Bayesian calibration of a model describing carbon, water and heat fluxes for a Swedish boreal forest stand

This study quantified major fluxes of carbon (C), heat and water, including uncertainty estimates, in a boreal forest in northern Sweden, using a process-based model (CoupModel) and Bayesian calibration methodology. Coupled C, water and heat fluxes were described together with estimated uncertainties for all major components of the simulated C budget. Simulated mean gross primary production was 641 ± 74 g C m−2 yr−1, total ecosystem respiration 570 ± 55 g C m−2 yr−1 and net ecosystem productivity 71 ± 37 g C m−2 yr−1. Most high-resolution measurements were well described but some interesting exceptions arose between model and measurements, e.g. latent heat flux was overestimated and field layer (understory) root litter production underestimated. Bayesian calibration reduced the assumed prior parameter ranges in 30 of 33 parameters, thus reducing the uncertainty in the estimates. There was a high degree of couplings between different sub-models and processes in the model, highlighting the importance of considering parameters not as singularities but in clusters.