The forest role in the process of carbon exchange between biosphere and atmosphere is ambiguous. Sometimes it is source of greenhouse gas during land use processes, and it could be a sink under sustainable forest management or even in natural conditions. After the AR4 (IPCC Fourth Assessment Report), the forest role in the global changes processes was highlighted, in special, the Amazonian forest. In any condition, reliable methods are necessary to estimate forest carbon stocks and changes over time; keeping in mind that any method must be subject of auditing process. This work was carried out under this background. The forest management on sustainable yield basis experiment of the National Institute for Research in the Amazon (INPA) was utilized as demonstrative area. This experiment has been carried out since 1980 in Manaus region, Amazonas State. The following data sets were utilized: aboveground fresh weight of primary forest (n = 494), belowground – coarse roots - fresh weight of primary forest (n = 131), aboveground fresh weight of secondary forests (n = 252 from the 14 years old e n = 341 from the 23 years old), belowground fresh weight of secondary forests (n = 65), 12 permanent sample plots of 1-ha each of a forest managed under four different logging intensities. Scientific hypotheses were tested for different allometric equations, as well as for different dimension relations and for the forest dynamics. The total fresh weight (aboveground + coarse roots) could be estimated using the following models for primary forests: PF = 2.7179 * DAP 1.8774 (r2 = 0.94 e sy.x = 3.9%) e PF = 0.5521 * DAP 1.6629 * HT 0.7224 (r2 = 0.95 e sy.x = 3.7%); where: PF = fresh weight in kg, DAP = DBH in cm, and HT = total height in m. Models for different components of biomass were developed too, either for primary and secondary forests. The results obtained for water and carbon contents were very similar to those found in the literature. For primary forest, for instance, the water and carbon contents averaged 41.6% (uncertainty = 2.8%) and 48.5% (uncertainty = 0.9%), respectively. The costs (running and personnel) for destructively collecting total biomass (aboveground + coarse roots) of an individual tree averaged from R$ 805.00 ± R$ 310.00 (CI 95%) considering only aboveground biomass, averaged from R$ 27.00 ± R$ 6.00 (CI 95%). There is no need to use the same sampling intensity used in this work; 100 trees (samples) are more than enough to keep the standard error of estimate below the limit of 10%. The annual aboveground biomass increment averaged 10.8 t.ha-1.year-1 and 10.6 t.ha-1.year-1 for 14 and 23 years old secondary forests, respectively. The annual aboveground carbon increments for logging treatments T0 (control), T1 (light logging), T2 (medium logging) and T3 (heavy logging) averaged 0.81 t.ha-1year-1, 2.22 t.ha-1year-1, 2.36 t.ha-1year-1 e 2.39 t.ha-1year-1, respectively. These results have indicated that the human induced activities could improve the productivity of the forests under sustainable forest management. Markov chain approach has shown a reliable performance in projecting the frequency distribution of diameter in managed forests.
