The aim of this study was to analyze different sections lengths for volume determination, fitting of volumetric models for timber production estimation in an area of forest management in the Tapajós National Forest (FNT). Six treatments for sectioning were tested in 152 logs of 12 commercial species. The obtained volumes were statistically compared by analysis of variance (ANOVA) for the choice of the best method of sectioning and calculating the actual volume of 2,094 sample trees in different diameter commercial classes. Ten mathematical models were fitted to the whole data and to the species Manilkara huberi (Ducke) Chevalier (maçaranduba) Lecythis lurida (Miers) Samori (jarana) and Hymenaea courbaril L. (Jatobá). The criteria to choose the best model were adjusted coefficient of determination in percentage (R²adj%), standard error of estimate in percentage (Syx%), significance of the parameters, normality of residuals, Variance Inflation Factor (VIF) and residuals graphic distribution. There was no statistical difference between the methods of sectioning and thus the total length of the logs was more operational in the field. The models in logarithmic form of Schumacher and Hall and Spurr were the best to estimate the volume for the species and for the whole sample set.
Dendrometry; Regression analysis; Amazon forest
