Multivariate analysis of sugar yield contributing traits in Sugarcane (Saccharum officinarum .L), in Ethiopia

Knowledge on performance of genotypes and interrelationships among traits is very important for sugarcane breeding programmes. The objectives of this study were to assess the phenotypic relationship among 49 sugarcane genotypes and the inter-relationships among traits considered. The cluster analysis demonstrated that the 49 sugarcane genotypes studied were clustered into nine groups and were highly different for Pol in juice, cane yield (tons ha-1m-1), number of tillers (ha-1), purity% and milleable stalk population (ha-1). The relationship among sugarcane genotypes was not dependent on geographic origin, suggesting that a high proportion of total genetic variation was retained within the groups of origin and active genetic ex-change was found between different origins. The principal component analysis indicated that cane yield, milleable stalk height and milleable stalk diameter were highly correlated with sugar yield while the correlation of quality traits with sugar yield was weak. In contrary, path and multiple regression analysis revealed that cane yield, recoverable sucrose percentage (%) and Pol contribute more to the variability of sugar yield; these are very important traits for high sugar yield that should be considered in sugarcane breeding programmes. Moreover, milleable stalk height and milleable stalk population via cane yield and Brix, Pol, purity and number of internodes via recoverable sucrose percentage had high indirect effects on sugar yield suggesting these traits should also be given consideration during selection for high sugar yield. Generally, similar and adequate information was generated following the use of cluster, principal component, linear discriminant, path coefficient and multiple regression analyses indicating the use of multivariate analyses was successful and results of the study were more substantial to give concrete recommendation. Key words: Clusters, genotypes, multiple regression, path coefficient, phenotypic correlation.