Temporal Variation of Ecosystem Scale Methane Emission From a Boreal Fen in Relation to Temperature, Water Table Position, and Carbon Dioxide Fluxes

We have analyzed decade-long methane flux data set from a boreal fen, Siikaneva, together with data on environmental parameters and carbon dioxide exchange. The methane flux showed seasonal cycle but no systematic diel cycle. The highest fluxes were observed in July-August with average value of 73 nmol m(-2) s(-1). Wintertime fluxes were small but positive, with January-March average of 6.7 nmol m(-2) s(-1). Daily average methane emission correlated best with peat temperatures at 20-35 cm depths. The second highest correlation was with gross primary production (GPP). The best correspondence between emission algorithm and measured fluxes was found for a variable-slope generalized linear model (r(2) = 0.89) with peat temperature at 35 cm depth and GPP as explanatory variables, slopes varying between years. The homogeneity of slope approach indicated that seasonal variation explained 79% of the sum of squares variation of daily average methane emission, the interannual variation in explanatory factors 7.0%, functional change 5.3%, and random variation 9.1%. Significant correlation between interannual variability of growing season methane emission and that of GPP indicates that on interannual time scales GPP controls methane emission variability, crucially for development of process-based methane emission models. Annual methane emission ranged from 6.0 to 14 gC m(-2) and was 2.7 +/- 0.4% of annual GPP. Over 10-year period methane emission was 18% of net ecosystem exchange as carbon. The weak relation of methane emission to water table position indicates that space-to-time analogy, used to extrapolate spatial chamber data in time, may not be applicable in seasonal time scales.