This study was carried out in order to propose a method for vertical stratification of trees in a heterogeneous natural forest stand with the use of multivariate analysis. Total height data were obtained from 10 permanent sample plots (each with 20 m x 50 m) (re)measured in June, 2000. Plots were established in 1994, in Viçosa (20º 45' S and 42º 55' W), Minas Gerais, Brazil. Total height data for individual trees were ranked and classified in one meter height classes. An X data matrix was organized with the total height information: each element x ij represents an individual observation on the ith tree located in the jth total height classes. This X matrix was the input for both the cluster and discriminant analysis. The Euclidean distance and the complete linkage method were used. Cluster analysis produced a hierarchy of distinct sequential clusters of total height, namely: low, average and high strata. Discriminant analysis has proven that the classification was correct. Distribution of trees among all three strata were as follows: low 1.068 (71.63%); average, 324 (21.73%); and high, 99 (6.64%) individuals. Anadenanthera macrocarpa, Apuleia leiocarpa, Pseudopiptadenia contorta and Piptadenia gonoacantha presented the largest absolute density (DA) per unit area and regular distribution in the three strata. The stratification method used in this study has proven to be an operationally viable technique for structural analysis of natural forest stands as well as for planning and implementation of sustainable forest management.
Secondary forest; vertical structure; multivariate analysis
