Portable NIR spectroscopy for nutrient profiling in rootstock and scion material: enhancing decision-making in the grafting industry

Abstract

The success of grafting in viticulture is deeply influenced by the nutrient composition of both rootstock and scion materials. Key components such as nitrogen and carbohydrates play a crucial role in graft compatibility, establishment, and overall plant vigor [1]. Traditional methods for assessing nutrient content are often time-consuming, destructive, and resource-intensive, limiting their application for rapid decision-making in commercial settings. This leads to grafting industries often adopting a non-selective grafting approach, where the grafting is attempted without prior assessment of viability (meaning whether or not the rootstock or scion has low nutrient content), relying solely on postgrafting establishment outcomes or vine vigour results to determine success. This method leads to inefficiencies, resource wastage, and variability in plant quality. Near-Infrared (NIR) spectroscopy has emerged as a non-destructive, rapid, and reliable tool for compositional analysis across various agricultural sectors. This study explores the application of NIR spectroscopy for assessing nutrient profiles in rootstock and scion materials, aiming to support data-driven grafting decisions that maximise resource use and optimise plant quality.

The study employed three NIR spectrometers: a benchtop unit, a commercially available portable spectrometer, and a handheld prototype device, to evaluate their performance in analysing nutrient content in grapevine rootstock and scion samples. Samples were collected from three grapevine nurseries, each providing material from different sites, cultivars, and clones, ensuring a diverse and representative sample set. The final dataset consisted of a thousand samples, which were analysed both on the outer layer and on the cross-sectional area of the shoot, following a fresh transverse cut, to avoid bark influence in the spectra. This diversity in samples allowed for a robust validation strategy using an independent dataset that maintained the proportion of different cultivars, ensuring the model’s reliability across varied genetic material. Multivariate calibration models were developed for each instrument using Partial Least Squares (PLS) Regression. However, due to the high volume of spectral data, not every sample underwent wet-chemical analysis. Instead, a recursive PLS approach was applied, where an initial subset of samples was chemically analysed, and the resulting model predictions were used to predict additional samples, which were then added to the model [2].

The study demonstrated promising results in predicting nitrogen and carbohydrate content using all three NIR spectrometers. The benchtop instrument exhibited the highest accuracy, but both the portable and handheld devices achieved acceptable predictive performance, making them suitable for industry applications. The resulting errors remained within acceptable thresholds for practical decision-making in the grafting industry. These findings highlight the potential of NIR spectroscopy as a rapid and non-destructive tool for nutrient profiling in rootstock and scion materials, enabling informed decisions that can improve graft success rates, optimise resource allocation, and ultimately enhance plant quality.

References

[1] Vršič, S., Pulko, B., & Kocsis, L. (2015). Scientia Horticulturae, 181, 168–173.

[2] van Wyngaard, E., Blancquaert, E., Nieuwoudt, H., & Aleixandre-Tudo, J. L. (2023). Biosystems Engineering, 232, 141-154.