Re-bugging biogeochemical models: Parsimoniously representing microbial enzymatic processes in models of soil organic matter decomposition

Background/Question/Methods

Black boxes are out of favor, but highly parameterized models that attempt to represent details of complex processes with poorly constrained parameter estimates can also lead us astray.  We know that the soil is a complex and heterogeneous matrix, populated by a wide diversity of organisms that produce similar and contrasting enzymes.  There is good evidence that microbial community composition varies temporally and spatially, and the reactive properties of extracellular enzymes also can probably be changed by microorganisms in response to environmental cues.  We know that enzymatic reaction rates are temperature sensitive when substrate is not limiting, and there is evidence that the affinities of enzymes for substrates are often temperature sensitive.  Because the soil is heterogeneous, substrate supply often, perhaps usually, limits enzymatic reaction rates.  The carbon, nitrogen, and phosphorus assimilation enabled by extracellular enzyme activity feeds back to affect the growth of microbial populations, their metabolism, and their enzyme synthesis and profoundly affects the biogeochemical cycles of these elements.  Do models need to represent all of these processes, including microbial population dynamics, enzyme production, substrate diffusion, and reaction kinetics?  Do these processes need to be modeled in four dimensions (space and time)?

Results/Conclusions

Ideally, the answer to those questions would be “yes,” but only if there is a viable approach to testing and validating the model structures and parameterizations representing each process.  When that is not possible, some aggregation or abstraction of processes is needed to arrive at a reasonable number of testable and measureable model structures and parameters.  A modular design also enables progress on model components without losing sight of the way that components fit together.  Admittedly, the Dual Arrhenius and Michaelis–Menten (DAMM) model does not yet attain all of these lofty goals, but it offers promise to build upon an integrated, modular approach to represent as parsimoniously as possible numerous key interacting processes in a heterogeneous matrix, and to keep making improvements until we get the DAMM thing right.