Wetland biogeochemical dynamics and resulting CO2 and CH4 emissions to the atmosphere depend on many biotic and abiotic processes. Mechanistically representing this complexity in models is very difficult, resulting in a wide range of approaches. For example, most regional- to global-scale models implicitly represent microbial processes, rather than explicitly representing microbial biomass or activity. Explicit representations of wetland plant processes and their interactions with soil biogeochemistry are also very simplified. While it remains to be demonstrated that more explicit representations of microbial and plant processes improve predictability at large scales, it is clear that a more mechanistic approach (a) can improve analysis of critical processes identified at the site level, (b) can allow use of new types of data (e.g., genomics) to explore underlying dynamics, and (c) may be more capable of accurately representing dynamics outside of the current climate training space.
In this talk I will discuss evidence for the need to more explicitly represent microbial and plant processes in wetland biogeochemical models using recent work from studies in Stordalen, Sweden and Utqiaġvik, Alaska. Using the ecosys model, we explain the observed hysteretic dependence of wetland CH4 emissions on temperature at these two sites based on acetoclastic and hydrogenotrophic methanogen activities driven by seasonality in substrate production and availability. Our results demonstrate that plant processes strongly affect the modeled soil thermal and hydrological conditions thereby altering biogeochemical responses under the same climate forcing. Using Morris sensitivity analysis, we identify the relative importance of various microbial functional groups to monthly CH4 emissions and identify needed measurements for model improvement. We then show, using the FLUXNET-CH4 eddy covariance dataset, that similar hysteretic patterns are present globally and consistent with the substrate availability hypothesis. Accounting for CH4 temperature hysteresis about halved the CH4 emission prediction bias across global wetland and rice paddy sites. Finally, evaluating the level of model complexity needed for accurate large-scale wetland CH4 emission estimates can be facilitated with benchmarking tools like the International Land Model Benchmarking (ILAMB) System. I will describe that tool and discuss results evaluating current global CH4 models against the FLUXNET-CH4 dataset.