THE ECONOMIC PROSPECTS OF CELLULOSIC BIOMASS
FOR BIOFUEL PRODUCTION

By
Subbu Kumarappan

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Agricultural, Food, and Resource Economics
2011

ABSTRACT
THE ECONOMIC PROSPECTS OF CELLULOSIC BIOMASS
FOR BIOFUEL PRODUCTION
By
Subbu Kumarappan
Alternative fuels for transportation have become the focus of intense policy debate and
legislative action due to volatile oil prices, an unstable political environment in many major oil
producing regions, increasing global demand, dwindling reserves of low-cost oil, and concerns
over global warming. A major potential source of alternative fuels is biofuels produced from
cellulosic biomass, which have a number of potential benefits. Recognizing these potential
advantages, the Energy Independence and Security Act of 2007 has mandated 21 billion gallons
of cellulosic/advanced biofuels per year by 2022. The United States needs 220-300 million tons
of cellulosic biomass per year from the major sources such as agricultural residues, forestry and
mill residues, herbaceous resources, and waste materials (supported by Biomass Crop Assistance
Program) to meet these biofuel targets.
My research addresses three key major questions concerning cellulosic biomass supply.
The first paper analyzes cellulosic biomass availability in the United States and Canada. The
estimated supply curves show that, at a price of $100 per ton, about 568 million metric tons of
biomass is available in the United States, while 123 million metric tons is available in Canada. In
fact, the 300 million tons of biomass required to meet EISA mandates can be supplied at a price
of $50 per metric ton or lower.
The second paper evaluates the farmers’ perspective in growing new energy crops, such
as switchgrass and miscanthus, in prime cropland, in pasture areas, or on marginal lands. My
analysis evaluates how the farmers’ returns from energy crops compare with those from other

field crops and other agricultural land uses. The results suggest that perennial energy crops
yielding at least 10 tons per acre annually will be competitive with a traditional corn-soybean
rotation if crude oil prices are high(ranging from $88-$178 per barrel over 2010-2019). If crude
oil prices are low, then energy crops will not be competitive with existing crops, and additional
subsidy support would be required. Among the states in the eastern half of US, the states of
Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South
Carolina, Tennessee, and Virginia are found to be economically more suitable to cultivate
perennial energy crops.
The third paper estimates the optimal feedstock composition of annual and perennial
feedstocks from a biorefinery’s perspective. The objective function of the optimization model is
to minimize the cumulative costs covering harvesting, transport, storage, and GHG costs, of
biomass procurement over a biorefinery’s productive period of 20 years subject to various
constraints on land availability, feedstock availability, processing capacity, contracting needs and
storage. The results suggest that the economic tradeoff is between higher production costs for
dedicated energy crops and higher collection and transport costs for agricultural residues; the
delivered costs of biomass drives the results. These tradeoffs are reflected in optimal spatial
planting pattern as preferred by the biorefinery: energy crops are grown in fields closer to the
biorefinery and agricultural residues can be sourced from fields farther away from the
biorefinery. The optimization model also provides useful insights into the price premiums paid
for annual and perennial feedstocks. For the parameters used in the case study, the energy crop
price premium ranges from $2 to $8 per ton for fields located within a 10 mile radius. For
agricultural residues, the price premiums range from $5 to $16 per ton within a 10-20 mile
radius.

Copyright by
SUBBU KUMARAPPAN
2011

ACKNOWLEDGEMENTS

I am deeply indebted to my major professor Dr. Satish Joshi. His unwavering support,
guidance, and help enabled me complete this dissertation. His insights and incisive analytical
skills sharpened my analysis and enabled me write a better dissertation. My regards are due to
Dr. Heather MacLean (University of Toronto) who supported me through various stages of my
dissertation research. The financial support provided from the Natural Sciences and Engineering
Research Council, Canada, and General Motors is gratefully acknowledged.
I also acknowledge the guidance from my committee members Drs. Soren T Anderson,
Robert J Myers, and Christopher H Peterson. Their comments and questions persuaded me to
think more deeply at these questions and other research ideas. They have made such a great
impact that I consider them as anonymous reviewers for my new research questions. Drs. Scott
Loveridge, Frank Lupi, and Scott Swinton helped me through various stages of my graduate
studies in MSU. I owe great admiration for the support staff in the AFRE department – they have
been wonderful people to work with. My gratitude goes to my family and friends whose
encouragement, faith, and patience saw me through the graduate school.

v

TABLE OF CONTENTS
1. INTRODUCTION
1.1 Introduction………………………………………………………………… ............ 1
1.2 References………………………………………………………………..… ............10
2. BIOMASS SUPPLY FOR BIOFUEL PRODUCTION: ESTIMATES FOR THE UNITED
STATES AND CANADA
2.1 Introduction ................................................................................................................12
2.2 Methods......................................................................................................................14
Agricultural Residues.................................................................................................14
Forest Residues ..........................................................................................................18
Municipal Solid Wastes..............................................................................................19
Energy Crops .............................................................................................................20
2.3 Results ........................................................................................................................25
US Biomass Supply ....................................................................................................25
Canadian Biomass Supply .........................................................................................26
Geographic Distribution ............................................................................................29
Comparison with Recent Estimates ...........................................................................31
2.4 Conclusions ................................................................................................................34
2.5 References ..................................................................................................................36
3. REGIONAL COMPETITIVENESS OF ENERGY CROPS
3.1 Introduction ...............................................................................................................41
Objectives and scope..................................................................................................44
3.2 Literature Review.......................................................................................................46
3.3 Methods......................................................................................................................52
Crop production budgets and Net Present Value (NPV) ...........................................52
NPV with uncertainty ................................................................................................56
Climatic factors and energy crop yields ....................................................................58
Crude oil price and agricultural input prices ...........................................................60
Crude oil price and agricultural output prices ..........................................................62
Simulating crude oil prices ........................................................................................64
Estimating crude oil price volatility ..........................................................................66
Simulating energy crop yields ....................................................................................71
Simulating field crop yields .......................................................................................72
Simulating energy crop production costs ..................................................................72
Simulating field crop production costs ......................................................................74
3.4 Role of Risk Aversion in Energy Crop Adoption ......................................................74
Comparison of NPV values for risk neutral farmers .................................................76
Comparison of CE values for risk averse farmers .....................................................77
3.5 Results and Discussion ..............................................................................................79
Miscanthus and corn-soybean (CS) rotation .............................................................82
Miscanthus and other crops .......................................................................................88
Impact of risk aversion...............................................................................................89
3.6 Conclusions ................................................................................................................90
vi

3.7 Appendices ................................................................................................................92
A. Real Options Value ................................................................................................93
B. Forecasting Historical Volatility...........................................................................96
C. Forecasting Implied Volatility ..............................................................................98
D. Temperature (GDD) and Precipitation Values.....................................................100
E. List of State Extension Websites ............................................................................106
F. Calculation of Break-even Risk Aversion Coefficient ...........................................107
G. Break-even prices for miscanthus replacing wheat ..............................................108
3.8 References .................................................................................................................109
4. CHOICE OF OPTIMUM FEEDSTOCK PORTFOLIO FOR A CELLULOSIC ETHANOL
PLANT – A MULTI-PERIOD LINEAR PROGRAMMING SOLUTION
4.1 Introduction ................................................................................................................119
4.2 Literature Review.......................................................................................................129
4.3 Model .........................................................................................................................134
Model Equations ........................................................................................................142
4.4 Case Study .................................................................................................................148
Harvest Shed around Hugoton, Kansas .....................................................................149
Decision Variables .....................................................................................................151
Cost Parameters.........................................................................................................152
4.5 Results ........................................................................................................................155
Shadow Prices ............................................................................................................160
4.6 Sensitivity Analyses ...................................................................................................163
4.7 Conclusions ................................................................................................................166
4.8 References ..................................................................................................................170

vii

LIST OF TABLES
Table 2.1: Biomass Supply Estimates for the US ..................................................................26
Table 2.2: Biomass Supply Estimates for Canada .................................................................28
Table 2.3: Biomass Feedstock Requirements and Expected Composition to meet Cellulosic
Ethanol Production Targets in the US ...................................................................................29
Table 2.4: Geographical Composition of Feedstock Mix to meet EISA Mandates...............30
Table 2.5: Canada Geographic Distribution ..........................................................................31
Table 3.1: Switchgrass supply curve – summary of different US national level studies.......48
Table 3.2: Comparison of switchgrass and miscanthus crop yields
and production costs – Illinois ...............................................................................................51
Table 3.3: Use of state level crop production budgets in computing NPV............................53
Table 3.4: Possible metrics to evaluate economic competitiveness of energy crops.............55
Table 3.5: Maximum potential yield of energy based on temperature and precipitation ......59
Table 3.6: Correlation between crude oil price and NPK fertilizers (1989-2008).................62
Table 3.7: Three scenarios of crude oil prices ($/barrel) .......................................................65
Table 3.8: Energy input use for energy crop production .......................................................73
Table 3.9a: Model Budget for corn and miscanthus crops in Nebraska (Year 4) ..................81
Table 3.9b: NPV values of corn and miscanthus crop for Nebraska (2010-2019) ................82
Table 3.10: Break-even price for miscanthus ($/ton) – Selected crops, selected states ........88
Table 3.D1: Growing Degree Days in US states ...................................................................101
Table 3.D2: Precipitation in US states (in inches) .................................................................103
Table 3.D3: Fixed costs of land in crop production in US states ..........................................104
Table 3.E1: List of state extension websites used to collect field crop production costs ......106
Table 3.G1: Break-even price for miscanthus when planted in wheat fields ($/ton) ............108
Table 4.1: Differences between agricultural residues and energy crops ...............................120

viii

Table 4.2: Alternative scenarios based on harvesting season
and harvest shed demarcation ................................................................................................126
Table 4.3: Lifecycle GHG Emissions from energy crops and agricultural residues .............140
Table 4.4: Land area available for harvesting cellulosic feedstocks
– case study assumptions .......................................................................................................150
Table 4.5: Seasonal cost variations based on quarterly diesel prices (1990-2009) ...............153
Table 4.6: Parametric values for scenarios A and B in preliminary results ...........................155
Table 4.7: Proportion of energy crops at different
material costs (percentage) – scenario B ...............................................................................164
Table 4.8: Cellulosic biomass raw material cost (cents per gallon) – scenario B .................165
Table 4.9: Impact of higher greenhouse gas prices on
optimal feedstock portfolio – scenario B ...............................................................................165
Table 4.10: Proportion of energy crops under
different land availability (percentage) – scenario B .............................................................166

ix

LIST OF FIGURES
Fig. 2.1: Biomass supply curves for the US...........................................................................26
Fig. 2.2: Biomass supply curves for Canada..........................................................................27
Figure 3.1: Framework of proposed simulation study ...........................................................45
Figure 3.2: Switchgrass production estimates – Summary of Polysys studies ......................47
Figure 3.3: NEMS forecast of crude oil prices 2010-2019 ($/barrel) ....................................65
Figure 3.4: Spot prices volatility of crude oil prices – January 1989-September 2009 ........67
Figure 3.5: Historical volatility forecast
– low, reference and high crude oil prices scenarios .............................................................69
Figure 3.6: Weighted volatility forecast
– low, reference and high crude oil prices scenarios .............................................................70
Figure 3.7: Relationship between NPV and Utility
for risk-neutral and risk-averse farmers .................................................................................77
Figure 3.8: Break even prices for miscanthus ($/ton) ............................................................84
Figure 3.9: Subsidy required for miscanthus ($/ton) .............................................................85
Figure 3.10: Probability that cellulosic biomass subsidies are not required (percentage) .....86
Figure 4.1: Concentric circular harvest shed area around the biorefinery (arrows represent
perpendicular roads used for transport) .................................................................................136
Figure 4.2: Acreage contracted (000 acres) with either feedstock– scenario B.....................157
Figure 4.3: Composition of biomass used for biofuel production (000 tons) – scenario B ...157
Figure 4.4: Spatial and temporal distribution of harvest shed – Scenario A (000 ac) ...........159
Figure 4.5: Spatial and temporal distribution of harvest shed – Scenario B (000 ac) ...........160
Figure 4.6: Price premiums paid for cellulosic biomass by zone and year – Scenario A ......161
Figure 4.7: Price premiums paid for cellulosic biomass by zone and year – Scenario B ......161

x

INTRODUCTION
1.1

Introduction
Alternative fuels for transportation have become the focus of intense policy debate and

legislative action due to volatile oil prices, an unstable political environment in many major oil
producing regions, increasing global demand, dwindling reserves of low-cost oil, and concerns
over global warming. A major potential source of alternative fuels is biofuels produced from
cellulosic biomass, which have a number of potential benefits. First, the raw materials used in the
production of cellulosic biofuels are largely waste materials from agriculture, forestry, or other
non-food crops. The use of wastes overcomes the problems of using food and feed grains, such
as corn, for biofuel production . Second, cellulosic biofuels help reduce greenhouse gas
emissions relative to fossil fuels and other biofuels (e.g. corn ethanol). Third, the biofuels from
cellulosic raw materials can be ‘drop-in fuels;’ that can be more easily integrated with the
existing fuel distribution infrastructure compared to ethanol produced from corn .

1

Recognizing these potential advantages, the Energy Independence and Security Act of
2007 has mandated 21 billion gallons of cellulosic/advanced biofuels per year by 2022 . One of
the major purposes of this act is to facilitate rapid growth of cellulosic and advanced biofuels.
EISA provides various support measures, such as subsidies, loan guarantees, and tax credits for
biorefineries construction, technology development, and cellulosic biomass supply.
Notwithstanding these support measures, the growth in cellulosic ethanol industry has rather
been slow . The major limiting factors are the difficulties and uncertainties associated with
reliable supply of cellulosic biomass . For a rapid development of cellulosic biofuels, it is

1

Advanced cellulosic biofuels (butanol, and other renewable fuels) do not have the corrosive
problem of ethanol.
1

important to create a robust supply of cellulosic biomass raw materials. In this dissertation, I
focus on three specific issues pertaining to cellulosic biomass supply raw materials.
The United States needs 220-300 million tons of cellulosic biomass per year to meet
2

EISA biofuel targets by 2022. Supply of such huge quantities of cellulosic biomass is
unprecedented and raises many key questions: ‘What are the types, their geographic distribution,
and supply costs of cellulosic biomass materials? What regional differences and advantages exist
for energy crops envisioned as a major feedstock by USDA? What price and contract incentives
are required to promote perennial energy crops to make them a reliable feedstock source
alongside annual feedstocks? How does crude oil price uncertainty affect the development of
cellulosic biomass industry? How do the differences between annual vs. perennial feedstocks
affect optimal supplies of multiple feedstock to a biorefinery? And finally, what are the
environmental and ecological implications of diverting biomass from their current uses?’
My research addresses three key questions among these: (i) How much cellulosic
feedstock is available in the US and Canada for biofuel production and at what prices? (ii) How
competitive are dedicated energy crops compared to current conventional crops after taking into
account uncertainties in costs and returns, on a regional basis? (iii) What is the optimal mix of
feedstocks for a given biorefinery, taking into account the differences between feedstocks in
terms of material costs, temporal yield patterns, density of biomass availability, transport costs,
lifecycle GHG emissions and contracting constraints? My analysis focuses on four sources of
cellulosic biomass currently promoted under the Biomass Crop Assistance Program : agricultural

2

assuming a conversion rate of 70-90 gallons per ton of cellulosic biomass
2

3

residues, forestry and mill residues, herbaceous resources, and waste materials. These four
biomass resources are currently available or can be cultivated to meet the biofuel production
targets. Agricultural residues, forestry and logging residues, and municipal solid waste streams
are readily available, are replenished annually, and generally do not have significant other uses.
Perennial biomass sources, such as herbaceous energy crops (e.g., miscanthus and switchgrass),
are said to have huge potential in the near term . The following paragraphs summarize the three
papers and their contributions.
Paper #1: Biomass supply for biofuel production: Estimates for the United States and
Canada
The first paper analyzes cellulosic biomass availability in the United States and Canada..
It estimates static supply functions for the four major cellulosic biomass feedstocks namely,
agricultural residues, dedicated energy crops, municipal solid waste and forest residues.
Agricultural residue quantities are estimated based on the amount of crop output from major
grain crops, including corn, wheat, barley, oats, and rice. The supply costs depend on harvest and
transport costs. Energy crop quantities are derived based on yield assumptions that vary by U.S.
states. The quantity of energy crops available depends on how competitive these crops are in
comparison to existing agricultural uses of land, while the regional production costs generate
their supply costs. Three, municipal solid waste (MSW) quantities are calculated based on the
organic portion that can be used for biofuel production. MSW supply costs depend on the state’s
landfill tipping fees and on sorting and processing costs. Four, forestry and logging residue
quantities are computed based on their current uses for fiber and combustion fuel. For logging
residues, the supply costs include the opportunity costs of harvesting and transporting costs; for
3

The BCAP program provides subsidies or loan guarantees for biofuel and other bioenergy
projects that use cellulosic biomass. BCAP program also supports other sources of biomass
(algae or perennial tree crops) for electricity generation.
3

mill residues, the supply costs include their opportunity costs in existing uses (fiber, fuel or other
feedstocks) and other costs of transporting them to the biofuel plant .
The estimated supply curves show that, at a price of $100 per ton, about 568 million
metric tons of biomass is available in the United States, while 123 million metric tons is
available in Canada. In fact, the 300 million tons of biomass required to meet EISA mandates
can be supplied at a price of $50 per metric ton or lower. At this price, the mix of feedstocks
would comprise mainly of agricultural residues and forest/ mill residues. The quantities of
municipal solid wastes are expected to be meager. Small quantities of energy crops will be
produced only in some southern U.S. states. The estimates of agricultural residue supply are very
sensitive to the assumed fraction of residues that can be sustainably removed from the field; the
potential of municipal solid waste as a feedstock is also found to depend on which components
can be economically converted into liquid biofuels. The results show that most of the cellulosic
biomass required in meeting the current biofuel mandates can be supplied without huge changes
in land use.
Paper #2: Regional competitiveness of energy crops
The second paper evaluates the farmers’ perspective in growing new energy crops, such
as switchgrass and miscanthus, in prime cropland, in pasture areas, or on marginal lands.
Switchgrass and miscanthus are two energy crops widely studied across the United States.
Switchgrass test plots exist in the states of Tennessee, Oklahoma, Maryland, Iowa, Nebraska,
and the Dakotas , while miscanthus is promoted in Illinois and a few other Midwestern states .
My analysis evaluates how the farmers’ returns from energy crops compare with those from
other field crops and other agricultural land uses. In particular, it studies whether diverting the
lands to energy crop production would increase or decrease expected farmers’ net returns over a

4

10 year period. The objective is to identify states in the eastern half of the United States where
the economic returns from energy crops are greater than or equal to the returns from existing
agricultural land uses. The economic returns from energy crops depend on crude oil prices
because they affect the prices for cellulosic biofuels and the derived demand for cellulosic raw
materials (energy crops, in this case), as well as costs of energy crop production. The fluctuations
in crude oil prices, at yearly intervals, are modeled to capture the impact of crude oil price
changes over the next 10 years. The simulated crude oil prices are used to derive empirical
probability distributions for the Net Present Values (NPV) of returns from energy crops and
other land uses. Similarly, I use Monte Carlo simulations to model the empirical distributions in
crop yields, output prices, and agricultural input prices for 2010-2019. Long run cointegration
relationships between crude oil prices and agricultural output and input prices drive the
simulation. Probability distributions of regional temperature and precipitation values in the long
run are employed to simulate energy crop yield distributions These simulated distributions of
returns are used to evaluate the competitiveness of energy crops. Specifically, the probability of
receiving higher returns from energy crops (compared to field crops) is estimated..
When the farmers are risk averse, they do not consider only the expected economic
returns (NPV) of energy crops, but also the risk around the expected returns. I estimate the state
level average farmer risk aversion coefficients using the stochastic power function approach. The
risk adjusted competiveness of energy crops is determined by comparing the certainty equivalent
NPVs of returns.
The results suggest that perennial energy crops yielding at least 10 tons per acre annually
will be competitive with a traditional corn-soybean rotation if crude oil prices are high(ranging
from $88-$178 per barrel over 2010-2019). If crude oil prices are low, then energy crops will not

5

be competitive with existing crops, and additional subsidy support would be required. Among
the states in the eastern half of US, the states of Alabama, Arkansas, Florida, Georgia, Kentucky,
Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia are found to be
economically more suitable to cultivate perennial energy crops.
Farmers in the above states may prefer to adopt energy crops because these crops have a
low dependence on crude oil inputs. Hence, their risk-averse nature might drive them to adopt
more energy crops when crude oil prices climb and stay consistently higher. A key feature of the
results is that the economic returns from major field crops such as corn, soybeans, and wheat are
sensitive to higher crude oil prices. Their returns decrease much more than the returns for energy
crops. This increases the probability of receiving higher NPV from energy crops, making it an
attractive crop alternative. Biomass Crop Assistance Program (BCAP) provides subsidies for
cellulosic biomass. The BCAP subsidy matches the cellulosic biomass market prices dollar-fordollar, up to a maximum of $45 per ton. Such matching subsidies effectively result in higher
subsidy payment when crude oil prices are high, even though there is less of a need for biomass
subsidies. The results from my study depict a clear need for counter cyclical biomass subsidies
tied to crude oil prices instead of the current practice of matching biomass prices.

4

Paper #3: Choice of optimum feedstock portfolio for a cellulosic ethanol plant – A multiperiod linear programming solution
The third paper estimates the optimal feedstock composition of annual and perennial
feedstocks from a biorefinery’s perspective. In identifying the optimal feedstock composition,
the model considers several objectives: (i) ensuring reliable supply of biomass over the entire
productive lifetime of the cellulosic biorefinery, (ii) minimizing procurement costs (including

4

When crude oil prices are low, the cellulosic biofuel prices are low as well, resulting in lower
subsidies.
6

harvest, baling, transport, storage, and seasonal costs), and (iii) maximizing the reduction in
greenhouse gas (GHG) emissions
The objective function of the optimization model is to minimize the cumulative costs
covering harvesting, transport, storage, and GHG costs, of biomass procurement over a
biorefinery’s productive period of 20 years subject to various constraints on land availability,
feedstock availability, processing capacity, contracting needs and storage. The model makes
several important contributions. First, existing studies treat the available biomass quantities in
the region as exogenously given and then try to minimize procurement costs. In comparison, my
proposed model treats biomass acreage to be harvested as an endogenous decision variable
5

subject to overall biomass availability constraints. In addition, transport costs, seasonal costs,
and environmental costs are also endogenously determined as a function of harvesting decisions
in my model. Second, my model explicitly considers changes in yield patterns that occur over a
perennial energy crop’s lifetime and the corresponding impacts on optimal feedstock mix. In
contrast, most other studies assume that average yields and feedstock mix remain constant over a
biorefinery’s lifetime. Third, my model also explicitly incorporates the flexibility with respect to
agricultural residue harvesting, and the necessity to enter into longer term contracts with energy
crop producers that assure harvesting each year, and the resulting cumulative effects on the
feedstock mix. This is accomplished by forcing the model to harvest all energy crop biomass
produced over its 10 year productive period; the lack of this restriction on agricultural residue
harvest is referred to as flexibility in biomass harvest in this model. Fourth and finally, the

5

Mapemba et al (2008) also model area harvested as an endogenous variable in an optimization
model. However, their analysis is specific to one region in Oklahoma. My model extends their
analysis by modeling a general harvest-shed area, with multiple concentric circles around the
biorefinery, which could be applied to any location.
7

impacts of GHG emissions on feedstock sourcing decisions are also evaluated. Thus, for these
reasons, my model provides better insights into the realities of biomass procurement.
The model is used to calculate the optimal feedstock mix for the proposed Abengoa
Biofuels biorefinery in Hugoton, Kansas, which is likely to use both agricultural residues and
dedicated energy crops for cellulosic biofuel production. Accordingly, the parameters of the
optimization model were calibrated for Hugoton, Kansas location, considering one annual
feedstock (corn stover) and one perennial feedstock (miscanthus). Average biomass raw material
costs are estimated at 60-70 cents per gallon of annual cellulosic biofuel production for this
biorefinery. At the optimum, up to two-thirds of the plant’s cellulosic biomass would come from
dedicated energy crops. Though more costly to produce, dedicated energy crops are preferred
due to their higher per-acre yields and wider harvest windows. Alternative harvesting scenarios
suggest that feedstocks that widen harvest windows or feedstocks that have higher biomass yield
densities will feature prominently in the optimal biomass feedstock portfolio.
The economic tradeoff is between higher production costs for dedicated energy crops and
higher collection and transport costs for agricultural residues. The delivered costs of biomass
(which is reduced as part of the cost minimization problem) drives the results. The ideal way to
reduce costs would be to generate higher biomass yields and incur lower transportation costs –
the feedstock that satisfies both these requirements would be ‘energy crops grown closer to the
biorefinery.’ The energy crops can offset their higher costs of production only when located
closely to the biorefinery thereby reducing the transport costs. The differences in transport costs
from different locations is a crucial determinant of optimal cropping pattern. To offset the higher
transport costs, agricultural residues characterized by lower production costs would be preferred.

8

These tradeoffs are reflected in optimal spatial planting pattern as preferred by the
biorefinery: energy crops are grown in fields closer to the biorefinery and agricultural residues
can be sourced from fields farther away from the biorefinery. The optimization model also
provides useful insights into the price premiums paid for annual and perennial feedstocks. For
the parameters used in the case study, the energy crop price premium ranges from $2 to $8 per
ton for fields located within a 10 mile radius. For agricultural residues, the price premiums range
from $5 to $16 per ton within a 10-20 mile radius. Over time, the premiums declined for
agricultural residues and rose for energy crops. In any given time period, the price premiums
increased with proximity to the biorefinery. The results suggest that optimal spatial planting
pattern depend on economic factors (harvesting, and transporting costs) which in turn depend on
the temporal patterns of biomass feedstock yields, their respective yield density (tons per acre)
and area available for sourcing cellulosic feedstocks within the harvest shed.

9

1.2 REFERENCES

10

References
BCAP. "Biomass Crop Assistance Program."
http://www.fsa.usda.gov/FSA/webapp?area=home&subject=ener&topic=bcap.
Cassida, K. A., J. P. Muir, M. A. Hussey, J. C. Read, B. C. Venuto, and W. R. Ocumpaugh.
"Biomass Yield and Stand Characteristics of Switchgrass in South Central U.S.
Environments." Crop Sci 45, no. 2 (2005): 673-81.
EISA. "Energy Independence and Security Act of 2007." http://frwebgate.access.gpo.gov/cgibin/getdoc.cgi?dbname=110_cong_bills&docid=f:h6enr.txt.pdf.
Guffey, Dora. "Chariton Valley Biomass Project." http://www.iowaswitchgrass.com/.
Heaton, Emity A, Frank G Dohleman, and Stephen P Long. "Meeting Us Biofuel Goals with
Less Land: The Potential of Miscanthus." Global Change Biology 14, no. 9 (2008): 200014.
Hill, Jason, Erik Nelson, David Tilman, Stephen Polasky, and Douglas Tiffany. "Environmental,
Economic, and Energetic Costs and Benefits of Biodiesel and Ethanol Biofuels." PNAS
103, no. 30 (2006): 11206-10.
Khanna, Madhu, Basanta Dhungana, and John Clifton-Brown. "Costs of Producing Miscanthus
and Switchgrass for Bioenergy in Illinois." Biomass and Bioenergy 32, no. 6 (2008): 48293.
Maryland Energy Administration. "Biofuel Potential of Switchgrass in Maryland."
www.agnr.umd.edu/maes_exe/dividends/staver1.pdf.
Perrin, Richard, Kenneth Vogel, Marty Schmer, and Rob Mitchell. "Farm-Scale Production Cost
of Switchgrass for Biomass." Bioenergy Research 1, no. 1 (2008): 91-97.
RFA. "Renewable Fuels Association Statistics." http://www.ethanolrfa.org/pages/statistics.
Tilman, David, and Jason Hill. "Corn Can't Solve Our Problem." The Washington Post, March
25, 2007.
USDA. "A Usda Regional Roadmap to Meeting the Biofuels Goals of the Renewable Fuels
Standard by 2022."
http://www.usda.gov/documents/USDA_Biofuels_Report_6232010.pdf.
Vadas, P. A., K. H. Barnett, and D. J. Undersander. "Economics and Energy of Ethanol
Production from Alfalfa, Corn, and Switchgrass in the Upper Midwest, USA." Bioenergy
Research 1, no. 1 (2008): 44-55.
Wallace, Robert, Kelly Ibsen, Andrew McAloon, and Winnie Yee. "Feasibility Study for CoLocating and Integrating Ethanol Production Plants from Corn Starch and Lignocellulosic
Feedstocks." 69. Golden, Colorado: National Renewable Energy Laboratory, 2005.
11

BIOMASS SUPPLY FOR BIOFUEL PRODUCTION: ESTIMATES FOR THE
1
UNITED STATES AND CANADA
2.1

Introduction
Alternative fuels, especially biofuels for transportation have become the focus of

intense policy debate and legislative action due to volatile oil prices, unstable
political/military environment in major oil production regions, rapidly increasing global
demand and dwindling reserves of oil, and concerns over global warming. Promoting
biofuel production is also viewed as means to reduce high agricultural program costs and
to promote rural incomes in North America. While ethanol from grains is expected to
account for most of the US/Canada biofuel production in the short run, ethanol from
lignocellulosic biomass is considered to be more attractive from a long term sustainability
perspective because of significantly lower life cycle greenhouse gas emissions compared
to grain ethanol, widespread domestic feedstock availability, and the potential to
ameliorate the conflict over food v/s fuel use of grains . Reflecting this view, the
renewable fuel standards under the US Energy Independence and Security Act of 2007
(EISA) set forth a phase-in for renewable fuel volumes beginning with 9 billion US
gallons (34 billion liters) in 2008 and growing to 36 billion gallons or 136 billion liters by
2022 . The conventional starch-based biofuel volumes are limited to 15 billion gallons
and advanced biofuel volumes are mandated to be 21 billion gallons including 16 billion
gallons of cellulosic ethanol by year 2022. However, achieving the cellulosic ethanol
mandate critically depends on the availability of biomass in sufficient quantities at
reasonable costs for conversion to liquid fuels.

1

published in BioResources Journal 4(3): 1070-1087 (with Dr. Satish Joshi and Dr.
Heather L. MacLean)
12

Cellulosic feedstocks include agricultural residues, forest and mill residues,
organic/lignocellulosic portion of municipal solid waste, and energy crops grown
purposely for conversion to fuels. Over the last decade, a number of studies, assessing
biomass potential at various regional, national and global scales have been published.
While few studies include all major feedstocks , many focus on a single feedstock, such
as agricultural residue assessments , and , and energy crop assessments . Comprehensive
biomass feedstock assessments at the national level have been completed for the US and
Canada . However, as conclude from their review of 17 studies of biomass energy
supply, the studies vary considerably in their assumptions, models and methodologies
employed, feedstocks covered, temporal and spatial scales, production technology
projections, and policy scenarios.
Four general approaches have been taken in estimating biomass supply potential
in these studies: (1) inventories of potential biomass sources with minimal attention to the
economics of actual supply and prices at which these quantities will be available; (2)
static supply curves which estimate quantities of biomass supplied at various
exogenously determined prices assuming everything else remains constant; (3)
projections of supply quantities in competition with other crops and uses, but under
current productivity and policy conditions; and (4) dynamic projections of supply
quantities in competition with other crops and uses, under projected/potential
productivity and policy conditions, ethanol/gas prices and quantity mandates. In addition,
some studies use a bottom-up approach where biomass potentials are assessed at
local/regional level, which are then aggregated into national level estimates, while some
other studies directly estimate national potentials based on national economic models.

13

Published studies have used a mixture of these approaches and often within a single
study, and the details of all the model and parameter assumptions are often not explicitly
reported. As a result, estimates of bioenergy potential vary significantly across studies,
and comparison and reconciliation of differences becomes difficult. A recent review of
fifteen North American studies, carried out by , also finds large variations in estimated
potential biomass supply across those studies for these reasons.
In this study, supply quantities for various biomass feedstocks in the US and
Canada are estimated using a bottom-up, static supply function approach. The relatively
simple, but consistent approach will provide more realistic estimates of short term supply
of biomass feedstocks. Such short-run estimates will be useful for assessing feasibility of
proposed cellulosic ethanol facilities. While similar supply functions have been
previously estimated for specific feedstocks or regions, a major contribution of this study
is providing comprehensive estimates for both US and Canada, covering all major
feedstocks.
2.2

Methods
Estimates of biomass supply from, agricultural residues, forest and mill residues,

cellulosic portions of municipal solid waste (MSW) and energy crops were developed.
The methods used for each feedstock are summarized below.
Agricultural Residues
For estimating supply of agricultural residues a procedure similar to Gallagher et
al (2003) was followed. Agricultural residue supply functions were estimated at the
individual county level for the US and at the census subdivision level for Canada. These

14

supply functions were then aggregated to estimate national level supply functions. The
steps in the estimation procedure are as follows.
The average crop output data for the years 2000-2002 was used to estimate the
total quantity of residue generated employing average residue factors i.e., residue
generated per ton of crop output. Assumed residue factors were 1.5, 1, 1 and 1.27 for
barley, corn, rice and wheat respectively. Four major crops: corn, wheat, rice and barley
are considered for the US estimates. The crop output data are from National Agricultural
Statistics Service . How much of this total quantity of residue generated will be supplied
to the market for conversion is a decision made by the farmers, which is governed by:
regulatory restrictions on residue removal and soil cover to prevent soil erosion, residue
harvesting, storage and transportation costs, opportunity cost of soil fertilization from
leftover residues, animal feed value of the residue, and other opportunity costs. The
residue that needs to be left on the field to prevent soil erosion depends on local
topographic, soil and wind conditions. It was assumed that recommended amounts of
crop residues, that keep the soil erosion below the threshold levels, are left on the field as
soil cover. Gallagher et al (2003) estimate that for corn, 0.65 tons (t) per acre of chopped
corn stover left in the fall fulfills soil erosion prevention requirements. Similarly for
wheat and other small grains 0.32 t/acre of fall residues satisfy the requirement including
the loss of residues during the winter . For winter wheat fallow, the winter loss occurs
twice, so the minimum fall residue would be 0.46 t/acre . An average of these two
estimates (0.39 t/acre) was used for wheat, barley and rice.
Next, the assumption was that if the price offered by ethanol conversion facilities
is lower than the feed value of the residues, farmers will first sell the residues as cattle-

15

feed until the local forage demand is met; however, at prices higher than the feed value,
this additional quantity of residues will be supplied for conversion to ethanol. The
estimates for the feed values are updated from Gallagher et al. (2003) and range from
$28.70/t (for wheat) to $56.74/t (for corn) in 2008 dollars. The forage demand was
estimated based on county livestock population and hay crop production using the
relation: Forage demand = County Cattle Population * Daily Feed Requirements * (365 –
Pasture feeding season length) – local hay production.
Excess residues after meeting soil conservation needs and the local animal feed
requirements will be available for conversion to ethanol only if the price offered is high
enough to compensate for the costs incurred for harvesting, storage and transporting the
residues. Further, residues left on the field have some fertilizer value, which subsequently
reduces fertilizer requirements for the next season. The price offered should also cover
this lost fertilizer value. The harvesting and transportation costs and lost fertilizer values
for different residues were estimated and it was assumed that farmers will supply
remaining residues at a price equal to the sum of these costs. The main operations in
harvesting include chopping, baling and hauling. The estimated chopping, baling and
hauling (within the farm) expenses were $25 per acre on an average. The fertilizer
nutrient value of barley straw, corn stover, rice straw and wheat straw were estimated at
$14.20, $12.30, $10.30 and $9.50 per metric ton respectively .
The farm to factory gate transportation costs depend on the average transportation
distance, which in turn depends on the density of crop residues and the size of the
conversion plant. For a plant with an annual capacity of Q (in metric tons) of residues, the
radius of the collection area (assuming a circular one) is (Q/πd)0.5, where d is the density

16

of residue availability in t/square mile. The average distance of collection is
0.67(Q/πd)0.5. The county residue density d was calculated by dividing the total quantity
of residue available after meeting soil conservation and cattle feed requirements, by the
total land area of the county. Harvesting and transportation costs were calculated by using
weighted average costs of different crops, where the proportion of the different crop
residue available after meeting soil conservation requirement was used as a weighting
factor. It was also assumed that the ethanol conversion plant had a processing capacity of
2000 metric tons per day (tpd) of residues and average transport costs were $0.40/ton
mile in estimating transportation costs.
Similar procedures were employed to estimate agricultural residue supply
functions for Canada. However, because Canada does not produce any rice, oats was
considered instead of rice. A dataset at the Census Sub Division (CSD) level for Canada,
similar to the county level dataset for the US was developed. Data on crop area harvested
and cattle population were collected from 'Statistics Canada' publications. Average crop
yield data (barley, corn, oats and wheat) for the years 1999-2003 are from Statistics
Canada reports . The data is reported at the CSD level. Other parameters such as residue
factors, feed value, fertilizer value, length of foraging season etc. were assumed to be
similar to those for the US. Because data on crop yields at the CSD level were not
available, the average yield and residue density at the Canadian provincial level were
assumed to be the same for all CSD within that province. The individual county/CSD
supply functions were then aggregated to derive national level agricultural residues
supply function for Canada.

17

Forest Residues
Forest residues comprise of logging residues that are generated during the
harvesting operations, and mill residues that are generated in saw mills, paper mills and
other wood processing units. Logging residues are currently left at harvesting sites and
hence need to be collected and transported from the forests, while mill residues are
available at processing facilities and currently being used either as fuel or as raw material
for other wood products.
reports the total quantities of logging residues and mill residues produced in
various states of the US, which are based on the USDA Forest Service’s Timber Product
Output database for 2002. Logging residue quantities for Canada were computed from
the total roundwood production reported at provincial level in Canadian national forestry
database for the year 2006 . For Canada, it was assumed that logging residue production
would be16% of the total roundwood production (Mabee, 2006). In comparison, logging
residue estimates for the US, vary between 4% and 28% of total roundwood production .
Data on the quantity of Canadian mill residues were drawn from .
It was assumed that all the logging residues produced in a US state or Canadian
province would be available at a price equal to the sum of grinding costs and
transportation costs. Mill residues were assumed to be available for conversion if the
price offered was greater than the opportunity cost of their current use as fiber, fuel or
other feedstocks. The US state level estimates of mill residues used as fiber, fuel and
other applications were from . The opportunity costs for the various types of residues
were estimated based on their current use: the mill residues used for fiber products
(pulpwood) were valued at $36/dry ton (dt), the fuel use was valued at $23.65/dt (i.e.

18

$1.25/million BTU based on coal price) and all other uses were valued at $16/dt. The
remaining residues that are not currently used were assumed to be available for free . It
was assumed that logging residues were uniformly distributed in the timberland area of
the region being considered and the average transport distance required to supply a 2000
tpd ethanol plant was calculated using similar methods as outlined above for agricultural
residues. The logging residues from the forests were valued at their fuel use value
($23.65/dt) which would be their opportunity costs in heat and power production. The
estimated average distance for collection of forest logging residues varied between 19 to
55 miles based on the state geographic area and density of forest residues. These distance
estimates were combined with the transportation cost of $0.40 per ton-mile as in the
previous case to compute the transportation costs. Since mill residues were readily
available at the processing facilities (like paper and pulp mills) a transport cost of $5/t
was assumed for mill residues. The estimated state/province level supply functions were
then aggregated to estimate the national supply functions.
Municipal Solid Wastes
The USEPA and the BioCycle magazine annually estimate the total quantities of
municipal solid waste (MSW) generated, recovered and discarded in the US. The USEPA
estimates are at an aggregate national level based on material flow analysis, while the
Biocycle estimates are based on an annual survey of state level MSW officials. In
Canada, Statistics Canada publishes biannual data on waste disposal and diversion at the
provincial level. We use state level estimates of MSW generated from the 15th annual
survey conducted by the BioCycle magazine and Earth Engineering Center of Columbia
University . It was assumed that 66% of the wastes were organic materials suitable for

19

cellulosic ethanol production, based on estimates from USEPA . An average moisture
content of 40% was assumed in deriving dry biomass equivalent of MSW feedstock,
based on estimated moisture content of various constituents in MSW which can vary
from 2 to 70% .
Since a well-established collection system for MSW is currently operating, and
tipping fees are paid to dispose MSW, all the lignocellulosic portion of MSW is
essentially available (albeit at the landfill) at a negative price equivalent to the current
landfill tipping fee. The tipping fees in 2006 ranged from $21/t to $123/t in the US and
from $40/t to $75/t in Canada . The total quantity of MSW currently landfilled in a
state/province was assumed to be available at a negative price equivalent to the average
tipping fee for the particular US state or Canadian province. Because the cellulosic
portion of MSW needs to be separated from other constituents before conversion to
ethanol, estimated separation/sorting and transportation costs of $55/wet ton was added to
the negative price to estimate the ethanol feedstock supply price. The processing cost
estimate is based on an update of a previous estimate . The state (province) level quantity
and price data were then combined to estimate the total quantity of MSW feedstock
supplied at various prices at national level.
Energy Crops
The potential supply of energy crops when they are competing with conventional
crops is a complex function of several factors. Farmers will switch to energy crops only if
expected returns from the energy crops are higher than returns from growing
conventional crops and/or keeping the land idle under conservation programs such as the
US Conservation Reserve Program (CRP) and collecting rental payments as well as the

20

government support payments. In addition to switching to energy crops, farmers will also
be switching between conventional crops based on relative expected returns. These
returns are governed by expected prices for different crops, yield and production cost
structure for energy crops compared to that of conventional crops. These costs and yields
also differ by the geographical location. Hence, the estimation of potential supply of
energy crops has to be carried out using an integrated model of agriculture that
incorporates the inter-dependencies across individual commodity grain, livestock, dairy,
consumption and international sectors, as well as agricultural policy variables. Most of
the current projections of energy crop supply in the US, are derived from an integrated
comprehensive model of US agriculture, POLYSYS, developed and maintained by the
Agricultural Policy Analysis Center (APAC), University of Tennessee, Knoxville . In
fact, most of the estimates of bioenergy crop supply published by the US Department of
Energy (USDOE) and US Department of Agriculture (USDA) draw on the POLYSYS
model estimates. For example, see Perlack et al. (2005) and De La Torre Ugarte et al.
(2000, 2003). However, estimates of energy crop supply using the POLYSYS model vary
significantly depending on variations in the underlying production/supply constraints, and
assumptions about energy crop productivity, relative profitability and policy variables.
To derive static supply curves for energy crops that were consistent with other
biomass feedstock estimates, a relatively simple approach was used. It was assumed that
farmers will potentially divert land from current traditional crops to energy crops (e.g.
switchgrass) if the ‘returns over variable costs’ (ROVC) for switchgrass were more than
the returns over variable costs for the traditional crops. Since the current ROVC were
adequate enough to cover the fixed costs such as land value and opportunity costs of

21

labor and overhead charges, and retain the land under production, a higher ROVC from
energy crop production would make switching to energy crops attractive.
Using county level crop production data from USDA - ERS, gross revenues per
acre for various crops namely – corn, soybeans, wheat, rice, barley, oats, and cotton at the
county level, were calculated using state average commodity prices and variable costs of
production . The variable costs of production included the costs of seeds, fertilizers,
pesticides, energy for machinery operations and custom work. The returns over variable
costs (ROVC – which is an estimate of fixed costs of agricultural production that are
being covered currently) were calculated by subtracting these variable costs from total
revenues. Average ROVC for the period 2002-05 for all major crops in each US County
along with harvested acreages were computed. Government payments were not included
in ROVC estimates because these payments have been effectively decoupled from
commodity production since the passage of Farm Bill 2002 .
Next ROVC for growing switchgrass at various switchgrass prices were estimated
at the county level. Counties in the eastern half of the US including the Dakotas,
Oklahoma and Texas, where switchgrass can be grown under rain-fed conditions were
considered. To account for differences in state level yields and costs of production, the
states were divided into three regions: south, central and north. In the southern region
(AL, AR, FL, GA, KY, LA, MS, NC, OK, SC, TN, TX, VA, WV) the energy crop yield
was assumed to be higher at 8 dt/acre at an average variable cost of $259.56 per acre . In
the central states (IL, IN, IA, KS, MO, NE, NY, OH, PA) the yield was assumed to be
4.45 dt/acre at an average variable cost of $211.71/acre . The northern states (CT, DE,
ME, MD, MA, MI, MN, NJ, RI, SD,VT, WI) are located in much colder climates and the

22

yields were found to average only 2.79 dt/ acre with variable costs of $128.32 per acre .
The variable costs include seed costs, initial establishment costs, fertilizer costs,
harvesting costs and baling costs. It was assumed that once established, switchgrass can
be harvested over a period of 10 years. These yield and production costs are based on
realized values, however improvements in switchgrass yields are likely in the future.
Although other energy crops such as miscanthus and energy cane have been considered
potentially attractive, the analysis is limited to switchgrass, to be consistent with previous
analyses .
The average annual ROVC of field crops were compared with ROVC from
switchgrass at various switchgrass prices and it was assumed that 10% the land currently
earning lower ROVC than that of switchgrass would be converted to energy crop
production.2 A variety of factors affect the crop-switching decision including subsequent
changes in relative crop prices and returns as a result of crop switching, land
characteristics, local weather/rainfall conditions, expectations, farmer expertise and risk
preferences. 10% land conversion was used as an indicative aggregate constraint resulting
from all these considerations. The estimates using this simplification are consistent with
the earlier work using POLYSYS models that estimated 5 to 14 per cent of crop land
being converted into energy crops . Transportation costs from the farm to factory gate
were estimated employing a similar approach as for agricultural residues.
2

To illustrate, Baldwin County (FIPS code 1003) in Alabama had an average of 4,250
acres under corn, 10,167 acres under soybeans and 7,933 acres under wheat crop during
2002-05. The net returns over variable costs for those crops during those years were
$140, $109 and $61 per acre respectively. If the energy crops were able to generate a
return of $50 per acre, then none of the land under these crops would be diverted to
energy crops. But if the ROVC from energy crops were $125 per acre, then we assume
that 10% the soybeans and wheat acreage will be converted to energy crops as the latter is
more profitable.
23

Canadian energy crop supply
The Whole Farm Database managed by Statistics Canada reports harvested acres
and total revenues for various crops at the Census Agricultural Region level (table series
C) in Canada . However the expenditure data are not available separately by individual
crops but aggregate expenditure per acre for the portfolio of crops is available. Hence the
average farm ROVC (dollars per acre) from the existing ‘portfolio’ of crops was
compared with the potential returns from energy crops. The Canadian crop portfolios
included the major crops such as wheat, oats, barley, rapeseed, soybeans, corn, small
grains and forage crops. The land area switching to energy crops at various energy crop
prices was estimated using similar procedures outlined above for the US energy crops.

24

2.3

Results

US Biomass Supply
Figure 2.1 and Table 2.1 summarize the biomass supply estimates for the US.
The supply curves are shown only up to the price level of $100 per dry metric ton (dt),
beyond which biomass use for ethanol conversion will likely be uncompetitive. The total
biomass potentially available in the US, at a price of $100/dt is 568 million dt,
comprising of about 250 million dt (44% of total supply) of agricultural residues, and 135
million dt (24%) of forest and mill residues. Under the assumption of planting a
maximum of 10 per cent of crop land in the eastern half of US, energy crops such as
switchgrass and miscanthus would approximately yield 107 million dt of biomass; of
which about 85 million dt will be available at a price of $100/dt.
From Table 2.1, it can be seen that 384 million dt of this biomass will be available
at price of $50 /dt in the US – primarily from agricultural residues and forest feedstocks.
Compared to that the minimum price at which switchgrass starts becoming available is
$67/dt. The reasons for higher prices for energy crops are two-fold: (i) the field crops
that are displaced by energy crops generate higher returns raising the break even prices of
energy crops and (ii) the yield of energy crops is currently low in temperate climates. In
fact, some studies consider energy crops as future (third generation) biomass feedstock
after corn grains and agricultural/forest residues . A total of 98 million dt of MSW is
potentially available for cellulosic ethanol production, however only less than a third of
MSW will be available at price of $50/dt, due to higher MSW processing costs.

25

@

Figure 2.1: Biomass supply curves for the US

Price $/dry metric ton

100

85
Energy Crops

90
80

250
Agricultural
Residues

135

98
MSW

Forest and Mill
Residues

70
60
50
40
30
20
10
0
-

50

100
150
200
Million dry metric tons

250

300

@

For interpretation of the references to color in this and all other figures, the reader is referred
to the electronic version of this thesis (or dissertation).

Table 2.1: Biomass Supply Estimates for the US
Price at biorefinery
a
gate ( $/dt)

30
40
50
60
70
80
90
100
a

Quantity supplied million dry metric tons
MSW

Agriculturalresidue

Forest and
Mill residues

15
32
36
58
63
75
86
98

110
204
234
245
248
249
250
250

12
80
114
130
133
135
135
135

Energy
crops

35
52
73
85

Total*
137
315
384
434
480
512
544
568

in 2008 US dollars

* Total quantities are different from the summed up values due to rounding
Canadian Biomass Supply
The potential total biomass supply in Canada is 123 million dt, with agricultural
residues (42 million dt) and forest/mill residues (43 million dt) constituting 60 per cent of
26

supply. As shown in table 2.2, the biomass in Canada is likely to be more expensive than
in the US. At a price of $50/dt, only one-fourth of all potential biomass would become
available in Canada, compared with nearly two-thirds in the US. The reasons are three
fold: (i) the lower agricultural cropping density in Canada (due to cooler climate) leads to
higher transportation costs; (ii) the yield of biomass is lower due to temperate climate and
(iii) the lower population resulting in lower MSW generation. Unlike in the US,
agricultural residues are more expensive than forestry feedstocks in Canada; the main
reason being increased transportation costs due to lower cropping density. MSW and
energy crops are costlier in the Canadian case as well. Energy crop supplies are 31
million dt at $100/dt. However, if switching to energy crops occurs in 20% of all
agricultural land with lower ROVC than switchgrass, then energy crop supply can nearly
double to 61 million dt in Canada.
Figure 2.2: Biomass supply curves for Canada
100
7
MSW

Price $/dry metric ton

90
80

Energy
Crops

30

42
Agricultural
Residues

70

43
Forest and
Mill
Residues

60
50
40
30
20
10
0
-

10

20

30
40
Million dry metric tons
27

50

Table 2.2: Biomass Supply Estimates for Canada
Biomass
Price $/dt*

30
40
50
60
70
80
90
100

Quantity Available (million dry metric tons)
MSW

1
2
3
4
5
6
7
7

Agricultural
residue

Forest and
Mill
residues

6
7
7
31
37
42
42
42

12
12
30
43
43
43
43
43

Energy
crops

26
30
31

Total**

20
22
40
79
85
117
121
123

* At biorefinery gate in 2008 US dollars
** Total quantities are different from the summed up values due to rounding
The renewable fuel standard provisions under the US Energy Independence and
Security Act, 2007 mandate using 21 billion gallons of advanced biofuels for
transportation by the year 2022, including 16 billion gallons of cellulosic ethanol . At a
conversion rate of 70-100 gallons/dt the biomass requirement to supply 21 billion gallons
is 210-300 million dt. These results suggest that this quantity of biomass (required by
2022) is readily available under current conditions at prices below $50/dt. Table 2.3
presents the amount of biomass required to produce 7, 14 and 21 billion gallons of
cellulosic ethanol and estimated composition of biomass supplied to meet these
requirements. Almost all (97%) of biomass required for meeting the advanced biofuel
provisions of EISA 2007 is likely to be various types of residues, with agricultural
residues accounting for 61%, followed by forest and mill residues (27%). Dedicated
energy crops are likely to be a minor source.

28

Table 2.3: Biomass Feedstock Requirements and Expected Composition to meet
Cellulosic Ethanol Production Targets in the US
Ethanol production target
billion gal (at 70 gal/dt)
Required biomass (million dt)

7

14

21

100

200

300

Agricultural Residues

67.3

147.0

180.0

Energy crops
Forest and Mill Residues
MSW

12.0
20.7

22.0
31.0

9.0
80.0
31.0

Geographical Distribution
Table 2.4 shows the geographical distribution of feedstocks required to meet
EISA 2007 mandates. The corn-belt states that supply agricultural residues account for
the largest quantities of biomass for biofuel production, followed by states like Texas,
Georgia and Oregon supplying significant quantities of forest and mill residues. It should
however be noted that the states near the bottom of Table 2.4, are unlikely to have
cellulosic ethanol plants due to low quantities of biomass available which would not be
adequate to exploit the significant economies of scale observed in biorefineries. Mill
residues account for almost all of the 80 million dt of forest and mill residues supplied.
Collecting logging residues was found to be uneconomical at prices below $50/dt. Energy
crops are competitive with field crops in south-eastern states, Carolinas, Tennessee,
Texas and Oklahoma at prices around $50/dt.
Table 2.5 presents the geographical distribution of biomass supply at a price of
$100/dt in Canada. Agricultural provinces of Saskatchewan, Alberta and Ontario account
for most of agricultural residues and energy crops, while forest and mill residues account
for most supplies in British Columbia and Quebec. Compared to US, a larger proportion

29

Table 2.4: Geographical Composition of Feedstock Mix to meet EISA Mandates
(Biomass Quantity Target: 300 million dt)
State

Agricultural Energy
Residues
Crops

Illinois
Iowa
Minnesota
Nebraska
Indiana
Michigan
Washington
Ohio
Texas
Georgia
Oregon
Wisconsin
Florida
Idaho
Alabama
South Dakota
California
Missouri
Pennsylvania
North Carolina
Mississippi
Virginia
Louisiana
Kentucky
Arkansas
Kansas
Tennessee
South Carolina
Oklahoma
Montana
Maryland
New Jersey
West Virginia
Colorado
New Hampshire
New York
Massachusetts
North Dakota
Delaware

34.79
38.27
18.95
20.09
16.06
4.34
2.31
8.70
3.21
0.09
0.06
5.47
2.18
0.01
5.92
0.92
4.63
0.08
0.18
0.39
0.24
0.80
2.40
0.31
3.25
0.83
0.86
0.46
0.65
1.37
0.07
1.01
0.46

3.36
0.66
0.09
0.66
0.85
0.07
0.17
0.09
0.58
0.58
0.51
1.09
0.46
30

Forest
and Mill
Residues
0.33
0.16
1.04
0.07
0.65
1.40
5.68
0.91
2.23
7.33
6.54
1.69
2.03
4.42
5.91
0.15
5.02
1.11
1.49
4.02
4.58
2.21
3.61
1.49
3.66
0.05
1.63
2.51
0.66
1.95
0.17
0.08
0.82
0.22
0.94
1.18
0.17
0.01
0.02

MSW

Total

5.91
0.84
4.51
1.94
0.87
4.57
3.52
2.06
1.36
1.98
0.58
0.39
0.98
0.35

41.02
38.43
20.83
20.16
16.71
10.25
9.93
9.61
8.80
8.08
7.47
7.16
6.69
6.60
6.59
6.07
5.94
5.73
5.08
5.04
5.04
4.68
4.50
4.46
3.96
3.30
3.04
3.02
2.60
2.41
2.18
2.06
1.86
1.60
1.33
1.25
1.14
1.01
0.83

Table 2.4 (continued)
State
Agricultural Energy
Residues
Crops
Maine
Other States
US Total*

0.13
180.00

9.00

Forest
and Mill
Residues
0.44
0.96
80.00

MSW

Total

0.36
0.78
31.00

0.79
1.88
300.00

* Sum of state biomass supply may not equal US totals due to rounding
Table 2.5: Canada Geographic Distribution

Saskatchewan
Alberta
British Columbia
Ontario
Manitoba
Quebec
Other Provinces

MSW
0.27
0.86
0.67
2.46
0.35
2.06
0.32

Agricultural
residues
16.68
13.27
5.54
6.17
0.16

Forest
residues
0.54
2.73
8.70
2.91
0.24
4.55
2.32

Mill
residues
0.52
2.63
8.40
2.81
0.23
4.40
2.24

Energy
crops
16.82
7.17
0.14
1.53
4.52
0.14
0.01

Canada Total*

*

7.00
41.81
21.99
21.23
30.33
Sum of state biomass supply do not equal US totals due to rounding error

of crop land is expected to convert to switchgrass production on account of relatively
lower profits from traditional row crop production Canada.
Comparison with Recent Estimates
Walsh (2008) estimates that a total of 283 million dry short tons (256 million dt)
of biomass will be available at a price of $100/dry short ton ($110/dt) in 2010, consisting
of 101 million dt of agricultural residues, 18 million dt of urban wastes, 54 million dt of
forest residues, and 52 million dt of mill residues and 31 million dt of switchgrass. These
estimates are substantially lower than our estimates. The reasons for the differences are
discussed below. Since comparable studies of Canadian biomass supply curves were not
available, this comparative analysis is limited to the US.

31

Walsh estimates that 101 million dt of agricultural residues will be available by
2010 at a price of $110/dt compared to our estimate of 250 million dt. The difference
arises mainly from assumed percentage of available residues that are collected for ethanol
conversion. Our analysis assumes the recommended level of residues to reduce soil
erosion below tolerable levels, will be left on the field, which results in collection rate of
76% of available residues. In comparison, estimates by Walsh (2008) implicitly assume a
collection of only 33-45% of available residues. Walsh estimates the amounts of residues
that should be left on the field to maintain soil carbon and organic matter levels in
addition to soil erosion control. Our estimates only include mandatory soil erosion
control requirements, but incorporate opportunity costs of soil fertility maintenance.
What is the sustainable residue removal rate is a subject of debate . Our estimates are
consistent with the ‘Billion ton’ study by Perlack et al (2005) which assumes removal of
70-80% of residues from the fields. If the collection is limited to about 40% of available
residues, our estimates are similar to those by Walsh (2008). Moreover, Walsh’s
estimates consider only corn-stover and wheat straw while our estimates include residues
from other crops namely rice, barley and oats.
For estimating urban wastes, Walsh only considers wood portions of urban
wastes; but, in this study all organic components of MSW including paper, wood, yard
trimmings and food waste are included as potential feedstocks for ethanol conversion.
Paper and food wastes accounted for 32.7% and 12.5% respectively of total MSW
generated in the US compared wood and yard trimmings which accounted for 5.6% and
12.8% respectively (USEPA 2008). Technologies for converting all organic fractions of
MSW into biofuels are reportedly available; e.g. gravity pressure vessel process from

32

Genesyst Inc . Walsh further adjusts the wood waste quantities downward by 53% to
account for contamination by paints, chemical treatment, adhesives etc.
Our estimates of forest and mill residues consist of 55 million dt of logging
residues and 80 million dt of mill residues. The logging residue estimates are very similar
to those by Walsh. Walsh’s assumption that 10% of mill residues are not usable because
of too small particle size helps explain some of the difference in mill residue estimates.
Additional differences may arise because Walsh’s residue estimates are based on
projected harvest rates for 2010, while our estimates are based on actual harvests in 2002.
Enough details are not available to reconcile all the differences.
With regard to energy crops, Walsh projected 31 million dt by 2010 at a price of
$110/dt. Our estimate is 85 million dt of energy crops at $100/dt. Her estimates are based
on dynamic projections from the POLYSYS model. While POLYSYS model compares
energy crop returns with field crop returns (similar to our approach), it also imposes
various ‘flexibility constraints.’ For example, changes in acreages under different crops
are limited to a maximum of 20% from previous year, and conversion of pasture lands to
energy crops is accompanied by corresponding increase in hay crop output to meet
animal feed requirements. In comparison, we employ a much simpler and cruder
assumption that 10% of all cropland with current economic returns lower than
switchgrass returns will shift to switchgrass. Further, in POLYSYS crop prices are
determined endogenously within the model and hence are likely to be higher in the
future, which makes energy crops relatively unattractive. Our ROVC estimates are based
on historical costs. As a result, our energy crop supply estimates are higher and represent
more optimistic estimates. The above comparison also demonstrates the sensitivity of the

33

estimates to underlying assumptions and modeling, which can lead to large variations in
biomass supply estimates across studies as pointed by ).
2.4

Conclusions
The analysis suggests that more than 500 million dt of biomass is available at

price of $100/dt in the US, while Canadian supplies are limited to 123 million dt.
Assuming ethanol yields of 70-95 gallons of ethanol/dt, biomass quantity is sufficient to
displace 27-37 billion gallons of gasoline in the US and 5.8-7.8 billion gallons of gasoline
in Canada. Biomass quantities necessary to meet advanced biofuel provisions of EISA
2007 will be available at prices around $50/dt. Agricultural residues such as corn stover
are the cheapest and most abundantly available sources of biomass followed by mill
residues. Forest residues and energy crops are likely to play minor role in meeting EISA
2007 mandates. At current productivity levels, energy crops are not competitive with
conventional crops in the prime agricultural areas of the US. However in some southern
states such as Texas, Oklahoma and Tennessee energy crops may be able to compete with
conventional crops and hence dedicated energy crop plantations are likely to appear first
in these states. For example, a 1000 acre switchgrass plantation is being developed by
Oklahoma Bioenergy Center, and University of Tennessee in contracting with farmers to
grow switchgrass for the proposed Dupont-Danisco cellulosic ethanol plant . Energy
crops may be more attractive in Canada because of relatively lower returns from
traditional crops. Saskatchewan and Alberta will be major sources of both agricultural
residues and energy crops, while British Columbia and Quebec will be major sources of
forest and mill residues.

34

The estimates of agricultural residue supply are very sensitive to the assumed
fraction of residues that can be sustainably removed from the field. Similarly the potential
of MSW as a feedstock depends on which components (e.g. food, paper, wood) can be
economically converted into liquid biofuels. Yields of energy crops need to improve
significantly from current levels to make them competitive with conventional crops.
Hence future research is needed in these areas. Finally the static supply function approach
taken in this study, while is relatively simple and transparent, inadequately accounts for
several factors that influence future biomass supplies such as productivity gains,
harvesting and conversion technology improvements, inter-temporal variations in yields,
costs and returns, policy interventions, and international trade effects. These limitations
have to be kept in mind while interpreting and drawing policy conclusion from the
reported results.

35

2.5 REFERENCES

36

References
Ahearn, Mary C., Robert Collender, Mitchell Morehart, and Michael Roberts. "How Do
Decoupled Payments Affect Resource Allocations within the Farm Sector?"
Amber Waves 2, no. 5 (2004): 1.
Andrews, Susan S. "Crop Residue Removal for Biomass Energy Production: Effects on
Soils and Recommendations." Soil Quality National Technology Development
Team USDA-Natural Resource Conservation Service,
http://soils.usda.gov/sqi/management/files/agforum_residue_white_paper.pdf.
Berndes, Göran , Monique Hoogwijk, and Richard van den Broek. "The Contribution of
Biomass in the Future Global Energy Supply: A Review of 17 Studies." Biomass
and Bioenergy 25, no. 1 (2003): 1-28.
Bradley, Doug. "Canada Biomass-Bioenergy Report." Climate Change Solutions,
http://www.climatechangesolutions.net/pdf/canada_country2006.pdf.
BRDB. "Sustainability - National Biofuels Action Plan."
http://www1.eere.energy.gov/biomass/pdfs/nbap.pdf.
BWPRR. "New York City Msw Composting Report - Summary of Research Project and
Conceptual Pilot Facility Design (Chapter 7)."
http://www.nyc.gov/html/nycwasteless/html/recycling/waste_reports.shtml.
Chartwell Solid Waste Group. "Solid Waste Group." Solid Waste Digest 16, no. 10
(2007).
City of Guelph. "Residential Tipping Fees."
http://guelph.ca/living.cfm?itemid=46456&smocid=1471.
City of Windsor. "Tipping Fees."
http://www.citywindsor.ca/DisplayAttach.asp?AttachID=9837.
De La Torre Ugarte, D.G. . "The Economic Competitiveness of, and Impacts on the
Agricultural Sector, of Bioenergy Crop Production. ."
http://beag.ag.utk.edu/summary/competativeness.pdf.
De La Torre Ugarte, D.G., M.E. Walsh, H. Shapouri, and S.P. Slinsky. "The Economic
Impacts of Bioenergy Crop Production on Us Agriculture." Report prepared for
the US Department of Energy and the US Department of Agriculture.
De La Torre Ugarte, Daniel G, Marie E Walsh, Hosein Shapouri, and Stephen P Slinsky.
"The Economic Impacts of Bioenergy Crop Production on U.S. Agriculture." Oak
Ridge National Laboratory,
http://www.usda.gov/oce/reports/energy/AER816Bi.pdf.

37

EISA. "Energy Independence and Security Act of 2007."
http://frwebgate.access.gpo.gov/cgibin/getdoc.cgi?dbname=110_cong_bills&docid=f:h6enr.txt.pdf.
English, Burton C., Daniel G. De La Torre Ugarte, Kim Jensen, Chad Hellwinckel,
Jamey Menard, Brad Wilson, Roland Roberts, and Marie Walsh. "25%
Renewable Energy for the United States by 2025: Agricultural and Economic
Impacts." http://www.agpolicy.org/ppap/REPORT%2025x25.pdf.
EPA. "Basic Information - Municipal Solid Waste (Msw)."
http://www.epa.gov/msw/facts.htm.
Gallagher, Paul , Mark Dikeman, John Fritz, Eric Wailes, Wayne Gauther, and Hosein
Shapouri. "Biomass from Crop Residues: Cost and Supply Estimates." U.S.
Department of Agriculture, Office of the Chief Economist, Office of Energy
Policy and New Uses, http://www.usda.gov/oce/reports/energy/AER819.pdf.
Genesyst. "Company Website." http://www.genesyst.com.
Gronowska, Magdalena, Satish Joshi, and Heather McLean. "A Review of U.S. And
Canadian Biomass Supply Studies." BioResources 4, no. 1 (2009).
Haq, Zia, and James L. Easterly. "Agricultural Residue Availability in the United States."
Twenty-Seventh Symposium on Biotechnology for Fuels and Chemicals (2006):
3-21.
Kalogo, Y, S Habibi, H MacLean, and S Joshi. "Environmental Implications of
Municipal Solid Waste Derived Ethanol." Environmental Science and Technology
41, no. 1 (2007): 35-41.
Mabee, Warren, Evan Fraser, Paul McFarlane, and John Saddler. "Canadian Biomass
Reserves for Biorefining." Applied Biochemistry and Biotechnology 129, no. 1
(2006): 22-40.
McLaughlin, S.B, De La Torre Ugarte, D.G., Garten, C.T., Lynd,L.R., Sanderson, M.A.,
Tolbert, V.R., Wolf, D.D. . "High-Value Renewable Energy from Prairie
Grasses." Environmental Science and Technology 36, no. 1 (2002): 2122-29.
Milbrandt, A. "A Geographic Perspective on the Current Biomass Resource Availability
in the United States." http://www.nrel.gov/docs/fy06osti/39181.pdf.
NFDP. "Net Merchantable Volume of Roundwood Harvested by Category, Species
Group, and Province/Territory, 1970-2006."
http://nfdp.ccfm.org/compendium/products/tables_index_e.php.
OBC. "Oklahoma Set to Plant First-Ever 1,000 Acre Switchgrass Field."
http://okbioenergycenter.org/.
38

Perlack, RD, LL Wright, AF Turhollow, RL Graham, BJ Stokes, and DC Erbach.
"Biomass as Feedstock for a Bioenergy and Bioproducts Industry: The Technical
Feasibility of a Billion-Ton Annual Supply." Oak Ridge National Laboratory,
http://feedstockreview.ornl.gov/pdf/billion_ton_vision.pdf.
Perrin, Richard, Kenneth Vogel, Marty Schmer, and Rob Mitchell. "Farm-Scale
Production Cost of Switchgrass for Biomass." BioEnergy Research 1, no. 1
(2008): 91-97.
Petrolia, Daniel R. . "The Economics of Harvesting and Transporting Hardwood
Forest Residue for Conversion to Fuel Ethanol: A Case Study
for Minnesota." Dept. of Applied Economics, Univ. of Minnesota,
http://agecon.lib.umn.edu/cgi-bin/pdf_view.pl?paperid=23488&ftype=.pdf.
Reinhart. "Estimation of the Moisture Content in Typical Msw."
www.msw.cecs.ucf.edu/ExerciseEstimationoftheMoistureContentinTypicalMSW.ppt.
Simmons, Phil, Nora Goldstein, Scott M Kaufman, Nickolas J Themelis, and James Jr
Thompson. "The State of Garbage in America." BioCycle 47, no. 4 (2006): 26.
Smith, W. Brad, Patrick D Miles, John S Vissage, and Scott A. Pugh. "Forest Resources
of the United States, 2002 - General Technical Report Nc-241." St. Paul, MN:
U.S. Dept. of Agriculture, Forest Service, North Central Research Station,
http://ncrs.fs.fed.us/pubs/viewpub.asp?key=1987.
Statistics Canada. "Whole Farm Database Reference Manual - Catalogue No.
21f0005gie."
http://www.statcan.ca/bsolc/english/bsolc?catno=21F0005G&CHROPG=1.
Tiller, K. "Switchgrass Production Costs (Personal Communication)."
http://www.utbioenergy.org/TNBiofuelsInitiative/Switchgrass/.
USDA - ERS. "Data Sets - Commodity Costs and Returns: U.S. And Regional Cost and
Return Data." http://www.ers.usda.gov/Data/CostsAndReturns/testpick.htm.
USDA - NASS. "Quick Stats (Ag Statistics Database)."
http://www.nass.usda.gov/Data_and_Statistics/Quick_Stats/index.asp.
UTK. "Tennessee Biofuels Initiative." http://www.utbioenergy.org/.
Vadas, P., K. Barnett, and D. Undersander. "Economics and Energy of Ethanol
Production from Alfalfa, Corn, and Switchgrass in the Upper Midwest, USA."
BioEnergy Research 1, no. 1 (2008): 44-55.
Walsh, Marie E. "U.S. Cellulosic Biomass Feedstock Supplies and Distribution."
www.ageconsearch.umn.edu/bitstream/7625/2/U.S.%20Biomass%20Supplies.pdf

39

Walsh, MARIE E., DG De La Torre Ugarte, HOSEIN Shapouri, and S.P. Slinsky.
"Bioenergy Crop Production in the United States." Environmental and Resource
Economics 24, no. 1 (2003): 313-33.
Wang, Michael, May Wu, and Hong Huo. "Life-Cycle Energy and Greenhouse Gas
Emission Impacts of Different Corn Ethanol Plant Types." Environ Res Lett 2, no.
024001 (2007): 13.
Wilhelm, W. W., J. M. F. Johnson, J. L. Hatfield, W. B. Voorhees, and D. R. Linden.
"Crop and Soil Productivity Response to Corn Residue Removal a Literature
Review." Agronomy Journal 96, no. 1 (2004): 1-17.
Wischmeier, W. H, and D.D. Smith. "Predicting Rainfall Erosion Losses." Agriculture
Handbook 537, no. 1 (1978).
Wood, Susan M, and David B. Layzell. "A Canadian Biomass Inventory: Feedstocks for
a Bio-Based Economy."
http://www.biocap.ca/images/pdfs/BIOCAP_Biomass_Inventory.pdf.

40

REGIONAL COMPETITIVENESS OF ENERGY CROPS
3.1

Introduction
Cellulosic feedstocks such as agricultural residues, forestry residues, and dedicated energy
1

crops (DEC) are predicted to become preferred feedstocks for biofuel production. Crop and
forestry residues are annually produced and available for immediate use; as a result their potential
supply can be estimated. However, the potential for dedicated herbaceous energy grass crops such
as miscanthus, and switchgrass is less certain because they are yet to be grown in large scale.

2

With a growing emphasis on energy crops to meet the cellulosic biomass requirements, one of the
key questions is ‘whether farmers will divert their lands from existing annual cropping practices to
planting perennial energy crops?’ This paper evaluates the economic competitiveness of energy
crops with field crops and other land uses.
This study estimates the economic returns for farmers who divert their lands to energy crops
versus the returns for farmers who retain their lands with the current crop uses. Since perennial
energy crops supply biomass for 10 years, the economic returns from energy crops and other crops
are compared over 10 years to account for economic fluctuations during that time frame. These
fluctuations in economic returns arise primarily from changes in input and output prices which are
affected by crude oil prices. In this study, I adopt a Monte Carlo simulation framework to analyze
the impacts of crude oil prices on energy crop adoption. The simulation framework incorporates the

1

Currently, corn grains and soybean oil are used as primary inputs to produce biofuels; increased
use of food and feed products for fuel production has generated interest in advanced biofuels such
as cellulosic ethanol produced from cellulosic feedstock
2
Perennial energy crops normally supply cellulosic biomass for 10 years
41

main factors that affect economics such as regional differences in crop yields, and differences in
cultivation practices of field crops across various U.S. states.
Current literature compares energy crops with other agricultural crop enterprises largely
based on point estimates (average values) of returns. But average returns do not capture the
variability in energy and conventional crop profitability arising from crude oil price fluctuations
and aggregate planting decisions of farmers. Fluctuating crude oil prices cause considerable
uncertainty in the success of investments in cellulosic biofuels and the growth of energy crops.
About 20 to 50 per cent of crop production costs are directly or indirectly tied to crude oil price
level making it as a critical factor in the success of energy crops (Shoemaker, et al., 2006). Crude
oil prices have varied between $20.87 and $127.34 per barrel (2010 dollars) between
years….and… and preliminary evidence suggests that the cellulosic ethanol industry, and possibly
energy crops, become economically attractive only if crude oil price stays consistently above $70
per barrel (Creyts, et al., 2007; EIA, 2010; Tyner and Taheripour, 2008). Schnitkey et al (2007)
argue that the potential for biofuel feedstocks (including corn) should be evaluated by taking crude
oil price fluctuations into consideration. With evidence that agricultural output prices – particularly
corn, soybeans and cotton – have long run relationship with crude oil prices, it is important to study
the impact of crude oil prices on the competitiveness of energy crops (Harri, et al., 2009; Modi,
2009). Crude oil prices affect the prices paid for energy crop biomass making it as a crucial
determinant of the success of energy crops (Mark, et al., 2009).
I first develop state level crop production budgets for various field crops as well as energy
crops over a 10 year planning period. Crude oil prices (COP) are a critical driver of relative returns
because they affect the market prices of biofuels, energy crops and major field crops (corn and

42

soybeans). Crude oil prices also affect the energy cost component (fuel and fertilizer) of crop
3

production. Prior literature suggests co-integrating long run relationships between crude oil prices,
field crop prices and natural gas which is the main input in nitrogen fertilizer production (GAO,
2003; Harri, et al., 2009; Hartley, et al., 2007). Hence I first empirically estimate the historical
relationships between crude oil prices, crop prices and the energy component of crop production
costs (natural gas). I consider three crude oil price scenarios – reference case, high and low oil
prices – as projected by USDOE over the next 10 years. Crop output prices and crop production
costs are projected for each of these scenarios. Energy crop yields are modeled as a function of
state level average climatic conditions (precipitation and temperature). Yields for conventional
crops are projected using historical trends.
Second, I use Monte-Carlo simulation techniques to incorporate uncertainty into the static
budgets developed above. I estimate the volatility (coefficient of variation) of projected crude oil
prices as a weighted average of historical and implied volatility. The crude oil price volatility is
assumed to drive the variations in crop production costs and crop prices. I simulate variations in
energy crop yields assuming historical variances in precipitation and temperature continue into the
future.
In the third step I consider crop switching decisions by farmers under different assumptions
regarding their risk preferences. The results of the analyses are then used to predict the breakeven
biomass prices necessary to make switching from current crops to energy crops attractive, and the
probabilities that such switching will occur under the various crude oil price and farmer risk
preference scenarios.

3

Crude oil prices affect natural gas prices which in turn determines nitrogen fertilizer prices
43

Objectives and Scope:
The primary objective of this analysis is to compare the economic returns of farmers who
convert their fields to energy crops with the economic returns of farmers who retain their fields
under existing crops. Another major objective is to identify the US states that have the greatest
potential for energy crop production in terms of the average level and variability of returns. In
particular, the research in this paper seeks to answer the following questions:
•

How do the distribution of economic returns of energy crops compare with that of field
crops across different states?

•

What role do crude oil price fluctuations play in determining the economics of energy crops
relative to traditional crops?

•

How do state level differences in temperature (growing degree days, GDD) and
precipitation affect energy crop (miscanthus and switchgrass) yield and returns?

•

Which croplands (or fallow) are more likely to be converted to energy crops based on
economic returns?

•

How do risk preferences of farmers affect their energy crop production decisions?

These analyses are carried out at state level. The analytical steps in evaluating the relative
economics of energy crops and conventional crops are summarized in Figure 3.1.

44

Figure 3.1: Framework of proposed simulation study:
Energy crop yield –
simulated based on
temperature (GDD) and
precipitation

Energy crop prices –
calculated based on
crude oil and biofuel
prices

Energy crop
production costs –
simulated based on
crude oil and biofuel
prices

Field crop yield – modeled
using historical growth rates

Simulated
empirical
distribution of
energy crop’s
economic returns

Simulated
empirical
distribution of
field crop’s (or
alternative land
uses) economic
returns

Comparison of empirical
distributions to evaluate
competitiveness

45

Field crop prices – modeled
using cointegration/correlation
properties of agricultural
commodities with crude oil
prices

Field crop production costs –
differentiated into energy and
non-energy costs; energy costs
simulated based on simulated
crude oil price

3.2

Literature Review
A number of national level studies – that use a linear programming model called Polysys –

evaluate the potential allocation of land to energy crops. Polysys models switchgrass yields based
on hay crop yields at an agricultural statistical district level (ASD which is a collection of 9 or 10
geographically adjacent counties). These studies compare the economic returns of energy crops
(switchgrass in particular) and field crops (De La Torre Ugarte and Hellwinckel, 2004; De la Torre
Ugarte, et al., 2000; Ray, et al., 1998a, 1998; Walsh, et al., 2003). The Polysys studies predict that
switchgrass will be competitive with field crops in 7 to 15 per cent of U.S. crop lands, supplying
anywhere from 25 to 334 million tons of biomass per annum.The wide range of biomass estimates
is a result of diverse assumptions used in Polysys modeling – see figure 3.2 and table 3.1 for a
summary of major Polysys studies and their assumptions. By comparing Net Present Values (NPV)
of energy crops and field crops, Polysys predicts a price of $32 – 74 per metric dry ton for
switchgrass delivered at the biorefinery.
While the Polysys studies help generate estimates of crop switching and biomass supply at
the level of individual agricultural statistical districts, these estimates suffer from many limitations.
The analyses are based on deterministic (point) estimates of production costs and crop yields
failing to include the fluctuations in production costs and returns. Polysys studies also suffer from
data mismatch while computing NPV: the gross revenues from field crops are calculated using
ASD data (groups of counties) but the production costs are estimated using USDA’s Farm
Production Regions (groups of states, see (De la Torre Ugarte and Ray, 2000; De la Torre Ugarte,
et al., 2000).

46

Figure 3.2: Switchgrass production estimates – Summary of Polysys studies
Price in 2004
Dollars/ MT

Switchgrass Production Estimates - Various Studies

90
80
12
70
11
60
14

50

8

5

10

6

7
40

1

13

30

9

4

2

3

20
10
Million Metric Tons
0
0

50

100

150

200

47

250

300

350

400

Table 3.1: Switchgrass supply curve – summary of different US national level studies
Point

Quantity (Million
Metric Tons)

Price (2004$)

Line

0-139

0 – 80

2004 Farm Bill Provisions

1

133

40.00

Moderate crop yield with land use changes

2

334

40.00

High crop yield with land use changes

3

54.9

32.70

Wildlife Management Scenario, $1.83/GJ

4

170.9

43.60

Production Management Scenario, $2.44 /GJ

5

129

38.15

Production Management Scenario, $2.14/GJ

38.15

Production Management Scenario, $2.14/GJ,
Switchgrass yield increases by 25%

Study
De La Torre
Ugarte and
Hellwinckel
(2004)

Perlack, et
al (2005)

Walsh, et al
(2003)

6

202

Assumptions

7

67.6

38.15

Production Management Scenario, $2.14/GJ,
Switchgrass yield decreases by 25%

8

24.2

38.15

$2.14/GJ, Cultivation costs increase by 25%

9

205.3

38.15

$2.14/GJ, Cultivation costs decrease by 25%

48

Table 3.1 (Continued):

Point

Quantity (Million
Metric Tons)

Price (2004$)

10

34

42.70

Yield = 11 Mg/ha

11

158

62.00

Yield = 9.4 Mg/ ha

12

191

73.80

Yield = 9 Mg/ ha

13

60

39.60

1999 USDA crop yields, costs and prices

14

172

49.50

1999 USDA crop yields, costs and prices

Study

McLaughlin
S.B (2002)

De la Torre
Ugarte, et al
(2000)

49

Assumptions

Polysys studies also use arbitrary land allocation constraints (called flexibility constraints) to
4

avoid corner solutions resulting from the linear programming algorithm. Polysys considers only
switchgrass as the major energy crop while there are alternative energy crops such as miscanthus
which yield higher than switchgrass and may be more competitive (Khanna, et al., 2008).
Few other studies have been conducted at the state level. These state level studies also
compare the static net present values of energy crops with that of field crops – (Bangsund, et al.,
2008; Cassida, et al., 2005; Guffey, 2006; Heaton, et al., 2008; Khanna, et al., 2008; Maryland
Energy Administration, 2003; Perrin, et al., 2008; Vadas, et al., 2008). Except Griffith, et al.,
(2009), the existing studies assume a certain average level of energy crop yield and production
costs in a state or region. Most of those studies implicitly treat energy prices to remain constant
over the entire productive life period (10 years) of energy crops. These studies predict that
energy crops can compete with field crops or other land uses if switchgrass prices (delivered at
the biorefinery) are at least $59.52-120 per dry ton in Tennessee, $60-90 per dry ton in Iowa or
$54 per ton in Alabama and $35-74 in North Dakota (Bangsund, et al., 2008; Guffey, 2006;
Jensen, et al., 2005). These price estimates have a wide range due to different assumptions used
in these state level studies.
These state level studies do capture regional variations by modeling the state level crop
production practices and state level energy crop yields; but they cannot be generalized or
aggregated to derive national level estimates due to variations in their underlying assumptions.

4

Polysys results are derived using ‘flexibility constraints’ that involve arbitrary assumptions on
land allocation to energy crops (e.g. crop switching is limited at 15-20% of crop land – see
section 2.5 in De la torre Ugarte, D. G., et al (2000) and section 3.1 in De la Torre Ugarte, D.,
and D. E. Ray (2000).
50

Most of the studies consider only switchgrass as the major energy crop while a few other studies
evaluate miscanthus as a potential energy crop in the US. Farmers in some states (Tennessee and
Oklahoma) have already adopted switchgrass as their prime energy crop due to lower
establishment costs and easier seed based propagation. But other energy crops such as
miscanthus may be more suitable due to higher yields even though their establishment costs are
5

higher due to difficulties in propagation with rhizomes. The differences in switchgrass and
miscanthus yields and production costs are depicted in table 3.2 (Khanna, et al., 2008).
Table 3.2: Comparison of switchgrass and miscanthus crop yields and production costs in
Illinois
Switchgrass

Miscanthus

Cumulative yield over 10
years (dry tons per acre)

48.75

189.38 *

Establishment costs during
years 1 and 2 ($/acre)

553.27

1099.19

Break-even price including
#
transportation costs ($/ton)

64.84

49.58

* Adapted for 10 year time frame; # excludes land rental costs; Source: Khanna, et al (2008)
The above discussion shows that there is a need to use a uniform framework while
evaluating energy crops that accounts for fluctuations in crop revenues, yields and returns in a
similar manner across all states. This study simulates budgets that is consistent across the states
while incorporating the fluctuations in energy crop returns.
5

Agronomists and conservation experts suggest to grow switchgrass since they are native
species to the US; miscanthus are native to European continent and has been successfully grown
in many European countries.
51

3.3

METHODS

Crop production budgets and Net Present Value (NPV)
I use field crop production budgets developed by state extension agencies as a primary
data source for my calculations. In particular, these budgets are used to derive the agricultural
input quantities required for major field crops in every state (table 3.3). The agricultural inputs
are divided into two groups – energy (s) and non-energy (x) inputs. The distinction between
energy and non-energy inputs is made to extract the impact of energy (crude oil) price
fluctuations on crop budgets. Energy inputs include diesel and fertilizers (N, P and K). Note that,
fertilizers contain large quantities of embodied energy and their prices are closely linked to fossil
fuel price movements. Non energy input costs include expenses on seeds, herbicides, pesticides,
and fixed costs associated with buildings and machinery. Similar budgets for energy crops were
developed by estimating the quantities of energy and non-energy inputs according to the
maximum yield potential of energy crops at the state level.
A single planting of energy crop has an average productive period of 10 years. To match
with this time frame, the returns from annual field crops are also estimated over 10 year time
frame. This assumes that the particular field cropping pattern would be maintained over the next
10 years. To account for crop rotations, a weighted average of returns from two or more monocropping sequences is computed. The weights are determined by the relative frequency of the
particular field crop in the crop rotation mix over a 10 year interval. The proportional changes in
input quantities over 2000-2009 are used to compute the energy and non-energy inputs over the
study period 2010-2019.

52

Table 3.3: Use of state level crop production budgets in computing NPV
X

This study (for years 1 through10)

Field crops or existing land uses
NASS data on state level yields are projected
Yield
according to their historic trend yields
Long run cointegrating relationships
Crop Prices
between crop and crude oil prices are used;
in NPV calculations, crude oil prices are
derived based on DOE-NEMS data
USDOE – NEMS projections for crude oil
Costs
prices are used
Long run cointegrating relationships
Crop
between crude oil prices and agricultural
inputs/Energy (sn)
inputs (diesel, N, P and K) are used to
predict energy input cost component
Assumed to be constant within a state but
Non-energy
allowed to vary across the states except for
(xm)
land rental values
Economic returns
in year t (Ft)
NPV
Energy crops
Yield

Crop Prices

Price * Yield –
(Energy costs + Non energy costs)

Methods adopted in
literature

Yield trends are incorporated
in Polysys models
Derived as projections based
on econometric or
optimization models
Energy prices are assumed to
remain constant
Impacts of energy prices are
generally ignored

No state level comparisons
exist, generally non energy
inputs are assumed to be
constant
No distinction made between
energy and non-energy costs

∑t Ft(x1, x2, x3,…xm; s1, s2, s3…. sm)

Derived based on a statistical model with
state level temperature and precipitation as
regressor
Derived as the residual value (defined as the
revenues from biofuel in excess of
6
processing and transporting costs)

Based on state level plot
studies or hay crop yields
Break even prices are
estimated based on costs of
production

Costs
Energy (sn)
Non-energy (xm)
6

Energy inputs are assumed to vary with yield
potential; energy input prices are derived
similar to above
Assumed to be constant across states except
for land rental rates

The residual values indicate the dollar value left to pay for cellulosic biomass after paying the
processing and transport costs.
53

Table 3.3 (Continued)
Price*Yield – (Energy costs + Non energy
Economic returns
costs)
NPV
∑t Gt(x1, x2, x3,…xm; s1, s2, s3…. sm)

No distinction made between
energy and non-energy costs

Comparison of crop budgets over a 10 years requires a metric such as Net Present Value
(NPV) to account for the time preferences of farmers. The traditional NPV without any
uncertainty is computed as
t

NPVi = ∑t δ [Priceit $ per ton * Yieldit tons per acre – Production Costsit $ per acre]
Or

NPVi = ∑t Ft(x1, x2, x3,…xm; s1, s2, s3…. sm; δ)

(A1)
(A2)

where subscript i indicates energy crops (miscanthus, switchgrass), field crops (corn, soybeans,
wheat, cotton) and other land uses (pasture, CRP); t stands for years t = 1, 2, … 10 (2010-2019),
δ is the discount factor = 1/(1+θ). Table 3.3 describes how the input cost components x and s are
7

modeled in this analysis. Many studies use these cost components to compute the break-even
prices for energy crops (table 3.4 – A).
The Net Present Value is a common metric used to compare investment alternatives
(Magni, 2002; table 3.4 - B). But NPV metric has limitations because there are additional sources
of economic value that may arise from the flexibility to delay planting of energy crops. The
ability to delay energy crop planting introduces additional value, termed as real option value.
7

Discount rate θ is assumed to be 2.01%; the discount rate is computed based on the yield of 10
year US treasury bonds (average of recent five year interest rate of inflation indexed bonds – (US
Treasury, 2009). The discount rate captures the time preference of economic returns received by
farmers over a 10 year time frame. The inflation indexed bond rates can be used to derive an
estimate for θ since I compute NPV in terms of constant dollars (2009 dollars). I use 10 year US
treasury bonds because energy crops normally have a productive life of 10 years. Note that the
subscript ‘i’ can also represent a crop rotation (for example, combination of corn and soybean
crops such as CS or CCS) repeated over a ten year period
54

Such additional economic values will alter farmers’ land allocation decisions (table 3.4 – C).
However, the current approach of using NPV with uncertainty remains useful for a number of
reasons. The option value of delaying planting energy crops will be similar for farmers in
different US states for the following reasons.
Table 3.4: Possible metrics to evaluate economic competitiveness of energy crops:
Criterion
A

B
C
D

E

F

Evaluation Metric
Remarks
Energy crops alone
Break even prices for energy crop
Calculated over 10 year time period
production costs
of energy crop’s productive life
(e.g. Brummer et al)
Comparison with other land uses
Net Present Values (NPV) of energy crops
No uncertainties accounted for (e.g.
compared with that of field crops
Vadas et al, Polysys studies)
Expanded NPV criterion with option to
e.g. (Song, et al., 2009)
delay (real options framework)
Empirical distributions of energy crop NPV
Uncertainty in input and output
and field crop NPV
factors (e.g. fluctuating crude oil
prices – this study)
Others
Account for available price hedging
Contracting terms between biofuel
mechanisms for energy crops in comparison plants and farmers
with field crops
Account for current and potential
Not included in the current analysis
government incentives and policy variables.

Once a biorefinery has been planned for a site, the biorefineries promote and support
immediate establishment of energy crop biomass which attenuates real option values in energy
crop plantations. The biorefineries are located based on the ability for local farmers to supply
feedstocks consistently (Robb, 2007). The efforts of the biorefinery in writing long term
contracts, providing on-farm cultivation support and ensuring proper supply chains for energy
crop production delivery reduce the real option values associated with delaying the planting of
energy crops. Other reasons are explained in appendix A. The analysis does not include hedging
55

because hedging is a short run risk management option and less relevant for long term crop
switching decisions. The role of government policies are excluded from the model because the
goal of the analysis is first to figure out how the market emerges in the absence of such support
policies. However the impacts of such support systems are discussed in the conclusions.
NPV with Uncertainty
With no uncertainties, the cost components (both x1, x2, x3…xm and s1, s2, s3…sn) are
well known and NPV turns into a deterministic metric. Such deterministic measures have been
the focus of the existing studies in evaluating the suitability of energy cops but they are not very
informative given the random fluctuations encountered in crop production costs especially the
energy cost component. This problem can be overcome by modeling fluctuations in some or all
cost components and simulating the distribution of NPV. I introduce such fluctuations in this
analysis by modeling energy cost (s1, s2, s3…sn) fluctuations based on crude oil price
movements. Dimakos, et al., (2006) shows the conceptual background of joint estimation of
stochastic variables to derive NPV with uncertainty (table 3.4 – D). This method is particularly
helpful when uncertainty is a critical factor in the success of a proposed new enterprise such as
energy crops. For instance, Grove, et al., (2007) use such stochastic budgeting methods to
evaluate the profitability of converting crop-animal farm into a hunting range where the
economic returns from the proposed hunting range is not known with certainty. These stochastic
budgets are also useful in computing the consistency (in terms of probability) of generating
higher economic returns from switching to energy crops (Kleiber, 2009).

56

In my analysis, stochastic budgeting technique works as follows: a specific value for
crude oil price (dollars per barrel) is simulated for each iteration. The simulated crude oil price is
used to simulate estimates of energy costs, commodity prices, and NPV of energy crops and field
crops. The crude oil prices variations are generated for three different scenarios – reference case,
high and low crude oil price cases (EIA, 2009). The empirical distribution of crude oil prices are
derived based on log normal distributional assumption. The crude oil prices are simulated using a
log-normal distribution with mean values derived from EIA estimates. The standard deviation of
crude oil price distribution is derived from the volatility estimates using a weighted average of
historic and implied volatility. The weighted average approach helps utilize most of the available
information both from the past (historic volatility) and about the future (implied volatility). The
energy cost components of crop production (diesel, N, P, K fertilizer prices denoted by s1, s2,
s3…sn) are simulated based on their long run cointegrating relationship with crude oil prices
(where applicable) and their historical correlations. Other non-energy inputs (x1, x2, x3…xm) are
maintained constant with every iteration. This helps isolate the impacts of energy cost variations.
But the non-energy inputs are allowed to vary across states to reflect the differences in
production practices and land quality. Repeated simulations of the budgets for various draws of
crude oil prices generate an empirical distribution for NPVi. This enables the comparison of
empirical distributions of energy crops’ NPV with that of field crops.
I use probability measures to compare the distributions and to measure the consistency of
generating higher returns with energy crops. It compares the consistency of receiving higher
returns from energy crops with that of existing crop rotations. The consistency of economic
57

returns from energy crops is important because the study simulates crude oil prices along with
other factors. The consistency is measured by comparing the empirical distributions of NPV
values of both energy crops and field crops. If the probability of receiving higher economic
returns from energy crops is higher than the existing land uses, then energy crops can be declared
to be competitive in that state compared with the traditional crops. The probability measure thus
becomes a measure of the variance of NPV values.
Climatic Factors and Crop Yields
One of the prime determinants of returns is yield of energy crops and field crops .
Surveys show that energy crop yield is a key drivers of energy crop adoption (Anand, et al.,
2009; Jensen, et al., 2005). Switchgrass yields in Tennessee average 8 to 10 tons per acre per
year while Iowa switchgrass yields are only half of that, due to differences in climatic factors
(temperature (GDD) and precipitation). Since climate plays a major role in determining energy
crop yields, I use a statistical model that estimates energy crop yields as a function of state level
temperature and precipitation data. It is common to model crop yields based on climatic factors
(Lobell, et al., 2006). With regard to energy crop yields, Ivanic (2008) uses a statistical model of
regional temperature and precipitation to estimate the potential yields of switchgrass and
8

miscanthus in any location. The estimates of the regression model are presented in table 3.5; all
coefficient estimates are significant at 1 per cent level showing the high impact of temperature
8

Miscanthus is high yielding crop non-native to the US propagated through rhizomes at higher
production costs; switchgrass is a native prairie grass already grown in CRP lands and easily
propagated through seeds at lower production costs. Ivanic (2008) study uses yield data from 16
plots for switchgrass: Arkansas (1 plot), Illinois (5), Louisiana (1), Mississippi (1), North Dakota
(4), Texas (3), and Tennessee (1); for miscanthus, yield data from 24 plots are used: Illinois (6),
Mississippi (1), and international locations (Austria (1), Canada (2), Denmark (1), Germany (5),
Ireland (1), Netherlands (2), Poland (3 plots), Sweden (1), and Turkey(1)).
58

and precipitation on energy crop yields. Although simplistic, Ivanic’s model provides a
reasonable way to predict energy crop yields; the temperature and precipitation factors explain
2

70% of energy crop yield variations (see the last column on R values in Table 3.5). In
comparison, advanced biophysical models such as ALMANAC that use micro (field) level
information on temperature, rainfall, soil characteristics such as soil water balance, soil and plant
nutrient balance, and interception of solar radiation explain about 80% of energy crop yield
variations which is only a slight improvement over Ivanic’s model (Kiniry, et al., 2005; Kiniry,
9

et al., 1996; Kiniry, et al., 2008). These statistical models can yield representative estimates of
energy crop yields at state level and have already been used by other researchers to conduct state
or county level studies (Khanna, 2009). Table 3.5 results show that miscanthus yields will be
consistently higher than that of switchgrass in all states – which reiterates the results presented in
table 3.2.
Table 3.5: Maximum potential yield of energy based on temperature and precipitation
2

Energy Crop Yield (metric
dry tons per acre per year)

Regression of yield on temperature and
precipitation

R value

Switchgrass

2.2 + 0.02 * (GDD * Precipitation) / 1000

0.69

Miscanthus

6.6 + 0.04 * (GDD * Precipitation) / 1000

0.71

Source: Ivanic ( 2008); all coefficient estimates are significant at 1 per cent level

9

Agricultural Land Management Alternatives with Numerical Assessment Criteria
(ALMANAC) model use information on temperature, rainfall, soil characteristics such as soil
water balance, soil and plant nutrient balance, and interception of solar radiation to predict
energy crop yields at micro (field) level
59

Crude oil price and agricultural input prices
The economics of energy crops or field crops have to be studied over a 10 year time
frame.

10

Crude oil prices fluctuate within a 10 year period which also leads to fluctuations in

energy costs (e.g. diesel and fertilizer inputs) in agriculture. The amount of energy inputs vary
across crops, leading to differential impact of crude oil price fluctuations on crop production
costs. Higher diesel prices result in higher transportation costs in agricultural operations (Huang,
2009; Olczyk, et al., n.d.). There is growing evidence that natural gas prices exhibit cointegration
(‘stable long run (positive) relationship’) with crude oil prices (Hartley, et al., 2007; Shoemaker,
et al., 2006); since natural gas price accounts for 70-90 per cent of nitrogen fertilizer production
costs, crude oil price fluctuations indirectly affect nitrogen prices as well (GAO, 2003; Huang,
2009). To incorporate these long run cointegration relationships, I empirically estimate diesel
and nitrogen prices as a function of crude oil price (COP, 2009 dollars) and natural gas price
respectively as given below (standard errors are in parentheses).To forecast diesel prices based
on crude oil prices, the weekly data on diesel and crude oil prices from January 1997 to January
2009 was used. To forecast nitrogen prices based on natural gas (which is cointegrated with
crude oil prices), the annual data from 1997 to 2005 was used. Both regressions used ordinary
least squares method.
Real Diesel price (2009 $/gallon) = 1.305 + 0.021 * COP
(0.11) (0.0002)
10

2

(R = 0.91)

The energy crop yield pattern gradually increases in the first two years, stabilizes and reaches
a potential maximum during years 3 through 7, and gradually declines in years 8 through 10
(Angelinia, et al., 2009). Over this 10 year time frame, the revenue uncertainty is high for both
field crops and energy crops. Even though there are futures markets for field crops to hedge
against uncertainty, those markets extend only 2 or 3 years into the future; no such markets exist
for energy crops.
60

The cost of N fertilizer is calculated based on natural gas prices:
Real Nitrogen price (2009 $/lb) =

0.253 +
(0.044)

2

(R = 0.51)

0.019 * NG price
(0.007)

where NG price refers to natural gas price ($/million BTU). The relationship between natural gas
and crude oil prices, as reported by Hartley et al. (2007- pp: 24), is given below:
ln Pt

NG

= -0.1872 + 0.8460 ln Pt
NG

where t is the time subscript, P

WTI

– 3.0032 ln [HRt

NG

RFO

/HRt

WTI

is natural gas price ($/MMBTU), P

(West Texas Intermediate type), HR

NG

is heat rate of natural gas, HR

RFO

]
is crude oil price
is heat rate of residual

fuel oil (a derivative of crude oil that is a substitute for natural gas). I use the above equation to
derive natural gas price based on the simulated crude oil price. This price relationship holds in
the ‘long run’ (italics as given in the Hartley et al.) which suits my purposes of forecasting NG
prices over 2010-2019. The heat rates for 2010-2019 are forecast using an ARIMA(1,1,0) model
at annual intervals. In spite of high correlation coefficients reported in table 3.6, there is no
evidence for long run cointegrating relationship between Crude oil prices, Phosphorus (P) and
Potassium (K) fertilizer prices (Huang, 2009). I simulate P and K prices based on their historical
correlations with N fertilizer prices; i.e. there is no causal relationship between crude oil price
fluctuations and P or K fertilizer prices. I implicitly assume that the historical probability
distribution of NPK fertilizer prices will continue into the future. The mean and standard
deviation of P and K prices are derived from their time series price data. The temporal
correlations between the three major fertilizers are presented in table 3.6.

61

Table 3.6: Correlation between crude oil price and NPK fertilizers (1989-2008)
N price
N price
P price
K price
COP

P price

K price

COP

1.000
0.925

1.000

0.938

0.990

1.000

0.959

0.933

0.956

1.000

Crude oil and agricultural output prices
Crude oil prices determine the output prices paid for energy crop cellulosic biomass and
field crop grains. The maximum prices paid for dedicated energy crops depend directly on
ethanol prices, which in turn depend on gasoline prices, crude oil prices and biofuel processing
costs (EIA, 2008a; Kumarappan and Gustafson, 2009; Mark, et al., 2009). Gasoline prices is
derived based on simulated crude oil prices as following; the monthly US DOE – EIA data from
January 1986 to December 2008 was used to estimate the following regression estimates:
2

Gasoline price (2009 $/gallon) = 0.915 + 0.027 * COP - 0.000032 * (COP)
(0.040) (0.001)
(0.00001)

Ethanol price (2009 $/gallon) is computed as 0.70 * predicted gasoline price. Then, energy crop
biomass price (2009 $/ton) is computed as [Ethanol price ($/gallon) * Ethanol yield (gallons/ton) –
costs for pretreatment, processing, and transport ($/ton)]. The above equations show how the prices
paid for energy crop biomass depend on the simulated crude oil price. Ethanol yield from cellulosic
biomass is assumed to be 80 gallons per ton. The pretreatment and processing costs are assumed to
be $1/gallon. This is slightly above the target processing costs for cellulosic biofuels to be
competitive with crude oil (DOE-EERE, 2010). Khanna et al (2009) uses a similar assumption; note
62

that the results are sensitive to this assumption. I discuss the implications of alternative processing
costs in the results section. The transport costs were assumed to be $10/ton.
Harri et al., (2009) show that crude oil prices affect (Granger-cause) corn and soybean
prices in the US especially after 2006. Other studies also find evidence of co-movement of
energy prices and agricultural commodity prices in the US (Benjamin, et al., 2009; Campiche, et
al., 2007; Du, et al., 2009; Zhang and Reed, 2008).

11

Such cointegrated price movements and

correlations between crude oil (or energy) and agricultural commodity prices are likely to
continue in the future given the continued reliance on corn ethanol over the next 15 years (EISA,
12

2007).

The cointegrating long run regression relationships between corn, soybean and cotton

prices and crude oil prices are estimated using monthly data after April 2006. The super
consistent estimates from ordinary least squares (OLS) regression of commodity prices on crude
oil prices and currency exchange rates are used for my calculations. The prices for other crops
(wheat, sorghum, pasture, fallow or CRP), which are not found to be cointegrated with crude oil
prices are simulated as correlated time series. It is a valid approach for the following reasons: as
explained for P and K fertilizers, the historic correlations between these crops and crude oil
prices can be expected to continue in the future; it ensures consistency that crude oil price
fluctuations are incorporated as part of the other crops; and using correlated simulations is a
common practice to model temporal data.

11

Although these results are found in the US, some international studies find no relationship
between crude oil prices and commodity prices in other countries – (see Benjamin et al, 2009;
Zhang and Reed, 2009)
12
Corn ethanol production is currently at 11 billion gallons per year – EISA requires corn
ethanol to be capped at 15 billion gallons per year during 2015-2022
63

A major limitation of this approach is that the past correlations are expected to continue
in the future without any changes. It is difficult to predict crude oil prices and even more difficult
to predict the correlation of crude oil prices with agricultural, and energy commodities in the
future. One reasonable approach is to modify the historical correlations and use it for projecting
in the future. Since the cointegration between crude oil prices and agricultural commodities is a
recent phenomenon (starting in 2006), there is a lack of data to predict correlation values in the
future. So, the historical correlations are assumed to continue into the future; the results have to
be interpreted recognizing this limitation.
Simulating crude oil prices
For the purposes of this study, crude oil prices over the next 10 year time frame (20102019) are required at annual intervals. I use the crude oil prices projected by US-EIA’s National
Energy Modeling System (NEMS – see figure 3.3). NEMS data provides long term forecasts
13

based on international energy market supply and demand conditions (EIA, 2009, 2003).

NEMS model is widely used for national and international policy making, including that of
14

USDA analysis (USDA, 2009).

Its forecasts are reliable because they anticipate alternative

economic growth scenarios and project crude oil prices corresponding to three scenarios: low,
reference case and high price scenarios (see figure 3.3, and table 3.7). In the reference case, the
crude oil price ranged from $49/barrel in 2010 to 112/barrel in 2019; in low price scenario, the
crude oil prices are expected to decline slightly from $55 to $47/ barrel over 2010-2019; in high

13

NEMS forecast are classified under high/low economic growth, high/low world oil price to
reflect various levels of crude oil prices. I used the former classification.
14
USDA analysis drives the policy behind subsidy for energy crop biomass (BCAP, EIA 2007)
64

price scenario, crude oil prices are expected to increase from $88 to $178/barrel over 201015

2019.

The NEMS model provides only the average price levels; the price fluctuations or

volatility around these average price levels have to be modeled for conducting simulations
underlying this analysis.
Figure 3.3: NEMS forecast of crude oil prices 2010-2019 ($/barrel):
200.0
180.0
160.0
140.0
120.0
100.0
80.0
60.0
40.0
20.0
0.0
2009

2010

2011

2012

Reference case

2013

2014

2015

2016

High economic growth

2017

2018

2019

2020

Low economic growth

Table 3.7: Three scenarios of crude oil prices ($/barrel)
Scenario

Crude Oil Price Range
2010-2019 (2009 dollars) *

Crude Oil Price Volatility (in
terms of coefficient of variation)
#

$55 - $47
From 36% in 2010 to 26% in 2019
Low economic growth (I)
$49 - $113
From 36% in 2010 to 24% in 2019
Reference case (II)
$88 - $178
From 36% in 2010 to 23% in 2019
High economic growth
(III)
* these values are depicted in figure 3.6; Source: DOE National Energy Modeling System
#
these estimates are based on historical and implied volatility figures explained below.
15

Inflation index between 2007 and 2009 is 1.04; i.e. $1 in 2007 is equivalent to $1.04 in 2009
65

Estimating crude oil price volatility
For my purposes of projecting crude oil price volatility over 10 years, I use a weighted
average of two volatility measures (historical volatility as described above and implied volatility)
that contain information about long term volatility movements in crude oil prices. The weighting
procedure is explained in Appendix B. Such a combined estimate is considered to be a more
reliable predictor compared to single estimator (Armstrong, 2001; Poon and Granger, 2003).
Volatility of any price series is defined to be “the standard deviation of the continuously
compounded rates of return… over a specified period” which is “the same as the standard
deviation of the differences in the natural logarithms of… prices… over the period” (FASB,
2004). A commonly used volatility measure is derived using historical data. For example, if
crude oil prices changed from C1 to C2 at a continuous compound rate r over a period of time t;
rt

i.e. C2 = C1 e , then volatility, expressed in percentage terms, is defined as the standard
deviation of the series r. It is interpreted as coefficient of variation (volatility (%) = standard
deviation/mean).

16

16

Hence, the continuously compounded rate of return (r) is r = (1/t) * [ln (C2 / C1)]. The
volatility of price series C over the period t is given by the standard deviation of the series r, i.e.,
(1/√t)*stdvn[ln (C2 / C1)], where stdvn stands for standard deviation. To express it in annualized
terms (t = 1 year), I multiply the above expression with √t to get the expression stdvn[ln (C2 /

C1)] or stdvn[ln (C2) – ln (C1)]. Volatility in terms of standard deviations (dollars per barrel) is
obtained by multiplying the volatility estimate (%) for a particular year with the mean crude oil
price estimate given in figure 3.5.
66

Figure 3.4: Spot prices volatility of crude oil prices – January 1989-September 2009

Crude Oil Prices and Volatility using Monthly Data
from January 1989 - September 2009
70
60
Volatility

50
40
30
20
10
-

20

40
60
80
100
Crude Oil Price (2009 Dollars)

120

140

The distribution of crude oil prices is heteroskedastic since volatility (a measure of
variance) varies with the level of crude oil prices. The crude oil price fluctuations vary inversely
with crude oil prices; i.e. volatility is more when crude oil prices are low and vice-versa (figure
3.4). The time series data for historical volatility estimates is forecast for 2010-2019 using
GARCH models.

17

Following Agnolucci (2009) I model the historical price volatility as a

GARCH(1,1) process; this process is shown to perform better than other competing processes, in
terms of lowest mean absolute error (MAE), and mean square error (MSE, for more information
see (Agnolucci, 2009; Manfredo, et al., 2001). This regression is used to forecast historical
17

There are other ways to model volatility of commodity prices that include stochastic
volatility, Monte Carlo Markov Chain (MCMC) methods, implied volatility, crack spread
futures, regime switching stochastic volatility and pattern matching methods (Andersen, et al.,
2005; Fan, et al., 2008; Figlewski, 2004; Murat and Tokat, 2009; Sadorsky, 2006; Vo, 2009; Yu,
et al., 2008; Zhang, et al., 2008). Many of these methods focus on explaining past regime shifts,
or rapid increases in prices, or sudden changes in price volatility rather than focusing on the
future.
67

volatility estimates (HVf) for the years 2010-2019. Inclusion of crude oil price as an exogenous
regressor in the GARCH model also helps differentiate the projected historical volatility based
on the level of crude oil prices (see figure 3.5). Historical volatility explained above contains
only past information. Information about the future crude oil price variations can be derived
using an implied volatility measure.
Implied volatility (IV) is computed using options on crude oil futures contracts. IV refers
to the underlying volatility that results in the observed premium price for options. It is obtained
by inverting the Black equation used for pricing options on futures contracts. Options on crude
oil futures are actively usually traded for three years (2010-2012) into the near future . Hence, I
can estimate the implied volatility measures based on options premiums for all months in 20102012 (IVf). But, the implied volatility estimates are required for all the years between 2010 and
2019. To forecast the IV estimates for the years of 2013 through 2019, I project the IV data from
1993 through 2012 using a GARCH(1,1) model. Agnolucci (2009) adopted this method while
comparing alternative procedures to forecast crude oil price volatility. The time series data
obtained from CRB (2009) is used for forecasting implied volatility which is explained in
Appendix C.

68

Figure 3.5: Historical volatility forecast (percentage) – low, reference and high crude oil prices
scenarios

90
80
70
60
50
40
30
20
10
1990

1995

2000

2005

2010

2015

Scenario Crude Oil Price $148/barrel
$51/barrel
$92/barrel

The volatility estimate used to simulate crude oil prices are derived by weighting IVf and
HVf values. The weights are derived using the procedure suggested by Granger and Ramanathan
(1984): the weights are the OLS coefficient estimates of the regression of realized spot price
volatility on the historical time series data of historical volatility (HVp) and implied volatility
(IVp). The data for this regression pertains to monthly data during 1993-2009.

18

Agnolucci

found that including a constant in the above regression is appropriate for modeling crude oil
price volatility. The procedure is explained below. The weights are actually the least squares
18

The past data on implied volatility is available starting from 1993.
69

coefficient estimates ω0^, ω1^, ω2^ from the regression rt = ω0 + ω1 IVp + ω2 HVp + ε, where
2

rt is the spot price volatility estimate calculated as ln(St/St-1) using daily data for years 1993 –
2009, St is the spot market price of crude oil at time t, IVp is the past data on implied volatility
(based on options on crude oil futures contract) for years 1993 – 2009, HVp is the past data on
historical volatility (based on futures contract prices) for years 1993 – 2009, and ε is the White’s
noise ~ N(0,1).

19

The time series data on IVp and HVp are obtained from Commodity Research

Bureau, CRB (2009). The weighted volatility measure (expressed in percentage) can also be
interpreted as coefficient of variation, i.e. standard deviation divided by mean (figure 3.6).
Figure 3.6: Weighted volatility forecast – low, reference and high crude oil prices scenarios
40.0

percentage

38.0
36.0
34.0
32.0
30.0
28.0

Low crude oil price

26.0
24.0
22.0
20.0
Feb-08

19

High crude oil price

Jul-09

Nov-10 Apr-12 Aug-13 Dec-14 May-16 Sep-17 Feb-19 Jun-20

Among GARCH models, single equation ARMA-GARCH model is found to be better
predictor of historical volatility than other GARCH models (see Padorsky, 2006); Actual IV data
exist for years 2010-2012 – so the forecast for those years were not used in the analysis
70

For illustrative purposes, standard deviation of crude oil price for 2014 is derived as
following: standard deviation for 2014 = weighted volatility measure for 2014 * mean price of
crude oil for 2014. Figure 3.6 shows that the price volatility is (slightly) higher when crude oil
prices are low confirming the inverse relationship depicted in figure 3.4. Implied volatility
computations (based on Black equation) assume that the future crude oil prices possess lognormal distribution. I retain the same distributional assumption for simulating crude oil prices
using mean price levels (figure 3.3) and volatility (figure 3.6) described above.
Simulating energy crop yields
The yield potential of energy crops is simulated using historic state level data on
temperature and precipitation. Since the climatic factors do not vary considerably within a state
or region, historic information is useful data to project or simulate future temperature and
precipitation values. The range (minimum and maximum) of values for temperature and
precipitation during 1970-2008 was used for simulation purposes (Appendix D). With regard to
climatic variables, this is rather a large time span over a large geographic area such as state.
Hence, any value from this range of values (between the minimum and maximum values) will
appear in the future. Under this assumption, a uniform distribution is valid to simulate
temperature and precipitation. I simulate state level temperature and precipitation data at annual
intervals which satisfy the above criteria, i.e. assuming uniform distribution (Dempster, 1999;
Mehrotra, et al., 2006).

20

20

The historical correlations between temperature and precipitation are

Note that uniform distributions do not fit smaller geographic locations (a single farm or zip
code) or smaller time frame (within an hour) – but it is a reasonable approximation over large
geographic areas (state) at annual intervals that use non-seasonal data. The use of 30 year
historical data ensures that the simulation of temperature and rainfall over 10 years into the
71

maintained in simulated data. The yield of miscanthus and switchgrass is computed using the
simulated temperature and precipitation data and the parametric estimates in table 3.5. The actual
yield of energy crops would be low at the start and end of the 10 year productive time frame:
30% of potential yield in year 1, 70% in year 2, 100% in years 3-7, and 80% in years 8-10
(Angelinia, et al., 2009).
Simulating field crop yields
Historical field crop yield data for individual states are available from USDA.. The time
series data for field crop yields corresponding to years 1970-2009 are found to be trend
stationary time series processes. Hence, I use trend projections to forecast the field crop yields
during 2010-2019. The growth in field crop yields is modeled using Compound Annual Growth
Rates (CAGR) because they are found to be better than linear or exponential extrapolation
procedures when the increase in yields are deterministic (Reilly and Fuglie, 1998; Taher and
Shadmehri, 2008). Reilly and Fuglie (1998) note that CAGR computed using only one set of
initial and final time points can be misleading. To overcome this problem, I derive CAGR for
every 10 year time slots: i.e. a CAGR estimate for 1970-79, another estimate for 1971-1980, and
so on. The average of all CAGR estimates is used to forecast field crop yields over 2010-2019.
My approach implicitly assumes that there will be no major technological breakthrough altering
the yield of field crops over the next ten years.
Simulating energy crop production costs
Nitrogen fertilizer is a major input in energy crop production. The maximum yield for
energy crops is attainable when nitrogen fertilizer is applied at a rate of 10 pounds per ton of
future will remain valid. Non-uniform distributions are needed only in micro-field level
simulations, which is not the case here.
72

potential cellulosic biomass yield (Perrin, et al., 2008; Vadas, et al., 2008); but the fertilizer
responsiveness of energy crop yields to nitrogen application flattens out beyond 70 lbs/acre (Jain
and Erickson, 2008) – hence, the amount of nitrogen applied for energy crops is 70 lbs or less
depending on the potential yield in a state. The other fertilizers (P and K) are required only in the
initial years of energy crop establishment. The energy inputs required for energy crop production
are given in table 3.8. The input rates for diesel, phosphorus and potassium are assumed to
remain the same across all states. The nitrogen input rate alone is varied according to the
maximum biomass yield possible within the particular state.
Table 3.8: Energy input use for energy crop production
Energy Input
Diesel (gallons/ton)
Nitrogen (lbs/ton)
Phosphorus (lbs/ac)
Potassium (lbs/ac)
Source: Ivanic (2008)

Quantity
2.1
10 (capped at a maximum of 70 lbs/ac)
18
36

The non-energy inputs such as seeding, reseeding, machinery, chemicals (pesticides and
herbicides) and harvesting are assumed to be constant: the seed material costs were assumed to
be $22.40 per acre in the first year and $2.40 per acre for subsequent years for reseeding in all
states; the fixed costs associated with machinery and non-energy inputs were estimated to be
$164.44 per acre for all years 1 – 10 (Perrin, et al., 2008). The land rental rates for energy crops
will be the same as that estimated for field crops (USDA - NASS, 2007).

21

21

In simulation studies, it is common practice to assume simpler distributions that require fewer
parameters when there is limited information available (such as land values, in this case).
73

Simulating field crop production costs
The energy and non-energy inputs of field crop production were derived from crop
production budgets generated by the state extension agencies. For every state in the study region,
the energy input quantities were multiplied by their corresponding simulated prices to derive the
crop production cost. The quantities of energy inputs are assumed to remain the same over 20102019. This assumption eliminates any input substitution that could result due to changing crude
oil prices. The non-energy cost component (seeds, chemicals and pesticides) are fixed constant
22

for every state (see table 3.3).

The land rental value for field crops are derived from NASS

reports (USDA - NASS, 2007). At the national level, the rental value for pasture is found to be
one third of prime cropland rental value. The list of crop production budgets from different state
extension agencies is given in Appendix E.
3.4

Role of Risk Aversion in Energy Crops Adoption

Aggregate utility function
Risk averse farmers derive different levels of utility depending on the riskiness of energy
crops vs. field crops. To study the impact of risk aversion on a representative farmer as in this
study, I transform the NPV measure to reflect farmers’ risk aversion level. I use concave utility
function to model the behavior of a representative risk averse farmer. The study of a
representative farmer generalizes the results for the entire state. Hence, the concave utility
functions can also be interpreted as aggregate risk averse utility functions commonly used in
macroeconomic studies (Pindyck, 1986; Saltari and Ticchi, 2007).
22

Only the states in the eastern half of US were considered; the western states were not
considered due to moisture restrictions and irrigation requirements that could make energy crops
non-competitive.
74

Any concave von-Neumann Morgenstern utility function can be used to study farmers
risk aversion. I use power utility function characterized by constant relative risk aversion
(CRRA). The power utility function uses one coefficient (CRRA) which reflects farmers’ risk
aversion level. It is computed using average economic returns from major crops such as corn,
wheat, and soybeans at state level as following:

23

the net returns over variable costs per acre

($/acre) were computed for corn and soybeans for each state during 1980-2009. The historic
economic returns were used to derive the cumulative distributions of corn and soybeans returns.
The CRRA coefficient is estimated based on the intersection point(s) of these cumulative
distributions. McCarl, et al. (1987) suggested that, the farmers prefer one major crop (corn or
soybeans, in this case) to either side of the intersection point. Hence, at the point of intersection,
the farmers would be indifferent between the returns from either crop. They developed a
procedure that translates this point of intersection into a RAC metric. Hardaker, et al. (2004a)
implemented this procedure for spreadsheet applications. I used this procedure for every state to
compute a CRRA metric. Using this CRRA metric with the power utility function provides a
way to model the impacts of farmers’ risk aversion.
Hardaker, et al. (2004a) also demonstrated how to use such aggregate utility functions to
calculate certainty equivalent (CE) values. The CE value can be defined as the guaranteed payoff
that makes the average farmer indifferent between a guaranteed payoff from energy crops
(through contracts) and a higher but uncertain payoff from field crops. The CE values are derived
through a monotonic transformation of NPV. I used the aggregate power utility functions as

23

The average economic returns for corn, soybeans and wheat are estimated at state level. This
measure varies with the state thus enabling a way to compare multiple states.
75

described above to compute CE values. Hardaker et al. (2004a) showed that CE values can be
used to rank different enterprises based on the risk aversion levels of farmers (CRRA).

24

Note

that the CE values are derived based on average returns for a farmer in every state – hence, they
are to be taken as evaluating the preferences of a representative farmer within a state.
Comparison of NPV values for risk-neutral farmers
For risk neutral farmers, the von Neumann – Morgenstern utility function is linear (line U
in figure 3.7). This leads to a linear correspondence between farmer’s utility and NPVi. Let the
field crops have an expected NPV of Dj with utility L and energy crops have an expected NPV
of Dk with utility H. The change in utility from converting lands from field crops to energy crops
is represented by (H-L or A1 = Dk – Dj). The empirical distribution for NPV can be used to
derive an additional probability metric which measures the probability of getting higher NPV
from energy crops compared to that of field crops (Dk), in notation, Pr(Dk > Dj) = Pr(A1 ≥ 0).
The calculation of Pr(A1 ≥ 0) is relatively straightforward. The cumulative distribution function
(CDF) is derived for variable A1. The proportion of area corresponding to the positive (or nonnegative) values of A1 gives the probability measure Pr(A1 ≥ 0). It is analogous to computing

24

Only the risks associated with agricultural energy input price risks are modeled here; other
risks related to crop yields, financial or institutional risks are not included.
76

the probability that energy crop’s NPVj has first degree stochastic dominance (FDSD) over field
25

crop NPVk or other land uses (Hansen, et al., 1978).

Figure 3.7: Relationship between NPV and Utility for risk-neutral and risk-averse farmers
Utility (U, V)

U
V

H1
L1
H
L

CEj CEk

Dj

Dk

NPV

Comparison of CE values for risk-averse farmers
For risk averse farmers, the correspondence between NPV and utility is non-linear as
depicted by the concave function V in figure 3.7 (Chibnik, 1981; Collins and Gbur, 1991; Lai, et
al., 2003; Myers, 1989; Pei, et al., 2005; Ramaratnam, et al., 1986). When the risk-averse farmer
shifts land from field crops (expected NPV Dj and utility H1) to energy crops (expected value of

25

For two revenue variables NPVj and NPVk with cumulative distribution functions F(NPVj)
and G(NPVk), respectively, returns from crop j (NPVj) stochastically dominates the other in first
degree if G(NPVk) ≥ F(NPVj), with at least one strict inequality for some value of NPV.
77

Dk and utility L1), the utility changes from H1 to L1. At the NPV level of Dj, the certainty
equivalent value for a risk averse farmer is CEj; similarly, at the NPV level of Dk, the certainty
equivalent value for a risk averse farmer is CEk. Hence, the change in utility for risk averse
farmers (H1 – L1) can be expressed in terms of difference in CE values: B2 = CEk – CEj. The
advantage of using CE (and B2) metric is that it can be interpreted in monetary terms. The
probability that CEk of energy crops exceeds CEj of field crops, Pr(CEk > CEj) is derived from
the cumulative distribution for B2, i.e. Pr(B2 > 0). The simulation software (@risk) is used to
calculate the probability measures explained above.
Calculation of CE values in numerical terms is based on the concept of Stochastic
Efficiency with Respect to a Function (SERF; Hardaker, et al., 2004a; Richardson, 2004;
Richardson, et al., 2004). I use power utility function to model the risk-averse behavior of
farmers. Power utility functions are commonly used to describe risk averse preference of
economic agents at a macro (state or aggregate) level (Pindyck, 1986; Saltari and Ticchi,
26

2007).

The power utility function also has the desirable property of decreasing absolute risk

aversion and constant relative risk aversion (CRRA). The CRRA measure reflects a constant
level of risk aversion which is plausible at an aggregate state level (Meyer and Robison, 1991;

26

Alternative approaches to eliciting individual farmer risk preferences would be surveys that
directly compares the returns of energy crops and field crops – see Jensen et. al. (2005) for a
similar study that focused on Tennessee farmers; such an approach may require considerable
resources to contact farmers in different states
78

Popp, et al., 2005).

27

The power utility function also has the benefit of yielding a good fit to the

data in many domains and is characterized by homotheticity (Danyang, 2002; Wakker, 2008).

28

The procedure to estimate CRRA (ρ) at state level is explained in appendix F.
The estimated CRRA value is substituted into the following power utility function
(1-ρ)

expression V(NPV) = [1/ (1- ρ)] * NPV

, where V(NPV) is the power utility function of risk

averse farmers. The empirical distribution of NPV is used to derive an empirical distribution for
the utility function, V(NPV). Let the mean value of the empirical distribution of V(NPV) be
#

denoted by V (NPV). Then the utility function is inverted to obtain the certainty equivalent
#

1/(1- ρ)

value as CE = (1- ρ) * [V (NPV)]
3.5

.

Results and Discussion
A sample budget generated for corn and miscanthus crops in Nebraska for the year 2013,

i.e. 4

th

year in a 10-year rotation, is given in the following table 3.9a. Similar budgets were

forecast for all field crops and energy crops corresponding to years 2010-2019. In 2013, the yield
of corn could be expected to average 122.5 bushels/acre in Nebraska. The net economic returns
for corn was calculated at $ 153.58, 286.56 and 428.33 per acre for low, reference and high crude
oil price scenarios respectively. The yield of miscanthus would be 9.9 tons/acre with net returns

27

Aggregate risk aversion at the level of state or multiple counties – since ‘a group’ of farmers
in that region have to plant energy crops if energy crops were to become a major feedstock for
cellulosic ethanol production in that region. The finest resolution of data for costs of production
and commodity prices are available only at state level (rather than county level or farm level); an
implicit assumption here is that farmers are considered to be homogeneous within a state.
28
Homotheticity: Multiplication of the same constant σ (> 0) with NPV of either enterprise do
not change the preferential order derived based on CE values, among alternative enterprises
79

of $-180.38, 331.44 and 1043.57 per acre under those three scenarios. The returns from one year
are not sufficient to compute the overall change in farmers’ income due to shifting to energy
crops. Hence, similar budgets were created for all years during 2010-2019, and used to compute
an estimate of NPV for energy crops and field crops. The NPV estimates for Nebraska
corn/soybean-miscanthus comparison are given in table 3.9b. The average NPV values differ
with every scenario. Energy crop NPV is less than that of corn when crude oil prices are low
(scenario I), comparable under reference case (scenario II) and higher than that of corn crop
when crude oil prices are high (scenario III). The direct implication of these results is that high
yielding energy crops (miscanthus) can be expected to compete with field crops if crude oil
prices stay high in the next 10 years. Since, crude oil prices fluctuate, NPV values also fluctuate
resulting in an empirical distribution for NPV.

80

Table 3.9a: Model Budget for corn and miscanthus crops in Nebraska (Year 4)

Expected Crude oil price in 2013
($/barrel)

Crude oil price scenario
Low
Reference
High
economic
case
economic
growth
growth
49.20
81.00
128.90

Field crop: Corn
Yield in 2013 (bu/ac)
Projected price ($/bushel)
Production costs ($/ac)
Energy input costs
Diesel
N fertilizer
P fertilizer
K fertilizer
Non energy input costs
Fixed costs
Land rent
Labor
Net returns for 2013 ($/ac)

122.50
3.14

122.50
5.64

6.82
30.86
3.25
117.19

9.29
37.91
8.56
117.19

12.73
46.95
13.61
117.19

66.71
5.72
153.58

66.71
5.72
268.56

66.71
5.72
428.33

9.93
24.40

9.93
78.58

9.93
153.88*

49.93
35.39
235.65

68.03
43.47
235.65

93.17
53.84
235.65

66.71
35.00
-180.38

Energy crop: Miscanthus
Yield (ton/ac)
Residual value ($/ton)
Production costs ($/ac)
Energy input costs
Diesel
N fertilizer
P fertilizer
K fertilizer
Non energy input costs
Fixed costs
Land rent
Labor
Net returns for 2013 ($/ac)

122.50
4.20

66.71
35.00
331.44

66.71
35.00
1043.57*

* the values are very high due to very high residual value that could be paid for
energy crop biomass under higher crude oil prices; the actual prices paid by the
biorefinery will be less than these residual values. Also note that miscanthus yield
81

levels are very high in Nebraska making it highly profitable if there is demand for
miscanthus biomass.
Table 3.9b: NPV values of corn and miscanthus crop for Nebraska (2010-2019)
NPV value
Corn-Soybean rotation
Miscanthus crop

Crude oil price scenario
Low
Reference case High
$ 1538
$ 2547
- $ 1839
$ 2096

$ 3949
$ 7587

Miscanthus and corn-soybean (CS) rotation
The following results are derived based on the empirical distributions of NPV for risk
neutral farmers based on 1000 simulations for every state. The break-even prices required for
miscanthus to compete with CS rotation – to cover energy crop production costs and the
opportunity costs (lost profits) – are given in figure 3.8. The three figures correspond to three
high, reference case and low crude oil price scenarios. The subsidies required for cellulosic
biomass are summarized in figure 3.9. If the break-even price was more than the residual value
of biomass (biofuel revenues minus processing costs), then subsidies for cellulosic biomass are
required as required in low crude oil price scenario. But if the residual value is more than the
break-even price, then subsidies are not required as can be seen in high crude oil price scenario.
The probability of biomass residual value exceeding the break-even price (corresponding to CS
29

rotation) is given in figure 3.10.

Figures 3.8, 3.9 and 3.10 correspond to planting miscanthus in

fields that are traditionally planted with corn and soybeans in yearly rotation.
When crude oil prices are low, the break even prices required for miscanthus range from
$27-71/ton. The residual value of energy crop biomass in the corresponding low crude oil price
scenario was estimated at $22 per ton. Since the break even prices are higher than the residual
29

Similar tables are available for other field crops (cotton, wheat) and other energy crop
switchgrass. Only the results comparing miscanthus with corn are discussed here.
82

value of biomass, additional subsidy support is needed to convert lands from corn-soybean
rotation to energy crops (miscanthus in this case). In the absence of subsidy support over the
entire 10 year time frame, it will not be economical to divert corn-soybean fields to miscanthus
production in low crude oil price scenario because it leads to a reduction of expected NPV.
Hence farmers would not convert their lands from field crops to energy crops when crude oil
prices are low. To illustrate, the break-even price for miscanthus in Alabama is $31/ton. Only
$22/ton would be covered by the residual value. Hence, a subsidy of $9/ton is required for
Alabama (figure 3.9). The subsidy of $9/ton for Alabama is required for all biomass produced
during 2010-2019.

30

That is, if 84 tons of miscanthus were produced per acre over a 10 year

time-period, then the total amount of subsidy required would be $9/ton * 84 tons = $756 per acre
over 10 years. Such large amounts of required subsidies indicate that energy crops are
uncompetitive against corn-soybeans rotation when crude oil prices are low. The results derived
from the empirical distributions on the probability of receiving higher returns from miscanthus
also reflect the same result (figure 3.10). Compared to miscanthus, switchgrass yields would be
even lower leading to higher break even prices and requiring more subsidies. Hence, conversion
of corn-soybean cropland to switchgrass (or any other energy crop with similar yield potential) is
much more unlikely when crude oil prices are low. The amount of subsidies reported above is
very sensitive to processing costs. When processing costs are assumed to increase from the target
rate of $1/gallon to the current levels of $1.50/gallon, the required subsidies are found to
increase substantially.

30

There is a federal subsidy of $45/ton under USDA-BCAP program for all biomass including
energy crops during the first two years of planting (USDA, 2010).
83

Figure 3.8: Break even prices for miscanthus replacing corn-soybean rotation ($/ton)
I.
Low Crude Oil Price
II. Reference Case
68

68

84
84

35
35

56
56

44
44

71
71

70 59
70 59

52
52

36

36

4242

32
32

47
47

88
88
66
66

72
72

33
33

29

48
48
34
34
46

49
49

41
41

37

46

37

38
38

III.

High Crude Oil Price

63
63

50
50

56
56

40
40

29
31
31

47
47

86
86

37
37

40
40

43
43

51
51

69
69

58
58

54
54

31

44
44

Legend

104
104
70
70
87
87

71
71

109
109
80
80
61
61

60
60

78
78

90
90

107
107

46 - 55

69
69

51
51

49
49
56
56

29 - 45

45
45

56 - 65
41
41

66 - 75
76 - 85

59
59

52
52

86 - 95
96 - 105
106 - 125

31

Figures 3.8, 3.9 and 3.10 compare the competitiveness of miscanthus against that of cornsoybeans rotation
84

Figure 3.9: Subsidy required for miscanthus replacing corn-soybean rotation ($/ton)
I.

Low Crude Oil Price

II. Reference Case

46
46

15
15
14
14
6
6

34
34
36
36

49
49

38
38

33
33

25
25

22
22

30
30

15
15

18
18
1
1

10
10
19
19

11
11

88
17
17

9
9

0
0

20
20
16
16

III.

High Crude Oil Price

Legend

0-5
6 - 10
11 - 25
26 - 30
31 - 35

0
0
36 - 40
41 - 45
46 - 50

85

Figure 3.10: Probability that cellulosic biomass subsidies are not required for miscanthus
replacing corn-soybean rotation (percentage)
I.

Low Crude Oil Price

II. Reference Case
0.35
0.35

0.22

0.8
0.8

0.22

0.02
0.02

0.29
0.29

0.16
0.16

00

0.78
0.78

0.57
0.57
0.52
0.52

0.54
0.54
0.63
0.63

0.13
0.13

0.68
0.68

0.85
0.85

0.96
0.96
0.98

0.23
0.23

0.89
0.89

0.11

0.03
0.03

III.

0.92

0.92

0.11
0.01
0.01

0.06
0.06

0.98

0.86
0.86

High Crude Oil Price

1
0.93 1
0.93
0.91
0.91

0.88
0.88

Legend

0.81
0.81

ProbLow

0.94
0.94
0.89
0.89

0.96

0.00 - 0.10

0.96

0.11 - 0.20

0.8
0.8
0.77 0.9
0.77

0.98
0.98

0.21 - 0.30

0.9

0.97
0.97

0.31 - 0.40
0.41 - 0.50

0.99
0.99

0.51 - 0.60
0.61 - 0.70

1
1

0.71 - 0.80
0.81 - 0.90
0.91 - 1.00

86

In reference case scenario (II), the state level analysis suggest that miscanthus is
competitive with CS rotation in the states Alabama, Arkansas, Florida, Georgia, Kentucky,
Louisiana, Mississippi, North Carolina, South Carolina, Tennessee and Virginia. When the crude
oil prices fluctuate in the range of $49-113 per barrel in scenario II, the break even prices for
miscanthus increase by 17% to 45% compared to scenario I. The reasons for an increase in
break-even prices are two-fold: (i) higher miscanthus production costs due to higher agricultural
input costs and (ii) higher opportunity costs associated with higher profits that were lost with
replacing corn-soybeans rotation with miscanthus. In scenario II, there is no need for cellulosic
biomass subsidies because the break even prices are lower than residual value estimated at
$70/ton (except the Midwestern states). The high probability estimates (figure 3.10) shows that
the economic returns are consistently higher with miscanthus. The reasons for better consistency
are two-fold: (i) the residual value of cellulosic biomass is higher in the reference crude oil price
scenario which consistently increased the returns from energy crops; (ii) while the NPV for
miscanthus and corn–soybeans increased in scenario II compared to scenario I, the increase in
the former is much higher than for the latter; i.e. the opportunity cost of planting miscanthus and
replacing corn-soybean crops declined. In scenario III (crude oil prices ranged from $88178/barrel), miscanthus is found to be attractive in all states including the Midwestern cornsoybean belt. The residual value available to pay for cellulosic biomass is above $150/ton
eliminating the need for any cellulosic biomass subsidies. That is, energy crop cultivation would
be very competitive with corn-soybeans rotation when the crude oil prices stay high over the 10
year duration, 2010-2019.

87

Miscanthus and other crops
Table 3.10 presents a comparison of miscanthus with other major crops such as wheat,
sorghum, and cotton or land uses such as pasture for selected states. The pattern of results is
similar to those discussed above. The major difference is with the specific
Table 3.10: Break-even price for miscanthus ($/ton) – selected crops, selected states
State
Indiana
Missouri
Texas
Kansas
Tennessee

Crop
Corn-soybeans
Wheat
Cotton
Sorghum
Pasture

I. Low*
59 (37)
42 (20)
41 (19)
49 (27)
32 (10)

Crude oil price scenario
II. Reference case
72
50
48
59
35

III.

#

High
87
59
56
69
38

* The amount of subsidies required (given in the parenthesis) is calculated as Max[0, Breakeven price – residual value ($22/ton in low crude oil price scenario)]. Under the assumption of
$1/gallon processing costs, subsidies are not required for reference case and high crude oil price
scenario.
#
More results on break-even prices for miscanthus when planted in place of other crops are
given in Appendix G
values estimated for break-even prices of miscanthus. Since wheat, cotton and sorghum crops are
less profitable than corn and soybeans, the break-even prices are lower. This also reduces the
amount of required subsidies. These results show that the competitiveness of energy crops is
heavily dependent on the type of land that is planted with miscanthus and the type of crop that is
replaced.
The energy crops would be more competitive in pasture lands or other marginal lands
such as Conservation Reserve Program (CRP) land, because of low economic returns from
pasturelands. Although field crop productivity is low in pasture or marginal lands, pastures are
more suitable to grow herbaceous grasses such as miscanthus. The low land opportunity cost of
diverting them to energy crop production is another reason that contributes to low break-even
88

price for miscanthus and low subsidies. These results for other states were found to follow the
same pattern and remain consistent with all crops in other states and not presented here. The
probability of getting higher economic returns increased substantially in crude oil price scenarios
II and III.
Impact of risk aversion
For risk-averse farmers, the comparison of NPV values is replaced with a comparison of
certainty equivalent values. Even when the farmers are modeled as risk averse, the economic
returns from miscanthus compare favorably with field crop returns when crude oil prices are
high. The results remain the same in terms of geographic suitability and economic viability in all
three scenarios. The risk-averse farmers would view energy crops less favorably (favorably)
when the crude oil prices are low (high). The major difference is in value of probability estimate.
Due to greater dependence on fossil fuel inputs for field crop production, the fluctuations in
crude oil prices translate into higher fluctuations in the NPV values of corn-soybeans returns.
This reduces the certainty equivalent value derived from corn-soybeans which in turn increases
the probability of receiving higher returns from energy crops for risk-averse farmers. The results
are not very different for risk neutral (NPV) and risk-averse (CE) farmers because the estimated
constant relative risk aversion coefficients (CRRA) are relatively low. CRRA estimates can
range from zero to infinity. The break-even risk aversion coefficient procedure following in this
study gave CRRA estimates in the range of 0.99-1.003. This range corresponds to lower end of
risk aversion. Hence, the impacts of risk can be expected to be low at the aggregate (state) level.

89

3.6

Conclusions
The potential for dedicated herbaceous energy grass crops such as miscanthus, and

switchgrass is less certain because they are yet to be grown in large scale. The demand for
energy crops and the economics of energy crops are affected by crude oil prices. To evaluate the
role of crude oil prices, a multi-state study was conducted over a 10 year 2010-2019 time frame.
First, crude oil prices were simulated using EIA estimates and a weighted average of historical
and implied price volatility of crude oil prices. The cost of energy inputs (diesel, fertilizers
NPK), prices for agricultural outputs (corn, soybeans, wheat, sorghum, cotton) were derived
based on the simulated crude oil price. The yield of field crops were projected using the
historical trend levels. Similarly, the crop production budgets for energy crops (miscanthus and
switchgrass) were generated for the years 2010-2019. The energy crop input costs and output
prices were derived based on crude oil prices; the energy crop yields were derived based on the
state level temperature and precipitation data. The repeated simulation of budgets over 1000
iterations enabled estimation of empirical distribution for the NPV values for field crops and
energy crops. The NPV values for field crops and energy crops were compared to model the
decision of risk neutral farmers. The NPV values were monotonically transformed to derive
certainty equivalent (CE) values for modeling risk averse farmer decisions. The CE values were
derived by assuming a power utility function characterized by constant relative risk aversion
(CRRA). The CRRAs were derived for each state using SERF procedure described in Hardaker
et al (2004).
The results suggested that for energy crops to be competitive with existing land uses, the
crude oil prices would have to be high over the period 2010-19. The probability of getting higher
90

economic returns from energy crops such as miscanthus increased substantially when crude oil
prices, biofuel and cellulosic biomass prices were high. In particular, the southeastern states such
as Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina,
South Carolina, Tennessee and Virginia would be more suitable for energy crop cultivation due
to higher yields and higher economic returns. Cellulosic biomass subsidies for energy crops
would be required only when crude oil prices remained low. The amount of subsidies estimated
for energy crops was found to be sensitive to biofuel processing cost assumption Switchgrass
might not become a major cellulosic feedstock unless its yield levels are comparable to that of
miscanthus or other high yielding energy crops. The above results derived for a risk neutral
farmer by comparing NPV values remained relatively similar for risk averse farmers as well. The
low level of risk aversion level at the aggregate (state) level was the main reason for lack of
differences between risk neutral and risk averse farmers. The key determinants of energy crop
competitiveness were found to be the yield potential of energy crop which depends on the
regional climatic differences, and the residual value of cellulosic biomass that depends on crude
oil prices, processing and transport costs. This study confirms that subsidies for energy crops are
required when the crude oil prices are low. That is, the subsidies have to be counter-cyclical to
crude oil price movements. This is markedly different from the proposed Biomass Crop
Assistance Program (BCAP) which pays matching subsidies where the subsidies are equal to the
payment received from the biomass processing facility. Under the BCAP program, the subsidies
will be higher when the crude oil prices are higher because the latter increases the ability to pay
more for energy crops. The results of this study suggest an alternative way to administer the
biomass subsidy program.
91

3.7 APPENDICES

92

Appendix A Real options value
I account for the important factors that cause uncertainty in energy crop returns: crude oil
prices, and climatic parameters. In addition to crude oil price fluctuations, farmers face other
types of uncertainties that may add or reduce value and affect their utility derived from planting
energy crops. One of the most common forms of additional economic value emerge from the
ability to delay energy crop planting till a later date (Dixit and Pindyck, 1994; Price and
Wetzstein, 1999; Song, et al., 2009). Other sources of uncertainties that affect energy crop
planting include (i) delay in energy crop adoption in wait of better energy crop varieties, (ii)
delay in energy crop adoption in search of more information about local markets for energy
crops and prices (which implicitly depends on crude oil prices), (iii) uncertainty due to public
policy changes that affect monies (prices and subsidies) available for energy crop biomass and
cellulosic biofuels, (iv) lack of perspectives on preferred feedstock or technology for cellulosic
biofuel production, and (v) unknown demand for the yet-to-be planted energy crops.

32

All these

factors delay immediate planting of energy crops due to inherent economic value (termed as real
options value – see criterion D in table 3.2). I exclude these factors for the following reasons:
(a) The uncertainties such as ‘information on preferred feedstock, demand for feedstock,
better seed varieties,’ are faced equally by all farmers in all states. Hence, the economic value
arising with delaying the adoption of energy crops is likely to be similar for farmers across the
US states. In other words, the real option value arising from these factors will not be the
distinguishing factor across the states. In fact, the farmers will not grow energy crops if there is
no biorefinery within a 50-100 mile radius, because there will be no demand for energy crop
biomass. This study is applicable to cases where biorefinery is planned creating sufficient
32

Agricultural lands can be converted from energy crop plantations to field crop production as
already seen with returning CRP lands into row crop production. Although such a conversion
potentially incurs high costs of land reclamation, the farmers would require a higher price for
energy crop biomass rather than delay planting of energy crops to avoid land irreversibility.
93

demand for cellulosic biomass for biofuel production purposes. Alternatively, if energy crop
production is competitive in a location, it will reduce the constraints faced by the biorefinery
with regard to sourcing biomass feedstocks.
(b) The farmers do not delay energy crop planting when there is demand for energy crops
resulting in higher prices for energy crops or higher residual value after accounting for biofuel
processing costs. The farmers subscribed to growing switchgrass in Tennessee pilot project, and
in Charitan Valley Biomass Project in Iowa because the markets and prices were assured.
Switchgrass planting was immediate in both cases when the farmers have assured contracts for
energy crops – hence, the main question remains what prices and revenues would farmers require
as part of the contract to plant energy crops instead of existing land uses.
(c) If the biorefineries consider energy crops as the preferred feedstocks, then
biorefineries would promote planting of energy crops. Biorefineries would rather not delay
planting of energy crops because energy crops require an establishment period of 1-2 years
before achieving maximum yield potential. Hence, biorefineries have incentives to reduce the
value arising due to additional real options value to secure cellulosic biomass for biorefinery
operations. In fact, the farmers who sign up first would be the preferred suppliers since those
farmers help reduce uncertainty faced by the biorefinery. This study is applicable to evaluating
energy crop competitiveness when there is sufficient demand for biomass. The results from this
study can help decide whether to locate a biorefinery, that prefers energy crop biomass, in a
particular geographic region or not.
(d) Crude oil price fluctuations affect the ability to construct a biorefinery and create
demand for energy crop biomass. The uncertainties faced in the construction of biorefineries are
much bigger than the uncertainties faced by the farmers in growing energy crops. The results
from this analysis are applicable when a biorefinery has been decided in a location removing the
uncertainties described above.
94

e) Long term contracts are needed to procure entire biomass produced from energy crop
fields. These long term contracts will include terms of farm production support to farmers in
establishing, and growing energy crops immediately (Kumarappan and Gustafson, 2009). That
is, the contract terms will be designed to reduce market uncertainties and promote planting of
energy crops immediately once the biorefinery is located in the vicinity. These contract terms
essentially function to diminish the real options value in delaying the planting of energy crops by
farmers. Although individual efforts of biorefineries to encourage farmers plant energy crops
would vary with the location and management, all biorefineries will have to provide comparable
incentives. This study evaluates those incentives in terms of break-even prices, subsidies and
probability of receiving higher returns from energy crops. A simulation study that derives an
empirical distribution of NPV for energy crops is an useful analysis – this is the focus of my
research. Due to the above reasons, the real options value to delay energy crop planting is
deemed low and excluded from this study.

95

Appendix B Forecasting Historical Volatility
Step 1: Calculation of weights (based on 1993-2009 data)

33

(Granger and Ramanathan, 1984):

ω0^, ω1^, ω2^ are the estimated weights (regression coefficient estimates) derived from
the regression rt = ω0 + ω1 IVp + ω2 HVp + ε, where
rt is the spot price volatility estimate calculated as ln(St/St-1)2 using daily data for years 1993 –
2009
St is the spot market price of crude oil at time t
IVp is the past data on implied volatility (based on options on futures) for years 1993 – 2009
HVp is the past data on historical volatility (based on futures contracts) for years 1993 – 2009
ε is the error term ~ N(0,1)
Step 2: Forecast of crude oil price volatility (for years 2010-2019)
SVy = ω0^ + ω1^ IVf + ω2^ HVf, where
SVy is the weighted volatility forecast for every year y = 2010…2019 (t = 1, 2, …10)
34
IVf is the implied volatility forecast (using IVp) for years 2010 – 2019
HVf is the historical volatility forecast (using HVp) for years 2010 – 2019

35

The series IVp and HVp are available from Commodity Research Bureau, CRB (2009). These
two series IVp and HVp are used to forecast future values (IVf and HVf) for years 2010-2019
using GARCH type models.
Forecasting historical volatility (HVf):
I use HVp estimates provided by CRB (2009) to forecast HVf. Similar to IVp series, the
historical volatility data series (HVp) is assembled by averaging daily estimates corresponding to
33

Data on historic volatility is available from 1993 – hence that is the reason for the starting
year in 1993.
34
For years 2016 through 2018, less number of options are traded in NYMEX. Hence, the 2015
value for implied volatility will be multiplied by √T, where T = 2 for year 2016, 3 for year 2017
and so on.
35
Among GARCH models, single equation GARCH(1,1) model is found to be better predictor
of historical volatility than other GARCH models (see Padorsky, 2006)
96

crude oil futures contracts. To illustrate, consider the crude oil futures contract that would expire
in September 1995. This contract would have been actively traded on a daily basis during the
months of September 1994 – August 1995. Thus, there are about 250 estimates of HVp for the
futures contract expiring in September 1995. A single historical volatility estimate for the
September 1995 contract is derived as an equally weighted average of all those 250 estimates.
Using equally weighted averages ensures consistency with the procedure used for assembling the
dataset on IVp. This procedure was repeated for all months during the years 1989-2009 to
assemble the monthly time series data for HVp. This series is fitted with an ARMA-GARCH
model to forecast HVf.
Based on the correlogram, and information criteria (AIC), an ARMA(3,1) was found to
account for autocorrelation in HVp. An ARMA(3,1)-GARCH(1,1) model was chosen to forecast
HVf. To account for the fact that crude oil price volatility is heteroskedastic (see figure 3.5), I
included crude oil prices as an exogenous regressor in the ARMA-GARCH estimation. I
compared two GARCH models where crude oil prices entered only the mean equation or only
the variance equation as a regressor. Using crude oil price as a regressor in the mean equation
yielded normally distributed errors (Null Hypothesis: Normally distributed errors; Jarque Bera
statistic = 2.778483 with a p-value of 0.2492) – this overcomes the problem noted by Agnolucci
(2009) that the residual error terms were not normally distributed for any of the GARCH type
models fitted for crude oil price volatility. The forecast HVf values are presented in figure 3.5.
Since the projections are made using monthly data, there are 12 volatility estimates for every
year during 2010-19. I compute a simple average of first 12 observations and use it as the
projected HVf estimate for 2010; the forecast observations numbered 13 to 24 are used to derive
the HVf forecast for year 2011 and so on.

97

Appendix C Forecasting implied volatility
The IV data for the years 1993-2015 (only a few selected months for the years 201036

2015) is available from Commodity Research Bureau (CRB, 2009, 2009b).

The daily IV

estimate for a contract is computed using eight call and put options on crude oil price futures
contracts:
two call options that are just in-the-money;
two call options that are just out-of-the-money;
two put options that are just in-the-money and
two put options that are just out-of-the-money.
There are many advantages to using only the contracts that are just in or out-of-the-money
contracts (in other words, nearest to the money contracts): they are found to be the least biased
when volatility is time-varying; using only the nearest to the money options provides implied
volatility forecast estimates that are robust to non-normal errors; and nearest to the money
options yield estimates that are least affected by non-simultaneity (Corrado and Miller, 1996;
Fleming, et al., 1995); also see footnote 1 in Agnolucci (2009).
Since options contracts are traded daily till expiry, options contract for every month has
daily estimates of implied volatility expressed in annualized terms. I derive one representative
estimate from this time series data. For example, the options contract expiring in June 2011 is
actively traded during the prior 30 months (2008-2011 for this example). There are
approximately 660 daily observations (22 days per month * 30 months = 660 daily observations)
for June 2011 alone the options contract was traded. I compute the equally weighted mean of
these estimates and use it as the volatility estimate for June 2011. I adjust the time to expiry to

36

Options on crude oil futures are actively traded for 4 to 5 years (2010-2015). Beyond that,
they are less liquid and those options do not provide much information on future implied
volatility
98

compute the equally weighted mean values (Hull, 2006). Uniformly weighted measures are
found to be the least biased estimates compared to other weighting procedures (Corrado and
Miller, 1996; Ederington and Guan, 2002). This procedure is repeated for all months between
January 1993 and December 2012. Since there are no estimates of IVf for the months during
2013-2019, I use ARMA-GARCH model to project IVp (table 3.6). The forecast values are in
monthly terms; since I need only annual estimates of IVf, I compute non-overlapping average of
IVf estimates from January till December for the years 2013-2019.

99

Appendix D Temperature (GDD) and Precipitation Values
GDD and precipitation values in various states are given in tables 3.D1 and 3.D2,
respectively. Both GDD and precipitation vary across states affecting energy crop yields. Even
within a state, they vary over time. These intra-state variations are captured using a uniform
distribution. The simulated values of temperature and precipitation are plugged into equations in
table 3.5 with minimum and maximum values. To illustrate a representative energy crop yield
calculation, consider the state of Alabama: the average GDD is 5535, precipitation is 57.3 inches
resulting in switchgrass yield of 2.2 + 0.02 * 5535 * 57.3/1000 = 8.5 tons per acre per year and
miscanthus yield of 6.6+0.04 * 5535 * 57.3/1000 = 19.3 tons per acre per year.

100

Table 3.D1: Growing Degree Days in US states
FIPS
State
Min Average Median
1 Alabama
4,978
5,535
5,671
5 Arkansas
4,334
4,995
5,224
4 Arizona
2,804
4,861
4,747
6 California
1,163
4,333
4,515
8 Colorado
2,451
3,185
3,228
9 Connecticut
3,321
3,381
3,381
10 Delaware
3,921
3,967
3,967
12 Florida
5,998
6,836
6,770
13 Georgia
5,044
5,599
5,678
19 Iowa
2,862
3,193
3,182
16 Idaho
2,345
2,830
2,675
17 Illinois
3,287
3,628
3,633
18 Indiana
3,229
3,658
3,621
20 Kansas
3,528
4,083
4,133
21 Kentucky
4,030
4,406
4,514
22 Louisiana
5,726
6,155
6,319
25 Massachusetts
2,706
2,903
2,845
24 Maryland
3,858
3,911
3,911
23 Maine
1,903
2,215
2,238
26 Michigan
1,724
2,543
2,548
27 Minnesota
1,723
2,370
2,387
29 Missouri
3,358
3,962
3,942
28 Mississippi
5,125
5,651
5,655
30 Montana
1,846
2,378
2,387
37 North Carolina
3,805
4,739
4,714
38 North Dakota
2,172
2,432
2,441
31 Nebraska
3,028
3,332
3,243
33 New Hampshire
290
1,531
1,531
34 New Jersey
3,879
3,889
3,889
35 New Mexico
3,358
4,453
4,574
32 Nevada
2,795
3,773
3,599
36 New York
2,518
2,933
2,933
39 Ohio
3,004
3,423
3,419
40 Oklahoma
4,695
4,985
5,054
41 Oregon
1,593
2,471
2,511
42 Pennsylvania
2,173
3,150
3,208
44 Rhode Island
3,344
3,344
3,344
45 South Carolina
4,929
5,417
5,356
46 South Dakota
2,469
2,750
2,749
101

Max
5,914
5,363
7,381
6,861
4,019
3,441
4,012
8,122
5,977
3,702
3,347
4,066
4,291
4,488
4,639
6,463
3,159
3,963
2,524
3,259
3,082
4,570
6,315
2,887
5,335
2,684
3,792
2,771
3,899
5,316
6,592
3,436
3,868
5,179
3,177
4,085
3,344
5,902
3,098

Table 3.D1 (Continued)
FIPS
State
47 Tennessee
48 Texas
49 Utah
51 Virginia
50 Vermont
53 Washington
55 Wisconsin
54 West Virginia
56 Wyoming
Source: (NOAA, 2009)

Min Average Median
3,916
4,656
4,771
4,253
5,952
5,979
2,768
3,258
3,350
3,830
4,325
4,282
2,382
2,554
2,554
1,067
2,312
2,406
2,181
2,678
2,694
2,779
3,491
3,547
1,693
2,376
2,480

102

Max
5,356
7,504
3,550
4,767
2,725
3,199
3,059
3,992
2,630

Table 3.D2: Precipitation in US states (in inches)
FIPS
State
Min Average Median
1 Alabama
53.4
57.3
55.9
5 Arkansas
40.9
54.8
49.3
4 Arizona
3.2
10.7
8.0
6 California
15.0
13.2
8 Colorado
7.6
11.8
11.2
9 Connecticut
41.7
42.9
42.9
10 Delaware
40.8
40.8
40.8
12 Florida
39.6
52.8
51.8
13 Georgia
44.6
48.3
49.5
19 Iowa
25.9
32.8
33.4
16 Idaho
12.1
12.2
12.1
17 Illinois
30.4
36.0
18 Indiana
34.8
39.2
39.5
20 Kansas
18.2
26.6
28.8
21 Kentucky
44.4
47.0
46.9
22 Louisiana
46.1
55.9
57.9
25 Massachusetts
41.5
46.1
47.8
24 Maryland
40.8
40.8
40.8
23 Maine
36.6
40.5
40.5
26 Michigan
28.3
32.1
32.6
27 Minnesota
24.4
28.0
28.3
29 Missouri
37.5
39.3
38.3
28 Mississippi
55.4
56.0
55.9
30 Montana
11.0
13.8
14.3
37 North Carolina
41.4
48.5
47.6
38 North Dakota
13.7
16.2
15.5
31 Nebraska
15.3
23.8
25.0
33 New Hampshire
36.4
67.7
67.7
34 New Jersey
40.3
42.1
42.1
35 New Mexico
8.9
12.2
12.6
32 Nevada
4.1
8.0
8.2
36 New York
32.0
40.0
38.9
39 Ohio
33.0
36.9
36.8
40 Oklahoma
33.4
37.0
37.0
41 Oregon
10.0
33.5
36.0
42 Pennsylvania
36.2
39.9
40.6
44 Rhode Island
45.5
45.5
45.5
45 South Carolina
48.5
50.3
50.6
46 South Dakota
16.6
19.8
19.3
103

Max
64.0
74.1
22.8
37.0
16.2
44.1
40.8
65.7
51.0
38.4
12.4
39.1
43.1
35.2
49.7
61.9
49.0
40.8
44.3
36.0
30.0
43.0
56.7
16.5
56.1
19.5
29.9
99.0
44.0
15.1
10.1
47.3
39.7
40.6
66.4
43.5
45.5
51.5
23.9

Table 3.D2 (Continued)
FIPS
State
Min Average Median
47 Tennessee
40.7
49.1
49.7
48 Texas
8.8
29.0
29.5
49 Utah
9.8
13.0
13.0
51 Virginia
39.9
41.9
41.1
50 Vermont
34.5
34.5
34.5
53 Washington
36.5
37.2
55 Wisconsin
28.8
30.8
30.7
54 West Virginia
41.0
42.5
42.0
56 Wyoming
12.5
13.6
13.7
Source: (Utah Weather Center, 2009)

Max
53.8
57.2
16.2
44.6
34.5
105.2
32.9
44.8
14.5

Table 3.D3: Fixed costs of land in crop production in US states
Land Rental Value ($/ac/year)
Compound
Annual Growth
State
Minimum Average Maximum
Rate
20
40
80
0.97%
AL
98
0.00%
19
61
AR
32
172
236
2.78%
AZ
13
342
369
CA
5
24
33
CO
28
67
122
DE
21
39
82
2.59%
FL
26
45
96
-0.60%
GA
38
140
216
0.45%
IA
70
-0.07%
5
59
ID
37
140
213
0.43%
IL
32
118
183
0.83%
IN
15
41
70
2.21%
KS
22
81
125
0.97%
KY
19
68
106
-0.18%
LA
28
65
120
MD
29
69
127
1.92%
MI
21
93
135
0.43%
MN
28
82
138
0.04%
MO
18
65
101
0.74%
MS
6
21
32
MT
26
56
109
-2.26%
NC
12
42
65
-0.08%
ND
13
78
105
0.56%
NE
104

Table 3.D3 (Continued)

State
NJ
NM
NV
NY
OH
OK
OR
PA
SC
SD
TN
TX
UT
VA
WA
WI
WV
WY

Land Rental Value ($/ac/year)
Compound
Annual Growth
Minimum Average Maximum
Rate
28
51
106
-2.46%
47
2
43
54
6
43
28
42
97
-1.67%
32
90
154
0.87%
10
30
50
-3.44%
15
107
138
105
0.95%
28
48
21
31
74
-2.12%
13
54
81
2.70%
20
71
112
-0.68%
9
24
42
-0.71%
11
67
89
20
43
84
2.25%
237
15
207
41
76
157
-1.56%
22
31
76
-1.26%
4
45
54

Source: (USDA - NASS, 2007); the growth rates are computed only for major states in the
eastern half of US which suit energy crop production (Graham, et al., 2007).

105

Appendix E List of State Extension Websites
Table 3.E1: List of state extension websites used to collect field crop production costs
State
State Extension Budget link
http://www.ag.auburn.edu/agec/pubs/budgets/
Alabama
Arkansas http://www.uaex.edu/depts/ag_economics/crop_budgets.htm
http://nfrec.ifas.ufl.edu/programs/enterprise_budgets.shtml
Florida
http://www.ces.uga.edu/Agriculture/agecon/printedbudgets.htm
Georgia
http://www.extension.iastate.edu/agdm/crops/pdf/a1-20.pdf
Iowa
http://www.farmdoc.illinois.edu/manage/newsletters/fefo08_13/fefo08_13.html
Illinois
http://www.agecon.purdue.edu/extension/pubs/crop_guide_05.pdf
Indiana
http://www.ksre.ksu.edu/library/DesktopDefault.aspx?tabid=16#FarmRanch
Kansas
http://www.uky.edu/Ag/AgEcon/pubsperson.php?searchterm=halig&db=faculty
Kentucky &ppage=halich.php
Louisiana http://www.lsuagcenter.com/en/money_business/farm_business/budgets/
Minnesota http://www.apec.umn.edu/faculty/wlazarus/tools.html
http://www.agecon.msstate.edu/what/farm/generator/
Missouri
North
http://www.ag-econ.ncsu.edu/faculty/bullen/bullen.htm
Carolina
North
http://www.ag.ndsu.edu/pubs/ecguides.html
Dakota
New
http://aesop.rutgers.edu/~farmmgmt/ne-budgets/conventional_practices.html
Jersey
http://aede.osu.edu/Programs/FarmManagement/Budgets/crops-2009/
Ohio
Oklahoma http://agecon.okstate.edu/budgets/sample_pdf_files.asp
Pennsylva
http://agguide.agronomy.psu.edu/cm/sec12/sec12toc.cfm
nia
South
http://cherokee.agecon.clemson.edu/crop_bud.htm
Carolina
South
http://econ.sdstate.edu/Extension/otherlinks.htm
Dakota
Tennessee http://economics.ag.utk.edu/budgets.html
http://agecoext.tamu.edu/resources/crop-livestock-budgets/by-commodity.html
Texas
http://www.ext.vt.edu/pubs/agecon/446-047/446-047.html
Virginia

106

Appendix F Calculation of Break-even Risk Aversion Coefficient
McCarl, et al., (1987) developed a procedure to use Break-even Risk Aversion
Coefficient (BRAC) to estimate CRRA. The concept behind BRAC metric (ρ) is that farmers
prefer to grow field crop j1 rather than j2 when the utility from j1 (R1) is higher and vice versa.
At a particular level of returns, the farmers will be indifferent between growing either crop j1 or
j2. Hardaker et al (2004a) argued that the level of returns can be estimated as the certainty
equivalent value which I the basis of SERF procedure. Hardaker, et al., (2004a) estimate ρ as the
constant that drives the difference B3 = (CEj1 – CEj2) to zero, where j1 and j2 refer to two major
1/(1- ρ)

#

field crops, CEj1 = (1- ρ)* [V (Rj1)]

#

1/(1- ρ)

and CEj2 = (1- ρ)* [V (Rj1)]

; Rj1 and Rj2 are

#

the returns from the two major crops; V (Rj) represents the average utility achieved from either
crop. I choose corn and soybeans as the two major crops since they are grown in most states in
the study region. Hardaker, et al., (2004a - Appendix B) illustrate the procedure to calculate ρ for
power utility functional form (V).

37

Repeating this procedure for every US state generates

different risk aversion coefficients for farmers across various US states.

37

See (Griffith, et al., 2009) for an alternative procedure to compute the risk aversion coefficient
for negative exponential utility function (using Risk Root software).
107

Appendix G Break-even prices for miscanthus replacing wheat
Table 3.G1: Break-even price for miscanthus when planted in wheat fields ($/ton)*
Crude Oil Price Scenario
State
I. Low
II. Reference
III.High
51.66
53.75
57.92
FL
51.68
55.70
49.66
GA
66.10
70.86
63.70
IL
41.36
43.74
48.51
KY
31.21
32.82
36.04
MN
47.78
49.86
54.02
MS
50.83
52.72
56.48
NC
44.01
47.26
42.37
ND
49.50
51.16
54.45
NE
56.68
59.25
64.35
OH
44.38
45.71
48.36
OK
55.69
57.98
62.54
PA
38.95
40.72
44.25
SC
46.23
47.75
50.78
SD
49.44
53.56
47.38
TN
37.60
38.80
41.20
TX
46.78
49.04
53.55
WI

* These dollar values refer to the average break-even price required for
miscanthus to replace continuous wheat crop in selected states.
These break-even prices for planting wheat fields are similar to the break-even price values
presented in table 3.10 (various crops) and figure 3.8 (corresponding to corn-soybean rotation).
These results confirm that energy crops have to be supported in selective US states because the
break-even prices (and profitability) vary by state and by the crude oil price scenario.

108

3.8 REFERENCES

109

References
Agnolucci, P. 2009. "Volatility in crude oil futures: A comparison of the predictive ability of
GARCH and implied volatility models." Energy Economics 31(2):316-321
Anand, M., et al. 2009 Farmers’ Willingness to Produce Biofuel Feedstocks: Examining the
Case for Cover Crops in the Southeast in Agricultural and Applied Economics
Association Milwaukee, WI July 25-28, Available at
http://www.aaea.org/UserFiles/file/Posters.pdf
Andersen, T. G., et al. 2005. "Volatility Forecasting." NBER Available at
http://www.nber.org/papers/w11188 (accessed October 2009)
Angelinia, L. G., et al. 2009. "Comparison of Arundo donax L. and Miscanthus x giganteus in a
long-term field experiment in Central Italy: Analysis of productive characteristics and
energy balance." Biomass and Bioenergy 33(4):635-643
Armstrong, J. S. (2001) Selecting Forecasting Methods, ed. J. S. Armstrong, Principles of
Forecasting, Kluwer Academic Publishers, pp. 365-386.
Bangsund, D. A., E. A. DeVuyst, and F. L. Leistritz. 2008. "Evaluation of Breakeven Farm-gate
Switchgrass Prices in South Central North Dakota." Available at
http://purl.umn.edu/37845 (accessed October 2009)
Bangsund, D. A., E. A. DeVuyst, and F. L. Leistritz. 2008. "Evaluation of Breakeven Farm-gate
Switchgrass Prices in South Central North Dakota-Summary." North Dakota State
University Available at http://purl.umn.edu/42253 (accessed August 2009)
Benjamin, C., M. Houee-bigot, and C. Tavera 2009 What are the long-term drivers of food
prices? Investigating improvements in the accuracy of prediction intervals for the
forecast of food prices in Agricultural and Applied Economics Association - 2009 Annual
Meeting Milwaukee, Wisconsin July 26-28, 2009, Available at http://purl.umn.edu/49436
Campiche, J. L., et al. 2007 Examining the Evolving Correspondence Between Petroleum Prices
and Agricultural Commodity Prices in American Agricultural Economics Association 2007 Annual Meeting Portland, Oregon July 29-August 1, Available at
http://purl.umn.edu/9881
Cassida, K. A., et al. 2005. "Biomass Yield and Stand Characteristics of Switchgrass in South
Central U.S. Environments." Crop Sci 45(2):673-681
Chibnik, M. 1981. "Small Farmer Risk Aversion: Peasant Reality or Policymakers'
Rationalization?" Culture & Agriculture 1(10):1-7

110

Collins, R. A., and E. E. Gbur. 1991. "Borrowing Behavior of the Proprietary Firm: Do Some
Risk-Averse Expected Utility Maximizers Plunge?" Western Journal of Agricultural
Economics 16(2):251-258
Corrado, C. J., and T. W. Miller. 1996. "Efficient option-implied volatility estimators." Journal
of Futures Markets 16(2):247–272
CRB. 2009. CRB DataCenter: Final Markets. Available at http://www.crbtrader.com/ (accessed
August 2009)
CRB. 2009b. Option Implied Volatility and Futures Historical Volatility. Available at
http://www.crbtrader.com/support/options.asp (accessed August 2009)
Creyts, J., et al. 2007. "Reducing U.S. Greenhouse Gas Emissions: How Much at What Cost?"
McKinsey & Company (pp 47) Available at
http://www.mckinsey.com/clientservice/ccsi/pdf/US_ghg_final_report.pdf (accessed
August 2009)
Danyang, X. 2002. Power Risk Aversion Utility Functions. Available at
http://ideas.repec.org/p/wpa/wuwpif/0207006.html (accessed January 2009)
De La Torre Ugarte, D., and C. Hellwinckel. 2004. "The Economic Competitiveness and Impacts
on the Agricultural Sector of Switchgrass as a Bioenergy Dedicated Crop." Agricultural
Policy Analysis Center, University of Tennessee Available at (accessed
De la Torre Ugarte, D., and D. E. Ray. 2000. "Biomass and bioenergy applications of the
POLYSYS modeling framework." Biomass and Bioenergy 18(4):291-308
De la Torre Ugarte, D. G., et al. 2000. "The Economic Impacts of Bioenergy Crop Production on
U.S. Agriculture." Available at http://bioenergy.ornl.gov/papers/wagin/index.html
(accessed January 2009)
Dempster, J. P. 1999. Insect Populations - in Theory and in Practice (pp 303). 1 ed: Springer
Dimakos, X. K., L. R. Neef, and K. Aas. 2006. "Net Present Value with uncertainty." Norsk
Regnesentral (Norwegian Computing Center, NR) Available at
http://publications.nr.no/samba0506.pdf (accessed
Dixit, A. K., and R. S. Pindyck. 1994. Investment under Uncertainty. Princeton, N.J.: Princeton
University Press
DOE-EERE. 2010. Biomass - Multi-year Program Plan (Appendix Table C-3). Available at
http://www1.eere.energy.gov/biomass/pdfs/biomass_mypp_november2010.pdf (accessed
December 2010)
111

Du, X., C. L. Yu, and D. J. Hayes 2009 Speculation and Volatility Spillover in the Crude Oil and
Agricultural Commodity Markets: A Bayesian Analysis in Agricultural and Applied
Economics Association - 2009 Annual Meeting Milwaukee, Wisconsin July 26-28,
Available at http://purl.umn.edu/49276
Ederington, L. H., and W. Guan. 2002. "Measuring Implied Volatility: Is an average better?
Which average?" The Journal of Futures Markets 22(9):811-837
EIA. 2009. Annual Energy Outlook 2009. Available at http://www.eia.doe.gov/oiaf/aeo/
(accessed August 2009)
EIA. 2003. The National Energy Modeling System: An Overview 2003. Available at
http://www.eia.doe.gov/oiaf/aeo/overview/overview.html (accessed August 2009)
EIA. 2008a. A Primer on Gasoline Prices. Available at
http://www.eia.doe.gov/bookshelf/brochures/gasolinepricesprimer/ (accessed August
2009)
EIA. 2010. Short-Term Energy Outlook – Real Petroleum Prices (Imported Crude Oil Prices).
Available at http://www.eia.doe.gov/emeu/steo/pub/fsheets/real_prices.html (accessed
January 2010)
EISA. 2007. Energy Independence and Security Act of 2007. Available at
http://frwebgate.access.gpo.gov/cgibin/getdoc.cgi?dbname=110_cong_bills&docid=f:h6enr.txt.pdf (accessed January 2009)
Fan, Y., Q. Liang, and Y.-M. Wei. 2008. "A generalized pattern matching approach for multistep prediction of crude oil price." Energy Economics 30(3):889-904
FASB. 2004. Appendix E: Glossary - Volatility pp: 281. Available at
http://www.fasb.org/pdf/fas123r.pdf (accessed January 2010)
Figlewski, S. 2004. "Forecasting Volatility." New York University Stern School of Business
Available at http://pages.stern.nyu.edu/~sfiglews/Docs/Forecasting%20Volatility.pdf
(accessed October 2009)
Fleming, J., B. Ostdiek, and R. E. Whaley. 1995. " Predicting stock-market volatility: a new
measure." Journal of Futures Markets 15(2):265–302
GAO. 2003. "Domestic Nitrogen Fertilizer Production Depends on Natural Gas Availability and
Prices." Government Accountability Office Available at
www.gao.gov/new.items/d031148.pdf (accessed August 2009)
Graham, R. L., et al. 2007. "Current and Potential U.S. Corn Stover Supplies." Agron J 99(1):111
112

Granger, C. W. J., and R. Ramanathan. 1984. "Improved methods of combining forecasts,
Journal of Forecasting." Journal of Forecasting 3(2):197–204
Griffith, A. P., et al. 2009 Stochastic Dominance Analysis of Bioenergy Crops as a Production
Alternative on an East Tennessee Beef and Crop Farm in Southern Agricultural
Economics Association -2009 Annual Meeting Atlanta, Georgia January 31-February 3,
2009, Available at http://purl.umn.edu/46811
Grove, B., P. R. Taljaard, and P. C. Cloete. 2007. "A Stochastic Budgeting Analysis of Three
Alternative Scenarios to Convert from Beef-Cattle Farming to Game Ranching." Agrekon
46(4):
Guffey, D. 2006. Chariton Valley Biomass Project. Available at
http://www.iowaswitchgrass.com/ (accessed April 2009)
Hansen, L. P., C. A. Holt, and D. Peled. 1978. "A note on first degree stochastic dominance."
Economics Letters 1(4):315-319
Hardaker, J. B., et al. 2004a. "Stochastic efficiency analysis with risk aversion bounds: a
simplified approach." The Australian Journal of Agricultural and Resource Economics
48(2):253-270
Harri, A., L. Nalley, and D. Hudson. 2009. "The relationship between oil, exchange rates and
commodity prices " Journal of Agricultural and Applied Economics 41(2):501–510
Hartley, P., K. B. M. III, and J. Rosthal. 2007. "The relationship between crude oil and natural
gas prices." James A Baker III Institute for Public Policy - Rice University Available at
http://www.rice.edu/energy/publications/naturalgas_na.html (accessed August 2009)
Heaton, E. A., F. G. Dohleman, and S. P. Long. 2008. "Meeting US biofuel goals with less land:
the potential of Miscanthus." Global Change Biology 14(9):2000-2014
Huang, W.-y. 2009. "Factors Contributing to the Recent Increase in U.S. Fertilizer Prices, 200208." Economic Research Service/USDA
Available at (accessed August 2009)
Hull, J. (2006) Options, Futures, and Other Derivatives pp 464.
Ivanic, R. (2008) Economic potential of dedicated energy crops: What is possible and how to get
there (Personal Communication), Mendel Biotechnology.
Jain, A., and M. Erickson 2008 Biofuel Crop Yield Estimation: An Integrated Biogeochemical
Cycle Modeling in (ed.) Linking Biophysical and Economic Models Chicago, IL
113

November 13-14, 2008, Available at http://greatlakesbioenergy.org/wpcontent/uploads/2008/11/Jain_Nov13.pdf (accessed
Jensen, K., et al. 2005. "Tennessee Farmers’ Views of Producing Switchgrass for Energy." BioBased Energy Analysis Group Available at
http://beag.ag.utk.edu/pp/switchgrasssurvey.pdf (accessed August 2009)
Khanna, M. (2009).
Khanna, M., B. Dhungana, and J. Clifton-Brown. 2008. "Costs of producing miscanthus and
switchgrass for bioenergy in Illinois." Biomass and Bioenergy 32(6):482-493
Kiniry, J. R., et al. 2005. "Switchgrass simulation by the ALMANAC model at diverse sites in
the southern US." Biomass and Bioenergy 29(6):419-425
Kiniry, J. R., et al. 1996. "Simulating Alamo Switchgrass with the ALMANAC Model."
Agronomy Journal 88(4):602-606
Kiniry, J. R., et al. 2008. "Switchgrass Biomass Simulation at Diverse Sites in the Northern
Great Plains of the U.S." Bioenergy Research 1(2):259-264
Kleiber, K. 2009 The Effect of Ethanol-Driven Corn Demand on Crop Choice in (ed.) AAEA Annual Meetings Milwaukee, WI Available at
http://ageconsearch.umn.edu/bitstream/49616/2/AAEA_610228.pdf (accessed May 2009)
Kumarappan, S., and C. Gustafson 2009 Biomass supply for biofuel production - contracting
issues in BIO sixth world congress Montreal, Canada July 19-22, Available at
Lai, J.-Y., R. J. Myers, and S. D. Hanson. 2003. "Optimal On-Farm Grain Storage by RiskAverse Farmers." Journal of Agricultural and Resource Economics 28(03):
Lobell, D. B., K. N. Cahill, and C. B. Field. 2006. "Weather-based yield forecasts developed for
12 California crops." California Agriculture 60(4):211-215
Magni, C. A. 2002. "Investment decisions in the theory of finance: Some antinomies and
inconsistencies." European Journal of Operational Research 137(1):206-217
Manfredo, M. R., R. M. Leuthold, and S. H. Irwin. 2001. "Forecasting Fed Cattle, Feeder Cattle,
and Corn Cash Price Volatility: The Accuracy of Time Series, Implied Volatility, and
Composite Approaches." Journal of Agricultural and Applied Economics 3(3):523-538
Mark, T., P. Darby, and M. Salassi 2009 A Comparison of Pricing Strategies for Cellulosic
Ethanol Processors: A Simulation Approach in Agricultural and Applied Economics
Association>2009 Annual Meeting Milwaukee, Wisconsin July 26-28, 2009, Available at
http://purl.umn.edu/49527
114

Maryland Energy Administration. 2003. Biofuel Potential of Switchgrass in Maryland. Available
at www.agnr.umd.edu/maes_exe/dividends/staver1.pdf (accessed April 2009)
McCarl, B. A., et al. 1987. "Stochastic Dominance over Potential Portfolios: Caution regarding
Covariance." American Journal of Agricultural Economics 69(4):804-812
McLaughlin S.B, D. G. de la Torre Ugarte, C.T. Garten Jr, L.R. Lynd, M.A. Sanderson, V.R.
Tolbert and D.D. Wolf. 2002. "High-Value Renewable Energy from Prairie Grasses."
Environmetal Science Technology 36(2122-2129
Mehrotra, R., R. Srikanthan, and A. Sharma. 2006. "A comparison of three stochastic multi-site
precipitation occurrence generators." Journal of Hydrology 331(1-2):280-292
Meyer, J., and L. J. Robison. 1991. "The Aggregate Effects of Risk in the Agricultural Sector."
American Journal of Agricultural Economics 73(1):18-24
Modi, A. Business Standard (2009) Biodiesel looking up as crude prices rise June 24, 2009;
Available at http://www.business-standard.com/india/news/biodiesel-lookingas-crudeprices-rise/361933/
Murat, A., and E. Tokat. 2009. "Forecasting oil price movements with crack spread futures."
Energy Economics 31(1):85-90
Myers, R. J. 1989. "Econometric testing for risk averse behaviour in agriculture." Applied
Economics 21(4):541 - 552
NOAA. 2009. Growing Degree Days. Available at
http://www.cpc.noaa.gov/products/analysis_monitoring/cdus/degree_days/grodgree.txt
(accessed January 2009)
Olczyk, T., et al. n.d. Updates on Fertilizer Prices. Available at
http://www.google.com/url?sa=t&source=web&ct=res&cd=1&url=http%3A%2F%2Fmia
midade.ifas.ufl.edu%2Fpdfs%2Ftropical_fruit%2FUpdates%2520on%2520fertilizer%2520
prices.pdf&ei=c7mFSu3DLoKGQbGlYS7CQ&usg=AFQjCNE34WkaR9JjVGfeBtKNngCBA2_RCQ&sig2=uv7wCe5Ox
M1gis2zFnHtQQ (accessed August 2009)
Pei, X., et al. 2005. "Effects of Farmers' Risk Attitudes and Personality Types on Production and
Marketing Decisions." Purdue University, College of Agriculture, Department of
Agricultural Economics, Aug Available at http://ideas.repec.org/p/pae/wpaper/05-10.html
(accessed

115

Perlack, R. D., et al. 2005. "Biomass as Feedstock for a Bioenergy and Bioproducts industry:
The Technical Feasibility of a Billion-Ton Annual supply." Available at
www.feedstockreview.ornl.gov/pdf/billion_ton_vision.pdf (accessed
Perrin, R., et al. 2008. "Farm-Scale Production Cost of Switchgrass for Biomass." Bioenergy
Research 1(1):91-97
Pindyck, R. 1986. "Risk aversion and determinants of stock market behavior." National Bureau
of Economic Research Available at http://www.nber.org/papers/w1921.pdf (accessed
October 2009)
Poon, S.-H., and C. W. J. Granger. 2003. "Forecasting Volatility in Financial Markets: A
Review." Journal of Economic Literature 41(2):478-539
Popp, M., M. Rudstrom, and P. Manning. 2005. "Spatial Yield Risk Across Region, Crop and
Aggregation Method." Canadian Journal of Agricultural Economics 53(2-3):103-115
Price, J., and M. E. Wetzstein. 1999. "Irreversible Investment Decisions in Perennial Crops with
Yield and Price Uncertainty." Journal of Agricultural and Resource Economics 24(173185
Ramaratnam, S. S., et al. 1986. "Risk attitudes and farm/producer attributes: A case study of
Texas coastal bend grain sorghum producers." Southern Journal of Agricultural
Economics 18(2):85-95
Ray, D. E., et al. 1998a. Chapter 6: Crop Supply Module - The POLYSYS Modeling Framework:
A Documentation. Available at http://apacweb.ag.utk.edu/poly/ch6.pdf (accessed
November 30)
Ray, D. E., et al. 1998. POLYSYS - Policy Analysis System. Available at
http://www.agpolicy.org/polysys.html (accessed August 2009)
Reilly, J. M., and K. O. Fuglie. 1998. "Future yield growth in field crops: what evidence exists?"
Soil and Tillage Research 47(3-4):275-290
Richardson, J. 2004. "Simulation for applied risk management with an introduction to the Excel
simulation add-in: Simetar." Department of Agricultural Economics, Texas A & M
University Available at (accessed
Richardson, J., K. Schumann, and P. Feldman. 2004. "Simetar: Simulation for Excel to analyse
risk." Department of Agricultural Economics, Texas A & M University Available at
(accessed
Robb, T. 2007. Biomass derived energy Southwest KS Project. Available at
www.kcc.ks.gov/energy/kreeec/presentations/A2_Tom_Robb.ppt (accessed April 2009)
116

Sadorsky, P. 2006. "Modeling and forecasting petroleum futures volatility." Energy Economics
28(4):467-488
Saltari, E., and D. Ticchi. 2007. "Risk aversion, intertemporal substitution, and the aggregate
investment-uncertainty relationship." Journal of Monetary Economics 54(3):622-648
Shoemaker, R., D. McGranahan, and W. McBride. 2006. "Agriculture and Rural Communities
Are Resilient to High Energy Costs." AmberWaves - April
Song, F., J. Zhao, and S. Swinton 2009 Switching to Perennial Biofuel Crops under Uncertainty
and Irreversibility in Agricultural and Applied Economics Association Milwaukee,
Wisconsin July 25-28, Available at
Taher, M., and A. Shadmehri. 2008. "Estimating Growth Rates and Decomposition Analysis of
Agricultural Production in Iran (1970-2000)." Trends in Agricultural Economics 1(1):1426
Tyner, W. E., and F. Taheripour. 2008. "Policy Options for Integrated Energy and Agricultural
Markets." Review of Agricultural Economics 30(3):387–396
US Treasury. 2009. U.S. government securities/Treasury constant maturities/Inflation indexed.
Available at
http://www.federalreserve.gov/releases/h15/data/Annual/H15_TCMII_Y10.txt (accessed
August 2009)
USDA - NASS. 2007. Land Values and Cash Rents 2007 Summary. Available at
http://usda.mannlib.cornell.edu/usda/nass/AgriLandVa//2000s/2007/AgriLandVa-08-032007.txt (accessed December 2008)
USDA. 2009. USDA Long-term Agricultural Projection Tables. Available at
http://www.ers.usda.gov/Briefing/Baseline/summary.htm (accessed August 2009)
Utah Weather Center. 2009. Normal Monthly Precipitation, Inches. Available at
http://www.met.utah.edu/jhorel/html/wx/climate/normrain.html (accessed December
2008)
Vadas, P. A., K. H. Barnett, and D. J. Undersander. 2008. "Economics and Energy of Ethanol
Production from Alfalfa, Corn, and Switchgrass in the Upper Midwest, USA." Bioenergy
Research 1(1):44-55
Vo, M. T. 2009. "Regime-switching stochastic volatility: Evidence from the crude oil market."
Energy Economics 31(5):779-788
Wakker, P. P. 2008. "Explaining the characteristics of the power (CRRA) utility family." Health
Economics 17(12):1329–1344
117

Walsh, M. E., et al. 2003. "Bioenergy Crop Production in the United States -- Potential
Quantities, Land Use Changes, and Economic Impacts on the Agricultural Sector."
Environmental and Resource Economics 24(4):313-333
Yu, L., S. Wang, and K. K. Lai. 2008. "Forecasting crude oil price with an EMD-based neural
network ensemble learning paradigm." Energy Economics 30(5):2623-2635
Zhang, Q., and M. Reed 2008 Examining the Impact of the World Crude Oil Price on China's
Agricultural Commodity Prices: The Case of Corn, Soybean, and Pork in Southern
Agricultural Economics Association - 2008 Annual Meeting Dallas, Texas February 2-6,
Available at http://purl.umn.edu/6797
Zhang, X., K. K. Lai, and S.-Y. Wang. 2008. "A new approach for crude oil price analysis based
on Empirical Mode Decomposition." Energy Economics 30(3):905-918

118

CHOICE OF OPTIMUM FEEDSTOCK PORTFOLIO FOR A CELLULOSIC
ETHANOL PLANT – A MULTI-PERIOD LINEAR PROGRAMMING SOLUTION

4.1

Introduction
Future cellulosic biorefineries are expected to be large-scale facilities using

multiple sources of feedstocks. Determining the optimal combination of feedstocks for
these biorefineries involves a number of considerations: (i) assuring a reliable supply and
uniform quality of biomass over the entire productive lifetime of cellulosic biorefinery; (ii)
lowering procurement costs (harvest, baling, transport, storage, and seasonal costs); and
(iii) reducing in greenhouse gas (GHG) emissions to qualify as a cellulosic biofuel under
the federal renewable fuels standard or similar regulations, and possibly for tradable GHG
credits . To ensure that cellulosic biomass feedstock meets these requirements, biorefineries
take a lead role in developing their feedstock supply. Since the biorefineries control the
extent of acreage contracted with one or more feedstocks, their feedstock procurement
process would resemble a vertically integrated operation where the suppliers deliver
cellulosic feedstock as per the needs of the biorefinery which essentially gets control of
fields planted and feedstocks supplied to the biorefinery. In this paper, I evaluate the spatial
and temporal patterns of such vertically integrated biomass procurement that include both
perennial and annual agricultural feedstocks.
Biorefineries constructed in the Midwest are likely to draw from two major types of
agricultural feedstocks: (1) energy crop biomass derived from perennial grasses such as
switchgrass, miscanthus, and mixed grasses and (2) agricultural residue biomass derived
from annual field crops such as corn, wheat, and barley. While agricultural residues are
already produced along with feedgrains, perennial grasses are not yet grown commercially

119

Table 4.1: Differences between agricultural residues and energy crops
Feedstock Type

Annuals

Perennials

Crops

Corn stover, Wheat or Sorghum
straw

Miscanthus, Switchgrass

Available biomass for
harvest (dry tons/ac)

1 – 1.25
(harvest limited quantities
every year or larger quantities
once every 2 - 3 years)

10 maximum
(30% year 1; 70% year 2;
100% years 3 – 7;
80% years 8 – 10)

Standard deviation of
yield in percent terms

Relatively stable – depends on
how much is collected

25%

Typical harvest span

July – December

November– February

Contracting

Farmers allow
4
periodic harvesting

Need to be harvested every
5
year or season

Cost of biomass raw
materials

$20 - $25 per ton

$30-$45 per ton

Harvest costs

Low

High

Transport costs

High, due to low biomass density
and large collection area

Slightly lower, due to high
biomass density

Theoretical ethanol
conversion rate

99 – corn stover
6
96.4 – wheat straw

106.1

69.3

74.3

Byproduct residues are not
allocated any GHG emissions
associated with corn grain
production, only conversion
related GHGs are considered

48,500 g of GHG per ton of
cellulosic biomass (511-653
tons of GHG per million
gallon of cellulosic ethanol)

#

1

2

3

7

(gallons/dry ton)
Conversion rate at 70%
of theoretical maximum
GHG emissions

8

(based on actual
conversion rate)
1

Assuming a collection rate of 33% of total straw produced as a coproduct with
grains (footnotes continued in the next page)
120

2

Maximum potential yield based on miscanthus yield

3

Annual crop harvest starts with spring wheat harvesting in May/June and extends
till corn harvesting in Oct/Nov/Dec ; energy crops are expected to be harvested by
the end of the growing season and possibly into winter months
4
Contracts can be written with an option (not) to harvest depending on the needs of
the biorefinery
5
If energy crops are grown in a field, harvesting biomass would be the only source
of income from that piece of land – hence, the farmers growing energy crops would
require harvesting every year
6
Corn stover chemical composition: Arabinan (2.54% mass), Xylan (18.32%),
Mannan (0.4%), Galactan (0.95%) and Glucan (34.61%) – corresponding to 44
Corn stover Zea mays Stalks and Leaves without cobs ; wheat straw composition:
Arabinan (2.35% mass), Xylan (19.22%), Mannan (0.31%), Galactan (0.75%) and
Glucan (32.64%) %) – corresponding to 154 Wheat Straw Triticum aestivum
Thunderbird whole plant
7
Switchgrass chemical composition: Arabinan (3.19%), Xylan (23.27%), Mannan
(0.22%), Galactan (1.05%) and Glucan (33.04%) – corresponding to 126 Cave-inrock high yield variety ; ethanol yield from switchgrass and miscanthus are will be
comparable because both are herbaceous energy crops with similar physiological
traits and chemical composition.
8
Based on the default parametric assumptions based on GREET model version 1.8
on a large scale. Thus, biorefineries need to choose the total acreage and locations
contracted for agricultural residues and perennial grass production/harvest. To address this
question, I develop a cost minimization model that identifies the optimal temporal and
geographical combination of perennial and annual feedstocks surrounding a biorefinery.
The optimization problem accounts for the differences in characteristics between the two
feedstocks (table 4.1). As a demonstration, I use this model to calculate the optimal
feedstock mix for a representative biorefinery located in Hugoton, southwest Kansas.
The two feedstocks – perennial energy crops and agricultural residues – differ on
various characteristics. Energy crops have higher biomass yields, which increase the
density of biomass availability (more tons per unit area) and shrink the extent of biomass
collection area. The climate hardiness of energy crops allows flexible harvest that extends

121

into winter months. The disadvantages with the perennial crops are higher establishment
costs, longer time delay to achieve higher yields, long term contracts necessary due to lack
of alternative markets, and potential problems in clearing the lands planted with perennial
crops. While the long-term contracts provide a secure revenue stream for farmers, they are
more constraining for biorefineries.

1

In contrast, agricultural residues have a different set of feedstock characteristics.
The revenues from agricultural residues are secondary to revenues from feedgrains – hence,
farmers are likely to be flexible with harvesting agricultural residues. Farmers also have the
option of not harvesting agricultural residues depending on whether feedstock prices more
than compensate for the added production costs and the value of leaving crop residues in
the field to maintain soil quality. More importantly, biorefineries may prefer agricultural
residues as feedstock because they result in greater reduction of greenhouse gas (GHG)
emissions. The reduction in GHG emissions is typically greater with agricultural residues
because residues are allocated little or none of primary crop production related GHG
emissions. A major disadvantage with agricultural residues is the lower yield of biomass
per acre than that of energy crops which increases the collection area radius and transport
costs. A cellulosic biorefinery also has to consider production costs, harvest costs, transport
costs, short-term versus long-term contractual commitments, life-cycle GHG emissions,
and other factors, such as losses in biomass storage and ethanol yield differences across the
two feedstocks. These factors will affect biorefinery’s spatial and temporal choice of
feedstocks within the feedstock collection area termed as ‘harvest shed.’ Typically, a

1

While the long-term contracts enable biorefineries to secure biomass supply, the prices at
which they will be purchased is expected to change over time depending on their
production costs.
122

cellulosic biomass harvest shed encompasses farm fields within a 50-mile to 100-mile
radius around the biorefinery. A harvest shed would be optimal if it can help reduce costs
based on where (spatial) and when (temporal) the feedstocks are grown.
I first develop a general mathematical programming model that can serve as a
decision tool for any biorefinery and address such tradeoffs across multiple feedstocks.
The objective is to minimize the cumulative discounted costs of biomass procurement over
the biorefinery’s productive life. I model the spatial distribution by subdividing the harvest
shed into multiple concentric circular zones. I estimate the optimal acreage to contract
within each zone and the optimal time to plant, harvest, and replant over the life of the
biorefinery. Quarterly intervals are chosen to account for seasonal cost variations. The
transport costs vary depending on the density of availability of the feedstocks. Hence, the
transport costs are endogenously determined based on the amount of biomass per unit area
– for instance, total tons per square mile. The seasonal costs are endogenously determined
based on transport costs. The environmental costs are endogenously determined based on
the acreages planted with various feedstocks. The harvest costs are exogenous but different
across the feedstocks; the studies that modeled harvest costs endogenous found no
appreciable difference between keeping harvest costs constant versus determining them
endogenously in the model . The storage costs were kept exogenous because they are likely
to be the same irrespective of the feedstock. Other exogenous parameters include biomass
yield, ethanol conversion rates, production costs, harvesting costs, long-term contractual
commitment needed to procure perennial feedstocks, life-cycle GHG emissions, and the
rate of storage loss.

123

The novel features of this modeling approach compared with existing studies are
summarized below. (i) Existing studies treat the available biomass quantities in the region
as exogenously given and then try to minimize procurement costs. In comparison, the
proposed model treats biomass acreage to be harvested as an endogenous decision variable
2

subject to overall biomass availability constraints; (ii) transport costs are endogenously
determined as a function of harvesting decisions which has not been done in any existing
studies; (iii) the temporal yield patterns of energy crops are modeled explicitly that also
affect feedstock acreage decisions unlike many other studies which use steady state average
yields; (iv) the current model incorporates the flexibility available with agricultural residue
harvest and the restrictions due to long term contracts with energy crops; (v) the possible
impacts of GHG emissions on feedstock sourcing decisions are also incorporated in the
model. As a result, this model provides better insights into the realities of biomass
procurement.
The model’s constraints include the land available for sourcing either feedstock. For
example, a certain proportion of geographic area is assumed to be potentially available for
supplying biomass within each concentric circular zone around the biorefinery. This
assumption is required because the land around the biorefinery includes non-agricultural
lands and farmlands not capable of supplying feedstocks. I assume that agricultural
residues will be procured only from these prime croplands within each concentric zone

2

Mapemba et al (2008) model acreage harvested as an endogenous variable. However,
their analyses and results are specific to Oklahoma. This model extends their analysis by
modeling a generic harvest-shed with a number of concentric circles around the biorefinery
for any location.
124

because the prime croplands are usually planted with field crops, such as corn and wheat.

3

In contrast, I assume that dedicated energy crops can be grown either on prime cropland or
on marginal croplands (e.g., pasture, fallow lands, and Conservation Reserve Program
(CRP)). The perennial energy crops will displace field crops and be grown in prime
croplands if their economic returns are competitive with the existing land uses.
I demonstrate the usefulness of the model using the case study of Abengoa
Bioenergy’s pilot plant in Hugoton, south-west Kansas. I illustrate the feedstock
composition for a proposed cellulosic ethanol biorefinery of 53 million gallons or 200
4

million liters of annual capacity. To simplify the analysis, only two feedstocks are
considered: corn stover (an annual crop residue) and miscanthus (a perennial dedicated
energy crop). Corn stover is produced every year along with corn grains; for miscanthus, I
assume that cellulosic biomass will be supplied for 10 years before replanting is required.
The biorefinery is assumed to operate for 20 years (80 quarters).
The potential harvest shed around the biorefinery is divided into six concentric
circular zones with outer radii of 5, 10, 15, 20, 30, and 50 miles. Surrounding Hugoton, KS,
prime croplands account for 12% of the geographic area, while marginal crop lands
5

account for 10% of geographic area. I evaluate two scenarios where the two feedstocks

3

This assumption will not be binding because, as the results show, the agricultural residues
are only second choice feedstocks. The chances of growing row crops on marginal lands
are rather slim because those lands are not profitably operable with row crops currently. I
maintain the same assumption that row crops will not be economical in marginal lands over
the next 10 years as well.
4
Abengoa Bioenergy’s initial production capacity will be 18 MGY which will later be
expanded to 25-75 million gallons
5
The remaining 78% of geographic area consists of agricultural lands where biomass is
not harvested from and non-agricultural lands. For the case study, I assume that residues
125

are harvested in the same (simultaneous) season and in different or subsequent (staggered)
seasons. These two scenarios help analyze the potential effect of harvest timing on the
optimal feedstock composition (Table 4.2). In scenario A, both feedstocks are harvested in
the third quarter of every year (July-Sept). In scenario B, agricultural residues and energy
crops are harvested in subsequent third quarter (July-September) and fourth quarter
(October-December), respectively.
Table 4.2: Alternative scenarios based on harvesting season and harvest shed
demarcation
Scenario
A
B
Agricultural residues from
Both feedstocks
Agricultural residues
prime croplands;
harvested during the
harvested in the third quarter;
same season (third
Energy crops harvested in the
Energy crops from prime
quarter)
fourth quarter
and marginal croplands
of every year
The results from the optimization suggest that the cellulosic biomass raw material
costs range from 60 to 70 cents per annual gallon of cellulosic biofuel. This estimate is in
the ballpark of estimates from other biomass feedstock studies . Biorefineries would prefer
to source a larger proportion of biomass from dedicated energy crops such as miscanthus in
spite of their higher establishment and production costs. The proportion of energy crops
was about 70% and 80% of cellulosic biomass raw materials in scenarios A and B,
respectively. The higher proportion of energy crops was due to the benefits of higher yields
and dense availability of biomass (tons per acre or tons per square mile around the
biorefinery). The increase in energy crops proportion in scenario B, compared to scenario
A, suggests that the ability to spread biomass harvest during lean seasons would be a
preferred characteristic. In both scenarios the spatial distribution turned out to be similar:

can be contracted only from prime croplands, while energy crops can be grown from either
on prime or on marginal croplands.
126

the energy crops were grown closer to the biorefinery, while the agricultural residues were
transported from fields farther from the biorefinery. For energy crops, the higher density of
biomass availability (in scenarios A and B) and staggered harvesting (in scenario B) offset
higher production costs. The staggered harvesting reduced biomass raw material costs
because it increased the proportion of energy crops in fields closer to the biorefinery which
in turn reduced transport costs.
The proportion of energy crops and agricultural residues depended on two factors:
the extent of marginal croplands available to grow energy crops, and the costs of sourcing
either feedstock from their ‘outer margins.’ To illustrate, let the energy crops be grown
within 15 miles radius and agricultural residues be grown within a 30 miles radius around
the biorefinery. Energy crops would be part of the optimal feedstock mix as long as the
material, transport and other costs of transporting it from a 15 mile radius were lower than
the total cost of acquiring agricultural residues from a 30 mile radius. Hence, the delivered
costs of the two feedstocks, which in turn depend on the acreages planted, would be used to
determine the optimal combination of feedstocks. These results for an individual
biorefinery reflect the outcome from other studies that estimate optimal feedstock
composition of annual and perennial feedstocks at national level . The environmental costs
did not seem to affect the optimal biomass portfolio much. Even when environmental costs
rose, the proportion of energy crops declined only slightly from 73% to 69%.

6

The optimization results generated shadow prices for binding land acreage
constraints. These shadow prices represent the value that biorefineries place on an
additional acre (or ton) of energy crops or agricultural residues grown within the harvest

6

by increasing the prices of GHG emissions from $15 to $50 per ton of carbon dioxide
127

shed. These shadow prices can be interpreted as the maximum amount that the biorefineries
are willing to pay to have one additional acre of a particular feedstock. The biorefineries
can pay a few more dollars not exceeding the shadow price amount. If the additional
payment brings in another acre or retains the last acre, then the biorefinery would still
reduce its biomass raw material cost. The shadow prices ranged from $2-8 per ton for
energy crops grown in fields located within a 10 mile radius. The shadow prices ranged
from $5-16 per ton for agricultural residues grown within a 10-20 mile radius. The salient
feature of the shadow prices was that the shadow prices varied significantly over time. The
shadow prices for agricultural residues were higher whenever energy crop output is lower.
For example, the additional value (shadow price or premium) placed on agricultural
residues was $16 per ton during the first year of operations but it declined gradually over
7

20 years. The shadow prices for energy crops were low in the beginning years; it
gradually increased over time. The yield patterns of both feedstocks and land restrictions
determined these results. The shadow prices for the zones with binding land acreage
constraints showed similar temporal pattern. The shadow prices gradually declined with an
increase in the distance of the fields/zones from the processing plant. Hence, the
biorefineries could adopt a differential pricing strategy depending on the location or
distance of the fields from the biorefinery. The lower material costs could justify a larger
price premium (shadow price) paid for agricultural residues. Adding the shadow prices to
the corresponding material and delivery costs would make both feedstocks comparable at

7

For instance, in case of scenario B, the price premiums for increasing agricultural residue
by an acre fell to $6 per ton in years 2, and gradually declined with increased supply of
biomass from energy crops.
128

the margin in every zone and every year, in other words these would also represent free
market feedstock prices.
The shadow prices for annual feedstocks declined as the supply of energy crop
biomass production ramped up over time. This shows that annual feedstocks are sought
after only as buffer feedstocks to meet biomass demand when the energy crop output is low
due to yield pattern differences. The shadow prices for energy crops were lower due to
their yield patterns and contracting limitations. During the first two years of establishment
phase, the energy crop yields would only be one-third and two-thirds of the maximum
potential yield (10 tons per acre per year). So, the benefits from an additional acre of
energy crop would accrue slowly over many years. Hence it would not be valued as much
as an additional acre of agricultural residues. Moreover, the restriction that energy crop
should be harvested during all 10 years created inflexibility and reduced the amount of
shadow price for energy crops..
4.2

Literature Review
There is a growing body of literature on issues surrounding the supply of cellulosic

biomass feedstock for biofuel production. These studies vary significantly in scope. Some
studies focus on supplying a single biorefinery with a single feedstock, while other studies
analyze the total potential supply of single feedstock within a region. Other studies analyze
supply of multiple feedstocks to a single biorefinery or to a number of refineries within a
region. These studies employ various methods, such as enterprise budgeting, supply curve
analysis, simulation modeling, and mathematical optimization.
Single feedstock for a single biorefinery: These studies typically focus on low cost
delivery of individual feedstocks based on mathematical programming models. Wang et al.

129

used a mixed integer linear programming model to study switchgrass harvest sheds. They
evaluated how the harvest shed expands with an increase in biorefinery size (from 25
million gallons to 50 million gallons) and the impacts of weather on harvesting season,
storage loss, and other related biomass supply issues such as type of baling operations
(rectangular vs. round bales), transport and storage costs. Their results show that
harvesting, baling and storage costs have to be included while modeling optimal feedstock
combinations for biorefineries.

8

Sokhansanj et al. (2006) developed a simulation-based optimization model called
Integrated Biomass Supply and Logistics (IBSAL) model to study the supply of
agricultural residues for day-to-day biorefinery operations. Kumar and Sokhansanj
modified IBSAL to study switchgrass supply in alternative forms such as circular vs.
rectangular bales, loaves or ensiled loafs. The IBSAL model identifies the optimal
sequence of activities to harvest, transport and deliver cellulosic biomass to the biorefinery
at a low cost. The biomass raw material costs were estimated at 70 – 73 cents per gallon of
cellulosic ethanol. This estimate was substantially higher than US Department of Energy
estimates of 40 – 45 cents per gallon reported in techno-economic studies . Currently, the
IBSAL model is being expanded to evaluate supply decisions of multiple cellulosic
feedstocks .
Single feedstock in a region: Many studies estimate the costs of crop establishment,
management, harvest and transport costs of energy crops. These cost estimates for a
particular enterprise, known as enterprise budgets, are commonly used to estimate the

8

Wang et al argued that rectangular bales will preferably be used immediately after the
harvest due to shorter shelf life while the round bales would be stored with plastic cover
and used during lean seasons
130

supply costs of energy crops such as miscanthus and switchgrass . These studies used
agricultural input data derived from trial plots. The results from these studies showed that
the potential for energy crops varied across regions depending on energy crop yields and
production costs. While this approach has been adopted for many states, their results are
largely limited to the region or state where the test plot sites are located. Other studies used
formulas to estimate the transport costs and logistics of supplying biomass in various
forms, such as chopped, rectangular bales, and round bales (Atchison and Hettenhaus,
2003; Gallagher, et al, 2003). They estimated harvest and baling costs to range from $11 to
$20 per ton depending on feedstock and regional conditions. A few other studies evaluated
the regional potential within the region, such as the number of biorefineries that can be
supported within a state based on feedstock composition and available biomass quantity .
Multiple feedstocks for a single biorefinery: Dunnett et al. and Jacobson et al.
argued that biomass yield levels of alternative feedstocks should be considered while
determining the optimal supply of multiple feedstocks to a biorefinery. Epplin et al.,
(2007), and Mapemba et al., (2007, 2008) developed a series of linear programming models
which studied the optimal combinations of naturally grown grasses and agricultural
residues. Their objective was to analyze the combination of multiple feedstocks for a single
biorefinery as well as multiple biorefineries in the state of Oklahoma by choosing the
number of acres planted with grasses and other feedstocks as well as the number of
harvesting units/machines required supplying cellulosic biomass.
Their mixed integer mathematical programming model maximized the net present
value of profits for a biorefinery that used the saccharification and fermentation process
over a 20 year time frame. They evaluated two major types of feedstocks: perennial grasses

131

naturally grown on Conservation Reserve Program (CRP) lands and agricultural residues
collected from prime croplands. They found that agricultural residues had a cost advantage
over cultivating perennial grasses. The energy crops did not feature prominently due to low
cellulosic biomass yield levels when their yield levels were at 3 to 4 tons per acre. These
yields are much lesser than the potential yields of 8 to 10 tons per acre when energy crops
are grown using intensive cultivation practices. According to their model, agricultural
residues were preferred more to energy crop biomass due to low raw material costs.
I analyze a different set of questions compared to Mapemba et al (2007, 2008) and
Epplin et al (2007). My model focuses on the optimal proportion of energy crops when
energy crops are cultivated intensively (Table 4.1), specifically, on the spatial and temporal
distribution of energy crops and agricultural residues within concentric circles around the
biorefinery. By changing the parameters, my model can be applied to multiple locations or
even different bioenergy outputs (generation of electricity from biomass).
Multiple feedstocks in a region: McCarl et al., evaluated the supply of agricultural
residues and forestry biomass for electricity generation purposes. Their mathematical
programming model (FASOM) was designed to maximize the objective of U.S. national
social welfare defined as the net present value of the integral of biomass demand curves
minus the integral of supply curves for the U.S. The FASOM model included biomass
supply and harvest in agricultural and forestry sectors, the amount of land used for biomass
harvesting, and shifting of lands between agriculture and forestry. The results showed that
large amounts of biomass could be sourced within the United States to displace coal.
Although the FASOM model does not have direct implications for an individual

132

biorefinery’s operations, it presents a set of constraints useful to model available land and
required biomass for energy production.
Khanna, et al., used a mathematical programming called Biofuel and
Environmental Policy Analysis Model (BEPAM) to evaluate the optimal composition of
multiple feedstocks at the U.S. national level. They estimated the supply potential of
perennial and annual feedstocks (energy crops and agricultural residues respectively) based
on economic returns from row crops, dairy operations and available farmland in 41 states to
supply one billion ton of cellulosic biomass by 2030. The BEPAM model predicted that
energy crops would be economically more suitable in marginal croplands. The state level
potential for energy crops varied with the regional characteristics, biomass yields (tons per
acre), and relative price of alternative feedstocks.
Research Gaps and Contributions: The literature review shows that many studies
focus on regional or national level biomass supply potential. While these studies are useful
for policy analysis, there is a gap in identifying the optimal feedstock combination for
individual biorefineries operations. The optimization models employed by Epplin et al.
(2007) and Mapemba et al. (2007, 2008) partially address this issue. The results from these
studies cannot be generalized because they emphasize a particular processing technology
(saccharification and fermentation process), type of output (liquid biofuel), and their
analysis is largely confined to the state of Oklahoma. I develop a more general model
which evaluates biomass feedstock supply potential for multiple outputs, with an emphasis
on the spatial and temporal patterns of harvest sheds, optimal acreage decisions and
additional price premiums, if any, payable for cellulosic biomass.

133

The existing literature largely treats the feedstocks costs constant or exogenous.
While it is a simpler approach, the biomass feedstock costs depend on acreage planting
decisions and density of biomass availability. The major cost components such as harvest,
baling and transport costs can potentially vary with harvest shed pattern. In this model I
compute transport costs, seasonal costs, and environmental costs endogenously based on
the decision variables (acreage and yield density).

9

Biorefineries could pay a higher price to achieve a desired spatial and temporal
pattern of harvest shed. The existing studies do not provide a reliable method to compute
such price premiums. Using this model, I derive such measures using the shadow prices of
binding land acreage constraints. These shadow values give an upper bound for price
premiums payable for a feedstock in a zone at a particular time.

10

This model evaluates

how GHG emission requirements affect the optimal composition of biomass feedstocks.
4.3

Model
Consider a generic biorefinery with ethanol production capacity PC (million

gallons/quarter). Its biomass raw material requirements can be met with multiple
feedstocks that include annually produced crop residues (s = 1, 2, … S) and perennials (g =
1, 2, … G). Agricultural residue yield levels are low (about 1.5 tons per acre per year); they

9

The models by Mapemba et al., and Epplin et al., endogenize only the harvesting costs by
choosing the number of harvesting units. Their results show that the number of harvesting
units was cut in half when harvesting costs are determined endogenously. But the
harvesting costs remained at $11/ton irrespective of whether they are determined
endogenously or whether they were treated exogenous. A possible reason for this result is
the change in other assumptions: for instance, the harvesting units would have been
assumed to work 24 (12) hours a day when the harvesting costs were endogenously
determined (maintained exogenous).
10
These shadow prices show the cost savings realized by adding or retaining one more
acre for biomass feedstock supply.
134

are annually produced as coproducts with feedgrains. Perennial energy crop yields are
relatively high; they are productive for τg years (normally 10 years). Since the
establishment costs of perennial crops are high, it is not economical to remove a perennial
crop soon after establishment. Hence, the farmers would seek assured contracts to sell all
energy crop biomass produced for τg years . This farmer requirement alters how
biorefineries design their harvest shed and enter into contracts for biomass. The
commitment of harvesting perennial energy crops for τg years is imposed as a constraint in
the model. This restricts that the land allocated to perennial energy crops would be retained
under energy crops for τg years. All biomass produced in those fields are assumed to be
purchased by the biorefinery.

11

The model is formulated over quarterly intervals (q) to

study how seasonal cost differences affect biomass supply and storage. The quarterly
intervals match the harvesting pattern of feedgrains and cellulosic biomass that usually
extends over three months during a crop year.

12

The quarterly intervals also help include

storage and seasonal costs that help maintain regular supply of biomass during the peak and
lean seasons of biomass harvests.

11

The farmers do not have any other alternative markets within the harvest sheds and rely
entirely on the single biorefinery which requires energy crops as raw materials.
12
Monthly intervals were not chosen due to lack of sufficient information on monthly
differences in perennial energy crop yields.
135

Figure 4.1: Concentric circular harvest shed area around the biorefinery (arrows
represent perpendicular roads used for transport):

The harvest shed is assumed to be circular with the biorefinery located at the center
(figure 4.1). The harvest shed is divided into concentric circular production zones (z =
1,2,… Z), each zone corresponding to a concentric circular zone of outer radius of Rz and
inner radius of Rz-1 miles. Each zone consists of both agricultural and non-agricultural
lands. I eliminate the non-agricultural and land unsuitable for producing cellulosic biomass
by estimating the available fraction of the area for energy crop production or agricultural
residue collection. The available area is modeled as a fraction of total geographic area in
each zone and is denoted by the symbol σ (σsz for annual feedstocks (s) and σgz for
perennial feedstocks (g)). The harvested acreage is assumed to be distributed uniformly
within every zone of the harvest shed.
Total transport costs (CT) and transport distance depends on the density of biomass
availability i.e. CT is a function of (acreage planted * yield / zone area). Thus the transport
136

costs are determined endogenously in the model. Transport costs include loading,
unloading, and trucking costs. French gave an expression for transport cost calculations for
circular harvest sheds. For a circular harvest shed, the total costs for transporting biomass
can be written as TC = N a0 +

a1 D r

2

	

where TC is the total transport

cost during a quarter (in dollars), N is the total amount of biomass required by the
biorefinery (in tons), a0 is fixed costs of transport equipment that do not depend on
distance, loading and unloading (in $/ton), a1 is variable costs ($/ton-mile), w is a constant
parameter to convert air distance to road distance, D is the density of biomass within the
circular harvest shed, and R is the outer radius of the circular harvest shed (French,
13

1960).

For a concentric circular harvest shed, I modify the above equation. The total

amount of biomass from zone z is set to Nzq; range of radii is set to Rz and Rz-1; and
biomass density within zone z is set to Dz to reflect the concentric circular zone variables.
TC

= Nzq a0 +

a1 Dz r

2

	

(1)

In equation (1), Dz is the density of biomass availability in tons per square mile in zone z. It
2

is substituted with another equivalent expression Dz = Nzq / π Rz . Similarly, the total
amount of all cellulosic biomass from zone z can be expressed as Nzq = [∑s Aszq * Yszq +
∑g ∑t Agztq * Ygtq]. Substituting both expressions for Nzq and Dz, the cost of transporting
biomass from zone z is calculated as following:
TCz = Nzq [a0 + a1
13

2

/3 w π (Rz3-Rz-13)/ (Rz2-Rz-12)]

This is obtained by combining equations (2) and (5) in French (1960)
137

(2a)

TCzq = [∑s Aszq * Yszq + ∑g ∑t Agztq * Ygtq] *
[a0 + a1

2

/3 w π (Rz3-Rz-13)/ (Rz2-Rz-12)]

(2b)

Summing the transport costs across all zones, the total transport costs CTq in quarter q
CTq = ∑z [∑s Aszq * Yszq + ∑g ∑t Agztq * Ygtq] *
[a0 + a1

2

/3 w (Rz3-Rz-13)/( Rz2-Rz-12)]

(3)

Equation (3) is added to other costs that are minimized in the objective function. Note that
CTq is dependent on the acreage decision variables Aszq and Agztq. Assuming
transportation is done through perpendicular roads, the value of w can be approximated at
√2.

14

Current developments in pilot cellulosic plants indicate that biomass will largely be
stored on field and transported to the biorefinery as and when needed for processing.
Moreover, moving the entire harvest of biomass during the harvest season is difficult due to
logistical issues and storage capacity limits . Transporting biomass during different seasons
leads to seasonal costs. The seasonal cost fluctuations arise due to changes in diesel fuel
and labor costs. The differences in labor costs are not considered for the following reason.
Harvesting cellulosic biomass requires skilled labor that operates expensive harvesting
equipment (tractors, collectors). The cost of skilled labor is relatively steady during peak

14

With perpendicular roads, a2 + b2 = c2, where a = distance traveled north-south, b =
distance traveled east-west, and c = air distance. Upon normalizing both a and b, c2 = 2 or c
= √2. That is, the sum (a+b) is also equivalent to multiplying c with √2. For example,
consider the air distance of a field from the biorefinery is 10 miles. With perpendicular
roads, the actual distance traveled would be 6 + 8 = 14 miles, derived from the relationship
62 + 82 = 102. Multiplying 10 with √2, gives 14.14, a close approximation to the actual
travel distance of 14 miles.
138

and lean season; hence, there will not be much difference in costs across different
seasons.

15

The fuel costs do change over seasons affecting biomass procurement costs . The
seasonal costs (CL) are computed by multiplying the transport costs and harvesting costs
with a factor ωq. This factor estimates the increase or decrease in costs over the four
seasons. The base or reference season is taken to be the second quarter extending from
April-June. The seasonal costs are endogenous because it depends on transport costs.
Material costs (CM), harvest costs (CH), and storage costs (CS) are assumed to be
exogenous and treated as constants in the model. All these costs are maintained the same
across all zones, expressed in terms of dollars per ton. This assumption is reasonable
because these cost components are relatively the same irrespective of the field location
within the harvest shed. Other parameters include biomass yield patterns of annual and
perennial feedstocks (Ys, Yg respectively), storage costs (ds, dg), proportion of biomass lost
in storage (εs, εg), the amount of biomass to be maintained in the inventory for continuous
functioning of the biorefinery (minimum inventory required, MIR), ethanol yield per ton of
annual and perennial feedstocks (Ks, Kg), fixed and variable cost components of transport
costs per ton mile (a0, a1), and the fraction of area available to plant either feedstock within
each zone (σsz, σgz).
The use of cellulosic biomass for bioenergy production has environmental benefits
such as reduced GHG emissions and environmental costs such as, increased soil erosion,

15

Unskilled labor wages is more likely to fluctuate over seasons
139

and greater use of chemicals (herbicides and insecticides). The promotion of biofuels to
reduce GHG emissions has led to accounting of GHG emissions at different life cycle
stages such as biomass feedstock production, biomass conversion to ethanol, and final
distribution. Leading models such as Argonne National Laboratory’s GREET model use
this approach to compare the environmental implications of using alternative feedstocks for
bioenergy production . I use GREET model’s GHG emissions data to compute the
environmental costs (or benefits) of using alternative feedstocks.
Table 4.3: Lifecycle GHG Emissions from energy crops and agricultural residues
Category
GHG emissions
In grams per million
BTU of cellulosic
2
ethanol
In tons of CO2e per

1

Fermentation process
Energy crops
(herbaceous biomass)
4589

Agricultural residues
(corn stover)
-6999

350

-534

3

million gallon
Net change in emissions
4
due to energy crops
884
0
compared to agricultural
residues (CO2e tons per
million gallon)
Source: GREET model v1.8c
1
Emissions embodied in using farming equipment, fertilizers, pesticides and
transportation of biomass are included
2
These numbers are taken from GREET model (see the underlying excel sheet,
EtOH worksheet, summary table 4). These data are sum of two forms of emissions
– CO2 and other forms of GHG emissions from biofuels. A positive (negative)
number indicates an overall increase (decrease) in GHG emissions
3
Ethanol energy content = 76,300 BTU/gallon; grams per gallon is converted into
tons per million gallon by multiplying and dividing by 1,000,000
4
Emissions from agricultural residues are normalized at zero
Table 4.3 data correspond to GHG emissions from the entire life-cycle (biomass
cultivation, harvest, transport, processing, and distribution of fuels) in producing and
140

processing biomass for cellulosic ethanol. The positive (negative) numbers refer to an
increase (decrease) in GHG emissions due to that processing pathway.

16

Since the

objective of my study is to compare two alternative feedstocks, I consider the emissions
from agricultural residues as the baseline. That is, the environmental cost (CEs) of using
agricultural residues is normalized to zero.
As seen from table 4.3, the environmental costs vary with the processing
technology. Energy crops emit more GHG than agricultural residues (+884 tons per million
gallons) under fermentation processing technology. I compute the environmental costs by
multiplying the GHG credits in table 4.3 by an expected GHG price. A reduction of GHG
emissions by one metric ton results in one GHG credit. The environmental costs depend on
the amount of annual versus perennial feedstocks processed which in turn depends on the
acreage decision variables. Hence, the environmental costs are also endogenously
determined together with transport and seasonal costs. I exclude the ‘indirect’ GHG
emissions associated with land use changes due to lack of scientific consensus on how to
estimate them .
The biorefinery’s decision problem is to minimize the net present value of
cumulative biomass procurement costs over the time period of its entire operations (e.g. 1520 years). The total biomass procurement costs include payments made directly to farmers
for biomass material (CM); payments made to contractors for harvesting (CH), transport
costs (CT), seasonal costs (CL); payments made to maintaining on-site storage structures
(CS), and internalized environmental costs (CE). The decision variables are (i) the acreage

16

Even though the emissions data are positive for certain biomass feedstocks or processes,
biofuels help reduce GHG emissions compared to GHG emissions from fossil fuels.
141

Aszq contracted to harvest agricultural residue s in quarter q in zone z, and the acreage Agztq
contracted to plant energy crop g in year t in zone z and, (ii) the amount of feedstock (s, g)
processed during every quarter q. The storage quantities are implicitly determined by
subtracting the amount of biomass processed from the amount produced during each
quarter. Note that, if some acreage is planted with energy crops in year t, then that acreage
will be retained with energy crops for the next τg years. This restriction does not apply to
agricultural residues.
Model Equations
The symbolic notations of the model are explained below:
Subscript notation:
s

= Annual agricultural residue feedstocks such as straw or stover [s = 1, 2, … S]

g

= Perennial grass feedstocks such as miscanthus, switchgrass [g = 1, 2, … G]

z

= Concentric circular production zone [z = 1, 2, … Z]

q

= The production/harvesting time period (quarter) [q = 1, 2... Q]

t

= Year in which perennial crops are planted [t = 1, 2, … T]. Perennial crop g is
assumed to supply biomass for τg years following establishment; hence, the
perennial crop g established in year 3 (t=3) will supply biomass starting in year 3
until 3 + τg

Parameters:
CMs, CMg
farmers)

= Unit material cost of feedstocks s and g (dollars per ton, price paid to

CHs, CHg

= Unit harvest cost of feedstocks s and g (dollars per ton)

CTz

= Unit transport cost of feedstock from zone z to the biorefinery located at
the center (dollars per ton)
142

CSsq, CSgq
quarter)

= Unit storage cost of feedstocks s and g in quarter q (dollars per ton per

CEgq

= Unit incremental environmental cost of perennial feedstock g in quarter q
(dollars per ton)

CXszq

= Total exogenous costs of annual feedstocks s processed in quarter
(CMs + CSs + (1+ωq) CHs, dollars per ton)

CXgzq

= Total exogenous costs of perennial feedstocks g processed in quarter
(CMg + CSg + (1+ωq) CHg + CEg, dollars per ton)

Ygtq

= Yield of perennial feedstock g, planted in year t, for quarter q [Fixed pattern of
yields in tons per acre per quarter; e.g. in scenario B, miscanthus crop planted in
year t = 3 will yield 3.33 tons/acre in quarter 12, 6.67 tons/acre in quarter 16, 10
tons/acre every fourth quarter during quarters 20 – 36, 8 tons/acre every fourth
quarter during quarters 40 – 48, and 0 tons in all other quarters If miscanthus crop
were planted in year t = 5, then the same yield pattern will be shifted from quarters
20 through 56. The amount of biomass available in quarter q depends on the
planting year (t) of miscanthus]

Ysq

= Yield of annual agricultural residues s that remains constant – harvested only once
in a year either during the third or during the fourth quarter)

Ψsq, Ψgq= Quantity of feedstock (s, g) produced within the entire harvest shed during
quarter q (tons)
Dsq, Dgq = Quantity of feedstock s and g processed at the biorefinery during quarter q
(tons)
ωq

= Factor to compute seasonal costs related to transporting; second quarter is taken
as the reference season, i.e. ωq=2 is normalized at 1 (see table 4.5)

δ

= Quarterly discount factor

d

= Storage cost parameter (dollars per ton per quarter)

εs

= Rate of loss of agricultural residue due to storage (percentage per quarter)

εg

= Rate of loss of perennial grasses due to storage (percentage per quarter)

143

PCq

= Quarterly ethanol processing capacity (gallons)

Ks, Kg = Ethanol output for feedstock s and g respectively (gallons per ton)
MIR

= Minimum Inventory Requirement (tons)

Q

= Terminal time period

PGHG = Price for one ton of greenhouse gas ($ per ton of CO2equivalent)
GCg

= Greenhouse gas credit for using energy crops, in comparison to using agricultural
residues (tons of GHG per million gallon of cellulosic ethanol)

a0

= Fixed component of transport costs ($ per ton of feedstock)

a1

= Variable component of transport costs ($ per ton-mile)

σsz

= Fraction of total land area available in zone z to harvest annual feedstock s (in
percentage)

σgz
= Fraction of land area available in zone z to harvest all perennial feedstocks g (in
percentage)
ZAz

= Total geographic area within zone z (acres)

Rz

= Outer radius of zone z (miles)

w

= factor to convert radial distance to road distance; with perpendicular road
network, w equals √2

Objective function:
Minimize discounted cumulative feedstock procurement costs over Q quarters:
q

∑q δ * [∑g ∑z ∑t CXg * Ygtq * Agztq) + ∑s ∑z CXs * Ys * Aszq)
+ (1+ωq) CTq
+ d * ∑s Xsq + d* ∑g Xgq + CEgq]
where CX refers to exogenous costs of cellulosic biomass, CT refers to endogenously
determined transport costs and d*X refers to storage costs

144

with respect to decision variables:
Aszq = Acreage contracted to harvest annual feedstock s in quarter q, zone z (in acres)
Agztq = Acreage planted with perennial feedstock g in year t, zone z (in acres; yield pattern
of perennial feedstocks is described in tables 4.1 and 4.5)
Xsq, Xgq = Storage levels (stock variable, either at the biorefinery or on farm fields) of
feedstock s and g at the end of quarter q (in tons)
Dsq, Dgq = Quantity of feedstock (stover s, grasses g) processed/demanded in quarter q –
which are implicitly determined as residuals upon choosing Xsq, and Xgq
subject to the following accounting relationships (E1-E4) and constraints (E5-E10):
Accounting relationships:
E1: Zone area ZAz (in acres) around the biorefinery extending from zonal radius Rz-1
to zonal radii Rz (in miles); the constant 640 converts square miles of area to acres
2

2

ZAz = 640 π (Rz - Rz-1 )
E2: Total biomass produced during every quarter (Ψq) is computed by multiplying the
acreage harvested (Aszq, Agztq) with yield (Ysq, Ygtq)
Ψsq = ∑z Ysq * Aszq
Ψgq

= ∑z ∑t Ygtq * Agztq

Ψq = ∑s Ψsq + ∑g Ψgq
E3: Transport costs (equation (3) from section 3):
CTq = ∑z [a0 + a1

2

/3 w (Rz3-Rz-13)/( Rz2-Rz-12)] *
[∑s Aszq * Yszq + ∑g ∑t Agztq * Ygtq]

145

E4: Environmental costs (CEgq) of perennial feedstocks are computed based on
expected GHG prices (PGHG) and GHG credit (GCg). In case of fermentation technology,
this term will be positive (additional costs due to using energy crops).

17

CEgq = PGHG * GCg * Dgq * Kgq/1000000
Constraints:
E5: Land availability constraints for perennial feedstocks:
The acreage harvested with grasses (Agztq) and agricultural residues (Aszq) should be
less than the available area from crop lands (σsz ZAz) and marginal (σgz ZAz)
croplands. This constraint has to be satisfied in every quarter q across all zones z.

18

∑g ∑t Agztq ≤ σgz ZAz
Land availability constraints for annual feedstocks
∑s Aszq ≤ σsz ZAz

for all q and z

E6: Biomass mass balance constraints: Biomass supplied from fields and storage
should equal the sum of biomass processed and inventoried in each quarter:
Biomass produced in quarter q (Ψq) + Stocks from previous quarter (q-1) = Biomass
used for biofuel conversion in quarter q (Dgq + Dsq) + Ending stock for quarter q
Ψq + [(1 – εs) * ∑s Xs q-1 + (1 – εg) * ∑g Xg q-1]
= Dgq + Dsq + [∑s Xsq + ∑g Xgq]
17

The division by 1000000 converts ethanol gallons to million gallons.

18

A different formulation of land allocation is where both feedstocks can be harvested
from all available lands. The restriction to source agricultural residues from prime
croplands and energy crops from marginal croplands can be relaxed in the following
manner. When all feedstocks can be grown in both prime and marginal croplands, the
constraint E5 is replaced with the following. The total proportion of available (prime
and marginal) cropland in every zone will be σz where σz = σsz + σgz. The summation
over years (t) adds up the acreage allotted to energy crops that are planted at different
times during the years 1 – 11. This constraint should be satisfied in every quarter q
across all zones z.
∑s Aszq + ∑g ∑t Agzt ≤ σz ZAz for all q and all z
146

E7: Biofuel produced has to meet or exceed the processing capacity (PCq) in every
quarter:
∑s Ks * Dsq + ∑g Kg * Dgq ≥ PCq

for all q

E8: Biomass stored at the biorefinery has to meet the minimum inventory required
(MIR) at the biorefinery – only this quantity of biomass incurs storage costs. The
excess biomass, if any, would be stored on field without storage costs.
∑s Ksq * Xsq + ∑g Kgq* Xgq ≥ MIR * PCq

for all q

E9: Terminal conditions for the last quarter (Q) are imposed by restricting the final
period storage to zero after meeting the biomass processing requirements
Biomass supplied from the fields in final quarter Q + supply from the storage in quarter
(Q-1) – Biomass used for conversion in Q = Ending stock for quarter Q = 0
ΨQ + ∑s (1 – εs)Xs Q-1 + ∑g (1 – εg) Xg Q-1 – DsQ – DgQ
= ∑g ∑s (XsQ + XgQ) = 0
E10: Non negativity constraints of acreage and storage decision variables:
Aszq ≥ 0; Agztq ≥ 0; Xsq ≥ 0; Xgq ≥ 0
The cost minimization problem is coded in GAMS and solved using MINOS solver.
The chosen solver helps achieve globally optimal solutions when the objective function and
constraints are convex sets ; in this model, they are linear yielding globally optimal
solutions. The results from the optimization model include: (i) the minimized total cost of
biomass, expressed in terms of dollars per annual gallon of ethanol, (ii) acreages of all
feedstocks (annuals and perennials) harvested in each quarter in each zone, (iii) variations
in biomass quantities processed versus maintained in storage, and (iv) shadow prices or
price premiums to expand land acreage within each zone. Additional sensitivity analyses
are conducted to analyze the impact of changes in exogenous parameters (e.g., land
availability, change in material costs).
147

4.4

Case Study
For the case study, I consider a biorefinery with a capacity of 53 MGY. The

geographic parameters are calculated for Hugoton, Kansas where Abengoa Bioenergy is
19

building one of its pilot cellulosic ethanol plants.

There are similarities among various

annual feedstocks in terms of harvesting flexibilities and stable biomass yield levels.
Similarly, there are also similarities among various perennial grasses in terms of plantation
establishment, productive life period (10 years) and yield patterns. Hence, I choose to
represent annual feedstocks (s) with corn stover and perennial grass feedstocks (g) with
miscanthus. Although only one annual and one perennial feedstock are used for this case
study, the model results can be extended to other feedstocks.
The objective of the optimization problem is to minimize (cumulative) biomass
feedstock procurement costs over 20 years (2011-2030), subject to following constraints:
biomass availability by type and season, biomass requirements for processing and storage,
land allocation restrictions for perennial energy crops, and the type of lands available to
grow energy crops and corn stover (see generic model and table 4.2). At an average
conversion rate of 70 gallons per ton, about 190,000 tons of biomass from both corn stover
and miscanthus would be required during each quarter. To supply the required biomass,
multiple harvests of corn stover and energy crops would be required.

19

Abengoa Bioenergy plans to use 75% of biomass from corn stover and 25% from energy
crops, provided the latter becomes economically available
148

Harvest shed around Hugoton, Kansas
I focus on the harvesting decisions within a 50 mile radius harvest shed since
previous estimates indicate availability of sufficient biomass potential within that area.

20

The circular harvest shed around the biorefinery is sub-divided into six concentric zones (z)
with outer radii of 5, 10, 15, 20, 30 and 50 miles. The geographic area within a 50 mile
radius around Hugoton, KS include the counties of Stevens, Morton, Seward, Stanton,
Grant, Haskell in Kansas; the counties of Texas, Beaver and Harper in Oklahoma; the
counties of Dallam, Hansford, Ochil-tree and Sherman in Texas; and the counties of Baca,
Bent, Kiowa and Prowers in Colorado.
An average of 36% of the geographic area in these counties is classified as prime
cropland. Crops such as corn, wheat, sorghum and barley that can supply agricultural
residues make up two-thirds of cropland area in these counties. Hence, agricultural residues
can potentially be collected from 24% of geographic area (two thirds of 36%). Since
agricultural residue collection is rotated among fields, only a portion of these fields would
be harvested in any given year. I assume that agricultural residues are harvested once every
two years. Thus, 12% of geographic area (one half of 24%) would potentially be available
to harvest agricultural residues during each year within 50 mile radius around Hugoton,
KS. Although it is possible to supply agricultural residues from marginal lands, it is not
likely to be significant because the additional returns from agricultural residues may not be
sufficient enough to warrant growing row crops in marginal lands where the economic
returns are low in general. For simplicity, I assume that feedstock acreage is distributed

20

As the subsequent results show, this restriction is not binding – i.e. there is surplus
unused land within the harvest shed of 50 mile radius
149

uniformly within each concentric circular zone. This assumption matches the assumption
underlying equation (3) used for calculating transport costs .
Table 4.4: Land area available for harvesting cellulosic feedstocks – case study
assumptions
Harvest shed area is divided
into two sections for the two feedstocks
Agricultural residues
Energy crops sourced from
sourced from prime
prime and marginal
croplands
croplands
@ 12% of geographic
@ 22% of geographic area
Zone –
area
(σgz=0.22)
Geographic
Outer
(σsz=0.12)
Radius
Area
Miles
Thousand Acres
Z1 – 5

50

6

11

Z2 – 10

201

18

33

Z3– 15

453

30

55

Z4 – 20

806

42

77

Z5 – 30

1,813

121

221

Z6 – 50

5,035

426

709

Energy crops can be grown in marginal croplands. The latter includes permanent
pastures, rangelands, and Conservation Reserve Program (CRP) acres. Mapemba et al.,
(2008) estimated that 10% of geographic area can be classified as marginal cropland in the
south-west Kansas and Oklahoma Panhandle areas (around Hugoton, KS). Energy crops
can also be grown in prime croplands provided they are economically competitive; there is
21

a possibility that farmers replant some of the prime cropland with energy crops.

The

above calculations show that 12% of prime cropland could be available for energy crop
production (where agricultural residues are not harvested from). Hence, a total of 22% of
21

To grow energy crops in prime croplands, their economic returns have to be comparable
with that of field crop returns. Since it is a farmer’s decision to plant or not plant a field
with energy crops, it is not explicitly modeled in my optimization model for the
biorefinery. An extension of this research is to model a multi-objective optimization
problem that models both biorefinery and farmers’ decisions.
150

geographic area – 10% marginal lands and 12% prime croplands – is assumed to be
available for planting energy crops. These assumptions are retained for all zones: σsz = 12%
and σgz = 22% – see table 4.4. These assumptions are relaxed by conducting sensitivity
analysis by changing the parametric values for σsz and σgz.
Decision variables
There are two sets of decision variables: (1) acreage planted with annual and
perennial feedstocks, and (2) quantity of biomass processed versus retained in storage
during each quarter.
(1a) For agricultural residues, there are potentially 480 acreage decision variables, i.e. 20
years * 4 quarters/year * 6 zones/quarter. However, agricultural residues are only harvested
during the third quarter of every year (table 4.2). The yield is restricted to zero when
st nd
th
biomass is not supplied from the fields during 1 , 2 and 4 quarters of each year.

(1b) For energy crops, the yield pattern is different; once established, energy crops can be
harvested during the next 10 years (τg=10 years or 40 quarters). That is, energy crop
established in year 1 will supply biomass during years 1-10, energy crop established in year
2 will supply biomass during years 2-11 and so on. If the energy crops are planted in each
of the 20 years, then there are potentially 120 choice variables related to energy crop
acreage (20 years * 6 zones/year). But as the biorefinery nears its shut down at the end of
20 years, the farmers may be unwilling to establish new acreages of perennial energy crops.
To reflect this unwillingness, establishment of new perennial crops is restricted to years 1
through 11; this ensures that the crop planted in year 11 will be fully harvested by year 20.
In this case, there are 66 choice variables associated with planting energy crops (11 years *
151

6 zones/year). The energy crop biomass may be harvested simultaneously with agricultural
residues during the third quarter of every year (scenario A) or in a staggered manner during
the fourth quarter of every year (scenario B).
(2) The second set of decision variables chooses the quantity of biomass processed versus
the kept in storage, either at the biorefinery or on field. There are a total of 79 storage
variables corresponding to all quarters except the last quarter (20 years * 4 quarters/year –
1). For the last time period (q = 80, denoted by Q) the storage level is restricted to zero to
reflect the terminal condition.
Cost Parameters
The following section describes how the model parameters are estimated for the
case study location in Hugoton, KS. The material or product costs refer to prices that
farmers receive for agricultural residues and energy crops. For agricultural residues, the
material costs cover opportunity costs such as potential revenues lost from selling residues
as animal bedding material and the lost fertilizer value when residues are removed. I
assume the material costs of agricultural residues at $22 per ton.

22

For energy crops, the

material costs cover crop production costs and opportunity costs such as potential lost
revenues (e.g., revenues lost from CRP payments or economic returns from earlier land
uses). I assume the material costs of energy crops at $30 per ton.

23

I conduct sensitivity

analyses to evaluate the impacts of changes in material costs on the optimal proportion of
energy crop feedstock.

22

In 2009, farmers in south central Nebraska were paid an average of $22 per ton as
material cost for agricultural residues supplied to Energy Grains Biomass LLC .
23
Mooney, et al. and Wang, et al. estimated the material costs of energy crops at $25-35
per dry metric ton.
152

Harvest costs include fixed costs of machinery equipment and variable costs of
chopping, raking, collecting, baling, hauling, and staging biomass within the farms. I
estimate harvesting costs (HC) of agricultural residues and energy crops at $14 and $16 per
ton, respectively, based on USDA reports . The harvesting costs for energy crops are
slightly higher because of intensive use of machinery in handling energy crop biomass.
These harvesting cost estimates are within the range of $11 to $20 per ton estimated by
other studies (Epplin, et al., 2007; Sokhansanj, et al., 2006; Thorsell, et al., 2004; Wang, et
al., 2009). Treating the harvest costs as an exogenous parameter is not limiting. The results
from Epplin et al (2007) and Mapemba et al (2007, 2008) showed that harvest costs
remained at about $11 per ton whether they were endogenized or held constant.
Transportation cost per ton mile is assumed at $0.28 per ton-mile based on previous
estimates . The transportation costs are endogenously determined based on the acreages
planted with either feedstock as shown in equation E3. For seasonal cost variations, I
consider the (real) price fluctuations of diesel fuel at quarterly intervals – Jan-Mar, AplJun, Jul-Sep, Oct-Dec. The first quarter serves as the reference point (normalized at 100%).
The seasonal price differences (ωq) during other quarters are given in table 4.5. The
seasonal costs are computed by multiplying the average percentage changes for a particular
season with harvest costs (CH) and transport costs (CT). The seasonal costs are also
endogenously determined due to their dependence on transport costs.
Table 4.5: Seasonal cost variations based on quarterly diesel prices (1990-2009)
1994-2010 Average
Source: EIA

Q1 – Jan-Mar
100%

Q2 – Apr-Jun
105%

153

Q3 – Jul-Sep Q4 – Oct-Dec
108%
109%

Cellulosic biomass harvest is usually limited to a short harvest window, during the
third or fourth quarter of every year. This requires construction of storage facilities,
associated costs and biomass quantity losses in inventory . Mapemba et al. estimated that
on-site storage would retain storage to supply cellulosic biomass for three weeks. In this
case study modeled in quarterly terms, this is equal to storing 25 percent of quarterly
biomass requirement (3 weeks out of a 12 week quarter). Any additional biomass would be
stored on field eliminating storage costs or losses for the biorefinery. The storage will be
maintained during all quarters except the final time period (Q = 80). The storage costs and
the biomass lost in storage (both on-field and on-site) are estimated at $3/ton/quarter and
3% per quarter respectively .
The environmental costs of cellulosic biomass feedstocks are computed based on
the amount of GHG emitted with processing agricultural residues or energy crops. Energy
crops result in higher GHG emissions when processed using fermentation and
saccharification technology (last row of table 4.3). The economic costs due to GHG
emissions (CEg) for energy crop feedstock are added to the objective function. I treat the
emissions from agricultural residues as baseline. The GHG prices ranged from $0.10 – $5
per ton of GHG in the U.S. compared to $20-30 per ton in the European markets.

24

I use

the expected GHG price level of $15 per ton of GHG; this is a reasonable estimate of GHG

24

Over the past 2 years, GHG prices ranged from $0.25 to $5 per metric ton in Chicago
Climate Exchange in the US and $10 - $25 in over-the-counter (OTC) exchanges in
Europe. Note that revenues from GHG reduction will be realized in dollars only if the
biorefineries become eligible to trade them. Firms such as Agra and AgRefresh working
towards creating this new market for biofuels sector . Even if the biofuel projects become
eligible for GHG trading, only a small part of the biofuels and biomass would satisfy the
additionality requirement and become eligible for tradable GHG credits . Including
environmental costs in the cost minimization problem ensures that the additional
environmental value created for the society due to cellulosic biofuel use is accounted for.
154

price and frequently quoted by North American emissions management funds . To compute
the environmental costs (or benefits), I multiply the GHG price of $15 per ton with the
relative GHG emissions from energy crops.
4.5

Results
The raw material costs of cellulosic biomass for biofuel production are estimated

for the parametric values reported in table 4.6. In scenario A, where both agricultural
residues and energy crops are harvested simultaneously in the third quarter, the raw
Table 4.6: Parametric values for scenarios A and B in preliminary results
Level in base
Parameter
case scenario
Costs of storage, d (per metric dry ton/year)

$12

Storage Loss (per year)

12%

Energy crops and stover grown in separate
fields
Land available for perennial energy
crop cultivation
Land available for stover collection

22%
12%

Discount rate

2%

Minimum inventory maintained at the
biorefinery facility (biomass worth 3 weeks
25%
of storage in a total of 12 weeks)
Material costs ($/dry ton, modeled using Atchison and Hettenhaus, 2003)
Energy crop
$ 30
Stover

$ 22

Harvesting costs ($/ dry ton)
Grasses

$ 16

Stover

$ 14

Transport costs per ton mile
Seasonal costs

Environmental costs

$0.28/ton mile
Reference season: quarter 1;
seasonal cost factor (table 4.5)
multiplied with transport and harvest costs
$15/ton of GHG;
quantities given in table 4.3
155

25

material costs are estimated at 64.5 cents per gallon.

In scenario B, where the agricultural

residues are harvested in the third quarter and energy crops are harvested in the fourth
quarter (staggered harvesting), the biomass raw material costs are estimated at 60.6 cents
per annual gallon. The raw material costs are lower in scenario B because staggered
harvesting helps reduce storage requirements and associated costs. These biomass raw
material cost estimates are within the range of 60-75 cents per gallon reported in other
studies.
Under the set of land and cost assumptions used in this study, the results suggest
that energy crops can be expected to supply a significant portion of cellulosic biomass. The
proportion of energy crops in the optimal supply stood at 70% and 73% of total biomass
requirements in scenarios A and B respectively. In spite of higher material and harvesting
costs, energy crops appeared in the optimal portfolio of raw materials due to two reasons.
(1) Staggered harvesting reduce storage requirements in scenario B compared to scenario A
(see table 4.2); hence, any feedstock that extends the harvest window and reduce storage
requirements would be suitable as part of cellulosic biomass portfolio. (2) Reserving
marginal croplands for exclusive cultivation of energy crops also contributed to a higher
proportion of energy crop biomass.
The temporal distribution of the acreage planted/harvested for energy crops and
agricultural residues is given in figure 4.2. The acreage under energy crops was less but
remained steady over the 20 year period. In spite of less acreage, up to 70% of biomass was
derived from energy crops. The long term land allocation restrictions and the higher yields

25

Cost per annual gallon
= (cumulative discounted cost of biomass over 20 years)/(total amount of biomass
processed over 20 years)
156

of energy crops contributed to these results. In contrast, the acreage under agricultural
residues fluctuated heavily. More acreage was allocated to agricultural residues whenever
the energy crop biomass output was low. Hence, energy crops turned out to be the preferred
feedstock. This result confirms the industry assertion that more cellulosic biomass would
be derived from energy crops .
Figure 4.2: Acreage contracted (000 acres) with either feedstock – Scenario B
600
400
200
0
1 2

3 4 5 6 7

8 9 10 11 12 13 14 15 16 17 18 19 20 Years

Figure 4.3: Composition of biomass used for biofuel production (000 tons)
– Scenario B
800
600
400
200
0
1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20

Even though a larger proportion of land (acres) corresponded to harvesting agricultural
residues, the amount of agricultural residues (tons) was much less as shown in figure 4.3.
The spatial and temporal distribution of the harvest shed is given in figures 4.4 and
4.5 for scenarios A and B, respectively. The number on the right shows the sum of prime
and marginal croplands in each zone (in thousands of acres). The results show that the
157

harvest shed extends only up to 30 mile radius around the biorefinery (zone Z5) in scenario
A; the spread is even lower at 20 mile radius (Z4) in scenario B. The geographic spread
shrank in scenario B because the higher proportion of energy crops that produce more
biomass per unit area. These results confirm the notion that whenever high yielding energy
crops are included as part of feedstock portfolio, then the fields within a 20-30 mile radius
would be sufficient to supply the biomass requirements for a biorefinery of about 50
million gallon capacity.
The optimal composition of feedstock depends on the economic trade-off between
delivered costs, biomass density (tons per acre) and land available for harvesting alternative
feedstocks. As an instance, consider how agricultural residues and energy crops are planted
in various zones. Energy crops are grown in fields located closer to biorefinery.
Agricultural residues are predominantly collected from fields/zones farther from the
biorefinery. In the model, the fields closer to the biorefinery are ideally planted with energy
crops and this forces collection of agricultural residues from fields that are farther away
from the biorefinery. The lower material costs of agricultural residues enable a higher
payment for transport costs from fields farther from the biorefinery as long as the combined
material and transport costs of agricultural residues is lower than that of energy crops.
Agricultural residues provide larger quantities of biomass whenever energy crop supply is
low in years 1-2, and 10 when energy crops are established, and reestablished. The supply
of agricultural residues is also higher during the middle years 4-7, and 14-17 due to lower
costs of transporting agricultural residues from 10-20 mile radius rather than growing
energy crops within a 10 mile radius around the biorefinery (and other associated land
constraints).
158

Figure 4.4: Spatial and temporal distribution of harvest shed – Scenario A (000 ac)*
Zone
Radius
0-5
5
5-10
10-15
15-20
20-30
30-50

* Even though the bars in figure 4.2 and 4.3 look the same, note that the acreage (denoted
by the number on the right) is significantly different across the zones.
The results show an interesting pattern of spatial distribution of feedstocks. All available
area is allotted to only one of the two feedstocks within a 10 mile radius (zones z1 and z2).
This appears like a corner solution within each zone . However, when the entire harvest
shed is considered, the solution is in fact an interior solution. The harvest shed includes
both energy crops and agricultural residues within 5-15 radius (zones z2 or z3) but during
different years. The major implication of this result is that optimal biomass supply varies
not only spatially but also temporally. The answer for the question ‘which feedstocks
would feature in the interior solution?’ is simple and intuitive – the feedstock that has lower
delivery costs (sum of all cost components explained above) would appear as part of the

159

Figure 4.5: Spatial and temporal distribution of harvest shed – Scenario B (000 ac)
Zone
Radius
0-5
5
5-10
10-15
15-20
20-30
30-50

optimal feedstock portfolio – such a cost comparison and trade-off of land acreages to
either feedstock occurs in fields 10-20 mile radius around the biorefinery (zones z3 or z4).
Substitution of energy crops with agricultural residues becomes more apparent (in the
initial and final years) in scenario B. The higher proportion of energy crops in scenario B
show that staggered harvesting is a preferred characteristic while developing feedstock
portfolios since they help reduce costs and storage requirements. In spite of lower material
costs, the proportion of agricultural crops is low due to lower yields per acre and inability
to extend harvest into winter months.
Shadow Prices
The land acreage constraints are binding only in fields closer to the biorefinery i.e.
within 0-15 mile radius around the biorefinery. That is, the biorefineries could reduce
160

Figure 4.6: Price premiums paid for cellulosic biomass by zone and year – Scenario A

Price Premium ($/ton)
16.00

Stover Z3

Stover Z4

14.00

Grass Z1

Grass Z2

12.00
10.00
8.00
6.00
4.00
2.00
1

3

5

7

9

11

13

15

17

19

Year

Figure 4.7: Price premiums paid for cellulosic biomass by zone and year – Scenario B

Price Premium ($/ton)
Stover Z3

16.00
14.00
12.00
10.00
8.00
6.00
4.00
2.00
-

Grass Z1
Grass Z2

1

3

5

7

9

11

161

13

15

17
19
Year

biomass raw material costs if more acres were to become available within this radius.
Expanding the harvest shed area by a few acres is equivalent to relaxing the constraints that
are binding. The biorefinery would be indifferent between ‘expanding the acreage by one
more acre by paying a premium’ that is equal to cost savings not exceeding the shadow
prices and ‘sourcing biomass from a field farther from the biorefinery.’ See figures 4.6 and
4.7 for the shadow prices of binding land constraints in scenarios A and B respectively. The
price premiums (dollars per ton) are obtained by dividing the shadow price (in dollars per
acre) by the amount of biomass harvested from one acre of harvest shed with any one of the
feedstocks (tons). In case of agricultural residues, the yield per acre is relatively simple at
1.25 tons per acre. In case of energy crops, the biomass yield from one acre would be 84
tons per acre (3.3 tons per acre in year 1; 6.7 tons per acre in year 2; 10 tons per acre in
years 3 through 7; 8 tons per acre in years 8, 9 and 10).
The shadow prices are a result of land constraints of the model; they are interpreted
as the maximum amount of price premium payable for either feedstock over the estimated
material prices. The shadow prices for agricultural residues are generally higher than that
of energy crops. That is, a higher premium could be paid for agricultural residues until the
delivered costs of both feedstocks become equal, at the margin.

26

In scenario A, the

biorefinery could pay a premium ranging from $5-16 per ton of agricultural residues and
$1-2 per ton of energy crops (figure 4.6). These premiums are warranted only for
feedstocks grown closer to the biorefinery: 0-10 mile radius for energy crops and 10-20
mile radius for agricultural residues. The premiums decline with increasing distance from
the biorefinery because of the interplay of increasing transport costs and varying density of
26

A larger premium for agricultural residues would indirectly result in a larger output of
feedgrains as well.
162

biomass (tons per square mile). The shadow prices for agricultural residues are higher in
the initial years but decline gradually when larger quantities of energy crops become
available. On the other hand, the shadow prices for energy crops increase over time as the
proportion of energy crop biomass increases.
These shadow price results hold true for both scenarios A and B (figures 4.4 and
4.5). The main difference is with the dollar amounts of shadow prices. Energy crops
command a higher price premium in scenario B because of the benefits of staggered
harvesting. The shadow prices are choppier for agricultural residues in scenario B, because
the reliance on agricultural residues in general decreases in scenario B where energy crop
become more preferable.
4.6

Sensitivity Analysis
This section focuses on the impacts of relaxing the parametric assumptions on the

proportion of energy crops and cost of cellulosic biomass raw materials. All results pertain
to staggered harvesting in scenario B. Table 4.7 presents the proportion of energy crops
under different material cost (CM): $30, 33, 36, and 39 per ton of energy crops and $22,
24.2, 26.4, and 28.6 per ton of agricultural residues. These values correspond to 10%-30%
increase in their respective material costs. Consider the ratio of material costs: CMg
divided by CMs. An increase in this ratio corresponds to energy crops becoming more
expensive than agricultural residues (moving along a column from top to bottom). When
this relative price ratio (CMg / CMs) increased, the proportion of energy crops decreased.
The rate of substitution of agricultural residues for energy crops differed with the absolute

163

values of material costs. For instance, the substitution was at a faster rate when the material
cost of agricultural residues was lower.
Table 4.7: Proportion of energy crops at different material costs (percentage) –
Scenario B
Agricultural Residues ($/ton)
22
24.2
26.4
28.6
Energy crops ($/ton)
percentages
72.9
74
80
81
30
67
71
74
81
33
56
64
70
73
36
51
64
70
48
39
To illustrate, consider the case where the material costs of agricultural residue are
$22/ton and $28.60/ton. In the first case, when the energy crop material costs increased by
30%, the proportion of energy crops decreased from 72.9% to 48% (by 24% - first column
in table 4.7). In the second case, for the same increase in energy crop prices, the proportion
of energy crops decreased from 81% to 70% (only by 11% - last column). These results
show that the proportion of any feedstock and its substitution depends on both absolute and
relative costs of delivering alternative feedstocks. The ratio of material costs remain same
along the leading diagonal cells (i.e. same relative price ratio, highlighted by grey
background in table 4.7), the proportion of energy crops was relatively stable. Table 4.8
presents the biomass raw material costs per annual gallon of biofuel when the material
costs of energy crops and agricultural resides increase by 10% to 30%. The raw material
costs range between 60 and 70 cents per gallon of cellulosic ethanol and remain in the
ballpark of estimates from previous studies.

164

Table 4.8: Cellulosic biomass raw material cost (cents per gallon) – Scenario B
Agricultural Residues ($/ton)
22
24.2
26.4
28.6
Energy crops ($/ton)
cents per gallon
60.6
61
62
62
30
63
64
64
65
33
66
67
68
65
36
68
69
70
67
39
Table 4.9 presents the impacts of greenhouse gas prices on optimal feedstock
portfolio. When the GHG prices are increased from $15 to $50 per ton of carbon dioxide
equivalent, the changes in feedstock mix are only minimal. The proportion of energy crops
decreased slightly from 73% to 69%. Hence, GHG emissions do not seem to be a major
factor determining the optimal feedstock mix or the proportion of energy crops. The
optimal biomass portfolio may not vary with revenues from emissions markets or the lack
of it. The differences in environmental benefits across energy crops and agricultural
residues may not be sufficient enough to warrant an alternative feedstock composition.
Table 4.9: Impact of higher greenhouse gas prices on optimal feedstock portfolio –
Scenario B
Technology
GHG Price
Proportion of
Biomass raw
($/ton)
energy crops
material cost
(percentage)
(cents per gallon)
72.9
60.6
15
Fermentation
70.7
61.1
25
69.0
62.3
50
Table 4.10 presents the changes in energy crop proportion with an increase in the
land acreage available to source either feedstock. The results are intuitive. An increase in
the acreage of marginal lands increases the proportion of energy crops (along the columns
from top to bottom); an increase in the acreage of prime croplands decreases the proportion
of energy crops (along the rows from left to right). In either case, the proportion of energy
crops stayed relatively high at 70% or more. These results reinforce that biorefineries

165

would prefer a higher proportion of energy crops as part of their feedstock portfolios, in
spite of higher material costs (or production costs). The advantages of energy crops include
higher yields (tons per acre) and ability to harvest in a staggered manner play a crucial role
in making them an attractive source of feedstock for the biorefineries.
Table 4.10: Proportion of energy crops under different land availability (percentage)
– Scenario B
Prime + Marginal cropland
Prime cropland supplying
supplying energy crops
agricultural residues
5% of
10% of
15% of
geographic area geographic area geographic area
78.6
72.0
NA
14% of geographic area
80.6
72.9
71.2
22% of geographic area
81.9
77.2
74.2
30% of geographic area
4.7

Conclusions
Cellulosic biorefineries face a challenge in developing low cost biomaterial raw

supplies including multiple feedstocks. The biorefineries use waste materials that do not
have much competing uses currently. This gives them the opportunity to design the harvest
shed or collection area that can reduce biomass procurement costs and ensure reliable
supply. Particularly, the biorefineries need to decide how, where and how much of
alternative feedstocks to be grown within the harvest shed. In this paper, I evaluate the
spatial and temporal design of such a harvest shed that include multiple agricultural
feedstocks such as perennial energy crops and annual agricultural residues.
I develop a cost minimization model for a Midwestern biorefinery that will use
agricultural feedstocks such as agricultural residues and energy crops. The model is
developed to estimate the demand for these two feedstocks from the biorefinery’s
perspective. The optimization model minimize the cumulative costs associated with
feedstock production, harvest, transport, storage, seasonal, and any environmental costs of

166

choosing a feedstock. The objective is to identify the criteria that affect the preference for
annual versus perennial feedstocks. The choice between annual and perennial feedstocks is
important because they impose significantly different restrictions and planning
requirements for the biorefinery. The two feedstocks differ in terms of yield levels, yield
pattern, cost of production, and GHG emissions.
The constraints facing the biorefinery include available land area for sourcing
feedstock, suitability of those lands, and biomass quantity required for continuous
functioning of the biorefinery. The biorefinery could face contractual inflexibilities with
perennial feedstocks (energy crops) because the energy crop farmers would require long
term contracts that ensure demand for all cellulosic biomass that do not have many
alternative uses. There is flexibility in harvesting or not harvesting agricultural residues
because the revenues from agricultural residues are only secondary compared to the
primary output of feedgrains. The biorefinery has to choose the acreage where energy crop
and agricultural residues are planted. The transport costs, seasonal costs and environmental
costs are modeled endogenous where they depend on the acreage decision variables. The
spatial patterns around the biorefinery and temporal decisions are included in the
optimization model.
The model parameters are calibrated for a case study location in Hugoton, KS
where Abengoa Bioenergy is constructing its cellulosic biorefinery. The feedstock
procurement decisions are modeled using one representative annual feedstock (corn stover)
and one perennial feedstock (miscanthus). For the case study, the energy crops are assumed
to supply biomass over 10 years. The feedstock harvest shed is divided into six concentric
circular zones with outer radii of 5, 10, 15, 20, 30 and 50 miles. The agricultural residues

167

are estimated to be harvested from prime cropland extending up to 12% of geographic area
around the biorefinery. The energy crops are assumed to be harvested from either prime
croplands (12%) or marginal croplands (10% of geographic area). The objective of the
model is to minimize the cumulative discounted biomass procurement cost over the 20 year
productive period of the biorefinery. The optimization results show that the biomass raw
material cost would range from $0.60 to $0.70 per annual gallon of cellulosic biofuel. The
energy crops are found to be preferred over the agricultural residues. The proportion of
energy crop biomass ranged from 70-80% of total cellulosic biomass quantity used over 20
year time period. Although energy crops are more expensive and they impose contractual
limitations on biorefineries, the benefits of energy crops’ higher density (up to 10 tons per
acre per year at its maximum production level) weighed in their favor. The energy crops
are preferably grown within a 10 mile radius around the biorefinery. Beyond that radius, it
would be cheaper to harvest agricultural residues than growing energy crops. These results
are robust to many parametric assumptions including the land area available to harvest
either feedstock.
The temporal pattern of harvest shed suggests that agricultural residues would be
used as buffer feedstock to meet where there is a shortage of energy crop biomass. Such
shortages happen in the initial years and middle years when the energy crops are being
established or reestablished. The proportion of energy crops increased when they can offer
other benefits such as extended harvest into winter seasons. Any feedstock that expands the
harvest shed could ideally be part of the biomass raw material portfolio for the biorefinery.
The shadow prices for the binding land constraints show that biorefineries could pay up to
$16 per ton of additional agricultural residues in certain years depending on the location.

168

This model presents shadow prices for binding land acreage constraints that can be
interpreted as the upper bound of price premiums payable to either feedstock. The shadow
prices declined for agricultural residues over time; in contrast, the shadow prices for energy
crops grew over time from $2-8 per ton. The shadow prices declined as the fields are
located farther from the biorefinery. The environmental costs (due to GHG emissions in
feedstock cultivation, harvest and transport) associated with energy crops are not found to
be substantial to cause a change in the optimal biomass portfolio of the biorefinery.

169

4.8 REFERENCES

170

References
Aden, A., M. Ruth, K. Ibsen, J. Jechura, K. Neeves, J. Sheehan, B. Wallace, L. Montague,
A. Slayton, and J. Lukas. "Lignocellulosic Biomass to Ethanol Process Design and
Economics Utilizing Co-Current Dilute Acid Prehydrolysis and Enzymatic
Hydrolysis for Corn Stover." NREL, 2002.
AgraGate. "Selling Your Credits." http://www.agragate.com/selling/cDetails.aspx.
AgRefresh. "U.S. Liquid Biofuels Carbon Markets Monetization: A Strategic Synthesis &
Quantification."
http://www.agrefresh.org/userfiles/Strategic_Synthesis_Executive_Summary_Excer
pt.pdf.
Angelinia, Luciana G., Lucia Ceccarinia, Nicoletta Nassi o Di Nassoa, and Enrico Bonarib.
"Comparison of Arundo Donax L. And Miscanthus X Giganteus in a Long-Term
Field Experiment in Central Italy: Analysis of Productive Characteristics and
Energy Balance." Biomass and Bioenergy 33, no. 4 (2009): 635-43.
Atchison, J.E., and J. R. Hettenhaus. "Innovative Methods for Corn Stover Collecting,
Handling, Storing and Transporting." NREL, 2003.
Bitsch, Vera. 2010. Personal Communication.
Brechbill, Sarah C, and Wallace E Tyner. "The Economic of Biomass Collection,
Transportation, and Supply to Indiana Cellulosic and Electric Utility Facilities Dept of Agricultural Economics, Purdue University Working Paper #08-03."
http://ageconsearch.umn.edu/bitstream/6148/2/wp080003.pdf.
Brummer, E. C. , C. L. Burras, M. D. Duffy, and K. J. Moore. "Switchgrass Production in
Iowa: Economic Analysis, Soil Suitability, and Varietal Performance." 140:
Bioenergy Feedstock Development Program, ORNL, 2002.
Ceres. "Biofuel Economics - Higher Yields Significantly Decrease the Land Required."
http://www.ceres.net/Products/Products-Ovw-BiofuelCalc.asp.
Dunnett, Alex J, Claire S Adjiman, and Nilay Shah. "A Spatially Explicit Whole-System
Model of the Lignocellulosic Bioethanol Supply Chain: An Assessment of
Decentralised Processing Potential." Biotechnology for Biofuels 1, no. 13 (2008).
EERE. "Biomass Feedstock Composition and Property Database."
http://www1.eere.energy.gov/biomass/feedstock_databases.html.
———. "Theoretical Ethanol Yield Calculator."
http://www1.eere.energy.gov/biomass/ethanol_yield_calculator.html.
EIA - DOE. "Diesel Fuel Retail Prices - Monthly."
http://www.eia.gov/emeu/steo/realprices/index.cfm.
171

EIA. "Monthly Us No 2diesel Retail Sales by All Sellers."
http://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=EMD_EPD2D_PTE
_NUS_DPG&f=M.
Energy Grains Biomass. "Stover Collection for Pellet Production."
http://contractyourstover.com/.
Epplin, Francis M., D. Clark Christopher, K. Roberts Roland, and Hwang Seonghuyk.
"Challenges to the Development of a Dedicated Energy Crop." American Journal of
Agricultural Economics 89, no. 5 (2007): 1296-302.
Ethanol RFA. "U.S. Cellulosic Ethanol Projects - under Development and Construction."
http://www.ethanolrfa.org/resource/cellulosic/documents/CellulosicPlantMap.pdf.
French, Ben C. "Some Considerations in Estimating Assembly Cost Functions for
Agricultural Processing Operations." Journal of Farm Economics 42, no. 4 (1960):
767-78.
Genera Energy. "Feedstock Operations." http://www.generaenergy.net.
Gov of Alberta. "Albertans and Climate Change."
http://environment.gov.ab.ca/info/faqs/faq3-Albertans_and_Climate_Change.asp.
Jacobson, Jacob J., Erin Searcy, Dave Muth, Erin Wilkerson, Shahabaddine Sokhansanj,
Bryan Jenkins, Peter Tittman, Quinn Hart, and Richard Nelson5. "Sustainable
Biomass Supply Systems."
http://aiche.confex.com/aiche/s09/techprogram/P145544.HTM.
Kang, Seungmo, Hayri Önal, Yanfeng Ouyang, Jürgen Scheffran, and Ü Deniz Tursun.
"Optimizing the Biofuels Infrastructure: Transportation Networks and Biorefinery
Locations in Illinois." In Handbook of Bioenergy Economics and Policy, 151-73,
2010.
Kerwin, Kristin. "Final Environmental Impact Statement for the Proposed Abengoa
Biorefinery Project near Hugoton, Stevens County, Kansas (Doe/Eis-0407)."
http://www.biorefineryprojecteis-abengoa.com/Final_EIS.html.
Khanna, Madhu, Xiaoguang Chen, Haixiao Huang, and Hayri Onal. "Supply of Cellulosic
Biofuel Feedstocks and Regional Production Patterns." Paper presented at the
Invited Session at the Agricultural & Applied Economics Association 2010 AAEA,
CAES, & WAEA Joint Annual Meeting, July 25-27, 2010, Denver, Colorado, 2010.
Khanna, Madhu, Basanta Dhungana, and John Clifton-Brown. "Costs of Producing
Miscanthus and Switchgrass for Bioenergy in Illinois." Biomass and Bioenergy 32,
no. 6 (2008): 482-93.

172

Kim, Seungdo, Bruce Dale, and Robin Jenkins. "Life Cycle Assessment of Corn Grain and
Corn Stover in the United States." The International Journal of Life Cycle
Assessment 14, no. 2 (2009): 160-74.
Kumar, Amit, and Shahab Sokhansanj. "Switchgrass (Panicum Vigratum, L.) Delivery to a
Biorefinery Using Integrated Biomass Supply Analysis and Logistics (Ibsal)
Model." Bioresource Technology 98, no. 5 (2007): 1033-44.
Kumarappan, Subbu, and Cole Gustafson. "Biomass Supply Chains for Biofuel Production
- Contracting Issues." In BIO world congress. Montreal, Canada, 2009.
Kumarappan, Subbu, and Satish Joshi. "Ghg Trading Framework for the U.S. Biofuels
Sector." Paper presented at the Transition to a Bioeconomy: Environmental and
Rural Development Impacts, Proceedings of Farm Foundation/USDA Conference,
St. Louis, Missouri, October 15-16 2008 2009a.
Mapemba, Lawrence D., Francis M. Epplin, Raymond L. Huhnke, and Charles M.
Taliaferro. "Herbaceous Plant Biomass Harvest and Delivery Cost with Harvest
Segmented by Month and Number of Harvest Machines Endogenously
Determined." Biomass and Bioenergy 32, no. 11 (2008): 1016-27.
Mapemba, Lawrence D., Francis M. Epplin, Charles M. Taliaferro, and Raymond L.
Huhnke. "Biorefinery Feedstock Production on Conservation Reserve Program
Land." Review of Agricultural Economics 29, no. 2 (2007): 227-.
McCarl, B.A, D.M. Adams, R.J. Alig, and John T. Chmelik. "Competitiveness of BiomassFueled Electrical Power Plants." Annals of Operations Research 94, no. 1 (2000):
37-55.
Mooney, Daniel F., Roland K. Roberts, Burton C. English, Donald D. Tyler, and James A.
Larson. "Switchgrass Production in Marginal Environments: A Comparative
Economic Analysis across Four West Tennessee Landscapes." In American
Agricultural Economics Association>2008 Annual Meeting, July 27-29, 2008.
Orlando, Florida, 2008.
Murtagh, Bruce A., Michael A. Saunders, and Philip E. Gill. "Minos."
http://www.gams.com/dd/docs/solvers/minos.pdf.
Perkis, David, Wallace Tyner, Paul Preckel, and Sarah Brechbill. "Spatial Optimization and
Economies of Scale for Cellulose to Ethanol Facilities in Indiana." Paper presented
at the Risk, Infrastructure and Industry Evolution - Proceedings of a conference
June 24-25, 2008, Berkeley, California, 2008.
Perrin, Richard, Kenneth Vogel, Marty Schmer, and Rob Mitchell. "Farm-Scale Production
Cost of Switchgrass for Biomass." Bioenergy Research 1, no. 1 (2008): 91-97.
Poet Biomass. "Poet Announces New Division Devoted to the Biofuels Feedstocks of the
Future." http://www.poet.com/news/showRelease.asp?id=170.
173

Robb, Tom. "Biomass Derived Energy Southwest Ks Project."
www.kcc.ks.gov/energy/kreeec/presentations/A2_Tom_Robb.ppt.
Sokhansanj, S , S Mani, A Turhollow, A Kumar, D Bransby, L Lynd, and M Laser.
"Large Scale Production, Harvest and Transport of Switchgrass (Panicum
Vigratum, L.) - Current Technology and Visioning a Mature Technology." Biofuels,
Bioproducts and Biorefining 3, no. 2 (2009): 124-41.
Song, Feng, Jinhua Zhao, and Scott Swinton. "Switching to Perennial Biofuel Crops under
Uncertainty and Irreversibility." In Agricultural and Applied Economics
Association. Milwaukee, Wisconsin, 2009.
Tiller, K. "Switchgrass Production Costs (Personal Communication)."
http://www.utbioenergy.org/TNBiofuelsInitiative/Switchgrass/.
Tursun, Umit Deniz, Seungmo Kang, Hayri Onal, Yanfeng Ouyang, and Jurgen Scheffran.
"Optimal Biorefinery Locations and Transportation Network for the Future Biofuels
Industry in Illinois." http://purl.umn.edu/53502.
UNFCCC-CDM. "Tool for the Demonstration and Assessment of Additionality."
http://cdm.unfccc.int/EB/029/eb29_repan05.pdf.
Vadas, P. A., K. H. Barnett, and D. J. Undersander. "Economics and Energy of Ethanol
Production from Alfalfa, Corn, and Switchgrass in the Upper Midwest, USA."
Bioenergy Research 1, no. 1 (2008): 44-55.
Wallace, Robert, Kelly Ibsen, Andrew McAloon, and Winnie Yee. "Feasibility Study for
Co-Locating and Integrating Ethanol Production Plants from Corn Starch and
Lignocellulosic Feedstocks." 69: USDA and US-DOE, 2005.
Wang, Chenguang, James A. Larson, Burton C. English, and Kim Jensen. "Cost Analysis
of Alternative Harvest, Storage and Transportation Methods for Delivering
Switchgrass to a Biorefinery from the Farmers’ Perspective." Paper presented at the
Southern Agricultural Economics Association 2009 Annual Meeting, , Atlanta,
Georgia, January 31-February 3, 2009 2009.
Wang, Michael, May Wu, and Hong Huo. "Life-Cycle Energy and Greenhouse Gas
Emission Impacts of Different Corn Ethanol Plant Types." Environmental Research
Letters 2 no. 024001 (2007): 13.

174