The objective of this research was to apply multivariete techniques analysis to separate ecological groups. Data of 37 tree species, in area without intervention, obtained in ten years of survey by the Experiment of Sustainable Production in Secondary Forest of Transition, established in 1986, at Rio Vermelho and Serra Azul de Minas, Minas Gerais State, Brazil were used. The species were separated in pioneers, early secondary and old secondary. The considered variables were: number of trees per hectare, number of ingrowth, mortality, basal area, volume, mean diameter, increment in diameter, increment in basal area, increment in volume, index of value of importance and natural regeneration. Principal components analysis (PCA); cluster analysis (CA) and the discriminant analysis (DA) were used. By PCA it was possible to reduce the dimension to three-dimensional with variance explanation above 79%. In the CA, seeking a classification at posteriori, it was observed that group formation did not correspond to the classification at priori. With the DA, 92.86 and 57.14% of classification at posteriori and at priori respectively were correct. In conclusion: the use of the principal components analysis, cluster analysis and discriminant analysis allowed the identification of tree species that should be classified in a larger number of ecological groups; and the application of PCA, CA and DA in the evaluation of at priori classification confirms most researchers' subjectivity in classifying ecological groups of tree species.
Cluster analysis; principal components analysis; discriminant analysis; semideciduos seasonal forest
