English Abstract The Kalman filter is used as a data-assimilation
technique to calculate methane concentrations and emissions in Europe. The
filter combines the results of model calculations and measurements to
achieve an optimal estimate using knowledge of the system or model noise and
the measuring noise. The results of the Eulerian 3 D Euros model, which
calculates emissions, transport and concentrations of methane all over in
Europe, and the results of five methane measuring stations in the
Netherlands and one in Ireland are used simultaneously in the filter. The
filter is very time consuming on a computer and therefore Kriging and a
Reduced Rank Square Root (RRSQRT) algorithm are implemented to reduce
calculation time. The results show that it is very important to carefully
select the optimal parameters in the model: the system and measurement
noise, the Kriging interpolation factors and the number of significant
Eigenvalues in the RRSQRT algorithm to get reliable outcomes. Systematic
differences between measurements and model calculations are caused by errors
in the emission input. A fixed point smoother has been implemented - in
combination with the filter - to be able to get an optimal estimate of
methane emissions in Europe. The working of the smoother and the influence
of the smoother parameters on the model results have been
analysed.