Articles 3.3 and 3.4 of the Kyoto protocol provide Annex I countries the possibility to reduce greenhouse gas emissions through the sequestration of carbon (C) in their terrestrial ecosystems. For such accounting, the 1990 flux is needed and, therefore, a correct knowledge of the baseline (1990) C stocks is necessary. In addition, a correct methodology should be used to investigate the capacity of ecosystems to sequester C through changes in land use or management by the end of the first commitment period (2008-2012). At national and regional scales, formulation of baseline C stocks in terrestrial ecosystems is difficult and uncertain. Differences in method of analysis, sampling depth of soil, lack of sufficient C data and the necessity to extrapolate C data to total soil organic C stocks, provide problems when comparing databases with each other. In this study, three extrapolation models were compared with the classical layer-based method to determine the model with the best fit. The model with the best predictions, in relation to the classical layer-based model, uses recent soil C profiles for estimating the parameter k, which represents the decrease in the proportion of soil organic C with depth, and for extrapolating the C data available for 1990 and 2000 to a depth of 1 m. The other two models gave large underestimates.