Modeling °Brix and pH in Wine Grapes from Satellite Images in Colchagua Valley, Chile

Abstract

To monitor the ripeness and composition of wine grape berries and establish an optimal harvest date, the determination of degrees Brix and pH is vital. This research studies two harvest seasons of Cabernet Sauvignon wine grapes: 2017 and 2018. Field data were periodically collected to follow the phenological state of the fruits. In parallel, eight bands and four spectral indices from Sentinel-2 image time series were used, which are directly related to the foliage properties and activity, and indirectly to the fruit evolution. They were related to the variables measured from field samples: degrees Brix and pH. The degrees Brix models obtained with the spectral indices presented an R-2 of 69% and 73% in the 2017 and 2018 seasons, respectively. In pH modeling, the 2017 season had low R-2 results, reaching 43%, improving considerably in the 2018 season, reaching 63.8%. Estimated Brix and pH maps were obtained, expressing the spatial variability in the evolution of the fruit, which is useful for zoning the plots and to improve the sampling task prior to harvest. They are therefore a valuable tool to monitor the maturation, to improve the efficiency of harvest and subsequently, the quality of the wine.

To monitor the ripeness and composition of wine grape berries and establish an optimal harvest date, the determination of degrees Brix and pH is vital. This research studies two harvest seasons of Cabernet Sauvignon wine grapes: 2017 and 2018. Field data were periodically collected to follow the phenological state of the fruits. In parallel, eight bands and four spectral indices from Sentinel-2 image time series were used, which are directly related to the foliage properties and activity, and indirectly to the fruit evolution. They were related to the variables measured from field samples: degrees Brix and pH. The degrees Brix models obtained with the spectral indices presented an R-2 of 69% and 73% in the 2017 and 2018 seasons, respectively. In pH modeling, the 2017 season had low R-2 results, reaching 43%, improving considerably in the 2018 season, reaching 63.8%. Estimated Brix and pH maps were obtained, expressing the spatial variability in the evolution of the fruit, which is useful for zoning the plots and to improve the sampling task prior to harvest. They are therefore a valuable tool to monitor the maturation, to improve the efficiency of harvest and subsequently, the quality of the wine.