Neural networks and ft-ir spectroscopy for the discrimination of single varietal and blended wines. A preliminary study.

The first demonstration of the interest of carbon isotope composition of sugars in grapevine, as an integrated indicator of vineyard water status, dates back to 2000 (Gaudillère et al., 1999; Van Leeuwen et al., 2001). Thanks to the isotopic discrimination of Carbon that takes place during plant photosynthesis, under hydric stress conditions, it is possible to accurately estimate the photosynthetic activity. Ever since, δ13C has been widely applied with success to zonation, terroir studies and vine physiology research, but is still not widely used by viticulturists. This is quite astonishing by considering the impact of global warming on viticulture and the need to improve water management, that would justify a widespread use of δ13C.
The lack of private laboratories proposing the analysis, the cost of the technology, as well as the long analytical delays, have been detrimental to its development. Some laboratories tried to overcome the analytical difficulties of isotopic analysis by using fourier transformed infrared spectroscopy, as a fast and cheap alternative to the official OIV method (IRMS). These claimed FTIR models have never been published or peer reviewed and cannot be considered robust. In this work, thanks to the recent acquisition of IRMS technology, new modern and robust applications of δ13C for viticulture are proposed. This includes the use of the analysis to make parcel separations at harvesting, the possibility to increase the precision of hydric stress cartography and the potential cost reduction when compared with Scholander pressure bomb analysis.

Climate influence on grapevine physiology is prevalent and this influence is only expected to increase with climate change. Although governed by a general determinism, climate influence on grapevine physiology may present variations according to the terroir. In addition, these site-specific differences are likely to be enhanced when climate influence is studied using farm data. Indeed, farm data integrate additional sources of variation such as a varying representativity of the conditions actually experienced in the field. Nevertheless, there is a real challenge in valuing farm data to enable grape growers to understand their own terroir and consequently adapt their practices to the local conditions. In such a context, this article proposes a framework to site-specifically study climate influence on grapevine physiology using farm data. It focuses on improving the analysis of time series of weather data. The analytical framework includes the synchronization of time series using site-specific thermal indices computed with an original method called Extended Growing Degree Days (eGDD). Synchronized time series are then analyzed using a Bayesian functional Linear regression with Sparse Steps functions (BLiSS) in order to detect site-specific periods of strong climate influence on yield development. The article focuses on temperature and rain influence on grape yield development as a case study. It uses data from three commercial vineyards respectively situated in the Bordeaux region (France), California (USA) and Israel. For all vineyards, common periods of climate influence on yield development were found. They corresponded to already known periods, for example around veraison of the year before harvest. However, the periods differed in their precise timing (e.g. before, around or after veraison), duration and correlation direction with yield. Other periods were found for only one or two vineyards and/or were not referred to in literature, for example during the winter before harvest.

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.

Making high quality but affordable Pinot noir (PN) wine is challenging in most terroirs and New Zealand’s (NZ) situation is no exception. To increase the probability of making highly typical PN wines producers choose to grow grapes in cool climates on lower fertility soils while adopting labour intensive practices. Stringent yield targets and higher input costs necessarily mean that PN wine cost is high, and profitability lower, in line-priced varietal wine ranges. To understand the reasons why higher yielding vines are perceived to produce wines of lower quality we have undertaken an extensive study of PN in NZ. Since 2018, we established a network of twelve trial sites in three NZ regions to find individual vines that produced acceptable commercial yields (above 2.5kg per vine) and wines of composition comparable to “Icon” labels. Approximately 20% of 660 grape lots (N = 135) were selected from within a narrow juice Total Soluble Solids (TSS) range and made into single vine wines under controlled conditions. Principal Component Analysis of the vine, berry, juice and wine parameters from three vintages found grape berry mass to be most effective clustering variable. As berry mass category decreased there was a systematic increase in the probability of higher berry red colour and total phenolics with a parallel increase in wine phenolics, changed aroma fraction and decreased juice amino acids. The influence of berry size on wine composition would appear stronger than the individual effects of vintage, region, vineyard or vine yield. Our observations support the hypothesis that it is possible to produce PN wines that fall within an “Icon” benchmark composition range at yields above 2.5kg per vine provided that the Leaf Area:Fruit Weight ratio is above 12cm2 per g, mean berry mass is below 1.2g and juice TSS is above 22°Brix.

In viticulture, a warming climate can have a very significant impact on grapevine development and therefore on the quality and characteristics of wines across different spatial scales, ranging from global to local. In order to adapt wine-growing to climate change, global climate models can be used to define future scenarios, but only at the scale of major wine regions. Despite the huge progress made over the last ten years in terms of the spatial resolution of climate models (now downscaled to a few square kilometres), they are not yet sufficiently precise to account for the local climate variability associated with such parameters as local topography, in spite of these parameters being decisive for vine and wine characteristics. This study describes a method to downscale future climate scenarios to vineyard scale. Networks of data loggers have been used to collect air temperature at canopy level in the Waipara winegrowing region (New Zealand) over five growing seasons. These measurements allow the creation of fine-scale geostatistical models and maps of temperature (at 100 m resolution) for the growing season. In order to model climate change at pilot site scale, these geostatistical models have been combined with regional climate change predictions for the periods 2031-2050 and 2081-2100 based on the RCP8.5 climate change scenario. The integration of local climate variability with regionalized climate change simulations allows assessment of the impacts of climate change at the vineyard scale. The improved knowledge gained using this methodology results from the increased horizontal resolution that better addresses the concerns of winegrowers. The results provide the local winegrowers with information necessary to understand current processes, as well as historical and future viticulture trends at the scale of their site, thereby facilitating decisions about future response strategies.