Dealing with Uncertainties in Analyzing Holocene Northern Peatland Carbon Dynamics
Thursday, 18 December 2014
Northern peatlands store about 500 Gt of carbon (C), constituting one of the largest continental C pools and thus a major component of the global C cycle. Due to the growing recognition of their importance, long-term changes in peatland C sequestration have been inferred from peat-core data to provide insights into the sensitivity to climate changes of the peatland C sink, to estimate C stock from local to global scales, and to incorporate peatlands into global C cycle models. In the analysis and synthesis of data from multiple sites/cores, chronological uncertainties are often ignored or treated simplistically, which may lead to over-interpret insignificant results. Also, additional problems may arise when analyzing and calculating peat C accumulation rates, which are strongly affected by age-depth modeling results, that is, small uncertainties in ages may lead to large differences in accumulation rates. Here we propose a new method, combining Bayesian age-depth modeling and Monte-Carlo procedure, to analyze peat C rate datasets by systematically addressing both chronological and rate calculation uncertainties. This method consists of the generation of a large number of dataset simulations from the possible modeled C rates, each with different Gaussian noise related to analytical error of C bulk density measurements. These dataset simulations can then be iteratively analyzed, allowing the systematic estimation of uncertainties related to age and density measurements. We applied this approach to the analysis of the latest northern peatland C database (Loisel et al., 2014. The Holocene 24) to generate regional and global synthesis curves, to refine the northern peatland C stock and net C balance during the Holocene, and to derive significant spatiotemporal patterns using Empirical Orthogonal Functions (EOF). For the synthesis curves and the C balance calculations, our method adds an extra ~10% to ~40% of error to the common error estimates (standard error of the means), whereas uncertainties on the EOF temporal patterns are mostly controlled by the number of sites in each time bin. The large amount of uncertainty inherent to age measurements emphasizes the need for more detailed northern peatland C rate records to better understand their dynamics and climate sensitivity, especially for the early Holocene.