When testing nine different sunflower cultivars in 32 South African environments (location x year), an Additive Main Effects and Multiplicative Interaction analysis (AMMI) identified sizeable genotype by environment (GxE) interaction. The first two Interaction Principal Components Axes (IPCA1 and IPCA2) were highly significant (p<0.001), but all the factors responsible for the GxE interaction could not be identified, as the causes of interaction seems to be of complex nature. IPCA1 captured 34% of the interaction SS with only 15.3% of the degrees of freedom, while IPCA2 captured 22.5% of the interaction SS with 14.5% of the degrees of freedom. This indicates that the AMMI2 model fits the data well and is parsimonious. Both cultivars and environments grouped together in quadrants according to their length of season when their respective IPCA1 and IPCA2 scores were plotted against each other. Environments from the warmer dry Western and Northern regions, including the Dry Highveld Grassland, Northern Arid and Central Bushveld, grouped in opposite quadrants, 1 and 3, while environments from the cooler moist Eastern regions, including the Moist Highveld Grassland grouped in opposite quadrants, 2 and 4. The factors responsible for the division between quadrants 1 and 3, as well as those responsible for the division between quadrants 2 and 4 could not be identified. The long-season cultivars were better adapted to the Northern and Western environments, while the medium-season cultivars were better adapted to the Eastern environments. Each quadrant was dominated by a different cultivar. Because the environments and cultivars could not be sufficiently described according to the factors responsible for the observed GxE interaction, cultivars can not be advised for specific environments. It is therefore presently recommended that cultivars which are more widely adapted to South African conditions, be selected.
Dissertation (MSc (Genetics))--University of Pretoria, 2007.