Non-destructive ripening evaluation of chilean grape by a portable vis/NIR system

Grape composition at harvest is a key factor determining the quality of future wine. For this reason, measurement of grape characteristics in field during the ripening is a requirement to evaluate the right moment for the harvest. Classical chemical analyses, common at the moment in wine industry, have several limits: these take time, are wasteful, destructive, and ultimately inadequate to test fruit in field. Wineries need new practical and quick instruments, non-destructive, and able to quantitatively estimate, in field and on a large scale, the interesting parameters. The goal of the present study was to test an optical portable system for prediction of ripening parameters of fresh berries of grape (Vitis vinifera L. ‘Cabernet Sauvignon’ and ‘Sauvignon Blanc’). In particular, the evaluation of the correlation between vis/NIR spectra and classical destructive ripening parameters was targeted, for samples of both red and white grapes. Each sample was obtained in a different date during the last period of ripening, by averaging the spectral acquisitions of several individual berries. A chemometric regression model was created for each parameter considered. Vis/NIR spectra were correlated with ripening parameters (total soluble solids content TSS, titratable acidity, pH, weight of 200 berries, potential alcoholic degree PAD and sugar/acidity ratio) and with phenolic ripening parameters (extractable anthocyanins EA, total anthocyanins TA and tannins) using the partial least square (PLS) regression algorithm. Principal component analysis (PCA) was performed on spectra too. For both red and white grapes, PCA showed a significant sample grouping for the different acquisition times. PLS models on red grape gave good predictive skills for TSS, acidity and PAD (0.7 < R2 < 0.8, low values of RMSE), and for EA (R2cv = 0.74), while less accurate was the model elaborated for TA. Similar results were obtained for white grape (RMSE < 0.9 °Brix for TSS).