Use of Monte Carlo Analysis to Characterize Nitrogen Fluxes in Agroecosystems SHELIE A. MILLER,* AMY E. LANDIS, AND THOMAS L. THEIS Institute for Environmental Science and Policy, University of Illinois at Chicago, 2121 West Taylor Street, Chicago, Illinois 60612 Intensive agricultural systems are largely responsible for the increase in global reactive nitrogen compounds, which are associated with significant environmental impacts. The nitrogen cycle in agricultural systems is complex and highly variable, which complicates characterization in environmental assessments. Appropriately representing nitrogen inputs into an ecosystem is essential to better understand and predict environmental impacts, such as the extent of seasonally occurring hypoxic zones. Many impacts associated with reactive nitrogen are directly related to annual nitrogen loads, and are not adequately represented by average values that de-emphasize extreme years. To capture the inherent variability in agricultural systems, this paper employs Monte Carlo analysis (MCA) to model major nitrogen exports during crop production, focusing on corn-soybean rotations within the U.S. Corn Belt. This approach yields distributions of possible emission values and is the first step in incorporating variable nutrient fluxes into life cycle assessments (LCA) and environmental impact assessments. Monte Carlo simulations generate distributions of nitrate emissions showing that 80% of values range between 15 and 90 kg NO 3 - N/ha (mean 38.5 kg NO 3 - N/ha; median 35.7 kg NO 3 - N/ha) for corn fields and 5-60 kg NO 3 - N/ha (mean 20.8 kg NO 3 - N/ha; median 16.4 kg NO 3 - N/ha) for soybean fields. Data were also generated for grain and residue nitrogen, N 2 O, NO x , and NH 3 . Results indicate model distributions are in agreement with available measured emissions. Introduction Since the early 19th century, human activities have increased the rate of conversion of nonreactive atmospheric nitrogen to reactive nitrogen eleven-fold (1). The substantial increase in the flux of reactive nitrogen contributes to several environmental problems, including global climate change, eutrophication and hypoxia, acid deposition, and production of ground-level ozone (2, 3). The amount of reactive nitrogen is projected to increase with continued population growth. Combustion of fossil fuels contributes only 15% of the anthropogenic reactive nitrogen in the environment, while food production accounts for 75% of anthropogenic nitrogen due to the manufacture of synthetic fertilizer and biological nitrogen fixation from cultivated crops such as legumes and rice (3). Significant quantities of nitrogen fertilizer are applied to crops in attempts to increase global and regional food supplies and to enhance nutritional quality of available foods (4). In addition to emissions of N2O and NOx from the combustion of fossil fuels associated with energy use from fertilizer production and on-farm processes, the agricultural industry is also responsible for NH3, NO, N2O emissions evolving from fields, and NO3 - in surface water runoff. The applied nitrogen is either acquired by crops or released to the environment through a variety of pathways. Figure 1 shows the major fluxes of the agricultural nitrogen cycle relevant to corn and soybean production. Life Cycle Assessment (LCA) is a tool used to evaluate the environmental impact of a process or product. The analysis examines the entire life cycle of the process or product, from creation to material acquisition through disposal or reuse. LCA documents all relevant mass and energy inputs and emissions through a process called life-cycle inventory (LCI) to determine the intensity of material uses and identify possible areas of improvement. LCA is usually used to compare alternatives and offer policy-makers quantitative information to inform environmentally significant decisions. With the recent rise of bio-based materials offered as alternatives to petroleum products, numerous LCIs have been conducted on agricultural systems (5-9). While these studies acknowledge that nutrient fluxes are a significant issue, they focus primarily on air emissions, and have not quantified aqueous emissions in a comprehensive manner, due to high variability of nonpoint emissions. LCA was initially developed for industrial systems, where processes are usually carefully structured and controlled, with known or measurable material and energy fluxes. Traditional LCAs often use average data to generate generic depictions of material and energy fluxes to simplify an analysis and generate a representative inventory; however, large deviations should be documented to provide a complete system description. While average data may be acceptable for industrial systems with low operational variability, such values seldom represent the inherent fluctuations of natural and agricultural systems. Concentrating solely on extreme values is equally problematic, as it can bias the results of the analysis, and not adequately depict the impacts for the majority of years. Agricultural processes differ from many industrialized processes because of inconsistent material fluxes due to nonpoint emissions, uncertain input variables (e.g., nitrogen fixation and soil mineralization), the temporal scale needed to produce a product, the interdependence of crops in rotation, and the high degree of system variability which depends on geography, weather patterns, soil type, and agricultural practices. If average data are used to characterize agricultural systems, “extreme” data, such as wet or drought years, may not be captured. Average data may be appropriate for use in situations where aggregate emissions cause long- term environmental impacts (e.g., carbon dioxide emissions leading to global climate change); however, impacts associ- ated with nitrogen emissions are temporally relevant. For example, the size of a hypoxic zone for a given year will vary proportionately with nitrate emissions into the watershed that year (10, 11). These complexities underscore the importance of conducting a comprehensive and reproducible LCI for agricultural systems. To reduce variability among parameters when examining agricultural systems, researchers often prefer to limit the analysis to a relatively small geographic region, assuming that the relative uniformity of system variables, such as climate and soil type, will allow greater precision in inventory * Corresponding author phone: 312-996-1081; fax: 312-355-0760; e-mail: smille24@uic.edu. Environ. Sci. Technol. 2006, 40, 2324-2332 2324 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 7, 2006 10.1021/es0518878 CCC: $33.50  2006 American Chemical Society Published on Web 02/23/2006