Constraining North American Wetland Methane Emissions: Evaluation of Existing CH4 Wetland Models and a Novel Global Wetland, Lake, and Reservoir Emissions Dataset Using a Top-Down Approach

Abstract:

Global wetlands are the largest natural emitters of methane (CH₄) and are coupled to global climate by important feedback processes. Thus, understanding the main drivers controlling wetland CH₄ fluxes has important implications for their role in the atmospheric CH₄ budget and global warming. Bottom-up models, which simulate emissions, can be useful tools in understanding how and what regional emission sources control atmospheric CH₄ concentrations. However, previous studies attempting to simulate wetland CH₄ emissions are poorly constrained, largely due to differences in wetland representation between models, inhibiting robust conclusions on the wetland contribution to atmospheric CH₄. Conversely, top-down approaches allow us to quantify the ability of bottom-up models to explain atmospheric CH₄ concentrations, evaluate flux estimate accuracy, and identify important environmental factors contributing to CH₄ emissions.

In this study, we evaluate differences in spatiotemporal trends between 6 bottom-up wetland models - 5 models from the Wetland and Wetland CH₄ Inter-comparison of Models Project (WETCHIMP) and a newer data product, WetCHARTs v1.0 - and a novel speciated wetland, lake, and reservoir (WLR) dataset. We find that all models predict a similar seasonality, although with different amplitudes, with large CH₄ fluxes during the summer that diminish by winter, especially in the northern latitudes. Using WRF-STILT sensitivity footprints generated using airborne and NOAA tower atmospheric CH₄ concentration measurements, we then compute bottom-up model estimates of atmospheric CH₄ concentrations. Comparing these estimates to measurements, we find that all models overestimate summer CH₄ concentrations, indicating that summer wetland emissions in these regions are largely overestimated. Compared to the 6 bottom-up models, the novel WLR dataset performs similarly, although it appears to fall on the lower end of overestimates. These findings are promising for the improvement of future wetland models.