Beyond Static Fluxes: Constraining parameters of a wetland methane model in a new fully coupled CH4DAS using Satellite Concentrations and In-Situ Fluxes

Methane emissions from natural wetlands are the largest and most uncertain component in the global methane budget. Conventionally they are estimated using two main methods. Top-Down methods work by inverting atmospheric concentrations of methane into fluxes, using a chemistry transport model. This is often done by optimizing the scaling factors of the inventory flux maps, but this approach decouples the fluxes from their physics and has limited predictive capabilities. Conversely, Bottom-up methods use biogeochemical models to directly estimate the fluxes, but they are severely affected by the scarcity of in-situ flux observations to calibrate them. To bridge this gap, we present the development and validation of a new fully coupled Methane Data Assimilation System (CH4DAS). This system integrates these two techniques, and provides constraints to the bottom-up model (ie., optimizing its main parameters) from both satellite concentrations and site-level fluxes. Such integration ensures that flux estimates remain consistent with physical drivers, simultaneously addressing data scarcity and enabling predictive capability.

CH4DAS is developed within the Community Inversion Framework (Berchet et al., 2021) and couples SatWetCH4 (Bernard et al., 2025), a simple bottom-up wetland methane model with the LMDz-SACS chemistry-transport model. This system can simultaneously assimilate both satellite concentrations (GOSAT) and site-level in-situ fluxes (FLUXNET-CH4) within a variational assimilation scheme to constrain the model parameters. To address the scale challenges while simultaneously assimilating observations of different streams, we run two instances of SatWetCH4. The first, driven by global forcing is coupled with LMDz-SACS and constrained by Satellite observations. While the second instance driven by site-level forcing is constrained by in-situ fluxes. This way, the shared internal temperature sensitivity parameter Q₁₀⁰ is jointly constrained by two data streams, while site-level and regional base rate parameter K account for data-specific variability.

We mathematically validate the system using an Identical Twin Observing System Simulation Experiment (OSSE), demonstrating its capacity to constrain the control variables. Further, we apply the system to real-world data to demonstrate that the system can successfully reduce the mismatches in the prior to match the spatiotemporal gradients observed by GOSAT, enabling insights on regional CH4 budgets.