Heterogeneity in freshwater wetland inundation zones drive carbon fluxes: A process-based modeling approach

Heterogeneity in wetland hydrology and vegetation influences biogeochemical processes, and since wetlands store significant global organic carbon and contribute substantially to CH₄ emissions, accurately modeling these dynamics over space and time is critical. This study used the process-based model ecosys to simulate ecosystem carbon gas fluxes (CO₂ and CH₄) in two hydrologically distinct areas of a temperate freshwater marsh in Ohio, USA: a deeper main wetland pool and a shallower cove. Our aim was to improve flux predictions by modeling the wetland using multiple grid cells that varied in hydrology and vegetation, representing these two zones. We evaluated the modeled carbon gas fluxes against both eddy covariance data from the main wetland pool (2015 to 2023) and chamber-measured flux data from the cove (2022 to 2023). We used the Bayesian Optimization for Anything framework to conduct four sequential optimization runs, deriving specific parameter sets for CO₂ and CH₄ processes in each zone. After optimization, ecosys adequately reproduced (R² = 0.34 and 0.58 for CO₂ and CH₄, respectively) the magnitude and timing of hourly diurnal flux patterns in the main pool. However, initially applying the main pool parameter values to the cove resulted in poor flux predictions, notably overestimating CO₂ uptake and CH₄ release. Implementing site-specific parameterization for the cove led to modeled fluxes that aligned more closely with chamber measurements, though non-vegetated cell fluxes were less variable than vegetated cell fluxes. Overall, these findings emphasize that improved modeling of carbon gas fluxes in heterogeneous wetlands requires a multi-grid cell approach and localized parameterization.