The Markov chain Monte Carlo (MCMC) method has been widely used to estimate terrestrial ecosystem model parameters. However, inverse analysis is now mainly applied to estimate parameters involved in terrestrial ecosystem carbon models, and yet not used to inverse terrestrial nitrogen model parameters. In this study, the Bayesian probability inversion and MCMC technique were applied to inverse model parameters in a coupled carbon-nitrogen model, and then the trained ecosystem model was used to predict nitrogen pool sizes at the Duke Forests FACE site. We used datasets of soil respiration, nitrogen mineralization, nitrogen uptake, carbon and nitrogen pools in wood, foliage, litterfall, microbial, forest floor, and mineral soil under ambient and elevated CO₂ plots from 1996-2005.

Results/Conclusions

The initial values of C pools in leaf, wood, root, litter, microbial and forest floor were well constrained. The transfer coefficients from pools of leaf biomass, woody biomass, root biomass, litter, forest floor were also well constrained by the actual measurements. The observed datasets gave moderate information to the transfer coefficient from the slow soil carbon pool. The parameters in nitrogen parts, such as C: N in plant, litter, and soil were also well constrained. In addition, parameters about nitrogen dynamics (i.e. nitrogen uptake, nitrogen loss, and nitrogen input through biological fixation and deposition) were also well constrained. The predicted carbon and nitrogen pool sizes using the constrained ecosystem models were well consistent with the observed values. Overall, these results suggest that the MCMC inversion technique is an effective method to synthesize information from various sources for predicting the responses of ecosystem carbon and nitrogen cycling to elevated CO₂.