Nitrous Oxide Emissions From Cropland: A Procedure for Calibration of the Biogeochemical Model Daycent Using An Inverse Modeling Technique.

DayCent is a biogeochemical model of intermediate complexity widely used to simulate soil organic carbon (SOC) and nutrients in crop, grassland, forest and savanna ecosystems. DayCent has been used extensively to simulate the impacts of climate and land use changes on ecosystems around the world, and is currently used for the annual U.S. inventory of greenhouse gas (GHG) emissions and sinks compiled by the U.S. EPA. Although, this model has been applied to a wide range of ecosystems, it is still parameterized through a traditional “trial and error” approach and has not been calibrated using statistical inverse modeling (i.e., algorithmic parameter estimation). Lamers, M. et al. (2007) and Hunt R J et al. (2007), among others, suggest that inverse modeling often leads to superior results. The aim of this study was to establish a procedure for calibration of the DayCent model to improve estimation of GHG emissions and sinks. We coupled the DayCent model with the PEST - parameter estimation software for universal inverse modeling. The PEST software can be used for calibration through regularized inversion as well as for evaluating model sensitivity and uncertainty analysis. The DayCent model was analyzed and calibrated using two years of data from the Iowa State University Agronomy and Agricultural Engineering Research Center (Ames, IA, USA) study site. Crop year 2003 data were used for calibration and 2004 data were withheld for validation. The coupling of DayCent and PEST yielded a useful tool for biogeochemical DayCent model calibration. The inverse modeling approach improved the model performance, reducing differences by about 40% between measured data and the model outputs, but equally important, the use of sensitivity and uncertainty tools provided valuable insight into the model structure, allowing for determination of the simplest and most realistic parameter field that is compatible with and makes full use of the information available in the data. This insight is valuable for guiding model development and making the best use of models of this type for estimation of GHG emissions.