Contents lists available at ScienceDirect Food Chemistry journal homepage: www.elsevier.com/locate/foodchem Towards on-line monitoring of phenolic content in red wine grapes: A feasibility study Jose Luis Aleixandre-Tudo a,b, ⁎ , Helene Nieuwoudt c , Wessel du Toit a a Department of Viticulture and Oenology, Stellenbosch University, Private Bag X1, Matieland, 7602, South Africa b Departamento de Tecnologia de Alimentos, Universidad Politécnica de Valencia, Camino de Vera s/n, 46022 Valencia, Spain c Institute for Wine Biotechnology, Stellenbosch University, Private Bag X1, Matieland, 7602, South Africa ARTICLE INFO Keywords: Spectroscopy PLS regression On-line monitoring Conveyor belt Phenolic compounds Wine grapes ABSTRACT Spectroscopy techniques to efficiently measure phenolic composition in grape berries may be a suitable ana- lytical practice, provided that robust calibrations are established. A contactless FT-NIR instrument was used for on-line spectral data collection from grapes transported on a conveyor belt system. Spectral data was also col- lected on static samples using the same NIR instrument. Spectral measurements of crushed berries captured from the conveyor belt system and the use of the homogenate extraction protocol as reference method provided the most accurate prediction models. Values obtained for errors in prediction (RMSEP%) and RPD were 12% and 2.37, 12.3% and 3.37, 7.8% and 3.2, 16.7% and 2.84 for tannins (mg/g) and anthocyanins (mg/g) on a fresh weight basis, total phenols and colour density (AU), respectively. The results observed in this study show the ability of NIR spectroscopy to monitor the phenolic composition of grape berries transported on a conveyor belt system online. 1. Introduction The biosynthesis and accumulation of phenolic substances in grapes take place throughout the ripening season. Anthocyanins and flavan-3- ol polymers, the so called proanthocyanidins, are the major flavonoids found in the inner thick-walled layers of the grape berry skin tissue (Adams, 2006). Additionally, proanthocyanidins are also found in large concentrations in the soft parenchyma of the grape seed coat tissue (Adams, 2006). Flavan-3-ols and their polymeric forms, proanthocya- nidins, start accumulating in the berry from blooming, reaching a maximum at véraison with a subsequent decrease during the last stages of the ripening period (Teixeira, Eiras-Dias, Castellarin, & Gerós, 2013). In contrast to the flavan-3-ol phenolics, anthocyanins start accumu- lating in the skin tissue from véraison and reach a maximum during the late stages of ripening, when the synthesis stops (Teixeira et al., 2013). The strong influence of phenolic substances on wine quality is due to its importance in the colour and mouthfeel attributes. Anthocyanins in their different forms and structures are mainly responsible for red wine colour (He et al., 2012), while flavan-3-ols and proanthocyanidins play a major role in wine bitterness and astringency (McRae & Kennedy, 2011). It is still very common to harvest wine grapes taking only the sugar to acid ratio into account. Often, only sugar concentration is used as a harvesting parameter, as tartaric acid additions during winemaking can be used to adjust this ratio if allowed. It is also not common to take other wine components, such as phenolic compounds, into account when deciding on the optimum harvesting date, for benchmarking purposes or for decision-making practices prior to the fermentation process (Nogales-Bueno, Hernández-Hierro, Rodríguez-Pulido, & Heredia, 2014). The measurement of phenolics in grape berries is a lengthy process, as the phenolic compounds first need to be extracted from the berry material before analysis. Despite recent advances in the field with the inclusion of hand-held devices, analytical techniques for fast and easy-to-operate methods for the quantification of phenolic le- vels in grape berries are still not readily available in the wine industry (Dambergs, Gishen, & Cozzolino, 2015). To overcome this, the use of spectroscopy with chemometrics tools may provide a suitable solution (Dambergs et al., 2015). Spectroscopy applications combine chemical information obtained from the emission of light (UV–Vis, IR and Raman, among others) with the measurement of the compounds or parameters of interest following established reference methods. This information is then correlated, in most cases using regression techni- ques, with the final goal to obtain prediction algorithms that are used to estimate phenolic levels in new samples using only spectral information (Aleixandre-Tudo, Buica, Nieuwoudt, Aleixandre, & Du Toit, 2017). A substantial reduction in the analysis time is thus achieved, with this https://doi.org/10.1016/j.foodchem.2018.07.118 Received 18 March 2018; Received in revised form 17 July 2018; Accepted 17 July 2018 ⁎ Corresponding author at: Department of Viticulture and Oenology, Stellenbosch University, Private Bag X1, Matieland, 7602, South Africa. E-mail address: joaltu@upvnet.upv.es (J.L. Aleixandre-Tudo). Food Chemistry 270 (2019) 322–331 Available online 18 July 2018 0308-8146/ © 2018 Elsevier Ltd. All rights reserved. T