Potential in detection of cereal yellow dwarf virus in cereals through VisNIR spectroscopy

Plant disease management often involves timely strategies and practices to prevent the spread and establishment of viral pathogens within plant populations. This study investigated the potential of using proximal sensing technologies for early detection of the yellow dwarf virus (YDV) disease. Four cereal cultivars (two wheat cultivars, Revenue and Mace, and two oat cultivars, Bass and Eurabbie) were grown under glasshouse conditions and inoculated with a YDV strain (Cereal yellow dwarf virus; CYDV) using aphid vectors. Spectral measurements of the third leaf were taken 3 weeks after infection using a FieldSpec spectrometer covering the wavelength range 325–1075 nm. Two machine learning (ML) classification algorithms, namely the eXtreme gradient boosting (XGBOOST) and the support vector machine (SVM), were used following the leave-one-cultivar-out-cross-validation (LOCOCV) approach. According to the results, virus infection in its early stages could be detected for susceptible oats (Eurabbie) with reasonable accuracy but no other cultivars. Best overall accuracy (OA) and kappa coefficient values were achieved for Eurabbie using XGBOOST built with vegetation indices (VIs) (OA = 0.79; kappa = 0.57). Further investigation into cultivar differences and growing conditions would be beneficial. More work is required to confirm the potential of VisNIR spectroscopy for detecting CYDV, especially in cultivars such as Eurabbie.