A Parsimonious Modular Approach to Building a Mechanistic Belowground C and N Model

Monday, 15 December 2014
Eric A Davidson1, Rose Zheng Abramoff2 and Adrien Finzi2, (1)Woods Hole Research Center, Falmouth, MA, United States, (2)Boston University, Boston, MA, United States
Ecosystem models simulate belowground processes of element cycling, such as soil carbon dynamics, through numerical representations of our mechanistic understanding of these processes. This poses a great challenge, because many mechanistic models are so highly parameterized and complex that the reasons for their behavior are often difficult to understand, while others are so simplistic that they are unlikely to capture potentially important complexities, such as C-N interactions. Here we propose a mechanistic, but parsimonious modular approach to building a belowground C and N model, with a core processes-level representation of microbial and exoenzymatic activity. We have merged the Dual-Arrhenius Michaelis Menten (DAMM) model of Davidson et al. (2012) with the Microbial Carbon and Nitrogen Physiology (MCNiP) model of Drake et al. (2013). DAMM explicitly simulates the effects of temperature, soil moisture and substrate supply on the kinetics of soil organic matter (SOM) depolymerization. MCNiP is a stoichiometrically constrained model of microbial maintenance and growth respiration, exoenzyme production, and C and N uptake and mineralization. Here we present the results of this model merger, focusing on how temperature, soil moisture, substrate limitation, and the stoichiometry of microbial processes individually and interactively regulate gains or losses of C and N from SOM at the Harvard Forest. We compare model performance running DAMM-MCNiP with and without C-N linkages. We find that substrate limitation is a major constraint over SOM decomposition and that N availability can be a strong regulator of decomposition. The C-only formulation of DAMM-MCNiP generally results in greater losses of SOC than the coupled C-N formulation. Future data-model assimilation and parameter estimation will help determine which aspects of the merged model can be well constrained and which aspects will need further refinement through new data collection or meta-analysis.