UV-Vis spectroscopy for the estimation of variety and chemical parameters of olive oils

Olive oils produced in different years from different varieties were studied with UV-Vis spectroscopy for classification and prediction. Multivariate models were created with second derivative spectral data, and tested with external validation sets. For varietal classification, orthogonal partial least square discriminant analysis resolved oil samples into various classes with correct classification rate more than 89% for validation set (n = 20). A sample of fresh and stored oils were also classified with a correct classification rate more than 90% for validation set (n = 20). In the predictions of chemical parameters (70 for calibration, 30 for validation), the combination of UV-Vis spectroscopy with orthogonal partial least square regression models showed potential for simultaneous quantification of chlorophylls (0.6-5.6 mg/kg; R-val(2) , 0.79; RPD, 1.97); carotenoids (0.6-3.3 mg/kg; R-val(2), 0.80; RPD, 2.38); ratio of mono to polyunsaturated fatty acids (3.6-8.8; R-val(2) , 0.77; RPD, 1.90), oleuropein derivatives (1.2-62.3 mg/kg; R-val(2) , 0.66; RPD, 1.77), and total phenol content (62.2-505 mg/kg; R-val(2) , 0.67; RPD, 1.74), although showed poor to moderate results for the quantification of free fatty acid (0.3-5.4%; R-val(2), 0.67; RPD, 1.64); monounsaturated fatty acids (66-76.5%; R-val(2) , 0.71; RPD, 1.67); polyunsaturated fatty acids (8.6-18.2%; R-val(2) , 0.73; RPD, 1.65). The models were unable to estimate oxidative stability, saturated fatty acids, and individual phenolics such as hydroxytyrosol, pinoresinol, luteolin, total phenolic acids (R-val(2) , 0.26-0.64; RPD, 0.60-1.52). Results showed the capacities of UV-Vis spectroscopy for classification of olive oils, and prediction of total pigments and phenol content and ratio of mono to polyunsaturated fatty acids.

Automated summary

This study utilized UV-Vis spectroscopy to classify and predict olive oils from different years and varieties. Multivariate models were successfully created for classification and prediction of chemical parameters.