Pre-anthesis canopy greenness and stomatal traits predict grain yield stability in Canadian heritage wheat panel

This study evaluated a Canadian heritage bread wheat panel over three rainfed growing seasons (2021–2023) in Saskatoon, SK, using the Stress Tolerance Index (STI) to identify traits associated with year-to-year yield stability. The STI identified the key, consistent morpho-phenological, physiological, and yield-attributing traits based on performance in contrasting years: 2021 and 2022. These traits were validated across three field seasons. The moderate to high broad-sense heritability (h2bs) for the identified traits underscores their importance as selection criteria. Principal component analysis based on STI grouped the genotypes in three distinct clusters, while the GYT (grain yield × trait) method ranked varieties using a Superiority Index. Further, consistent traits linked to yield stability over years were used as key independent variables in multiple linear regression and in a comparative study of three machine learning regression models. The regression was analyzed in three combinations, using the two-year dataset to predict the grain yield in the remaining year. Multiple linear regression multivariate models achieved prediction accuracies up to 0.786, and the best predictors were identified over the years. Overall, our prediction model identified NDVI before heading, abaxial stomatal aperture length, and file distance as potential early-stage yield predictors under contrasting seasons, while grain width and peduncle length were late-stage screening traits.