CANCELLED - Using global data to test microbial activity in litter decomposition models

Background/Question/Methods

Decomposition plays a major role in the global carbon (C) and nitrogen (N) cycles, providing plant available N and releasing ten times more CO₂ to the atmosphere than fossil fuel and industrial sources. Thus, our ability to predict atmospheric CO₂ and ecosystem properties depends on accurately modeling decomposition. Despite this, our understanding of decomposition remains relatively rudimentary. This uncertainty is reflected in the diversity of approaches used to model decomposition, and in debate about if, and how, microbes should be explicitly represented. To examine the relevance of explicitly incorporating microbial activity into large-scale C and N decomposition models, we compiled a large-scale, long-term database of litter decomposition data. We used these data and likelihood methods to compare a set of models that varied only in how microbial activity was modeled (implicitly or explicitly). A base model was developed following the Adair et al. (2008) model that consists of three C pools (labile, cellulosic, and lignin), which only implicitly incorporates microbial activity by allowing climate and litter chemistry to modify pool decomposition rates. Using this microbially implicit model as a base, we developed four models that explicitly incorporated microbial processes by including the formation of recalcitrant microbial products and/or pools representing microbial populations. We hypothesized that while microbially implicit models may accurately represent C loss over time, because they are not readily adapted to represent N flows (i.e., allowing N mineralization but providing no mechanism for immobilization), a microbially explicit model would better predict litter N dynamics.

Results/Conclusions

The microbially implicit base model explained 64% of the variation in litter C loss but only 12% of the variation in litter N loss. This base model was unable to account for slow rates of N loss or immobilization and dramatically over-predicted N loss from litter. Incorporating microbial processes did not substantially improve C loss prediction, but did improve the prediction of N loss over time. Only models that incorporated microbial populations and allowed those populations to take up N from the environment were able to account for both slow N loss and the immobilization of N into litter. Thus, while the best model for describing patterns of C loss was the simpler, microbially implicit model, the best model for describing N loss explicitly incorporated microbial populations. Our results suggest that incorporating microbial activity into mechanistic decomposition models will improve our ability to predict patterns of N loss through decomposition.