Comparison of models for predicting nitrification, denitrification and nitrous oxide emissions in pastoral systems

Intensification of agricultural systems has resulted in remarkable increases in productivity. However in grazed systems only about 10-20% of the N ingested by grazing animals is retained in animal products. The reminder is returned in excreta to the paddock in a spatially non-uniform fashion as dung and urine and then either accumulates in the soil or is lost via leaching or gaseous emissions. The amount of N lost from these patches via leaching and gaseous emissions is highly dependent on site-specific factors. Routine direct measurements are impractical given the scale and variability so simulation models are an essential alternative to estimate likely N losses. Various simulation approaches are in use or are being developed to predict N leaching and N2O emissions. The models vary in the level of detail or number of nitrogen pools and transformation processes considered, as well as on how the processes are described. Other processes within the models, such as water and heat transport within the soil also affect the modelled N transformations and losses. While most models have been tested and validated for some processes or outputs in a particular range of systems, soils and climates, there is a lack of information on how models compare in other aspects. We discuss key differences and similarities between the N transformation components of the APSIM and DNDC models with respect to their simulation of nitrification, denitrification and N2O emissions under pastoral systems. Agreement between daily N2O emissions from urine patches simulated by APSIM and DNDC was variable, Figure 1. For the Horotiu soil the index of agreement (IA) between measurements and APSIM or DNDC were 0.4 and 0.13, and for the Templeton soil 0.77 and 0.47. Agreement between total emissions over the experimental period was better for the APSIM model than the DNDC, based on default model parameters. For the Horotiu soil, N2O emissions over 3 months were 4.9 kg/ha, and simulations with APSIM and DNDC gave values of 7.4 and 13.7 kg/ha. For the Templeton soil over 1 year measurements were 2.8 kg/ha and simulations with APSIM 1.9 kg/ha and with DNDC 16.3 kg/ha. Adjusting default values to NZ conditions can, however, improve the prediction capacities of the models.