Importance of vegetation classes in modeling CH4 emissions from boreal and subarctic wetlands in Finland.

ABSTRACT

Boreal/arctic wetlands are dominated by diverse plant species, which vary in their contribution to CH4 production, oxidation and transport processes. Earlier studies have often lumped the processes all together, which may induce large uncertainties into the results. We present a novel model, which includes three vegetation classes and can be used to simulate CH4 emissions from boreal and arctic treeless wetlands. The model is based on an earlier biogeophysical model, CH4MODwetland. We grouped the vegetation as graminoids, shrubs and Sphagnum and recalibrated the vegetation parameters according to their different CH4 production, oxidation and transport capacities. Then, we used eddy-covariance-based CH4 flux observations from a boreal (Siikaneva) and a subarctic fen (Lompolojänkkä) in Finland to validate the model. The results showed that the recalibrated model could generally simulate the seasonal patterns of the Finnish wetlands with different plant communities. The comparison between the simulated and measured daily CH4 fluxes resulted in a correlation coefficient (R2) of 0.82 with a slope of 1.0 and an intercept of -0.1mgm-2h-1 for the Siikaneva site (n=2249, p<0.001) and an R2 of 0.82 with a slope of 1.0 and an intercept of 0.0mgm-2h-1 for the Lompolojänkkä site (n=1826, p<0.001). Compared with the original model, the recalibrated model in this study significantly improved the model efficiency (EF), from -5.5 to 0.8 at the Siikaneva site and from -0.4 to 0.8 at the Lompolojänkkä site. The simulated annual CH4 emissions ranged from 7 to 24gm-2yr-1, which was consistent with the observations (7-22gm-2yr-1). However, there are some discrepancies between the simulated and observed daily CH4 fluxes for the Siikaneva site (RMSE=50.0%) and the Lompolojänkkä site (RMSE=47.9%). Model sensitivity analysis showed that increasing the proportion of the graminoids would significantly increase the CH4 emission levels. Our study demonstrated that the parameterization of the different vegetation processes was important in estimating long-term wetland CH4 emissions.