Interpretation of genotype × environment interactions of sugarcane: Identifying significant environmental factors

An understanding of the causes of genotype × environment (G × E) interactions is essential for the implementation of efficient selection and evaluation networks. Currently, studies involving the interpretation of sugarcane (Saccharum spp.) G × E interactions are limited. The objective of this study was to investigate the relative influence of environmental factors on the G × E interactions of sugarcane under rainfed conditions in South Africa through a comprehensive analysis of a multi-environment trial (MET) dataset. Fifteen commercial cultivars were evaluated over 147 environments (trial × ratoon combinations) across the coastal (C), hinterland (H) and midlands (M) regions of the sugar industry. Environments were characterized according to five site covariates (soil depth, clay percentage, organic matter percentage, nitrogen mineralization category, and total available moisture) and nine seasonal covariates (time of harvest, age at harvest, average daily heat units, solar radiation, rainfall, evaporation, and three derived water stress indices).Additive main effects and multiplicative interaction (AMMI) biplots for cane yield (TCANE), estimated recoverable crystal percent (ERC) and tons ERC (TERC) revealed overlapping of C and H environments, while M environments formed unique clusters characterized by specific cultivar adaptabilities. Principal components analysis (PCA) allowed visualization of the covariates determining the regional separation patterns. AMMI interaction principal components axes (IPCA) 1 and 2 scores were correlated to the covariates and showed that harvest age, temperature, and water stress were mainly responsible for separation of M environments from C and H environments on the TCANE and TERC biplots. Time of harvest was identified as an important covariate influencing ERC G × E patterns in the C and H regions. The third water stress index (based on a ratio of observed yields to simulated irrigated yields) was a dominant factor influencing G × E patterns within the C and H regions and was identified as a superior indicator of water deficient environments for future studies. The M trials were characterized by shallower soils with lower total available moisture and greater variability in this regard compared with the C and H trials. Nitrogen mineralization category, organic matter percent, and clay percent were not significantly correlated to IPCA scores, while soil depth was identified as a major site selection criterion in the M region. The M region should be treated as a single mega-environment, while the C and H regions could be combined for future interpretive studies, where covariates should be summarized within growth phases. The results of this study will assist in restructuring the MET network through exploitation and targeting of the relevant environmental factors within the different regions.