Variability in crop responses as a function of environment affects the NDVI relationship with grain yield in wheat

Advancing wheat breeding requires reliable digital traits that capture genotype × environment interactions and improve yield prediction across diverse growing conditions. Although vegetation indices such as the normalized difference vegetation index (NDVI) are widely used, their performance relative to yield variability and environmental stress remains underexplored in multi-environment trials. This study utilized unmanned aerial vehicle multispectral imagery to derive NDVI and assess its relationship with grain yield in 34 spring and winter wheat variety trials. These trials included data across seven Washington State locations in different precipitation zones, five years (2019 to 2023), and some irrigated trials. Environments were grouped into high-, moderate-, and low-stress clusters based primarily on precipitation and temperature. Variability was quantified using the coefficient of variation, and correlations between grain yield and NDVI were evaluated within and between varieties across environments based on market classes (hard and soft spring and winter wheat). Across all environments and varieties, NDVI strongly correlated with grain yield (r = 0.79–0.82, p < 0.001) in each market class. Stress-prone environments exhibited greater variability in both traits, and higher yield variability could be associated with higher NDVI variability, particularly in spring wheat (r = 0.72 in hard spring, r = 0.53 in soft spring). These conditions also improved discrimination between varieties. Although heritability patterns were not clearly differentiated by stress clusters, environments with higher genetic control of yield also tended to show stronger NDVI heritability. Overall, NDVI reliably captured wheat grain yield, which is governed by the genotype × environment driven variability, with its predictive value strongest in stress-prone conditions. These findings underline NDVI's usability as a practical digital trait for improving variety testing and guiding breeding decisions in challenging environments.