Process-based modelling of NH₃ exchange with grazed grasslands

In this study the GAG model, a process-based ammonia (NH₃) emission model for urine patches, was extended and applied for the field scale. The new model (GAG_field) was tested over two modelling periods, for which micrometeorological NH₃ flux data were available. Acknowledging uncertainties in the measurements, the model was able to simulate the main features of the observed fluxes. The temporal evolution of the simulated NH₃ exchange flux was found to be dominated by NH₃ emission from the urine patches, offset by simultaneous NH₃ deposition to areas of the field not affected by urine. The simulations show how NH₃ fluxes over a grazed field in a given day can be affected by urine patches deposited several days earlier, linked to the interaction of volatilization processes with soil pH dynamics. Sensitivity analysis showed that GAG_field was more sensitive to soil buffering capacity (β), field capacity (θ_fc) and permanent wilting point (θ_pwp) than the patch-scale model. The reason for these different sensitivities is dual. Firstly, the difference originates from the different scales. Secondly, the difference can be explained by the different initial soil pH and physical properties, which determine the maximum volume of urine that can be stored in the NH₃ source layer. It was found that in the case of urine patches with a higher initial soil pH and higher initial soil water content, the sensitivity of NH₃ exchange to β was stronger. Also, in the case of a higher initial soil water content, NH₃ exchange was more sensitive to the changes in θ_fc and θ_pwp. The sensitivity analysis showed that the nitrogen content of urine (c_N) is associated with high uncertainty in the simulated fluxes. However, model experiments based on c_N values randomized from an estimated statistical distribution indicated that this uncertainty is considerably smaller in practice.

Finally, GAG_field was tested with a constant soil pH of 7.5. The variation of NH₃ fluxes simulated in this way showed a good agreement with those from the simulations with the original approach, accounting for a dynamically changing soil pH. These results suggest a way for model simplification when GAG_field is applied later at regional scale.