Predictive models incorporating environmental covariates for genotype × environment × management (G×E×M) interactions applied to sorghum agronomy trials

Abstract

Researching the management (M) of genotypes (G) in agronomic experimentation is essential to help farmers maximise grain yield, though the approach is complicated by interactions emerging from changing environmental (E) factors across sites and seasons. Available statistical methods for modelling the G×E interaction are limited as they do not provide a functional understanding of how environmental factors influence the G×E interaction, nor assess how different management practices (M) influence the G×E interaction. A predictive linear mixed model is proposed that incorporates site/season-specific environmental covariates into a standard G×E interaction framework. The model is extended to include continuously varying agronomic management practices whilst allowing for non-linear trait responses and complex variance structures. The methodology was applied to a multi-environment data set associated with GRDC’s optimising sorghum agronomy program. The analysis identified key environmental drivers and management strategies that explained the G×E×M interaction, enhancing the biological understanding of the analysis results and allowing for the development of more robust recommendations for agronomic practices.