Resumo. The aim of this study was to adjust equations to estimate the carbon stock per hectare in stems of trees in function of forest variables, obtained by horizontal point sampling (Bitterlich method). Were sampled 55 points with a basal area factor K = 1, in natural forest fragments, located at Viçosa, Minas Gerais. After adjustment and evaluation of different regression models, was verified that the model 8 (non-linear) were the most precise to estimate the carbon stock per hectare in the stems of trees.