The nature and causes of sugarcane genotype x environment interactions: integrated approaches to analysis and interpretation
Information on the nature and causes of G x E interactions in the rainfed parts of the South African sugar industry were lacking. The aim of this study was to systematically analyse, identify causes, and explore more comprehensive methods of analysing and understanding the G x E interactions of sugarcane, in order to optimize future MET networks. Data from plant breeding selection trials and post-release evaluation trials were systematically analysed using various statistical approaches combined with the use of soil and climatic data and crop models. Statistical methods based on multivariate methodologies such as GGE biplot, AMMI, and pattern analysis, were used to explore the effects of different environmental factors on sugarcane performance and agronomic traits. The age at harvest was the main factor causing different genotypic responses in the midlands region, which was unique in comparison with the coastal and hinterland regions that shared similar characteristics. In the midlands region, two testing sites were identified as being redundant and were recommended for removal from the trial network in favour of a testing site in a frost pocket. In the coastal/hinterland region, three sites were identified as being redundant. Along the coast, time of harvest influenced G x E interactions, with yields showing stronger correlations to stalk population in the early season and stalk weight in the late season. In all regions, site x ratoon interactions accounted for more variation in yield than genotype x ratoon interactions, suggesting that variation in ratoon performance is influenced more by site differences than genotype differences. The repeatable component of G x E interaction (genotype x site) was larger than the non-repeatable components (genotype x ratoon and genotype x site x ratoon) on the long cycle program on the coast, however, this was not the case on the short cycle. This suggests that more effort be placed on identifying more diverse test sites on the short cycle. In addition to providing direct recommendations for the industry selection programs, this study also illustrated novel methods of understanding sugarcane growth in different environments. The benefits of using a crop growth model to characterize sugarcane METs for water stress were illustrated throughout the study. The further use of the crop model to establish sugarcane growth phases also proved useful, and is likely to be more valuable when diverse datasets are analysed. Trials were characterized in terms of basic climatic and soil variables, which proved to be invaluable in understanding the causes of G x E interactions. The characterisation of the current sites ensures that future site selection will be more rigorous, as plant breeders will be more aware of the conditions to select for or against. The study showed that the integration of empirical and analytical statistical approaches was more valuable than using either approach in isolation, as is conventionally done in sugarcane. Additionally, these techniques were applied across many trial series and shown to produce repeatable results. The different strategies used to investigate sugarcane trait relations in this study have not been reported elsewhere, and future sugarcane studies dealing with similar traits (or other traits associated with lower level plant processes) may benefit from applying these methodologies. Furthermore, the integration of these multivariate methods with the largest ever simulation of sugarcane METs has opened new doors for the combined use of crop and statistical models in sugarcane research – an area not previously explored for this crop. The study illustrated novel methods of identifying factors responsible for sugarcane G x E interactions and introduced new ways of characterizing sugarcane METs through the use of crop growth models and supplementary environmental data.