Wheat phenology is a major determinant of the adaptation of wheat to diverse environments and management, and lowering risk of yield reducing climatic factors (e.g. frost, heat and drought). However, it is impacted by interactions of genotype, environment and management (GxExM) from agronomy perspectives or interaction of genotype and environment (GxE) from genomics perspectives.

Over the decades, genomic predictions and crop models have been separately developed to deliver genetic gain within limited environments, or determine environmental responses within limited genotypes. Here we describe an integrated model combining genomic prediction (GP) and a crop growth model (CGM) to predict wheat phenology (i.e. timing of flag leaf emergence, heading and flowering) in diverse environments across the Australian grain belt.

A wheat population was selected to represent the current phenological diversity of cultivars in Australia and genotyped for single nucleotide polymorphisms (SNPs) with the wheat 90K Illumina SNP array.

A random forest model was used to predict genotypic parameters for the crop model APSIM Next Generation, which was used to further predict wheat phenology. The observed datasets are assembled from controlled environment and field experiments and historical datasets used to calibrate and validate the new hybrid model (GP-CGM) in four scenarios with one calibration and three validations for unobserved genotypes and/or environments.

GP-CGM can accurately predict wheat phenology in unobserved environments and genotypes with prediction accuracy of more than 0.7. The new GP-CGM model can now be used to predict phenology for current cultivars across environments and new cultivars with just SNP and controlled environment validation to explore the complex interaction between genotype and environment. It can also be extended to other crop models, -omics prediction and phenotypic data sourced from remote and proximal sensing.