Predicting nitrous oxide (N2O) fluxes from agricultural soil requires simulating the primary processes of nitrification and denitrification involved in N2O production, and of gas transport in soil. While most ecosystem models represent these mechanisms, it is still highly challenging to accurately capture the dynamic of N2O fluxes. We used on-field N2O measurements with automated chambers on a highly productive grassland in Switzerland during a 6-year period (2002-2007), to assess the accuracy of the ecosys model to predict N2O flux dynamics at hourly and daily time scales as part of the NITROEUROPE project. The model, which had not been previously applied to temperate productive grassland, accurately captured the dynamics of plant growth and soil environment that were pre-requisites to estimate soil microbial processes. The timing of N2O emission peak events was well predicted and, most of the time, the intensity of the flux peaks were within the range of measurement errors. The model-data evaluation helped us in understanding the underlying processes controlling the N2O fluxes and we demonstrated that the dynamics of secondary processes, such as root nitrogen uptake and fertilizer pellet solubilization, need also to be well predicted to capture N2O fluxes. Some missing processes such as the dynamic change of soil porosity during drying/wetting are missing in ecosys and their mechanistic formulation in the model could improve the modelled N2O flux dynamic. Finally, we showed that the weak relationships between N2O emissions and soil environmental variables did not allow us to propose simple generalizations to estimate N2O emissions from our grassland site.