Predicting yellow rust in wheat breeding trials by proximal phenotyping and machine learning.

Area of Science:

• Plant pathology
• Agricultural science
• Remote sensing

Background:

• High-throughput plant phenotyping (HTPP) accelerates crop breeding through automated, scalable methods.
• HTPP is crucial for disease resistance breeding, overcoming visual phenotyping limitations.
• This study focuses on predicting yellow rust in winter wheat using HTPP.

Purpose of the Study:

• To evaluate the use of HTPP data for predicting yellow rust severity in winter wheat.
• To assess the efficacy of spectral vegetation indices (SVIs) and Random Forest (RF) modeling for this prediction.

Main Methods:

• Proximal phenotyping was employed to collect spectroradiometer data in a winter wheat field trial.
• Random Forest (RF) modeling was used to predict yellow rust scores.
• Recursive feature elimination (RFE) identified key spectral vegetation indices (SVIs) for prediction.

Main Results:

• 40-42 SVIs derived from spectral data were sufficient for predicting yellow rust scores.
• RF models achieved prediction accuracies (r_s) between 0.50-0.61.
• Key spectral features included Plant Senescence Reflectance Index (PSRI), Photochemical Reflectance Index (PRI), Red-Green Pigment Index (RGI), and Greenness Index (GI).

Conclusions:

• Combining SVI data with RF models offers a viable HTPP method for scoring yellow rust.
• This approach has potential for deployment in wheat breeding programs to enhance disease resistance selection.