Carbon Cycling of Forest Ecosystems As a Fuzzy System: An Attempt to Assess Uncertainties.

We discuss a methodology of the full verified carbon account of forest ecosystems (FCA) that attempts at reliable and complete assessment of the uncertainties. In essence, the FCA is a fuzzy (underspecified) system, of which the membership function is inherently stochastic. This predetermines principle impossibility of assessment of structural uncertainties within any individually used method of carbon account because the “within method” uncertainty is inevitably incomplete. The methodology integrates results independently received by basic methods of terrestrial ecosystems carbon accounting (landscape-ecosystem approach, process-based models, eddy covariance, inverse modelling, direct assessment of important components of the FCA by remote sensing) following the principles of applied systems analysis. Landscape-ecosystem approach (LEA) presents the empirical background of the account, containing spatially distributed relevant information about ecosystems and landscapes and serves for strict designing the studied system. Information background of LEA is presented by an Integrated Land Information System as multi-layer and multi-scale GIS by polygons of a hybrid land cover (HLC). The HLC is developed using diversity of relevant remote sensing products, geographically-weighted regression and validation by Geo-Wiki tools. By-pixel parametrization of forest cover is provided based on multi-sensor remote sensing concept, data of forest inventory, soil and landscape characteristics etc. based on a special optimization algorithm. Major carbon fluxes are calculated based on databases of measurements in situ with correction for seasonal weather conditions by sets of regionally distributed models. Uncertainties within LEA are calculated based on error propagation theory and Monte-Carlo simulations. In order to assess the “full uncertainties”, the LEA results are mutually constrained with those presented by other methods of carbon account using the Bayesian approach. Strengths and weaknesses of such a methodology are illustrated by the FCA for Russian forests for 2010-2012. The uncertainty of yearly Net Ecosystem Carbon Budget is in range of 20-25% (CI 0.9).