This study was carried out at the Forest Management Unit (FMU) in Tracajás farm (02º35'53" S and 47º47'10"W), owned by Nova Era Agroflorestal, in Paragominas, Pará, Brazil. The objective was to accomplish the vertical stratification of an unexplored terra firme dense ombrophylous forest using cluster and discriminant analysis. The forest was stratified in homogeneous areas, denominated classes of volumetric stock I, II and III. In each class, five 100 x 100 m plots were randomly set up, where the total heights of the individual trees (dbh > 15.0 cm) were used for vertical forest stratification. The trees were organized in increasing order of total height and classified in 1 m amplitude classes, from the minimum to the maximum total height. A total height Xmatrix was elaborated, with each variable x ij representing the total height of the i-th tree classified into j-th class of total height. This X-matrix was the input for the cluster and discriminant analysis. The application of cluster analysis resulted into hierarchical and sequential clusters of height classes into total height strata (inferior, medium, and superior). Discriminant analysis demonstrated a 100% correct classification. The vertical structure subdivision of forest into total height strata, by multivariate statistics proved the method efficient for uneven-aged forest vertical stratification.
Tropical forest; vertical stratification; multivariate analysis