled in each watershed increases or decreases by one StDev...... 207

Supplemental Table 4.2. Additional LCI data inputs and sources for the CS ethanol LCA. .219 xi

VI. DISSERTATION ABSTRACT

The Life Cycle Assessment of Cellulosic Ethanol Production in the Wisconsin and Michigan Agricultural Contexts: The Influence of LCA Methods and Spatial Variability on Environmental Impact Assessment

By Julie C. Sinistore

Under the supervision of Professor Douglas J. Reinemann

At the University of Wisconsin-Madison

Life Cycle Assessment (LCA) is a commonly used tool for the evaluation of the environmental impacts of the production of a good or service. The application of this tool requires further methodological development to handle the production specifications of cellulosic ethanol from agricultural feedstocks. Allocation methods, transparency and the spatial variability in crop production have all been identified grand challenges in this field. The goal of this research is to examine these challenges in the literature, test their impact on biofuel LCA results, develop and test methods for solutions to these problems and use these methods to conduct an LCA on cellulosic ethanol production. This dissertation first presents a case study on the ability of cellulosic ethanol to meet the United States Renewable Fuel Standard when various allocation methods are applied. Next, a methodological review of recent ethanol production studies is presented with recommendations for standardizing and presenting LCA methods to facilitate study comparison. Finally, a spatially-explicit LCA on the production of cellulosic ethanol from switchgrass and corn stover in southern Wisconsin and Michigan under 104 different combinations of location, agricultural management practices, ethanol production technologies and allocation methods is presented. One major finding of this research is that allocation choices and spatial variation in yields and agricultural emissions can be the deciding factor as to whether xii a fuel meets a regulatory requirement. Moreover, the allocation method used can limit the influence of agricultural production and mask the effects of spatial variability on the ethanol production LCA. It is recommended that a range of values be presented instead of single environmental impact results. Another major conclusion of this research is that transparency in the presentation of LCA methods is still lacking in the literature and this impedes the direct environmental impact comparison of the many available transportation fuel choices. This research develops the LCA of biofuels field by examining LCA methodology, proposing means to facilitate the comparison of studies, demonstrating that spatial variability can be accounted for in LCA and testing the effects of such variability in concert with allocation choices for their combined influence on environmental outcomes. 1

CHAPTER 1

Introduction

1.1. INTRODUCTION

On October 31, 2011, the United Nations reported that the seven-billionth person was born on planet Earth (Roberts, 2011). The birth of this child raised the volume of the chorus of voices sounding questions on how humanity will feed, house and otherwise satisfy the needs and wants of seven billion people. What will this child eat, and where will his or her food come from? Will this person drive someday, and, if he or she does drive, what type of fuel will propel the vehicle?

These questions, and many others, do not have simple answers, but these answers will have complex consequences for the delicate balance between human and ecological needs on this one and only planet we share.

If we are to continue to live and thrive as a species, we must focus on making our current food and fuel systems sustainable. A sustainable system must be one that ensures the prolonged, if not perpetual, stability of both the Earth’s ecological and climatological systems and humanity’s social and economic systems. Agriculture is uniquely positioned at the intersection of food and fuel as well as ecology, sociology and economics. The idea of converting plant matter from agriculture into fuels has been proposed to meet many of the goals of sustainability such as climate regulation, energy independence and security, water quality and rural economic growth

(EISA, 2007 and WIDNR, 2008).

According to the Billion-Ton Study and the recently released U.S. Billion-Ton Update, there is enough biomass available in the United States (from cropland and marginal lands) to meet the goal of the Energy Independence and Security Act of 2007 (EISA 2007) revised Renewable Fuel 2

Standard (also called RFS2) that 36 billion gallons of renewable fuels be produced domestically by 2022 (Perlack et al., 2005 and USDOE, 2011). The RFS2 further specifies that 16 billion gallons per year of this fuel should be cellulosic biofuels. Cellulosic biofuel is any fuel, including ethanol, which is made from the cellulose polymer found in the cell walls of plants. This standard further specifies that a cellulosic biofuel must achieve a 60% reduction in Greenhouse

Gas (GHG) emissions through its life-cycle as compared to a 2005 average baseline level of

GHG emissions for the fossil fuel that the cellulosic fuel replaces (e.g. ethanol replaces gasoline)

(USEPA, 2010a and USEPA, 2010b). The United States Environmental Protection Agency

(USEPA) recommends the use of Life Cycle Assessment (LCA) to assess if a fuel meets this

60% GHG emissions reduction goal. Life Cycle Assessment is a tool that can be used to evaluate if the production of a good or service satisfies an environmental law, certification standard or societal goal (ISO 14040, 2006). An important distinction must be made, however, to clarify that

LCA studies do not evaluate the fuel itself, but rather the production and use of the fuel.

Preliminary assessments have been conducted by the USEPA, but they acknowledge that the fields of biofuel production and LCA are changing and that the methods for assessing fuels will

have to evolve with the state of scientific knowledge in both fields (USEPA, 2010c). Many of

these studies have concluded that biofuels, produced from a variety of feedstocks and production

pathways, can be energy efficient, greenhouse gas neutral or negative and otherwise

environmentally advantageous. Comparable evaluations of fuel production systems are needed to

guide policy-makers, scientists and the general public to select, research and support the most

environmentally sustainable fuels for production.

Cellulosic ethanol, produced from biomass such as agricultural residues and dedicated energy

crops, has the potential to meet the goals of a sustainable fuel and improve the overall

sustainability of our food and fuel systems. On the surface, it makes sense that producing fuels domestically from a renewable plant-based resource would reduce the overall GHG emissions from transportation fuel production, but no two production systems or landscapes are the same and GHG emissions are not the only measure of sustainability. We must also use tools like LCA

in a spatially-explicit way to assess the impacts of this fuel production on the air, water and land

on which we all rely.

1.2. INTRODUCTION TO LIFE CYCLE ASSESSMENT

Life Cycle Assessment aims to evaluate the production, use and disposal of a product or service,

not the product or service itself. Production, use and disposal are the three major phases of a

product’s life cycle. There are many ways in which a single product or several products, that all

serve the same functional purpose, can progress through these life cycle phases. An LCA can be

used to evaluate which life cycle pathway satisfies a law, certification process or societal goal.

Laws and certification processes for biofuels generally apply to the production phase more so

than the use or disposal phases.

As defined by the ISO document 14040, Life Cycle Assessment is the “compilation and

evaluation of the inputs, outputs, and the potential environmental impacts of a product system

throughout its life cycle” (ISO 14040, 2006). It is one of many tools that can be used to evaluate

the overall environmental impact of a product, service or project. It is a relative approach in that

the analysis must be structured around a functional unit, such as one liter of ethanol or all of the

liters of ethanol that can be produced at one biorefinery in one year. All of the relevant impact 4 category metrics must reflect quantities based on that functional unit. Life Cycle Assessment is also an iterative process because it is conducted in phases that are not finite. As each phase is completed, the previous phase may need to be revised and repeated to reflect the new level of complexity revealed in subsequent phases. Additionally, the full four-phase process may be repeated several times under various system assumptions and time scales to achieve the most comprehensive results. The four phases and their connections are illustrated in Figure 1.1. As defined by ISO standard 14040, the four phases of LCA are:

1. Goal and scope definition

2. Inventory analysis

3. Impact assessment

4. Interpretation

Figure 1.1. Iterative connections between the four phases of LCA.

(based on ISO 14040, 2006) 5

Defining the goal and scope of the analysis is the most important part of LCA. The goal of an

LCA is generally to evaluate a specific set of environmental impacts from the life cycle of a product or service. The scope definition part of this LCA phase also includes identifying what parts of the life cycle will be included in the analysis and which environmental impacts will be measured. If the LCA includes all life cycle phases, it is called a “cradle to grave” analysis. If the assessment focuses more narrowly on the production phase, it is called “cradle to gate”. The examples that will be used throughout this discussion will be the “cradle to gate” production of corn-grain and cellulosic ethanol. Energy usage, GHG emissions and other emissions to the environment will be the environmental impacts of interest in this example.

It is in this first phase that the system boundary must be set. The system boundary encompasses all relevant steps in the production of the product or service. In the case of corn-grain and cellulosic ethanol production, the system boundary often begins with the production of all relevant inputs to biomass production (e.g. nitrogen, phosphorus and potassium fertilizer, lime, seed, diesel, machinery, etc.), and carries through to the feedstock production practices (tillage, seeding and harvesting), the transportation of the biomass to the ethanol plant, the production of the necessary inputs to ethanol production (e.g. yeast, enzymes, natural gas, etc.) and the production of the ethanol fuel itself. Some studies include the transportation of the fuel to a blending facility and then to a fuel station and the combustion of that fuel in one or many types of conventional or modified engines, but these are the use and disposal phases.

Another extremely important part of the Goal and Scope Definition phase of LCA is the consideration of allocation methodology. The ISO standard defines allocation as, “partitioning the input or output flows of a process or a product system between the product system under study and one or more other product system.” The need for allocation arises from the inherent 6 multifunctionality of many industrial processes (Guinée, 2002). For example, corn-grain ethanol production produces more than one product. It produces ethanol and Distillers Grains (DG). The

DG is a valuable product which is commonly used as animal feed, but could also be used to produce biogas in an anaerobic biodigester or as a soil amendment. Since the ethanol production process produces more than one product, the environmental burdens of the production process must be apportioned between the two products. Furthermore, the need to allocate between different products can arise within an LCA of cellulosic ethanol derived from corn stover. In order to produce corn stover (the aboveground leaves, stalk and cob of the corn plant), corn grain must also be produced. Thus, the burdens and benefits of corn production must also be apportioned between these two products and the associated effects of corn stover must be accounted for in a cellulosic ethanol evaluation. There are many methods for dealing with this problem, but there is no consensus on the most appropriate method for biofuels applications

(McKone et al., 2011; Reap et al., 2008a and Reap et al., 2008b).

An inventory analysis for an LCA in this example would include collecting data on corn grain or cellulosic feedstock production, as well as ethanol production data. It is in this second phase that the analyst may discover that insufficient, out-of-date or no data are available for certain aspects of the production of the product. If this occurs, the analyst may choose to return to the first phase and redefine the system boundary to exclude or otherwise account the processes for which little or no reliable data are available. Alternatively, the analyst may choose to estimate more current data based on available sources and test the sensitivity of the analysis to this estimation in the final phase of the LCA (Baumann and Tillman, 2004).

The impact assessment phase of the analysis aims to enumerate the environmental impacts of the entire process. For this step computer models and spreadsheets are used to aggregate the data 7

collected in the inventory analysis and group different it into different impact categories. For

example, soil tillage method affects both the energy usage and GHG emissions associated with

feedstock production for ethanol, but in different ways. The act of tilling the soil with machinery

uses fossil fuels. The use of fossil fuel must be added to the fossil energy use total and the release

of GHGs must be added to the GHG emissions total. Additionally, the act of tilling the soil can

contribute to GHG emissions from the soil itself and these emissions must also be calculated

with dynamic equations and added to the GHG emissions total in the model. Accounting

decisions such as these must be completely transparent to ensure that calculations reflect actual

conditions and to prevent under- or over-counting.

The final phase, interpretation, includes evaluation, sensitivity analysis, consistency checks and

other methods of validation. Assumptions made in the previous phases are tested for their

relative influence on the final results of the study. It is in this stage that the analyst may find that

underlying assumptions about the system boundary, data sources, calculation methods or

categorizations of impacts must be changed in order to reflect the sensitivity of the entire system

to those decisions. It is also in the interpretation phase that the analyst comes to conclusions

about the environmental impacts of the product or service and makes recommendations for

improvements in the process.

1.3. INTRODUCTION TO CELLULOSIC ETHANOL PRODUCTION

Ethanol is an alcohol produced by fermentation of sugars. These sugars can be derived from any number of sources such as sugar cane, corn grain and cellulose. If ethanol is produced from sugars derived from cellulose, it is generally called cellulosic or lignocellulosic ethanol.

Cellulose is a structural material found in all plants and consists of a linear polymer of glucose 8

which is linked together with a beta-glucosidic linkage. This linkage allows the cellulose

polysaccharide to form long linear chains that stack with little space in between each cellulose

strand. This is different from the starch polysaccharide (also a polymer of glucose) in that the

starch has an alpha-glucosidic linkage. Starch is an energy storage material in plants. The alpha-

glucosidic linkage makes starch non-linear or branched. This linear versus non-linear structural

difference is important because the non-linear nature of starch allows glucose (sugar) to be freed

from the structure much more easily than the linear structure of cellulose. The ease with which

sugar can be liberated from starch is one reason why starch to ethanol technology (e.g. corn grain

ethanol) has developed to commercialization much more rapidly than cellulose to ethanol

technology.

The term lignocellulosic ethanol refers to the fact that cellulose does not exist alone as a

structural material in plants. Cellulose, hemicelluloses and lignin form the complex structural

matrix of plant cell walls which provide strength and resistance to bacterial, fungal or insect

invasion to the plant. Hemicelluloses are a group matrix polysaccharides made up of different

five- and six-carbon sugar (pentoses and hexoses). Hemicellulose forms a complex network of

crossing and linked fibers that bind cellulose and lignin together. Lignin is a polymer of phenylpropane units. There are many different types of these units (hydroxphenyl, guaiacyl, syringly and many more), but the main feature of lignin in plants is the abundance and strength of carbon-carbon bonds, especially with double bonds between carbons. These three components together form the incredibly strong and nearly impenetrable fortress of plant cell walls that allow plants, like giant sequoia and redwood trees (Sequoiadendron giganteum and Sequoia

sempervirens), to grow to be some of the largest, tallest and oldest living things on Earth. 9

Ethanol can be produced from numerous feedstocks via various pathways. Each feedstock has a

different amount of available sugars for ethanol production. The efficiency and production levels

of a cellulosic ethanol plant hinge on the feedstock chosen. The feedstock for a cellulosic ethanol

plant is generally referred to as biomass, but there are dozens of potential biomass feedstocks.

Commonly studied feedstocks include corn stover (Zea mays), switchgrass (Panicum virgatum),

mixed prairie grasses and Miscanthus, but various other studies have also considered cassava

(Manihot esculenta), woody biomass, rice straw (Oryza spp.), sorghum (Sorghum bicolor), citrus

peels, wheat straw (Triticum spp.) and many others (Worldwatch Institute, 2007). Moreover,

cellulosic ethanol production hinges on the type of pretreatment method used to release the

valuable sugars from the complex, strong and recalcitrant lignocellulosic matrix.

The most common pathway of from biomass to cellulosic ethanol is the sugar fermentation

pathway. This begins with a stringent form of pretreatment which exposes the cellulose in the

lignocellulosic plant wall matrix. Pretreatment uses heat, acids, bases, pressure, mechanical

methods or a combination of part or all of the above to break apart this matrix. Common methods

include steam explosion, dilute acid, and Ammonia Fiber Expansion (AFEX), though there are

many other pretreatment methods. Each pretreatment method produces variable quantities of

hydrolysable cellulose and hemicelluloses, but each method can also produce unwanted

byproducts like fermentation inhibitors, recalcitrant lignin and condensed lignin particles

(Brown, 2003).

The next step in cellulosic ethanol production is enzymatic hydrolysis. The goal of this step is to break down cellulose and hemicelluloses into individual sugar monomers (hexoses or pentoses) by using enzymes. The recalcitrant lignin potentially formed during pretreatment, however, can bind to and block the active sites of the expensive enzymes which render them useless. 10

Furthermore, condensed lignin particles which form on the cellulose fibers can block an

enzyme’s access to specific locations on the cellulose fiber. This is a problem since different

enzymes act to break up cellulose in very specific locations (e.g. the oxidizing or reducing end of

the molecule). Hydrolysis is conducted with various different cocktails of enzymes that also

produce variable amounts of five- and six- carbon sugars based on the type of feedstock,

pretreatment method and specific enzyme cocktail (Brown, 2003).

The final ethanol production step is fermentation. Fermentation uses yeast or bacteria to produce

ethanol from sugar. The ethanol is a by-product of the metabolism of these micro-organisms.

There are many different organisms that can be used to produce ethanol, but the most commonly used in grain ethanol production is common baker’s yeast (Saccharomyces cerevisea). Among the other organisms that could be used for fermentation are Zymomonas mobilis, Escherichia coli and genetically modified baker’s yeast. The limitation of regular baker’s yeast is that it can only ferment six-carbon sugars, but the pretreatment and hydrolysis of biomass also produces five-

carbon sugars from hemicelluloses. This represents a significant loss in the full potential of

producing ethanol from lignocellulosic biomass. Furthermore, many ethanol-producing organisms are inhibited from producing ethanol by degradation products potentially produced during pretreatment. The inhibitors include hydroxymethylfurfural (HMF), levulinic acid, formic

acid and furfural. These compounds can slow down or completely stop the fermentation process

and are expensive to remove from the hydrolysate before fermentation. Thus, the feedstock and

pretreatment method can further alter the amount of ethanol produced as a final product (Brown,

2003).

The final steps performed at the ethanol refinery prepare the ethanol for commercial delivery and

they are distillation, dehydration and denaturation. Distillation uses a series of heated columns to 11

separate most of the water from the ethanol and collect any remaining solids from fermentation.

This results in a water-ethanol azeotrope of approximately 95% ethanol and 5% water by

volume. This mixture is commonly referred to as hydrous ethanol and it cannot be mixed with

gasoline for use in a conventional spark-ignition engine in the United States. Therefore, the water and ethanol must be separated. This is generally done with a molecular sieve that allows ethanol

to pass through, but not water. The resulting anhydrous ethanol is a suitable vehicle fuel, but it

must be denatured (commonly with gasoline) to prevent human consumption of this pure

alcohol. These steps do not vary much between ethanol plants.

There is a potential in cellulosic ethanol to combine many, if not all, of the ethanol production

steps into one process. The most commonly assumed pathway outline above is called Separated

Hydrolysis and Fermentation (SHF) because the hydrolysis and fermentation occur in separate

reaction tanks at different times. The advantage of this production method is that the hydrolysis

and fermentation can take place at their most optimal temperature and pH ranges. This method,

however, takes more time and energy to produce ethanol and also requires a higher enzyme

loading. One alternative to this SHF method is Simultaneous Saccharification and Fermentation

(SSF). In this production method, both the hydrolysis and fermentation occur together in the

same tank and at the same time. This method has the potential of producing higher ethanol yields

than SHF while requiring less time, lower enzyme loading (lower cost) and reducing inhibition.

The disadvantage of SSF, however, is that neither fermentation nor hydrolysis will occur at their

optimum temperature or pH. There are several other methods under development for processing

biomass into ethanol and other liquid fuels and some do not even involve a biological catalyst,

but the focus of this research is the conventional SHF conversion of cellulose to ethanol. 12

1.4. CHALLENGES IN CELLULOSIC ETHANOL PRODUCTION LCA

With all of these manifold pathways from biomass to ethanol and with all of the variables in

energy use, time and ethanol yield, it can be incredibly difficult to evaluate the overall

environmental burdens and benefits of cellulosic ethanol production. This difficulty is

compounded by the fact that there are no commercial scale cellulosic ethanol plants in full

production in the world today. There are some pilot and demonstration scale cellulosic ethanol

plants, but getting operational data from these highly proprietary sources is difficult and scaling

these data to commercial scale would be problematic. This means that the process as it could

actually be conducted cannot be studied directly. Fortunately, there are some studies that put

forth both technically and economically feasible production pathways from feedstock to fuel

(Aden et al., 2002 and Humbird et al., 2011).

The LCA of cellulosic ethanol production, however, is further complicated by the flexible nature

of LCA methodology. Life Cycle Assessment methodological choices have been shown to

greatly influence the results of an ethanol production evaluation such that the same fuel could

exceed, meet or fail to meet a policy benchmark depending on analysis choices (Curran, 2007;

Farrell et al., 2006 and Kaufman et al., 2010). There is no one pre-defined and agreed upon set of

LCA methods for ethanol production evaluation. Life cycle analysts can choose from a variety of potential functional units, system boundaries, allocation methods and impact category metrics.

This flexibility is advantageous because it allows the methodology to be shaped to answer specific LCA questions, but this flexibility can also result in many analyses of the same product which all reach very different conclusions. Moreover, there exists an abundance of information and data sources for necessary aspects of the LCA of cellulosic ethanol production, and also a dearth of information and data sources in other aspects. For example, spatially-explicit and 13 small-scale feedstock production data is not readily available, but has been identified potentially necessary for accurate LCAs (McKone et al, 2011 and Reap et al., 2008b). Thus, there is a need for some consensus on cellulosic ethanol production methods that could be achieved at the commercial scale in the near-term and the most accurate, representative and comparable methods and data sources for evaluating cellulosic ethanol produced by these methods.

In recent years, there have been a number of ethanol production studies evaluating impacts of fuel production and the impacts of methodological choices on the results of such analyses (Beer and Grant 2007; Curran 2007; Farrell et al., 2006; Fu et al., 2003; Hill et al., 2009; Hill et al.,

2006; Kaufman et al., 2010; Laser et al., 2009; Liska et al., 2009; Luo et al., 2009; Sinistore and

Bland, 2010; Spatari et al., 2010 and Spatari et al. 2005). The most commonly evaluated impact categories are net energy balance and GHG emissions. These metrics alone, however, are insufficient indicators of a fuel’s environmental impact. Other impact categories, such as air pollution emissions, acidification, eutrophication and many more, must also be evaluated before a final judgment on the sustainability of a fuel can be made (Hill et al., 2009; Luo et al., 2009 and Curran, 2007).

On the whole, the results of an LCA on cellulosic ethanol production hinge on the feedstock and production path chosen to produce ethanol as well as the data sources used for conversion efficiency, energy usage and feedstock production, the LCA methods employed and the impact category metrics evaluated. It is the goal of this dissertation to evaluate not only a broad range of environmental impacts of cellulosic ethanol production, but also to assess and build upon the methods used to evaluate cellulosic ethanol production. 14

1.5. DISSERTATION OBJECTIVES AND FORMAT

The goal of this dissertation was to conduct a comprehensive Life Cycle Assessment of

cellulosic ethanol produced in two very specific agricultural contexts (southern Wisconsin and

Michigan) that builds upon the current state of scientific knowledge by addressing data-gaps and

methodological challenges identified in literature.

This goal can be broken down into three objectives:

Objective 1: To contribute to the development of the science and methods of Life Cycle

Assessment as it relates to the study of agricultural products such as biofuels.

Objective 2: To determine the life-cycle environmental impacts of biomass production in the

Wisconsin and Michigan Regionally Intensive Modeling Areas (RIMAs) by using spatially-

explicit cropping system data.

Objective 3: To determine the life-cycle environmental impacts of cellulosic ethanol production

from corn stover and switchgrass with dilute acid or AFEX pretreatment in Wisconsin and

Michigan and test the sensitivity of the results to allocation methods and spatial variation.

The first objective of this study was designed to develop and enrich my understanding of the

current state of biofuel Life Cycle Assessments and the sustainability questions answered and

unanswered by literature on biofuels and biofuel LCAs. Completing this objective before

tackling the second two objectives allowed me to identify new and different ways of approaching

this type of research and prepared me to anticipate data needs and avoid methodological pitfalls.

The second objective of this study was designed to integrate spatially-explicit cropping-system data and cradle-to-gate Life Cycle Inventory (LCI) agricultural input production data into one 15

LCA evaluation of the GHG emissions, water quality (acidification and eutrophication) and

energy use impacts from the production of switchgrass and corn stover for cellulosic ethanol.

Accomplishing this objective helped me differentiate my work from previous studies in three

ways: 1) by using dynamically modeled crop production data that varies over space and time at a

small spatial scale instead of static and large-scale agricultural census data; 2) by including the full life-cycle impacts of the production of all agricultural and industrial inputs; 3) by including the assessment of the eutrophication and acidification potential of switchgrass and corn stover production. This was an essential step towards completing a full cellulosic ethanol production

LCA.

The final objective of this study was to integrate the first two objectives and add the final ethanol production piece of the puzzle to produce one comprehensive cellulosic ethanol LCA. This research integrated data from many different sources and to remedy some of the methodological pitfalls identified in the first objective (such as the static spatial treatment of agricultural inputs to ethanol production and nationally-generalized electricity production data). In addition to the spatially-explicit treatment of the cropping system, this work differed from previous studies by reflecting the current state of technology such as the most up-to-date information available on the

Ammonia Fiber Expansion (AFEX) cellulose pretreatment method. Furthermore, this LCA included region-specific energy production data. The culmination of this research is a comprehensive and rigorous examination of the life-cycle environmental impacts of cellulosic ethanol production.

This dissertation follows the format of an introductory chapter, three body chapters consisting of published or publishable journal articles and a conclusion. The first body chapter is an in-depth analysis of the ability of a fuel to meet policy standards based on the allocation method applied 16

in the Life Cycle Assessment. The objective of this study was to investigate and assess different allocation methods to illuminate their effect on policy. This article was published in the journal

Energy Policy in 2010 and served as a valuable case study to help me identify the most important methodological choices I would need to make in my LCA research. The second body chapter is a methodological review of seven recently-published ethanol production studies. It examines the

similarities and differences in methods used to evaluate the same fuel (ethanol) through the lens

of LCA. The objective of this study was to bring together all of the various functional units,

system boundaries, allocation methods and impact category metrics used in recently published

studies, identify barriers to the comparison of results based on these methods and suggest ways to facilitate comparison of studies. This paper was submitted to the International Journal of Life

Cycle Assessment and accepted with revisions in January of 2012. The research conducted for the first two studies significantly informed the methodology employed in the final LCA produced as a part of this dissertation research. It was this research that identified openings for

research contributions in the field ethanol production LCA and shed light on potential analysis

pitfalls. The final body chapter of this dissertation is a full LCA of cellulosic ethanol production

in southern Wisconsin and southern Michigan from corn stover and switchgrass (under several

agricultural production treatments) via either dilute acid or AFEX pretreatment. This article will

be submitted to the journal Biofuels, Bioproducts and Biorefining. The conclusion of this

dissertation will summarize major research findings, conclusions and recommendations to both

improve cellulosic ethanol production and the methods used to evaluate it. 17

1.6. REFERENCES

Aden, A., Ruth, M., Ibsen, K., Jechura, J., Neeves, K., Sheehan, J., and Wallace, B. 2002. Lignocellulosic biomass to ethanol process design and economics utilizing co-current dilute acid prehydrolysis and enzymatic hydrolysis for corn stover (NREL/TP-510- 32438); National Renewable Energy Laboratory: Golden CO, 2002. Available at http://www.nrel.gov/docs/fy02osti/32438.pdf. Baumann, H. and Tillman, A. 2004. The Hitch Hiker’s Guide to LCA: An orientation in life cycle assessment methodology and application. Lund, Sweden: Studentliterature. Brown, R. C. 2003. Biorenewable Resources: Engineering new products from agriculture: Engineering new products from agriculture. Ames, I.A.: Blackwell Publishing. Curran, M. 2007. Studying the Effect on System Preference by Varying Coproduct Allocation in Creating Life-Cycle Inventory. Environmental Science & Technology. 41(20): 7145- 7151. Energy Independence and Security Act of 2007. Pub. L. no. 110-140. 110th Congress. 1st session. 2007. http://www.gpo.gov/fdsys/pkg/BILLS-110hr6enr/pdf/BILLS-110hr6enr.pdf. Accessed: November 2011 Farrell, A. E., Plevin, R. J., Turner, B. T., Jones, A. D., O’Hare, M. and Kammen, D. M. 2006. Ethanol can contribute to energy and environmental goals. Science. 311: 506-508. Fu, G. Z., Chan A.W., and Minns, D.E.. 2003. Life Cycle Assessment of Bio-ethanol Derived from Cellulose. International Journal of Life Cycle Assessment. 8(3): 137-141. Guinée, J. B. 2002. Handbook on Life Cycle Assessment. Dordrecht, The Netherlands: Kluwer Academic Publishers. Hill, J., Nelson, E., Tilman, D., Polasky, S. and Tiffany, D. 2006. Environmental, economic, and energetic costs and benefits of biodiesel and ethanol biofuels. Proceedings of the National Academy of Sciences, 103: 11206-11210. Hill, J., Polasky, S., Nelson, E., Tilman, D., Huo, H., Ludwig, L., Neumann, J., Zheng, H. and Bonta, D. 2009. Climate change and health costs of air emissions from biofuels and gasoline. Proceedings of the National Academy of Sciences. 106(6): 2077-2082. 18

Humbird, D., Davis, R., Tao, L., Kinchin, C., Hsu, D., Aden, A., Schoen, P., Lukas, J., Olthof, B., Worley, M., Sexton, D. and Dudgeon, D. 2011. Process Design and Economics for Biochemical Conversion of Lignocellulosic Biomass to Ethanol: Dilute-Acid Pretreatment and Enzymatic Hydrolysis of Corn Stover. National Renewable Energy Laboratory. Technical Report: NREL/TP-5100-47764. Golden, CO. International Organization for Standardization (ISO). 2006a. ISO 14040: Environmental management – Life cycle assessment – Principles and framework. Geneva, Switzerland. International Organization for Standardization (ISO). 2006b. ISO 14044: Environmental Management – Life cycle assessment, Life cycle impact assessment. Geneva, Switzerland. Kaufman, A. S., Meier, P. J., Sinistore, J. C., and Reinemann, D. J. 2010. Applying life-cycle assessment to low carbon fuel standards – How allocation choices influence carbon intensity for renewable transportation fuels. Energy Policy. 38: 5229-5241. Laser, M., Haimin, J., Jayawardhana, K. and Lynd, L. R. 2009. Coproduction of ethanol and power from switchgrass. Biofuels, Bioproducts and Biorefining. 3: 195-218. Liska, A. J., Yang, H. S., Bremer, V. R., Klopfenstein, T. J., Walters, D. T., Erickson, G. E. and Cassman, K. G. 2009. Improvements in Life Cycle Energy Efficiency and Greenhouse Gas Emissions of Corn-Ethanol. Journal of Industrial Ecology. 13(1): 58-74.

Luo, L., van der Voet, E., Gjalt, H. and Udo de Haes, H. A. 2003. Allocation issues in LCA methodology: a case study of corn stover-based fuel ethanol. International Journal of Life Cycle Assessment. 14: 529-539.

Perlack, R., Wright, L., Turhollow, A. Graham, R., Stokes, B., and Erbach, D. 2005. Biomass as a feedstock for a bioenergy and bioproducts industry: The technical feasibility of a billion-ton annual supply. (Tech. Rep ORNL/TM - 2006/66). Oak Ridge, TN: Oak Ridge National Laboratory. 19

Reap, J., Roman, F., Duncan, S. and Bras, B. 2008a. A survey of unresolved problems in life cycle assessment; Part 1: goal and scope and inventory analysis. International Journal of Life Cycle Assessment. 13: 290-300. Reap, J., Roman, F., Duncan, S. and Bras, B. 2008b. A survey of unresolved problems in life cycle assessment; Part 2: impact assessment and interpretation. International Journal of Life Cycle Assessment. 13: 374-388. Roberts, S. 2011. “U.N. Says 7 Billion Now Share the World”. The New York Times. Published Online: 31 Oct 2011. http://www.nytimes.com/2011/11/01/world/united-nations-reports- 7-billion-humans-but-others-dont-count-on-it.html. Accessed: 13 Dec 2011. Sinistore, J. C. and Bland, W. L. 2010. Life-Cycle Analysis of Corn Ethanol Production in the Wisconsin Context. Biological Engineering. 2(3): 147-163. Spatari S., Zhang Y. and MacLean, H. L. 2005. Life Cycle Assessment of Switchgrass- and Corn Stover-Derived Ethanol-Fueled Automobiles. Environmental Science & Technology. 39(24): 9750-9758. Spatari, S., Batley, D. M., and MacLean, H. L. 2010. Life cycle evaluation of emerging lignocellulosic ethanol conversion technologies. Bioresource Technology. 101: 654-667.

U.S. Environmental Protection Agency (EPA). 2010a. Regulation of Fuels and Fuel Additives: Changes to the Renewable Fuel Standard Program; Final Rule. USEPA. Federal Register. 75(58): 14669-15320.

U.S. Environmental Protection Agency (EPA). 2010b. EPA Finalizes Regulations for the National Renewable Fuel Standard Program for 2010 and Beyond. Office of Transportation and Air Quality. USEPA. http://www.epa.gov/otaq/renewablefuels/420f10007.pdf. Accessed: 2 Jan 2012.

U.S. Environmental Protection Agency (EPA). 2010c. EPA Lifecycle Analysis of Greenhouse Gas Emissions from Renewable Fuels. Office of Transportation and Air Quality. USEPA. http://www.epa.gov/otaq/renewablefuels/420f10006.pdf. Accessed: 2 Jan 2012. 20

Wisconsin Department of Natural Resources (WIDNR). 2008. Governor’s Task Force on Global Warming Final Report.Wisconsin’s Strategy for Reducing Global Warming.. July 2008. Available online at: http://www.climatestrategies.us/library/library/view/313. Accessed: 13 Dec 2011.

Worldwatch Institute. 2007. Biofuels for Transport: global potential and implications for sustainable energy and agriculture. Sterling, V.A.: Earthscan. 21

CHAPTER 2

Applying life-cycle assessment to low carbon fuel standards – How allocation choices

influence carbon intensity for renewable transportation fuels

2.1. PUBLICATION AND CONTRIBUTION DETAILS

In 2010, the paper “Applying life-cycle assessment to low carbon fuel standards – How

allocation choices influence carbon intensity for renewable transportation fuels” was published

in the journal Energy Policy. I co-authored this paper with Andrew Kaufman, Paul Meier and

Doug Reinemann. My contribution to this research included 50% of the methodology which

included data sourcing, explanation of terms, calculation method development and data

validation. I contributed to approximately 20% of the paper’s final text and 50% of the revisions

completed for the response to the reviewers’ comments and concerns which preceded

publication. The conclusions of this paper prompted me to dive further into the depths of LCA

methodology.

Authors: Andrew S Kaufman1, Paul J Meier1,2, Julie C Sinistore1,3, Douglas J Reinemann1,3

1) Great Lakes Bioenergy Research Center, University of Wisconsin-Madison

2) Energy Institute, University of Wisconsin-Madison

3) Biological Systems Engineering, University of Wisconsin-Madison,

2.2. ABSTRACT

The Energy Independence and Security Act of 2007 (EISA) requires life-cycle assessment

(LCA) for quantifying greenhouse gas emissions (GHG) from expanded U.S. biofuel production.

To qualify under the Renewable Fuel Standard, cellulosic ethanol and new corn ethanol must demonstrate 60% and 20% lower emissions than petroleum fuels, respectively. A combined corn-grain and corn stover ethanol system could potentially satisfy a major portion of renewable fuel production goals. This work examines multiple LCA allocation procedures for a hypothetical system producing ethanol from both corn grain and corn stover. Allocation choice is known to strongly influence GHG emission results for corn-ethanol. Stover-derived ethanol production further complicates allocation practices, because additional products result from the

same corn production system. This study measures the carbon intensity of ethanol fuels against

EISA limits using multiple allocation approaches. Allocation decisions are shown to be paramount. Under varying approaches, carbon intensity for corn ethanol was 36 to 79% that of gasoline, while carbon intensity for stover-derived ethanol was -10 to 44% that of gasoline.

Producing corn stover ethanol dramatically reduced carbon intensity for corn-grain ethanol, because substantially more ethanol is produced with only minor increases in emissions.

Regulatory considerations for applying LCA are discussed.

Keywords: LCA, Ethanol, Bioenergy

2.3. INTRODUCTION AND BACKGROUND

The Energy Independence and Security Act (EISA) of 2007 sets a Renewable Fuel Standard of

36 billion gallons of biofuel production by the year 2022, of which 21 billion gallons must be from a cellulosic or advanced biofuel process. Also according to EISA, the renewable fuels are subject to greenhouse gas (GHG) limits, requiring LCA to demonstrate that GHG emissions from cellulosic ethanol and new corn ethanol are 60% and 20% lower, respectively, than petroleum fuels (Sissine, 2007). This study performed an LCA for an integrated corn-grain ethanol and stover-ethanol system, in order to test whether resulting GHG emissions satisfy EISA’s proposed

GHG limits. Using four common LCA allocation methods, we illustrate how reasonable LCA allocation methods divide the GHG emissions among corn ethanol and corn stover ethanol, as well as their associated co-products. Specifically, we report GHG emission per ethanol energy output (carbon intensity) for corn ethanol and stover ethanol under each allocation procedure and compare against carbon intensity for gasoline.

Corn stover has garnered considerable interest as a promising feedstock for cellulosic ethanol.

Recent work by James et al. (2010) compared seven cellulosic crops against a corn and stover system. Given this study’s assumptions for price and quality, Miscanthus returns break even with the corn and stover system at $45/ton for biomass, assuming the cost for Miscanthus rhizomes drops significantly. No other cellulosic crop returns break even unless biomass exceeds $110/ton.

Largely for this reason, most other cellulosic crops are not widely grown, while there is a ready supply of corn stover.

For corn ethanol LCA, it is well acknowledged that allocation practices strongly influence

results, by determining how emissions are assigned among the primary ethanol product and other

co-products. Production of stover-derived ethanol further complicates allocation practices,

because additional fuel and co-products result from the same corn production system. For

practical purposes, both stover-ethanol and corn-ethanol production pathways require integrated

LCA. Because both grain and cellulose feedstocks emanate from the same plant, accurate LCA for one ethanol product requires an understanding of the other. For example, in order to allocate emissions to the corn grain, the LCA practitioner must be aware of the quantity of co-produced

stover. Without an integrated LCA, inconsistent system boundaries or input assumptions would

most likely result in either over-counting or under-counting emissions.

For a process that results in two or more products, an LCA must define how environmental

impacts will be assigned to products. Two accounting procedures are most commonly applied to

biofuel LCA studies. The first is an allocation procedure, which divides inputs and impacts

among products based on a ratio of physical (e.g. mass) or other (e.g. market value)

characteristics. The second approach is termed system expansion and co-product displacement

(hereafter referred to as system expansion). For system expansion, one product is typically

defined as the primary product, and the other products are defined as co-products. The co-

products are assumed to replace other market products whose production is outside the original

system boundary. For example, dry-mill corn ethanol production typically produces a distillers

grain co-product which can replace animal feeds such as corn grain or soybean meal. In system

expansion, the LCA boundary is expanded to include the production of the displaced market product, and the primary product receives credit for the avoided environmental impact. In the corn ethanol refinery example, the system boundary is expanded to include soybean meal production. Allocation is then avoided by assigning all emissions to the primary corn-ethanol product, less the avoided emissions from displacing soybean meal with dry distillers grain with solubles (DDGS).

The International Standards Organization (ISO) states that allocation should preferably be avoided by expanding system boundaries to account for the impact of displaced products. If system expansion is not possible, then the ISO recommends allocation should be based on physical parameters, such as mass or energy content. (ISO, 2006a and 2006b). Associated nomenclature is inconsistent throughout the literature. A particular source of potential confusion is whether “system expansion” should be referred to as a type of allocation procedure. Our interpretation of ISO guidance and selected literature is that allocation is the partitioning of environmental impacts across products based on physical or economic ratios, and that system expansion is actually a means to avoid allocation. Given its prevalent use in the literature, however, we use the term “allocation” to collectively refer to system expansion as well as allocation by partitioning.

The ISO preference for system expansion is not uniformly adopted by practitioners. Arguments against system expansion contend that data used to expand the system in a LCA are of questionable origin, thus significantly alter results (Bjorklund, 2002; Ekvall and Finnveden, 2001

and Farrell et al., 2006). Work by Ekvall and Finnveden (2001) includes sensitivity studies and suggests that allocation by physical properties is satisfactory and has only a small influence over results. Some studies rely on more than one approach to allocation. Work by Hill et al. (2006) uses an economic method for the primary products of the modeled system as well as an energy– based method to account for co-products that do not fit into their economic method. Guinée et al.

(2009) apply a substitution method to remove the additional functions of co-products for a hypothetical wood furniture and electricity production system. Several other studies discuss LCA allocation for bioenergy and have been recently summarized by Mendoza et al. (2008).

While several previous bioenergy LCA studies discuss the significance of allocation methods, very few studies report results using multiple allocation choices (Mendoza et al., 2008).

Mendoza et al. (2008) test multiple allocation approaches for bio-electricity production chains, concluding that allocation on a physical basis is preferable when the priority is consistency, but that economic-based allocation or co-product substitution (system expansion) is more valid, though vulnerable to data inadequacies. Curran (2007) compared five allocation approaches across different fuel production systems, concluding that the relative impacts for gasoline and ethanol-blended fuels are the same (e.g. ethanol-blended gasoline produces reduced GHG emissions compared to regular gas when all of the same allocation methods are applied to both systems). Beer et al. (2007) considers economic and system expansion allocation procedures for a molasses-ethanol system. Kodera’s (2007) review does not test bio-ethanol LCA allocation procedures, but suggests that system expansion be used for consistency with ISO standards, with

sensitivity analysis to measure the impact of the allocation procedures. Malca and Freire (2006) assert that without proper consideration of allocation or system expansion methods, the LCA results may not be indicative of what is actually occurring in the system. Tilman (2000) asserts that the overall goal of a LCA study can influence the results by initially influencing the choice of allocation method used.

Sound allocation decisions require appreciating the differences between attributional and consequential LCA approaches. An attributional LCA (also called descriptive or retrospective) looks at a system with static output and calculates the environmental burdens considering that specific level of production. A consequential LCA (also called change-oriented or prospective

LCA) evaluates the incremental changes to environmental burdens (e.g. net change in GHG emissions) in response to a decision or change in product output (e.g. additional ethanol production). The distinction between consequential and attributional approaches to LCA has received increasing attention, as cited by Mendoza et al. (2008), including Curran et al., (2005),

Ekvall and Andrae, (2006), Ekvall and Tillman, (1997), Thomassen et al., (2008), Weidema,

(2001). In a study of corn and stover production, Kim et al. (2009) use a consequential approach and assign stover only the incremental impacts from additional fuel and nutrient requirements, as well as altered soil carbon and nitrogen dynamics.

The allocation procedures compared in this report demonstrate both attributional and consequential approaches. The allocation examples that partition emissions based on physical or economic ratios employ an attributional approach. System expansion is necessarily

consequential, because the environmental impact is compared with and without co-product

substitution. Here we examine a hybrid approach, where consequential system expansion is

applied to co-product production with emission impacts credited to an otherwise attributional

LCA. Finally, we examine a more truly consequential LCA, where we measure the change in the

system’s GHG emissions based on the addition of the corn stover ethanol production. For LCA

applied to EISA, the distinction between attributional and consequential approaches is potentially

confusing. Different approaches may be appropriate depending on whether the LCA goal is to

evaluate the policy’s future impact (e.g. net benefit analysis) or whether the goal is to measure

carbon intensity of renewable fuels. While the consequential approach is likely preferable for

prospectively assessing the policy’s future impact, we contend that attributional approaches may

be preferable for measuring carbon intensity, particularly if corroboration with historic emission

inventories is important.

2.4. METHODS

This study evaluates the effects of LCA allocation practices on GHG emission results for ethanol

produced from corn grain and corn stover. The methods section below provides representative

LCA values for a hypothetical corn-ethanol and stover-ethanol system. The results section

subsequently compares carbon intensity values under various allocation practices relative to the

benchmarks established by EISA.

In order to clearly present results for the most common allocation practices, important

simplifications were made to the LCA. First, sensitivity analysis was not performed. LCA studies

often measure variability around changes in key parameters. Varying both parameters and

allocation procedures, however, would result in an array of results with great potential for

confusion. We believe that input parameter variability would not significantly alter the relative

comparison of allocation approaches. Secondly, co-product allocation was simplified by assuming that the corn and stover refineries produce only one co-product, and that each co-

product replaces only one other market product. In real-world systems refineries would produce

many co-products, each with the potential to replace multiple market products (Kim and Dale

2002). Finally, potential indirect land use change (ILUC) impacts were not considered. EISA

requires consideration of secondary agricultural sector impacts including land use change. By

defining our system to include only existing corn-producing land, we avoid the need to consider

direct conversion of land to agriculture. The appropriate inclusion of ILUC is unclear for the

studied system, which holds corn production constant. The methodology to quantify ILUC for

existing corn-ethanol systems is the subject of considerable debate. For stover-derived ethanol,

ILUC impacts are even less clear. The California Air Resources Board (2009) stated in their low

carbon fuel standard that “No currently available model is capable of estimating the land-use-

change effects of plant-based feedstocks that do not displace agricultural commodities”. LCA

protocol for incorporating ILUC is clearly important and warrants separate and significant

consideration.

2.4.1. System description and boundaries

This study considers a system producing corn for animal feed, conventional ethanol from

refining corn grain, and cellulosic ethanol from refining the corn stover. Corn-grain and corn

stover feedstock procurement was based on a conservative estimate of the existing corn

producing land area in four Wisconsin counties: Columbia, Dane, Iowa, and Sauk. Data from the

U.S. Dept. of Agriculture’s (2008) National Agricultural Statistics Service’s database were used

to determine how much land area was devoted to corn production in 2005. It was assumed that

one-third of the available corn grain was used for ethanol production, with the remainder

dedicated to animal feed. The EPIC/APEX model was used to provide a biogeochemical

simulation of feedstock production (Izaurralde et al., 2006 and Williams et al., 2008). The

modeling targeted yields of 8,490 kg/ha, resulting in 362 million kg of corn grain for ethanol

feedstock, and 734 mil. kg corn grain for animal feed. Stover was assumed to be uniformly

procured from all corn grown, both for feed and ethanol. It was assumed that the dry mass ratio

of stover to grain of the harvested plant was 0.8 (Pordesimo et al., 2004). The EPIC simulation

assumed 56% of stover was collected from the 4-county area, resulting in 491 mil kg feedstock.

The performance of the corn stover biorefinery was based on works by Aden et al. (2002) and

Sheehan et al. (2004), for which the reported inputs and outputs were scaled relative to the

ethanol production assumed in this study. Based on a cellulosic refinery yield of 0.34 L/kg

(Sheehan et al., 2004), the assumed stover production would support 193-million liters per year of cellulosic ethanol. The process described by Aden et al. (2002) and Sheehan et al. (2004)

produces high-pressure steam for electricity production and process heat by combusting solids

from distillation, along with concentrated syrup from the evaporator, and biogas from anaerobic

digestion. The combined heat and power system ultimately converts7% of the stover energy

content to electricity, of which roughly one-third is required for the refinery’s auxiliary load. The

remaining 4.5% of the stover’s energy content is considered “excess electricity”, potentially

available for export to the power grid. Recent reporting by Luo et al.(2009) identified that

electricity associated with enzyme production was not internalized within the Aden and Sheehan

studies. Luo et al. (2009) assume that all excess electricity is required for enzyme production, but

state that assuming no electricity is exported to the grid may be unrealistic. We assume that

electricity may indeed be exported to the grid and evaluate three sensitivity cases representing various levels of electricity co-product. Our reference case assumes that 3% of the stover energy content is ultimately converted to excess electricity and exported to the regional grid. This assumes that future enzyme requirements diminish as a result of improved pre-treatment processes and advances in cellulose hydrolysis enzymes and ethanol-fermenting organisms.

Cases representing 1.5% and 0% of stover converted to excess electricity are evaluated to allow for comparison against less optimistic scenarios. Figure 2.1. illustrates the process flow, while

Table 2.1. summarizes annual quantities of intermediate feedstocks, primary ethanol products, and associated co-products.

In addition to cellulosic ethanol, one-third of corn grain production was assumed to provide refinery feedstock for producing 145 million liters of corn-grain ethanol annually. All of the

corn-grain ethanol produced in the four-county region studied was assumed to be the product of the dry milling process, forming a major co-product of 92.6 mil. kg of DDGS. Production efficiencies and ethanol and co-product yields were based on national dry mill plant averages from the Greenhouse gases, Regulated Emissions, and Energy use in Transportation (GREET)

Model (Wang 1999).

Figure 2.1. Process flow for system producing corn for animal feed, conventional ethanol from refining corn grain, and cellulosic ethanol from refining the corn stover.

Table 2.1. System intermediate and final products1

Grain for Ethanol Feedstock [mil kg / year] 362 Grain for Animal Feed [mil kg / year] 734 Stover for Ethanol Feedstock [mil kg / year] 491 Corn Ethanol [mil liter / year] 145 Distillers Grain Co- product [mil kg / year] 92.6 Stover- derived Ethanol [mil liter / year] 193 3% Electricity Co-product [mil kWh / year] 70.4 Sensitivity Cases 3% Excess Electricity Co-product [mil kWh / year] 35.2 0% Excess Electricity Co-product [mil kWh / year] 0

1Grain mass reported assumes 15.5% moisture content. Stover mass reported is dry.

2.4.2. Process steps and emissions estimation

The processes within the system boundary include corn production and associated agricultural inputs, feedstock transport, and ethanol refining. Total annual GHG emissions are shown in

Table 2.2.

Table 2.2. Annual Greenhouse Gas emission for combined corn and stover ethanol system

GHG Process Step mil kg CO 2-eq.

Input & machinery production, transport 107 Corn Field Production 65.4

Terrestrial Emission N 2O flux 318 Subtotal - Feedstock Production 490

Second Pass Harvest 5.5 Stover Replacement Nutrients 3.6 Subtotal - Stover-Only Production 9.0

Grain Feedstock Transport 1.7 Corn-Ethanol Refining 85.6 Subtotal - Corn-Grain Ethanol Refining 87.3

Stover Feedstock Transport 3.1 Stover Refining1 0.0 Subtotal - Corn-Stover Ethanol Refining 3.1

System Total 589

1 For the cellulosic refinery, process heat and electricity is fueled from residual stover. The CO2

released during stover combustion is negated by the CO2 captured during plant growth,

resulting in no net CO2 impact.

This case study is intended to simulate future operations coinciding with the time frame of the

EISA mandates through 2022. Where possible, forward-looking estimates of energy-related

emission factors were incorporated. The use of historic energy intensity and emission factors reflects the assumption that recent practices are still in place in the future. This assumption is largely due to the retrospective nature of readily available data. The following process steps were considered and the literature identified was used to estimate energy requirements and associated emissions:

1) Farming-machinery depreciable capital, farming fertilizer inputs, and transportation

of fertilizer/farming machinery to farm (Economic Research Service, 2008, Sinistore,

2008 and West and Marland, 2002);

2) Feedstock production steps, including tillage, planting, fertilizing, pesticide

application, harvest, and second pass harvesting for corn stover collection (West and

Marland, 2002);

3) Transportation of corn and corn stover-to-ethanol producing facilities (Sinistore, 2008

and Wang, 1999);

4) The dry-mill ethanol plant and the corn stover cellulosic biorefining stage (Aden et

al., 2002, Sheehan et al., 2004 and Wang, 1999).

Emissions from transporting farm inputs (e.g. fertilizer, pesticide) were based on the transportation mode and nationally averaged transport distances from GREET. model (Wang,

1999). For feedstock transport, average distances of 55 and 38 km were estimated to transport corn stover and corn grain feedstocks to their respective biorefineries. The ratio of corn grown

for ethanol, relative to corn grown for feed, is assumed to increase near the refinery. Stover

feedstock is assumed collected from all corn in the study area, both for ethanol and for feed, and

therefore, has a longer average transport distance. Feedstock transport mode was assumed to be

medium heavy-duty diesel truck, with associated fuel and emission rates supplied by GREET

(Wang, 1999).

The EPIC/APEX model was used to simulate feedstock production, providing representative

values for corn grain and stover yields, as well as terrestrial emission of GHG from soil carbon

and nitrogen fluxes (Izaurralde et al., 2006 and Williams et al., 2008). Based on the

biogeochemical modeling, terrestrial emission of nitrous oxide comprised 64% of total GHG

emission from feedstock production. Indirect land–use-change emissions (ILUC) were not

considered. To compensate for nutrients contained in the stover that are lost with stover removal,

modeling assumed a 25% increase in nitrogen, an 18% increase in phosphorus and a 50%

increase in potassium (Elmore, 1982).

Emissions calculation for corn ethanol production assumed that refinery process heat was fueled

-1 -1 by natural gas at a rate of 9.8 MJ L , with an associated emission rate of 0.056 kgCO2eqMJ

(Wang, 1999). Electricity requirements for the corn-ethanol refinery were assumed to be 0.198 kWh L-1 (Liska et al., 2007). The emission rate for regional electricity was based on a 2020 fuel mix forecast for the five states in the east north central census division (Wisconsin, Illinois,

Michigan, Indiana, Ohio). The average emission rate for the future electricity mix was 0.725

-1 kgCO2eq kWh , based on results from the U.S. EPA 9-Region MARKet ALlocation

(MARKAL) model and technology database, version 1.1 (U.S. EPA, 2009). This modeling

incorporated recent U.S. policy including EISA. These emission factor projections would change

should federal GHG rules limit emissions from the electricity sector.

Sensitivity analysis was performed to consider how the emission factors for displaced electricity

influences overall results. Excess electricity from the stover refinery would reduce electricity

required from regional power plants and associated emissions. An important question is whether

avoided emissions should be estimated based on the average emission rate for electricity, or the

anticipated change in emissions, i.e. the marginal emission rate (Meier, 2005). One emission rate

was based on the regional average of all electricity sources as described above (0.725 kgCO2eq kWh-1). Given that natural gas power plants typically provide load following services, a second

emission rate assumed that only natural gas generation was displaced. A gas-generation emission

-1 rate of 0.396 kgCO2eq kWh was based on the weighted average of gas turbine and combined- cycle generation from the same 2020 regional MARKAL forecast.

2.4.3. Allocation approaches

Multiple LCA allocation approaches were evaluated including: 1) allocation approaches based on mass, energy, and market value, 2) system expansion approaches with co-product crediting,

3) consequential approach assessing the incremental impact of stover production and refining.

The first method used to account for GHG emissions was allocation based on the physical properties of mass and energy, as well as market value. For feedstock production, Equation 2.1.

illustrates a mass-allocation ratio. Amass,CGE represents the fraction of corn stover relative to total

feedstock mass, where MCS is the mass of collected corn stover from the field, MCGE is the mass

of corn grain devoted to corn ethanol production, and MCGF is the mass of the corn grain devoted

to feed. The resulting mass-based allocation ratio is used to assign a portion of the corn-

production emissions to the corn-grain-ethanol feedstock.

M CGE Amass,CGE = (2.1) M CS + M CGE + M CGF

Similar to feedstock production, an allocation ratio can be calculated for the refinery products. In

the example shown in Equation 2.2., Amass, CE is the mass-allocation fraction used to assign refining-related emissions to corn ethanol, where MCE is the mass of corn ethanol and MDDGS is

the mass of DDGS. Mass allocation is impractical for the stover refinery, because the electricity

co-product is effectively without mass.

M CE Amass,CE = (2.2) M CE + M DGS

Calculation of energy and market value-based allocation ratios for feedstock production are

analogous to Equations 2.1. and 2.2. For the energy-based approach, farming emissions were

allocated using the feedstock energy content, assuming 15.8 MJ kg-1 for corn grain and 17.2 MJ

kg-1 for air-dried corn stover (Penn State Univ. Cooperative Extension, 2009 and U.S. DOE,

2009a). For the refining step, the energy content of the main product (ethanol) and the co-

products (e.g. electricity or DDGS) were estimated based on a lower heating value basis (Wu et al., 2006).

Table 2.3. summarizes allocation ratios for intermediate and final products. Market value allocation was based on monthly prices of corn grain, ethanol, electricity, and DDGS from 2007 and 2008 (Data Transmission Network, 2009 and U.S. DOE, 2009b). In order to obtain more consistent results, we used average prices to account for temporary price drops and surges

(Malca and Freire, 2006). A 47 $ ton-1 value was assumed for corn stover. There is not yet a

market for corn stover reflecting its potential value as an ethanol feedstock. For this study, its

future value was estimated based on the price a cellulosic biorefinery would be willing to pay

given the following assumptions (in millions 2008$): Initial Capital Cost: 150M, Annual

Operating & Maintenance Cost: 6M, Annual Feedstock Transport and Storage Cost: 20.7M,

Annual Revenue from Ethanol Sales: 68.8M, Annual Revenue from Electricity Sales: 9.25M.

Under these assumptions, a corn stover price of 47 $ ton-1 provided a 9.6% return on the

investment over a 30-year refinery life.

Table 2.3. Allocation ratios for intermediate and final products1

Allocation Method Mass Energy Market Value Products Mil kg % PJ % Mil $ % Feedstock Production Corn Grain for Feed 621 44% 11.6 45% 118 53% Corn Grain for Ethanol 306 22% 5.7 22% 58 26% Corn Stover 491 35% 8.4 33% 45 20% Corn-Grain Refining Ethanol 114 55% 3.1 55% 77 77% Distillers Grain 93 45% 2.5 45% 23 23% Corn-Stover Refining Ethanol 152 NA 4.1 94% 102 94% Electricity NA NA 0.25 6% 6 6% 1For reference case scenario assuming that 3% of the stover energy content is ultimately converted to excess electricity for export to the regional grid. 1Grain mass values differ from Table 2.1. because they have been adjusted to account for 15.5% moisture content for comparison to dry stover mass.

A system expansion approach was also used to consider co-product displaced emissions,

summarized in Tables 2.4. and 2.5. Although numerous co-products exist for both corn-ethanol

and stover-ethanol processes, this study included only the single most significant, in order to

clearly illustrate the implications of allocation methods. For corn-grain ethanol, the dry milling

process produced DDGS co-product which can be fed to cattle to displace soybean meal from the diet. Emissions avoided by using DDGS in place of the soybean meal were calculated assuming

2.5 kg of DDGS substitutes for 1 kg of soybean meal (0.4 and a soybean GHG emission rate of

-1 0.424 kgCO2eq kg ) (Kaiser, 2002; Robertson et al., 2000 and Sinistore, 2008). Other changes

in the diet are also required to substitute DDGS, but only the first-order effects of displacing

soybean meal were considered in this study. For the corn stover ethanol refinery, the co-product considered was electricity. As discussed previously, sensitivity cases were developed around the amount of excess electricity exported to the grid and associated emission rate for displaced electricity. Our reference case scenario assumes that 3% of the stover’s energy content is converted to excess electricity, exported to the upper-Midwest power grid, and displaced emissions based on the forecasted average rate of regional 2020 electricity production. As shown in Table 2.4., four additional scenarios were developed to examine lower estimates of excess electricity and both averaged and marginal emission rates. For the 0% excess electricity case, the emission factor is irrelevant, as no electricity or emissions are displaced. The avoided GHG emissions were based on the same regional emission factors from the MARKAL modeling discussed earlier.

Table 2.4. Annual DDGS co-product displaced GHG emissions1, 2

Distillers Grain (DDGS) Total DDGS (Bone Dry Produced) [mil kg] 92.6 DDGS to Soy Meal Conversion Ratio [-] 0.400 Total Soy Meal Displaced by DDGS [mil kg] 37.0 Soymeal Emission Factor [kg/CO2eq/kg soy] 0.424 Displaced Emissions [mil kgCO2eq] 16

1Soy Meal Displaced [mil kg] = 92.6 * 0.4 = 37.0 2 Displaced Emissions [mil kgCO2eq] = 37.0 * 0.424 = 16

Table 2.5. Annual electricity co-product displaced GHG emissions1, 2

CO2-eq Displaced

Excess Electricity Rate CO2-eq Scenario stover % mil kWh kg / kWh mil kg

Refererence Case 3.0A 3.0% 70.4 0.725 51

Sensitivity Case 1.5A 1.5% 35.2 0.725 26

Sensitivity Case 3.0M 3.0% 70.4 0.396 28

Sensitivity Case 1.5M 1.5% 35.2 0.396 14

Sensitivity Case 0.0 0.0% 0.0 NA 0.0

1Excess Electricity [mil kWh] = 8.45 PJ Stover Input * stover % * 278 [mil kWh PJ-1]

2 -1 Displaced CO2eq [mil kg] = Excess Electricity [mil kWh] * CO2eq Rate [kg kWh ]

A consequential approach was used to quantify the system-wide GHG emission changes as a

result of adding the corn stover refining system. Using identical life-cycle inventory data described above, emissions were estimated for the system absent the corn stover ethanol refining.

By comparing system-wide emissions with and without stover refining, the marginal impact of stover refining was quantified. While this approach would normally be applied to prospectively consider the impact of stover refining, it could arguably be used to assign emissions impacts solely to corn stover ethanol for the purposes of calculating carbon intensity. Such an approach is

included in the results section.

2.5. RESULTS

For the system described in the previous section, life-cycle GHG emission results were

- quantified in terms of grams CO2 equivalent emission per MJ ethanol energy output gCO2eq MJ

1. These carbon intensity results are also presented as a percentage relative to petroleum fuel,

-1 assuming a gasoline carbon intensity of 96 gCO2eq MJ (CARB, 2009). Results for the reference case scenario are summarized in Table 2.6.

Table 2.6. Carbon intensity for ethanol products and % GHG emission relative to gasoline using multiple LCA allocation approaches1

Allocation Method Corn-Stover Ethanol Corn Ethanol

Carbon CO2eq Carbon CO2eq Feedstock Intensity Percent of Intensity Percent of

Production Refining [gCO 2eq/MJ] Gasoline [gCO 2eq/MJ] Gasoline Attributional approach assigning all emissions from a single system scenario to all products by partitioning based on physical or other attributes.. Energy Energy Allocation Allocation 40 41% 35 36% Market Value Market Value Allocation Allocation 26 27% 54 56%

System expansion credits refinery with net emissions change from two scenarios, with and without co-product displacement. Allocation is applied to feedstock production step. Energy System Allocation Expansion 30 31% 58 61% Market Value System Allocation Expansion 15 15% 65 68% Consequential approach examines system with and without corn-stover ethanol refining, assigning net emissions change to stover-ethanol product only. Feedstock production accounting is effectively a subdivision, because only stover harvest and make-up nutrients are assigned to stover. System Subdivision Expansion -9 -9.9% -- --

Corn-ethanol emissions based on the system without stover refining and using system expansion accounting for the refining process step. System NA Expansion -- -- 76 79% 1For reference case scenario assuming that 3% of the stover energy content is ultimately converted to excess electricity and exported displacing emissions at the average rate of

the regional grid. Calculations supporting these results are illustrated in Figures 2.2. and -1 -1 2.3. CO2eq Percent of Gasoline [%] = Carbon Intensity 96 [gCO2eq MJ ]

For the reference case scenario, the energy allocation and market allocation procedures are

illustrated in Figures 2.2.a. and 2.2.b. respectively, with allocation fractions corresponding to

Table 2.3. Comparing Figures 2.2.a. and 2.2.b., we see that when the allocation fractions are

applied to feedstock production, the energy-based approach assigned 33% of corn-growing emissions to the stover, compared to 20% using market-value allocation. These emissions were carried forward to the respective refining steps for corn-grain and stover. For stover refining, the allocation fractions were similar for energy and market value approaches. For corn-grain refining, the energy allocation assigned 45% emissions to the DDGS co-product, compared to

23% by market-value allocation. The energy allocation approach resulted in a substantially lower

-1 corn-ethanol carbon intensity (35 gCO2eq MJ ), than reached by the market value approach (54

-1 gCO2eq MJ ). For stover ethanol, market-value allocation resulted in a lower carbon intensity

-1 -1 (26 gCO2eq MJ ), compared to energy allocation (40 gCO2eq MJ ). The energy allocation method shifted emissions from the corn-grain ethanol to the corn stover ethanol, substantially raising carbon intensity for the stover-derived fuel, while lowering carbon intensity for the grain- derived fuel.

The system expansion approach is illustrated in Figures 2.3.a. and 2.3.b. For the combined

stover-grain-ethanol system, allocation at the feedstock production step was seemingly

unavoidable. System expansion requires evaluating an expanded system with and without co- product substitution. In the feedstock production step, however, corn feed is the co-product.

Replacing corn feed with another product is not only difficult to conceive, but would preclude the existence of the stover. For the refining process step, system expansion is readily applicable and it significantly altered carbon intensity results, as shown in Table 2.6.

a. energy-content-based allocation b. market-value-based allocation

Figure 2.2.a. The energy-based approach allocates greenhouse gas emissions based on the embodied energy of intermediate and final products. The total greenhouse gas emissions

for corn growing (490), stover harvest (9.0), grain refining (87), and stover refining (3.1)

are 589 million kgCO2eq. and equal to the sum of emissions attributed to final products for grain feed (221), grain ethanol (108), distillers grains (88), stover-based ethanol (163), and electricity (10).

Figure 2.2.b. The market-value-based approach allocates greenhouse gas emissions based on the market value of intermediate and final products. The total greenhouse gas emissions for corn growing (490), stover harvest (9.0), grain refining (87), and stover refining (3.1) are 589 million kgCO2eq. and equal to the sum of emissions attributed to final products for grain feed (262), grain ethanol (166), distillers grains (50), stover-based ethanol (105), and electricity (6).

Values shown correspond to Tables 2.1. – 2.5. and calculations correspond to Equations 2.1 and 2.2. -1 -1 Carbon Intensity (CI) [gCO2eq MJ ] = GHG [million kgCO2eq] Energy [PJ] Units of [g MJ-1] and [mil kg PJ-1] are equivalent.

a. system expansion with b. system expansion with energy- allocated feedstock market value-allocated feedstock

Figure 2.3.a. The energy-based approach allocates greenhouse gas emissions to intermediate feedstock production products. For the refining step, system expansion attributes all greenhouse greenhouse gas emissions to a primary product, and credits displaced emissions from the co-product system. Total greenhouse gas emissions for corn growing (490), stover harvest (9.0), grain refining (87), stover refining (3.1), DGS offsets (-

14) and electricity offsets (-51) are 523 million kgCO2eq. and equal to the sum of emissions attributed to final products for grain feed (221), grain ethanol (180), and stover-based ethanol (122).

Figure 2.3.b. The energy-based approach allocates greenhouse gas emissions to

intermediate feedstock production products. For the refining step, system expansion attributes all greenhouse greenhouse gas emissions to a primary product, and credits displaced emissions from the co-product system. Total greenhouse gas emissions for corn growing (490), stover harvest (9.0), grain refining (87), stover refining (3.1), DGS offsets (-

16) and electricity offsets (-51) are 523 million kgCO2eq. and equal to the sum of emissions attributed to final products for grain feed (262), grain ethanol (201), and stover-based ethanol (60).

Values shown correspond to Tables 2.1. – 2.5. and calculations correspond to Equations 1 and 2. -1 -1 Carbon Intensity (CI) [gCO2eq MJ ] = GHG [million kgCO2eq] Energy [PJ] Units of [g MJ-1] and [mil kg PJ-1] are equivalent.

As shown in Figure 2.3.a. and 2.3.b., results were influenced by the unavoidable allocation at the feedstock production step. Stover ethanol carbon intensity was higher with energy allocated feedstock production than with market-value-allocated feedstock production, at 30 and 15 gCO2eq/MJ respectively. Conversely, corn ethanol carbon intensity was lower with energy-

allocated feedstock production than with market-value-allocated feedstock production, at 58 and

-1 65 gCO2eqMJ respectively. Relative to purely allocation approaches, system expansion at the

refining step increased corn-ethanol’s carbon intensity and decreased stover ethanol’s carbon

intensity. Using system expansion, all emissions are assigned to the primary product, less a co-

product substation credit. The DDGS co-product credit reduced corn ethanol’s emissions by only

8 - 9% through avoided soybean meal production. In contrast, electricity co-product reduced stover ethanol’s emissions by 42 - 85% by avoiding CO2-intensive electricity production.

Finally, a consequential approach was used to quantify the impact of adding the corn stover

ethanol production. Earlier we clarified a distinction between attributional approaches, which

assign emissions based on static system output, and consequential approaches, which consider

the net change in emissions under varying levels of system output. System expansion is

necessarily a consequential approach, because the emissions are measured with, and without, co-

product substitution. System expansion and consequential analysis, however, are not

synonymous. In the previous examples, system expansion is used to account for co-products by

consequentially examining the system with differing amounts of soymeal production. The

remainder of the analysis, however, remained consistent with the earlier attributional assignment

of emissions. Next, we examine a more fully consequential approach in which we measure the

incremental environmental impacts associated with the addition of the stover refinery itself. By

comparing the change in total system emissions, no explicit allocation decision was made for the

feedstock production step. In the parlance of LCA researchers, the feedstock production

allocation is effectively a “subdivision”, meaning the process is sub-divided and sub-process emissions assigned to separate products. By using this method, emissions from the sub-processes of stover harvest and make-up nutrients were effectively assigned to stover with emissions from all other feedstock sub-processes assigned to the grain. The results of the consequential approach are shown in Table 2.7. By adding the stover-ethanol system, GHG emission decreased by 54 million kgCO2eq, due mainly to emissions avoided from the electricity co-product. This net

emissions reduction could arguably be applied to the 4.1 PJ of stover-ethanol product, resulting

-1 in a negative carbon intensity of –9 gCO2eq MJ . The negative emissions result occurs because

the avoided GHG emissions from displaced electricity were larger than the incremental

emissions increases from stover production (second-pass stover harvest and additional nutrients).

In reality, stover-ethanol production would not truly function as a carbon sink. Rather, both the stover-derived ethanol and stover-derived electricity are actually GHG sources; however, their associated emissions are much lower than their fossil fuel counterparts. using a consequential approach to examine the system with and with-out the stover refinery.

It is interesting to examine the corresponding change in corn ethanol’s carbon intensity, as a result of adding the stover-ethanol system. Table 2.7. summarizes system expansion approach for only corn-grain ethanol, absent stover production. In this case, grain production for feed can be ignored and feedstock allocation avoided, because all grain produced supplies the corn ethanol

-1 refinery. The resulting corn ethanol carbon intensity is 76 gCO2eq MJ based on 233 million

kgCO2eq of GHG emission and 3.1 PJ of ethanol output. This carbon intensity can then be

compared to corn ethanol’s carbon intensity (also by system expansion) with stover refining in

place. As shown in Table 2.6., the addition of the stover-ethanol system significantly reduces

corn ethanol’s carbon intensity to 58 using energy-allocated feedstock, or 65 using market-value

allocated feedstock. This effect is expected and results from shifting a fraction of feedstock

production emissions from corn grain to stover.

Table 2.7. Annual system GHG emission for corn-ethanol refining absent corn feed and stover-ethanol production

Corn Ethanol by System Expansion GHG Process Step mil kg CO 2-eq.

Input & machinery production, transport 35 Corn Field Production 21.6

Terrestrial Emission N 2O flux 105 Subtotal - Feedstock Production 162

Grain Feedstock Transport 1.7 Corn-Ethanol Refining 85.6 DGS Co-product Credit -15.7 Subtotal - Corn-Grain Ethanol Refining 71.6

Total 233

This fully consequential approach demonstrates how system emissions might change with, and

without, the existence of the cellulosic refinery and associated ethanol production. This question

and the consequential approach are reasonable given that conventional corn ethanol production is

in practice, and that stover-ethanol production is not. Using the resulting change in system

emissions to calculate stover-ethanol’s carbon intensity, however, is most likely inappropriate.

The regulatory benchmarks tested in this study, would only apply after such a refinery were

actually in place. If the goal of the LCA study was not carbon intensity measurement, but instead

to measure total GHG mitigation potential, then the consequential approach would be appropriate and would preferably include gasoline displacement within the system boundary.

Table 2.8. Annual system GHG emission with and without corn stover refining1

GHG mil kg CO2-eq. Absent With Process Step Stover-Ethanol Stover-Ethanol

Feedstock Production 490 499 Feedstock Transport & Refining 87.3 90.4 Co-product Credits -15.7 -66.8 System Emission Sub- total 561 523

Net System Change -39

1For reference case scenario assuming that 3% of the stover energy content is ultimately converted to excess electricity and exported displacing emissions at the average rate of the regional grid.

Figure 2.4. illustrates the results of the sensitivity analysis. For each allocation procedure, sensitivity analysis compares the impact of various levels of excess electricity for the stover refinery, with avoided emissions estimated from either average grid emission rate or a marginal grid emission rate. The amount of excess electricity is only significant for the approaches using system expansion, because the other approaches do not consider co-product displaced emissions

(refer to Figures 2.2. and 2.3.). Figure 2.4. quantifies a readily anticipated impact. Stover ethanol’s carbon intensity increases as the excess electricity and associated emissions decrease.

Logically, carbon intensity increases when avoided emissions are estimated at the lower marginal rate, rather than the higher average rate.

Figure 2.4. Carbon intensity as a percentage of gasoline: For each allocation procedure, sensitivity analysis compares the impact of various levels of excess electricity for the stover refinery, with avoided emissions estimated from either average grid emission rate or marginal grid emission rate.1 At far right, a consequential approach considers the net change in system emissions from adding stover ethanol production. The emission reduction is attributed only to stover ethanol, while corn ethanol emissions are based on the system absent stover ethanol.

2.6. DISCUSSION

The results of this study demonstrate that allocation procedures are paramount decisions for LCA

-1 studies. Corn-ethanol carbon intensity varied from 35 to 76 gCO2eq MJ (36% – 79% of

gasoline) under the examined allocation approaches. Stover ethanol varied from between -9 to 39

-1 gCO2eq MJ (-10% – 41% of gasoline) for the reference case scenario. Energy allocation

resulted in the highest carbon intensity for stover ethanol and the correspondingly lowest carbon

intensity for corn ethanol. As previously discussed by Meier, treating different forms of energy

equally potentially diminishes the meaningfulness of results (Meier, 2002). Using energy

allocation, the energy content of readily convertible sugars and starches was treated as equivalent

to the energy content of recalcitrant plant cellulose. Allocation by market value is arguably more

appropriate, because the feedstock prices should reflect the relative capital investments and

operating costs necessary to convert each feedstock to an identical ethanol product. The

drawback to market allocation in this study, however, is that it requires speculating hypothetical

stover prices.

Despite the wide range of results, the carbon intensity values for both corn-grain-derived and

stover-derived ethanol were dramatically lower than that of gasoline. For the reference case, the

stover ethanol complied with the 60% reduction target established by EISA (i.e. 40% or lower

CO2eq as a percentage of gasoline) under each of the allocation approaches examined. The

exception was for energy allocation, which barely exceeded the target with 41% of gasoline’s

emissions. EISA’s requirement for new corn ethanol to have less than 80% of gasoline’s CO2eq

is not relevant here, because we assumed corn grain was supplied to existing ethanol refineries.

The 20% GHG reduction target still provides an interesting benchmark. When corn ethanol was

examined without the stover refinery present, corn-ethanol’s carbon intensity was 79% that of

gasoline. This is within the typical range of corn ethanol LCA results using system expansion in

areas with high-productivity corn production, without consideration of indirect land use change.

The addition of the corn stover refinery dramatically lowered corn ethanol’s carbon intensity

under all allocation approaches examined, because the stover-ethanol system results in more than twice the ethanol output from the same area of corn production, but with only minor increases in emissions. By distributing slightly more emissions across substantially more product, corn ethanol’s carbon intensity was reduced to between 36 to 68% that of gasoline, depending on

allocation approach.

Sensitivity analysis considered less optimistic assumptions for the excess electricity exported to

the grid by the stover refinery. The least optimistic scenario, with no excess electricity, resulted

in a slightly higher range of stover ethanol carbon intensity, between 3.1 % to 43.7% that of

gasoline. Emission rate assumptions for displaced electricity were also shown to be important.

Switching from an average electricity emission rate to a marginal electricity emission rate

(reflecting natural gas generation) increased stover ethanol’s carbon intensity when using system expansion approaches. This increase was uniform, adding 5.9% to carbon intensity values

(percentage relative to gasoline) for scenarios with 3% excess electricity, and adding 2.9% for scenarios with 1.5% excess electricity. The influence of allocation methods on results was far more important than the influence of sensitivity analysis around this one key input parameter.

While this would seem to support an argument that varying allocation methods should be a standard part of LCA sensitivity analysis, the resulting array of results may be more confusing and less useful. Rather than testing many allocation approaches, it is likely more valuable to carefully deliberate and justify the appropriate practices based on the goal of the study.

The ISO preference for system expansion should not be taken for granted. ISO guidance recommends using system expansion to credit products for emissions displaced by co-products, instead of using simple allocation to partition emissions between product and co-product based on their ratio of mass, energy or market value. The latter may be preferable at times, because it simply distributes the system’s entire emissions inventory (historic or projected) to associated products. In the system expansion approach, some emissions are instead deducted from the inventory based on the difference between the baseline scenario and what may otherwise occur.

In other words, the system expansion approach is subtracting “actual” emissions from the system inventory, to account for “virtual” emissions offsets. In the example illustrated in Figure 2.2.,

estimated emission offsets (shown in red) are based on hypothetical production of soybean meal

or electricity that did not occur, but may have occurred in the absence of the co-products. Even if

the anticipated product displacement is extremely likely, the sum of product-assigned emissions

no longer matches the actual historic emission inventory, a potentially undesirable effect for

regulatory purposes.

For measuring carbon intensity, simply allocating emissions based on physical or economic

partitioning (an attributional approach) may be preferable, particularly if corroboration with

historic emission inventories is important. The system expansion approach may not be consistent

with the long-term goals of these regulations. Allocation by partitioning is readily scalable, so if

carbon intensity rules were expanded to other fuels, product-assigned emissions would remain consistent with emission inventories. In contrast, extending the system expansion requires a myriad of co-product considerations, leading to further divergence from emission inventories.

Inconsistencies with the system expansion approach are further demonstrated in Figure 2.2., where GHG emissions are assigned to one energy product, ethanol, but not to another energy product, electricity. In this example, the ability to accurately exchange GHG offsets between electricity and transportation-fuels sectors would be limited. However, if the LCA goal is net benefits analysis (e.g. evaluating the realized or potential impact of a policy), then consequential approaches, including system expansion, is likely preferable. In such cases, marginal emission rates may be preferable to average emission rates for estimating avoided emissions (Meier,

2005).

Given the wide range of carbon intensity results, the allocation methods require careful consideration in context of the specific GHG policy objectives, such as those established by

EISA. Otherwise, the regulatory approach could diverge from the legislative intent due to the strong influence of allocation methods on results. Legislative proposals requiring LCA would preferably be developed in conjunction with analysis of regulatory implementation. In which case, explicit allocation approaches can ensure that the allocation used for implementation is consistent with the allocation used to establish policy objectives. In all cases, the allocation choices, as well as their justification and implications, should be clearly conveyed.

Acknowledgements

This work was part of the DOE Great Lakes Bioenergy Research Center

(www.greatlakesbioenergy.org) supported by the U. S. Department of Energy, Office of Science,

Office of Biological and Environmental Research, through Cooperative Agreement DE-FC02-

07ER64494 between The Board of Regents of the University of Wisconsin System and the U. S.

Department of Energy. We thank R. Cesar Izaurralde, Joint Global Change Research Institute, for the biogeochemical modeling of feedstock production which provides valuable input to this work.

2.7. REFERENCES

Aden, A., Ruth, M., Ibsen, K., Jechura, J., Neeves, K., Sheehan, J. and Wallace, B. 2002. Lignocellulosic biomass to ethanol process design and economics utilizing co-current

dilute acid prehydrolysis and enzymatic hydrolysis for corn stover (NREL/TP-510- 32438); National Renewable Energy Laboratory: Golden CO, 2002. Available at: http://www.nrel.gov/docs/fy02osti/32438.pdf.

Beer, T. and Grant, T. 2007. Life-cycle analysis of emissions from fuel ethanol and blends in Australian heavy and light vehicles. Journal of Cleaner Production. 15: 833-837.

Bjorklund, A. 2002. Survey of approaches to improve reliability in LCA. International Journal of Life Cycle Assessment. 7: 64-72.

California Environmental Protection Agency Air Resources Board (CARB). 2009. Proposed regulation to implement the low carbon fuel standard: vol. 1; CaARB: Sacramento, CA. http://www.arb.ca.gov/regact/2009/lcfs09/lcfsisor1.pdf. Accessed: Nov 2009.

Curran, M. A., Mann, M. and Norris, G. 2005. The international workshop on electricity data for life cycle inventories. Journal of Cleaner Production 13(8): 853-862.

Curran, M.A. 2007. Studying the effect on system preference by varying coproduct allocation in creating life-cycle inventory. Environmental Science and Technology. 41: 7145-7151.

Data Transmission Network. 2009. Ethanol Center; DTN: Omaha NE. http://www.dtnethanolcenter.com/. Accessed: Feb 2009.

Economic Research Service (ERS). 2008. Farm Business and Household Survey Data. http://www.ers.usda.gov/data/arms/app/Crop.aspx. Accessed: Nov 2008.

Ekval, T. and Tillman, A. 1997. Open-Loop Recycling: Criteria for Allocation Procedures. International Journal of Life Cycle Assessment. 2: 155-162.

Ekvall, T. and Andrae, A. 2006. Attribuitional and consequential environmental assessment of shift to lead-free solders. International Journal of Life Cycle Assessment. 11: 344-353.

Ekvall, T. and Finnveden, G. 2001. Allocation in ISO 14041—a critical review. Journal of Cleaner Production. 9: 197-208.

Elmore, R. 1982. How a corn plant develops (SR0048). Iowa State Extension publication: Ames, IA.

Farrell, A., Plevin, R., Turner, B., Jones, A., O'Hare, M., Kammen, D. 2006. Ethanol can contribute to energy and environmental goals. Science. 311: 506-508.

Guinée, J., Heijungs, R. and van der, Voet, E. 2009. A greenhouse gas indicator for bioenergy: some theoretical issues with practical implications. The International Journal of Life Cycle Assessment. 14: 328-339.

International Organization for Standardization (ISO). 2006a. ISO 14040: environmental management—life cycle assessment—principals and framework. Geneva, Switzerland.

International Organization for Standardization (ISO). 2006b. ISO 14044: environmental management—life cycle assessment—requirements and guidelines. Geneva, Switzerland.

Izaurralde, R. C., Williams, J. R., McGill, W. B., Rosenberg, N. J., and Quiroga Jakas, M. C. 2006. Simulating soil C dynamics with EPIC: Model description and testing against long- term data. Ecological Modeling. 192: 362-384.

James, L., Swinton, S. and Thelen, K. 2010. Profitability Analysis of Cellulosic Energy Crops Compared with Corn. Agronomy Journal. 102: 675–687.

Kaiser, R. 2005. Utilizing the Growing Local Supply of Distillers Grains. http://www.uwex.edu/ces/dairynutrition/pubs.cfm. Accessed: Feb 2008.

Kim, S. and Dale, B. E. 2002. Allocation procedure in ethanol production system from corn grain. International Journal of Life Cycle Assessment. 7: 237-243.

Kim, S., Dale, B. E. and Jenkins, R. 2009. Life cycle assessment of corn grain and corn stover in the United States. International Journal of Life Cycle Assessment. 14: 160-174.

Kodera, K. 2007. Analysis of allocation methods of bioethanol LCA. MSc Thesis. CML, Leiden University, Netherlands.

Luo, L., van der Voet, E., Huppes, G. and Udo de Haes, H.A. 2009. Allocation issues in LCA methodology: a case study of corn stover-based fuel ethanol. The International Journal of Life Cycle Assessment. 14: 529-539.

Malca, J. and Freire, F. 2006. Renewability and life-cycle energy efficiency of bioethanol and bio-ethyl tertiary butyl ether (bioETBE): Assessing the implications of allocation. Energy. 31: 3362-3380.

Meier, P. 2002. Life-Cycle Assessment of Electricity Generation Systems and Applications for Climate Change Policy Analysis. Ph.D. Dissertation. The University of Wisconsin – Madison, Madison, WI, pp. 59 -66.

Meier, P., Erickson, J. and Mapp, J. 2005. Measuring Emission Benefits with Integrated Resource Models - Why Simple Estimates of Emission Reductions May Be Wrong. Proceedings of the International Energy Program Evaluation Conference (IEPEC). 659- 670.

Mendoza, A., van Ruijven, T., Vad, K. and Wardenaar, T. 2008. The Allocation Problem in Bio- Electricity Chains. MSc. Industrial Ecology. Leiden, The Netherlands.

Pordesimo, L., Edens, W., Sokhansanj, S. 2004. Distribution of aboveground biomass in corn stover. Biomass and Bioenergy. 26: 337-343.

Robertson, G., Paul, E. and Harwood, R. 2000. Greenhouse Gases in Intensive Agriculture: Contributions of Individual Gases to the Radiative Forcing of the Atmosphere. Science. 289(5486): 1922-1925.

Sheehan, J., Aden, A., Paustian, K., Killian, K., Brenner, J., Walsh, M., and Nelson, R. 2004. Energy and environmental aspects of using corn stover for fuel ethanol. Journal of Industrial Ecology. 7: 117.

Sinistore, J. C. 2008. Corn ethanol production in the Wisconsin agricultural context: Energy efficiency, greenhouse gas neutrality and soil and water implications. M.S. Thesis. The University of Wisconsin – Madison, Madison, WI.

Sissine, F. 2007. Energy independence and security act of 2007: A summary of major provisions (RL34294). CRS Report for Congress: Washington D.C. http://assets.opencrs.com/rpts/RL34294_20071221.pdf. Accessed: Nov 2009.

Thomassen, M., Dalgaard, R., Heijungs, R. and de Boer, I. 2008. Attributional and consequential LCA of milk production. International Journal of Life Cycle Assessment. 13: 339-349.

Tillman, A. 2000. Significance of decision-making for LCA methodology. Environmental Impact Assessment Review. 20: 113-123.

U.S. Department of Agriculture (USDA). 2008. National Agriculture Statistics Service. http://www.nass.usda.gov. Accessed: Nov 2008.

U.S. Department of Energy (DOE). 2009a. Energy Efficiency and Renewable Energy, Biomass Feedstock Composition and Property Database. http://www1.eere.energy.gov/biomass/feedstock_databases.html. Accessed: March 2009.

U.S. Department of Energy (DOE). 2009b.Energy Information Administration. Average Retail Price of Electricity to Ultimate Customers: Total by End-Use Sector. http://www.eia.doe.gov/fuelelectric.html. Accessed: Feb 2009.

U.S. Environmental Protection Agency (EPA). 2009. U.S. EPA MARKAL 9-Region Database (version 1.1); U.S. EPA Office of Research and Development: Research Triangle, NC. 2009.

Wang, M. 2009. The Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) Model (Version 1.8b). Center for Transportation Research; Energy Systems Division, Argonne National Laboratory/Department of Energy: Argonne, IL.

Weidema, B. P. 1993. Market aspects in product life cycle inventory methodology. Journal of Cleaner Production. 1: 161-166.

Weidema, B. 2001. Avoiding Co-Product Allocation in Life-Cycle Assessment. Journal of Industrial Ecology. 4(3): 11-33.

West, T. O. and Marland, G. 2002. A synthesis of carbon sequestration, carbon emissions, and net carbon flux in agriculture: comparing tillage practices in the United States. Agriculture, Ecosystems & Environment. 91: 217-232.

Williams, J., Izaurralde, R.C. and Steglich, E. 2008. Agricultural Policy/Environmental eXtender Model. Theoretical Documentation, Version 0604 (BREC Report # 2008-17). AgriLife Research, Texas A&M System, Blackland Research and Extension: Temple, TX, p.130.

Wu, M. Wang, M., and Huo, H. 2006. Fuel cycle assessment of selected bioethanol production pathways in the United States (ANL/ESD/06-7). Energy Systems Division, Argonne National Laboratory: Argonne, IL.

CHAPTER 3

Methodological review of recent ethanol production studies and recommendations to

facilitate comparison

3.1. PUBLICATION AND CONTRIBUTION DETAILS

In the summer of 2011, I completed a meta-analysis on seven major corn and cellulosic environmental impact assessments as a part of my annotated bibliography work and wrote the paper “Methodological Review of Recent Ethanol Production Studies and Recommendations to

Facilitate Comparisons”. This paper was submitted to the International Journal of Life Cycle

Assessment in August of 2011 and was accepted with major revisions in early January of 2012.

The paper was revised and re-submitted by the end of January, 2012. My contribution to the following text included 100% of the background research and methodology, approximately 95% of the text and 95% of the revisions completed for the response to the reviewers’ comments. The conclusions of this study shaped the development and presentation of the LCA methods used in the next chapter of this dissertation.

Authors: Julie C Sinistore1,2 and Douglas J Reinemann1,2

1) Great Lakes Bioenergy Research Center, University of Wisconsin-Madison

2) Biological Systems Engineering, University of Wisconsin-Madison

3.2 ABSTRACT

Purpose. The environmental impact assessment of corn-grain and cellulosic ethanol production has been the subject of many published studies in recent years because such evaluations are

needed to inform policy and guide future research. Many researchers, however, either do not use

Life Cycle Assessment (LCA) for their evaluations or they apply LCA methods inconsistently.

Four key LCA components vary widely: functional units, system boundaries, allocation methods

and impact category metrics.

Methods. To illustrate the consequences of a lack of agreement in methodology on the

comparability of results, we compare these four key components as they are employed in seven

representative studies. Previous studies have illustrated the effects of these choices on results, but

we draw attention to just how incommensurate the methods used in ethanol studies are. We begin

with a brief summary of each article. Following that, we compare the key LCA components of

each study side-by-side to shed light on common methodological choices and facilitate the

development of more standardized methods for future transportation fuel production studies.

Results and discussion. The results of this study show the broad and inconsistent application of

LCA methodology in the life-cycle energy and emissions assessments of corn-grain and

cellulosic ethanol production. We aver that LCA is the most appropriate framework for biofuel

environmental impact analysis. Therefore, if the underlying components of biofuels LCAs are incongruent, comparison of the results of such studies is difficult or not possible. Particular attention is paid to the choice of functional unit and system boundary that will facilitate the

comparison of biofuels to fossil fuels and other developing alternative transportation fuels such as electricity.

Conclusions and recommendations. To facilitate the comparison of transportation fuel production LCAs, we recommend: 1) the inclusion of a summary table that clearly states the key

LCA components used in the study with the abstract or methods section of published fuel production LCAs; 2) the use of the energy (MJ) functional unit; 3) the inclusion of depreciable capital and all relevant material production unit processes in and the exclusion of blended ethanol combustion from the system boundary; 4) the presentation of all major allocations methods if allocation avoidance is not possible.

Keywords: Life Cycle Assessment, biofuel, ethanol, corn ethanol, cellulosic ethanol, transportation fuel

3.3. INTRODUCTION AND BACKGROUND

Biofuels, produced from a variety of feedstocks and production pathways, have been touted as energy efficient, greenhouse gas (GHG) neutral or negative and otherwise environmentally advantageous. Life Cycle Assessment (LCA) has emerged as a highly effective tool for the evaluation of biofuels and the comparison of biofuels to their fossil counterparts. An important distinction must be made however, to clarify that LCA studies do not evaluate the fuel itself, but rather the production and use of the fuel. The goal of an LCA is to assess the environmental impacts of the production of a good or service (ISO 14040, 2006 and ISO 14044, 2006).

Comparable evaluations of fuel production systems are needed to guide policy-makers, scientists

and the general public to select, research and support the most environmentally sustainable fuels for production.

Ethanol can be produced from numerous feedstocks via various pathways. Corn-grain ethanol production has already been widely implemented at the commercial scale in the US; therefore, this biofuel system can be studied directly. Nevertheless, corn ethanol production methods are not uniform from facility to facility. Corn-grain ethanol can be produced by dry-milling, dry grind and wet-milling processes which all have different energy requirements, conversion efficiencies and co-products. Furthermore, the age of the facility can impact energy requirements and conversion efficiency (Brown, 2003). Some LCAs study a handful of actual plants with real operational data (Liska et al., 2009 and Sinistore and Bland 2010) while others generalize production over large geographic areas such as one state or country (Hill et al., 2009; Hill et al.,

2006; Luo et al., 2009 and Spatari et al., 2010).

Cellulosic ethanol LCAs are further complicated by the fact that there are no commercial scale cellulosic ethanol facilities in operation today. There are several pilot and demonstration scale

cellulosic ethanol plants, but it is very difficult to get actual operational data from them and

scaling up those data to perform accurate analysis on large scale production is tenuous.

Moreover, the efficiency and production levels of a cellulosic ethanol plant hinge on the

feedstock and the pretreatment method. Commonly studied feedstocks include corn stover,

switchgrass, mixed prairie grasses and Miscanthus, but various other studies have also

considered cassava, woody biomass, rice straw, sorghum, citrus peels, wheat straw and many

others (Worldwatch Institute 2007).

The most common method of cellulosic ethanol production is the sugar fermentation pathway

which generally begins with a stringent form of pretreatment aimed at exposing the cellulose in

the lignocellulosic plant wall matrix. Pretreatment methods include steam explosion, dilute acid,

and Ammonia Fiber Expansion (AFEX), though there are many other pretreatment methods.

Each pretreatment method produces variable quantities of hydrolysable cellulose and

hemicelluloses. Hydrolysis can be performed with various different cocktails of enzymes that

also produce variable amounts of five- and six- carbon sugars for fermentation. Even the final

sugar fermentation step is unpredictable due to many different factors such as the presence and concentration of fermentation inhibitors formed during pretreatment and the type of fermentation organism used (Brown, 2003). On the whole, there are numerous ways to produce ethanol fuel and all of these production pathways need to be assessed and compared for their relative environmental impacts. Furthermore, these impacts must be compared to other biofuels, conventional fuels and alternative transportation energy carriers.

In recent years, there have been numerous papers published in notable journals that assess the environmental impacts of corn-grain and cellulosic ethanol biofuel production (Beer and Grant,

2007; Curran, 2007; Farrell et al., 2006; Fu et al., 2003; Hill et al., 2009; Hill et al., 2006;

Kauffman et al., 2010; Laser et al., 2009; Liska et al., 2009; Luo et al., 2009; Spatari et al., 2010 and Spatari et al., 2005). The various functional units, system boundaries, allocation methods, and impact category metrics of these articles are as numerous as the number of papers itself. All

of the variation in ethanol production pathways makes it hard enough to compare studies on

different ethanol fuels internally (e.g. corn grain and cellulosic ethanol) or to other transportation

fuels (e.g. ethanol, gasoline, diesel and electricity). This difficulty is only compounded by the

great variation in analysis methodology. A study by Liska and Cassman (2008) notes that the wide variety of metrics used in LCA studies of various biofuels (not just ethanol) makes it difficult to compare the fuels on the same playing field. The authors further assert that a generic framework for standardizing analysis is needed so that future biofuel LCAs can be conducted in a way that ensures comparable results. Many other recent studies note that the net energy balance and GHG emissions of a fuel production alone are insufficient indicators of a fuel’s environmental impact. Other impact categories, such as air pollution emissions, acidification, eutrophication, water usage and many more, must also be evaluated before a final judgment on the sustainability of a fuel can be made (Hill et al., 2009; Luo et al., 2009 and Curran, 2007).

Our objective is to bring together several representative US ethanol production studies and compare their methods side-by-side to illuminate common methodological choices and facilitate the development of more standardized methods for future biofuel studies. These articles were selected because either they are oft-cited articles or they present a unique method for biofuels analysis. Furthermore, these seven studies illustrate the range of methods used to present the four key LCA components of interest to our study and all were published in or after 2006 which is the year that the revised ISO 14040 Standards series on LCA principles, framework, requirements and guidelines was published. Not all of the included studies claim to be LCAs, but they all evaluate the environmental impacts of ethanol fuel production in various ways, and many of these articles are cited for their environmental analysis of biofuel production. Since LCA has been indentified as an appropriate tool for the environmental impact analysis of the production of goods, it is necessary to examine these ethanol production studies through the lens of LCA.

We do not aim to compare the results of the studies in terms of the net energy balance, GHG emissions or other impacts from the production of corn-grain or cellulosic ethanol since other existing papers have already done this (Farrell et al., 2006; Liska and Cassman, 2008 and von

Blottnitz and Curran, 2007). Instead, we intend to show the variation in four key LCA

components used in recent biofuel production studies, discuss the merits of and reasons for these

various choices and illustrate how difficult it is to compare studies that use different versions of

these four components. The four key components are: functional units, system boundaries,

allocation methods and impact category metrics. We also endeavor to shed light on the lack of

agreement on these four components and further the discussion of more uniform methods and

metrics. Finally, we will recommend ways to communicate the key components of a biofuel

study so that readers can easily determine if studies are comparable.

3.4. METHODS

A review of all of the publications on this topic in the last five years alone could occupy several

volumes; therefore, we selected a relevant subset for this study (Table 3.1.). First, a brief

overview of each article is given. The discussion that follows compares and contrasts the four

key LCA components (functional unit, system boundary, allocation and impact category metrics)

employed in the seven studies. For ease of side-by-side comparison, the methods used by each

study for all four components are presented in Table 3.2. at the end of this section. The location

of the four key LCA components in each of the seven studies is presented in the Supplemental

Table 3.1.

Table 3.1. Articles reviewed in this study and the feedstocks considered

Study Title Ethanol feedstock reference

Curran, Studying the effect on system preference by Corn grain

2007 varying co-product allocation in creating life-

cycle inventory

Hill et al., Environmental, economic, and energetic costs Corn grain

2006 and benefits of biodiesel and ethanol biofuels

Hill et al., Climate change and health costs of air Corn grain, corn stover,

2009 emissions from biofuels and gasoline switchgrass, diverse prairie

biomass and Miscanthus

Laser et al., Coproduction of ethanol and power from Switchgrass

2009 switchgrass

Liska et al., Improvements in Life Cycle Energy Corn grain

2009 Efficiency and Greenhouse Gas Emissions of

Corn-Ethanol

Luo et al., Allocation issues in LCA methodology: a Corn stover

2009 case study of corn stover-based fuel ethanol

Spatari et Life cycle evaluation of emerging Corn stover and switchgrass

al., 2010 lignocellulosic ethanol conversion

technologies

Curran (2007) presents a theoretical study not on the effect of co-product allocation methods on

the internal LCA results for one product (like ethanol or gasoline), but on the effect of varying

co-product methodologies in the same way across the LCAs of three products (gasoline with

ethanol 8.7% by volume, E85 and conventional gasoline). This was done to see if the allocation

method changes the relative results. The ethanol feedstock for the compared fuels was corn grain

and the conversion processes included wet and dry milling. The researcher used a combination of

Excel spreadsheets, and the Environmental Protection Agency (EPA) models MOBILE 6 (model

for generic passenger automobiles) and TRACI (EPA’s Tool for the Reduction and Assessment

of Chemical and other Impact) to conduct her analysis. The functional unit was clearly stated to be per 1000 gallons (3785 liters) of energy equivalent conventional gasoline (1014 gallons (3838

L) of gasoline with 8.7% ethanol and 1380 gallons (5224 L) of E85). The system boundaries for each production process are clearly depicted in figures and include corn grain production, ethanol production, crude oil extraction, petroleum refining, bulk terminal storage, fueling and final combustion of the fuels for vehicle operation. It is unclear, however, if the production of the inputs to the corn production system (e.g. fertilizers) were included in the analysis. The results of this study quantified nine environmental impacts: acidification, ecotoxicity, eutrophication, global warming, human health cancer, human health noncancer, ozone depletion and

photochemical smog. This study did not evaluate the energy balances of the considered products.

The study considered the following co-product allocation methods: mass, volume, energy,

economic and demand. The major finding of this study was that allocation methodology does not

alter the final LCA result when comparing different ethanol fuel products. The results indicated that E85 had lower impact than gasoline on global warming, ozone depletion and human health

(cancer and noncancer). Gasoline blended with 8.7% ethanol had a lower impact than gasoline in

all of the same categories except human health cancer.

Hill et al. (2009) constitutes a comprehensive evaluation of the life-cycle GHG and Particulate

Matter (of 2.5 μm diameter) (PM2.5) emissions from corn-grain ethanol by three different production methods and cellulosic ethanol by four different production methods as they compare to a gasoline baseline. Corn-grain ethanol was assumed to be produced in the same way across scenarios, but with different sources of thermal heat for the biorefinery including coal, natural gas (conventional), natural gas (mature case with fewer emissions in the production of the gas) and corn stover. The two natural gas scenarios were counted as one production method. The four cellulosic ethanol production methods varied not by heat source but by feedstock. Cellulosic

ethanol was assumed to be produced from corn stover, switchgrass, diverse prairie biomass and

Miscanthus. The Greenhouse gases, Regulated Emissions, and Energy use in Transportation

(GREET) model was used to model GHG emissions on a per-gallon and per-liter of fuel basis.

This model does not allow the user to model PM2.5, so the Response Surface Model (RSM)

developed by the US EPA was used to calculate particulate matter emissions. According to the

GREET model and Figure 2 in the paper, the functional unit is one gallon or liter of ethanol

produced (or gasoline for the baseline). The abstract of this paper, however, notes that the

authors “quantify and monetize the life-cycle climate-change and health effects … for each

billion ethanol-equivalent gallons of fuel produced and combusted in the US”. The impact

category metric presented was dollars of human health cost per gallon or liter of fuel. Estimates

were made regarding the financial cost to society of GHG emissions (relating to global warming

impacts) and PM2.5 emissions relating to health costs. Values for the GHG emissions costs were

based on existing carbon markets in the European Union and estimates on the cost of carbon

mitigation. The system boundaries for the analysis were not given in a figure, but sources of

emissions from various processes were noted in a list. From the methods section of the paper and

the online supporting materials, the system boundaries include corn grain production (and

upstream inputs), cellulosic biomass production (and upstream inputs), ethanol production,

electricity generation, gasoline production and fuel combustion in vehicles. Also included were

the GHG and particulate matter emissions from the combustion of biomass at the biorefinery and

ethanol fuel. The allocation methods used are referred to as the “GREET defaults methods”.

From the GREET model, these defaults appear to be mass, energy and economic allocation, but it is difficult to assess exactly how these are calculated from the model (Wang, 2007). A co- product GHG credit of 17% of the ammonia emissions from fertilizer production is mentioned in the supporting materials, but the calculation method for this credit is not given.

Hill et al. (2006) is a landmark study that evaluated the conventional production pathways of

corn-grain ethanol and biodiesel fuels for their ability to meet the goals of net energy gain,

environmental benefit and economic viability. Data from the USDA were used to quantify on-

farm impacts and data from other studies were used to quantify biorefinery impacts on the life-

cycle of fuel production. Though the system boundary is not explicitly stated in the text or with a

figure, it can be deduced from the methods section that it begins with the farm (and the upstream

production of farm inputs), continues through the biorefinery and ends with the combustion of

the biofuel. This study is unique in that it also included the energy use of what it calls “sustaining

production facility workers and their households”, which includes many energy streams not

included in other biofuel LCAs. A figure caption explains that the functional unit for analysis is

1 MJ from ethanol and every other measure is normalized to be comparable with that unit. The

methods section clearly states that the allocation methods used were economic displacement,

mass balance, energy content and market value, but the details about how these values were

calculated can only be found in the supporting materials. The metrics employed in this study

were Net Energy Benefit (NEB), Net Energy Benefit Ratio (NEB Ratio) and GHG emissions per

NEB. Environmental impacts from the application of fertilizers to corn and soybean agricultural

lands were summarized in the article. The authors also compared the release of other air

emissions from the production and combustion of the fuel, but this study did not generate these

data. Instead, this study compiled values from other studies and analyzed the aggregated effects.

The economic competitiveness of biofuels with conventional transportation fuels was also

evaluated in this study.

Laser et al. (2009) used the Aspen Plus model to evaluate three process-technology scenarios for

the co-production of ethanol and electricity from switchgrass. It referred to the “base case” technology scenario as dilute acid pretreatment with simultaneous sacharrification and fermentation (SSF). The two “mature” technology scenarios both used Ammonia Fiber

Expansion (AFEX) with consolidated bioprocessing (CBP), but used different electricity and heat generation technologies (Rankine and gas turbine combined cycle). The system boundary is

not clearly defined, although several complicated process diagrams are given. From these

diagrams and a careful examination of the methodology section, it was determined that the

system boundary included switchgrass feedstock storage and handling, ethanol production,

utilities production and waste treatment. The upstream inputs to the production of switchgrass, however, do not appear to be factored in to the life cycle of ethanol or electricity production.

There was also no specific statement about the functional unit of the study or if allocation was employed. The functional unit appears to be the Lower Heating Value (LHV) of a switchgrass feedstock evaluated in another article. This study mainly assessed the relative costs of these production methods as they relate to base-case cellulosic ethanol technology, but also included the overall energy efficiency and potential fossil fuel displacement. The metrics used for the energy efficiency results were process efficiency, fossil fuel displacement ratio, percent reduction in process power and percent reduction in process steam demand.

Liska et al. (2009) includes four large scale surveys of US corn-grain ethanol plants. These surveys collected data pertaining to corn grain and energy use (natural gas and electricity) and other inputs to wet and dry milling ethanol plants as well as ethanol production data. These data were analyzed with the Biofuel Energy Systems Simulator (BESS) Model, developed at the

University of Nebraska Lincoln, to assess three primary metrics of corn grain ethanol fuel production: Net Energy Yield (NEY), Net Energy Ratio (NER) and GHG intensity. The authors also used actual agricultural production data from the USDA Economic Research Service for the major US corn-producing states. Various production scenarios involving thermal energy for production from natural gas, coal and biogas produced from the manure of animals fed the

Distillers Grains with Solubles (DGS) co-products of corn-grain ethanol production (“closed-

loop” scenario) were evaluated. The functional unit was not clearly defined in the text, but could be found in the BESS manual as one liter of corn-grain ethanol. The system boundary of this

LCA was from “seed-to-fuel” in the authors’ terminology or from agricultural production to the fuel as the end product. Ethanol combustion was not considered. To deal with allocation, the authors developed a unique co-product crediting system specific to cattle (beef) feeding which they called “displacement,” but it is similar to the system expansion method. The goal of this study was to evaluate the most current state of corn-grain ethanol production technology to determine if it met the Energy Independence and Security Act (EISA) 2007 GHG reduction mandates and the potential (at the time) California Low Carbon Fuel Standard.

Luo et al. (2009) set out to quantify and analyze the effects of the four major co-product crediting methods accepted in LCA methodology (mass-allocation, energy-allocation, economic- allocation and systems expansion as described in ISO 14040-44) on a corn-stover ethanol fuel production system. This study evaluated the GHG impact of ethanol production, but did not evaluate the energy balance. It did, however, evaluate many other environmental metrics, including Abiotic depletion (ADP), Ozone layer depletion (ODP), Photochemical oxidation

(POCP), Human and ecotoxicity (HTP and ETP), Acidification Potential (AP) and

Eutrophication Potential (EP). Part of the reason this study was able to evaluate many additional environmental impact categories is that the researchers used data from the US Life Cycle

Inventory Database (from the National Renewable Energy Labs or NREL) and datasets from

EcoInvent that include this information in addition to GHG impacts. The software tool used in this analysis was Chain Management by Life Cycle Assessment. Unlike many studies that evaluate environmental impacts on a per-liter of ethanol or per MJ of energy-produced basis, the

functional unit used in this study was 1 km of driving of a midsized car. Thus, the system

boundary for analysis was from agricultural production to combustion in a midsized car. The

authors also define four additional functional units which incorporate the energetic values of the

corn grain and corn stover in order to accommodate the allocation methodology comparison

undertaken in the study. The study found that environmental impact results differ greatly when the allocation method is varied.

Spatari et al. (2010) compared different near- and mid-term potential technologies for cellulosic ethanol production in the US. These technologies included dilute acid pretreatment, AFEX,

Simultaneous Sacharrification and Co-Fermentation (SSCF) and Combined Bio-Processing

(CBP). Illustrated with a figure and stated in the text, the system boundary is clearly defined to include feedstock production and handling, pretreatment, hydrolysis and fermentation ethanol recovery, wastewater treatment and lignin separation. This study is unique in that also included enzyme production analysis, which had not previously been factored into many analyses. The research also accounted for the impacts of chemical and nutrient production required for ethanol production and used data from GaBi 4 software databases. The researchers used Aspen simulations to model ethanol production for the future “nth” plant. As a way to allocate and account for the removal of stover from corn fields, this study used nutrient replacement rates from previous studies in terms of the mass of the nutrients (N, P and K) contained in the dry mass of removed corn stover. Another paper is cited as the source of the method to allocate emissions between the harvested corn stover and that which is left on the field. An electricity co- product of ethanol production credit was also calculated by the system expansion method. The functional unit was defined as the volumetric flowrate of ethanol produced in the refinery. The

impact category metrics were also clearly defined in the methods section and included fossil energy input, petroleum input, GHG emissions and air pollutant emissions for different pollutant species and solid waste generation. This article focused on biofuels production and not agricultural activities. The study also evaluated electricity as a co-product from burning the lignin fraction and other biosolids left over from ethanol production.

Table 3.2. Functional units, system boundaries, allocation methods, and impact category metrics in reviewed studies

Study Functional Unit(s) System Boundary (unit Allocation Method(s) Impact category metrics reference processes included)

Curran, 1000 gallons (3785 Corn grain production, ethanol Weight-based, volume- Net Energy, TRACId scores for air 2007 liters) of conventional production, crude oil extraction, based, market value- impacts: acidification, ecotoxicity, gasoline (and the petroleum refining, bulk based, energy-based eutrophication, global warming, energy equivalent terminal storage, fueling and and demand-based human health cancer, human health volumes of gasoline vehicle operation. criteria, human health noncancer, with 8.7% ethanol and ozone depletion, photochemical smog E85) and water impacts: ecotoxicity, eutrophication, human health cancer and human health noncancer

Hill et 1 MJ of ethanol Corn grain productiona,b, ethanol Economic Net Energy Benefit, Net Energy al., 2006 productionc and combustion displacement, mass Benefit Ratio, GHG emissions per balance, energy content NEB and market value

Hill et One gallon or liter of Corn grain productiona,b, GREET default Dollars of human health cost from a al., 2009 ethanol produced (or cellulosic biomass production , methods (mass, energy GHG and PM2.5 emissions per gallon gasoline for the ethanol production, electricity and economic or liter of fuel baseline) and billion generation, gasoline production allocation) ethanol-equivalent and fuel combustion gallons

Laser et Switchgrass feedstock Switchgrass feedstock storage None Process efficiencye, fossil fuel al., 2009 Lower Heating Value and handling, ethanol displacement ratio, percent reduction (LHV) production, utilities production in process power, percent reduction in (by Rankine cycle and gas process steam demand turbine combined cycle 82

coproduction of power and steam) and waste treatment

Liska et One liter of corn-grain Corn grain productiona,b, ethanol Displacement Net Energy Yield, Percent reduction al., 2009 ethanol productionc, feeding of co- of GHG emissions relative to gasoline products to beef animals and life-cycle emissions, Net Energy anaerobic digestion of manure Ratio, GHG intensity, Ethanol yield from beef animals

Luo et 1 kilometer driven on Corn stover productiona, ethanol Mass, economic, Abiotic Depletion, ozone layer al., 2009 E10, E85, pure productionc, transportation of system expansion depletion, photochemical oxidation ethanol or gasoline by ethanol to storage and refinery, (POCP), human & ecotoxicity (HTP & a midsized car and gasoline production, ETP), acidification potential (AP), combined driving and transportation and storage and eutrophication potential (EP), and co-product energy car driving global warming potential (GWP) content

Spatari Conversion facility Feedstock productiona and Nutrient replacement Fossil energy input and petroleum et al., volumetric flowrate of handling, pretreatment, rate for corn stover input, GHG emissions and air 2010 ethanol produced hydrolysis and fermentation removed from the field pollutant emissions for different ethanol recovery, wastewater and cited methods from pollutant species and solid waste treatment and lignin separation another paper, generation and enzyme production electricity co-product credit by system expansion a Feedstock production included the upstream production of farming inputs (e.g. fertilizer, lime, seed, pesticides, electricity and fossil fuels); b Depreciable capital is included.; c Ethanol production for this study included the upstream production of conventional energy sources such as: electricity, natural gas and gasoline; d United States Environmental Protection Agency’s Tool for the Reduction and Assessment of Chemical and other Impact; e Process efficiency is defined as the ratio of product energy out to feedstock energy in, which is similar to the NER used in other studies. 83

3.5. RESULTS AND DISCUSSION

3.5.1. Functional unit

A functional unit defines the relevant quantity of a product or service needed to perform the

function of the product. This unit is used as a reference that relates inputs and outputs and is

necessary for comparing the results of different LCAs (ISO 14040, 2006 and ISO 14044, 2006).

If two studies on the same fuel use different functional units, then the results of the studies are

not directly comparable. All units must be converted to be based on the same functional unit in

order to compare the results. If the functional unit and how it is calculated are not made

completely transparent, then such a unit conversion may not be possible and, therefore, the

results cannot be compared (Baumann and Tillman, 2004). In ethanol LCAs, this unit conversion

can be further complicated if the volumetric and energy densities used in calculations are not

stated clearly.

The choice of the functional unit is not uniform across the seven reviewed studies. Liska et al.

(2009) and Hill et al. (2009) chose the volumetric measurement of one liter or gallon of ethanol

fuel, while each of the other five studies chose a different functional unit. Therefore, we pose the

question: what is the most appropriate functional unit for ethanol production evaluation? The

answer to that question should be found in the answer to this question: what is the function of

ethanol? The function of ethanol varies with context. To some, the function of ethanol is to drive

cars, yet, to others, the function of ethanol is to displace gasoline in the transportation fuel

system and reduce foreign oil imports. Moreover, to other groups the function of ethanol is

simply to be a liquid energy carrier. Unfortunately, these answers do not give rise to the same

functional unit.

Luo et al. (2009) assert that the functional unit should be one kilometer driven on the blended

ethanol fuel (E10 or E85) because this is both the end use (end of the life-cycle) of the fuel and

the function of the fuel (to propel cars). This is an acceptable functional unit because propelling a

vehicle is a function of ethanol, but it makes comparison to other fuels difficult. There are many

different cars on the road in the United States and abroad, and the energy efficiency, GHG

emissions and other emissions from combustion depend on the vehicles’ make, model, year and

maintenance history. Certain vehicles that have been designed to run on gasoline, and are not

optimized for ethanol, will use more ethanol to travel the same distance traveled on gasoline, yet

the difference is not due to the fuel choice, but rather the car chosen. The driven distance by a car

on a particular fuel is also influenced by the weight and the aerodynamic profile of the car. A

heavier car will use more fuel to travel the same distance as a lighter car. Similarly, a car

designed to reduce wind resistance will travel farther on the same fuel than a car with sub- optimal aerodynamic design. There is no one vehicle on the market today that could be tested for distance traveled on gasoline, ethanol (blended or pure), diesel and electricity for controlled comparison of how far the car would travel on each fuel. Without a similar test-car, variables that have nothing to do with the fuel choice will influence the results of the study. Furthermore, from

a renewable-fuels policy standpoint, there is little that an ethanol producer can do to improve the efficiency or emissions of a car. This is one reason why studies such as Liska et al. (2009), Hill et al. (2009) and Laser et al. (2009) do not use a functional unit based on the operation of vehicles.

If a researcher or policy-maker needed to compare LCA results of one study that used the MJ

functional unit to another study that used the km functional unit, he or she would need to find the

vehicle fuel efficiency assumptions used in the km functional unit study. Then, a conversion

from distance traveled to energy or energy to distance traveled would need to be performed. It can be difficult, however, to find these fuel efficiency assumptions. For example, in the Luo et al. (2009) study, the functional unit is 1-km of driving in a midsized car, but the fuel efficiency

(km liter fuel-1) of the assumed mid-sized car is not specified. Therefore, a reader would not be

able to convert metrics of interest to a per MJ form to facilitate comparison with other studies.

Further, if the researcher or policy-maker needed to compare the results of two studies which

used the distance traveled functional unit, he or she would first need to ensure that the same

vehicle specifications were used in both studies, otherwise, the results would not be comparable.

If the distance traveled functional unit is to be the preferred functional unit, then a standardized

set of vehicle characteristics should be proposed. The selection of a general vehicle for this

purpose would be very difficult for several reasons such as the wide range of ages, sizes and

shapes of vehicles in the U.S. automotive fleet. In contrast, if a researcher or policy-maker

needed to convert the results of an LCA conducted with a MJ functional unit into another unit

such as volume or distance, this could be accomplished by using a standard volumetric energy

content of the fuel (Oakridge National Laboratory, 2011) and the fuel-efficiency assumptions of

the vehicle of interest to the research or policy.

The functional unit employed in Laser et al. (2009) was based on the LHV of the switchgrass

feedstock because one of the goals of the study was to determine the potential efficiency of

converting switchgrass to ethanol and electricity as opposed to directly burning the biomass. This

is another example of a functional unit that does not accommodate direct comparison since many transportation fuels, bio-based or conventional, are made from different feedstocks. Curran

(2007) used the unconventional functional unit of 1000 gallons (3785 liters) of conventional gasoline and the energy equivalent volumes of gasoline with 8.7% ethanol and E85. This unit suited her study of the comparison between the conventional gasoline and two commonly- available ethanol-blends of gasoline. Still, it does not allow for the comparison of ethanol production with other transportation fuel alternatives such as biodiesel or electricity. From the viewpoint of an ethanol plant that is using LCA to meet a legal standard or identify areas of process improvement, a more useful functional unit might be the one chosen by Spatari et al.

(2010), which was the volumetric flowrate of the ethanol production facility. These types of functional units help the ethanol company assess the efficiency of their process and hone in on areas for improvement. Even so, the volumetric flowrate functional unit would not facilitate comparison of fuel production types, as it may vary greatly from plant to plant. The unit used by

Hill et al. (2006) was the energy content in the fuel. This unit does allow for uncomplicated comparison to other transportation fuels because all fuels have an energy content which can be measured easily. Furthermore, the energy functional unit (MJ) represents the potential of a fuel to propel a vehicle regardless of what form the fuel takes.

The functional unit chosen depends on the goal of the LCA. If an LCA is used to certify that fuel production at a particular ethanol plant meets a state or federal fuel production standard, then the most useful functional unit must be tied to a physical property of the fuel. Furthermore, a physical property functional unit would also be appropriate if the goal of an ethanol study is to identify areas in ethanol production where improvement would greatly reduce the environmental

impact of the fuel. A fuel physical property functional unit is also the most appropriate for

comparing ethanol production to the production of other transportation fuels. This seven study

comparison of functional units illustrates that functional unit selection is influenced by the

researchers’ perceived function of the fuel, but this selection must be guided by the end-use of the LCA study. Incompatible functional units in transportation fuel production LCAs will make

comparing studies difficult, and, in some cases, not possible. Therefore, one functional unit must

be identified to facilitate fuel production impact comparison.

3.5.2. System boundary

A system boundary defines where the LCA of a production system begins and ends, as well as all

of the relevant production steps included in the assessment (ISO 14040, 2006 and ISO, 14044

2006). Often, elements within the system boundary of analysis can be subdivided into smaller

and clearly identifiable processes that have quantifiable inputs and outputs. These sub-processes

are called unit processes (ISO 14040 2006 and ISO 14044 2006). An LCA analyst chooses which

unit processes to include in or exclude from the system boundary based on the availability of

data on the process, knowledge about employed production technology and the relative impact of

including or excluding the process on the results of the LCA.

An illustrative diagram of a sample system boundary with unit processes for ethanol production

analysis is given in Figure 3.1. The impact of equipment and infrastructure on a biofuel

production LCA is often called “depreciable capital” (represented by octagons in the figure).

This label signifies that the burden of the production of this capital has been amortized over its

useful lifetime rather than realized in the year it was produced. This is a simple system boundary

diagram in that it does not elaborate on the production processes for each input to a unit process.

Some studies do not include the upstream production of inputs such as fossil fuels or seed in the

LCA of a biofuel. Many studies do include these unit processes within the system boundary by modeling them individually in the LCA. Other studies avoid this added modeling by using data that include the “cradle-to-gate” LCA of the production of the input. For simplicity, Figure 3.1. does not depict the upstream production of inputs.

Figure 3.1. Sample system boundary diagram for ethanol production LCA [solid black line encompassing the figure is the system boundary, ovals are material inputs and outputs,

rectangles are production processes, parallelograms are conventional transportation of inputs and outputs by truck, train or ship, large boxes encompassing production processes and materials denote the division of the unit processes within the system boundary, octagons are non-physical impacts to the system, such as the depreciation of equipment and infrastructure (capital)].

The system boundary decision is one of the most difficult, and simultaneously most important, decisions made in conducting an LCA. There are two common points of divergence in the ethanol production system boundary in the reviewed studies. First, there are differing assumptions for how far back the analysis goes into the production of inputs to the system. Two examples of this are the inclusion or exclusion of the production of equipment and infrastructure in the agricultural and ethanol production unit processes and enzyme production for biomass feedstock hydrolysis. Second, there is a lack of consensus on whether the system boundary of a biofuel LCA should end with the biofuel at the refinery gate or with the combustion of the fuel in a vehicle.

The first example of divergence at the starting point of the system boundary is commonly referred to as depreciable capital since, like a financial cost, the impacts of equipment production are amortized over the useful life of the equipment. Liska et al. (2009) and Hill et al. (2006) agree that the depreciable capital energy and GHG emissions associated with the production of farm and biorefinery equipment should be included within the system boundary of an ethanol

LCA. The consideration of the “manufacture, maintenance and decommissioning of capital equipment” is noted in ISO Standard 14040 (2006) an example of an attribute of a production system which should be included in the system boundary. Both include depreciable capital costs because the corn grain feedstock could not be produced if the machinery was not produced.

These two studies disagree, however, on both the data sources and the calculation methods that should be used to estimate these values. Thus, the two studies estimate the contribution very differently. The BESS model was used to conduct the Liska et al. (2009) study, and this model estimates depreciable capital energy costs to be 320 MJ ha-1yr-1. The number used in the Hill et

al. (2006) study was 1,538 MJ ha-1yr-1. A significant portion of this discrepancy stemmed from

different data sources used for steel production and machinery assembly. The other significant

part of this difference came from energy costs for farm-related transportation, personal commutes and farm household energy use included in Hill et al. (2006). The BESS model analysis did not include these energy costs, since such costs are not tied to the amount of corn produced by a farm in the way that fertilizers, pesticides, seed, machinery and diesel fuel are. By

default, the Hill et al. (2009) study employed the depreciable capital assumptions programmed

into the GREET model because this model was used for the analysis. The GREET model’s

assumptions for depreciable capital are very similar to those of the BESS model (Plevin, 2009).

Another study, which compared the results of many ethanol LCAs, noted that the age of the data

used to calculate depreciable capital inputs can also dramatically affect these numbers (Farrell et

al., 2006). With time, steel and machinery manufacturing technologies improve and become

more energy efficient; therefore, forty-year-old data may not accurately reflect the production of

ten- or twenty-year-old equipment.

Other studies omit depreciable capital entirely. A recent LCA study on the production of corn

grain and corn stover (not ethanol production) asserts that such inputs are relatively small

compared to other inputs such as fertilizer and fuel (Kim et al., 2009). Luo et al. (2009), Spatari

et al. (2010) and Curran (2007) do not mention the inclusion of depreciable capital energy or

GHG emissions embodied in farm equipment, but they also do not specify whether these factors

should be included. Laser et al. (2009) is the only study that did not include the agricultural

production of the feedstock used in the production of ethanol and energy; therefore, the on-farm

depreciable capital costs are not captured. It can be argued that studies of the biorefinery alone

are useful for comparing the efficiency of different ethanol production technologies and biological pathways, and such information could guide future research in biorefining technology development. Variation in the choices of unit processes included in the system boundary,

however, makes the comparison of ethanol production studies difficult.

Another interesting system boundary input production difference between the reviewed studies is

the inclusion or exclusion of enzyme production. Before fermentation, both corn-grain and

cellulosic ethanol production require enzymes to break down starch or cellulose into sugars in a

process called hydrolysis (Brown, 2003). Spatari et al. (2010) included enzyme production while

no other study explicitly stated the inclusion of the production of this input unless it was

produced in the ethanol facility. Enzymes can be produced in the ethanol facility, but most

enzymes are purchased from a supplier. In Laser et al. (2009), the base case scenario included

the production of enzymes at the ethanol plant and the mature case scenario assumed the

Combined Bioprocessing (CBP) unit produced enzymes in the tank where the hydrolysis occurs.

This enzyme production assumption was included to illustrate one of the benefits of developing

the CBP technology, but, at this time, this technology does not exist for commercial scale use.

Luo et al. (2009) also assumed that enzyme production occurs at the ethanol plant, and accounted

for the electricity required to produce enzymes. If the enzyme production occurs in the ethanol

plant, then it is absolutely within the system boundary of ethanol production. Nevertheless,

enzymes are an input to the ethanol production system, and so their production should be

included as well. MacLean and Spatari (2009) found that, in the case of conventional corn-grain

ethanol production, enzyme production did not contribute significantly to overall GHG

emissions, but the opposite was true of cellulosic ethanol production. Even with the evidence

presented in the MacLean and Spatari (2009) study, enzyme production is not uniformly

included in all cellulosic ethanol production LCAs because of decisions made by researchers to

not include it. Such disparities in boundary decisions complicate the comparison of biofuel

production studies.

The end point of the analysis is another non-trivial system boundary decision. The two most common ending points of an ethanol LCA are the denatured ethanol product at the biorefinery gate and the combustion of the final blended ethanol product in a vehicle. The latter boundary requires the study to incorporate assumptions about the type of vehicle engine that burns the fuel.

As discussed earlier, this choice of vehicle can affect the energy efficiency and emissions results

of the study. As noted in the functional unit section, studies such as Curran (2007), Luo et al.

(2009) and Hill et al. (2006 and 2009) include combustion in some way. Laser et al. (2009),

Liska et al. (2009) and Spatari et al. (2010) do not include combustion of ethanol, though Laser

et al. does include biomass combustion for electricity generation. If the LCA aims to evaluate

only the net effects of the production of a particular type of ethanol, then the system boundary

generally ends at the biorefinery gate. Such an analysis is often called a Well-to-Gate (WTG)

analysis as opposed to a Well-to-Wheels (WTW) analysis, which includes the combustion of the

fuel in one or many vehicles (Wang, 2007). Both WTG and WTW analyses are perfectly valid,

but cannot be compared easily. If the goal of the analysis is to determine the life-cycle emissions

of criteria pollutants such as particulate matter (Hill et al., 2009) or other human health and

environmental factors (Curran, 2007 and Luo et al., 2009), then combustion must be included in

the system boundary because the combustion of ethanol releases compounds that affect these

categories.

One very common use of biofuel LCA is to determine the net energy balance of fuel production.

When an ethanol LCA evaluates the net energy balance, the analysis should not include combustion since the energy released during combustion depends on the energy content of the fuel, not the net energy of the fuel’s production. That is to say that if two fuels have the same energy content, but different net energy balances, the differing net energy balances will have no effect on the energy released during combustion. Furthermore, there are ways of communicating a fuel’s potential to move a vehicle that allow many fuel production systems to be compared without making assumptions about the type of vehicle used. For example, the LHV of ethanol and other liquid transportation fuels are not affected by the raw material (agricultural or fossil) from which it was made. Ethanol fuel made from corn grain will have the same LHV as ethanol made from switchgrass, corn stover, wood or any other biomass. The net energy balances of these ethanol fuels will change with the choice of feedstock, but not their energy content.

Another common use of biofuel LCA is to evaluate air quality effects. If the goal of the study is to examine whether the fuel production methods meet air quality standards for a geographical region (such as an air pollution non-attainment zone), then it is not appropriate to include fuel combustion. Combustion could occur outside of the location in which the ethanol is produced, and determining the geographic location of combustion can be very difficult. Researchers must pay careful attention to the system boundaries of different studies so as not to compare similar results, such as particulate emissions per volume of ethanol, from a study that included combustion to one which did not. Overall, disparities in system boundary choices, such as the two mentioned in this section, can compromise the comparison of different ethanol production studies.

3.5.3. Allocation

When a production system produces more than one product or uses an input whose production

also resulted in the production of other products, allocation is used to partition and assign the

relevant environmental burdens to the various products (Guinée, 2002). In ethanol production, it

is rare that ethanol is the only product. Animal feed, heat, power and other economically

valuable products are often produced in addition to ethanol and are referred to as ethanol co- products because they have value and are not waste streams. These products not only have economic value, but also have energetic and GHG emissions values because they can displace the production of other similar products in the global marketplace. For example, if electricity is a co-product of ethanol production, and this electricity is exported to the general grid, it can be argued that electricity from other sources (such as coal or natural gas) is displaced. Allocation is one method used to account for these impacts. Examples of allocated inputs and outputs are shown in Figure 3.2.

Figure 3.2. Ethanol production examples of allocated inputs (A) the impacts of corn production must be divided between the two products (corn grain and stover) that result from corn production and are inputs to various production systems (ethanol and animal).

The potential allocated outputs (B) of ethanol production include ethanol, animal feed and electricity

According to the ISO Standards 14040 and 14044 (2006), allocation should be avoided whenever possible. Two methods for avoiding allocation are sub-division (SD) and system expansion (SE).

The SD method removes the processes exclusively associated with the production of a co- product from the production of the primary product on which the LCA focuses (Ekval and

Finnveden, 2001, Kaufman et al., 2010 and Kraatz et al., 2011). In the case of corn-grain ethanol, distillers grains (DG) is one potential co-product. The SD method would remove from the corn ethanol LCA any processes carried out for only the production of DG, such as drying the DG to

reduce moisture and improve storage. The SE method involves identifying which products

outside of the study-system (ethanol production) are affected by the co-products. The DG

primarily displaces soybean meal and corn grain in the diets of livestock. The LCA of these

displaced products is then incorporated into the primary LCA study to evaluate the relative

energy-usage, GHG-emissions or other relevant impacts of the displaced products (Kaufman et al., 2010; Luo et al., 2009; Sinistore and Bland, 2010 and Spatari et al., 2010). Alternatively, a literature source can be cited for these values. Finally, all of the avoided energy-use, GHG emissions or other environmental burdens associated with the production of the displaced products are credited back to the primary system according to the amount which is displaced. In the corn-grain ethanol example, the LCA energy-use and GHG emissions for the production of corn grain and soybean meal are credited to the ethanol production system according to the amount of DG fed to livestock. It is often the case that neither the SD or SE methods of allocation avoidance can be applied. The SD method cannot be used if the co-product and the

primary product share all of the unit processes of production. The SE method may be rendered

untenable if existing data on the life cycle impacts of the displaced products is insufficient or non-existent.

When allocation is unavoidable, the ISO standards recommend using a ratio based on the mass, energy or economic value of the co-product to calculate a relevant energy, GHG or other impact burdens or credits to the primary system. For example, an economic ratio of the value of the DG to value of the corn-grain ethanol multiplied by the total life cycle GHG emissions could be used to credit some avoided GHG emissions back to the ethanol for the production of DG. In addition to crediting co-products, it is also often necessary to divide inputs between two or more

feedstocks. For example, the agricultural production of corn produces corn grain and stover

(Figure 3.2.A.). Corn grain can be used for feed or ethanol, and corn stover can be used for bedding or ethanol. The relative percentages (by mass, energy or economic ratio) of each product that goes into each process could be used as an allocation ratio. This ratio is then used to determine the distribution of agricultural-input environmental-burdens to the final products

(ethanol, feed or bedding). Some obvious choices can be made between mass, energy and economic ratios since some co-products have no mass (such as electricity) and some co-products are not generally measured in terms of their energy content (bedding). Even with these clearly defined methods from the ISO standards, there is great discrepancy in the application of allocation methodology in the literature.

Allocation methodology appears to be the least agreed upon component of LCA in the seven studies. Curran (2007) notes that previous studies on allocation methods “fail to identify a single, logically defensible approach for allocation”. An example of this arises in Luo et al. (2009) in which the authors claim to evaluate mass, energy and economic allocation in addition to a system expansion. However, the authors assert that the mass ratio (mass of corn grain to corn stover) and the energy ratio (energy in corn grain and corn stover) are equal. They therefore apply the same value for both and report the results of this method as just mass allocation. Luo et al. (2009) make this assumption based on methods used in another study by different authors. According to many other studies, however, the ratios of mass and energy for corn grain and corn stover are not the same (Penn State University Cooperative Extension, 2009; Pordesimo et al., 2004 and

USDOE, 2009). This is a very influential assumption since many studies also note that the choice of allocation ratio can change the LCA results significantly (Curran, 2007; Farrell et al., 2006

and Kaufman et al., 2010). There appears to be some consensus around the idea that all allocation methods should be evaluated and then compared for internal consistency within the

LCA in the literature. Allocation, however, has been called one of the most controversial issues

in LCA and one of the “grand challenges” facing LCA analysts today since the methods used can

dramatically alter LCA results (Reap et al., 2008 and McKone et al., 2011).

Most of the seven studies (Laser et al., 2009; Liska et al, 2009; Luo et al., 2009 and Spatari et al.,

2010) did not compare all available methods, and none of the studies use the SD method. Both

Hill et al. studies (2006 and 2009) included more than one method. Hill et al. (2006) used

economic displacement, mass balance, energy content and market value. The economic

displacement method is a form of SE in that the authors calculated “the energy required to

generate the products for which each [co-product] serves as a substitute in the marketplace (i.e.,

corn and soybean meal for DDGS and synthetic glycerol for soybean-derived glycerol)”. In this

way, the authors expand their system boundary to calculate the energetic cost of producing

products, like corn and soybean meal, which are displaced by a corn ethanol co-product (DDGS)

in the market place. All of the calculations for these allocation methods are presented in the

supporting materials, but their individual effects on the biofuel analysis results are not compared

in the article text. Hill et al. (2009) employed the GREET model for its calculations, and this

model’s default allocation methods are mass, energy and economic ratios. The 2009 article also

uses the word allocation to talk about the distribution of particulate emission over spatial areas,

but this is not the same as the allocation methods outlined in the ISO standards. The results of the

2009 study are not presented for different non-spatial allocation methods. Therefore, it is

difficult to compare the results of this study to studies which employ allocation strategies.

100

Laser et al. (2009) avoided allocation by focusing the study on the biorefinery and not the

upstream switchgrass production and by keeping the electricity co-product inside the system

boundary. The electricity is assumed to be consumed by the facility to reduce the overall external input of electricity. Any excess electricity is sold to the grid. According to LCA methodology, once the electricity leaves the ethanol system boundary and enters the electricity grid, some analytical method, such as allocation, should be applied. In terms of the metrics examined in this study, which are all focused around energy use, this electricity export is simply subtracted. It could be argued that, since it took energy to make that electricity, the total energy exported cannot simply be deducted, but instead should be allocated according to its net energy content.

Liska et al. (2009) developed a unique method for dealing with corn-grain ethanol production co-

products that they call “displacement”, but this method is very similar to system expansion. They

expanded the system boundary to account for the production of corn grain, soybean meal and

urea for cattle feed. They then displaced these three feeds in a conventional beef-cattle diet with

DGS and credited the energy and GHG emissions required to produce the three feeds back to the ethanol system. This level of detail arguably produces a more accurate co-product credit than a simple mass, energy or economic credit, since the co-product credit is based on the function of the DGS. The authors also developed several “closed-loop” scenarios which included animal feeding and manure bio-digestion to prevent the DGS from leaving the system boundary. Still, in reality, it can be difficult to track which kind of animal is consuming what quantity of the DGS and accurately estimate the displacement of other feeds. This becomes even more cumbersome if many ethanol plants, such as all of the plants in the United States, are included in the analysis.

Therefore, this method is more suitable for an LCA of individual plants.

101

Spatari et al. (2010) used various system expansion methods to deal with different points in the

analysis where allocation was needed. At the point when the burdens of corn production needed

to be divided between the corn grain and stover, nutrient replacement rates were used to account

for the nitrogen, phosphorus and potassium removed from the field with the corn stover. These

nutrients had to be compensated for with the application of synthetic fertilizers to ensure

adequate corn yields in the next crop. System expansion was used again to account for the

electricity co-product from burning the lignin co-product of corn-stover ethanol production.

These are both completely valid methods that achieve the goal of avoiding allocation set forth in the ISO standards. It is difficult, however, to compare the results of this study with one that used

the same functional unit, system boundaries and other LCA components, but did not use system

expansion because no other allocation method was considered in Spatari et al. (2010).

If all ethanol studies were to examine the effects of all allocation methods and present the

variation in results due to use of these different methods, then results could be more directly

comparable. It may be a laborious task, but all biofuel studies should attempt to avoid allocation by using SD or SE or, if allocation avoidance is not possible, they should include all common allocation methods to ensure that results are comparable across studies.

3.5.4. Impact category metrics

Impact category metrics are the units used to quantify environmental impacts and represent results. Often, when these metrics are presented in simple units, such as total MJ of energy or total kg of GHG emissions per functional unit, they are too abstruse to convey meaningful information about the system (e.g. when results are presented per driven km with no information

102

on the fuel efficiency of the assumed vehicle ). Biofuel studies have developed some unique metrics for illustrating the benefits and detriments of biofuel production. Most biofuel studies focus on metrics such as net energy balance, fossil energy balance and GHG intensity. The net energy balance compares the total output of energy to the total input energy in the entire production process. Fossil energy balance is different from net energy balance in that it compares the total amount of fossil energy required for fuel production (from coal, natural gas and petroleum) to the amount of energy in the fuel. This is a particularly important distinction for cellulosic biofuels since the lignin fraction of the biomass can be burned to provide process heat and power to the ethanol production. This lignin heat and power reduces the use of fossil energy in the production system. Greenhouse Gas intensity expresses the total emission of GHGs, in mass of CO2eq, per unit of energy in the fuel. This can be a highly informative metric that is

useful for comparing biofuels to fossil fuels. It is often unclear; however, if the unit of energy

used to calculate GHG intensity is based on the net energy of producing the fuel or the energy

content of the fuel.

Many studies used similar impact category metrics, but gave them different names. For example,

the Net Energy Ratio (NER) is the total energy output of the fuel (plus allocated co-product

energy) divided by the total energy input to make the fuel (Liska et al., 2009). This is the same

definition given in Hill et al. (2006) for their Net Energy Benefit (NEB) Ratio. The unitless NER

or NEB Ratio is the number often cited in discussion of the energy gains or losses from ethanol

production. Laser et al. (2009) presented a metric similar to NER called “process efficiency”,

which was defined as the ratio of product energy out to feedstock energy. Another metric used in

Hill et al. (2006) was the NEB (not the ratio) which is calculated as the energy output by the

103

biofuel and its co-products minus the energy input to make the biofuel. A 2008 study on life-

cycle metrics for biofuels noted that this is the same calculation for the Net Energy Value (NEV)

of a biofuel (Liska and Cassman, 2008). Liska et al. (2009) put forth yet another interesting

energy-related metric called the Net Energy Yield (NEY) which was defined as the net energy output of the biofuel per acre of land it took to produce the feedstock for the ethanol. This metric

allows the comparison of the total agricultural land-use impact of producing various biofuels under various agricultural conditions. Liska and Cassman (2008) found that twelve out of nineteen studies on biofuels include an NER-like calculation as their energy impact category metric. Four of the nineteen studies used the energy per mass of feedstock metric while two used the energy per land area metric. Only one study reported energy per driven distance in a vehicle.

Liska and Cassman concluded that one of major issues in biofuel LCA is the standardization of impact category metrics for the ease of comparison of studies.

Greenhouse gas metrics show just as much variation in the literature as energy metrics. As set forth the by the Intergovernmental Panel on Climate Change (IPCC), GHG emissions can be measured in grams of carbon dioxide equivalents to account for the three major greenhouse gases

(carbon dioxide, methane and nitrous oxide), which each have different global warming potentials (GWPs). To consolidate these GWPs into one number, the masses of methane and nitrous oxide emissions are converted into GWP equivalent amount of carbon dioxide (IPCC

2006). Most studies do not report the total GWP as a raw number. Instead, studies such as Liska et al. (2009) reported GHG intensity (the mass of carbon dioxide equivalents per unit of energy in the biofuel). This study, and other studies, also reported the life-cycle emissions of ethanol production in terms of the relative percent reduction in GHG emission compared to an energy

104

equivalent amount of gasoline. For example, Liska et al. (2009) reported that the corn ethanol

production systems modeled in their study produced a 48% to 59% reduction in GHG emissions

-1 compared to gasoline (if gasoline production is assumed to produce 94 gCO2eq MJ ). Spatari et

al. (2010), however, did not report this GHG metric. Instead, they reported the gCO2eq per liter

of fuel. Both metrics are valid, but conversions must be made in order to compare the results of

both studies.

Among the nineteen reviewed articles in Liska and Cassman (2008), there is less agreement on

GHG metrics than there is on energy metrics. Out of the nineteen studies, only four employ the

GHG intensity metric, while six use a mass of GHG emissions per mass of feedstock metric.

Four studies reported the mass of GHG emissions per area of agricultural land, and another four reported the mass of GHG emissions per distance driven using a biofuel. Again, all of these

metrics are valid for the evaluation of biofuels, but they are not directly comparable. The process

of converting one metric into another becomes tenuous if energy and volumetric density

assumptions for the biofuel are not clearly stated, as these values can vary with temperature and pressure. Moreover, even if the metrics are presented with the same units and assumptions are

stated, system boundary and allocation decisions are bound to have affected the results. The

standardization and agreement on GHG in biofuel LCA is essential to the development and

implementation of GHG emission reduction policies.

While net energy usage and GHG emissions have been the focus of the majority of biofuel

studies, there is burgeoning interest in the presentation of other metrics. Hill et al. (2006)

mentioned the emissions of particulate matter, sulfur oxides, nitrogen oxides, volatile organic

105

compounds and carbon monoxide, and cited studies that evaluated the emissions of these pollutants from the combustion of E85 versus gasoline. Hill et al. (2006) did not, however, evaluate these other emissions in the published study. Hill et al. (2009) did evaluate the emissions of particulate matter (2.5μm diameter), or PM2.5, from the production and combustion

of biofuels as they compare to gasoline. Curran (2007), Luo et al. (2009) and Spatari et al. (2010)

all mention the importance of including metrics that shed light on other environmental impacts,

and they all evaluated such impact categories. Future biofuel LCAs will need to consider

environmental impacts other than energy-use and GHG emissions to effectively evaluate

biofuels.

3.6. CONCLUSIONS

The use of many different functional units, system boundaries, allocation methods and impact

category metrics in ethanol production studies causes confusion in the interpretation and

comparison of the results of such studies. This review touched on only four key components, but

there are several other methodologies that can differ in biofuel production assessments. For

example, each study used a different computer model or user-created Excel spreadsheet to

calculate the environmental impacts. In addition, while there was some overlap in data sources,

the relative comparability of studies based on data sources could be the topic of a paper in and of

itself.

Not all of the reviewed studies claim to be LCAs, and perhaps, due to the lack of clearly defined

key LCA components, some of them are not LCAs. Nevertheless, they all evaluate ethanol fuel

production for some of its environmental impacts, and many have been cited for their

106

environmental analysis of ethanol fuel production. We maintain that LCA should be the tool for this sort of evaluation because it has consistent and well-defined methods that are internationally accepted. Furthermore, the structure of LCA ensures the ease of comparison of studies as long as they employ similar key LCA components. Even studies that explore the impacts of methodological choices on the results of an LCA need to employ consistent functional units, system boundaries, allocation methods and impact category metrics in order to be comparable.

A functional unit based on a physical property is the most appropriate functional unit for biofuel

LCA. We find that the most appropriate functional unit for an ethanol production LCA is energy content (MJ) because this unit facilitates comparisons between biofuels and other transportation fuels without introducing confounding variables tied to the design of the car (e.g. weight, aerodynamics, make, model, engine type and maintenance history). Further, the energy functional unit allows a researcher or policy-maker to easily convert the results of a fuel study to the functional unit needed for the research or policy application. Electric vehicles are emerging as a serious alternative to liquid transportation fuel, but electricity has no mass or volume. An energy functional unit would facilitate the comparison of liquid fuel production to electricity production for the vehicle propulsion.

If the system boundaries of two biofuel production studies are not the same, the results are not directly comparable. The clear statement of an analysis system boundary must include a list or diagram of all included unit processes. We assert that biofuel LCAs should include the production of all necessary materials, such as enzymes, within the system boundary.

Furthermore, we recommend that the environmental burdens of equipment and infrastructure be

107

included in the system boundary as depreciable capital as a way of including the environmental impacts of capital equipment as per ISO 14040. Whether or not the ethanol combustion unit process is included in the ethanol production system boundary is a major different among the seven studies. To ensure clear comparison of the environmental impacts difference of fuel production from many fossil and non-fossil sources, we recommend that the system boundary end at the refinery gate. Assumptions about the average fleet vehicle on the road today (model, age, maintenance and engine efficiency), which are required if the fuel combustion unit process is to be included, make biofuel study comparison difficult, and it introduces unnecessary variability in the results.

It has been shown that the allocation method employed in a study weighs heavily on the results of a biofuel LCA. Therefore, we affirm that sub-division and system expansion should be attempted to avoid allocation by ratios whenever possible. If avoidance is not possible, then we recommend that all three major allocation methods (mass, energy and economic ratios) be applied, compared for internal consistency within the study and presented in the results section of the study. The presentation of all allocation methods will aid in the comparison of biofuel production LCAs from a variety of feedstocks.

There is also a need for consistent impact category metrics among studies. The most common impact categories are energy use and GHG emissions. We conclude that the most comparable impact category metrics are NER, NEY and GHG intensity. We recommend that future biofuel

LCAs evaluate additional environmental impact categories such as water use, other emissions to air, water and land, biodiversity and human health impacts whenever possible.

108

The ISO 14040 series standards call for transparency as defined as “open, comprehensive and

understandable presentation of information” (ISO 14040, 2006). It is interesting to note that this

standard was published in 2006 and each of the reviewed studies was published in or after 2006.

Still, our descriptions of each of the seven studies show that essential information is difficult to

find in the articles, or it is found outside of the articles in other papers and models. Furthermore,

the presentation of this information is often confusing. We aver that this methodological

comparison of seven recent ethanol production studies illustrates that presentation of key

information is still a problem. With respect to lack of transparency, the two main problems we

found in these studies were that it is often difficult to find the four identified key LCA

components in the studies and the reader can easily misinterpret the definitions of these

components from the text. We therefore recommend that biofuel production assessment articles state the functional units, system boundaries, allocation methods (and any formulas used to calculate allocation) and impact category metrics clearly and directly in the methods section and provide a diagram of the system boundary. This is similar to the Liska et al. (2008) finding that there exists a need for future studies to be very clear about systems boundaries and metrics.

Table 3.3. illustrates the type of information which is critical to the interpretation of a biofuel

LCA. The inclusion of such information in an easy-to-read table in a paper’s abstract or methods section would make determining if the results of several studies are potentially comparable based on the congruity methods employed fast and simple for researchers and policy-makers. We acknowledge that the presentation of this information in a table is simple, but we believe that it is more likely to be adopted by authors and used by readers because it is simple and straight- forward.

109

Table 3.3. Example of a table to be included with the abstract of a biofuel production LCA

Primary product Corn grain Functional Energy content, MJ ethanol Unit

Co-product(s) Distillers System Cradle to Gate: Input material Grains Boundary: production (fertilizer, pesticide, fossil included unit fuels, electricity, yeast, enzymes, Allocation Sub-division, processes chemicals), depreciable capital, avoidance system transportation, agricultural production, method(s) expansion ethanol production (liquefaction, hydrolysis, fermentation, distillation, Allocation Mass ratio, dehydration and denaturing) method(s) energy ratio, economic ratio Impact Net Energy Ratio: energy output category energy input-1 metric(s) GHG Intensity: net mass CO2eq MJ of fuel-1

PM2.5 emissions: mass PM2.5 MJ of fuel-1

The implementation of the methodological consistency discussed herein will facilitate proper

comparisons of studies. There is a wealth of information on the environmental impacts of bio and

conventional transportation fuel production in the literature, but very little of it is easily

comparable. The comparability of fuel production environmental impact studies is essential if we are to make clear decisions about the environmental sustainability of fuel production.

110

Acknowledgements

This work was funded by the Department Of Energy Great Lakes Bioenergy Research Center

(DOE BER Office of Science DE-FC02-07ER64494 and DOE OBP Office of Energy Efficiency and Renewable Energy DE-AC05-76RL01830).

3.7. REFERENCES

Beer, T. and Grant, T. 2007. Life-cycle analysis of emissions from fuel ethanol and blends in Australian heavy and light vehicles. Journal of Cleaner Production. 15: 833-837.

Baumann, H. and Tillman, A. 2004. The Hitch Hiker’s Guide to LCA: An orientation in life cycle assessment methodology and application. Lund, Sweden. Studentlitteratur AB. Brown, R. C. 2003. Biorenewable Resources: Engineering New Products from Agriculture. Ames, IA: Blackwell Publishing Professional. Curran, M. A. 2007. Studying the Effect on System Preference by Varying Coproduct Allocation in Creating Life-Cycle Inventory. Environmental Science & Technology. 41 (20): 7145- 7151. Ekvall, T. and Finnveden, G. 2001. Allocation in ISO 14041—a critical review. Journal of Cleaner Production. 9: 197-208. Farrell, A. E., Plevin, R. J., Turner, B. T., Jones, A. D., O'Hare, M., and Kammen, D. M. 2006. Ethanol Can Contribute to Energy and Environmental Goals. Science. 311(5760): 506- 508. Fu, G. Z., Chan, A. W. and Minns, D. E. 2003. Life Cycle Assessment of Bio-ethanol Derived from Cellulose. International Journal of Life Cycle Assessment. 8(3): 137-141. Guinée, J. B. 2002. Handbook on Life Cycle Assessment. Dordrecht, The Netherlands: Kluwer Academic Publishers.

111

Hill, J., Nelson, E., Tilman, D., Polasky, S. and Tiffany, D. 2006. Environmental, economic, and energetic costs and benefits of biodiesel and ethanol biofuels. Proceedings of the National Academy of Sciences. 103: 11206-11210. Hill, J., Polasky, S., Nelson, E., Tilman. D, Huo, H., Ludwig, L., Neumann, J., Zheng, H. and Bonta, D. 2009. Climate change and health costs of air emissions from biofuels and gasoline. Proceedings of the National Academy of Sciences. 106(6): 2077-2082.

Intergovernmental Panel on Climate Change (IPCC). 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge.

International Organization for Standardization (ISO). 2006a. ISO 14040: Environmental management – Life cycle assessment – Principles and framework. Geneva, Switzerland. International Organization for Standardization (ISO). 2006b. ISO 14044: Environmental Management – Life cycle assessment, Life cycle impact assessment. Geneva, Switzerland. Kaufman, A. S., Meier, P. J., Sinistore, J. C., and Reinemann, D. J. 2010. Applying life-cycle assessment to low carbon fuel standards – How allocation choices influence carbon intensity for renewable transportation fuels. Energy Policy. 38: 5229-5241. Kim, S., Dale, B. E. and Jenkins, R. 2009. Life cycle assessment of corn grain and corn stover in the United States. International Journal of Life Cycle Assessment. 14: 160-174.

Kraatz, S. Sinistore, J.C. and Reinemann, D.J. 2011. Methods for Solving the Multifunctionality

Problem in Life Cycle Assessment of Ethanol Production from Corn Grain. Proceedings of the American Society of Agricultural and Biological Engineers Annual International Meeting. Louisville, KY, August 7-10. 1110828.

Laser, M., Haimin, J., Jayawardhana, K. and Lynd, L.R. 2009. Coproduction of ethanol and power from switchgrass. Biofuels, Bioproducts and Biorefining. 3: 195-218.

112

Liska, A.J. and Cassman, K.G. 2008. Towards standardization of life-cycle metrics for biofuels: greenhouse gas emissions mitigation and net energy yield. Journal of Biobased Materials and Bioenergy. 2:187-203. Liska, A.J., Yang, H.S., Bremer, V.R., Klopfenstein, T.J., Walters, D.T., Erickson, G.E., Cassman, K.G. 2009. Improvements in Life Cycle Energy Efficiency and Greenhouse Gas Emissions of Corn-Ethanol. Journal of Industrial Ecology. 13(1): 58-74.

Luo, L., van der Voet, E., Huppes, G. and Udo de Haes, H.A. 2009. Allocation issues in LCA methodology: a case study of corn stover-based fuel ethanol. International Journal of Life Cycle Assessment. 14: 529-539. MacLean, H.L., and Spatari, S. 2009. The contribution of enzymes and process chemicals to the life cycle of ethanol. Environmental Resource Letters. 4: 14001-14010. McKone, T.E., Nazaroff, W.W., Berck, P., Auffhammer, M., Lipman, T., Torn, M.S., Masanet, E., Lobscheid, A., Santero, N., Mishra, U., Barrett, A., Bomberg, M., Fingerman, K., Scown, C., Strogen, B. and Horvath, A. 2011. Grand Challenges for Life-Cycle Assessment of Biofuels. Environmental Science and Technology. 45: 1751-1756.

Oakridge National Laboratory (ONL). 2011. Transportation Energy Data Book, Appendix B, Table 4 Heat Content for Various Fuels. United States Department of Energy. http://cta.ornl.gov/data/index.shtml. Accessed 15 January 2012.

Penn State University Cooperative Extension. 2009. Coping with high energy prices heat energy content of shelled corn. State College, PA. http://energy.cas.psu.edu/energycontent.html.

Accessed: 26 March 2011.

Plevin, R.J. 2009. Modeling Corn Ethanol and Climate: A Critical Comparison of the BESS and GREET Models. Journal of Industrial Ecology. 13(4): 495-507.

Pordesimo, L., Edens, W. and Sokhansanj, S. 2004. Distribution of aboveground biomass in corn stover. Biomass and Bioenergy. 26: 337-343.

113

Reap, J., Roman, F., Duncan, S. and Bras, B. 2008. A survey of unresolved problems in life cycle assessment; Part 1: goal and scope and inventory analysis. International Journal of Life Cycle Assessment. 13: 290-300. Sinistore, J. C. and Bland, W. L. 2010. Life-cycle analysis of corn ethanol production in the Wisconsin context. Biological Engineering. 2(3): 147-163. Spatari, S., Zhang, Y. and MacLean, H. L. 2005. Life Cycle Assessment of Switchgrass- and Corn Stover-Derived Ethanol-Fueled Automobiles. Environmental Science & Technology. 39 (24): 9750-9758. Spatari, S., Batley, D. M. and MacLean, H. L. 2010. Life cycle evaluation of emerging lignocellulosic ethanol conversion technologies. Bioresource Technology. 101: 654-667.

U.S. Department of Energy (DOE). 2009. Energy Efficiency and Renewable Energy, Biomass Feedstock Composition and Property Database. http://www1.eere.energy.gov/biomass/feedstock_databases.htmlS. Accessed: 20 March 2010. von Blottnitz, H. and Curran, M. A. 2007. A review of assessments conducted on bio-ethanol as a transportation fuel from a net energy, greenhouse gas and environmental life cycle perspective. Journal of Cleaner Production. 15: 607-619.

sustainable energy and agriculture. Sterling, V.A.: Earthscan.

114

Supplemental Table 3.1. Location of LCA component definitions in the seven studies.

Study Functional System Boundary Allocation Impact reference Unit(s) (unit processes Method(s) category included) metrics

Curran, Approach Figures 3 and 4, pg Approach Section: Approach 2007 section: pg 7147a pg 7146 and Section: pg 7146a Methodology for 7146a Five Allocation Schemes section pg 7146a

Hill et Figure 1, pg Methods: Energy Use Methods: Energy Results: Net al., 2006 11207b in Crop Production Yield from Biofuel Energy Balance and Energy Use in Production section, (NEB) section, Converting Crops to pg 11209a pg 11206 and Biofuels section, pg Life-Cycle 11209a Environmental Effects section, pg 11207b

Hill et Abstract: pg Methods: List of Methods: Fuel Abstract pg al., 2009 2077 and Fuel Life-Cycle Production and Use 2077 and

Figure 2, pg Emissions section, pg section, pg 2081a Figure 2 pg 2079b 2082b 2079b

Laser et Table 1, pg Process Description None Mass and al., 2009 196 and section, pg 197, energy Table 7, pg Figure 1, pg 198 and Balances 207b Figure 2, pg 199b Results summary

115

section, pg 205, Table 7, pg 207, Table 8, pg 208b

Liska et BESS Introduction: LCA of Introduction: LCA Abstract, pg 58, al., 2009 User’s Corn-Ethanol of Corn-Ethanol Methodology Guide (the Systems section, pg Systems section, pg section: GHG instructions 61 and BESS User’s 60 and Results and Emissions for the Guide, pg 19a Discussion: Factors section, modeled Coproduct Energy pg 64, Figure 2, used in this Credits and Impact pg 66, Figure 3, analysis) pg on GHG Emissions pg 68, Table 2, 11-12b section, pg 67a pg 69b

Luo et Methodolog Methodology: section Abstract: Methodology: al., 2009 y: section 2.2 System Discussion section, section 2.6 2.1 Boundary, pg 531 pg 529, Impact Functional and Figure 1, pg 532a Methodology: assessment and unit and section 2.5 interpretation, alternatives, Allocation pg 534 and pg 531, Methodology, pgs Figure 2, pg a

section 2.5 533-534 and Table 535 Allocation 1, pgs 533-534a methodolog y, pg 533 and Table 2, pg 534a

Spatari Methods: Figure 1, pg 657a Methods: section Methods, pg

116

et al., section 2.4 2.1 Feedstock 655b 2010 LCA model production, equations, collection and pg 659a transport (external source cited) pg 656 and Table 1, pg 655a

a The terms functional unit, system boundary (or unit processes), allocation method(s) (or allocated inputs/outputs) or impact category were used

b The terms functional unit, system boundary (or unit processes), allocation method(s) (or allocated inputs/outputs) or impact category were not used

117

CHAPTER 4

Part 1: Life Cycle Assessment of Switchgrass Cellulosic Ethanol Production in the Wisconsin and Michigan Agricultural Contexts: Testing LCA results for sensitivity to spatial variability and allocation decisions

Part 2: Life Cycle Assessment of Corn Stover Cellulosic Ethanol Production in Wisconsin and Michigan Agricultural Contexts: The interplay of spatial variability and allocation decisions

4.1. PUBLICATION AND CONTRIBUTION DETAILS

This chapter describes the background, justification, methods, results and conclusions of a complex life cycle assessment of cellulosic ethanol production in 104 different scenarios. Eight scenarios focus on the production of ethanol from switchgrass and 96 focus on the production of ethanol from corn stover. This chapter is broken into two parts to allow for adequate description and discussion of these two very disparate production systems. The first part was written with the help of Doug Reinemann (UW-Madison) and César Izaurralde (Pacific Northwest National

Laboratory (PNNL) and the University of Maryland-College Park) as a paper targeted for publication in the journal Biofuel, Bioproducts and Biorefining. This journal has a specific section for papers on modeling and analysis which present and discuss modeling issues that affect, among other things, the biofuels sector. The analysis on switchgrass was deemed to be sufficient to illustrate that accounting for spatial variability in LCA modeling is possible and can influence the LCA results significantly. The second part of this chapter adds the influence of allocation choices to spatial variability in a way that builds upon the findings of chapter 2. César

Izaurralde and his colleagues at PNNL and Oakridge National Labs contributed the spatially- modeled data from EPIC on switchgrass and corn stover production. Paul Meier contributed the region-specific electricity grid data and Bryan Bals and other members of Bruce Dale’s lab at

Michigan State University contributed the AFEX pretreatment data required for this analysis. My 118

contribution to this research included all of the Life Cycle Inventory data collection, LCA model

development and construction, data entry, scenario development and sensitivity analysis. César

Izaurralde contributed to the description of the EPIC model development, methods, results and uncertainties. Doug Reinemann contributed to the statistical analysis of the ethanol production

LCA results. My contribution to the following text included 100% of the background research and methodology, 100% of the LCA calculations and approximately 95% of the text.

119

CHAPTER 4

PART 1.

Life Cycle Assessment of Switchgrass Cellulosic Ethanol Production in the Wisconsin and

Michigan Agricultural Contexts: Testing LCA results for sensitivity to spatial variability

Short title: LCA of Switchgrass Ethanol Production in Wisconsin and Michigan

Authors: Julie C. Sinistore1,2, Douglas J. Reinemann1,2, R. César Izaurralde2,3

Affiliations: 1. Biological Systems Engineering Department of the University of Wisconsin-

Madison, 2. Great Lakes Bioenergy Research Center, 3. Joint Global Change Research Institute

of Pacific Northwest National Laboratory and the University of Maryland-College Park.

Brief biography of each

Dr. Julie C. Sinistore was a research assistant with the Great Lakes Bioenergy Research Center

and is a Ph.D. graduate from the Biological Systems Engineering Department of the University

of Wisconsin-Madison. She is currently the Senior Life Cycle Analyst at Virent Inc. in Madison,

Wisconsin. Her expertise is in the Life Cycle Assessment of bio-based fuels and products with particular attention paid to local and regional differences in agricultural means of production.

Dr. Douglas J. Reinemann is Professor of Biological Systems Engineering, University of

Wisconsin-Madison. He is a member of the sustainability group of the UW Great Lakes

Bioenergy Research Center examining environmental impacts of biofuels production

systems. He also leads the UW 'green cheese' team who are investigating synergies between 120

dairy and biofuels production systems in Wisconsin. He has been on the executive committee

of the Midwest Rural Energy Council, an organization of power suppliers addressing issues

related to energy supply to agricultural production and processing operations as well as

integrating renewable energy resources into the energy distribution grid for the past 20 years.

Dr. R. César Izaurralde is Laboratory Fellow at the Joint Global Change Research Institute

(JGCRI), a collaboration of the Pacific Northwest National Laboratory with the University of

Maryland- College Park. He is also adjunct professor in the Department of Geographical

Sciences, University of Maryland. Dr. Izaurralde is a soil scientist with expertise in agriculture

and ecosystem modeling. His current research aims at developing robust modeling methods for

evaluating the impacts of climate change on terrestrial ecosystems and water resources. His

research also examines mitigation options in agriculture and energy through soil carbon

sequestration, greenhouse gas emission reductions, and the sustainable production of biofuels.

Keywords: Panicum virgatum L., greenhouse gas emissions, soil carbon, nitrous oxide,

Environmental Policy Integrated Climate (EPIC), net energy ratio, acidification, eutrophication

4.2. ABSTRACT

Spatial variability in yields and emissions from soils has been identified as a key source of

variability in Life Cycle Assessments (LCAs) of agricultural product such as cellulosic ethanol.

This study aims to conduct an LCA of cellulosic ethanol production from switchgrass in a way that captures this spatial variability and tests results for sensitivity to using spatially averaged results. The EPIC model was used to calculate switchgrass yields, greenhouse gas emissions as 121 well as nitrogen and phosphorus losses from crop production in southern Wisconsin and

Michigan at the watershed scale. These data were combined with cellulosic ethanol production data via Ammonia Fiber Expansion and Dilute Acid pretreatment methods and region-specific electricity production data into an LCA model of eight ethanol production scenarios. Standard deviations from the spatial mean yields and soil emissions were used to test the sensitivity of Net

Energy Ratio, Global Warming Potential Intensity and Eutrophication and Acidification

Potential metrics to spatial variability. When watershed mean values were used for agricultural

GHG emissions, all of the ethanol production scenarios met the 60% reduction in GHG emissions (compared to gasoline) threshold set forth in the Energy Independence and Security

Act of 2007. When soil emissions were increased by one standard deviation from the spatial mean, however, two scenarios failed to meet this target. Substantial variation in the

Eutrophication Potential was also observed when nitrogen and phosphorus losses from soils were varied. This work illustrates the need for spatially-explicit agricultural production data in the

LCA of biofuels and other agricultural products.

122

4.3. INTRODUCTION

As it stands today, the corn grain ethanol biofuel sector is well established in the United States and many studies have evaluated basic sustainability aspects of first generation ethanol production (Farrell et al., 2006; Hill et al., 2006; Hsu et al., 2010; Liska et al., 2009; Pimentel,

2003; Sinistore and Bland, 2010 and Wu, et al., 2006). This level of understanding is due, in part, to the fact that studies can be based on agricultural and biorefinery data collected directly from existing producers (Liska et al., 2009 and Sinistore and Bland, 2010). The sustainability of second generation biofuel production, such as cellulosic ethanol, however, is not as well understood since many of the feedstocks for cellulosic ethanol production are not currently produced at commercial scale and there are no operational commercial-scale cellulosic ethanol plants. Life Cycle Assessment (LCA) is the widely accepted methodology for assessing the environmental impact of a product and it relies on complete and accurate data for the production system under consideration. The lack of actual production data on biomass and cellulosic ethanol production makes conducting an LCA of cellulosic ethanol difficult. The LCA of corn grain ethanol uncovered many key areas of the production of both feedstock and fuel with potentially negative environmental consequences and illuminated areas for improvement (Hill et al., 2006;

Hill et al., 2009; Farrell et al., 2006 and Sinistore and Bland, 2010). Consequently, similar studies must be applied to cellulosic ethanol production if we are to avoid negative environmental consequences and help to design agricultural landscapes that maximize environmental benefits.

A significant portion of the overall environmental benefits (such as GHG emissions reduction) and detriments (such as eutrophication and acidification) of cellulosic ethanol production stem from biomass feedstock production (Fu et al., 2003; Hsu et al., 2010 and Schmer et al., 2007). 123

The type of crop grown and the specific means of production for that crop (e.g. tillage type,

fertilizer and pesticide application, harvest method and frequency) can greatly alter the results of

a full cellulosic ethanol production LCA. Furthermore, these impacts vary significantly over

space and time, yet many LCA studies treat crop production as a static and spatially-averaged

(over large areas) input to ethanol production (Hsu et al., 2010; Spatari et al., 2010 and Laser et al., 2009). Agricultural LCA, unlike the LCA of an industrial system, must deal with the non- linearity and amorphous boundaries intrinsic to ecological systems (such as soil) over space and time to ensure accurate assessments of emissions to air, water and soils. In fact, spatial variability has been identified as a key source of variability in LCA (Johnston et al., 2009; McKone et al.,

2011 and Reap et al., 2008).

Spatial variability can manifest in an LCA in several ways. Potting and Hauschild (2006) classify three major types of LCAs with regards to their treatment of space. The most pertinent to this discussion is the “site-generic LCA” which is said to “lack spatial information and assume globally homogeneous effects” (Potting and Hauschild 2006). The use of spatially-homogeneous data does not invalidate a study, but it adds uncertainty which might not be suitable for the accurate evaluation of ethanol-producing regions/states for the purposes of meeting sustainability standards. In reality, the variability in agricultural production from state to state and even county to county is considerable. Nevertheless, new modeling tools and spatial datasets (e.g. Zhang et al., 2010) can replace the use of spatially homogeneous assumptions about crop production encountered in site-generic LCAs with site-specific crop production data.

Reap et al. (2008) also identify LCA problems related to space as defined by spatial variation and local environmental uniqueness. The authors define spatial variation as “differences in geology, topography, land cover (both natural and anthropogenic), and meteorological 124 conditions,” while local environmental uniqueness is defined as “differences in the parameters describing a particular place (i.e. soil pH).” Distinctions in the geology, topography and meteorology of a landscape rarely align with political boundaries such as county or state lines.

Instead, these landscape variables, plus variables like soil type, are more likely to conform to ecological boundaries such as watersheds. Ecological and political boundaries, however, rarely align (Figure 4.1.). This fact ties back into the spatial treatment classifications developed by

Potting and Hauschild (2006) and the problems associated with the homogeneous treatment of space encountered specifically in the LCA of agricultural products.

Figure 4.1. An example of ecological and political boundaries for two regions in Wisconsin and Michigan, after to be referred to as the study areas of this research (county and watershed by 10-digit Hydrologic Unit Codes boundaries are illustrated).

While crop yield data may be available at a county- or state-level, these political geographic boundaries are incongruent with the boundaries around ecological determinants of crop 125

production (e.g. soil type). A county line can encompass many watersheds, soil types and

microclimates. Moreover, local environmental uniqueness, such as soil pH or buffering capacity,

can not only affect factors like yields, but also the overall impact of an activity on the

acidification of soil or water. Soil pH, for example, is a key determinant of the amount of

agricultural lime applied to a field and a factor in biomass yield. Moreover, the impact of soil

acidification on a landscape with a high buffering capacity will be far less severe than it would

be on a soil with a low buffering capacity. Local environmental uniqueness can vary vastly over

areas as large as counties. All of these factors point to the need for spatially-explicit crop production data in the LCA of agronomic crops such as switchgrass (Panicum virgatum L.) for cellulosic ethanol production.

There have also been studies which give the agricultural production system a more dynamic treatment, but then treat the ethanol production step as one static conversion rate of biomass to ethanol (Schmer et al., 2007 and Zhang et al., 2010). There are many variables which determine

the life-cycle impacts from ethanol production such as the type of biomass and the pretreatment

method. Each type of biomass has a different concentration of cellulose, hemicelluloses, lignin

and ash per unit mass. Each pretreatment method produces varying amounts of hydrolysable cellulose and hemicelluloses and has the potential to produce inhibitors that reduce the overall

efficiency of fermentation (Brown, 2003). All of these variables affect the net yield of ethanol, and thus the potential net energy yield and they affect the overall LCA by determining the set of inputs to ethanol production. For example, Dilute Acid (DA) pretreatment requires an input of sulfuric acid, while Ammonia Fiber Expansion (AFEX) pretreatment requires an input of ammonia and an ammonia recovery process (Humbird et al., 2011and Laser et al., 2009). A recent study noted that, though the physical mass of inputs like enzymes and chemicals is small 126

compared to the mass of feedstock input into the system, the production of these materials is

significant to the overall ethanol LCA (MacLean and Spatari, 2009). These types of factors must

be taken into account in the full LCA of cellulosic ethanol production. Without an operational

plant to study, however, accounting for all of these production variables becomes a difficult task, which is why this has been identified as one of the “grand challenges” facing the LCA of biofuels (McKone et al., 2011).

Many biofuel studies have also noted that environmental impact factors, such as acidification and eutrophication potential, need to be evaluated in addition to the most commonly evaluated factors (net energy and GHG emissions) (Bright et al., 2009; Curran, 2007; Fu et al., 2003; Hill et al., 2009 and Luo et al., 2009). Agricultural system parameters (e.g. nitrogen and phosphorus application) and electricity production assumptions (e.g. the proportion of electricity generated from coal versus nuclear power) will influence the overall acidification and eutrophication potential of cellulosic ethanol production. These parameters, however, change with location, so efforts must be made to source and use location-specific data for these inputs.

In this study, we put forth a cellulosic ethanol LCA which considers spatial variation, local environmental uniqueness, and variations in biomass feedstock and ethanol production methods as well as the upstream impacts of the production of agricultural and industrial system inputs.

Accomplishing this task required integrating the efforts of many researchers into one comprehensive yet flexible LCA. We collaborated with many researchers under the umbrella of

the Great Lakes Bioenergy Research Center (GLBRC) to obtain biogeochemical, biophysical,

electricity grid and biorefinery data. The results of this LCA will be useful to policy-makers and

analysts who must certify that a biofuel meets standards such as the Energy Independence and

Security Act of 2007 (EISA) Renewable Fuel Standard (RFS) (EISA, 2007) which requires that 127

cellulosic ethanol achieve a 60% reduction in GHGs compared to gasoline. Moreover, this type

of research can aid in identifying the most significant environmental impacts of ethanol

production and spur further research on how to reduce these impacts. Biofuel production is a

field that is experiencing significant change and development. We believe that now is the time to

guide this field towards a more sustainable system.

The main objective of this study is to perform a spatially-explicit LCA on cellulosic ethanol production in order to estimate the net energy, global warming potential (GWP), eutrophication potential (EP) and acidification potential (AP) of this production by AFEX and DA pretreatment from switchgrass under different agricultural production scenarios in southern Wisconsin and

Michigan. A secondary object of this research is to quantify the effects of crop production spatial variability and other production variables on the environmental impact assessment and the ability

of switchgrass ethanol to qualify as a cellulosic biofuel.

4.4. METHODS

To conduct this research, we performed a cradle-to-refinery-gate LCA of switchgrass (SG)

cellulosic ethanol production. This LCA focused on biomass and ethanol production in the

southern agricultural regions of Wisconsin (WI) and Michigan (MI). First, Life Cycle Inventory

(LCI) data were collected for crop and ethanol production (including standard deviations

(StDevs) from watershed mean modeled values due to spatial variation in crop production) for

eight scenarios. Then these data were used to run LCA models in GaBi 4.4 (PE, 2006). Finally, a

sensitivity analysis was conducted using spatial StDevs to test the magnitude of spatial

variability on impact category metrics. The scenarios varied by nitrogen application rate (high

and medium), location (WI and MI) and pretreatment method (AFEX and DA). All key LCA 128

components and scenario variables are summarized in Tables 4.1. and 4.2. respectively. This

LCA was conducted as per the guidelines set forth in the ISO standards (ISO, 2006a and 2006b).

Table 4.1. SG ethanol production LCA summary table Primary Switchgrass Cellulosic Functional Energy content of ethanol fuel, MJ product Ethanol Unit Co-product Electricity System Cradle to Refinery Gate: Production of Allocation Displacement Boundary: agricultural inputs (fertilizer, seed, avoidance included fossil fuels and power), depreciable method unit capital, agricultural field techniques Impact Net Energy Ratio: energy processes (chisel tiller, field cultivator, spike category output (energy input)-1 harrow, fertilizer applicator, plant drill, metrics Global Warming harvester and baler), feedstock and Potential Intensity: net input transportation, ethanol production -1 gCO2eq MJ of fuel (pretreatment, hydrolysis, fermentation, Acidification Potential: distillation, dehydration, denaturation, -1 gSO2eq MJ of fuel enzyme and chemical production, Eutrophication Potential: wastewater treatment, lignin -1 gPO4eq MJ of fuel combustion, storage and feedstock handling)

Table 4.2. SG ethanol production LCA scenario variable combination abbreviations Pretreatment Method AFEX DA

Nitrogen a b a b High Med High Med Application Rate Wisconsin WIHN AFEX WIMN AFEX WIHN DA WIMN DA Michigan MIHN AFEX MIMN AFEX MIHN DA MIMN DA a 90 kg N ha-1 b 60 kg N ha-1

4.4.1. Functional unit

The functional unit for this analysis is one unit of energy (MJ) from the final ethanol fuel (Low

Heating Value, LHV). This unit was selected because it allows for direct comparison with other

transportation fuels, co-product energy credits and internal comparisons between the energy

yield from ethanol and the initial energy inputs from feedstocks. The LHV of SG is 16.9 MJ (kg

SG dry matter)-1 (Laser et al., 2009). 129

4.4.2. Analysis, geographic and temporal system boundaries

The system boundary for this analysis begins with the upstream production of agricultural inputs

and ends at the biorefinery gate with fuel-grade ethanol (Figure 4.2.). Agricultural inputs include: nitrogen and potassium fertilizer and agricultural lime (soil amendments), power, petroleum fuels and seed. Ethanol production inputs include: biomass, petroleum fuels, electric power, chemicals and enzymes. 130

Figure 4.2. Process flow and system boundary diagram for feedstock production and cellulosic ethanol production (solid lines are material flows, dashed lines are internal 131

biorefinery heat and power flows, rectangles are stationary unit processes, rhombuses are mobile unit processes that depend on distance or area and octagons are products).

The geographic scope of this LCA focuses on two Regionally Intensive Modeling Areas

(RIMAs). One is a nine-county region in southern Michigan which has been divided into 39 watersheds as they are identified by their unique 10-digit hydrologic unit code (HUC) (Figure

4.1.). The second region is composed of four counties in southern Wisconsin and includes 46 watersheds also identified by their unique HUC (Figure 4.1.). The geographic scope of the production of the agricultural inputs to the field operation is limited to US standard production assumptions. The time unit of the analysis is one annual average year of agricultural and refinery production in the very near future. The agricultural data were averaged over one 12-year period.

The average yields and emissions from these 12 years were used to limit the influence of inter- annual weather variability on the results. Twelve years was selected because SG was assumed to be grown in 12-year cycles with two years for establishment and ten years of productivity before replanting. The areas modeled for crop production were only the areas that were previously cropped in the two RIMAs (WI = 211,026 ha and MI = 429,632 ha).

4.4.3. Data sources, assumptions and sensitivity analysis

Biomass yield and emissions from soils (to air and water) were modeled with the Environment

Policy Integrated Climate (EPIC) model for the two RIMAs. These data were modeled at a 56m x 56m (0.31 ha) resolution with simulations running for 24 years with historical weather data and

soils data from SSURGO (NRCS, 2008). The specific methodology and input data sources for

this biophysical and biogeochemical modeling can be found in Zhang et al. (2010) and the

calibration, validation and summarization of model uncertainty for EPIC the model as it pertains

to site-specific data can be found in He et al., (2006), Izaurralde et al. (2007), Izaurralde, et al. 132

(2012) and Wang et al., (2012). The testing of EPIC model uncertainty was not a part of this

LCA. Sources of uncertainty in the EPIC model include the accuracy and precision of observed

data (e.g. historically recorded weather and yield data) and the age of certain datasets (i.e. the

soil characteristics stored in the SSURGO database are an approximate spatial distribution of soil

types and a snapshot of a past soil states and soils change constantly). Furthermore, while the

EPIC model has been tested and validated with National Agricultural Statistics Service (NASS)

data and field trials in several locations, the field trials for the validation of the data used in this

study are still underway and will not be completed for a number of years. Therefore, we

acknowledge that there are limitations in using these data.

Input amount assumptions for the application of nitrogen, potassium and lime were drawn from

EPIC input assumptions to maintain consistency between the models. While nitrogen and

potassium (90 kgN ha-1 for HN, 60 kgN ha-1 for MN and 34 kgK ha-1) rates were constant for

each scenario, liming rates were calculated based on crop requirements and soil pH. The EPIC

data were first temporally averaged over the second 12-year period (1997 to 2009), then spatially averaged over the area in each watershed in the study areas. The temporal average was that of the last 12 years of a 24-year model run to allow for the spin-up of the model parameters such as soil carbon. This spatial average is the average of all of the 56m x 56m pixels in each watershed and the StDevs of the within-watershed distribution of inputs (lime) and outputs (yield and emissions) were used in the sensitivity analyses as the basis for testing spatial variability. This

represents an exercise in setting boundaries around the estimate of environmental impacts in the

RIMAs.

The LCI data for the upstream production of inputs to the agricultural system were taken from

GaBi 4 (PE, 2006) which included the US LCI dataset produced by the National Renewable 133

Energy Laboratory (NREL, 2008), the PE International Professional database (PE, 2006) and

EcoInvent (EcoInvent Center, 2007). Data on the region-specific electricity production grid mix for the two RIMAs in 2010 were modeled by P. Meier of the GLBRC (P. Meier Pers. Comm.,

2011). Regional emission factors for electric power were developed using the MyPower electricity dispatch model (MER, 2011) with data from the U.S. EPA National Electric Energy

Data System (NEEDS) (EPA, 2010a) and associated documentation (EPA, 2010b). Regional emission factors for transportation and process-heating fuels were based on the Greenhouse gases, Regulated Emissions, and Energy use in Transportation (GREET) model (ANL, 2010).

For a full list of LCI inputs and data sources, see Supplemental Table 4.1. The GaBi v. 4.4 model was used as the platform to integrate data from these sources and test the sensitivity of the analysis to variations in parameters (PE, 2006).

The DA pretreatment specifications for heat, power, chemical and water loading as well as the conversion efficiencies from biomass to relevant components were taken from the 2011 NREL ethanol production process design study (Humbird et al., 2011). The same AFEX pretreatment specifications, plus data on the ammonia recovery system, were taken from personal communications with GLBRC scientists and the 2009 article by Laser et al. (B. Bals Pers.

Comm., 2011 and Laser et al., 2009). Separate enzymatic hydrolysis and fermentation (with recombinant Zymomonas mobilis) as well as wastewater treatment, distillation, dehydration, denaturation and lignin combustion for process power processes are modeled after Humbird et al.

(2011). The conversion efficiencies for hexoses and pentoses to ethanol are 90% and 80% respectively for DA and 95% and 95% respectively for AFEX (Humbird et al., 2011 and Laser et al., 2009). The SG harvest index was 95%, though GLBRC field trials achieved about 65%.

These conversion processes and efficiencies were chosen because they are all potentially feasible 134

in the near-term. Switchgrass is a new crop, but technology is likely to develop rapidly to make

its production and harvest as efficient as that of other commodity crops if demand for cellulosic ethanol continues to rise. Current harvesting is being conducted with machinery optimized to harvest other crops (e.g. hay), but our analysis is meant to represent a mature production system that could support the near-term large-scale production of cellulosic ethanol in the two RIMAs.

At the time of this writing, there are no operational commercial-scale cellulosic ethanol plants in the US. Though there are plans for commercial cellulosic ethanol plants and some are even under construction, the lack of actual production means that there is no current demand for feedstock production. Therefore, we assumed the occurrence of a pulse in the market that would lead to the rapid construction of two cellulosic ethanol plants (in southern WI and MI), which, in turn, spur direct land use change on area occupied by field crops (e.g. corn (Zea mays) and soybean

(Glycine max [L] Merr.). Thus, our analysis evaluates the life-cycle implications of the birth of the commercial cellulosic ethanol market in the near future. To account for the direct land use change (DLUC) which would result from this pulse, all emissions from agricultural lands were calculated as a change from a baseline agricultural scenario of a conventional corn-soybean rotation with high fertilizer N (155 kgN ha-1 for WI and 135 kgN ha-1 for MI), conventional

chisel tillage, no stover removal, 1kgAI ha-1 pesticide application, and phosphorus and potassium

application rates of 30 kgP ha-1 and 20 kgK ha-1 for WI and 24 kgP ha-1 and 34 kgK ha-1 for MI.

This rotation was chosen as the baseline for several reasons. First, it is a conservative estimate for potential DLUC. If land were transitioned out of a more intense cropping system (e.g. continuous corn, chisel till and high N), then almost any other set of management practices would appear beneficial. Conversely, if a comparatively benign system was used for the baseline

(e.g. a corn, soy, alfalfa, fallow rotation with no till, and limited nitrogen application), the results 135 might be skewed to look overly damaging. Second, the corn soy rotation is a common practice in both study areas (USDA NASS, 2009). Therefore, this management practice represents a reasonable middle-of-the-road assumption for a baseline scenario.

It was assumed that existing agricultural land in a baseline crop production system was converted to SG production. It has been noted that these sorts of DLUC can impact LCA results significantly, but that these effects vary substantially in the landscape and sophisticated biological models should be used to account for these changes (Lange, 2011 and Spatari et al.,

2005). Therefore, the EPIC model was used to model the baseline and bioenergy production scenarios’ carbon, nitrogen and phosphorus fluxes. Accounting methods for indirect land use change (ILUC) were considered outside of the goal and scope of this study because a clear and commonly accepted method for calculating ILUC is still a matter of great scientific debate (Dale and Kim, 2011; Gawel and Ludwig, 2011; Kim and Dale, 2011; Kline et al., 2011; Mathews and

Tan, 2009; O’Hare et al., 2011 and Searchinger et al., 2008).

The importance of tracking biotic versus abiotic carbon separately, rather than ignoring biotic carbon, has been a noted criticism of previous biofuel studies (Haberl et al., 2012). The difference between abiotic and biotic carbon is that abiotic carbon is a part of non-living things such as crude oil, natural gas and coal, while biotic carbon is a part of living things like plants, microbes and animals (Baumann and Tillman, 2004). The abiotic carbon contained in fossil fuels would have remained essentially sequestered underground if human beings had not removed these fuels and combusted them, thereby releasing the abiotic carbon into the ambient air where it can rise into the troposphere and exacerbate the greenhouse effect. One primary goal of biofuels production is to reduce the concentration of carbon dioxide in the ambient air so as to prevent this abiotic CO2 from making it into the atmosphere where it can exacerbate climate 136

change. There are two ways in which biofuels can accomplish this goal: 1) by reducing the

release of abiotic carbon from fossil storage by displacing fossil fuel use (primarily liquid

transportation fuels) and 2) by cycling the carbon already in the ambient air and sequestering

some of it in biotic pools such as plants and soils. Carbon dioxide in the ambient air is obviously

composed of abiotic and biotic carbon since some carbon comes from burning fossil fuels and

some comes from biological processes like respiration and decomposition. Nevertheless, the end

goal is to reduce all of the ambient carbon, regardless of its source. Therefore, for this analysis,

ambient carbon was categorized as abiotic so that any net removal of carbon from the air would

be considered a transition from abiotic to biotic carbon. Then, any release of the biotic carbon

(through fermentation, combustion or changes in soil) would balance out in the calculation of net

carbon impacts and, therefore, address the criticism that biotic carbon emissions cannot be

ignored (Haberl et al., 2012).

Ecosystem carbon gains and losses (e.g. net carbon dioxide absorbed from ambient air and synthesized into plant material or carbon lost by microbial respiration) were modeled with EPIC

(Izaurralde et al., 2006 and 2012). We also accounted for the emissions of biotic CO2 in the

ethanol production system (from fermentation and lignin and biogas combustion). The final

product leaving the system boundary is ethanol which has embodied carbon from the plant. Since

combustion is not in the system boundary, this carbon is not part of the biotic carbon release. Net

GHG emission values therefore account for fossil and biotic emissions and any net absorption of carbon in the system. The Global Warming Potential (GWP) Intensity, Eutrophication Potential

(EP) and Acidification Potential (AP) impact category metrics are characterized measurements as per the CML, 2001 characterization factors available in GaBi 4.4. These are all measures of the global effects of releasing particular compounds into the environment as a result of the life 137

cycle of a product. The GWP weights the potential for different gasses to contribute to global

warming according to an equivalent mass of carbon dioxide. There are many molecules that

contribute to GWP with differing intensity. For example, one gram of methane, nitrous oxide or sulfur hexaflouride emissions is equivalent to 25, 298 and 22,200 grams of CO2 emissions, respectively. Similarly, the EP weights the potential for different molecules to contribute to eutrophication if released into an environment that is vulnerable to eutrophication. This potential is measured in terms of phosphate equivalents such as, for example, ammonia, ammonium, nitric

acid or phosphoric acid (0.35, 0.33, 0.1 and 0.97 phosphate equivalents, respectively). Moreover,

the AP weights the potential for different molecules to contribute to acidification if released into

a suitable environment. This measure is based on sulfur dioxide equivalents such that, for example, one gram of nitric acid, phosphoric acid, sulfuric acid or nitrogen oxides (as NO2) is

equivalent to 0.51, 0.98, 0.65 or 0.7 grams of PO4 equivalents, respectively.

4.4.4. Allocation

Electricity generated from co-product lignin combustion at the ethanol refinery displaced the use

of conventional electricity generated in each RIMA at the refinery. Excess electricity was

exported to the grid to displace electricity production from conventional sources specific to each

region.

4.5. RESULTS AND DISCUSSION

The results of this are study presented in three sections. First, there are the results of the cropping

system section of the LCA in which only the agricultural production factors are considered in

order to show the spatial variation in results over the two RIMAs. Second, the full ethanol 138

production LCA results are presented for all eight scenarios. Finally, the sensitivity analysis

results show how spatial variations in agricultural production affected the final LCA results.

4.5.1 Cropping system results and spatial variation

Table 4.3. and Figure 4.3. show the variation in yield and impact category metrics by watershed for the cropping system portion of the LCA. Full-sized maps can be found in Appendix 6. The ranges of values observed in the crop production LCA indicate that there is great variation in all metrics across the watersheds due to differences in soils and weather and this impacts biomass yields and emissions to soil and water. Note that EPIC calculated SG yields may overestimate realistic, present-day, commercial-scale yields. The EPIC model has not been validated for the specific case of SG production in across the entire RIMAs considered in this analysis. Therefore, these yields are to be used as relative indicators for LCA modeling purposes, not to predict the exact potential SG yields in the two RIMAs. Furthermore, the EPIC model used an optimistic

95% harvest index which is higher than some studies have shown. Finally, the GWP Intensity,

AP and EP results are all presented as a change from the baseline crop production scenario to account for DLUC, therefore, these results are do not represent the environmental impacts of SG crop production alone. They represent a change in current crop production practices in these

RIMAs.

139

Table 4.3. RIMA average and watershed range and StDev of SG production metrics across the RIMA watersheds for SG production only and change from the baseline scenario -1 -1 -1 - (NER, MJ output MJ input ; GWP, gCO2eq MJ ; AP, gSO2eq MJ and EP, gPO4eq MJ 1) (n = 46 and n = 39 for WI and MI, respectively). Scenario WIHN MIHN Average Min Max StDev Average Min Max StDev a NER 31.1 10.2 65.8 7.4 32.1 23.3 35.3 2.28 GWP -1.03 -3.15 -0.46 0.35 -1.07 -1.32 -1.00 0.05 AP 0.29 0.13 0.84 0.10 0.28 0.25 0.38 0.02 EP -0.36 -1.87 0.68 0.49 0.07 -0.26 0.42 0.18 Scenario WIMN MIMN a NER 35.5 23.3 46.3 4.6 44.4 4.35 310 47.8 GWP -1.12 -1.40 -0.66 0.16 -1.47 -9.28 -0.13 1.50 AP 0.24 0.18 0.37 0.03 0.31 0.03 2.03 0.32 EP -0.89 -3.45 0.03 0.56 0.03 -0.27 1.02 0.22 a Here the NER is calculated with the energy output in the switchgrass biomass.

140

Figure 4.3. Yield, Net Energy Ratio, Global Warming Potential, Acidification Potential and Eutrophication Potential by watershed from the switchgrass production portion of the LCA as a change from the baseline scenario (Yield, kgSG dry matter ha-1; NER, MJ output 141

-1 -1 -1 -1 MJ input ; GWP, kgCO2eq ha ; AP, kgSO2eq ha and EP, kgPO4eq ha ). Red to green colors indicate poor to good environmental outcomes.

It is important to note the difference between actual and potential effects in these metrics. For example, eutrophication potential is not a measure of the eutrophication that has occurred due to some practice. Rather, it is an indicator of eutrophication that could occur because of the types and amounts of certain compounds released into the environment as a result of a practice. Here, a negative potential indicates a reduction in the possibility that impact will occur. A negative potential does not mean that the global warming, eutrophication or acidification effects were actually reduced in a physically observable way. These potentials are therefore more stochastic than deterministic indicators of environmental impacts since there is a probability of environmental impact if these compounds are emitted to air, water and soils. These impacts are

not deterministic because these compounds can be converted and transported in different ways

after being emitted that will not necessarily cause environmental impacts such as climate change,

eutrophication or acidification. In order to calculate such deterministic effects, a fate-and- transport model would need to be used.

The great variation in results for the switchgrass production portion of the LCA is evident in

Figure 4.3 and Table 4.3 (e.g. the six fold difference in the NER range for the WIHN scenario).

From these maps, areas that could produce SG and minimize negative environmental effects become evident. For example, the most environmentally favorable outcomes appear consistently in the northeast corner of the WI RIMA with the medium nitrogen application. It is interesting to note, however, that this is considered to be prime agricultural land in WI. Therefore, it is somewhat unlikely that SG would be grown in this area unless the price paid for SG can compete with other commodity crops. A better locational target for SG production could be watersheds 142

that achieve reasonable NERs (i.e. not prime agricultural land) and minimize other

environmental impacts such as the southwest corner of WI under medium nitrogen application.

What also becomes apparent from these maps is that the increased nitrogen application does not

significantly increase the NER by boosting yields, but it does increase the GWP, AP and EP in

most areas. This indicates that an optimal N application rate, from an LCA perspective, is closer

to 60kgN ha-1 than 90kgN ha-1. Decreasing N application according to the needs of the crop

would also boost the NER since nitrogen fertilizer application makes up just over 50% of the

energy inputs to the crop production system in each scenario.

4.5.2. Ethanol production LCA results

The overall results of the full ethanol production LCA have notable differences across all eight

scenarios (Table 4.4., Figures 4.4. and 4.5.). The highest NER occurs in the WIHN DA scenario,

in part, because the biomass yields for this scenario were quite high (12 Mg SG dry matter ha-1

on average) and a higher biomass yield from this scenario means that more ethanol was produced

for the same inputs. The DA pretreatment required less process energy than the AFEX

pretreatment because AFEX requires additional energy for ammonia recovery. Therefore, the

AFEX pretreatment had a lower NER and higher GWP compared to the DA pretreatment since

the additional energy required for ammonia recovery reduced the total excess electricity

available for export to displace grid electricity. Additionally, all of the MI scenarios had lower

GWP than the WI scenarios because the MI grid uses more coal and less natural gas for

electricity generation than the WI grid. Therefore, each unit of electricity displaced in MI avoids more GHG emissions than in WI. Similar effects were observed in the AP results due to relative emissions of NOx and SOx. 143

Table 4.4. Results for impact category metrics across all SG ethanol production scenarios -1 -1 -1 - (AP, gSO2eq MJ ; EP, gPO4eq MJ ; GWP, gCO2eq MJ and NER, MJ output MJ input 1) State WI MI Pretreatment AFEX DA AFEX DA N rate MN HN MN HN MN HN MN HN NER 5.76 5.86 7.93 8.09 5.74 5.66 7.88 7.72 GWP 13.2 16.7 10.5 14.7 8.10 12.9 4.67 10.2 AP 0.032 0.28 0.046 0.042 0.033 0.034 0.048 0.05 EP -0.017 -0.007 -0.023 -0.010 0.015 0.017 0.014 0.016

Figure 4.4. Net Energy Ratio (MJ output MJ input-1) versus Global Warming Potential -1 Intensity (gCO2eq MJ ) for all SG ethanol production scenarios. 144

-1 Figure 4.5. Acidification Potential (gSO2eq MJ ) versus Eutrophication Potential (gPO4eq MJ-1) for all SG ethanol production scenarios.

The high nitrogen application rate scenarios resulted in higher GWP from both the increased need to produce fertilizer and the increased N2O emissions from soils. Differences in EP were primarily a function of nitrogen application rate. Negative values for EP indicate that the EP from SG production was significantly less than that of the baseline scenario due to differences in fertilizer application rates and the effects on soils of producing annual versus perennial crops.

Therefore, the potential for eutrophication is lessened by this land use change from a conventional corn-soybean management to SG production. 145

There is no one biofuel production LCA in the literature that would exactly match the system

boundary, geographic and temporal scope or production data and assumptions of this study.

However, there are some studies with similar assumptions that provide a reasonable point of

comparison. A corn grain ethanol study on Wisconsin-specific production found an NER of 1.58

-1 and a GWP Intensity of 49.4 gCO2eq MJ (Sinistore and Bland, 2010). Corn grain ethanol

production, however, enjoys a substantial GWP reduction benefit from the displacement of

conventional feeds with co-products of ethanol production. If a biofuel is to be produced in

Wisconsin, this point of comparison suggests that appreciable NER and GWP Intensity benefits

could be obtained if switchgrass is grown for biofuels instead of corn grain. Another LCA found

-1 a GWP Intensity of 22.69 gCO2eq MJ for ethanol from switchgrass grown in Ontario, Canada

(Spatari et al., 2005), but the Canada study and this study differed in several ways. First, the

study focused on Canadian production which included Canadian assumptions for electricity

production, transportation and agriculture. Second, the Canada study did not take into account

the impacts of DLUC. Finally, the Canada study was based on older biorefinery technology

assumptions than this study. The difference in the results of the Canadian study and this study

support the notion that the location of agricultural and refinery production are important factors

in the overall ethanol LCA.

The NER and GWP Intensity of biofuel production are the most commonly evaluated impact

category metrics in biofuel LCAs, while evaluation of other metrics, such as AP and EP are less

common. A recent LCA on switchgrass ethanol production in the United Kingdom found an AP

-1 -1 of 0.415 gSO2eq MJ and an EP of 0.149 gPO4eq MJ (Bai et al., 2010). These values are greater than those found in our analysis by a factor of ten or more. One driver of this difference, especially for EP, is the assumption in Bai et al. (2010) that switchgrass received 100kgN ha-1 yr- 146

1. This amount of N application seems unrealistic as it approaches the levels of N application for

commodity crops like corn, which have relatively low Nitrogen Use Efficiency (Grace et al.,

2011; Grassini and Cassman, 2011 and Van Groenigen et al., 2010). Recent studies suggest that

the N application rates should be in the range of 65 to 69 kgN ha-1 in order to optimize both biomass and economic yields (Aravindhakshan et al., 2011 and Haque et al., 2009).

Furthermore, location-specific variables like the electricity production mix and transportation

distances affect the release of acidifying compounds in a way that can cause significant

differences in the measure of AP for a given production regime. These differences, plus

differences in data and LCA assumptions are the likely cause of disparities in results.

A Yates Computation is a statistical method for calculating values for the mean, main effects and

interactions between studied variables. The results of the ethanol production LCA were tested

with a Yate Computation according to the following treatment variables: RIMA, nitrogen rate and pretreatment. These varied in a binary fashion (WI or MI, high or medium N and AFEX or

DA). For the resulting metrics the following predictive equation was used:

Resulting Metric = Intercept + R + N + P +RxN + RxP + NxP (3.1)

Where R is RIMA, N is nitrogen application rate and P is pretreatment. The overall mean NER is

6.83 and the most important determinant of the NER was the pretreatment method (32% of the mean) with nitrogen application rate as the second most important effect (3% of the mean) (see

Appendix 10). The effect of the RIMA on NER is very small (<1%). There is only a small interaction between RIMA and N application rate. This indicates that the effect of nitrogen application rate in each RIMA is slightly different. The interaction between nitrogen application rate and pretreatment is very small, such that the effect is similar for both high and medium N 147

application rates. For GWP, the mean was 11.4 and the RIMA, nitrogen application rate and

pretreatment were the largest effects with 45%, 31% and 24% of the overall mean, respectively.

The effects of nitrogen application rate are different for the two RIMAs as signified by the

significant interaction between these two variables (12%). Conversely, there is weak interaction

between RIMA and pretreatment (6%) and nitrogen application rate and pretreatment (6%).

4.5.3. Sensitivity Analysis

The StDevs from the spatial average of all the EPIC model pixels in each watershed were provided from the EPIC model for the biomass yield, lime application rates (calculated for the demands of the soil and crop) and emissions to air and soils due to agricultural production.

These StDevs were used to test the sensitivity of the LCA results to possible spatial variation and

produce ranges for our results rather than just static values. Note that the whiskers on the

following figures do not represent the traditional statistical error around the calculation of a mean

value. Instead, these whiskers are to be interpreted as a range of possible values caused by spatial

variability alone. They do not include a measure of other sources of variability or uncertainty in the EPIC or LCA models. They indicate scenarios for each environmental metric when the spatially-explicit crop production inputs to the LCA model were decreased or increased by one

StDev from their spatial mean within a watershed. This analysis was conducted to quantify the range of environmental impact results that could arise from producing biomass on more or less productive land within a watershed. This range is an indicator of how ecological determinants of yield and emissions from crop production vary in space within a watershed.

148

4.5.3.1. GWP

The GWP Intensity across all eight scenarios was influenced by variation in the CO2 and N2O emissions from soils due to cultivation and by variations in yield (Figure 4.6.). Negative emissions in this figure indicate either the uptake of carbon into plant tissues and soils during agricultural production (including any change in soil carbon from the baseline scenario) or avoided emissions from the displacement of electricity production at the grid by the export of excess electricity from the biorefinery. The positive GHG emissions from ethanol production include emissions from the combustion of fossil fuels, emissions from soils due microbial processes, the release of carbon from the plant material during ethanol production (from fermentation and combustion) and any other emission of GHGs from the production of agricultural or industrial inputs. Recall that all carbon, whether it originated from fossil fuels

(abiotic) or living material (biotic), was tracked separately and fully throughout the analysis.

This is in contrast other ethanol LCAs which tracked only abiotic carbon and assumed, without testing this assumption, that the net biotic carbon must be equal to zero regardless of soil carbon change or the effects of DLUC (Hill et al., 2006; Liska et al., 2009; Pimentel, 2003; Sinistore and

Bland, 2010; Spatari et al., 2005 and Wu, et al., 2006). The net difference between the starting biomass carbon and the carbon released during fermentation and combustion was between 6% and 9% due to the carbon content of biomass ash, incomplete combustion and other solids remaining after wastewater treatment (Humbird et al., 2011).

The influence of spatial variations in GHG emissions from soils and yield is greater in WI than in MI. Biomass yield is an important determinant of overall efficiencies. For example, a larger

SG yield from the same area with the same amount of inputs and tillage will reduce the overall

GWP per mass of SG that is carried forward to the biorefinery. Across all scenarios, however, 149 spatial variability in GHG emissions from soils contributed four to five times more to the change in the GWP Intensity than variations in yield. It is clear that the absorption and emission of carbon during agricultural production is the primary driver of the overall GWP Intensity of SG ethanol. The GHG emissions from the production and transport of inputs to the system contribute far less to the overall GWP Intensity. Furthermore, the effect of DLUC on the overall ethanol

LCA provides a GWP Intensity benefit to the SG ethanol. From a policy standpoint, this could provide an incentive to transition agricultural land out of uses that are known to be large GHG emitters into uses that decrease emissions by building soil carbon. 150

-1 Figure 4.6. Net Global Warming Potential Intensity (black squares, gCO2eq MJ ) with relative contributions (bars) from each stage of ethanol production and net GWP Intensity with error introduced by spatial variability emissions from soils and yield as a change from the baseline crop production scenario.

151

This change in the final GWP Intensity value for the cellulosic ethanol production has important

policy implications. The EISA 2007 standard requires that a biofuel reduce emissions by 60%

compared to an energy equivalent amount of gasoline in order to be a cellulosic fuel. Whether or

not the SG ethanol produced in each scenario meets or misses the definition for cellulosic fuel

depends on if the yields and emissions from soils in each watershed vary by plus or minus one

StDev from the spatial mean (Table 4.5.). All of the scenarios meet the 60% reduction

requirement if the spatial mean is used, but WIHN AFEX and WIHN DA fail if emissions from

soils increase by one StDev even if yields increase by one StDev.

-1 Table 4.5. Percent reduction in GHG emissions compared to gasoline (94 gCO2eq MJ ) for each switchgrass ethanol production scenario and the new percent reduction value if the modeled emissions and yields increased or decreased by one StDev in each watershed.

WIMN WIHN WIMN WIHN MIMN MIHN MIMN MIHN Scenario AFEX AFEX DA DA AFEX AFEX DA DA

% reduction using spatial 86% 82% 89% 84% 91% 86% 95% 89% mean % reduction if increased by 1 61% 57% 62% 57% 83% 77% 88% 81% StDev % reduction if decreased by 111% 107% 116% 112% 99% 95% 102% 98% 1 StDev

152

4.5.3.2. NER

The NER for the full ethanol production LCA is not sensitive to variations in liming (less than

1% change in NER), but the energy inputs to the cropping system are sensitive to lime. In WI,

the NER changes by as much as 20% if the lime application rates is varied by one StDev from its

spatial mean, while, in MI, this value changes by less than 5%. The NER was also changed by less than 0.5% across all scenarios when yields were varied by one StDev from their spatial mean. The primary contribution the input energy was the energy used to produce inputs to production such as chemicals for agricultural production, pretreatment, cellulase production and denaturation (Figure 4.7.).

153

Figure 4.7. Relative contributions to energy use and production in ethanol production (without the energy from ethanol itself) measured on the left y-axis and Net Energy Ratio (MJ output MJ input-1) including energy in ethanol measured on the right y-axis.

154

4.5.3.3. AP and EP

The combustion of fossil fuels for field operations, transportation and the production of biorefinery chemicals and enzymes were the primary contributors to AP and emissions from soils were not a significant contributor (Figure 4.8.). The avoided combustion of fossil fuels (e.g. coal and natural gas) from electricity displacement provided a substantial benefit to the AP of the fuel. The final AP value for the LCA was not sensitive to spatial variation in SG production (less than 1% change). By contrast, EP was sensitive to spatial variations in the modeled nitrogen and phosphorus losses from soils via runoff, sediment, lateral subsurface flow and percolation below the root zone. The change from the baseline conventional corn-soybean production to SG production substantially reduced the EP of the ethanol production in WI, but was highly sensitive to spatial variation in the all scenarios (Figure 4.9.). 155

Figure 4.8. Relative contributions to Acidification Potential (gSO2eq MJ-1) from each stage of ethanol production on the left y-axis and net Acidification Potential with range of values due to spatial variability on the right y-axis. 156

Figure 4.9. Relative contributions to Eutrophication Potential (gPO4eq MJ-1) from each

stage of ethanol production on the left y-axis and net Eutrophication Potential with range

of values due to spatial variability in N and P movement in soils on the right y-axis.

4.6. CONCLUSIONS

The considerable spatial variations observed in this study indicate that spatial averages of yields and emissions over large areas (like counties, states or countries) introduce a large amount of variability into the results of LCAs that involve agricultural production. When spatial averages

are used for a county, state or country, data on the StDevs from these averages are not available, 157 therefore, the sensitivity of the results to actual spatial variation in these parameters cannot be tested. The maps in Figure 4.3. illustrate the large variation in yields and environmental impact metrics in a single county. By using modeled data with StDevs, we were able to test the sensitivity of the LCA results to spatial variations and show that, in all but two scenarios, the SG ethanol produced meets the EISA 2007 GHG emissions reduction target even with spatial variation. Moreover, we demonstrated that a fuel could meet the standard if mean values were used, but fail if emissions from soils are actually greater by only one StDev from the spatial mean. In addition to this, the location of the biorefinery determined the electricity grid mix displaced by the co-product electricity generation which affected the GWP and the AP of the fuel. Therefore, the types and ratios of fossil fuels used to produce power in each region are an important LCA factor for any system that also produces electricity. Finally, the statistical analysis revealed that the most important levers for controlling NER and GWP are pretreatment and a combination of location, nitrogen application rate and pretreatment respectively. This information will aid the work of producers, researchers and policy-makers in identifying production practices that boost the NER and reduce GWP Intensity of biofuels production.

From the NER vs. GWP intensity figure (Figure 4.4.), we can see that the scenario with the lowest GWP Intensity and the highest NER is MIMN DA, but on the AP vs. EP graph (Figure

4.5.), this scenario has one of the highest AP and EP combinations. The scenarios with the lowest

AP and EP are WIHN AFEX and WIMN AFEX, but these scenarios have some of the highest

GWP Intensities and lowest NERs. Still, the GWP for both WIHN AFEX and WIMN AFEX meet the EISA 2007 threshold for a 60% reduction in GHG emissions compared to gasoline

(Table 4.5.) and have higher NERs than corn grain ethanol (1.25 according to Hill et al., 2006).

The sensitivity analysis, however, shows that the spatial variability in emissions from soils and 158 yield could cause the WIHN AFEX scenario to miss the 60% emissions reduction target. The

WIMN AFEX scenario does not miss this target even with spatial variability, therefore, this scenario is has the best potential to simultaneously meet the EISA standard for cellulosic biofuel, provide a substantial net energy gain and minimize AP and EP. This analysis is just one example of how a spatially-explicit and flexible LCA model of cellulosic ethanol production could be used to evaluate the environmental impacts of several scenarios simultaneously and check that a fuel production scenario meets standards even with potential variations in metrics due to spatial variability.

We also acknowledge that the spatially-explicit crop production data from the EPIC model used in this analysis has many strengths and weaknesses. Among the weaknesses are limitations associated not only with these modeled data, but all modeled crop production data. For example, uncertainty is introduced into crop production modeling based on the accuracy and precision of historical weather and yield data used to calibrate the model. Furthermore, crop production models rely on soils data from databases like SSURGO which may be lacking in spatial precision and present only a snapshot in time of soil characteristics that change with time. In this particular analysis, the EPIC model is still in the process of being validated for its ability to predict yield and emissions from SG production in these two RIMAs. Switchgrass itself is a new crop that has yet to be grown at a large scale for harvest in these regions. Efforts must be made to evaluate the impacts of its production over large areas before it is introduced onto the landscape in order to avoid unintended consequences. Additionally, the EPIC modeled SG production represents a mature production system, similar to that of other commodity crops, that has been developed to be efficient and produce yields that would support a cellulosic ethanol plant. We believe that, if the demand for cellulosic ethanol continues to increase, the SG production system 159

will reach the efficiency levels attained my other crops, but we recognize that current SG

production has not reached this level.

Several voices in the LCA and biofuels communities have called for attention to spatial

variability and local environmental uniqueness in the LCA of biofuels (Reap et al., 2008 and

McKone et al., 2011). Modeled data is able to address these concerns in a way that observed crop

production data cannot. Field trials used to collect such data are conducted on small plots of land

in a limited number of locations. It is would be prohibitively expensive and unrealistic to attempt

to measure the yield, GHG emissions, leaching and runoff from every agricultural field in each

of the RIMAs. Field trial data could be extrapolated over the two RIMAs, but this would introduce sources of uncertainty into the overall LCA similar to that of modeled data (e.g. uncertainty from the accuracy and precision of observation tools). A major strength of the EPIC model was its ability to provide a tremendous amount of high-resolution spatially-explicit crop production data. These data facilitated comparisons of the effects of different land management practices across a large region. Due to the inherent uncertainty in the modeled data, however, we acknowledge that there are limitations in using the results of this study to draw conclusions about the environmental impacts of cellulosic ethanol production. As noted earlier, there are still no fully operational-commercial scale ethanol plants in the world and there is no large scale production of the feedstocks that would be needed to feed such an ethanol plant. Therefore, this study was intended to be an exploration into the potential consequences and benefits of commercial scale ethanol production in these regions, and not intended to be a precise summation of actual impacts from production. An assessment of actual impacts of the production of a good or service can only be conducted on an existing system. Therefore, LCAs should be conducted on future large-scale ethanol production systems. 160

In this study, we not only demonstrated the importance of spatially-explicit agricultural and electricity grid data in the LCA of biofuels, but we also demonstrated that it is possible to take these factors into account. It is important to evaluate the environmental impacts of spatial variation and a variety of land management practices together. We do not know where on the landscape crops will be grown for biofuels in a future with demand for cellulosic ethanol.

Depending on where and how the crop is grown, an analysis that includes spatial variation will indicate which location and management practice combinations will have the best or the worst impacts on the environment. Furthermore, testing the effects of increasing or decreasing EPIC modeled parameters by one standard deviation from their watershed spatial means gives us an indication of what could happen if crops are produced on the most productive or least productive lands within a watershed.

Future spatially-explicit LCA modeling could include Local Biomass Processing Depots

(LBPDs) placed in spatially-optimized locations on the landscape (Eranki et al., 2011) and the co-production of animal feeds from pretreatment (Dale et al., 2010). The addition of economic variables for the biomass selling price for “profit-oriented farmers” in a given region

(Egbendewe-Mondzozo et al., 2011) and techno-economic variables for the costs ethanol of production for a given technology (Kazi et al., 2010) would also enhance the analysis. The addition of these factors would make the model into a robust decision-support tool for evaluating the effects of the location of a biorefinery in terms of its economic viability, environmental sustainability and ability to meet policy standards. Advanced modeling techniques provide the data necessary to give agricultural production the spatially-explicit treatment required to conduct a thorough LCA and address the grand challenges identified in this field.

161

Acknowledgements

This work was funded by the Department Of Energy Great Lakes Bioenergy Research Center

(DOE BER Office of Science DE-FC02-07ER64494 and DOE OBP Office of Energy Efficiency

and Renewable Energy DE-AC05-76RL01830). The authors also gratefully acknowledge the

contributions of Bryan Bals, Bruce Dale, David Duncan, Pragnya Eranki, Shujiang Kang, David

Manowitz, Timothy D. Meehan, Paul Meier, Mac Post and Xuesong Zhang to this work.

4.7. REFERENCES Aravindhakshan, S. C., Epplin, F. M. and Taliaferro, C. M. 2011. Switchgrass, Bermudagrass, Flaccidgrass, and Lovegrass biomass yield response to nitrogen for single and double harvest. Biomass and Bioenergy. 35: 308-319. Argonne National Laboratory (ANL). 2010. Greenhouse gases, Regulated Emissions, and Energy use in Transportation Model, GREET Version 1.8C. Accessed: March 2010. Bals, B. Personal Communication. 26 January 2011. Michigan State University. Baumann, H. and Tillman, A. 2004. The Hitch Hiker’s Guide to LCA: An orientation in life cycle assessment methodology and application. p. 145. Lund, Sweden. Studentlitteratur AB. Bright, R. M. and Strømman, A. H. 2009. Life cycle assessment of second generation bioethanols produced from Scandinavian boreal forest resources: a regional analysis for Middle Norway. Journal of Industrial Ecology. 13(4): 514-531. Brown, R. C. 2003. Biorenewable Resources: Engineering New Products from Agriculture. Ames, IA: Blackwell Publishing Professional. Curran, M. A. 2007. Studying the Effect on System Preference by Varying Coproduct Allocation in Creating Life-Cycle Inventory. Environmental Science and Technology. 41(20): 7145- 7151. Dale, B. E., Bals, B. D., Kim, S. and Eranki, P. 2010. Biofuels Done Right: Land efficient animal feeds enable large environmental and energy benefits. Environmental Science and Technology. 44(22): 8385-8389. 162

Dale, B. E. and Kim, S. 2011. Response to comments by O’Hare et al., on the paper indirect land use change for biofuels: Testing predictions and improving analytical methodologies. Biomass and Bioenergy. 35: 4492-4493. Ecoinvent Centre. 2007. Ecoinvent Eata. v2.0. Ecoinvent Reports No.1-25. Swiss Centre for Life Cycle Inventories. Dübendorf. Egbendewe-Mondzozo, A., Swinton, S. M., Izaurralde, R. C., Manowitz, D. H. and Zhang, X. 2011. Biomass supply from alternative cellulosic crops and crop residues: A spatially explicit bioeconomic modeling approach. Biomass and Bioenergy. 35: 4636-4647. Energy Independence and Security Act (EISA) of 2007. Pub. L. no. 110-140. 110th Congress. 1st session. 2007. http://www.gpo.gov/fdsys/pkg/BILLS-110hr6enr/pdf/BILLS- 110hr6enr.pdf. Accessed: Nov 2011. Eranki, P. L. and Dale, B. E. 2011. Comparative life cycle assessment of centralized and distributed biomass processing systems combined with mixed feedstock landscapes. Global Change Biology: Bioenergy. 3(6): 1757-1707 Farrell, A. E., Plevin, R. J., Turner, B. T., Jones, A. D., O'Hare, M. and Kammen, D. M. 2006. Ethanol Can Contribute to Energy and Environmental Goals. Science. 311(5760): 506- 508. Fu, G. Z., Chan, A. W. and Minns, D. E. 2003. Life Cycle Assessment of Bio-ethanol Derived from Cellulose. International Journal of Life Cycle Assessment. 8 (3): 137-141. Gawel, E. and Ludwig, G. 2011. The iLUC dilemma: How to deal with indirect land use changes when governing energy crops? Land Use Policy. doi: 10.1016/j.landusepol.2011.03.003. Grace, P. R., Robertson, G. P., Millar, N., Colunga-Farcia, M., Basso, B., Gage, S. H., and Hoben, J. 2011. The contribution of maize cropping in the Midwest USA to global warming: A regional estimate. Agricultural Systems. 104: 292-296. Grassini, P. and Cassman, K. G. 2011. High-yield maize with large net energy yield and small global warming intensity. Proceedings of the National Academy of Sciences Early Edition. doi: 10.1073/pnas.1201296109. Haberl, H., Sprinz, D. Bonazountas, M., Cocco, P., Desaubies, Y., Henze, M. Hertel, O., Johnson, R. K., Kastrup, U., Laconte, P., Lange, E., Novak, P., Paavola, J., Reenbergy, A., van den Hove, S., Vermeire, T., Wadhams, P. and Searchinger, T. 2012. Correcting a 163

fundamental error in greenhouse gas accounting related to bioenergy. Energy Policy. 45: 18-23. Haque, M., Epplin, F. M. and Taliaferro, C. M. 2009. Nitrogen and harvest frequency effect on yield and cost for four perennial grasses. Agronomy Journal. 101: 1463-1469. He, X., Izzaurralde, R. C., Vanotti, M. B., Williams, J. R., and Thomson, A. M. 2006. Simulating long-term and residual effects of nitrogen fertilization on corn yields, soil carbon sequestration, and soil nitrogen dynamics. Journal of Environmental Quality. 35:1608- 1619. Hill, J., Polasky, S., Nelson, E., Tilman. D, Huo, H., Ludwig, L., Neumann, J., Zheng, H. and Bonta, D. 2009. Climate change and health costs of air emissions from biofuels and gasoline. Proceedings of the National Academy of Science. 106 (6): 2077-2082. Hill, J., Nelson, E., Tilman, D., Polasky, S. and Tiffany, D. 2006. Environmental, economic, and energetic costs and benefits of biodiesel and ethanol biofuels. Proceedings of the National Academy of Sciences. 103: 11206-11210. Hsu, D. D., Inman, D., Heath, G. A., Wolfrum, E. J., Mann, M. K. and Aden, A. 2010. Life Cycle Environmental Impacts of Selected U.S. Ethanol Production and Use Pathways in 2022. Environmental Science and Technology. 44: 5289-5297. Humbird, D., Davis, R., Tao, L., Kinchin, C., Hsu, D., Aden, A., Schoen, P., Lukas, J., Olthof, B., Worley, M., Sexton, D. and Dudgeon, D. 2011. Process Design and Economics for Biochemical Conversion of Lignocellulosic Biomass to Ethanol: Dilute-Acid Pretreatment and Enzymatic Hydrolysis of Corn Stover. National Renewable Energy Laboratory. Technical Report: NREL/TP-5100-47764. Golden, CO. International Organization for Standardization (ISO). 2006a. ISO 14040: Environmental management – Life cycle assessment – Principles and framework. Geneva, Switzerland. International Organization for Standardization (ISO). 2006b. ISO 14044: Environmental Management– Life cycle assessment, Life cycle impact assessment. Geneva, Switzerland. Izaurralde, R. C., Williams, J. R., McGill, W. B., Rosenberg, N. J. and Quiroga Jakas, M. C. 2006. Simulating soil C dynamics with EPIC: Model description and testing against long- term data. Ecological Modeling. 192: 362-384. 164

Izaurralde, R. C., Williams, J. R., Post, W. M., Thomson, A. M., McGill, W. B., Owens, L. B and Lal., R. 2007. Long-term modeling of soil C erosion and sequestration at the small watershed scale. Climatic Change. 80(1-2): 73-90. Izaurralde, R. C., McGill, W. B. and Williams, J. R. 2012. Development and application of the EPIC model for carbon cycle, greenhouse-gas mitigation, and biofuel studies. In Franzluebbers, A., Follett, R., and Liebig, M. (eds.) Managing Agricultural Greenhouse Gases: Coordinated agricultural research through GRACEnet to address our changing climate, Elsevier (in press). Johnston, M., Foley, J. A., Holloway, T., Kucharik, C. and Monfreda, C. 2009. Resetting global expectations from agricultural biofuels. Environmental Resource Letters. doi: 10.1088/1748-9326/4/014004. Kang, S. and Post, W. M. 2012. Personal Communication. Unpublished Data. Oakridge National Laboratory. Oakridge, TN. Kaufman, A. S., Meier, P. J., Sinistore, J. C., and Reinemann, D. J. 2010. Applying life-cycle assessment to low carbon fuel standards – How allocation choices influence carbon intensity for renewable transportation fuels. Energy Policy. 38: 5229-5241. Kazi, F. K., Fortman, J. A., Anex, R. P., Hsu, D. D., Aden, A., Dutta, A. and Kothandaraman, G. 2010. Techno-economic comparison of process technologies for biochemical ethanol production from corn stover. Fuel. 89: S20-S28. Kim, S. and Dale, B. E. 2011. Indirect land use change for biofuels: testing predictions and improving analytical methodologies. Biomass and Bioenergy. 35(7): 3235-3240. Kline, K. L., Oladosu, G. A., Dale, V. H. and McBride, A. C. 2011. Scientific analysis is essential to assess biofuel policy effects: In response to the paper by Kim and Dale on “Indirect land-use change for biofuels: Testing predictions and improving analytical methodologies.” Biomass and Bioenergy. 35: 4488-4491. Lange, M. 2011. The GHG balance of biofuels taking into account land use change. Energy Policy. 39: 2373-2385. Laser, M., Haimin, J., Jayawardhana, K. and Lynd, L. R. 2009. Coproduction of ethanol and power from switchgrass. Biofuels, Bioproducts and Biorefining. 3: 195-218. 165

Liska, A. J., Yang, H. S., Bremer, V. R., Klopfenstein, T. J., Walters, D. T., Erickson, G. E. and Cassman, K. G. 2009. Improvements in Life Cycle Energy Efficiency and Greenhouse Gas Emissions of Corn-Ethanol. Journal of Industrial Ecology. 13(1): 58-74. Luo, L., van der Voet, E., Gjalt, H. and Udo de Haes, H. A. 2009. Allocation issues in LCA methodology: a case study of corn stover-based fuel ethanol. International Journal of Life Cycle Assessment. 14: 529-539. MacLean, H. L., and Spatari, S. 2009. The contribution of enzymes and process chemicals to the life cycle of ethanol. Environmental Resource Letters. 4: 14001-14010. Matthews, J. A. and Tan, H. 2009. Biofuels and indirect land use change effects: the debate continues. Biofuels, Bioproducts and Biorefining. doi: 10. 1002/bbb.147. McKone, T. E., Nazaroff, W. W., Berck, P., Auffhammer, M., Lipman, T., Torn, M. S., Masanet, E., Lobscheid, A., Santero, N., Mishra, U., Barrett, A., Bomberg, M., Fingerman, K., Scown, C., Strogen, B. and Horvath, A. 2011. Grand Challenges for Life-Cycle Assessment of Biofuels. Environmental Science and Technology. 45: 1751-1756. Meier Engineering Research (MER), LLC. 2011. MyPower ™ electricity-sector model. National Renewable Energy Laboratory (NREL). 2008. U.S. Life-Cycle Inventory (LCI) Database. http://www.nrel.gov/lci/database/default.asp. Accessed: January 2010. Natural Resources Conservation Service (NRCS), Soil Survey Staff, United States Department of Agriculture. Soil Survey Geographic (SSURGO) Database for Wisconsin and Michigan. Available online at http://soildatamart.nrcs.usda.gov. Accessed: 2008. O’Hare, M., Delucchi, M., Edwards, R., Fritsche, U., Gibbs, H., Hertel, T., Hill, J., Kammen, D., Laborde, D., Marelli, L., Mulligan, D., Plevin., R. and Tyner, W. 2011. Comment on “Indirect land use change for biofuels: Testing predictions and improving analytical methodologies” by Kim and Dale: statistical reliability and the definition of the indirect land use change (iLUC) issue. Biomass and Bioenergy. 35: 4485-4487. PE International. 2006. GaBi 4: The Software System for Life Cycle Engineering. PE-Europe GmbH and IKP University of Stuttgart. Pimentel, D. 2003. Ethanol Fuels: Energy Balance, Economics, and Environmental Impacts are Negative. Natural Resources Research. 12(2): 127-134. 166

Potting, J. and Hauschild, M. Z. 2006. Spatial differentiation in life cycle impact assessment: a decade of method development to increase the environmental realism of LCIA. International Journal of Life Cycle Assessment. 11: 11-13. Reap, J., Roman, F., Duncan, S. and Bras, B. 2008. A survey of unresolved problems in life cycle assessment; Part 2: impact assessment and interpretation. International Journal of Life Cycle Assessment. 13: 374-388. Schmer, M. R., Vogel, K. P., Mitchell, R. B. and R. K. Perrin. 2007. Net energy of cellulosic ethanol from switchgrass. Proceedings of the National Academy of Sciences. 105(2): 464- 469. Searchinger, T., Heimlich, R., Houghton, R. A., Dong, F., Elobeid, A., Fabiosa, J., Tokgoz, S., Hayes, D. and Yu, T. 2008. Use of U.S. croplands for biofuels increases greenhouse gases through emissions from land-use change. Science. 319: 1238-1240. Sinistore, J. C. and Bland, W. L. 2010. Life-cycle analysis of corn ethanol production in the Wisconsin context. Biological Engineering. 2(3): 147-163. Spatari, S., Zhang, Y. and MacLean, H. L. 2005. Life Cycle Assessment of Switchgrass- and Corn Stover-Derived Ethanol-Fueled Automobiles. Environmental Science and Technology. 39(24): 9750-9758. Spatari, S., Batley, D. M. and MacLean, H. L. 2010. Life cycle evaluation of emerging lignocellulosic ethanol conversion technologies. Bioresource Technology. 101: 654-667. U.S. Department of Agriculture National Agriculture Statistics Service (USDA NASS). 2009. Cropland Data Layer. http://www.nass.usda.gov/Statistics_by_State/Wisconsin /Publications/Cropland_Data_Layer/index.asp. Accessed: Jan 2010. U.S. Environmental Protection Agency (EPA). 2010a. National Electric Energy Data System (NEEDS). Retrieved: June 2010. U.S. Environmental Protection Agency (EPA) .2010b. EPA's IPM Base Case 2006 (v3.0).Comprehensive modeling documentation. Retrieved: Oct 2010. Van Groenigen, J. W., Velthof, G. L., Onema, O., Van Groenigen, K. J. and Van Kessel, C. 2010. Towards an agronomic assessment of N2O emissions: a case study for affable crops. European Journal of Soil Science. 61: 903-913. 167

Wang, X., Williams, J. R., Gassman, P. W., Baffaut, C., Izaurralde, R. C., Jeong, J. and Kiniry, J.R. 2012. EPIC and APEX: Model use, calibration, and validation. Transactions of the ASABE. 55(3): 1-16. Wisconsin Department of Natural Resources (WIDNR). 2011. Surface Water Data Viewer. Available online at: http://dnrmaps.wi.gov/imf/imf.jsp?site=SurfaceWaterViewer. Accessed: 13 Dec 2011. Wu, M., Wang, M., and Huo, H. 2006. Fuel-cycle assessment of selected bioethanol production pathways in the United States. Argonne National Laboratory (ANL/ESD/06-7): Energy Systems Division. Center for Transportation Research. Argonne, IL. Zhang, X., Izaurralde, R. C., Manowitz, D., West, T. O., Post, W. M., Thomson, A. M., Bandaru, V. P., Nichols, J. and Williams, J. R. 2010. An integrative modeling framework to evaluate the productivity and sustainability of biofuel crop production systems. Global Change Biology: Bioenergy. 2: 258-277. Zhang, X. and Manowitz, D. Personal Communication: Nov 2011.

Supplemental Table 4.1. Life Cycle Inventory (LCI) data inputs and sources for SG ethanol Fossil Fuel Data Source Agricultural Data Source Transport Data Source Production LCI Input LCI Input LCI Input Natural Gas US LCI, 2008 Nitrogen US LCI, 2008 Train, diesel US LCI, 2008 Fertilizer powered Gasoline US LCI, 2008 Potassium US LCI, 2008 Barge, diesel US LCI, 2008 Fertilizer powered Diesel US LCI, 2008 Agricultural US LCI, 2008 Barge, US LCI, 2008 Lime residual fuel oil powered Bituminous US LCI, 2008 Seed EcoInvent, Combination US LCI, 2008 Coal 2007 Truck, diesel powered Residual Fuel US LCI, 2008 Soil PE, 2006 Truck PE, 2006 Oil Cultivation: (52,000lbs or Chisel Tiller 23 MT payload), diesel powered Liquefied US LCI, 2008 Fertilising: PE, 2006 Petroleum Mineral Gas fertilizing 168

(lime) Heat and Data Source Fertilising: PE, 2006 Chemical Data Source Power LCI Sprayer and Input Biorefinery LCI Input Residual Fuel US LCI, 2008 Soil PE, 2006 Sulphuric PE 2006 Oil Cultivation: Acid Combusted in Chisel Tiller Industrial Boiler Diesel US LCI, 2008 Sowing: Plant PE, 2006 Ammonia US LCI, 2008 Combusted in Drill Industrial Boiler Gasoline US LCI, 2008 Sowing: PE, 2006 Diammonium EcoInvent, Combusted in Spike Harrow Phosphate 2007 equipment Bituminous US LCI, 2008 Harvesting: PE, 2006 Ammonium EcoInvent, Coal straw/hay sulphate 2007 Combusted in harvester and Industrial baling Boiler Natural Gas US LCI, 2008 Switchgrass MI: Zhang, et AFEX B. Bals Pers. Combustion Yield al. 2010 pretreatment Comm, 2011; in Industrial WI: Kang and Laser et al., Boiler Post, 2012 2009 Liquefied US LCI, 2008 N loss via MI: Zhang, et DA Humbird et Petroleum surface runoff al. 2010 Pretreatment al., 2011 Gas WI: Kang and Combusted in Post, 2012 Industrial Boiler Regional MER, 2011; N loss via MI: Zhang, et Hydrolysis Humbird et emission EPA, 2010a sediment al. 2010 al., 2011 factors for and WI: Kang and electric power EPA, 2010b Post, 2012 Regional ANL, 2010 N loss via MI: Zhang, et Fermentation Humbird et emission lateral al. 2010 al., 2011 factors for subsurface WI: Kang and transportation flow Post, 2012 and process- heating fuels Power from PE, 2006 N loss via MI: Zhang, et Cellulase Humbird et Nuclear percolation al. 2010 Production al., 2011 Power Plant WI: Kang and Post, 2012 169

Power from PE, 2006 P loss via MI: Zhang, et Lignin Humbird et Coal surface runoff al. 2010 Combustion al., 2011 WI: Kang and and Utilities Post, 2012 Power from PE, 2006 P loss via MI: Zhang, et Wastewater Humbird et Natural Gas sediment al. 2010 Treatment al., 2011 WI: Kang and Post, 2012 Power from PE, 2006 P loss via MI: Zhang, et Distillation, Humbird et Hydropower percolation al. 2010 Dehydration al., 2011 WI: Kang and and Post, 2012 Denaturation Power from PE, 2006 Net Primary MI: Zhang, et Storage and Humbird et Wind Production al. 2010 Handling al., 2011 WI: Kang and Post, 2012 C content of MI: Zhang, et yield al. 2010 WI: Kang and Post, 2012 CO2 respired MI: Zhang, et al. 2010 WI: Kang and Post, 2012 Soil C lost MI: Zhang, et al. 2010 WI: Kang and Post, 2012 N2O MI: Zhang, et emissions al. 2010 from soils WI: Kang and Post, 2012

170

CHAPTER 4

PART 2.

Life Cycle Assessment of Corn Stover Cellulosic Ethanol Production in the Wisconsin and

Michigan Agricultural Contexts: The interplay of spatial variability and allocation decisions

4.8. ADDITIONAL OBJECTIVES

This text gives the additional information on the corn stover (CS) cellulosic ethanol production

LCA conducted in tandem with the switchgrass ethanol LCA. All of the methods, results and conclusions specific to the CS ethanol LCA are detailed herein. The primary variables that set this work apart from the SG work (other than the feedstock) are the increased number of agricultural management practices (rotation and tillage) and the inclusion of more complex allocation between corn grain and corn stover. This study builds on the findings of chapter two on the importance of allocation decisions in LCA by combining it with spatial variability. The objective of this study, like the SG study, is to determine the life cycle environmental impacts of the production of cellulosic ethanol with a spatially-explicit LCA. In addition to this, this research aims to compare the effects of additional variables (agriculture management practices and allocation choices) on the final metrics and the ability of CS ethanol to meet standards.

4.9. ADDITIONAL METHODS

To conduct this research, a cradle-to-refinery-gate LCA of the CS cropping systems and ethanol production in both Wisconsin and Michigan was performed. This LCA focused on biomass and ethanol production in the same RIMAs as the SG ethanol LCA. Life Cycle Inventory (LCI) data 171

were collected on corn (Zea mays) production for grain and stover (including standard deviations

from watershed mean modeled values due to spatial variation in crop production) for 32 different

ethanol production scenarios. Then, these data were used to run LCA models in GaBi 4.4 with

three different allocation methods to apportion the burdens of crop production between the corn

grain (primary economic product) and the stover (co-product of grain production). Finally,

sensitivity analysis was conducted on all 96 scenarios using the spatial standard deviations

(StDevs) to test the overall change in impact category metrics. The scenarios varied by nitrogen

application rate (high and medium), location (WI and MI), crop rotation (continuous corn and

corn-soybean), tillage type (chisel till and no till), pretreatment type (AFEX and DA) and

allocation or allocation avoidance method (mass (M), energy (E) and sub division (SD)). Full

lists of all cropping system combinations and ethanol production scenarios are given in

appendices 1 and 2. The methods for LCA modeling and data sources for this LCA are the same as for the SG ethanol LCA except where noted below. Key LCA components, scenario variables and N application rates are summarized in Tables 4.6., 4.7. and 4.8. respectively.

172

Table 4.6. CS ethanol production LCA summary table

Primary product Corn Stover Functional Unit Energy content, MJ Cellulosic Ethanol Allocation points Corn grain System Cradle to Refinery Gate: Production and Corn Boundary: of agricultural inputs (N, P and K Stover, included unit fertilizer, lime, pesticide, seed, fossil Ethanol and processes fuels and power), agricultural field Electricity techniques (chisel or no-till tiller, field Allocation Sub sdivision cultivator, spike harrow, fertilizer avoidance applicator, row planter, harvester and method baler), transportation, ethanol Allocation Mass ratio production (pretreatment, hydrolysis, methods and energy fermentation, distillation, dehydration, ratio denaturation, enzyme and chemical production, wastewater treatment, co- product combustion, storage and feedstock handling) Impact Net Energy Ratio: energy category output/energy input metrics Global Warming Potential Intensity: -1 gCO2eq MJ of fuel Acidification Potential: gSO2eq MJ of fuel-1 Eutrophication Potential: gPO4eq MJ of fuel-1

Table 4.7. CS ethanol production scenario variables Allocation Nitrogen Method Crop Tillage Pretreatment Landscape Application (between Rotation Type Method Rate grain and stover) Continuous Chisel Mass or WI or MI Medium or Dilute Acid or Corn or Till or Energy or High AFEX Corn-Soy No Till Sub division

173

Table 4.8. CS production nitrogen application rates (all units are kgN ha-1) Scenario Wisconsin Michigan Wisconsin Michigan Rotation Continuous Corn Tillage Chisel Chisel No Till No Till High N 185 175 185 175 Med N 148 125 148 125 Rotation Corn-Soy Tillage Chisel Chisel No Till No Till High N 155 135 155 135 Med N 124 110 124 110 Baseline scenario: Corn-Soy, Chisel Till, High N High N 155 135 - -

4.9.1. Functional unit

The functional unit for this analysis is one unit of energy (MJ) from the final ethanol fuel (Low

Heating Value, LHV). This unit was selected because it allows for direct comparison with other transportation fuels, co-product energy credits and internal comparisons between the energy yield from ethanol and the initial energy inputs from feedstocks (as concluded in chapter 3). The

LHV of CS is 16.5 MJ (kg dry CS)-1 (Domalski et al., 1986).

4.9.2. Analysis, geographic and temporal system boundaries

The system boundary for this analysis begins with the upstream production of agricultural inputs

and ends at the biorefinery gate with fuel-grade ethanol (Figure 4.10). Agricultural inputs

include: nitrogen, phosphorous and potassium fertilizer, agricultural lime, pesticide (Round-up

®), power, fossil fuels, and seed. This boundary therefore includes the entire agricultural

production of CS and corn grain. Ethanol production inputs include: biomass, fossil fuels, power,

chemicals and enzymes. The geographic and time system boundaries of this LCA are exactly the

same as the SG ethanol LCA (Figure 4.1.).

174

Figure 4.10. Process flow and system boundary diagram for feedstock production and cellulosic ethanol production (solid lines are material flows, dashed lines are internal 175

biorefinery heat and power flows, rectangles are stationary unit processes, rhombuses are mobile unit processes that depend on distance or area and octagons are products).

4.9.3. Data sources, assumptions and sensitivity analysis

Biomass yield, grain yield and emissions from soils (to air and water) were modeled with the

EPIC model for the two RIMAs in the same manner as the SG production. Therefore, the input amount assumptions for the application of nitrogen, phosphorus, potassium and lime were taken from the EPIC model input assumptions to maintain consistency between the modeling efforts.

While N rates were constant for each scenario, liming rates were calculated based on crop requirements and soil pH for each pixel and the averaged for each watershed (StDevs were provided with these data). Phosphorous, potassium and pesticide application rates for all

Wisconsin scenarios were 30 kgP ha-1, 40 kgK ha-1 and 1 kgAI pesticide ha-1 and for all

Michigan scenarios they were 24 kgP ha-1, 34 kgK ha-1 and 1 kgAI pesticide ha-1. All data from

EPIC were spatially averaged over each watershed and the standard deviations from this spatial

mean were used in the sensitivity analyses in the same way they were used in the SG ethanol

LCA. The CS harvest index was 50%. The same sources of EPIC model uncertainty noted in part

one of this chapter apply in this CS analysis. The EPIC model, however, has been tested for its

ability to predict corn yields with field trial data from a research station in the WI RIMA (Wang

et al., 2012). Therefore, we would expect less uncertainty in the CS model outputs. The effects of

EPIC model uncertainty, however, were not tested in this LCA.

Life Cycle Inventory (LCI) data for the upstream production of inputs to the agricultural system

were taken from the same sources as the SG ethanol LCA. The additional inputs to the CS

ethanol LCA were phosphorus, pesticide and field operations (chisel tiller or no till planter, field 176 cultivator, spike harrow, fertilizer applicator, row planter, hay harvester, baler). The sources of these additional data are summarized Supplemental Table 4.2. at the end of this chapter.

All of the pretreatment and ethanol production assumptions and data sources for this CS ethanol

LCA are the same as those used for the SG ethanol LCA. The primary difference between the ethanol production portion of the SG and CS ethanol LCAs is the relative concentrations of hexoses, pentoses and lignin in the two feedstocks. For CS, these data were taken from the

NREL cellulosic ethanol production process design study (Humbird et al., 2011).

The same market pulse assumption is made in this CS ethanol LCA since CS is not currently collected in-mass at the commercial scale in either of the study areas. Therefore, this analysis evaluates direct land use change (DLUC) in the same way as a change from the same baseline conventional corn-soybean, chisel tillage and high nitrogen application scenario. It was assumed that existing agricultural land in a baseline crop production system without CS removal was converted to corn production with CS removal. Biotic and abiotic carbon was also tracked throughout this study.

4.9.4. Allocation

In the corn production scenarios, corn grain is the primary product and corn stover is the co- product. The sub division method was used to assign only the life-cycle burdens which result directly from stover production and harvest to the CS as was described in Chapter 2 of this dissertation and the published paper that resulted from that chapter (Kaufman et al., 2010). For comparison, allocation of burdens between CS and corn grain was also conducted by mass and energy content. For the soybean production in the corn-soy rotation, all processes attributed only 177

to soy production were sub-divided out of the system boundary. Electricity generated from co-

product lignin combustion was accounted for in the same way as it was in the SG ethanol LCA.

4.10. RESULTS AND DISCUSSION

The results of this study are presented in three sections. The results of the cropping system

portion of the LCA are presented first and are followed by the results of the full corn stover

ethanol production LCA. The results of the sensitivity analysis with results from variations in

space and allocation method are presented last.

4.10.1. Spatial variation and cropping system results

Figures 4.11. through 4.15. present maps of the results of the crop production LCA (corn stover

yield, NER, GWP Intensity, AP and EP) without allocation between corn grain and stover (just the net effects of production). Full-sized maps can be found in Appendix 6. Abbreviations for scenarios are explained in Table 4.9. Table 4.10 gives the overall results of the crop production

LCA. The scenario labels appear in the following order: State, Crop Rotation, Tillage, and

Nitrogen Application. For example, MIContCornCTMN represents the Michigan, Continuous

Corn, Chisel Tillage and Medium Nitrogen scenario. Also, the same distinction between actual and potential effects noted in part one of this chapter applies here. Note that, the GWP Intensity,

AP and EP results are all presented as a change from the baseline crop production scenario to account for DLUC, therefore, these results are do not represent the environmental impacts of CS harvesting alone. They represent a change in current crop production practices in these RIMAs.

Table 4.9. CS feedstock production LCA variables and abbreviations.

Variable Type Abbreviation State Wisconsin WI Michigan MI Crop Rotation Continuous Corn ContCorn 178

Corn-Soybean CornSoy Tillage Chisel Tillage CT No Till NT Nitrogen Application Medium Nitrogen MN High Nitrogen HN

Figure 4.11. Corn stover yields (kg CS dry matter ha-1) across all scenarios. Red to green colors indicate poor to good yield outcomes.

179

Figure 4.12. Corn stover production Net Energy Ratio (MJ output MJ input-1) results across all scenarios. Red to green colors indicate poor to good environmental outcomes. 180

-1 Figure 4.13. Corn stover production Global Warming Potential (kgCO2eq ha ) results across all scenarios as a change from the baseline crop production scenario. Red to green colors indicate poor to good environmental outcomes. 181

-1 Figure 4.14. Corn stover production Acidification Potential (kgSO2eq ha ) results across all scenarios as a change from the baseline crop production scenario. Red to green colors indicate poor to good environmental outcomes. 182

-1 Figure 4.15. Corn stover production Eutrophication Potential (kgPO4eq ha ) results across all scenarios as a change from the baseline crop production scenario. Red to green colors indicate poor to good environmental outcomes.

183

Table 4.10. Average, range and StDev of CS production metrics across the RIMA

watersheds for CS production only and calculated as the change from the baseline crop

production scenario (Net Energy Ratio, MJ output MJ input-1; Global Warming Potential,

-1 -1 gCO2eq MJ ; Acidification Potential, gSO2eq MJ and Eutrophication Potential, gPO4eq

MJ-1). These values do not include allocation because the ending boundary of crop

production is the farm gate not the point of biomass use.

Scenario WIContCornNTHN WIContCornNTMN

Average Min Max StDev Average Min Max StDev

NERa 5.83 4.94 6.37 0.38 5.83 4.94 6.37 0.38 GWP 564 131 1550 278 564 131 1550 278 AP 1.50 1.37 1.76 0.11 1.50 1.37 1.76 0.11 EP 5.92 1.59 18.1 3.36 0.01 0.00 0.02 0.004 Scenario WIContCornCTHN WIContCornCTMN

NERa 0.11 0.00 0.50 0.11 5.40 4.30 5.90 0.36 GWP 695 239 2170 355 601 146 2100 351 AP 1.80 1.64 2.30 0.14 1.64 1.50 2.06 0.12 EP 11.8 5.31 37.8 6.38 10.3 4.43 35.3 6.27 Scenario WICornSoyNTHN WICornSoyNTMN

Average Min Max StDev Average Min Max StDev

NERa 8.32 6.62 34.3 3.94 7.21 5.32 8.66 0.79 GWP 12.7 -330 97.4 81.3 -22.2 -518 64.3 110 AP 1.12 0.22 1.33 0.15 1.25 1.03 1.68 0.14 184

EP -1.88 -8.97 -0.28 1.76 -0.13 -4.49 0.82 1.04 Scenario WICornSoyCTHN WICornSoyCTMN

NERa 5.42 4.09 6.21 0.46 5.52 4.05 6.62 0.60 GWP 273 161 820 145 262 129 887 162 AP 1.64 1.42 2.16 0.15 1.63 1.34 2.20 0.19 EP 3.82 0.85 15.8 3.25 4.65 0.68 20.4 4.12 Scenario MIContCornNTHN MIContCornNTMN

Average Min Max StDev Average Min Max StDev

NERa 5.38 4.54 6.10 0.36 4.28 3.56 4.91 0.28 GWP 501 129 1480 410 658 124 2370 621 AP 1.64 1.44 1.93 0.11 2.04 1.77 2.44 0.14 EP 8.38 5.52 12.8 1.40 13.4 9.41 20.0 2.23 Scenario MIContCornCTHN MIContCornCTMN

NERa 6.85 4.86 19.0 2.05 5.52 3.47 15.67 1.74 GWP 483 78.2 1770 471 588 62.6 2180 641 AP 1.32 0.46 1.79 0.18 1.64 0.55 2.49 0.26 EP 7.86 2.74 13.1 1.85 13.2 4.24 20.0 2.91 Scenario MICornSoyNTHN MICornSoyNTMN

Average Min Max StDev Average Min Max StDev

NERa 6.69 4.50 19.0 2.09 7.28 4.03 57.8 8.51 GWP 170 75.3 348 48.4 222 18.3 571 94.4 AP 1.37 0.46 1.95 0.20 1.50 0.15 2.17 0.31 EP 1.00 0.39 1.73 0.22 1.38 0.11 2.13 0.35 185

Scenario MICornSoyCTHN MICornSoyCTMN

NERa 7.89 5.33 22.3 2.46 6.84 4.60 19.4 2.14 GWP 157 57.2 240 36.8 219 88.2 470 67.2 AP 1.16 0.39 1.64 0.17 1.33 0.45 1.89 0.20 EP 0.79 0.22 1.08 0.18 1.21 0.33 1.94 0.32 a Here the NER is calculated with the energy output in the corn stover biomass.

By examining Figure 4.11, it becomes apparent that the potential for high corn stover yields is greater in WI than in MI. This is due in part to the rich and thick soils of southern WI. While both RIMAs have alfisols and mollisols as their dominant soil orders (known for being high nutrient and prairie soils), the Cation Exchange Capacity (CEC) of the WI Soils is generally higher in the major agricultural regions of the RIMA than that of MI (see soils CEC and pH maps in Appendix 9) (Brady and Weil, 2004 and NRCS, 2008). The effect of a low CEC on even a high nutrient soil is a decreased availability of nutrients to microorganism and plant root structures (Brady and Weil, 2004). Among the various agricultural production treatments

(rotation, nitrogen application rate and tillage), the highest yields occur in WI the continuous corn, high N and chisel till scenario, though the corn-soy, medium N and no till performs better in MI. This sort of watershed-scale yield mapping could aid in determining the most productive locations for corn stover harvest. Unlike with switchgrass, corn for grain production is more likely to already be occurring in prime agricultural areas of both WI and MI. Therefore, corn stover removal for ethanol production could be more feasible than a transition to switchgrass production in these highly productive areas.

Figure 4.12 shows the great spatial variation in NER across states and production methods. The high nitrogen application scenarios tend to have higher NERs than the medium nitrogen 186

scenarios. This shows that, even though the additional nitrogen production increases the energy

life cycle energy inputs to the system, the agricultural yield return on this input investment could

be worthwhile (provided that it does not severely worsen other environmental impacts). Still, this

yield bump from high nitrogen application does not hold for all watersheds. This type of

watershed scale mapping could aid in identifying areas that would and would not benefit from

higher nitrogen applications. For example, the NER in the MI, corn soy rotation, medium

nitrogen application and no till scenarios are high even with medium N application. This could

be due in part to the additional nitrogen fixed by soybeans in the rotation. Table 4.10 provides

additional information on the range of observed NER values among the watersheds in the two

RIMAs. In general, the StDevs do not depart drastically from the mean observed NER which

indicates that NER does not vary as greatly among the watersheds as the other impact categories.

The main driver of differences in NER is differences in yield which, as noted above, hinge on

soil quality and are location-specific.

Figure 4.13 indicates that agricultural management practices can influence GWP greatly and in noticeably different ways depending on location. For example, the change from the baseline scenario (corn-soy, high N and chisel till) to continuous corn production in WI results in higher

GWP per area than it does in MI. This could be due in part to the existing high soil carbon content of WI soils compared to MI soils. That is to say that if the soil has more carbon to begin with, then it has more to lose with the implementation of practices that disturb soil structure like chisel tillage than soils with lower beginning soil carbon content. This is yet another point of interest when selecting sites for feedstock production and determining the type of agricultural practices that will strike the delicate balance between high yields and low emissions. 187

One somewhat counter-intuitive result that can be observed in Figure 4.13 is that, in some areas,

higher N application actually resulted in lower GWP than medium N application. This is due in

part to the increased yield gained with increased N application and it appears to counteract the

increased N2O emissions from N fertilizer production and soil emissions. This result suggests

that, while increased nitrogen application is known to increase N2O emissions, there is a point at

which the boost that this nitrogen gives to the plants’ productivity (and therefore carbon

absorption) is worth the incremental addition of nitrogen. In addition to this, there appear to be

clear benefits to switching from the baseline corn production scenario to a corn-soy rotation in

WI regardless of tillage or nitrogen application rate. Table 4.10 shows a large range of values

across the watershed in all scenarios normalized by energy output (MJ from CS) rather than by

area as different way of viewing the data. Some scenarios even show StDevs close to the mean

GWP (e.g. MIContCornNTMN). This provides further evidence that GWP from crop production

can be location-dependent just as much as it can be dependent on management practice.

Nevertheless, the data in Table 4.10 indicate that the corn-soy rotation provides much lower

GWP than the continuous corn management regardless of other management variables.

Figure 4.14 demonstrates that there is much more spatial uniformity in AP across WI than there

is in MI. As noted earlier, WI has higher CEC than MI and CEC plays an important role in soil buffering capacity, soil pH and the loss of potentially acidifying molecules from soils (Brady and

Weil, 2004). The lack of discernible variation in AP across WI indicates that the majority of AP stems from the combustion of fuels (primarily diesel) in on-farm equipment for field operations.

The chisel till production scenario did require two additional field operations (field cultivator and seed drill) not required in the no till scenarios since no till equipment prepares a channel for seeds, drops the seeds in and closes the channel all in one pass. The combustion of additional 188 diesel in chisel till management, as compared to the no till management, therefore adds to the AP of chisel tills scenarios. This pattern of elevated AP with chisel till versus no till continues data presented in Table 4.10. The StDev of AP does depart more from the mean in MI than in WI and the range of values are greater in MI than in WI. Still, compared to GWP, the AP does not span as wide of a range. Therefore, spatial variation considerations may not be as crucial to the accurate calculation of AP as it is for other environmental impacts.

The variation in EP across locations and agricultural practices is illustrated in Figure 4.15.

Again, there are noticeable differences between the WI and MI RIMA which are likely related to

CEC and pH. The cation and anion exchange reactions that occur on the surface of soil colloids affect the ability of organic and inorganic molecules to move through the soils and be accessed by plants and microbes. Anion Exchange Capacity (AEC) is inversely related to CEC, but dependent upon pH. Soils with medium CEC that are slightly acidic will have a high AEC and will have improved control of the movement of anions (like nitrates and phosphates) though soils and be able to prevent them from entering groundwater. Furthermore, if nitrogen and phosphorus anionic compounds remain adsorbed to soil colloids longer, this increases the likelihood that they will be broken down by soil microbes into forms that are usable by plants (Brady and Weil,

2004). The crop production EP tends to be lowest in the northwest corner of the WI RIMA where this unique combination of CEC and pH occurs (Appendix 9). The occurrence of some negative

EP, such as in the corn-soy, medium N, no till scenarios, indicates that potential for eutrophication in this management scenario is greatly reduced when the land use is changed from the baseline corn production scenario. The effect of rotation, tillage and nitrogen application rate is also evident in Table 4.10. The continuous corn, high N and chisel till scenarios produce the highest EP, while corn-soy, medium N and no till produce the lowest EP. The range of values 189

and StDevs are also reduced in the corn-soy, medium N and no till scenarios. This indicates that more intense agricultural practices may exhibit more spatial variability than less intensive practices.

4.10.2. Ethanol production results

All scenario labels for the full CS ethanol LCA are explained in Table 4.11. Table 4.12 and

Figures 4.16 through 4.23 present the full ethanol LCA results for NER, GWP, AP and EP.

Labels appear in the following order: State, Tillage, Nitrogen Application, Pretreatment,

Allocation. For example, MICTMN AFEX SD stands for the Michigan, Chisel Tillage, Medium

Nitrogen, Ammonia Fiber Expansion and Sub Division scenario. The crop rotation will be

specified in the figure caption, column heading or legend.

Table 4.11. CS ethanol production LCA variable and abbreviation table.

Variable Type Abbreviation State Wisconsin WI Michigan MI Crop Rotation Continuous Corn ContCorn Corn-Soybean CornSoy Tillage Chisel Tillage CT No Till NT Nitrogen Application Medium Nitrogen MN High Nitrogen HN Pretreatment Ammonia Fiber AFEX Method Expansion Dilute Acid DA Allocation Method Mass M Energy E Sub Division SD

190

Table 4.12. Results for impact category metrics across all CS ethanol production scenarios

-1 -1 (Net Energy Ratio, MJ output MJ input ; Global Warming Potential, gCO2eq MJ ;

-1 -1 Acidification Potential, gSO2eq MJ and Eutrophication Potential, gPO4eq MJ )

ContCorn NER GWP AP EP ContCorn NER GWP AP EP WICTMN 5.08 55.04 0.041 0.10 MICTMN 5.42 56.06 0.046 0.13 AFEX M AFEX M WICTMN 4.98 59.36 0.055 0.17 MICTMN 4.85 59.30 0.058 0.19 AFEX E AFEX E WICTMN 6.92 49.85 0.025 0.02 MICTMN 6.94 50.10 0.025 0.02 AFEX SD AFEX SD WICTHN 5.78 55.64 0.037 0.11 MICTHN 5.11 57.63 0.052 0.20 AFEX M AFEX M WICTHN 5.46 60.71 0.052 0.19 MICTHN 4.48 61.67 0.066 0.30 AFEX E AFEX E WICTHN 7.21 49.56 0.019 0.01 MICTHN 6.91 50.20 0.025 0.03 AFEX SD AFEX SD WINTMN 5.78 55.10 0.040 0.07 MINTMN 5.73 55.50 0.041 0.12 AFEX M AFEX M WINTMN 5.08 59.50 0.052 0.12 MINTMN 5.19 58.69 0.051 0.18 AFEX E AFEX E WINTMN 6.93 49.86 0.025 0.02 MINTMN 6.97 50.08 0.025 0.02 AFEX SD AFEX SD WINTHN 5.78 55.12 0.040 0.07 MINTHN 5.45 56.27 0.046 0.19 AFEX M AFEX M WINTHN 5.08 59.53 0.052 0.12 MINTHN 4.84 59.84 0.058 0.29 AFEX E AFEX E WINTHN 6.93 49.86 0.025 0.02 MINTHN 6.94 50.13 0.025 0.02 AFEX SD AFEX SD WICTMN 7.03 60.72 0.057 0.11 MICTMN 6.55 61.92 0.063 0.15 DA M DA M WICTMN 5.82 65.70 0.073 0.19 MICTMN 5.61 65.65 0.077 0.22 DA E DA E WICTMN 9.36 54.74 0.038 0.02 MICTMN 9.42 55.05 0.038 0.02 DA SD DA SD WICTHN 7.12 61.47 0.054 0.13 MICTHN 6.04 63.73 0.070 0.23 DA M DA M WICTHN 5.77 67.32 0.071 0.22 MICTHN 5.06 68.38 0.086 0.34 DA E DA E WICTHN 9.88 54.45 0.033 0.01 MICTHN 9.35 55.15 0.039 0.03 DA SD DA SD WINTMN 7.16 60.79 0.055 0.08 MINTMN 7.07 61.27 0.057 0.13 DA M DA M 191

WINTMN 5.98 65.86 0.070 0.13 MINTMN 6.16 64.95 0.068 0.20 DA E DA E WINTMN 9.38 54.74 0.038 0.02 MINTMN 9.47 55.02 0.038 0.02 DA SD DA SD WINTHN 7.16 60.81 0.055 0.08 MINTHN 6.21 52.43 0.065 0.22 DA M DA M WINTHN 5.98 65.89 0.070 0.13 MINTHN 5.33 56.54 0.079 0.33 DA E DA E WINTHN 9.38 54.74 0.038 0.02 MINTHN 8.66 45.34 0.041 0.03 DA SD DA SD CornSoy NER GWP AP EP CornSoy NER GWP AP EP WICTMN 5.77 51.71 0.040 0.05 MICTMN 5.56 52.08 0.044 0.03 AFEX M AFEX M WICTMN 5.05 53.49 0.053 0.07 MICTMN 4.97 53.40 0.055 0.04 AFEX E AFEX E WICTMN 6.93 49.62 0.025 0.02 MICTMN 6.95 49.87 0.025 0.01 AFEX SD AFEX SD WICTHN 5.73 51.91 0.041 0.04 MICTHN 5.48 52.71 0.045 0.03 AFEX M AFEX M WICTHN 5.00 53.84 0.054 0.06 MICTHN 4.87 54.38 0.057 0.04 AFEX E AFEX E WICTHN 6.93 49.64 0.025 0.02 MICTHN 6.95 49.91 0.025 0.02 AFEX SD AFEX SD WINTMN 6.02 49.66 0.036 0.02 MINTMN 5.92 51.67 0.038 0.02 AFEX M AFEX M WINTMN 5.40 49.82 0.046 0.02 MINTMN 5.43 52.74 0.046 0.03 AFEX E AFEX E WINTMN 6.95 49.48 0.024 0.01 MINTMN 6.99 49.85 0.024 0.01 AFEX SD AFEX SD WINTHN 6.05 49.85 0.036 0.00 MINTHN 5.79 52.37 0.040 0.03 AFEX M AFEX M WINTHN 5.45 50.15 0.045 - MINTHN 5.26 53.85 0.049 0.04 AFEX E 0.01 AFEX E WINTHN 6.96 49.49 0.024 0.01 MINTHN 6.98 49.89 0.025 0.02 AFEX SD AFEX SD WICTMN 7.14 56.88 0.056 0.05 MICTMN 6.78 57.33 0.060 0.03 DA M DA M WICTMN 5.93 58.93 0.071 0.08 MICTMN 5.80 58.85 0.074 0.04 DA E DA E WICTMN 9.37 54.47 0.038 0.02 MICTMN 9.44 54.78 0.038 0.01 DA SD DA SD WICTHN 7.07 57.10 0.057 0.05 MICTHN 6.64 58.05 0.062 0.04 DA M DA M WICTHN 5.85 59.34 0.072 0.07 MICTHN 5.65 59.98 0.076 0.05 DA E DA E 192

WICTHN 9.37 54.49 0.038 0.02 MICTHN 9.42 54.82 0.038 0.01 DA SD DA SD WINTMN 7.59 54.52 0.052 0.02 MINTMN 7.42 56.85 0.054 0.02 DA M DA M WINTMN 6.51 54.70 0.063 0.02 MINTMN 6.56 58.09 0.063 0.03 DA E DA E WINTMN 9.43 54.31 0.038 0.01 MINTMN 9.51 54.75 0.038 0.01 DA SD DA SD WINTHN 7.65 54.73 0.051 0.00 MINTHN 7.19 57.66 0.056 0.03 DA M DA M WINTHN 6.59 55.08 0.062 - MINTHN 6.28 59.37 0.066 0.04 DA E 0.01 DA E WINTHN 9.43 54.32 0.038 0.01 MINTHN 9.48 54.80 0.038 0.01 DA SD DA SD

193

Figure 4.16. Net Energy Ratio versus Global Warming Potential Intensity for the

Wisconsin continuous corn scenarios with allocation methods.

Figure 4.17. Eutrophication Potential versus Acidification Potential for the Wisconsin continuous corn scenarios with allocation methods. 194

Figure 4.18. Net Energy Ratio versus Global Warming Potential Intensity for the Michigan continuous corn scenarios with allocation methods.

Figure 4.19. Acidification Potential versus Eutrophication Potential for the Michigan continuous corn scenarios with allocation methods.

195

Figure 4.20. Net Energy Ratio versus Global Warming Potential Intensity for the

Wisconsin corn soybean scenarios with allocation methods.

Figure 4.21. Acidification Potential versus Eutrophication Potential for the Wisconsin corn soybean scenarios with allocation methods. 196

Figure 4.22. Net Energy Ratio versus Global Warming Potential Intensity for the Michigan corn soybean scenarios with allocation methods.

Figure 4.23. Acidification Potential versus Eutrophication Potential for the Michigan corn soybean scenarios with allocation methods. 197

The ranges of values shown in Table 4.12 indicate significant differences in the overall environmental impacts of both the production scenarios and the allocation methods studied. For example, the highest NER found in all 96 scenarios occurs in the WIContCornCTHN DA SD scenario (9.88) while the lowest NER occurs in the MIContCornCTHN AFEX E scenario (4.48).

These two scenarios have the same rotation, tillage type and high N rate, but differ by location

(RIMA), pretreatment method and allocation method. The AFEX pretreatment method uses more of the energy produced from lignin combustion than the DA pretreatment method because of the added ammonia recovery process. This lowers the NER for AFEX scenarios since less electricity is available to be exported to the grid. The SD allocation method also limits the overall impact of the feedstock production on the full ethanol production LCA because it apportions most of the production burden to corn grain (the primary agricultural product of the system). The E allocation method divides the burdens between the corn grain and the corn stover almost equally, while the M allocation method shifts the burdens towards the corn grain, but not by as much as the SD method. Thus, any scenario with this combination of DA pretreatment and SD allocation is likely to have a higher NER than its counterpart with AFEX pretreatment and E allocation

The stark contrast between these two scenarios also illustrates the differences in yields between the two RIMAs. Just as the crop production portion of the LCA showed higher yields in WI versus MI, the full ethanol LCA carries that effect forward into the final metrics. In fact, the

NER range for the WI RIMA across all production and allocation scenarios is slightly higher overall than the NER range for MI RIMA (4.98 to 9.88 for WI and 4.48 to 9.51 for MI). This provides further evidence to support the notion that yield is a primary driver of overall energy efficiency in the ethanol production system. The primary input contribution to energy use in the ethanol production system was gasoline production for the ethanol fuel denaturant. The chemical 198 inputs to cellulase production and pretreatment combined roughly equaled the energy use contribution of the gasoline which is in line with the conclusion of MacLean and Spatari (2009) that these inputs are significant and should not be omitted from biofuels production LCA.

Like the NER results, the primary contributors to GWP in the form of abiotic emissions came from the production of chemicals used in the biorefinery. This contribution however hinged on the allocation method used. For the M allocation, the combined GWP effect of the process chemicals equal that of the corn stover production, but in the case of the SD method, the GWP effect of the process chemicals exceeded that of the corn stover production. In contrast to the

NER results, both the highest and the lowest GWP occur in the MI RIMA (68.38 and 45.34

-1 gCO2eq MJ , respectively) (Table 4.12). In this case, however, the difference hinges on the allocation method and the tillage type, not the pretreatment method. The highest GWP occurs in the MIContCornCTHN DA E scenario and the lowest occurred in the MIContCornNTHN DA

SD scenario. While the no till management did have a beneficial effect on the GHG emissions from soils as compared to the baseline scenario, the allocation method was the primary reason for this difference in GWP. For the MIContCornNTHN DA scenarios with M and E allocation

-1 methods, the GWPs were 54.43 and 56.54 gCO2eq MJ , respectively (Table 4.12). Therefore, when the allocation method reduces the GWP burden apportioned to the corn stover production, the overall GWP of the ethanol production is reduced as well. While this is an intuitive conclusion, it holds true only if the effect of the land use change on GHG emissions is an increase in GHG emissions. Interestingly, if the land use change decreases GHG emissions from soils, but the allocation method attributes more of the burdens to corn grain than to corn stover then the GHG emissions savings would not be passed on to the corn stover ethanol. The SD method severely limits the GHG emissions impacts of agricultural management practices on the 199

final ethanol production LCA. Management practices that reduce GHG emissions are not

rewarded and practices that increase GHG emissions are not penalized. From a policy standpoint,

if an allocation method like sub division is mandated for the ethanol production LCA, then this

policy would actually reduce the incentive to implement agricultural management practices that

reduce GHG emissions.

The lowest and highest APs occurred in the WIContCornCTHN AFEX SD and

-1 MIContCornCTHN DA E scenarios with 0.019 and 0.086 gSO2eq MJ , respectively (Table

4.12). As with the GWP results, the allocation method limits the influence of the agricultural

management practices. The combustion of fossil fuels, like diesel, is a large contributor to AP.

The chisel till management practice requires additional field operations not required in no till

management and, thus, results in more diesel combustion. Again, the SD method limits this

effect of agricultural management practices on the overall ethanol production LCA by reducing

the AP burdens apportioned to corn stover. For example, the WIContCornCTHN AFEX SD scenario may have the lowest AP of all the scenarios, but changing the allocation method to E

-1 almost triples the AP to 0.052 gSO2eq MJ (Table 4.12). This result has the same policy

implications noted for GWP.

Just as it had the highest GWP and AP of all the scenarios, the MIContCornCTHN DA E

-1 scenario had the highest EP as well (0.34 gPO4eq MJ , Table 4.12). The lowest EP occurred in

- both the WICornSoyNTHN AFEX E and WICornSoyNTHN DA E scenarios (-0.01 gPO4eq MJ

1, Table 4.12). In this impact category metric, the movement of N and P via leaching and runoff

during crop production was a large contributor to the total EP. Location, rotation and tillage type

all play an important role in the loss of eutrophying compounds from soils. No till management reduces the soil disturbance which reduces the breakup of soil aggregates. No till also reduces 200

the exposure of the colloids that make up soil aggregates to the air which can accelerate the

breakdown of organic colloids (colloids made of organic matter rather than clay minerals or

hydrous oxides). Cation and anion adsorption occurs on the surface of soil colloids, therefore, the

more colloids there are, the more surface area there is to hold cations and important eutrophying

- -3 anions such as nitrate (NO3 ) and phosphate (PO4 ) (Brady and Weil, 2004). Therefore, the combination of soil aggregate destruction and colloid exposure from chisel tillage with the additional nitrogen source from high nitrogen application (which is even higher in the continuous corn rotation than it is in the corn soy rotation) can increase the EP. Once again, the negative EP indicates that effect of DLUC is a decrease in EP and allocation choice is paramount.

Just as with the GWP and the AP, the allocation method chosen affects the crop production

benefits and burdens apportioned to corn stover, and therefore, can change the overall

assessment of the ethanol production. In the WICornSoyNTHN AFEX E and WICornSoyNTHN

DA E scenarios where the EP was negative, the EP using the SD method was positive (0.01

-1 gPO4eq MJ , Table 4.12). Moreover, in the MIContCornCTHN DA E EP drops by 91% when

the allocation method changes from E to SD. Thus, the same policy implications of allocation

choice on agricultural management practices arise with EP as with GWP and AP.

Examination of Figures 4.16 through 4.23 reveals that corn stover ethanol production results for

NER, GWP, AP and EP tend to cluster in groups by allocation method. In fact, some of the

points on these graphs overlap. In the case of NER versus GWP (Figures 4.16, 4.18, 4.20 and

4.22), all of the SD scenarios are located in the preferable region of high NER and low GWP

(bottom to middle right corner), while the E scenarios are located in the lower NER and higher

GWP region (top to middle left corner) with the M scenarios nestled between SD and E.

Therefore, comparisons of the overall impact of cellulosic ethanol production are more 201 appropriate within allocation method and across other production variables. Within the E allocation method, the DA pretreatment tends to have higher NER than the AFEX pretreatment method, but this comes at the expense of increased GWP. Within E allocation method, the corn-

-1 soy scenarios tend to have lower GWP (by as much as 20 gCO2eq MJ ) and NERs (by as much as 2.0). These differences also fall out between the WI and MI RIMAs within the E allocation, where the WI corn soy scenarios have higher NER and GWP than the MI continuous corn or corn soy scenarios. No one scenario has the highest NER and GWP combination, but within E allocation, the WICornSoyNTMN AFEX and WICornSoyNTHN DA scenarios have the lowest

GWP and highest NER respectively. For WICornSoyNTMN AFEX E scenario, the NER is 5.40

-1 and the GWP is 49.82 gCO2eq MJ and for the WICornSoyNTHN DA E scenario the NER is

-1 6.59 and the GWP is 55.08 gCO2eq MJ . These types of comparisons aid in the identification of corn stover ethanol production scenarios that optimize NER and GWP environmental impacts.

When EP versus AP is examined (Figures 4.17., 4.19, 4.21 and 4.23), a similar allocation-related pattern emerges. The SD allocation scenarios clump in the low AP and EP region (bottom left corner), while the E allocation scenarios gather in the high AP and EP region (top right corner).

Similar to the NER versus GWP comparison figures, these figures show that, among the E allocation scenarios, the WICornSoy scenarios give the lowest AP and EP, while the

-1 MIContCorn gives the highest AP and EP. These values range from 0.045 to 0.086 gSO2eq MJ

-1 for AP and -0.01 to 0.34 gPO4eq MJ for EP (Table 4.12). The scenario with the highest NER also happens to have the lowest EP (WICornSoyNTHN DA E). This lowest EP was also shared with the scenario that had the lowest AP (WICornSoyNTHN AFEX E). Moreover, the scenario with the worst GWP, AP and EP (among the E allocation scenarios) was the MIContCornCTHN

DA E scenario and this scenario’s NER was 23% less than the highest NER. Similar scenarios in 202

MI with continuous corn and DA are also located in the unfavorable regions of the NER vs.

GWP and EP versus AP figures.

Thus, from the perspective of minimizing all environmental impact, the WI scenarios with corn

soy rotations and no till are the best candidates. If NER is of primary concern, then the WI corn

soy, no till, high nitrogen and dilute acid pretreatment scenario is ideal (and it also minimizes

EP). Whereas, if GWP is of major importance, the WI corn soy, no till, medium nitrogen and

AFEX pretreatment scenario is the most favorable. Neither of these scenarios optimizes AP, but

middle of the road option could be the WI corn soy, no till high nitrogen and AFEX pretreatment

-1 because it provides the best AP and EP without a large increase in GWP (1.33 gCO2eq MJ ) and

a slightly below average NER (5.45). If these results were to be used to choose the production

variables for a cellulosic ethanol biorefinery in MI, then it is obvious that the worst combination

(from an environmental impact perspective) is the MI continuous corn, chisel till, high nitrogen

and dilute acid pretreatment scenario. The lowest GWP, AP and EP all occur in the MI corn soy,

no till, medium nitrogen and AFEX pretreatment scenario (within the E allocation method)

-1 -1 -1 (52.74 gCO2eq MJ , 0.046 gPO4eq MJ and 0.03 gSO2eq MJ , respectively). This scenario has a slightly above average NER for the MI scenarios of 5.43. This scenario’s dilute acid pretreatment counterpart, however, has the highest NER (6.56) and the same EP with only a 5.35

-1 -1 gCO2eq MJ and 0.017 gSO2eq MJ increase in GWP and AP respectively. Therefore, the

examination of many combinations of cellulosic ethanol production variables shows that

different production methods produce more or less favorable environmental outcomes in

different states. This information, combined with economic cost of production and transportation

infrastructure, information could be very useful to companies looking to site a cellulosic ethanol

biorefinery. 203

Statistical analysis of the effect size of the different scenario variables, similar to the analysis

performed on SG ethanol in part one of this chapter, was also conducted on the CS ethanol

production in section (see Appendix 10 for full results). The analysis was conducted within

allocation method and rotation type to test the effect size and interactions between RIMA,

nitrogen application rate, tillage type and pretreatment on all resulting metrics (NER, GWP, AP and EP). In both rotation types and across all allocation methods, the largest effect size for NER,

GWP and AP came from the pretreatment. This is due to a number of factors such as the differences in excess electricity production exported to displace grid electricity and the overall conversion efficiency differences between AFEX and DA (which affects the energy output in total ethanol). In addition to this, there were moderate to strong interaction effects on all metrics

between location (RIMA) and agricultural management variables (tillage and N application rate)

in the continuous corn rotation scenarios across all allocation types. These interaction effects

were stronger in the continuous corn scenarios than they were in the corn soy scenarios. This

suggests that environmentally taxing rotations (e.g. continuous corn) are affected differently by

location, tillage and N application rate than less taxing rotations (e.g. corn soy).

For EP, in the continuous corn rotation type and across all allocation methods, the largest effect

resulted from the strong interaction between RIMA and nitrogen application rate. This was the

second largest effect in the corn soy rotation type across all allocation methods. This is due to the location-specific effects of nitrogen application, since, as discussed earlier, the location-specific soils characteristics play a large role in the movement and transformation of nutrients like nitrogen. For the corn soy rotation type, across all allocation methods, the main effect for EP was tillage. A slightly weaker interaction effect was observed between the RIMA and the tillage type for EP in both rotation types and across all allocation methods. This also ties back into the 204

location-specific ramifications of different agricultural management practices and supports the

notion that different agricultural practices will not result in the same magnitude of environmental

impacts in different locations.

4.10.3. Sensitivity analysis

Figures 4.24 through 4.27 compare the CS ethanol production LCA results for NER, GWP, AP

and EP for all scenarios side by side. These figures also include error bars which indicate the

result of varying all of the crop production data points from EPIC by plus or minus one standard

deviation from the mean value for each watershed. Table 4.13 gives the percent reduction in

GWP Intensity compared to an energy equivalent amount of gasoline if the mean watershed

values are used and if these values are increased or decreased by one StDev from their spatial

mean. Note that the whiskers on the following figures do not represent the traditional statistical

error around the calculation of a mean value. Instead, these whiskers are to be interpreted as a range of possible values caused by spatial variability alone. They do not include a measure of

other sources of variability or uncertainty in the EPIC or LCA models. They indicate scenarios

for each environmental metric when the spatially-explicit crop production inputs to the LCA

model were decreased or increased by one StDev from their spatial mean within a watershed.

This analysis was conducted to quantify the range of environmental impact results that could

arise from producing biomass on more or less productive land within a watershed. This range is

an indicator of how ecological determinants of yield and emissions from crop production vary in

space within a watershed.

Figure 4.24. Corn stover ethanol LCA Net Energy Ratio (MJ output MJ input-1) across all scenarios and allocation methods with a range of values due to spatial variability.

-1 Figure 4.25. Corn stover ethanol LCA Global Warming Potential Intensity (gCO2eq MJ ) across all scenarios and allocation methods with a range of values due to spatial variability. 205

-1 Figure 4.26. Corn stover ethanol LCA Acidification Potential (gSO2eq MJ ) across all scenarios and allocation methods with a range of values due to spatial variability.

-1 Figure 4.27. Corn stover ethanol LCA Eutrophication Potential (gPO4eq MJ ) across all scenarios and allocation methods with a range of values due to spatial variability. 206

207

Table 4.13. Percent reduction in GHG emissions for CS ethanol compared to gasoline (94

-1 gCO2eq MJ ) for all CS ethanol production scenarios and the new percent reduction value if the emissions from soils and yield modeled in each watershed increases or decreases by one StDev.

% % % % % % reduct reduction reduction reduct reduction reduction ion if if ion if if using increased decrease using increased decreased spatial by 1 d by 1 spatial by 1 by 1 ContCorn mean StDev StDev ContCorn mean StDev StDev WICTMN MICTMN 41% 30% 53% 40% 32% 49% AFEX M AFEX M WICTMN MICTMN 37% 18% 56% 37% 25% 48% AFEX E AFEX E WICTMN MICTMN 47% 46% 48% 47% 46% 47% AFEX SD AFEX SD WICTHN MICTHN 41% 36% 45% 39% 29% 48% AFEX M AFEX M WICTHN MICTHN 35% 16% 54% 34% 20% 49% AFEX E AFEX E WICTHN MICTHN 47% 46% 48% 47% 46% 47% AFEX SD AFEX SD WINTMN MINTMN 41% 41% 42% 41% 33% 49% AFEX M AFEX M WINTMN MINTMN 37% 37% 37% 38% 25% 50% AFEX E AFEX E WINTMN MINTMN 47% 47% 47% 47% 46% 47% AFEX SD AFEX SD WINTHN MINTHN 41% 30% 52% 40% 30% 50% AFEX M AFEX M WINTHN MINTHN 37% 18% 56% 36% 22% 51% AFEX E AFEX E WINTHN MINTHN 47% 46% 47% 47% 46% 47% AFEX SD AFEX SD WICTMN MICTMN 35% 23% 47% 34% 25% 44% DA M DA M WICTMN MICTMN 30% 8% 52% 30% 17% 44% DA E DA E WICTMN MICTMN 42% 41% 43% 41% 41% 42% DA SD DA SD WICTHN 35% 21% 48% MICTHN 32% 21% 44%

208

DA M DA M WICTHN MICTHN 28% 5% 52% 27% 10% 44% DA E DA E WICTHN MICTHN 42% 42% 42% 41% 41% 42% DA SD DA SD WINTMN MINTMN 35% 35% 36% 35% 25% 45% DA M DA M WINTMN MINTMN 30% 30% 30% 31% 16% 45% DA E DA E WINTMN MINTMN 42% 42% 42% 41% 41% 42% DA SD DA SD WINTHN MINTHN 35% 22% 48% 44% 33% 55% DA M DA M WINTHN MINTHN 30% 9% 51% 40% 23% 57% DA E DA E WINTHN MINTHN 42% 41% 42% 52% 51% 52% DA SD DA SD % % % % % % reduct reduction reduction reduct reduction reduction ion if if ion if if using increased decrease using increased decreased spatial by 1 d by 1 spatial by 1 by 1 CornSoy mean StDev StDev CornSoy mean StDev StDev WICTMN MICTMN 45% 38% 52% 45% 44% 45% AFEX M AFEX M WICTMN MICTMN 43% 31% 55% 43% 42% 44% AFEX E AFEX E WICTMN MICTMN 47% 47% 48% 47% 47% 47% AFEX SD AFEX SD WICTHN MICTHN 45% 38% 52% 44% 41% 47% AFEX M AFEX M WICTHN MICTHN 43% 31% 55% 42% 39% 45% AFEX E AFEX E WICTHN MICTHN 47% 47% 48% 47% 47% 47% AFEX SD AFEX SD WINTMN MINTMN 47% 44% 51% 45% 43% 47% AFEX M AFEX M WINTMN MINTMN 47% 41% 53% 44% 40% 47% AFEX E AFEX E WINTMN MINTMN 47% 47% 48% 47% 47% 47% AFEX SD AFEX SD WINTHN MINTHN 47% 45% 49% 44% 42% 47% AFEX M AFEX M WINTHN MINTHN 47% 42% 51% 43% 39% 47% AFEX E AFEX E

209

WINTHN MINTHN 47% 47% 48% 47% 47% 47% AFEX SD AFEX SD WICTMN MICTMN 39% 31% 48% 39% 38% 40% DA M DA M WICTMN MICTMN 37% 23% 51% 37% 37% 38% DA E DA E WICTMN MICTMN 42% 42% 43% 42% 42% 42% DA SD DA SD WICTHN MICTHN 39% 31% 47% 38% 35% 41% DA M DA M WICTHN MICTHN 37% 30% 43% 36% 32% 41% DA E DA E WICTHN MICTHN 42% 41% 43% 42% 42% 42% DA SD DA SD WINTMN MINTMN 42% 40% 44% 40% 37% 42% DA M DA M WINTMN MINTMN 42% 35% 49% 38% 34% 42% DA E DA E WINTMN MINTMN 42% 42% 43% 42% 42% 42% DA SD DA SD WINTHN MINTHN 42% 39% 45% 39% 36% 42% DA M DA M WINTHN MINTHN 41% 36% 47% 37% 32% 41% DA E DA E WINTHN MINTHN 42% 42% 42% 42% 42% 42% DA SD DA SD

4.10.3.1. GWP

The greatest variation in environmental impact results due to spatial variation is observed in the

GWP Intensity (Figure 4.25.). This is due in part to the fact that many crop production factors that vary in space contribute to GHG emissions. For example, the emissions of CO2 and N2O from soils are large drivers of the overall GWP Intensity and they exhibited large variations over space (as noted in the crop production LCA discussion). In addition to this, yield contributes to the GWP Intensity in two ways. First, yield plays a role in the denominator by determining the efficiency of the production of the feedstock (i.e. how much energy is captured from the sun for each unit of energy put in to the agricultural system). Second, the yield plays a role in the amount

210 of carbon captured by the plant and stored in both above and below ground tissues and this is a driver of the potential for carbon absorption. In addition to this, spatial variations in the application of lime according to the soil requirements contributed to the overall variability in the

GWP intensity metric, but less so than emissions from soils and yield. Two other interesting trends become apparent from examination of Figure 4.25.

First, not only is the GWP intensity smallest in the scenarios with SD allocation, but the effect of the spatial variability is dramatically reduced in these scenarios compared to the M and the E allocation scenarios. Of course, the spatial variability comes from the agricultural production, so it follows that an allocation method that reduces the overall role of the agricultural system in the full LCA will therefore reduce the influence of spatial variability in the final GWP Intensity.

More importantly, this finding has relevant policy implications. If a policy on the assessment of fuels required spatially-explicit treatment of agricultural production, but also used an allocation method that unevenly apportions burdens between corn grain and corn stover, then this policy may cause the analysis to inaccurately represent the true spatial variability. For example, the error bars around the GWP Intensity in scenarios with SD allocation are very small compared to the error bars in scenarios with E or M allocation. This could be misinterpreted to mean that there is very little spatial variability in GHG emissions from agricultural production. The crop production portion of this LCA, however, demonstrated that there is appreciable spatial variability in GHG emissions.

The second noticeable trend in Figure 4.25. is that the error bars around the MI CornSoy scenarios (purple bars) are much smaller across all scenarios (more noticeably with M and E allocation than in SD allocation scenarios). The MI CornSoy scenarios did show more GWP spatial uniformity than the WI CornSoy or MI ContCorn scenarios. This suggests that the corn

211 soy rotation management practice reduces spatial variability in GHG emissions in certain locations (i.e. in MI, but not in WI). A potential driver of this trend could be a difference in how the MI soils respond to the fixation of nitrogen during soybean production as compared to WI soils or the absence of a soybean production. Overall, this analysis for the sensitivity of GWP

Intensity results to spatial variation illustrates that this metric is quite sensitive to spatial variation. Therefore, this factor should not be assumed to be unimportant or inconsequential in future study.

The percent reduction in GWP Intensity from CS ethanol production (compared to an energy equivalent amount of gasoline) across all 96 (Table 4.13.) does not yield a single scenario that meets the 60% reduction target set forth for cellulosic ethanol in EISA (2007) with or without the change in the GWP Intensity due to spatial variability. This CS ethanol would meet the 20% reduction threshold to be defined as a “renewable fuel” and a few scenarios come close to or meet the 50% reduction threshold to be defined as an “advanced biofuel” (EISA, 2007). One scenario that meets this 50% target regardless of spatial variability is the MIContCornNTHN DA

SD scenario. As noted earlier, however, the SD allocation method reduces the overall environmental impacts of agricultural production in the full ethanol production LCA. This scenarios M and E allocation method counterparts, do not meet the 50% cutoff when the spatial mean values are used, but do meet the cutoff when the spatial means are reduced by one StDev.

This example draws attention to an interesting interplay between allocation methodology and spatial variability. Perhaps, the only way to be sure that a fuel production scenario can meet the requirements of a renewable fuel standard is if the range of emissions results (due to allocation choices and spatial variation) falls within an acceptable range of values rather than just meeting one cutoff. Alternatively, maybe the midpoint of the range should have to meet the cutoff rather

212 than just one value. In this example, the emissions reductions range widely from 23% to 57% (a

34 percentage-point difference) and the midpoint between these two values is 39.5%. This set of scenarios would, therefore, not meet a 50% cutoff. This use of this calculation method could reduce the likelihood that agricultural production practices which increase spatial variability in environmental impacts are incentivized inadvertently. As noted above, there appears to be a reduction of spatial variability in the MI CornSoy scenarios. The same production scenario

(MINTHN DA) with a corn soy rotation gives a much smaller range of values across all allocation methods and with the inclusion of spatial variation (only a 10 percentage-point difference), but this scenario does not meet the 50% reduction target. One corn soy rotation scenario that would come close to the 50% reduction target with this midpoint approach is the

WINTMN AFEX which ranges more narrowly from 41% to 53% and has a midpoint of 46.5%.

This scenario was also identified as one of the best possible combinations of environmental outcomes from the ethanol LCA results analysis.

4.10.3.2. NER

The effect of spatial variability on the NER (Figure 4.24) was not as dramatic as it was on the

GWP. The two main sources of spatial variability that influence the NER are the yield and the liming application. The differences in the calculated lime application rate for each specific pixel affected the amount of lime needed to be produced for the locations. Increased lime production requires increased energy inputs to the crop production system. This input was not as large of a contributor to the overall analysis, however, as the yield. As noted earlier, yield is a substantial driver of the NER metric. Once again, the range of potential values due to spatial variability is reduced in the SD allocation scenarios and largest in the E allocation scenarios. Interestingly, the

WI ContCorn scenarios (blue bars) have some of the largest ranges. This presents noteworthy

213 contrast to the limited range of values observed for the MI CornSoy scenarios, discussed in the

GWP section above. This lends additional support to the idea that a more environmentally intensive rotation could increase the variation in environmental outcomes.

4.10.3.3. AP and EP

The range of AP values introduced by spatial variation (Figure 4.26.) appears uniform across the production scenarios, but not the allocation methods. Once again, the SD allocation method reduces the overall influence of spatial variability in the AP results. In the M and E allocation methods, however, the ranges are quite uniform across the production scenarios. Also, the ranges are very small compared to the total AP values, and compared to the variation observed in the

GWP Intensity. Hence, the spatial variability appears to have a minimal effect on this metric.

This is not the case for the EP metric. Many of the scenarios have wide ranges of potential EP results due to spatial variability (Figure 4.27.). This effect is reduced in the scenarios with the SD method (as is the total EP in general). In this metric, larger ranges occur in WI than in MI in both the corn soy and continuous corn rotations. This could be indicative of differences in the soil characteristics of the two regions. The ranges around less environmentally intensive practices

(e.g. no till with medium N application) are smaller than that of more intensive practices (e.g. chisel till with medium or high N application). This pattern, however, does not appear between the corn soy and continuous corn rotations. Still, some of the ranges are large enough to either eradicate the potential EP benefit of a production process (e.g. WICornSoyNTHN AFEX E) or create a negative EP in a scenario whose mean values did not produce one (e.g.

WICornSoyCTHN DA E). This variability stems primarily from the spatial variability in emissions of N and P through leaching and runoff. Therefore, the accounting of these emissions in an ethanol production LCA should be done in a spatially-explicit manner.

214

4.11. CONCLUSIONS

In the crop production portion of the LCA, the effects of differences in soil characteristics

became evident in relevant metrics such as GWP and EP. Since corn production is likely to

continue on prime agricultural lands (such as the northeast corner of the WI RIMA), identifying

management practices that would reduce negative environmental impacts when stover is

removed is key. No till and corn-soybean rotations are practices that markedly reduce GWP and

EP when land is transitioned out of the baseline scenario. While reduced N application

intuitively leads to reduced GWP and EP, the potential drop in CS yield counteracts that effect

on a per MJ of stover output basis. The identification of nitrogen application rates, based on the

location-specific needs of the soil, could optimize CS production and produce the most beneficial

NER, GWP and EP combination that would improve the overall CS ethanol environmental

impact. Additionally, the results of the crop production LCA varied less in space with less intensive agricultural practices (e.g. no till, medium N and corn soy rotation). The effect of spatial variation on the full ethanol production LCA GWP and EP metrics was large in many scenarios and appeared to be tied to agricultural management practices, but not so for the NER and AP metrics. Finally, the statistical analysis of the results showed strong interaction effects between location and agricultural management practices on environmental impacts.

The full ethanol production LCA results revealed interesting effects of allocation on the impacts of agricultural management practices. For example, the use of the SD allocation method set up a counter-productive incentive to boost yields by whatever means necessary (to increase the ethanol production NER), without passing on much of the associated GWP, AP or EP penalty of management practices that boost yields (e.g. increased nitrogen fertilizer application or chisel tillage). In addition to this, the SD allocation method limited the range of potential results due to

215 spatial variability in the sensitivity analysis. This made it appear as though there was not much spatial variation in crop production, even though there clearly was notable variation (especially in the GWP and EP metrics) when the M and E allocation methods were used. This finding has important policy implications if a renewable fuel standard were to mandate the methods used to conduct a cellulosic ethanol LCA. Policy makers must ensure that they do not set up perverse incentives that inadvertently increase negative environmental impacts from crop production. It is important to note that just because the GWP, AP and EP of the corn stover cellulosic ethanol production appear low as a result of the allocation choice, does not mean that impacts to climate change, acidification or eutrophication are kept low or even reduced. The climate changing, acidifying or eutrophying compounds were still released as a result of the agricultural management practices implemented for corn production.

If the goal of cellulosic ethanol production is to reduce any or all of these impacts, then perhaps the burdens must be evenly split between the corn grain and corn stover production so as not to incentivize practices that increase yields and negative environmental impacts. While there are some standards requiring LCA for cellulosic ethanol production in the US, there are no similar standards for commodity crop production. Furthermore, the allocation method chosen affects the influence of spatial variability in the environmental impact results. The SD method dramatically reduces the influence of spatial variability in the full ethanol LCA GWP Intensity by reducing the overall influence of the crop production on the full life cycle. The actual spatial variability from crop production, however, is not reduced; it just appears to have less of an effect on the overall LCA when the SD method is used. It is therefore recommended that an allocation method which evenly apportions burdens between corn grain and corn stover is used (such as energy allocation) in order to prevent the unintended consequence of promoting agricultural practices

216 that worsen environmental impacts from corn production through policies aimed at cellulosic ethanol production.

From the scenario comparison within the energy allocation method, clear environmental impact winners and losers became apparent. Unfortunately, none of the CS ethanol production scenarios met the EISA (2007) 60% reduction threshold for a cellulosic biofuel. There is encouraging evidence, however, that some scenarios could meet the 50% reduction target to be considered an advanced biofuel. The scenarios that combined WI corn soy rotations with no till and either high or medium nitrogen and DA or AFEX pretreatment gave the best combinations of environmental outcomes. Interestingly, the WICornSoyNTMN AFEX scenarios also came close to meeting the

50% threshold for an advanced biofuel using the midpoint method described in the results and discussion section. These scenarios also did not yield wide ranges of values for NER, GWP, AP or EP in the sensitivity analysis. These factors combined provide a strong basis of support for the assertion that this scenario could deliver the desired environmental outcomes. While the MI scenarios were not as environmentally beneficial as the WI scenarios, valuable insights on optimal production variable combinations specific to MI were clear and somewhat similar to that of the WI optimal scenarios. These scenarios combined corn soy rotations with no till, but only medium nitrogen application and a choice between AFEX or dilute acid pretreatment. The MI

CornSoy scenarios showed the least variation in final metrics in the sensitivity analysis which suggests that ethanol production from stover produced with these management practices could deliver the desired environmental outcomes.

From a biorefinery-siting perspective, this information, combined with economic assessments of potential business taxes and incentives, would be useful in choosing a location between, for example, Wisconsin and Michigan. From a technology selection standpoint, this information,

217

combined with economic estimates relative cost differences between AFEX and DA

pretreatment or social science surveys on farmers’ willingness-to-produce feedstock, would be useful in determining location and technology selections. Finally, from a policy standpoint, this information could be useful in crafting policies that incentivize positive environmental outcomes in a way that is state- or region-specific since one set of practices did not produce the best outcomes in both RIMAs.

This analysis was unique in that it quantified the environmental impacts of cellulosic ethanol production from corn stover in a way that included spatial variability, local environmental uniqueness and the effects of different agricultural management practices and pretreatment technologies on the overall results. This research also demonstrated three important and previously unexplored aspects of the LCA of biofuels: 1) it is possible to include spatial variability and spatially-explicit data in the LCA of a production system that includes agriculture,

2) the inclusion of this variability has substantial impacts on the LCA results, and 3) there allocation method implemented not only affects the final results, but it also limits the influence of spatial variability on the results in the sensitivity analysis. Perhaps the most significant finding is that there is a strong interplay between allocation and spatial variability when the allocation

point is between two agricultural products produced over large areas in different agricultural

contexts.

As noted in the conclusions of part one of this chapter, there are many strengths and weaknesses

associated with using the EPIC modeled crop production data for this analysis. Unlike SG,

however, the EPIC model has been validated with corn production field trials conducted at a

field station in the WI RIMA. An analysis published by Wang et al. (2012) showed that observed

yields were within the 5% and 95% confidence limits of EPIC modeled yields and the correlation

218 between the EPIC predicted and measured soil organic carbon values had an r2 of 0.89. We acknowledge that modeled crop production data has inherent sources of variability and uncertainty (e.g. the accuracy and precision of soil and weather measured data), but that, especially in the case of corn, the strengths outweigh the weaknesses. Crop production models, in general, also allow scientists to test the effects of agricultural management practices that could have devastating effects on large areas without actually altering the landscape. For example, a twelve-year large-scale field trial of continuous corn production with chisel tillage, high nitrogen application and corn stover removal could result in significant erosion, soil degradation and

GHG emissions. It is important to test the effects of different management practices on different landscapes before implementing or incentivizing any set of practices.

We do not currently know where corn stover will be harvested for ethanol; therefore, it is important to evaluate the potential impacts on the landscape from its production and removal under a variety of management strategies and across a number of locations. Additionally, the use of the standard deviations from the watershed mean values for yield and emissions allowed us to quantify some of the differences in environmental impacts that could result from producing the stover on more or less productive land within the watershed. While USDA NASS data are collected for crops like corn, these data represent a spatial average over an entire county and provide no indicator of deviation from this average within a county. Therefore, LCAs on crop or biofuel production that use USDA NASS data do not have a realistic basis on which to test for the effects of spatial variation within a county. These problems of homogeneous assumptions and the lack of attention to spatial variability and local environmental uniqueness have been identified as grand challenges and unresolved problems in the LCA of biofuels field (Reap et al.,

2008 and McKone et al., 2011). This analysis attempted to address these problems with some of

219

the most comprehensive and spatially-explicit data currently available. Our results indicate that

the combined effects of where and how the CS is grown can play a large role in the overall

environmental impacts from CS and cellulosic ethanol production.

This research also illustrated that the results of an ethanol production LCA are not single values.

They are in fact ranges of potential values. Yet, the renewable fuel standard set forth in EISA

(2007) bases its reduction targets on single values and many studies report only single values

(with no range). Ranges could be produced by calculating results with several allocation

methods, by including spatial variability and in a number of other ways. If an analysis is

conducted in such a way as to reduce the influence of crop production (with an SD allocation

method between corn grain and corn stover) or ignore spatial variability, then the result could

show that the fuel production meets the standard even if the range of potential outcomes (or the

midpoint between those outcomes) does not meet the standard. It is therefore imperative that

future LCAs of the production of a good that involves agriculture take into account these issues

to ensure that the environmental goals and standards set forth by society are met.

4.12. REFERENCES

These are references cited only in this part two of Chapter 4.

Brady, N. C. and Weil, R. R. 2004. Elements of the Nature and Properties of Soils, Second Edition. Prentice Hall: Upper Saddle River, NJ.

Domalski, E. S., Jobe, T. L., Milne Jr., T. A. 1986. Thermodynamic data for biomass conversion and waste incineration. Golden, Colorado: Solar Energy Research Institute.

Supplemental Table 4.2. Additional LCI data inputs and sources for the CS ethanol LCA. Agricultural Data Source LCI Input Phosphorus US LCI, 2008

220

Fertilizer Pesticide: EcoInvent, glyphosate 2007 Maize Seed EcoInvent, 2007 No Till PE, 2006 planter Corn Stover MI: Zhang, et Yield al. 2010 WI: Kang and Post, 2012 Corn Grain MI: Zhang, et Yield al. 2010 WI: Kang and Post, 2012

221

CHAPTER 5

Conclusions and Reflections

5.1. Conclusions and Reflections

Life Cycle Assessment (LCA) is like a jigsaw puzzle with hundreds of pieces and no picture on

the box. You have to imagine the whole picture yourself, examine the pieces you have and seek

the ones you lack. You might find that there are pieces missing or extra pieces that have nothing

to do with the picture you are trying to put together. Sometimes you have to create missing

pieces or painstakingly describe the pieces you need for someone else to make for you. Most

challenging of all is the mental task of keeping the big picture in focus in your mind while you

simultaneously drill into the detailed contours of every puzzle piece.

To learn how the puzzle pieces would fit together, I had to read everything I could about LCA and biofuels, and conduct a case study on the most crucial pieces. It was only after exploring the known world of biofuel LCA that I could begin creating my own novel LCA picture. To start forming this big picture, the outline and skeleton of the LCA picture must be built first. I

accomplished this in GaBi – a program I had never even heard of at the onset of this journey.

Then the blank spaces in the skeleton had to be filled in with data. The first source I turned to for this information was the literature. Soon it became obvious that there were pieces I could not find in the literature or create myself. That is when a long conversation with scientists from many disciplines began. Data from the EPIC model filled fundamental gaps in for crop production and emissions from agricultural systems. With no operational cellulosic ethanol plants to observe, I turned to fellow scientists to help identify the near-term methods for cellulosic ethanol production and specifications on how a cellulosic ethanol refinery would 222 operate. Power production for the specific study areas filled in the last open spaces and, finally, the full picture was out of my head and in the world for everyone to see.

Along the way, I wrestled with pieces that just did not seem to fit in their specified place. In some cases, their boundaries matched the blank space in the puzzle, but the picture was all wrong. I had to go back to the scientists who gave me the piece and ask why the picture looked so strange. In one case, it was because the piece had not been made correctly and had to be completely re-done. As years went by, new and improved pieces became available. I had to go back and take out old pieces based on older ethanol production and replace them with the new pieces that reflected new technologies and expanded understanding in the field.

During this time, the fundamental understanding of LCA itself changed around me. I needed to study how LCA methodological choices would affect the overall analysis and delve into some of the grand challenges identified in LCA. I took stock of the body of literature in ethanol production, LCA, biogeochemical modeling, biomass production, the impact of human activities on the global environment and many other related areas. I read several methodological exploration papers in preparation for this research, but two of these papers stood out among the rest. The first was a paper entitled “Grand Challenges for Life-Cycle Assessment of Biofuels” by

McKone et al. (2011). This paper identified seven major problems facing the LCA of biofuels, but the most interesting to me were: understanding feedstock options and land use, predicting biofuel production technologies and practices and incorporating spatial components into life cycle inventories and assessments. The second paper was actually a two part series on LCA problems entitled “A survey of unresolved problems in life cycle assessment” by Reap et al.

(2009a and 2009b). Part one dealt with goal and scope and inventory analysis LCA phases, while the part two addressed the impact assessment and interpretation phases of LCA. The list of 223

unresolved problems in LCA noted in this study is very long and, since this paper was not

focused on biofuels LCA, not all of these issues pertain to biofuels. The problem identified in

this paper that most piqued my interest was the homogeneous treatment of the landscape in

LCAs which involved the production of resources from ecological systems (such as biomass

production). The spatial variability and local environmental uniqueness issues identified in both

articles were something I was well aware of, but had yet to see a cellulosic ethanol LCA attempt

to deal with. I set forth with the rest of my research with these problems as a target on which to focus.

The first body chapter of this dissertation was a case study on the effects of allocation choices in a corn grain and stover ethanol production LCA. Allocation was identified in both Reap et al.

(2009a and 2009b) and McKone et al. (2011) as an LCA grand challenge and a topic great

controversy and debate. This study compared several allocation methods and ways of

approaching the LCA of a multifunctional system. Not surprisingly, this research showed that the

results of such an LCA can change dramatically based on the allocation methods applied. More

interestingly, the allocation choice could determine whether or not corn stover ethanol met or

missed the 60% emissions reduction compared to gasoline emissions target put forth by the

Energy Independence and Security Act (EISA) of 2007 Renewable Fuel Standard (RFS).

Furthermore, this study compared attributional and consequential approaches to ethanol LCA

and concluded that the attributional approach is more appropriate for studies based on historical

data such as GHG emissions data. This case study set the stage for the larger and more complex

ethanol LCA I would undertake for the rest of the dissertation because it made me aware of the

many LCA analysis choices that could affect the analysis and cause a fuel to pass or fail a

standard. 224

The number of conclusions about the Global Warming Potential (GWP) and net energy metrics

of ethanol are as numerous as the number of studies on the topic. Countless studies attacked this

question from various angles such as different ethanol production methods and feedstocks, data

sources and LCA methods. As I read more and more of these studies, one major theme emerged

– none of these studies were comparable! Furthermore, just trying to discern the basis of analysis

for many of the studies was a study in and of itself. For example, a switchgrass cellulosic ethanol

LCA by Bai et al. (2010) gave values for Acidification Potential (AP) and Eutrophication

Potential (EP) which I wanted to compare to the AP and EP values for my final body chapter.

The functional unit in Bai et al. (2010) by which all results were normalized was defined as

“power to wheels for 1-km driving of a midsize car”. The trouble with this was that the assumed km per liter of fuel was never stated in the paper. Instead, the energy requirements for driving were cited as “similar to the assumptions of Luo et al. (2009b)”. Of course, Luo et al. (2009b) also did not state this conversion factor and, instead, cited Fu et al. (2003) who, in turn, cited

Wang et al. (1997). Finally, Wang et al. (1997) cited the United States Energy Information

Administration for the fuel efficiency of a standard new (circa 1990s) midsized vehicle burning gasoline (not ethanol or a gasoline-ethanol blend) as 28mpg. It would have been much simpler for anyone to compare their results to the results of Bai et al. (2010) if this conversion factor was clearly stated in the text or a supplemental table, or if the results were presented per MJ in addition to per km.

This convoluted path seemed to fly in the face of the fundamental tenant of science that methods should be transparent and repeatable. This was the impetus for the second body chapter of this dissertation which set forth to illustrate this problem with seven representative studies on ethanol production. It compared and evaluated the functional units, system boundaries, allocation 225 methods and impact category metrics of all the studies. This study helped me to identify the key

LCA components I should use to facilitate comparison (e.g. the MJ functional unit and additional impact category metrics beyond the conventional GWP Intensity and NER) and inspired me to come up with a sort of LCA meta-data table to make the presentation of this information more transparent. I know that the inclusion of such a table will not remedy all of the repeatability and transparency problems encountered in LCA papers, but it could at least help a reader determine if the study they are reading is potentially comparable to another study they are reading or conducting. Plus, the use of such a table could prevent misinterpretation and inappropriate comparisons not only by scientists, but also by policymakers and the general public.

With a case study and methodological review under my belt, I dove head first into the final body chapter of this dissertation on the LCA of cellulosic ethanol production. It was in this research I aimed to address issues of spatial variability and local environmental uniqueness in agricultural and industrial production, as well as understand feedstock production choices, land use and ethanol production technology. This study showed that spatial variability could indeed change the final results of an ethanol production LCA in substantial ways. The ability of a fuel to meet the EISA 2007 target for consideration as a cellulosic biofuel hinged on the spatial variability in the case of switchgrass. Unlike the results of other studies, corn stover ethanol could not meet this benchmark. Additionally, an interrelation between spatial variability and allocation choices emerged. While each of these issues was addressed separately in the Reap et al. (2009a and

2009b) and McKone et al. (2010) papers, their interrelation was not noted. This research addressed the calls for LCAs to be spatially-explicit by showing that it is possible to do and that the results are impacted by spatial variability. It also employed the lessons learned in the previous chapters to come up with a more comprehensive analysis. Future research in the area of 226 spatially-explicit LCA could use this work as a baseline upon which to build further complexity and improve the overall LCA puzzle picture.

All throughout this process, I have been asked to single out winners and losers for cellulosic ethanol production. Should the feedstock be switchgrass because it is a perennial? Should we just use corn stover because we grow so much corn already? Which pretreatment method yields the best NER or GWP Intensity? I have always felt that it is not who wins or loses, but how you conduct the analysis. If I was forced to pick a winner, I think that my analysis shows that switchgrass as a feedstock has the best potential to meet the EISA (2007) GHG emissions reduction target to be considered a cellulosic biofuel. Furthermore, agricultural production of both feedstocks under many different combinations of management practices yielded better environmental outcomes in Wisconsin than in Michigan (and sensitivity analysis showed that this finding was not influenced by sports rivalries or the fact that badgers are cool). As noted in part one of chapter four, the Wisconsin switchgrass with medium nitrogen and AFEX pretreatment scenario had the most optimal combination of NER, GWP, AP and EP, so if I had to ‘pick a winner’, that would be it. If I had to name a ‘loser’, then, with all due respect to the

Spartans, the Michigan corn stover with continuous corn rotation, chisel till, high nitrogen and dilute acid pretreatment had the worst combination of environmental impacts.

Grand challenges in the LCA of biofuel production still exist. One such challenge that even this work could not tackle is the free availability of well-documented data. Much of the data used in my studies was sourced from GaBi 4.4. These data are well-documented and designed to have consistent system boundaries and analysis methods so that linking up many simple unit processes into one complex model is as seamless as possible. Nevertheless, only a limited number of people have access to these data and gaining access can be prohibitively expensive for many 227

interested parties. Furthermore, since the data are the proprietary property of PE International

(the makers of GaBi), they do not want these raw data published. I openly admit that this limits

the transparency and repeatability of my research in a way with which I am not pleased.

Therefore, I recommend the development of a similarly consistent and well-documented LCI database that could be freely available. I know that some work has been done in this field by

OpenLCA, but the data available on this website (http://www.openlca.org/index.html) are limited. If the results of the numerous published LCAs could be added to this database with the sort of, meta-data LCA table I suggest in chapter three, this could greatly improve the data availability, and therefore, transparency of future LCAs.

Finally, I would like to note that the learning, reading and research required to complete this dissertation left me with many useful rules that I will take with me and follow in future research endeavors. I believe that it is only by following these guidelines in research on the environmental impacts of the production of goods and services that we can hope to ensure the long-term sustainability of this one Earth we all share.

1. Know your system.

2. Know your data and data sources.

3. If you must allocate, allocate fairly and always remember why you are allocating.

4. Be consistent with data and allocation.

5. Clearly state all assumptions and the reasons for the assumptions.

6. Question everything.*

*This is a good rule for life too. 228

Works Cited

Aden, A., Ruth, M., Ibsen, K., Jechura, J., Neeves, K., Sheehan, J., and Wallace, B. 2002. Lignocellulosic biomass to ethanol process design and economics utilizing co-current dilute acid prehydrolysis and enzymatic hydrolysis for corn stover (NREL/TP-510- 32438); National Renewable Energy Laboratory: Golden CO, 2002. Available at: http://www.nrel.gov/docs/fy02osti/32438.pdf. Aravindhakshan, S. C., Epplin, F. M. and Taliaferro, C. M. 2011. Switchgrass, Bermudagrass, Flaccidgrass, and Lovegrass biomass yield response to nitrogen for single and double harvest. Biomass and Bioenergy. 35: 308-319. Argonne National Laboratory (ANL). 2010. Greenhouse gases, Regulated Emissions, and Energy use in Transportation Model, GREET Version 1.8C. Accessed March 2010. Bals, B. Personal Communication. 26 January 2011. Michigan State University. Baumann, H. and Tillman, A. 2004. The Hitch Hiker’s Guide to LCA: An orientation in life cycle assessment methodology and application. Lund, Sweden. Studentlitteratur AB. Beer, T. and Grant, T. 2007. Life-cycle analysis of emissions from fuel ethanol and blends in Australian heavy and light vehicles. Journal of Cleaner Production. 15: 833-837. Bjorklund, A. 2002. Survey of approaches to improve reliability in LCA. International Journal of Life Cycle Assessment. 7: 64-72. Brady, N. C. and Weil, R.R. 2004. Elements of the Nature and Properties of Soils, Second Edition. Prentice Hall: Upper Saddle River, NJ. Bright, R. M. and Strømman A. H. 2009. Life cycle assessment of second generation bioethanols produced from Scandinavian boreal forest resources: a regional analysis for Middle Norway. Journal of Industrial Ecology. 13(4): 514-531. Brown, R. C. 2003. Biorenewable Resources: Engineering New Products from Agriculture. Ames, IA: Blackwell Publishing Professional. California Environmental Protection Agency Air Resources Board (CARB). 2009. Proposed regulation to implement the low carbon fuel standard: vol. 1; CaARB: Sacramento, CA. 2009. http://www.arb.ca.gov/regact/2009/lcfs09/lcfsisor1.pdf. Accessed: Nov 2009. Curran, M. A., Mann, M. and Norris, G. 2005. The international workshop on electricity data for life cycle inventories. Journal of Cleaner Production 13(8): 853-862. 229

Curran, M. A. 2007. Studying the Effect on System Preference by Varying Coproduct Allocation in Creating Life-Cycle Inventory. Environmental Science and Technology. 41(20): 7145- 7151. Dale, B. E., Bals, B. D., Kim, S. and Eranki, P. 2010. Biofuels Done Right: Land efficient animal feeds enable large environmental and energy benefits. Environmental Science and Technology. 44(22): 8385-8389. Dale, B. E. and Kim, S. 2011. Response to comments by O’Hare et al., on the paper indirect land use change for biofuels: Testing predictions and improving analytical methodologies. Biomass and Bioenergy. 35: 4492-4493. Data Transmission Network. 2009. Ethanol Center; DTN: Omaha NE. http://www.dtnethanolcenter.com/. Accessed: Feb 2009. Domalski, E. S., Jobe T. L. and Milne Jr., T. A. 1986. Thermodynamic data for biomass conversion and waste incineration. Golden, Colorado: Solar Energy Research Institute. Ecoinvent Centre. 2007. Ecoinvent Eata. v2.0. Ecoinvent Reports No.1-25. Swiss Centre for Life Cycle Inventories. Dübendorf. Economic Research Service (ERS). 2008. Farm Business and Household Survey Data. http://www.ers.usda.gov/data/arms/app/Crop.aspx. Accessed: Nov 2008. Egbendewe-Mondzozo, A., Swinton, S. M., Izaurralde, R. C., Manowitz, D. H. and Zhang, X. 2011. Biomass supply from alternative cellulosic crops and crop residues: A spatially explicit bioeconomic modeling approach. Biomass and Bioenergy. 35: 4636-4647. Ekval, T. and Tillman, A. 1997. Open-Loop Recycling: Criteria for Allocation Procedures. International Journal of Life Cycle Assessment. 2: 155-162. Ekvall, T. and Andrae, A. 2006. Attribuitional and consequential environmental assessment of shift to lead-free solders. International Journal of Life Cycle Assessment. 11: 344-353. Ekvall, T. and Finnveden, G. 2001. Allocation in ISO 14041—a critical review. Journal of Cleaner Production. 9: 197-208. Elmore, R. 1982. How a corn plant develops (SR0048). Iowa State Extension publication: Ames, IA. Energy Independence and Security Act (EISA) of 2007. Pub. L. no. 110-140. 110th Congress. 1st session. 2007. http://www.gpo.gov/fdsys/pkg/BILLS-110hr6enr/pdf/BILLS- 110hr6enr.pdf. Accessed: November 2011. 230

Eranki. P, L. and Dale, B.E. 2011. Comparative life cycle assessment of centralized and distributed biomass processing systems combined with mixed feedstock landscapes. Global Change Biology: Bioenergy. 3(6): 1757-1707 Farrell, A.E., Plevin, R.J., Turner, B.T., Jones, A.D., O'Hare, M. and Kammen, D.M. 2006. Ethanol Can Contribute to Energy and Environmental Goals. Science. 311(5760): 506- 508. Fu, G. Z., Chan, A.W. and Minns, D.E. 2003. Life Cycle Assessment of Bio-ethanol Derived from Cellulose. International Journal of Life Cycle Assessment. 8(3): 137-141. Gawel, E. and Ludwig, G. 2011. The iLUC dilemma: How to deal with indirect land use changes when governing energy crops? Land Use Policy. doi: 10.1016/j.landusepol.2011.03.003. Grace, P.R., Robertson, G.P., Millar, N., Colunga-Farcia, M., Basso, B., Gage, S.H., and Hoben, J. 2011. The contribution of maize cropping in the Midwest USA to global warming: A regional estimate. Agricultural Systems. 104: 292-296. Grassini, P. and Cassman, K.G. 2011. High-yield maize with large net energy yield and small global warming intensity. Proceedings of the National Academy of Sciences Early Edition. doi: 10.1073/pnas.1201296109. Guinée, J. B. 2002. Handbook on Life Cycle Assessment. Dordrecht, The Netherlands: Kluwer Academic Publishers. Guinée J., Heijungs R. and van der Voet, E. 2009. A greenhouse gas indicator for bioenergy: some theoretical issues with practical implications. The International Journal of Life Cycle Assessment. 14: 328-339. Haberl, H., Sprinz, D. Bonazountas, M., Cocco, P., Desaubies, Y., Henze, M. Hertel, O., Johnson, R.K., Kastrup, U., Laconte, P., Lange, E., Novak, P., Paavola, J., Reenbergy, A., van den Hove, S., Vermeire, T., Wadhams, P. and Searchinger, T. 2012. Correcting a fundamental error in greenhouse gas accounting related to bioenergy. Energy Policy. 45: 18-23. Haque, M., Epplin, F. M. and Taliaferro, C.M. 2009. Nitrogen and harvest frequency effect on yield and cost for four perennial grasses. Agronomy Journal. 101: 1463-1469. He, X., Izzaurralde, R. C., Vanotti, M. B., Williams, J. R., and Thomson, A. M. 2006. Simulating long-term and residual effects of nitrogen fertilization on corn yields, soil carbon 231

sequestration, and soil nitrogen dynamics. Journal of Environmental Quality. 35: 1608- 1619. Hill, J., Polasky, S., Nelson, E., Tilman. D, Huo, H., Ludwig, L., Neumann, J., Zheng, H. and Bonta, D. 2009. Climate change and health costs of air emissions from biofuels and gasoline. Proceedings of the National Academy of Sciences. 106(6): 2077-2082. Hill, J., Nelson, E., Tilman, D., Polasky, S. and Tiffany, D. 2006. Environmental, economic, and energetic costs and benefits of biodiesel and ethanol biofuels. Proceedings of the National Academy of Sciences. 103: 11206-11210. Hsu, D. D., Inman, D., Heath, G. A., Wolfrum, E. J., Mann, M. K. and Aden, A. 2010. Life Cycle Environmental Impacts of Selected U.S. Ethanol Production and Use Pathways in 2022. Environmental Science and Technology. 44: 5289-5297. Humbird, D., Davis, R., Tao, L., Kinchin, C., Hsu, D., Aden, A., Schoen, P., Lukas, J., Olthof, B., Worley, M., Sexton, D. and Dudgeon, D. 2011. Process Design and Economics for Biochemical Conversion of Lignocellulosic Biomass to Ethanol: Dilute-Acid Pretreatment and Enzymatic Hydrolysis of Corn Stover. National Renewable Energy Laboratory. Technical Report: NREL/TP-5100-47764. Golden, CO. Intergovernmental Panel on Climate Change (IPCC). 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge. International Organization for Standardization. (ISO). 2006a. ISO 14040: Environmental management – Life cycle assessment – Principles and framework. Geneva, Switzerland. International Organization for Standardization (ISO). 2006b. ISO 14044: Environmental Management – Life cycle assessment, Life cycle impact assessment. Geneva, Switzerland. Izaurralde, R. C., Williams, J. R., McGill, W. B., Rosenberg, N. J., and Quiroga Jakas, M. C. 2006. Simulating soil C dynamics with EPIC: Model description and testing against long- term data. Ecological Modeling. 192: 362-384. Izaurralde, R. C., Williams, J. R., Post, W. M., Thomson, A. M, McGill, W. B., Owens, L. B and Lal., R. 2007. Long-term modeling of soil C erosion and sequestration at the small watershed scale. Climatic Change. 80(1-2): 73-90. 232

Izaurralde, R. C., W. B. McGill, and Williams, J. R. 2012. Development and application of the EPIC model for carbon cycle, greenhouse-gas mitigation, and biofuel studies. In Franzluebbers, A., Follett, R. and Liebig, M. (eds.). Managing Agricultural Greenhouse Gases: Coordinated agricultural research through GRACEnet to address our changing climate. Elsevier (in press). James, L., Swinton, S. and Thelen, K. 2010. Profitability Analysis of Cellulosic Energy Crops Compared with Corn. Agronomy Journal. 102: 675–687. Johnston, M., Foley, J. A., Holloway, T., Kucharik, C. and Monfreda, C. 2009. Resetting global expectations from agricultural biofuels. Environmental Resource Letters. doi: 10.1088/1748-9326/4/014004. Kaiser, R. 2005. Utilizing the Growing Local Supply of Distillers Grains. http://www.uwex.edu/ces/dairynutrition/pubs.cfm. Accessed: Feb 2008. Kang, S. and Post, W. M. 2012. Personal Communication. Unpublished Data. Oakridge National Laboratory. Oakridge, TN. Kaufman, A. S., Meier, P. J., Sinistore, J. C., and Reinemann, D. J. 2010. Applying life-cycle assessment to low carbon fuel standards – How allocation choices influence carbon intensity for renewable transportation fuels. Energy Policy. 38: 5229-5241. Kazi, F. K., Fortman, J. A., Anex, R. P., Hsu, D. D., Aden, A., Dutta, A. and Kothandaraman, G. 2010. Techno-economic comparison of process technologies for biochemical ethanol production from corn stover. Fuel. 89: S20-S28. Kim, S. and Dale, B. E. 2011. Indirect land use change for biofuels: testing predictions and improving analytical methodologies. Biomass and Bioenergy. 35(7): 3235-3240. Kim, S., Dale, B. E. and Jenkins, R. 2009. Life cycle assessment of corn grain and corn stover in the United States. International Journal of Life Cycle Assessment. 14: 160-174. Kim, S. and Dale, B. E. 2002. Allocation procedure in ethanol production system from corn grain. International Journal of Life Cycle Assessment. 7: 237-243. Kline, K. L., Oladosu, G. A., Dale, V. H., and McBride, A. C. 2011. Scientific analysis is essential to assess biofuel policy effects: In response to the paper by Kim and Dale on “Indirect land-use change for biofuels: Testing predictions and improving analytical methodologies.” Biomass and Bioenergy. 35: 4488-4491. 233

Kodera, K. 2007. Analysis of allocation methods of bioethanol LCA. MSc Thesis. CML, Leiden University, Netherlands.

Kraatz, S. Sinistore, J. C. and Reinemann, D. J. 2011. Methods for Solving the Multifunctionality Problem in Life Cycle Assessment of Ethanol Production from Corn Grain. Proceedings of the American Society of Agricultural and Biological Engineers Annual International Meeting. Louisville, KY, August 7-10. 1110828.

Lange, M. 2011. The GHG balance of biofuels taking into account land use change. Energy Policy. 39: 2373-2385. Laser, M., Haimin, J., Jayawardhana, K. and Lynd, L. R. 2009. Coproduction of ethanol and power from switchgrass. Biofuels, Bioproducts and Biorefining. 3: 195-218. Liska, A. J., Yang, H. S., Bremer, V. R., Klopfenstein, T. J., Walters, D. T., Erickson, G. E. and Cassman, K. G. 2009. Improvements in Life Cycle Energy Efficiency and Greenhouse Gas Emissions of Corn-Ethanol. Journal of Industrial Ecology. 13(1): 58-74.

Liska, A. J., Yang, H. S., Bremer, V. R., Walters, D. T., Erickson, G. E., Klopfenstein, T., Kenney, D., Tracy, P., Koelsch, R. and Cassman, K. G. 2007. BESS: Biofuel Energy Systems Simulator; Life-Cycle Energy and Emissions Analysis Model for Corn-Ethanol Biofuel (Version 2008.3.0 for non-commercial use). University of Nebraska-Lincoln. Luo, L., van der Voet, E., Huppes, G. and Udo de Haes, H. A. 2009. Allocation issues in LCA methodology: a case study of corn stover-based fuel ethanol. International Journal of Life Cycle Assessment. 14: 529-539. MacLean, H. L., and Spatari, S. 2009. The contribution of enzymes and process chemicals to the life cycle of ethanol. Environmental Resource Letters. 4: 14001-14010. Malca, J. and Freire, F. 2006. Renewability and life-cycle energy efficiency of bioethanol and bio-ethyl tertiary butyl ether (bioETBE): Assessing the implications of allocation. Energy. 31: 3362-3380. Matthews, J. A. and Tan, H. 2009. Biofuels and indirect land use change effects: the debate continues. Biofuels, Bioproducts and Biorefining. doi: 10. 1002/bbb.147. 234

McKone, T. E., Nazaroff, W. W., Berck, P., Auffhammer, M., Lipman, T., Torn, M.S., Masanet, E., Lobscheid, A., Santero, N., Mishra, U., Barrett, A., Bomberg, M., Fingerman, K., Scown, C., Strogen, B. and Horvath, A. 2011. Grand Challenges for Life-Cycle Assessment of Biofuels. Environmental Science and Technology. 45: 1751-1756. Meier Engineering Research (MER), LLC. 2011. MyPower ™ electricity-sector model. Meier, P., Erickson, J. and Mapp, J. 2005. Measuring Emission Benefits with Integrated Resource Models - Why Simple Estimates of Emission Reductions May Be Wrong. Proceedings of the International Energy Program Evaluation Conference (IEPEC). 659- 670. Meier, P. 2002. Life-Cycle Assessment of Electricity Generation Systems and Applications for Climate Change Policy Analysis. Ph.D. Dissertation. The University of Wisconsin – Madison, Madison, WI. pp. 59 -66. Mendoza, A., van Ruijven, T., Vad, K. and Wardenaar, T. 2008. The Allocation Problem in Bio- Electricity Chains. MSc. Industrial Ecology. Leiden, The Netherlands. National Renewable Energy Laboratory (NREL). 2008. U.S. Life-Cycle Inventory (LCI) Database. http://www.nrel.gov/lci/database/default.asp. Accessed: January 2010. Natural Resources Conservation Service (NRCS), Soil Survey Staff, United States Department of Agriculture. Soil Survey Geographic (SSURGO) Database for Wisconsin and Michigan. Available online at http://soildatamart.nrcs.usda.gov. Accessed: 2008.

O’Hare, M., Delucchi, M., Edwards, R., Fritsche, U., Gibbs, H., Hertel, T., Hill, J., Kammen, D., Laborde, D., Marelli, L., Mulligan, D., Plevin., R., and Tyner, W. 2011. Comment on “Indirect land use change for biofuels: Testing predictions and improving analytical methodologies” by Kim and Dale: statistical reliability and the definition of the indirect land use change (iLUC) issue. Biomass and Bioenergy. 35: 4485-4487. PE International. 2006. GaBi 4: The Software System for Life Cycle Engineering. PE-Europe GmbH and IKP University of Stuttgart. 235

Penn State University Cooperative Extension. 2009. Coping with high energy prices heat energy content of shelled corn. State College, PA. http://energy.cas.psu.edu/energycontent.html. Accessed: 26 March 2011. Perlack, R., Wright, L., Turhollow, A. Graham, R., Stokes, B., and Erbach, D. 2005. Biomass as a feedstock for a bioenergy and bioproducts industry: The technical feasibility of a billion-ton annual supply. (Tech. Rep ORNL/TM - 2006/66). Oak Ridge, TN: Oak Ridge National Laboratory. Pimentel, D. 2003. Ethanol Fuels: Energy Balance, Economics, and Environmental Impacts are Negative. Natural Resources Research. 12(2): 127-134.

Plevin, R. J. 2009. Modeling Corn Ethanol and Climate: A Critical Comparison of the BESS and GREET Models. Journal of Industrial Ecology. 13(4): 495-507.

Pordesimo, L., Edens, W. and Sokhansanj, S. 2004. Distribution of aboveground biomass in corn stover. Biomass and Bioenergy. 26: 337-343. Potting, J. and Hauschild, M. Z. 2006. Spatial differentiation in life cycle impact assessment: a decade of method development to increase the environmental realism of LCIA. International Journal of Life Cycle Assessment. 11: 11-13. Reap, J., Roman, F., Duncan, S. and Bras, B. 2008a. A survey of unresolved problems in life cycle assessment; Part 1: goal and scope and inventory analysis. International Journal of Life Cycle Assessment. 13: 290-300. Reap, J., Roman, F., Duncan, S. and Bras, B. 2008b. A survey of unresolved problems in life cycle assessment; Part 2: impact assessment and interpretation. International Journal of Life Cycle Assessment. 13: 374-388. Roberts, S. 2011. “U.N. Says 7 Billion Now Share the World”. The New York Times. Published Online: 31 Oct 2011. http://www.nytimes.com/2011/11/01/world/united-nations-reports- 7-billion-humans-but-others-dont-count-on-it.html. Accessed: 13 Dec 2011. Robertson, G., Paul, E. and Harwood, R. 2000. Greenhouse Gases in Intensive Agriculture: Contributions of Individual Gases to the Radiative Forcing of the Atmosphere. Science. 289(5486): 1922-1925. 236

Schmer, M. R., Vogel, K. P., Mitchell, R. B. and Perrin, R. K. 2007. Net energy of cellulosic ethanol from switchgrass. Proceedings of the National Academy of Sciences. 105(2): 464- 469. Searchinger, T., Heimlich, R., Houghton, R. A., Dong, F., Elobeid, A., Fabiosa, J., Tokgoz, S., Hayes, D. and Yu, T. 2008. Use of U.S. croplands for biofuels increases greenhouse gases through emissions from land-use change. Science. 319: 1238-1240. Sheehan, J., Aden, A., Paustian, K., Killian, K., Brenner, J., Walsh, M. and Nelson, R. 2004. Energy and environmental aspects of using corn stover for fuel ethanol. Journal of Industrial Ecology. 7: 117. Sinistore, J. C. and Bland, W. L. 2010. Life-cycle analysis of corn ethanol production in the Wisconsin context. Biological Engineering. 2(3): 147-163. Sinistore, J. C. 2008. Corn ethanol production in the Wisconsin agricultural context: Energy efficiency, greenhouse gas neutrality and soil and water implications. M.S. Thesis. The University of Wisconsin – Madison. Madison, WI. Sissine, F. 2007. Energy independence and security act of 2007: A summary of major provisions (RL34294). CRS Report for Congress: Washington D.C. http://assets.opencrs.com/rpts/RL34294_20071221.pdf. Accessed: Nov 2009. Spatari, S., Zhang, Y. and MacLean, H. L. 2005. Life Cycle Assessment of Switchgrass- and Corn Stover-Derived Ethanol-Fueled Automobiles. Environmental Science and Technology. 39(24): 9750-9758. Spatari, S., Batley, D. M. and MacLean, H. L. 2010. Life cycle evaluation of emerging lignocellulosic ethanol conversion technologies. Bioresource Technology. 101: 654-667. Thomassen, M., Dalgaard, R., Heijungs, R. and de Boer, I. 2008. Attributional and consequential LCA of milk production. International Journal of Life Cycle Assessment. 13: 339-349. Tillman, A. 2000. Significance of decision-making for LCA methodology. Environmental Impact Assessment Review. 20: 113-123. U.S. Department of Agriculture (USDA). 2008. National Agriculture Statistics Service. http://www.nass.usda.gov. Accessed: November 2008. U.S. Department of Agriculture National Agriculture Statistics Service (USDA NASS). 2009. Cropland Data Layer. http://www.nass.usda.gov/Statistics_by_State/Wisconsin /Publications/Cropland_Data_Layer/index.asp. Accessed: Jan 2010. 237

U.S. Department of Energy (DOE). 2011. U.S. Billion-Ton Update: Biomass Supply for a Bioenergy and Bioproducts Industry. Perlack, R. D. and Stokes, B.J. (Leads), ORNL/TM-2011/224. Oak Ridge National Laboratory, Oak Ridge, TN. 227p. U.S. Department of Energy (DOE). 2009a. Energy Efficiency and Renewable Energy, Biomass Feedstock Composition and Property Database. http://www1.eere.energy.gov/biomass/feedstock_databases.html. Accessed: March 2009. U.S. Department of Energy (DOE). 2009b.Energy Information Administration. Average Retail Price of Electricity to Ultimate Customers: Total by End-Use Sector. http://www.eia.doe.gov/fuelelectric.html. Accessed: Feb 2009. U.S. Environmental Protection Agency (EPA). 2010a. National Electric Energy Data System (NEEDS). Retrieved: June 2010. U.S. Environmental Protection Agency (EPA). 2010b. EPA's IPM Base Case 2006 (v3.0).Comprehensive modeling documentation. Retrieved: Oct 2010. U.S. Environmental Protection Agency (EPA). 2010c. Regulation of Fuels and Fuel Additives: Changes to the Renewable Fuel Standard Program; Final Rule. USEPA. Federal Register. 75(58): 14669-15320. U.S. Environmental Protection Agency (EPA). 2010d. EPA Finalizes Regulations for the National Renewable Fuel Standard Program for 2010 and Beyond. Office of Transportation and Air Quality. USEPA. http://www.epa.gov/otaq/renewablefuels/420f10007.pdf. Accessed: 2 Jan 2012. U.S. Environmental Protection Agency (EPA). 2010e. EPA Lifecycle Analysis of Greenhouse Gas Emissions from Renewable Fuels. Office of Transportation and Air Quality. USEPA. http://www.epa.gov/otaq/renewablefuels/420f10006.pdf. Accessed: 2 Jan 2012. U.S. Environmental Protection Agency (EPA). 2009. U.S. EPA MARKAL 9-Region Database (version 1.1); U.S. EPA Office of Research and Development: Research Triangle, NC. 2009. Van Groenigen, J. W., Velthof, G. L., Onema, O., Van Groenigen, K. J. and Van Kessel, C. 2010. Towards an agronomic assessment of N2O emissions: a case study for affable crops. European Journal of Soil Science. 61: 903-913. 238

von Blottnitz, H. and Curran, M. A. 2007. A review of assessments conducted on bio-ethanol as a transportation fuel from a net energy, greenhouse gas and environmental life cycle perspective. Journal of Cleaner Production. 15: 607-619.

Wang, X., Williams, J. R., Gassman, P. W., Baffaut, C., Izaurralde, R. C., Jeong, J. and Kiniry, J. R. 2012. EPIC and APEX: Model use, calibration, and validation. Transactions of the ASABE. 55(3): 1-16. Wang, M. 2009. The Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) Model (Version 1.8b). Center for Transportation Research; Energy Systems Division, Argonne National Laboratory/Department of Energy: Argonne, IL. Weidema, B. P. 1993. Market aspects in product life cycle inventory methodology. Journal of Cleaner Production. 1: 161-166. Weidema, B. 2001. Avoiding Co-Product Allocation in Life-Cycle Assessment. Journal of Industrial Ecology. 4(3): 11-33. West, T. O. and Marland, G. 2002. A synthesis of carbon sequestration, carbon emissions, and net carbon flux in agriculture: comparing tillage practices in the United States. Agriculture, Ecosystems & Environment. 91: 217-232. Wisconsin Department of Natural Resources (WIDNR). 2008. Governor’s Task Force on Global Warming Final Report. Wisconsin’s Strategy for Reducing Global Warming. July 2008. Available online at: http://www.climatestrategies.us/library/library/view/313. Accessed: 13 Dec 2011. Wisconsin Department of Natural Resources (WIDNR). 2011. Surface Water Data Viewer. Available at: http://dnrmaps.wi.gov/imf/imf.jsp?site=SurfaceWaterViewer. Accessed: 13 Dec 2011. Williams, J., Izaurralde, R. C. and Steglich, E. 2008. Agricultural Policy/Environmental eXtender Model. Theoretical Documentation, Version 0604 (BREC Report # 2008-17). AgriLife Research, Texas A&M System, Blackland Research and Extension: Temple, TX, p.130. Worldwatch Institute. 2007. Biofuels for Transport: global potential and implications for sustainable energy and agriculture. Sterling, V.A.: Earthscan. 239

Wu, M., Wang, M. and Huo, H. 2006. Fuel-cycle assessment of selected bioethanol production pathways in the United States. Argonne National Laboratory (ANL/ESD/06-7): Energy Systems Division. Center for Transportation Research. Argonne, IL. Zhang, X., Izaurralde, R. C., Manowitz, D., West, T. O., Post, W. M., Thomson, A. M., Bandaru, V. P., Nichols, J. and Williams, J. R. 2010. An integrative modeling framework to evaluate the productivity and sustainability of biofuel crop production systems. Global Change Biology: Bioenergy. 2: 258-277. Zhang, X. and Manowitz, D. Personal Communication: Nov 2011.

240

Works referenced

Aguirre-Villegas, H. A., Milani, F. X., Kraatz, S. and Reinemann, D. J. 2012. Life Cycle Impact Assessment and Allocation Methods Development for Cheese and Whey Processing. Transactions of the ASABE. 55(2): 1-16. Anex, R. and Reid L. 2009. Post Script to the Corn Ethanol Debate, Reaching Consensus? Journal of Industrial Ecology. 13(6): 996-999. Ashby, M. F. 2009. Materials and the Environment: Eco-Informed Material Choice. Burlington, M.A.: Elsevier Inc. Bals, B. D. and Dale, B. E. 2012. Developing a model for assessing biomass processing technologies within a local biomass processing depot. Bioresource Technology. 106: 161- 169. Bals, B., Wedding, C., Balan, V., Sendich, E. and Dale, B. E. 2011. Evaluating the impact of ammonia fiber expansion (AFEX) pretreatment conditions on the cost of ethanol production. Bioresource Technology. 102(2): 1277-1283. Bals, B., Rogers, C., Jin, M., Venkatesh, B. and Dale, B. 2010. Evaluation of ammonia fibre expansion (AFEX) pretreatment for enzymatic hydrolysis of switchgrass harvested in different season and locations. Biotechnology for Biofuels. 3: 1-11. Bremer, V. R., Liska, A. J., Klopfenstein, T. J., Erickson, G. E., Yang, H. S., Walters, D. T. and Cassman, K. G. 2010. Emissions Savings in the corn-ethanol life cycle from feeding coproducts to livestock. Journal of Environmental Quality. 39: 472-482. Burney, J.A., Davis, S.J. and Lobell, D.B. 2010. Greenhouse gas mitigation by agricultural intensification. Proceedings of the National Academy of Sciences. 107(26): 12052- 12057. Carvalheiro, F., Duarte, L. C. and Gírio, F. M. 2008. Hemicellulose biorefinieries: a review on biomass pretreatments. Journal of Scientific & Industrial Research. 67: 849-864. Cruse, R. M. and Herndl, C. G. 2009. Balancing corn stover harvest for biofuels with soil and water conservation. Journal of Soil and Water Conservation. 64(4): 286-291.

Crutzen, P. J., Mosier, A. R., Smith, K. A. and Winiwarter, W. 2007. N2O release from agro- biofuel production negates global warming reduction by replacing fossil fuels. Atmospheric Chemistry and Physics Discussions. 7: 11191-11205. 241

D’Agostino, A. L., Sovacool, B .K., Trott, K., Ramos, C. R., Saleem, S. and Ong, Y. 2011. What’s the state of energy studies research?: A content analysis of three leading journals from 1999 to 2008. Energy. 36: 508-519. Davis, S. C., Parton, W. J., Del Grosso, S. J., Keough, C., Marx, E., Alder, P. R., and DeLucia, E. H. 2011. Impact of second-generation biofuel agriculture on greenhouse-gas emissions in the corn-growing regions of the US. Frontiers in Ecology and the Environment. doi: 10.1890/110003. Demirba, A. 2009. Political, economic and environmental impacts of biofuels: A review. Applied Energy. 86: S108-S117. Dominguez-Faus, R., Powers, S.E., Burken, J.G., and Alvarez, P.J. 2009. The water footprint of biofuels: A drink or drive issue? Environmental Science and Technology. 43(9): 3005- 3010. Donner, S. D. and Kucharik, C. J. 2008. Corn-based ethanol production compromises the goal of reducing nitrogen export by the Mississippi River. Proceedings of the National Academy of Sciences. 105(11): 4513-4518. Egbendewe-Mondzozo, A., Swinton, S. M., Izaurralde, R. C., Manowitz, D. H. and Zhang, X. 2010. Biomass Supply from Alternative Cellulosic Crops and Crop Residues: A Preliminary Spatial Bioeconomic Modeling Approach. Staff Paper-Michigan State University Department of Agricultural, Food, and Resource Economics. M.I.: East Lansing. Fargione, J., Hill, J., Tilman, D., Polasky, S. and Hawthorne, P. 2008. Land Clearing and the Biofuel Carbon Debt. Science. 319: 1235-1238. Foley, J. A., DeFries, R., Asner, G. P., Barford, C., Bonan, G., Carpenter, S. R., Chapin, F. S., Coe, M. T., Daily, G. C., Gibbs, H. K., Helkowski, J. H., Holloway, T., Howard, E. A., Kucharik, C. J., Monfreda, C., Patz, J. A., Prentice, I. C., Ramankutty, N. and Snyder, P.

K. 2005. Global Consequences of Land Use. Science. 309: 570-574. Ford, A. 1999. Modeling the Environment: an introduction to System Dynamics Modeling of Environmental Systems. Washington, D.C.: Island Press. Frank, A. B., Berdahl, J. D., Hanson, J. D., Liebig, M. A. and Johnson, H. A. 2004. Biomass and Carbon Partitioning in Switchgrass. Crop Science. 44: 1391-1396. 242

Georgescu, M., Lobell, D. B. and Field, C. B. 2011. Direct climate effects of perennial bioenergy crops in the United States. Proceedings of the National Academy of Sciences. 108(11): 4307-4312. Gnansounou, E., Dauriat, A., Panichelli, L. and Villegas, J. 2008. Energy and greenhouse gas balances of biofuels: biases induced by LCA modeling choices. Journal of Scientific & Industrial Research. 67: 885-897. Grace, P. R., Ladd, J. N., Robertson, G. P. and Gage, S. H. 2006. SOCRATES – A simple model for predicting long-term changes in soil organic carbon in terrestrial ecosystems. Soil Biology & Biochemistry. 38: 1172-1176. Graffin, G. W. and Provine, W. B. 2007. Evolution, Religion and Free Will. American Scientist. 95: 294-297. Hong, J., Shaked, S., Rosenbaum, R. K. and Jolliet, O. 2010. Analytical uncertainty propagation in life cycle inventory and impact assessment: application to an automobile front panel. International Journal of Life Cycle Assessment. 15: 499-510. Horne, R., Grant, T. and Verghese, K. 2009. Life Cycle Assessment: Principles, Practice and Prospects. Collingwood, Australia: CSIRO publishing. Kanarek, M. S. and Anderson H. A. 2007. Environmental Epidemiology Practitioners: Looking to the Future October 11, 2006. Environmental Epidemiology Practitioners. 17(11): 911- 913. Keeney, D. 2009. Ethanol USA. Environmental Science and Technology. 43: 8-11. Kemppainen, A. J. and Shonnard, D. R. 2005. Comparative Life-Cycle Assessments for Biomass-to-Ethanol Production from Different Regional Feedstocks. Biotechnology Progress. 21: 1075-1084. Kikuchi, R., Gerardo, R. and Santos, S. M. 2009. Energy lifecycle assessment and environmental impacts of ethanol biofuel. International Journal of Energy Research. 33: 186-193. Kløverpris, J., Wenzel, H., Banse, M., Milà I Canals, L. and Reenberg, A. 2008. Global Land Use Implications of Biofuels: State-of-the-art; conference and workshop on modeling global land use implication in the environmental assessment of biofuels. International Journal of Life Cycle Assessment. 13(3): 178-183. 243

Lau, M. W. and Dale, B. E. 2009. Cellulosic ethanol production from AFEX-treated corn stover using Saccharomyces cerevisiae 424A(LNH-ST). Proceedings of the National Academy of Sciences. 106(5): 1368-1373. Licker, R., Johnston, M., Foley, J. A., Barford, C., Kucharik, C. J., Monfreda, C. and Ramankutty, N. 2010. Mind the gap: How do climate and agricultural management explain the ‘yield gap’ of croplands around the world? Global Ecology and Biogeography. doi: 10.1111/j.1466-8238.2010.00563.x Linquist, B., van Groenigen, K. J., Adviento-Borbe, M. A., Pittelkow, C. and van Kessel, C. 2012. An agronomic assessment of greenhouse gas emissions from major cereal crops. Global Change Biology. 18: 194-209. Lobell, D. B. and Asner, G. P. 2003. Climate and management contributions to recent trends in U.S. agricultural yields. Science Brevia. 299: 1032. Nair, S. S., Kang, S., Zhang, X, Miguez, F. E., Izaurralde, R. C., Post, W. M., Dietze, M. C., Lynd, L. R. and Wullschleger, S. D. 2012. Bioenergy crop models: descriptions, data requirements and future challenges. Global Change Biology Bioenergy. doi: 10.1111/j.1757-1707.2012.01166.x. Nielsen, P. H., Oxenbøll, K. M. and Wenzel, H. 2007. Cradle-to-Gate Environmental Assessment of Enzyme Products Produced Industrially in Denmark by Novozymes A/S. International Journal of Life Cycle Assessment. 12(6): 432-438. Mandel, J. 2011. Miscanthus, Switchgrass Show Promise as Corn Replacements – Study. The New York Times. 14 July 2011. McGeehin, M. A., Qualters, J. R. and Niskar, A. S. 2004. National Environmental Public Health Tracking Program: Bridging the Information Gap. Environmental Health Perspectives. 112(14): 1409-1413. McSwiney, C. P., Bohm, S., Grace, P. R. and Robertson, P. G. 2010. Greenhouse Gas Emissions Calculator for Grain and Biofuel Farming Systems. Journal of Natural Resources & Life Sciences Education. 39: 125-131. Monteith, J. L. 1996. The quest for balance in crop modeling. Agronomy. 88: 695-697. Mullins, K. A., Griffin, W. M. and Matthews, H. S. 2011. Policy Implications of Uncertainty in Modeled Life-Cycle Greenhouse Gas Emissions of Biofuels. Environmental Science and Technology. 45: 132-138. 244

Park, S., Croteau, P., Boering, K. A., Etheridge, D. M., Ferretti, D., Fraser, P. J., Kim, K. R., Krummel, P. B., Langenfelds, R.L., van Ommen, T. D., Steele, L.P. and Trudinger, C. M. 2012. Trends and seasonal cycles in the isotopic composition of nitrous oxide since 1940. Nature Geoscience. 5: 261-265. Patz, J. A., Campbell-Lendrum, D., Holloway, T. and Foley, J. A. 2005. Impact of regional climate change on human health. Nature Reviews. 438: 310-316. Piñeiro, G., Jobbágy, E. G., Baker, J., Murray, B. C. and Jackson, R. B. 2009. Set-asides can be better climate investment than corn ethanol. Ecological Applications. 19(2): 277-282. Rajagopal, D., Hochman, G. and Zilberman, D. 2011. Indirect fuel use change (IFUC) and the lifecycle environmental impact of biofuel policies. Energy Policy. 39: 228-233. Rettenmaier, N., Köppen, S., Gärtner, S. O. and Reinhardt, G. A. 2010. Life cycle assessment of selected future energy crops for Europe. Biofuels, Bioproducts and Biorefining. 4: 620- 636.

Ringer, D. U. 2010. Investigations on atmospheric CO2 impacts of the energetic use of biomass by global carbon cycle models. 17th International Drying Symposium. Magdeburg, Germany 3-6. October 2010. Robertson, G. P., Dale, V. H., Doering, O. C., Hambur, S. P., Melillo, J. M., Wander, M. M., Parton, W. J., Alder, P. R., Barney, J. N., Cruse, R. M., Duke, C. S., Fearnside, P. M., Follet, R. F., Gibbs, H. K., Goldemberg, J., Mladenoff, D. J., Ojima, D., Palmer, M. W., Sharpley, A., Wallace, L., Weathers, K. C., Wiens, J. A. and Wilhelm, W. W. 2008. Sustainable biofuels redux. Science, 322: 49-50. Sendich, E. N., Laser, M., Kim, S. Alizadeh, H., Laureano-Perez, L., Dale, B. and Lynd, L. 2008. Recent process improvements for the ammonia fiber expansion process and resulting reduction in minimum ethanol selling price. Bioresource Technology. 99: 8429-8435. Shapouri, H., Gallagher, P. W., Nefstead, W., Schwartz, R., Noe, S. and Conway, R. 2010. 2008 Energy Balance for the Corn-Ethanol Industry. USDA, Office of the Chief Economist and Office of Energy Policy and New Uses. Agricultural Economic Report number 846. June 2010. Tait, J. 2011. The ethics of biofuels. Global Change Biology Bioenergy. 3: 271-275. Tilman, D., Hill, J. and Lehman, C. 2006. Carbon-Negative Biofuels from Low-Input High- Diversity Grassland Biomass. Science. 314: 1598-1600. 245

Tilman, D., Socolow, R., Foley, J. A., Hill, J., Larson, E., Lynd, L., Pacala, S., Reilly, J., Searchinger, T., Somerville, C. and Williams, R. 2009. Beneficial biofuels – The Food, Energy and Environment Trilemma. Science. 325. www.sciencemag.org. Accessed: 20 July 2009. Varvel, G. E., Vogel, K. P., Mitchell, R. B., Follet, R. F. and Kimble, J. M. 2008. Comparison of corn and switchgrass on marginal soils for bioenergy. Biomass and Bioenergy. 32: 18-21. Wang, M. Q., Han, J., Haq, Z., Tyner, W. E., Wu, M. and Elgowainy, A. 2011. Energy and greenhouse gas emission effects of corn and cellulosic ethanol with technology improvements and land use changes. Biomass and Bioenergy. 35: 1885-1896. Wang, M., Wu, M. and Huo, H. 2007. Life-cycle energy and greenhouse gas emissions impacts of different corn ethanol plant types. Environmental Research Letters. doi: 10.1088/1748- 9326/024001. Wang, M., Lee, H. and Molburg, J. 2004. Allocation of energy use in petroleum refineries to petroleum products; Implications for Life-Cycle energy use and emission inventory of petroleum transportation fuels. International Journal of Life Cycle Assessment. 9(1): 34- 44. Weiss, M., Haufe, J., Carus, M., Brandão, M., Bringezu, S., Hermann, B and Patel, M. K. 2012. A review of the environmental impacts of biobased materials. Journal of Industrial Ecology. 16: S169-S181. White, M. K., Gower, S. T. and Ahl, D. E. 2005. Life cycle inventories of roundwood production in northern Wisconsin: Inputs into an industrial forest carbon budget. Forest Ecology and Management. 219: 13-28. Ximing, C., Zhang, X. and Wang, D. 2011. Land Availability for Biofuel Production. Environmental Science Technology. 45: 334-339. 246

Appendices

I. Appendix 1

Full cropping system variable table

Allocation Nitrogen Method Treatment Application Crop Tillage (between grain Number Feedstock Rate Landscape Rotation Type and stover)

1 switchgrass med WI N/A N/A N/A

2 switchgrass high WI N/A N/A N/A

3 switchgrass med MI N/A N/A N/A

4 switchgrass high MI N/A N/A N/A

corn- 5 corn stover med WI corn no-till mass

corn- 6 corn stover high WI corn no-till mass

7 corn stover med WI corn-soy no-till mass

8 corn stover high WI corn-soy no-till mass

corn- 9 corn stover med WI corn no-till energy

corn- 10 corn stover high WI corn no-till energy

11 corn stover med WI corn-soy no-till energy

12 corn stover high WI corn-soy no-till energy

corn- 13 corn stover med WI corn no-till sub-division

corn- 14 corn stover high WI corn no-till sub-division

15 corn stover med WI corn-soy no-till sub-division 247

16 corn stover high WI corn-soy no-till sub-division

corn- chisel 17 corn stover med WI corn till mass

corn- chisel 18 corn stover high WI corn till mass

chisel 19 corn stover med WI corn-soy till mass

chisel 20 corn stover high WI corn-soy till mass

corn- chisel 21 corn stover med WI corn till energy

corn- chisel 22 corn stover high WI corn till energy

chisel 23 corn stover med WI corn-soy till energy

chisel 24 corn stover high WI corn-soy till energy

corn- chisel 25 corn stover med WI corn till sub-division

corn- chisel 26 corn stover high WI corn till sub-division

chisel 27 corn stover med WI corn-soy till sub-division

chisel 28 corn stover high WI corn-soy till sub-division

corn- 29 corn stover med MI corn no-till mass

corn- 30 corn stover high MI corn no-till mass

31 corn stover med MI corn-soy no-till mass 248

32 corn stover high MI corn-soy no-till mass

corn- 33 corn stover med MI corn no-till energy

corn- 34 corn stover high MI corn no-till energy

35 corn stover med MI corn-soy no-till energy

36 corn stover high MI corn-soy no-till energy

corn- 37 corn stover med MI corn no-till sub-division

corn- 38 corn stover high MI corn no-till sub-division

39 corn stover med MI corn-soy no-till sub-division

40 corn stover high MI corn-soy no-till sub-division

corn- chisel 41 corn stover med MI corn till mass

corn- chisel 42 corn stover high MI corn till mass

chisel 43 corn stover med MI corn-soy till mass

chisel 44 corn stover high MI corn-soy till mass

corn- chisel 45 corn stover med MI corn till energy

corn- chisel 46 corn stover high MI corn till energy

chisel 47 corn stover med MI corn-soy till energy

chisel 48 corn stover high MI corn-soy till energy 249

corn- chisel 49 corn stover med MI corn till sub-division

corn- chisel 50 corn stover high MI corn till sub-division

chisel 51 corn stover med MI corn-soy till sub-division

chisel 52 corn stover high MI corn-soy till sub-division

250

II. Appendix 2

Full ethanol production system variable table

N Allocation Treatme Applica Method nt tion Land- Crop Tillage (between grain Pretreatment Number Feedstock Rate scape Rotation Type and stover) Method

1 switchgrass med WI N/A N/A N/A Dilute Acid

2 switchgrass high WI N/A N/A N/A Dilute Acid

3 switchgrass med MI N/A N/A N/A Dilute Acid

4 switchgrass high MI N/A N/A N/A Dilute Acid

5 switchgrass med WI N/A N/A N/A AFEX

6 switchgrass high WI N/A N/A N/A AFEX

7 switchgrass med MI N/A N/A N/A AFEX

8 switchgrass high MI N/A N/A N/A AFEX

9 corn stover med WI corn-corn no-till mass Dilute Acid

10 corn stover high WI corn-corn no-till mass Dilute Acid

11 corn stover med WI corn-soy no-till mass Dilute Acid

12 corn stover high WI corn-soy no-till mass Dilute Acid

13 corn stover med WI corn-corn no-till energy Dilute Acid

14 corn stover high WI corn-corn no-till energy Dilute Acid

15 corn stover med WI corn-soy no-till energy Dilute Acid

16 corn stover high WI corn-soy no-till energy Dilute Acid

17 corn stover med WI corn-corn no-till sub-division Dilute Acid

18 corn stover high WI corn-corn no-till sub-division Dilute Acid

19 corn stover med WI corn-soy no-till sub-division Dilute Acid 251

20 corn stover high WI corn-soy no-till sub-division Dilute Acid

21 corn stover med WI corn-corn no-till mass AFEX

22 corn stover high WI corn-corn no-till mass AFEX

23 corn stover med WI corn-soy no-till mass AFEX

24 corn stover high WI corn-soy no-till mass AFEX

25 corn stover med WI corn-corn no-till energy AFEX

26 corn stover high WI corn-corn no-till energy AFEX

27 corn stover med WI corn-soy no-till energy AFEX

28 corn stover high WI corn-soy no-till energy AFEX

29 corn stover med WI corn-corn no-till sub-division AFEX

30 corn stover high WI corn-corn no-till sub-division AFEX

31 corn stover med WI corn-soy no-till sub-division AFEX

32 corn stover high WI corn-soy no-till sub-division AFEX

chisel 33 corn stover med WI corn-corn till mass Dilute Acid

chisel 34 corn stover high WI corn-corn till mass Dilute Acid

chisel 35 corn stover med WI corn-soy till mass Dilute Acid

chisel 36 corn stover high WI corn-soy till mass Dilute Acid

chisel 37 corn stover med WI corn-corn till energy Dilute Acid

chisel 38 corn stover high WI corn-corn till energy Dilute Acid

chisel 39 corn stover med WI corn-soy till energy Dilute Acid 252

chisel 40 corn stover high WI corn-soy till energy Dilute Acid

chisel 41 corn stover med WI corn-corn till sub-division Dilute Acid

chisel 42 corn stover high WI corn-corn till sub-division Dilute Acid

chisel 43 corn stover med WI corn-soy till sub-division Dilute Acid

chisel 44 corn stover high WI corn-soy till sub-division Dilute Acid

chisel 45 corn stover med WI corn-corn till mass AFEX

chisel 46 corn stover high WI corn-corn till mass AFEX

chisel 47 corn stover med WI corn-soy till mass AFEX

chisel 48 corn stover high WI corn-soy till mass AFEX

chisel 49 corn stover med WI corn-corn till energy AFEX

chisel 50 corn stover high WI corn-corn till energy AFEX

chisel 51 corn stover med WI corn-soy till energy AFEX

chisel 52 corn stover high WI corn-soy till energy AFEX

chisel 53 corn stover med WI corn-corn till sub-division AFEX

chisel 54 corn stover high WI corn-corn till sub-division AFEX 253

chisel 55 corn stover med WI corn-soy till sub-division AFEX

chisel 56 corn stover high WI corn-soy till sub-division AFEX

57 corn stover med MI corn-corn no-till mass Dilute Acid

58 corn stover high MI corn-corn no-till mass Dilute Acid

59 corn stover med MI corn-soy no-till mass Dilute Acid

60 corn stover high MI corn-soy no-till mass Dilute Acid

61 corn stover med MI corn-corn no-till energy Dilute Acid

62 corn stover high MI corn-corn no-till energy Dilute Acid

63 corn stover med MI corn-soy no-till energy Dilute Acid

64 corn stover high MI corn-soy no-till energy Dilute Acid

65 corn stover med MI corn-corn no-till sub-division Dilute Acid

66 corn stover high MI corn-corn no-till sub-division Dilute Acid

67 corn stover med MI corn-soy no-till sub-division Dilute Acid

68 corn stover high MI corn-soy no-till sub-division Dilute Acid

69 corn stover med MI corn-corn no-till mass AFEX

70 corn stover high MI corn-corn no-till mass AFEX

71 corn stover med MI corn-soy no-till mass AFEX

72 corn stover high MI corn-soy no-till mass AFEX

73 corn stover med MI corn-corn no-till energy AFEX

74 corn stover high MI corn-corn no-till energy AFEX

75 corn stover med MI corn-soy no-till energy AFEX

76 corn stover high MI corn-soy no-till energy AFEX

77 corn stover med MI corn-corn no-till sub-division AFEX 254

78 corn stover high MI corn-corn no-till sub-division AFEX

79 corn stover med MI corn-soy no-till sub-division AFEX

80 corn stover high MI corn-soy no-till sub-division AFEX

chisel 81 corn stover med MI corn-corn till mass Dilute Acid

chisel 82 corn stover high MI corn-corn till mass Dilute Acid

chisel 83 corn stover med MI corn-soy till mass Dilute Acid

chisel 84 corn stover high MI corn-soy till mass Dilute Acid

chisel 85 corn stover med MI corn-corn till energy Dilute Acid

chisel 86 corn stover high MI corn-corn till energy Dilute Acid

chisel 87 corn stover med MI corn-soy till energy Dilute Acid

chisel 88 corn stover high MI corn-soy till energy Dilute Acid

chisel 89 corn stover med MI corn-corn till sub-division Dilute Acid

chisel 90 corn stover high MI corn-corn till sub-division Dilute Acid

chisel 91 corn stover med MI corn-soy till sub-division Dilute Acid

chisel 92 corn stover high MI corn-soy till sub-division Dilute Acid

chisel 93 corn stover med MI corn-corn till mass AFEX 255

chisel 94 corn stover high MI corn-corn till mass AFEX

chisel 95 corn stover med MI corn-soy till mass AFEX

chisel 96 corn stover high MI corn-soy till mass AFEX

chisel 97 corn stover med MI corn-corn till energy AFEX

chisel 98 corn stover high MI corn-corn till energy AFEX

chisel 99 corn stover med MI corn-soy till energy AFEX

chisel 100 corn stover high MI corn-soy till energy AFEX

chisel 101 corn stover med MI corn-corn till sub-division AFEX

chisel 102 corn stover high MI corn-corn till sub-division AFEX

chisel 103 corn stover med MI corn-soy till sub-division AFEX

chisel 104 corn stover high MI corn-soy till sub-division AFEX

256

III. Appendix 3

Abbreviations and acronyms

ACLCA American Center for Life Cycle Assessment

AFEX Ammonia Fiber Expansion

AP Acidification Potential

ASABE American Society of Agricultural and Biological Engineers

CS Corn Stover

CT Chisel Tillage

DA Dilute Acid

EISA Energy Independence and Security Act of 2007

EP Eutrophication Potential

EPIC Environmental Policy Integrated Climate

GHG Greenhouse Gas

GLBRC Great Lakes Bioenergy Research Center

GWP Global Warming Potential

HN High Nitrogen

HUC Hydrologic Unit Code

LCA Life Cycle Assessment

LCI Life Cycle Inventory

LHV Low Heating Value

MI Michigan

MN Medium Nitrogen

NER Net Energy Ratio 257

NT No Till

NREL National Renewable Energy Laboratory

RFS2 Renewable Fuel Standard 2

RIMA Regionally Intensive Modeling Area

SG Switchgrass

SOC Soil Organic Carbon

USDOE United States Department of Energy

USEPA United States Environmental Protection Agency

WI Wisconsin

IV. Appendix 4

Electricity Grid data (P. Meier Pers. Comm, 2011 and MER, 2011)

Electricity Grid Wisconsin 2010 Michigan 2010 Data* Delivered Conversion Delivered Conversion Electricity [%] Efficiency [%] Electricity [%] Efficiency [%] Coal 54.7 36.1 55.8 36.5 Natural Gas 18.6 48.0 8.1 42.5 Nuclear 20.7 32.8 30.6 32.8 Hydro 2.3 - 1.4 - Wind 2 - 1.3 - Biomass/Landfill 1.5 - 2 - Gas Solar 0.2 - 0.8 - * Note that these data are simulated not historic. Average transmission and distribution losses are 6.47% in both states.

258

V. Appendix 5

Nitrogen application rates for all scenarios (all units are kgN ha-1)

Scenario Wisconsin Michigan Switchgrass High N 90 90 Med N 60 60 Corn Stover Rotation Continuous Corn Tillage Chisel High N 185 175 Med N 148 125 Tillage No Till High N 185 175 Med N 148 125 Rotation Corn-Soy Tillage Chisel High N 155 135 Med N 124 110 Tillage No Till High N 155 135 Med N 124 110 Baseline scenario: Corn-Soy, Chisel Till, High N High N 155 135

259

VI. Appendix 6.

Full-size maps of all crop production scenarios and all results (biomass yield, Net Energy Ratio (NER), Global Warming Potential (GWP), Acidification Potential (AP) and Eutrophication Potential (EP)) for the crop production portion of the LCA depicted graphically by watershed.

Figure 1. Switchgrass yield in the high nitrogen scenario (kg SG dry matter ha-1)

Figure 2. Switchgrass yield in the medium nitrogen scenario (kg SG dry matter ha-1) 260

Figure 3. Switchgrass NER in the high nitrogen scenario (MJ output MJ input-1)

Figure 4. Switchgrass NER in the medium nitrogen scenario (MJ output MJ input-1) 261

-1 Figure 5. Switchgrass GWP in the high nitrogen scenario (gCO2eq ha )

-1 Figure 6. Switchgrass GWP in the medium nitrogen scenario (gCO2eq ha ) 262

-1 Figure 7. Switchgrass AP in the high nitrogen scenario (gSO2eq ha )

-1 Figure 8. Switchgrass AP in the medium nitrogen scenario (gSO2eq ha ) 263

-1 Figure 9. Switchgrass EP in the high nitrogen scenario (gPO4eq ha )

-1 Figure 10. Switchgrass EP in the medium nitrogen scenario (gPO4eq ha ) 264

Figure 11. Corn stover yield in the continuous corn, high N and chisel tillage scenario (kg CS dry matter ha-1)

Figure 12. Corn stover NER in the continuous corn, high N and chisel tillage scenario (MJ output MJ input-1) 265

Figure13. Corn stover GWP in the continuous corn, high N and chisel tillage scenario.

Figure 14. Corn stover AP in the continuous corn, high N and chisel tillage scenario. 266

Figure 15. Corn stover EP in the continuous corn, high N and chisel tillage scenario.

Figure 16. Corn stover yield in the continuous corn, medium N and chisel tillage scenario. 267

Figure 17. Corn stover NER in the continuous corn, medium N and chisel tillage scenario.

Figure 18. Corn stover GWP in the continuous corn, medium N and chisel tillage scenario. 268

Figure 19. Corn stover AP in the continuous corn, medium N and chisel tillage scenario.

Figure 20. Corn stover EP in the continuous corn, medium N and chisel tillage scenario. 269

Figure 21. Corn stover yield in the continuous corn, medium N and no till scenario.

Figure 22. Corn stover NER in the continuous corn, medium N and no till scenario. 270

Figure 23. Corn stover GWP in the continuous corn, medium N and no till scenario.

Figure 24. Corn stover AP in the continuous corn, medium N and no till scenario. 271

Figure 25. Corn stover EP in the continuous corn, medium N and no till scenario.

Figure 26. Corn stover yield in the continuous corn, high N and no till scenario. 272

Figure 27. Corn stover NER in the continuous corn, high N and no till scenario.

Figure 28. Corn stover GWP in the continuous corn, high N and no till scenario. 273

Figure 29. Corn stover AP in the continuous corn, high N and no till scenario.

Figure 30. Corn stover EP in the continuous corn, high N and no till scenario. 274

Figure 31. Corn stover yield in the corn-soy, medium N and chisel till scenario.

Figure 32. Corn stover NER in the corn-soy, medium N and chisel till scenario. 275

Figure 33. Corn stover GWP in the corn-soy, medium N and chisel till scenario.

Figure 34. Corn stover AP in the corn-soy, medium N and chisel till scenario. 276

Figure 35. Corn stover EP in the corn-soy, medium N and chisel till scenario.

Figure 36. Corn stover yield in the corn-soy, high N and chisel till scenario. 277

Figure 37. Corn stover NER in the corn-soy, high N and chisel till scenario.

Figure 38. Corn stover GWP in the corn-soy, high N and chisel till scenario. 278

Figure 39. Corn stover AP in the corn-soy, high N and chisel till scenario.

Figure 40. Corn stover EP in the corn-soy, high N and chisel till scenario. 279

Figure 41. Corn stover yield in the corn-soy, medium N and no till scenario.

Figure 42. Corn stover NER in the corn-soy, medium N and no till scenario. 280

Figure 43. Corn stover GWP in the corn-soy, medium N and no till scenario.

Figure 44. Corn stover AP in the corn-soy, medium N and no till scenario. 281

Figure 45. Corn stover EP in the corn-soy, medium N and no till scenario.

Figure 46. Corn stover yield in the corn-soy, high N and no till scenario. 282

Figure 47. Corn stover NER in the corn-soy, high N and no till scenario.

Figure 48. Corn stover GWP in the corn-soy, high N and no till scenario. 283

Figure 49. Corn stover AP in the corn-soy, high N and no till scenario.

Figure 50. Corn stover EP in the corn-soy, high N and no till scenario.

284

VII. Appendix 7.

Full size graphs of all corn stover ethanol LCA results (Net Energy Ratio (NER), Global Warming Potential (GWP), Acidification Potential (AP) and Eutrophication Potential (EP)). All scenario labels appear in the following order: Tillage, Nitrogen Application, Pretreatment, Allocation.

Variable Type Abbreviation State Wisconsin WI Michigan MI Crop Rotation Continuous Corn ContCorn Corn Soybean CornSoy Tillage Chisel Tillage CT No Till NT Nitrogen Application Medium Nitrogen MN High Nitrogen HN Pretreatment Ammonia Fiber AFEX Method Expansion Dilute Acid DA Allocation Method Mass M Energy E Sub Division SD

Figure 1. Corn stover ethanol LCA NER (MJ output MJ input-1) across all scenarios and allocation methods with range of values introduced by spatial variability.

-1 Figure 2. Corn stover ethanol LCA GWP Intensity (gCO2eq MJ ) across all scenarios and allocation methods with range of values introduced by spatial variability. 285

-1 Figure 3. Corn stover ethanol LCA AP (gSO2eq MJ ) across all scenarios and allocation methods with range of values introduced by spatial variability.

-1 Figure 4. Corn stover ethanol LCA EP (gPO4eq MJ ) across all scenarios and allocation methods with range of values introduced by spatial variability. 286

287

VIII. Appendix 8. Life Cycle Inventory (LCI) data input and source table

Fossil Fuel Data Source Agricultural Data Source Transport Data Source Production LCI Input LCI Input LCI Input Natural Gas US LCI, 2008 Nitrogen US LCI, 2008 Train, diesel US LCI, 2008 Fertilizer powered Gasoline US LCI, 2008 Potassium US LCI, 2008 Barge, diesel US LCI, 2008 Fertilizer powered Diesel US LCI, 2008 Phosphorus US LCI, 2008 Barge, US LCI, 2008 Fertilizer residual fuel oil powered Bituminous US LCI, 2008 Pesticide: EcoInvent, Combination US LCI, 2008 Coal glyphosate 2007 Truck, diesel powered Residual Fuel US LCI, 2008 Maize Seed EcoInvent, Truck PE, 2006 Oil 2007 (52,000lbs or 23 MT payload), diesel powered Liquefied US LCI, 2008 Agricultural US LCI, 2008 Petroleum Lime Gas Heat and Data Source SG Seed EcoInvent, Chemical Data Source Power LCI 2007 and Input Biorefinery LCI Input Residual Fuel US LCI, 2008 Maize Seed EcoInvent, Sulphuric PE 2006 Oil 2007 Acid Combusted in Industrial Boiler Diesel US LCI, 2008 No Till PE, 2006 Ammonia US LCI, 2008 Combusted in planter Industrial Boiler Gasoline US LCI, 2008 Soil PE, 2006 Diammonium EcoInvent, Combusted in Cultivation: Phosphate 2007 equipment Chisel Tiller Bituminous US LCI, 2008 Fertilising: PE, 2006 Ammonium EcoInvent, Coal Mineral sulphate 2007 Combusted in fertilizing Industrial (lime) 288

Boiler Natural Gas US LCI, 2008 Fertilising: PE, 2006 AFEX B. Bals Pers. Combustion Sprayer pretreatment Comm, 2011; in Industrial Laser et al., Boiler 2009 Liquefied US LCI, 2008 Soil PE, 2006 DA Humbird et Petroleum Cultivation: Pretreatment al., 2011 Gas Chisel Tiller Combusted in Industrial Boiler Regional MER, 2011; Sowing: Plant PE, 2006 Hydrolysis Humbird et emission EPA, 2010a Drill al., 2011 factors for and electric power EPA, 2010b Regional ANL, 2010 Sowing: PE, 2006 Fermentation Humbird et emission Spike Harrow al., 2011 factors for transportation and process- heating fuels Power from PE, 2006 Harvesting: PE, 2006 Cellulase Humbird et Nuclear straw/hay Production al., 2011 Power Plant harvester and baling Power from PE, 2006 Switchgrass MI: Zhang, et Lignin Humbird et Coal and corn al. 2010 Combustion al., 2011 stover WI: Kang and and Utilities biomass Yield Post, 2012 Power from PE, 2006 Corn Grain MI: Zhang, et Wastewater Humbird et Natural Gas Yield al. 2010 Treatment al., 2011 WI: Kang and Post, 2012 Power from PE, 2006 N loss via MI: Zhang, et Distillation, Humbird et Hydropower surface runoff al. 2010 Dehydration al., 2011 WI: Kang and and Post, 2012 Denaturation Power from PE, 2006 N loss via MI: Zhang, et Storage and Humbird et Wind sediment al. 2010 Handling al., 2011 WI: Kang and Post, 2012 N loss via MI: Zhang, et lateral al. 2010 subsurface WI: Kang and flow Post, 2012 N loss via MI: Zhang, et 289

percolation al. 2010 WI: Kang and Post, 2012 P loss via MI: Zhang, et surface runoff al. 2010 WI: Kang and Post, 2012 P loss via MI: Zhang, et sediment al. 2010 WI: Kang and Post, 2012 P loss via MI: Zhang, et percolation al. 2010 WI: Kang and Post, 2012 Net Primary MI: Zhang, et Production al. 2010 WI: Kang and Post, 2012 C content of MI: Zhang, et yield al. 2010 WI: Kang and Post, 2012 CO2 respired MI: Zhang, et al. 2010 WI: Kang and Post, 2012 Soil C lost MI: Zhang, et al. 2010 WI: Kang and Post, 2012 N2O MI: Zhang, et emissions al. 2010 from soils WI: Kang and Post, 2012

290

IX. Appendix 9. WI and MI Cation Exchange Capacity (CEC) and soil pH maps (first 10 cm depth). Map data sourced from SSURGO via the National Resource Conservation Service soils data mart (NRCS, 2008). Wisconsin maps were assembled by Tim Meehan and MI maps were assembled by David Duncan.

Figure 1. CEC (milli-equivalent of hydrogen per 110 grams of dry soil or meq+ 100g-1) map of WI RIMA. (Red indicates low/poor CEC while blue indicates high/good CEC). 291

Figure 2. CEC map of MI RIMA (meq+ 100g-1). (Red indicates low/poor CEC while blue indicates high/good CEC).

292

Figure 3. Soil pH map of WI RIMA. Yellow-green indicates moderately to slightly acidic (pH range 5.6-6.0 and 6.1-6.5) while teal indicates neutral (pH range 7.4-7.8) and light blue indicates slightly alkaline (pH range 7.4-7.8).

Figure 4. Soil pH map of MI RIMA (same scale as the WI RIMA pH map). Yellow-green indicates moderately to slightly acidic (pH range 5.6-6.0 and 6.1-6.5) while teal indicates neutral (pH range 7.4-7.8) and light blue indicates slightly alkaline (pH range 7.4-7.8).

293

X. Appendix 10 Yates computation results for SG and CS ethanol production LCA

Table 1. Effect size of pretreatment, nitrogen application rate and RIMA on the NER of SG ethanol according to the function NER = Intercept + R + N + P +RxN + RxP + NxP.

Effect Coefficient Coefficient as percent of overall Mean* Intercept 5.7625 Pretreatment (P) 2.175 32% Nitrogen application 0.175 3% rate (N) RIMA (R) -0.025 0% RxN -0.25 4% RxP -0.05 -1% NxP -0.05 -1% *Overall mean NER = 6.83.

Table 2. Effect size of pretreatment, nitrogen application rate and RIMA on the GWP of SG ethanol according to the function GWP = Intercept + R + N + P +RxN + RxP + NxP.

Effect Coefficient Coefficient as percent of overall Mean* Intercept 13.2 Pretreatment (P) -5.1 45% Nitrogen application 3.5 31% rate (N) RIMA (R) -2.7 24% RxN 1.3 12% RxP -0.72 6% NxP 0.68 6% *Overall mean GWP = 11.4.

Table 3. Effect size of pretreatment, nitrogen application rate, RIMA and tillage type on the NER, GWP, AP and EP of corn stover ethanol from continuous corn rotation by allocation method.

Continuous Corn Using averages instead of intercept for % calculations. NER GWP AP EP Coef.* COV** Coef. COV Coef. COV Coef. COV Sub Division Intercept 6.95 51.1 0.025 0.0186

RIMA (R) 0.07 1% -1.05 -2% -0.001 -4% 0.0012 6% Nitrogen (N) 0.01 0% -1.3 -3% -0.001 -4% -0.0014 -7% Tillage (T) -0.01 0% -1.31 -3% -0.001 -4% 0.0009 4% 294

Pretreatmen 2.39 29% 3.7 7% 0.014 44% 1% t (P) 0.0002 RxN -0.44 -5% -2.21 -4% 0.004 12% 0.0076 37% RxT -0.04 -1% 2.63 5% 0.002 7% 0.0011 5% NxT 0.39 5% 2.31 4% -0.003 -10% -0.0031 -15% Mass Allocation Intercept 5.8 56.28 0.04 0.066

RIMA (R) -0.01 0% -0.83 -1% 0.001 1% 0.055 41% Nitrogen (N) 0.06 1% -1.25 -2% -0.001 -2% 0.004 3% Tillage (T) -0.36 -6% -1.33 -2% 0.001 1% 0.036 28% PreTrt 1.28 21% 4.6 8% 0.017 32% 0.016 12% RxN -0.69 -11% -1.52 -3% 0.008 16% 0.072 54% RxT -0.12 -2% 3.21 6% 0.006 11% -0.028 -21% NxT 0.28 5% 3.19 5% -0.002 -3% 0.004 3% Energy Allocation RIMA (R) 5.16 60.67 0.052 0.106

Nitrogen (N) 0.17 3% -2.04 -3% -0.002 -3% 0.073 35% Tillage (T) 0.02 0% -1.15 -2% -0.001 -1% 0.009 4% Pretreatmen -0.11 -2% -1.33 -2% 0.003 4% t (P) 0.066 32% RIMA (R) 0.72 13% 5.21 8% 0.019 29% 0.027 13% RxN -0.63 -12% -1.3 -2% 0.01 15% 0.105 51% RxT -0.36 -7% 3.17 5% 0.006 9% -0.06 -29% NxT 0.17 3% 3.82 6% -0.002 -2% 0.009 4% *Coefficient

**Coefficient of Variance

Table 4. Effect size of pretreatment, nitrogen application rate, RIMA and tillage type on the NER, GWP, AP and EP of corn stover ethanol from corn soy rotation by allocation method.

Corn Soy Using averages instead of intercept for % calculations. NER GWP AP EP Coef.* COV** Coef. COV Coef. COV Coef. COV Sub Division Intercept 6.95375 49.456 0.02425 0.01418 RIMA (R) 0.00012 0.05625 1% 0.4088 1% 0% 0.00064 4% 5 Nitrogen (N) 0.00125 0% 0.0138 0% 0.00012 0% -0.001 -7% 295

5 Tillage (T) 0.00062 -0.0438 -1% 0.1537 0% 2% 0.00245 16% 5 Pretreatmen 0.01337 2.47625 30% 4.8738 10% 43% -0.0008 -5% t (P) 5 RxN -0.0175 0% 0.0275 0% 0.00025 1% 0.00115 8% RxT -0.0075 0% -0.133 0% -0.00025 -1% -0.0024 -16% NxT 0.0025 0% 0.0025 0% -0.00025 -1% 0.00044 3% Mass Allocation Intercept 6.12875 49.506 0.03575 0.01535 RIMA (R) -0.1775 -3% 2.1925 4% 0.0025 5% 0.00995 36% Nitrogen (N) 0.0025 0% 0.2225 0% 6.51E-19 0% -0.0147 -53% Tillage (T) 124 -0.3925 -6% 2.2275 4% 0.0045 9% 0.03481 % PreTrt 1.395 21% 5.145 10% 0.016 33% 0.00116 4% RxN -0.14 -2% 0.51 1% 0.0015 3% 0.0171 61% RxT - -0.065 -1% -1.805 -3% 0.001 2% -0.0333 119 % NxT -0.015 0% -0.035 0% 0.0005 1% 0.00631 23% Energy Allocation RIMA (R) 5.5 49.573 0.04525 0.01649 Nitrogen (N) -0.0175 0% 3.1963 6% 0.00075 1% 0.01528 40% Tillage (T) 0.0075 0% 0.3962 1% -0.00025 0% -0.0264 -69% Pretreatmen 164 -0.5225 -9% 3.9913 7% 0.00825 14% 0.06245 t (P) % RIMA (R) 0.9675 17% 5.3338 10% 0.01775 30% 0.00257 7% RxN -0.175 -3% 0.7575 1% 0.0025 4% 0.02938 77% RxT - -0.03 -1% -3.323 -6% 0.001 2% -0.0608 159 % NxT -0.015 0% -0.057 0% 0.0005 1% 0.01125 30% *Coefficient

**Coefficient of Variance