Partitioning and driver analysis of eddy covariance derived N2O emissions from a grazed and fertilized pasture

• Three years of N 2 O eddy covariance measurements at a grazed and fertilized pasture. • Random forest was suitable for partitioning of measured N 2 O fluxes. • The average emission factor for cattle excreta was 1.1 ± 0.5 %. • Maximum N 2 O fluxes were found after fertilizer application and under optimum soil moisture. • Additional N 2 O peaks occurred during the cold season and a drought period. Managed pastures are strong sources for the greenhouse gas nitrous oxide (N 2 O) through various nitrogen (N) inputs. So far, chamber measurements have been used to quantify N 2 O emissions and emissions factors of specific emissions sources like grazing cattle excreta. This study presents a three-year dataset of N 2 O emissions from a grazed and fertilized pasture measured by eddy covariance (EC) in eastern Switzerland. N 2 O fluxes were gap-filled and disaggregated into the emission sources (flux partitioning) by using random forest. The excreta N deposition in the pasture was estimated based on a cattle nitrogen budget approach using observed milk yield, body weight and feed intake of the cattle herd. Furthermore, a driver analysis was performed to quantify the relationship between N 2 O emissions and predictor variables. The observed annual N 2 O emissions amounted to 5.3 ± 0.8, 3.1 ± 0.5 and 4.4 ± 0.7 kg N 2 O-N ha -1 yr -1 and were disaggregated into background, fertilizer and excreta related N 2 O emissions with contributions of 27–46 %, 15–40 % and 30–51 %, respectively. Combining the excreta N 2 O fluxes with the excreta N inputs resulted in an average emission factor (EF) for cattle excreta of 1.1 ± 0.5 %, that tends to be higher than the IPCC default value of 0.6 % for wet climates. While maximum N 2 O emissions usually were observed after fertilizer application and under optimum soil moisture conditions as expected, distinct N 2 O emission peaks also occurred during a longer drought period in summer and could be parametrised as a function of precipitation and previous grazing activity. Moreover, peak N 2 O emissions occurred during the cold season at low temperatures and should be considered in future studies. Overall, we suggest that EC measurements under pasture conditions with subsequent flux partitioning by random forest are suitable for quantifying pasture N 2 O emissions of different sources.