Foliar and berry hyperspectral reflectance predicts winegrape berry composition across developmental stages and varieties

Grapevine berry chemistry varies across varieties and developmental stages and is sensitive to environmental conditions. Efforts have sought to leverage proximal remote sensing data to exploit potential covariation of foliar traits and berry composition to enable nondestructive and scalable methods for assessing berry chemistry. Traditionally, simple vegetation indices have been used based on the assumption of a relationship between foliar coverage and berry yield and composition. However, indices assessing foliar coverage, chlorophyll content, and biomass may not be sensitive enough to track subtle variations of berry chemistry across winegrape varieties and developmental stages. Thus, in this study, we seek to leverage full-range hyperspectral data (400–2500 nm) and a partial least squares regression (PLSR) model for assessing traditional metrics of berry composition. The benefit of PLSR models for analysis of hyperspectral data is that they can integrate subtle physiological and structural changes in plant reflectance to optimize a model. In this study, we measured leaf and berry spectra and berry composition (Brix, tartaric acid, and pH) across 23 winegrape varieties and developmental stages across 2 years from June to September in a common garden in Davis, California. Our results show that both foliar and berry hyperspectral PLSR models can be used to predict berry composition across phenological stages. This suggests that proximal remote sensing of foliage has the potential to enable rapid and nondestructive monitoring of berry developmental stages to aid in management decision-making and harvest timing.