In light of the climate change that jeopardizes future food security, genomic selection is emerging as a valuable tool for breeders to enhance genetic gains and introduce high-yielding varieties. However, predicting grain yield is challenging due to the genetic and physiological complexities involved and the effect of genetic-by-environment interactions on genomic prediction accuracy.

We utilized a chained model approach to address these challenges, breaking down the complex prediction task into simpler steps. A diversity panel with a narrow phenological range was phenotyped across three Mediterranean environments for various morpho-physiological and yield-related traits. In addition, canopy hyperspectral reflectance was captured by an unmanned aerial vehicle at heading.

The results indicated that a multi-environment model outperformed a single environment in prediction accuracy for most traits. However, no improvement in prediction accuracy was found for grain yield, ranging from 0.14 to 0.29. Spike number was assessed using machine learning from canopy spectra and exceed substances in a leave-one-environment-out validation.

The estimated spike number was utilized as a secondary trait in a multi-trait genomic selection and significantly improved grain yield prediction accuracy by 36% to 75% when only the calibration dataset includes the secondary trait.

This study emphasized the potential of hyperspectral-based high-throughput trait estimation as a secondary trait for improving genomic selection for wheat grain yield.