Aplicación de imágenes multiespectrales e indicadores clave monitorizados para optimizar la gestión eficiente de parcelas de viñedo ("Vitis vinifera" L.)

Abstract

According to ESA (European Space Agency), remote sensing is a way of collecting and analysing data to obtain information about an object, without the instrument used to collect the data being in direct contact with said object. This tool has proven useful in a wide range of fields, including agriculture, where the use of multispectral imagery has become widespread and could become an important tool to manage vineyards and fight against climate change. Furthermore, these images can be used alone or combined with other data for better results, providing helpful information on the state of crops. Four elements are essential in remote sensing: a platform, a target object, a sensor, and a way to use and store the information obtained. Nowadays, there are several platforms for obtaining information, such as satellites, drones, aircraft, and ground vehicles. Thus, data will be obtained with different spatial, temporal, spectral and radiometric resolution characteristics depending on the platform and sensor. Consequently, the cost will be different depending on the technology used. Remote sensing applications in agriculture are a recognised innovation with increasing potential. This tool can be used for various applications in a wide range of fields. In agriculture, the available information can be processed using vegetation indices. Similarly, it is possible to use a single image at a specific moment of the phenological cycle (usually veraison, which is related to the maximum amount of vegetation), or it is also possible to use all available images and work with time series. In viticulture, research studies show that remote sensing techniques allow the assessment of vineyard (Vitis vinifera L.) variability and the control of grape quality and quantity. Remote sensing has been successfully used to estimate several vineyard parameters, such as leaf area index (LAI). In this PhD thesis, Sentinel-2 satellite imagery was used to check if they were related to the agronomic and oenological parameters of several vineyards located in the Appellation of Origin Rueda, Valladolid. For this purpose, a time series of images was analysed, confirming that the phenological stage of veraison is a good moment for the use of the images. Field data was taken in each vineyard, and it was found that the satellite images were able to classify the vineyards according to their vegetative development, finding significant differences in several agronomic and quality parameters. In addition, a similar experiment was carried out on pistachio to check the applicability of the method, observing significant differences in yield. Finally, Landsat-8 images were used on several vineyards in Galicia. Field data related to yeast populations was compared using NDVI as an indicator of the amount of vineyard vegetation. As a result, significant differences were found concerning the plots and NDVI. On the other hand, to study the effect of mixed pixels in vineyards, an experimental trial was carried out in a vineyard where vines were progressively removed. Thus, satellite pixels were marked on the surface, and the removals were synchronized with the Sentinel-2 satellites imagery. The effect of the reduction of vegetation on the spectral information captured by the satellites was analysed (using NDVI). Then, the removed vegetation was carefully measured in the laboratory to check the exact leaf area, finding that for a trellised vineyard, every 20% reduction in the amount of vegetation meant a reduction of around 6% in NDVI. Additionally, before each vine removal, orthophotographs were taken with UAV and multispectral cameras to develop a novel method for estimating the leaf area of the vineyard (LAI) using the shadows of the plants projected on the ground. The flight time was carefully planned to maximise shadows, enabling pilots not only to use a new low-cost method with similar accuracy to other more expensive methods but also by providing flexibility when carrying out the work, as with this new method, pilots do not need to fly the drone in the solar midday. Finally, two comprehensive field studies were conducted in separate vineyards: one in the DO Rueda and the other in the DO Ribera del Duero in Spain. A sampling grid was created to try to capture the spatial variability of the vineyards, and Sentinel-2 imagery taken over the course of one year was employed to construct a time series and apply a functional principal component analysis (f-PCA). The results show that the two principal components explain most of the variability in the vineyard, and that from the third component onwards, the relationship between the components and the field parameters is not clear. On the other hand, it was found that f-PCA allowed better results than solely a veraison image, and each principal component explained the variability caused by different variables in the vineyard. In this doctoral thesis: i) the relationship between the spectral information obtained from the images and the vineyard parameters is quantified, ii) tools are implemented to establish differentiated vineyard management units, including those derived from Sentinel-2 images, iii) it is verified that the differences are transferred to the wines produced from these differentiated units, iv) the tools allow dynamic monitoring of the vineyards, v) they are remote sensing-based tools accessible to producers and low cost, and vi) they provide knowledge and present a useful product for the sector. The great challenge of this "digital era in viticulture" is to have professionals with sufficient training to take advantage of the immense opportunities of this technology and to offer practical solutions to farmers and winegrowers.