The purpose of this work was to study the genetic divergence among 99 accessions of the active genebank of elephantgrass (Pennisetum purpureum) of Embrapa, in Coronel Pacheco, MG, Brazil. The experimental design used was randomized blocks, with 99 treatments (97 genotypes and two controls), with five replications. The evaluation involved 17 quantitative traits. The following multivariate analyses were used: distance of Mahalanobis, Tocher's method, and canonical variables. By Tocher's method the 99 genotypes were grouped in 18 clusters. According to the analysis of canonical variables, 81.80% of the accumulated variance were explained by the first seven canonical variables. The relative importance of the characteristics was determined and those that contributed less to the genetic divergence were: diameter of the stem, width of the half blade of the medium complete leaf, width of the lower blade of the medium complete leaf, length of the awn, length of the spikelet, length of the medium complete leaf and width of the flag leaf. One test eliminating diameter of the stem altered the number of groups and the position of the genotypes established by the method of Tocher, as an illustration of the use of this technique. Several crossings were indicated, involving the control G98 and the promising divergent genotypes to attain increases in variability and the possibility of identification of transgressive segregants.
Pennisetum purpureum; genotypes; genetic distance; genetic variability; selection; forage
