COS 210-2 - Methane emissions from global freshwater wetlands

Abstract: Wetland CH4 emissions represent ~20-30% of global CH4 emissions, and these emissions are increasing due to ongoing climate warming. Uncertainties around wetland methane contribute to an under-constrained global CH4 budget and hinder strategic targeting of emissions mitigation opportunities. Furthermore, the rising methane emissions will exacerbate the warming climate because the radiative power of CH4 is ~30 times stronger than CO2 over a 100-year time horizon. Classic approaches to estimate wetland CH4 emissions used either process-based bottom-up (BU) models that directly simulated wetland biogeochemical processes or top-down (TD) transport models that indirectly inferred wetland CH4 emissions based on atmospheric CH4 concentrations.

The ongoing synthesis efforts at FLUXNET-CH4 sites provide useful data to parameterize BU models. Existing BU modeling studies have shown some progress in capturing the observed CH4 emissions at a handful of FLUXNET-CH4 sites after careful calibration. However, BU models still suffer from large parametric uncertainty and use incomplete biogeochemical theories.

In this work, we strive to better understand global wetland CH4 emissions through two avenues: 1) assimilating FLUXNET-CH4 site data into a BU model (E3SM land model) with parameter perturbations, and 2) calibrating multiple BU models (GCP-CH4 models) with machine learning and FLUXNET-CH4 site observations. The former offers a better understanding of the calibrated processes (e.g., production, oxidation, transport), while the latter provides a more efficient and effective calibration scheme.

Overall, we found that the discrepancy among BU models that participated in the recent GCP-CH4 project could be significantly reduced by considering constraints from FLUXNET-CH4 site observations, particularly over wetland hotspots. Thus, the model-data fusion for CH4 emissions provides a more consistent and robust estimate of global wetland CH4 emissions that can be used as a prior in TD models and provides a better understanding of the spatial-temporal dynamics of global freshwater wetland CH4 emissions.