Feasibility of near-infrared spectroscopy as a tool to estimate carotenoid content in ‘IAC Rurik’ specialty potato cultivar

The study explores the use of NIR spectroscopy with chemometric techniques as a non-destructive method to determine carotenoids in fresh ‘IAC Rurik,’ a new yellow potato rich in these compounds. Tubers were harvested in 2022 and in 2023. Reflectance FT-NIR spectra were acquired on the periderm of 200 tubers, in two positions. Spectra were pre-processed and regression models were developed using partial least square (PLS), support vector (SVR), ridge regression, k-nearest neighbors (KNN), interval-partial least squares (iPLS), and kernel-partial least squares (kPLS) regressions. Multivariate classification was carried out by applying principal component analysis with linear discriminant analysis (PCA-LDA) and partial least squares discriminant analysis (PLS-DA). Carotenoid content prediction was better obtained using mean-centered spectra and ridge regression (RMSEP = 0.0028 g kg−1, R2P = 0.90, RPD = 2.57, RER = 4.21 %). The classification of three groups (low = 0.031–0.045 g kg−1; average = 0.045–0.065 g kg−1; and high = 0.065–0.078 g kg−1) was possible by applying PLS-DA with a correct classification of 93 %, 70 %, and 86 %, respectively for low, average, and high carotenoid content. Thus, NIR spectroscopy can be used as a non-destructive method to predict carotenoids and classify ‘IAC Rurik’ tubers based on their carotenoid content.