Abstract:

The purpose of this study was to determine the variation pattern that exists in the East
African Highland bananas (Musa AAA) grown in Uganda, estimate levels of
dissimilarity caused by different growing conditions, and establish a flexible provisional
classification and identification system. Techniques of numerical taxonomy were
employed to determine the variation pattern and these included two different
coefficients, three different clustering methods, principal component analysis and
classificatory discriminant analysis. Sixty one morphological characters were employed
to determine differences among the 238 accessions available for the study; 192
accessions were from the national banana germplasm collections, 46 were from
farmers’ fields in selected sites. Phenetic classifications resulting from different
analyses were compared with an independent subjective classification. The phenetic
classifications agreed with the subjective classification with regard to the positions of
the majority of accessions. Accessions which were inconsistently placed in the cluster
analyses were classified by classificatory discriminant analysis and 84 clones were
identified. It was advantageous to compare different methods because they often gave
complementary results. For example the comparison of cluster analysis versus principal
component analysis revealed similar clusters of accessions in the phenograms and along the first four principal components.