Optimizing biomass partitioning in wheat using UAV-based hyperspectral phenomic and genomic prediction: kernel-based...

Area of Science:

• Agricultural Science
• Plant Physiology
• Remote Sensing

Background:

• Optimizing biomass partitioning is crucial for sustainable wheat yield improvement, especially under environmental stress.
• Key traits like spike partitioning index (SPI), harvest index (HI), and fruiting efficiency (FE) are difficult to phenotype manually.
• Understanding assimilate allocation is vital for wheat breeding programs.

Purpose of the Study:

• To evaluate the potential of UAV-based hyperspectral reflectance data for predicting biomass partitioning traits in wheat.
• To compare genomic prediction (GP), phenomic prediction (PP), and integrated multi-omic models.
• To assess the utility of these methods for in-season selection in wheat breeding.

Main Methods:

• Utilized UAV-based hyperspectral data from wheat trials (2022-2024).
• Developed genomic prediction, phenomic prediction, and integrated multi-omic models.
• Employed kernel-based BLUP, random forest regression, and partial least squares regression for predictive ability estimation.

Main Results:

• Phenomics-driven models significantly outperformed GP for traits like SPI (PA up to 0.61), FE (PA up to 0.56), grains/m² (GN, PA up to 0.71), and grain yield (GY, PA up to 0.66).
• Hyperspectral data showed higher accuracy than vegetation indices.
• Multi-omic integration offered a slight improvement in prediction accuracy for GN (PA up to 0.73).

Conclusions:

• UAV-based hyperspectral phenotyping effectively captures physiological signals for biomass partitioning.
• This approach offers a scalable, data-driven method for in-season selection in wheat breeding.
• It aids in optimizing biomass partitioning for enhanced yield resilience and genetic gain in wheat cultivars.