A Bayesian inversion estimate of N2O emissions for western and central Europe and the assessment of aggregation errors
Abstract. A Bayesian inversion approach was used to retrieve temporally and spatially resolved N2O fluxes for western and central Europe using in-situ atmospheric observations from the tall tower site at Ochsenkopf, Germany (50°01' N, 11°48' E). For atmospheric transport, the STILT (Stochastic Time-Inverted Lagrangian Transport) model was employed, which was driven with ECMWF analysis and short term forecast fields. The influence of temporal aggregation error, as well as the choice of spatial and temporal correlation scale length, on the retrieval was investigated using a synthetic dataset consisting of mixing ratios generated for the Ochsenkopf site. We found that if the aggregation error is ignored, then a significant bias error in the retrieved fluxes ensues. However, by estimating this error and projecting it into the observation space, it was possible to avoid bias errors in the retrieved fluxes. Using the real observations from the Ochsenkopf site, N2O fluxes were retrieved every 7 days for 2007 at 2 by 2 degrees spatial resolution. Emissions of N2O were strongest during the summer and autumn months, with peak emissions in August and September, while the regions of Benelux and northern United Kingdom had strongest annual mean emissions.