The objective of this work was to compare two multivariate methodologies in similarity studies among Atlantic Forest fragments. A bibliographical survey was carried out, and databanks were set up from 11 fragments of Atlantic Forest located in the States of Pernambuco, Rio Grande do Norte, Minas Gerais, São Paulo and Rio de Janeiro. The study of the floristic similarity was constructed using two methodologies of the multivariate analysis. For the usual methodology, a binary matrix (presence/absence) of 236 tree species present in the 11 fragments was used. A cluster analysis was applied using the simple linkage method and the Jaccard's coefficient. In the proposed methodology, the principal components analysis was used for reducing the size of the density matrix and absolute dominancy of the 236 tree species, using the principal component scores to apply the cluster analysis by the Euclidean distance single linkage method. Two groups were identified; one in the Northeastern Region (Pernambuco) and another in the Southeastern Region (Minas Gerais). The proposed methodology identified only one group with fragments in the Northeastern region (Pernambuco), standing out that the quantitative variables are of utmost importance for the association of forests in different regions. The proposal methodology has potential for use in the study of forest fragment similarity.
Cluster analysis; principal components analysis; association of multivariate techniques
