Methane emissions from biogenic sources, e.g. Arctic permafrost ecosystems, are associated with uncertainties due to the high variability of fluxes in both space and time. Besides the traditional eddy covariance method, we evaluated a method based on wavelet analysis, which does not require a stationary time series, to calculate fluxes. The occurrence of extreme methane flux events was strongly correlated with the soil temperature. They were triggered by atmospheric non-turbulent mixing.