Functional-Type Modeling Approach and Data-Driven Parameterization of Methane Emissions in Wetlands

Authors

Gil Bohrer1* (bohrer.17@osu.edu), Theresia Yazbeck1, Madeline Scyphers1, Justine Missik1, Joel Paulson1, Yvette Onyango1, William Riley2, Qing Zhu2, Jorge Villa3, Robert Bordelon3, Diana Taj3, Eric Ward4, Kelly Wrighton5

Institutions

1The Ohio State University, Columbus, OH; 2Lawrence Berkeley National Laboratory, Berkeley, CA; 3University of Louisiana, Lafayette, LA; 4Wetland and Aquatic Research Center, U.S. Geological Survey, Lafayette, LA; 5Colorado State University, Fort Collins, CO

Abstract

The role of natural wetland emissions in the recent sharp increase of global atmospheric methane concentrations is hotly debated. Difficulties in modeling methane emissions are driven, in part, by the spatially heterogeneous nature of wetland ecosystems. This variability in fluxes is a result of the underlying within-wetland heterogeneity, combined with the complex and interconnected below- and aboveground processes of methane production, consumption, and transport. The goal of this project is to improve the simulation of methane emissions from coastal wetlands using E3SM Land Model (ELM). Researchers will improve ELM along three spatial and conceptual axes: (1) patch resolution: expand the sub-grid surface-tile approach to represent ecohydrological wetland patch types, such as open-water, mudflats, floating vegetation, cattail marsh, swamp forest; (2) vertical resolution: provide detailed observations of belowground dissolved methane concentration gradients and utilize ELM’s patch-level vertical soil column to resolve methane production, oxidation, and transport at high vertical resolution per patch; and (3) process resolution: provide observations of process-specific parameters that will be used to improve the resolution of processes within the complex system that drive net methane fluxes.

The team has completed the development of a revised version of the methane module in ELM. The new version resolves wetlands as a land unit with the ability to resolve multiple different eco-hydrological patches within a wetland. The team allowed prescribing maximal inundation depth as a forcing. Researchers prescribed patch area distribution from classified remote sensing images. The team developed the Bayesian Optimization for Anything (BOA) package and wrapped ELM with this optimization engine. Researchers used eddy covariance observations of carbon fluxes to parameterize photosynthesis, respiration, and methane production. They prescribed observed plant aerenchyma conductivity to methane and oxygen transport and used vertical profiles of methane concentration in the soil along with chamber flux measurements to parameterize methane production. The team conducted simulations of two sites in Louisiana (US-LA2, US-LA3) where they conducted observations over at least 2 years, running with two to four patch types in each site. The revised ELM model is capable of matching carbon and methane fluxes and soil concentration profiles more accurately than previous models.