Evaluating CENTURY and Yasso soil carbon models for CO2 emissions and organic carbon stocks of boreal forest soil with Bayesian multi-model inference

We can curb climate change by improved management decisions for the most important terrestrial carbon pool, soil organic carbon stock (SOC). However, we need to be confident we can obtain the correct representation of the simultanous effect of the input of plant litter, soil temperature and water (which could be altered by climate or management) on the decomposition of soil organic matter. In this research, we used regression and Bayesian statistics for testing process-based models (Yasso07, Yasso15 and CENTURY) with soil heterotrophic respiration (Rh) and SOC, measured at four sites in Finland during 2015 and 2016. We extracted climate modifiers for calibration with Rh. The Rh values of Yasso07, Yasso15 and CENTURY models estimated with default parameterization correlated with measured monthly heterotrophic respiration. Despite a significant correlation, models on average underestimated measured soil respiration by 43%. After the Bayesian calibration, the fitted climate modifier of the Yasso07 model outperformed the Yasso15 and CENTURY models. The Yasso07 model had smaller residual mean square errors and temperature and water functions with fewer, thus more efficient, parameters than the other models. After calibration, there was a small overestimate of Rh by the models that used monotonic moisture functions and a small generic underestimate in autumn. The mismatch between measured and modelled Rh indicates that the Yasso and CENTURY models should be improved by adjusting climate modifiers of decomposition or by accounting for missing controls in, for example, microbial growth.Peer reviewe