A Spatial-Dynamic Model of Bioenergy Crop Introduction in Illinois

1 A Spatial-Dynamic Model of Bioenergy Crop Introduction in Illinois Jürgen Scheffran Adjunct Associate Professor Program in Arms Control, Disarmament and International Security (ACDIS) and Center for Advanced Bioenergy Research (CABER) University of Illinois at Urbana-Champaign, Champaign, IL 61820 scheffra@uiuc.edu Todd BenDor* Assistant Professor Department of City and Regional Planning University of North Carolina, Chapel Hill, NC 27599 bendor@unc.edu Yun Wang Ph.D. Candidate Department of Geography University of Illinois at Urbana-Champaign, Urbana, IL 61801 yunwang@uiuc.edu Bruce Hannon Professor Department of Geography University of Illinois at Urbana-Champaign, Urbana, IL 61801 yunwang@uiuc.edu The 25th International Conference of the System Dynamics Society Boston, MA July 29 – August 2, 2007 *Corresponding author: Department of City and Regional Planning, University of North Carolina at Chapel Hill, New East Building CB #3140, Chapel Hill, NC 27599. Email: bendor@unc.edu, Ph: 919-962-4760, fax: 919-962- 5206 2 Abstract Growing concern about climate change and energy security has led to increasing interest in developing renewable, domestic energy sources for meeting electricity, heating and fuel needs in the United States. Illinois has significant potential to produce bioenergy crops, including corn, soybeans, miscanthus (Miscanthus giganteus), and switchgrass (Panicum virgatum). However, land requirements for bioenergy crops place them in competition with more traditional agricultural uses, in particular food production. Additionally, environmental and economic conditions, including soil quality, climate, and variable agricultural costs, vary significantly across Illinois. The intent of this study is to examine the spatial and economic conditions necessary for introducing bioenergy crops into the Illinois landscape. In this paper, we develop a spatial dynamic model to explore the process by which individual farmer agents optimize profits through crop selection and cost minimization. This dynamic agent-based modeling approach will allow us to determine the optimal spatial arrangement of crops throughout Illinois as it is influenced by several factors, including the use of subsidies, changes in travel costs and crop demand, and the introduction of new ethanol production plants. This article discusses model development and specification, and outlines future calibration procedures and scenario tests that will be formalized in future work. Keywords: Land use change, bioenergy crops, renewable energy, spatial dynamic modeling, geographic information systems (GIS), agent-based modeling Administrator 文本下划线工具 Administrator 文本高亮工具 3 Introduction Biomass can contribute to a variety of energy uses, including electricity production through incineration and refinement into biogas biofuels, including ethanol and biodiesel (Rosillo-Calle et al. 2006). On the global level, about 79% of all renewable energy is generated from biomass, corresponding to 10.4% of global energy use (in comparison, nuclear power provides 6.5%; The Economist 2007). Bioenergy is intended to be climate-neutral since the carbon emitted during energy use has been initially sequestered by plants from the atmosphere. Our study examines the feasibility of introducing alternative biomass energy sources in Illinois. In particular, we examine two promising high-yield perennial grasses, miscanthus (Miscanthus giganteus) and switchgrass (Panicum virgatum), both of which are expected to play major roles as energy crops in the Midwestern United States. In order to examine patterns of land allocation among competing agricultural uses, we create and implement a spatially extended agent-based model that simulates the decisions of individual farmer agents throughout Illinois. Here, we intend to simulate the cultivation of four crops: miscanthus and switchgrass, which are harvested exclusively for bioenergy production, and corn and soybeans, which are used for both traditional and bioenergy purposes. To provide a tool for decision-making in multi-actor environments, the project builds on an agent-based modeling approach developed for applications in economic and environmental management (Scheffran 2000; Scheffran and Pickl 2000; Ipsen et al. 2001; Scheffran 2002; Billari et al. 2006; Scheffran and Leimbach 2006). Additionally, this agent-based perspective on markets for biomass crops can simulate the boom and bust associated with changing agricultural prices and the over- or under-supply that often ensues. We begin with a discussion of the bioenergy potential in Illinois and background information on the proposed bioenergy crops. Next, we discuss our data collection and processing efforts, followed by the creation, implementation, and testing of the farmer agent model. Future work will outline a series of policy tests, and will include a section detailing our conclusions and the implications of this research. Bioenergy in Illinois A study by Bournakis et al. (2005) analyzed economic impacts of 1% annual increases in the fraction of electricity generated from renewable resources, reaching at least 8% in 2012 and 16% in 2020. This study concluded that meeting these targets by 2020 would require construction of renewable energy facilities capable of delivering about 12.5 Terawatt hours (TWh) in 2012 and about 28 TWh in 2020. While these figures are quite high, Bournakis et al. (2005) noted that Illinois has considerable wind energy, biomass, and biowaste resources that could potentially be used to meet these targets. Brower et al. (1993) has also concluded that “homegrown biomass energy could create jobs in Illinois, keep energy dollars in state, reduce air pollution and soil erosion, and provide many other environmental benefits, all at competitive costs.” Bioenergy crops also have the potential to not only displace coal in power plants and thereby reduce carbon emissions, but they also have a significant potential to sequester carbon in the soil in Illinois (Dhungana 2007). One popular method of bioenergy usage is the conversion of biomass into ethanol, an option that carries considerable economic and political weight within Illinois and is likely to experience rapid future growth. By the end of 2006, there were 110 ethanol refineries in operation and 73 under construction in the U.S. (Renewable Fuels 4 Association 2006). By end of 2008, the ethanol production capacity will be an estimated 42 billion liters per year and by 2030, it is estimated that ethanol may replace 30% of the current total petroleum consumption (Perlack et al. 2005). Energy crops comprise a variety of perennial grasses and trees and are often produced using conventional agricultural practices. Perennial crops have the potential to improve environmental quality due to lower fertilizer requirements than corn and soybeans farmed under traditional practices. Additionally, extensive perennial root systems and winter harvest may also improve water quality, decrease soil erosion and increase soil organic matter. Here, we discuss two potential bioenergy sources that have been proposed to augment the use of corn in ethanol production, switchgrass and miscanthus. Switchgrass (Panicum virgatum), also known as tall panic grass, Wobsqua grass, wild redtop, or thatchgrass, is a warm season grass that has historically been a dominant species of the central North American prairie. Switchgrass was determined to be a strong candidate crop for bioenergy production based on its resilience in poor soil and climate conditions, rapid growth characteristics and low fertilization and herbicide requirements (McLaughlin and Kszos 2005). According to a recent review, biomass productivity of switchgrass ranges from 9.9-23.0 t ha-1 in research trials, with an average of 13.4 t ha-1. Several studies (McLaughlin and Kszos 2005; Perlack et al. 2005) assume that the rapid increases in switchgrass yields will continue, with innovative breeding efforts generating 20 t ha-1 switchgrass yields, an increase of 60 percent, by 2030. Miscanthus (Miscanthus giganteus) is a perennial grass from East Asia that is genetically similar to sugar cane. The crop can photosynthesize well at low temperatures and attain high yields with low amounts of nitrogen input. Like switchgrass, miscanthus has been shown to be effective at carbon sequestration and soil quality improvement. Its utility for energy production has been explored in extensive test trials (Heaton et al. 2006) which indicate harvestable miscanthus yields range from 10-40 t ha-1 throughout Europe. In 2004 and 2005 miscanthus tests trials in Illinois, dry matter per unit area was significantly greater than for switchgrass. Peak dry biomass production of Miscanthus was highest in central Illinois (60.8 t ha-1 average), and decreasing to an average of 48.5 t ha-1 in southern Illinois, and 38.1 t ha-1 average in northern Illinois (Heaton et al. 2006). Earlier trials with miscanthus demonstrated little nitrogen contribution to runoff water and an overall decrease in water use (Beale and Long 1997; Beale et al. 1999). The considerable variation in miscanthus yields is largely due to Illinois’ North-South orientation, which leads to high levels of heterogeneity in soil quality, climatic conditions, and precipitation. For example, high soil temperatures and soil moisture coupled with few frost days, make southern and central Illinois generally more suitable to biomass crop production than northern Illinois (Heaton et al. 2006). Although it may be environmentally suitable, the attractiveness of biomass crops may be lower in central Illinois since the region produces corn and soybean yields that are much higher than in southern Illinois. This leads to land competition within the region. Additionally, the cost of transporting biomass from production regions to local power or ethanol plants may be significant and needs to be considered. Therefore, the production of biomass is more likely to be profitable (and therefore successful) in areas closer to the demand centers. The profitability of cultivating specific crops varies significantly throughout space. Given this spatial heterogeneity, any study examining the viability of bioenergy sources must recognize that choosing among the alternatives is not an “either/or” alternative, but rather a 5 matter of finding a mix of biomass crops that can be successfully harvested in a given area. Finding this mix means determining the spatial pattern of land that should be allocated to traditional agriculture and to biomass crops. Our study examines the feasibility and dynamics of introducing alternative biomass energy sources in Illinois using an agent-based, spatially explicit, model of farmer decisions to produce and harvest bioenergy crops. Using geographic information systems (GIS) data on crop yields, agricultural land availability, and agricultural costs, we simulate the profitability of farmers based on their selected mix of crops. This mix generates revenue based on crop prices (determined by the relative supply from all other farmers simulated in the model), as well as costs associated with cultivating certain crops within different regions in Illinois. Farmers can then optimize their profit potential by changing their crop mix on a yearly basis in order to take advantage of more profitable crops. This model allows us to identify areas where it may be profitable to switch from conventional agriculture to bioenergy crop production. We focus our attention on the potential for cultivating miscanthus, as previous work has shown that it can be grown productively throughout Illinois and that it would be cost-effective to transport miscanthus yields to local processing plants. Our goal is to assess the conditions under which this may be the case, including an analysis of the role of market price and critical transportation distance to the next power or ethanol plant. We begin by discussing the data used to inform and parameterize our model. Data Collection and Processing Geographically referenced agricultural data were collected for the State of Illinois using a variety of sources. We begin by selecting an analysis unit size that facilitates the simulation of farmer behavior while maintaining computational tractability. Ideally, this would involve the collection of high resolution geographic data on individually controlled farms. However, data delineating farms or farm ownership is not available for the entire state of Illinois. Given the resolution of several of our datasets, we select a unit of analysis corresponding to the size of one township. A township is a land unit originally created by the Public Land Survey System under Thomas Jefferson and is commonly established as a 6 x 6 mile area.1 Although this unit size (which becomes our model cell size as described in the next section) is somewhat large, it stands as an important starting point for modeling farmer behavior. Agricultural Land Use Data We began by generating a land use map for the State of Illinois using U.S. Department of Agriculture (USDA) National Agricultural Statistics Service (NASS) Cropland Data Layer. Like most land use and land cover information, this data is collected annually using satellite imagery, specifically from the Thematic Mapper instrument on Landsat 5 and the Enhanced Thematic Mapper on Landsat 7 (Jensen 2000). The layer is aggregated to 13 standardized categories with an emphasis on agricultural land cover. Classification decisions are based on extensive field observations collected during the annual NASS June Agricultural Survey (NASS 2006). NASS uses broad land use categories to define land that is not under cultivation, such as non- 1 Townships can be divided into 1 by 1 mile sections, which can be further subdivided into quarter sections and quarter-quarter sections. This is commonly the basis of legal definitions of land delineation throughout the Midwest and parts of the Western United States. 6 agricultural, pasture/rangeland, waste, wooded, and farmstead lands. Here, the classification accuracy has been found to be approximately 85% to 95% correct for agricultural-related land cover categories. The reference list of the categorization codes and land covers for IL is shown in Table 1. Although this data contains accurate information on land cover, no data identifying the land holdings of individual farmers is reported or derivable from this data layer. Table 1: Land classification of USDA-NASS Cropland Data Layer for Illinois Classification Code Land Cover Potential Land for Energy Crops 0 No Data 1 Corn X 4 Sorghum X 5 Soybeans X 24 Winter Wheat X 25 Other Small Grains & Hay X 26 Double-Cropped Win Wheat/Soybean X 28 Oats X 36 Alfalfa X 43 Potatoes X 44 Other Crops X 54 State Code 564, Other Crops X 61 Idle Cropland/Fallow X 62 Pasture/range/ Non Agriculture 63 Woodland 81 Clouds 82 Urban 83 Water 87 Wetlands 88 Grassland Due to the discrepancy of the ground resolution of this base agricultural land use map (30 by 30 meters) and the resolution selected for this project (6 by 6 miles), the land cells on the base map were aggregated using the ESRI ArcToolbox GIS software (ESRI 2006). Here, the fraction of the agricultural land that can be used for growing bioenergy crops (marked with ‘x’ in Table 1) is estimated for each 36 square mile aggregate cell. Figure 1 reveals the heavy row crop agriculture in Central Illinois. 7 Figure 1: Agricultural Land in Illinois Illinois Crop Yield Data Soybean and corn production data were taken from the Illinois Crop Yields Historical NASS Database, which lists wheat, corn, and soybean yields for each county in Illinois between 1972 and 2004 (Sherrick 2005). A geographical distribution of yields was taken by taking a five year average of yields during a representative period between 1997 and 2001. This data was converted from bushels per acre measurements to metric tonnes per hectare using conversion constants from the Canada Grains Council 1999 Statistical Handbook (Canada Grains Council 1999). Since this data was collected at the County level, which is a lower resolution than our analysis, we performed a GIS spatial interpolation through a geostatistical analysis known as Kriging (spherical model; O'Sullivan and Unwin 2003). This technique interpolates the value of a variable at unobserved locations from values at nearby known locations, thereby allowing us to create a fairly accurate and continuous map of soybean and corn yields throughout Illinois at the township resolution (Figure 2). 8 Figure 2: Illinois Corn and Soybean Yields (1997-2001 Avg.) Soybean raw county data Soybean spatially interpolated data Corn raw county level data Corn spatially interpolated data Given that switchgrass and miscanthus have not been planted extensively in Illinois (or elsewhere in the United States, for that matter), information on their growth patterns and harvesting costs is still relatively sparse. Potential miscanthus and switchgrass yield estimates, based on soil quality, climate, and other environmental conditions, were obtained from recent work by Khanna et al. (2005), and were aggregated to the township resolution. Likewise, estimates of the harvested biomass fraction actually taken off the field (rather than left to 9 enhance soil quality) are 67% for miscanthus, 80% for switchgrass, and 98% for both corn and soybeans. Figure 3: Miscanthus and Switchgrass Yields Original miscanthus/switchgrass data Aggregated miscanthus/switchgrass data Source: Khanna et al. (2005) Harvest Production and Cost Data The value of production of corn and soybeans is estimated from the Illinois Crop Yields Historical NASS Database (NASS 2007), from which the average ‘value of production,’ the average amount of soy and corn sold multiplied by the average selling price, was calculated for 2000 through 2006. During this period, Illinois produced an average $2.604 billion worth of soybeans $4.072 billion worth of corn. Harvest cost data for corn and soybeans were collected from the Illinois Farm Business Farm Management Association through the University of Illinois Farm Decision Outreach Council (FARMDOC 2007), which maintains cost records for corn and soybean back to 2001. These costs were divided into direct (fertilizer, pesticides, seed, storage, drying, crop insurance), power (machine use/lease/depreciation, utilities, fuel), and overhead (labor, building repair/rent/depreciation, insurance) costs. Here, harvest costs within Illinois vary by region, with the data being separated into northern, central (high and low productivity), and southern regions within Illinois (Figure 4). A six year average for each region was taken using 2001-2006 cost data, while high and low productivity areas were averaged in central Illinois. 10 Figure 4: Illinois Harvest Cost Regions Source: FARMDOC (2007) Miscanthus and switchgrass costs were estimated from the Illinois Interactive Agronomy Handbook (Hoeft and Nafziger 2006) based on similar, well established crops and were broken into capital, labor, and material cost categories. This data collection and adjustment process yields four GIS maps showing expected yields for miscanthus, corn, switchgrass, and soybean crops throughout the State of Illinois. These maps are then used as geographically c