The informative potential of remote and proximal sensing application on vertical- and overhead-trained vineyards in Northeast Italy

Abstract

Context and purpose of the study

The application of remote and proximal sensing in viticulture have been demonstrated as a fast and efficient method to monitor vegetative and physiological parameters of grapevines. The collection of these parameters could be highly valuable to derive information on associated yield and quality traits in the vineyard. However, to leverage the informative potential of the sensing systems, a series of preliminary evaluations should be carried out to standardize working protocols for the specific features of a winegrowing area (e.g., pedoclimate, topography, cultivar, training system). This work aims at evaluating remote and proximal sensing systems for their performance and suitability to provide information on the vegetative, physiological, yield and qualitative aspects of vines and grapes as a function of different training systems in the Valpolicella wine region (Verona, Italy).

Material and methods

Five vineyards in the Valpolicella wine region were investigated for their intra-parcel variability during 2022 growing season. Three vineyards were trained with cane pruning vertical shoot positioning system (Guyot), while the other two were trained with cane pruning overhead system (Pergola). Blocks presenting intra-parcel variability were selected and monitored in each vineyard. The Normal Difference Vegetation Index (NDVI) was calculated using both the data of remote sensors such as the satellite Sentinel-2 and UAV-mounted multispectral camera, and a proximal handheld NDVI device. Further proximal sensor evaluation was carried out employing a handheld thermal camera, which estimates the Crop Water Stress Index (CWSI). The data collected from the sensors was then compared with that of direct measurements on the vines and the berries (e.g., bud fertility, shoot growth kinetics, leaf area, yield, berry skin thickness and technological berry ripening parameters). Multivariate and correlation analyses were applied to determine the relationship between the sensor data and the direct vine and berry measurements and to further evaluate the nature of these relationships as a function of the vine training system.

Results

Multivariate analyses on the whole dataset distinguished the Guyot-trained blocks from the Pergola-trained blocks. Positive correlations emerged between the NDVI values obtained from the satellite images, the UAV images and the proximal NDVI sensor, which were ground-truthed by obtaining high positive correlations with a series of direct measurements, among which the bud fertility, the shoot growth kinetic, the leaf area and the crop yield. The vigor data correlated negatively with quality berry parameters such as the sugar and the polyphenolic content. The strength of the detected relationships varied as a function of the training system, suggesting different informative potential of the tested sensor systems for Guyot and Pergola.