Evaluation of uncrewed aerial system‐derived vegetation indices for forage trait phenomic prediction in oat–pea intercrop breeding trials

Abstract Accelerating genetic gain in intercrop forage breeding requires scalable phenotyping approaches that can capture both yield and quality variation across complex genotype combinations. This study evaluated the use of uncrewed aerial system (UAS)‐derived multispectral vegetation indices (VIs) for predicting forage yield and nutritive quality in oat ( Avena sativa L.)‐pea ( Pisum sativum L.) mixtures. Field trials across three seasons included 24 oat genotypes and five pea varieties evaluated in monocrops and all possible mixture combinations. Key traits included biomass yield (YLD), crude protein (CP), fiber fractions (acid detergent fiber [ADF] and neutral detergent fiber [NDF]), total digestible nutrients (TDN), and relative feed value (RFV). Five VIs (normalized difference vegetation index, green normalized difference vegetation index, normalized difference red edge index, excess green, excess green minus red index) derived from near‐infrared (NIR) and visible bands were incorporated as secondary traits in bivariate and multivariate mixed models to evaluate predictive ability in cross‐validation studies. VIs showed moderate to high heritability ( H 2 = 0.14–0.82) and significant correlations with forage traits in mixtures ( r = 0.25–0.71). Predictive ability varied across models, from −0.16 to 0.68, and was consistently higher for multivariate models (mean ∼0.41) compared to bivariate models (mean ∼0.22). Models integrating all five VIs provided the highest accuracy for YLD, CP, and ADF, whereas NIR‐based indices were most effective for NDF, TDN, and RFV. Green‐band indices were strongly associated with quality‐related traits, while NIR‐based indices better captured yield and fiber variation. These results demonstrate that combining spectrally complementary VIs improves phenomic prediction in oat‐pea systems by capturing both species composition and biomass structure. UAS‐based phenotyping provides an effective framework for high‐throughput selection in intercrop breeding.