Partitioning methane flux by the eddy covariance method in a cool temperate bog based on a Bayesian framework

The responses of CH4 fluxes to environmental drivers are known to be complex in wetlands and are not easily interpreted due to their nonlinear nature. To better understand the observed CH4 flux, we developed a method to partition this flux into CH4 production, oxidation, and three transport pathways. Based on a Bayesian method with six-year eddy covariance measurements from a cool temperate bog in northern Japan, we estimated the parameters of a simple two-layer model, which considered the processes in surface oxic and deep anoxic layers. The constrained model explained 87% of the variation in the observed CH4 flux at the daily to seasonal timescales. The model estimated that 64% of CH4 was transported by ebullition compared with 36% by plantmediated transport during snow-free periods. The model predicted that CH4 was mostly emitted from the deep anoxic layer rather than from the surface layer. The model explained 56% of the interannual variations in the annual CH4 flux, which was mostly controlled by CH4 production. Posterior distributions of the parameters depended on the data coverage that constrained the model, strongly indicating that long-term data are indispensable for constraining process models. Even when using the six-year data, all parameters were not well constrained probably because the data did not contain enough information to constrain the processes. Thus, the method must be tested in various wetlands with additional long-term data to evaluate its applicability and limitations.