Modeling denitrification in a tile-drained, corn and soybean agroecosystem of Illinois, USA

• Published in: Biogeochemistry 2009-03, Vol.93 (1-2), p.7-30
• Main Authors: David, Mark B, Del Grosso, Stephen J, Hu, Xuetao, Marshall, Elizabeth P, McIsaac, Gregory F, Parton, William J, Tonitto, Christina, Youssef, Mohamed A
• Format: Article
• Language: English
• Subjects: Agricultural ecosystems; Agricultural soils; Air temperature; Biogeosciences; Corn; Crop yield; Denitrification; Earth and Environmental Science; Earth Sciences; Earth, ocean, space; Ecosystems; Environmental Chemistry; Exact sciences and technology; Fluctuations; Geochemistry; Greenhouse gases; Harvesting; Hydrologic models; Leaching; Life Sciences; Mineralization; Modeling; Nitrates; Nitrification; Nitrogen fixation; Nitrous oxide; Simulations; Soil and rock geochemistry; Soil moisture; Soil water; Soil water content; Soils; Soybeans; Surficial geology; Watersheds
• ISSN: 0168-2563; 1573-515X
• DOI: 10.1007/s10533-008-9273-9

Denitrification is known as an important pathway for nitrate loss in agroecosystems. It is important to estimate denitrification fluxes to close field and watershed N mass balances, determine greenhouse gas emissions (N₂O), and help constrain estimates of other major N fluxes (e.g., nitrate leaching, mineralization, nitrification). We compared predicted denitrification estimates for a typical corn and soybean agroecosystem on a tile drained Mollisol from five models (DAYCENT, SWAT, EPIC, DRAINMOD-N II and two versions of DNDC, 82a and 82h), after first calibrating each model to crop yields, water flux, and nitrate leaching. Known annual crop yields and daily flux values (water, nitrate-N) for 1993-2006 were provided, along with daily environmental variables (air temperature, precipitation) and soil characteristics. Measured denitrification fluxes were not available. Model output for 1997-2006 was then compared for a range of annual, monthly and daily fluxes. Each model was able to estimate corn and soybean yields accurately, and most did well in estimating riverine water and nitrate-N fluxes (1997-2006 mean measured nitrate-N loss 28 kg N ha⁻¹ year⁻¹, model range 21-28 kg N ha⁻¹ year⁻¹). Monthly patterns in observed riverine nitrate-N flux were generally reflected in model output (r ² values ranged from 0.51 to 0.76). Nitrogen fluxes that did not have corresponding measurements were quite variable across the models, including 10-year average denitrification estimates, ranging from 3.8 to 21 kg N ha⁻¹ year⁻¹ and substantial variability in simulated soybean N₂ fixation, N harvest, and the change in soil organic N pools. DNDC82a and DAYCENT gave comparatively low estimates of total denitrification flux (3.8 and 5.6 kg N ha⁻¹ year⁻¹, respectively) with similar patterns controlled primarily by moisture. DNDC82h predicted similar fluxes until 2003, when estimates were abruptly much greater. SWAT and DRAINMOD predicted larger denitrification fluxes (about 17-18 kg N ha⁻¹ year⁻¹) with monthly values that were similar. EPIC denitrification was intermediate between all models (11 kg N ha⁻¹ year⁻¹). Predicted daily fluxes during a high precipitation year (2002) varied considerably among models regardless of whether the models had comparable annual fluxes for the years. Some models predicted large denitrification fluxes for a few days, whereas others predicted large fluxes persisting for several weeks to months. Modeled denitrification fluxes were controlled mainly by