Geodaysit 2023

Peppino Sarasso


Monitoring Erbaluce and Nebbiolo vineyards by means of Sentinel-2 NDVI index maps
Enrico Borgogno-Mondino, Alberto Cugnetto, Giorgio Masoero, Peppino Sarasso

The advent of satellite technologies has made it possible to make georeferenced observations of the entire globe at periodic intervals of a few days and with high spatial resolutions.
ESA's Copernicus mission makes available open-source data from the Sentinel-2 constellation created to provide useful information for agricultural purposes thanks to appropriately calibrated multispectral images [2].
The NDVI (Normalized Vegetation Index) [1] can be correlated with some biophysical or agronomic variables of the vineyard [3].
The work presents the results of a two-year work carried out in the province of Turin in the Piedmont region, that involved six vineyards cultivated with different varieties (Nebbiolo, Erbaluce) and two vine training system (pergola and espalier). The NDVI georeferenced data were provided by the EOS Crop Monitoring web platform.
The experimental design divided the vineyards in three classes of vigor areas, defined through a pre-survey operated by comparing the series of georeferenced NDVI images collected the summer before.
In the different vineyards for each of the chosen vigor areas, five plants were identified and used as a ground reference to evaluate a series of vegetative-productive parameters. The total amount of plants monitored were 30 for Nebbiolo and 55 for Erbaluce.
All NDVI index showed significant predictability for the studied variables.
As expected, the trend of the quantitative variables was positively related to the NDVI while the qualitative variables were negatively related. As far as the percentage mean error was concerned a high predictability, (error 1÷7% respectively for Erbaluce and Nebbiolo vineyards). Considering the canopy architecture, the leaf layers were accurately predicted from the NDVI (R2 0,72 and 0,55 respectively for Erbaluce and Nebbiolo) with an error around 10%. Regarding the fruit compartment a strong difference emerged between the systems. The shaded cluster percentage in the Nebbiolo vines was highly predictable with (R2 0,57, error 6%). In Erbaluce the error was higher (36%) with a correlation index R2 of 0,42. This fact derives from the higher variability of the plants in the compared plots. The number of clusters were predicted with a minor error in Nebbiolo than in Erbaluce (9% and 29%, R2 0,70 and 0,16 respectively) and for the bud fertility (8% and 15%, R2 0,83 and 0,36 respectively). In sum, the true productive traits appeared as the less predictable in the Erbaluce vineyards, with 31% error in yield (R2 0,26) compared to a less erroneous prediction (error 22% and R2 0,63) in Nebbiolo vines. The pruning wood weight was similarly predicted from the NDVI with 21 and 23% error, with a correlation index R2 of 0,41 and 0,28 for Erbaluce and Nebbiolo respectively.
The PCA analysis, allowed discriminating observations based on vigor attributes and consistently with the measured variables, even when all the observations, for the different varietal combinations, are processed simultaneously with the same multivariate model.
The study confirmed the possibility to use Sentinel-2 NDVI output to map the vineyards variability also in small plots (< 1 ha), estimating the vineyard canopy density, the productive and wine most important technological parameters.

[1] Giovos, R., Tassopoulos, D., Kalivas, D., Lougkos, N., & Priovolou, A. (2021). Remote sensing vegetation indices in viticulture: A critical review. Agriculture, 11(5), 457.
[2] Sarvia, F., De Petris, S., Orusa, T., & Borgogno-Mondino, E. (2021). MAIA S2 versus sentinel 2: spectral issues and their effects in the precision farming context. In Computational Science and Its Applications–ICCSA 2021: 21st International Conference, Cagliari, Italy, September 13–16, 2021, Proceedings, Part VII 21 (pp. 63-77).
[3] Vélez, S., Rançon, F., Barajas, E., Brunel, G., Rubio, J. A., & Tisseyre, B. (2022). Potential of functional analysis applied to Sentinel-2 time-series to assess relevant agronomic parameters at the within-field level in viticulture. Computers and Electronics in Agriculture, 194, 106726.

AIT Contribution
Sala Biblioteca @ PoliBa