Quantitative prediction of grape ripening parameters combining an autonomous IoT spectral sensing system and chemometrics

• Introduction of an IoT spectral sensing system for remote vineyard monitoring. • Analysis of multivariate model development for predicting grape ripening parameters. • Investigation of optical data pre-processing techniques for enhanced model accuracy. • System performance evaluation against existing analytical approaches. • Advancements in vineyard management through fast, non-destructive ripening assessment. The research presented in this study offers a contribution to the field of viticulture by testing at lab scale an innovative approach for monitoring grape ripening using an autonomous proximal sensing technology. By leveraging an IoT spectral sensing system, termed i-Grape, the research aims to remotely monitor vineyards and provide real-time data on grape ripening status. This system, consisting of tailored optical, host, and controller modules, offers a novel solution for continuous monitoring throughout the crop season, overcoming limitations associated with traditional sampling methods. The study conducted comprehensive sampling in the viticulture area of the Douro Valley, collecting data from cv. Touriga Nacional and Touriga Franca. Both optical and wet-chemistry analyses were performed on the grape samples to develop predictive models for ripening parameters, including Total Soluble Solids (TSS), Potential Alcohol (PA), pH, Titratable Acidity (TA), Total Polyphenols (TP), and Extractable Anthocyanins (EA). Exploratory analysis of the optical data revealed insights into the behaviour of the spectral readouts over time, highlighting the evolution of grape ripening and the potential interference factors that need to be addressed for accurate modelling. Pre-processing techniques, including background subtraction and Log10 transformation, were employed to enhance the quality of the optical data and improve model performance. Overall, predictive PLS models with good performance were obtained for the estimation of the technological ripening parameters (RPD = 2.76 and R 2 = 0.86 for TSS; RPD = 2.58 and R 2 = 0.85 for PA; RPD = 3.65 and R 2 = 0.92 for TA; RPD = 2.27 and R 2 = 0.79 for pH), establishing a solid ground for the application of this sensing strategy in the field. For the phenolic parameters (TP and EA), the performance of the models is still insufficient (RPD = 1.28 and R 2 = 0.51 for TP; RPD = 1.55 and R 2 = 0.58 for EA). A comparison with existing literature highlighting the advancements achieved in terms of predictive performance and operational capabilities has been reported. The potential of the i-Grape system to revolutionize grape ripening monitoring by offering a cost-effective, non-destructive, and scalable solution for vineyard management has been demonstrated at lab scale. In conclusion, the research laid the groundwork for further advancements in optical sensing technology for viticulture, opening up avenues for future research in optimizing hardware design, data processing algorithms, and field implementation strategies to realize the full potential of IoT-based solutions in precision agriculture.