Grape variety identification and detection of terroir effects from satellite images

Currently, the main goal of agriculture is to promote the resilience of agricultural systems in a sustainable way through the improvement of use efficiency of farm resources, increasing crop yield and quality under climate change conditions. This last is expected to drastically modify plant growth, with possible negative effects, especially in arid and semi-arid regions of Europe on the viticultural sector. In this context, the monitoring of spatial behavior of grapevine during the growing season represents an opportunity to improve the plant management, winegrowers’ incomes, and to preserve the environmental health, but it has additional costs for the farmer. Nowadays, UAS equipped with a VIS-NIR multispectral camera (blue, green, red, red-edge, and NIR) represents a good and relatively cheap solution to assess plant status spatial information (by means of a limited set of spectral vegetation indices), representing important support in precision agriculture management during the growing season. While differences between UAS-based multispectral imagery and point-based spectroscopy are well discussed in the literature, their impact on plant status estimation by vegetation indices is not completely investigated in depth. The aim of this study was to assess the performance level of UAS-based multispectral (5 bands across 450-800nm spectral region with a spatial resolution of 5cm) imagery, reconstructed high-resolution satellite (Sentinel-2A) multispectral imagery (13 bands across 400-2500 nm with spatial resolution of <2 m) through Convolutional Neural Network (CNN) approach, and point-based field spectroscopy (collecting 600 wavelengths across 400-1000 nm spectral region with a surface footprint of 1-2 cm) in a plant status estimation application, and then, using Bayesian regularization artificial neural network for leaf chlorophyll content (LCC) and plant water status (LWP) prediction. The test site is a Greco vineyard of southern Italy, where detailed and precise records on soil and atmosphere systems, in-vivo plant monitoring of eco-physiological parameters have been conducted.

Monitoring vine water status is a major challenge for vineyard management because it influences both yield and harvest quality. It is also a challenge at the territorial scale for identifying periods of high water restriction or zones regularly impacted by water stress. This information is of major importance for defining collective strategies, anticipating harvest logistic or applying for irrigation authorisation. At this spatial scale, existing tools and methods for monitoring vine water status are few and often require strong assumptions (e.g. water balance model). This paper proposes to consider a collaborative collection of observations by winegrowers and wine industry stakeholders (crowdsourcing) as an interesting alternative. Indeed, it allows the collection of a large number of field observations while pooling the collection effort. However, the feasibility of such a project and its interest in monitoring vine water status at regional scale has never been tested.

The objective of this article is to explore the possibility of making a regional map of vine water status based on crowdsourcing observations. It is based on the study of the free mobile application ApeX-Vigne, which allows the collection of observations about vine shoot growth. This information is easy to collect and can be considered, under certain conditions, as a proxy for vine water status. This article presents the first results obtained from the nearly 18,000 observations collected by winegrowers and wine industry stakeholders during 2019, 2020 and 2021 seasons. It presents the vine shoot growth maps obtained at regional scale and their evolution over the three vintages studied. It also proposes an analysis of the factors that favoured the number of observations collected and those that favoured their quality. These results open up new perspectives for monitoring vine water status at a regional scale but above they provide references for other crowdsourcing projects in viticulture.

Optimizing the use of water and improving irrigation strategies has become increasingly important in most winegrowing countries due to the consequences of climate change, which are leading to more frequent droughts, heat waves, or alteration of precipitation patterns. Optimized irrigation scheduling can only be based on a reliable knowledge of the vineyard water status.

In this context, this work aims at the development of a novel methodology, using a contactless, miniaturized, low-cost NIR spectral tool to monitor (on-the-go) the vineyard water status variability. On-the-go spectral measurements were acquired in the vineyard using a NIR micro spectrometer, operating in the 900–1900 nm spectral range, from a ground vehicle moving at 3 km/h. Spectral measurements were collected on the northeast side of the canopy across four different dates (July 8th, 14th, 21st and August 12th) during 2021 season in a commercial vineyard (3 ha). Grapevines of Vitis vinifera L. Graciano planted on a VSP trellis were monitored at solar noon using stem water potential (Ψs) as reference indicators of plant water status. In total, 108 measurements of Ψs were taken (27 vines per date).

Calibration and prediction models were performed using Partial Least Squares (PLS) regression. The best prediction models for grapevine water status yielded a determination coefficient of cross-validation (r2cv) of 0.67 and a root mean square error of cross-validation (RMSEcv) of 0.131 MPa. This predictive model was employed to map the spatial variability of the vineyard water status and provided useful, practical information towards the implementation of appropriate irrigation strategies. The outcomes presented in this work show the great potential of this low-cost methodology to assess the vineyard stem water potential and its spatial variability in a commercial vineyard.

The research includes an analysis of the impact of weather conditions on phenological development of the vine and grape quality, through monitoring of four experimental cultivars (Chardonnay, Graševina, Merlot and Plavac mali) over two production years. In each experimental vineyard, which were evenly distributed throughout the regions of Slavonia and The Croatian Danube, Croatian Uplands,

A large varietal collection including over 1700 varieties was maintained in Conegliano, ITA, since the 1950s. Phenological data on a subset of 400 grape varieties including wine grapes, table grapes, and raisins were acquired at bud break, flowering, veraison, and ripening since 1964. Despite the efforts in maintaining and acquiring data over such an extensive collection, the data set has varying degrees of missing cases depending on the variety and the year. This is ubiquitous in phenology datasets with significant size and length. In this work, we evaluated four state-of-the-art methods to estimate missing values in this phenological series: k-Nearest Neighbour (kNN), Multivariate Imputation by Chained Equations (mice), MissForest, and Bidirectional Recurrent Imputation for Time Series (BRITS). For each phenological stage, we evaluated the performance of the methods in two ways. 1) On the full dataset, we randomly hold-out 10% of the true values for use as a test set and repeated the process 1000 times (Monte Carlo cross-validation). 2) On a reduced and almost complete subset of varieties, we varied the percentage of missing values from 10% to 70% by random deletion. In all cases, we evaluated the performance on the original values using normalized root mean squared error. For the full dataset we also obtained performance statistics by variety and by year. MissForest provided average errors of 17% (3 days) at budbreak, 14% (4 days) at flowering, 14.5% (7 days) at veraison, and 17% (3 days) at maturity. We completed the imputations of the Conegliano dataset, one of the world’s most extensive and varied phenological time series and a steppingstone for future climate change studies in grapes. The dataset is now ready for further analysis, and a rigorous evaluation of imputation errors is included.