Prediction of viability of oriental beechnuts, Fagus orientalis, using near infrared spectroscopy and partial least squares regression

Near infrared (NIR) spectroscopy, combined with multivariate calibration, was applied for determining viable and non-viable single oriental beechnuts. Initially, samples were sorted into viable and non-viable classes with X-ray and then NIR reflectance spectra were recorded on individual nuts using a fibre-optic probe. Calibration models were developed on raw and pretreated spectra with partial least squares (PLS) regression. Multiplicative signal correction (MSC) and orthogonal signal correction (OSC) were applied to remove systematic noise in the spectra. The resulting models separated viable and non-viable nuts in the test set with 100% accuracy. Moisture was the major source of spectral variation between viable and non-viable nuts, although lipid and protein moieties were also contributing factors for the separation of the two classes. We concluded that prediction models, based on fast and non-destructive NIR spectroscopy, have a high potential for the removal of non-viable nuts within oriental beechnut lots, thereby facilitating single nut sowing in the nursery.