Prediction of Olive Chemical Characteristics by FT-NIR Spectroscopy

Chemical characteristics of olive fruits change during the ripening period and definitely affect the quality of the oil. A rapid and nondestructive method for predicting these characteristics is of paramount importance for the quality design of the end product. However, spectroscopic determination of quality parameters in intact olives is less frequent than other fruits (Fernández-Espinosa, 2016). Thus, this work aimed at predicting water, oil, and total polyphenol content (TPC) for different cultivars of olives by means of FT-NIR spectroscopy. In particular, 267 olive samples belonging to 13 different cultivars and collected during three harvesting years were analysed in diffuse reflectance by an FT-NIR spectrometer (12,500–3,600 cm-1; 8 cm-1 resolution; 32 scans). Samples were analysed as single olives (20 olives per sample) by a fibre optic probe and as aliquots (100 g each) by an integrating sphere (2 aliquots per sample). Chemical analyses were performed as reported by Trapani et al. (2016). Spectra were sample-based averaged and pretreated to develop PLS regression models validated both by cross-validation and external prediction (30% of samples selected by Kennard-Stone algorithm). Moisture, oil, and TPC content ranges were 39.5–85.3%, 2.1–26.0%, and 2.5–60.6 g/kg, respectively. Good PLS models were obtained for all the chemical parameters, with prediction R2 ranging from 0.78 to 0.84 and maximum RMSEP values of 4.3%, 3.0%, and 8.5 g/kg for moisture, oil, and TPC, respectively. Similar results were obtained for both of the sample presentation forms, suggesting applicability of FT-NIR spectroscopy for chemical characterization of olive fruits both in-field and on-line.