Abstract

Grape ripening is typically monitored using laborious and destructive methods that should be soon replaced by automatic and remote alternatives. This study was conducted with grapes (cv Cabernet Sauvignon) obtained at a drip-irrigated commercial vineyard located in the Pencahue Valley of the Maule Region in Chile. The study aimed to develop and validate models based on spectral reflectance (SR) indices to estimate soluble solids (SS), titratable acidity (TA), pH, color index (CI), anthocyanins (ANTH), tannins (TAN), and phenols (PHEN) of grapes under regulated deficit irrigation (RDI). From veraison to harvest, berry samples were collected to measure grape quality (pH, TA, SS, CI, ANTH, TAN, and PHEN) every week from both shaded and illuminated clusters. SR measurements of berry samples were taken using a portable hyperspectral spectroradiometer, and correlations between SR indices and grape quality were established. In this case, the normalized difference spectral index (NDSI 29 and 20), damage-sensitive spectral index (DSSI 1), and simple ratio pigment (SRPI) indices were evaluated in this study. Results indicated that there were significant linear correlations between grape quality and SR indices with R2 values ranging from 0.59 to 0.78. The NDSI29 index had the strongest correlation with SS and DSS 1 presented the weakest linear association with ANTH1. Model validation indicated that the linear models based on NDSI29 and DSSI showed errors lower than 8 % to simulated SS, pH, TA, and ANTH while those based on NDSI20 and SRPI predicted TAN and PHEN with errors between 11 and 21 %.