Nitrogen cycling under urine patches: Model comparison and sensitivity analysis

Abstract

Agricultural greenhouse gas (GHG) emissions, including nitrous oxide (N2O), are a major contributor to New Zealand's GHG emissions, and animal excreta deposited onto pastures are a main source of these emissions. Reducing these N2O emissions requires a better understanding of the factors driving emissions and evaluation of mitigation strategies. Computer simulation models can provide an effective tool for these tasks. We compared two process-based models, APSIM and NZ-DNDC, on results of three experiments on N2O emissions following urine application to the Horotiu soil in the Waikato region of New Zealand in various seasons and years. Soil ammonium and nitrate concentrations were also determined in the experiments. With default parameter settings, both models predicted the daily pattern of N2O emissions poorly (Figure 1) with negative model efficiencies. The sensitivity of various model parameters was examined: For APSIM these included the nitrification rate, the optimum soil temperature on nitrification, the fraction of nitrified N emitted as N2O, the denitrification rate, and the rainfall intensity. For DNDC they were microbial activity, nitrification rate, denitrification rate, plant growth, ammonia volatilisation, rainfall intensity, and fraction of N2O produced during nitrification were varied. Changing some of the default model parameters improved the model agreement in some cases; e.g. for APSIM when the fraction of nitrified nitrogen emitted as N2O was increased or the optimum temperature for nitrification was decreased, and for DNDC when microbial activity was decreased or volatilization increased. However, none of the parameters investigated could improve predicted emissions so that they agreed reasonably with all three datasets. A sensitivity analysis which includes more parameters and model functions, as well as changing various parameters simultaneously is needed.