Hyperspectral imaging and machine learning for monitoring grapevine physiology

Herbicides are commonly sprayed in the vine row to prevent competition with vines for water and minerals and to keep weeds from growing into the bunch zone. Sprays are applied before budbreak and reapplied multiple times during the season to keep the undervine bare. There is growing concern about the negative effects of herbicides on humans and the environment, and weeds in New Zealand have developed resistance to herbicides. Therefore, it is imperative that we reduce our reliance on herbicides in viticulture and incorporate methods that do not engender resistance.

Endogenic diseases often arise from pathogens that exist within the plant tissue, including fungi, bacteria, and viruses, which can remain latent and then emerge under stress conditions or favorable environmental conditions, causing symptoms that weaken vines or can lead to plant death.

“Flavescence dorée” (FD) is a grapevine quarantine disease associated with phytoplasmas and transmitted to healthy plants by insect vectors, mainly Scaphoideus titanus. Infected plants usually develop symptoms of stunted growth, unripe cane wood, leaf rolling, leaf yellowing or reddening, and shrivelled berries. Since plants can remain symptomless up to four years, they may act as reservoirs of FD contributing to the spread of the disease. So far, conventional management strategies rely mainly on the insecticide treatments, uprooting of infected plants and use of phytoplasma-free propagation material. However, these strategies are costly and could have undesirable environmental impacts. Thus, the development of sustainable and noninvasive approaches for early detection of FD and its management are of great importance to reduce disease spread and select the best cultural practices and treatments. The present study aimed to evaluate if multispectral/hyperspectral technologies can be used to detect FD before the appearance of the first symptoms and if infected grapevines display a spectral imaging fingerprint. To that end, physiological parameters (leaf area, chlorophyll content and photosynthetic rate) were collected in concomitance to the measurements of plant reflectance (using both a portable apparatus and a remote sensing drone). Measurements were performed in two leaves of 8 healthy and 8 FD-infected grapevines, at four timepoints: before the development of disease symptoms (21st June); and after symptoms appearance (ii) at veraison (2nd August); at post-veraison (11th September); and at harvest (25th September). At all timepoints, FD infected plants revealed a significant decrease in the studied physiological parameters, with a positive correlation with drone imaging data and portable apparatus analyses. Moreover, spectra of either drone imaging and portable apparatus showed clear differences between healthy and FD-infected grapevines, validating multispectral/ hyperspectral technology as a potential tool for the early detection of FD or other grapevine-associated diseases.

Hydrogen sulfide and methanethiol are recognised as two of the most significant contributors to reductive off-flavours in wine.

Vine clone selection aims at the survival of clones with particularly desireable attributes for the production of high quality wines. The purpose of this research was to study the enological potential of the clones of Greek indigenous grape varieties over two vintages, 2018 and 2019.
METHODS: Two clones of the white grape varieties Moschofilero (E26 and E27), Assyrtiko (E11 and 16), Roditis (25E16 and 02E1E21) and two clones of the red grape varieties Xinomavro (19 and E2E30) and Agiorgitiko (03E40 and 41E47) were vinified under the same protocol for the white wines and common for the red wines in 2018 and 2019. The resulting products were studied for several enological parameters such as alcohol content, volatile acidity, pH, total phenolics, anthocyanins and tannins for the red wines, as well as browning tests for the white wines. The aroma profile of these ten samples was investigated through sensory analysis with intensity rating of individual attributes on a five-point scale by a trained panel.