Shelf life of white and rosé wines: mechanistic modelling and development of predictive approaches based on electrochemical and spectroscopic sensors

The management of the shelf-life of white and rosé wines is a critical challenge for winemakers. The stability of bottled wines is determined by a complex interaction between the intrinsic composition of the matrix, packaging characteristics and environmental storage conditions. Among the main degradation factors, chemical oxidation and photo-oxidation play a predominant role in compromising aromatic freshness and color stability. Despite extensive literature, the practical management of these phenomena still relies largely on empirical approaches. A unified kinetic framework describing how oxygen and light interact, and how wine-to-wine variability modulates susceptibility, is still lacking. At the same time, there is a growing industrial demand for rapid, non-destructive and predictive analytical tools able to anticipate wine evolution and support cellar decisions. This doctoral work addressed these gaps through two complementary research axes. The first focused on the mechanistic modelling of oxidative and photo-oxidative degradation using multifactorial experimental designs and metabolomic approaches. The second aimed at developing predictive models based on rapid sensors: cyclic voltammetry, Fourier transform infrared spectroscopy (FTIR) and in-bottle visible-near infrared spectroscopy (VIS-NIR), to classify the oxidative and photo-oxidative risk of individual wines. In the first part, the combined impact of the stress factors light, oxygen, temperature and time on rosé wines was studied using a response surface methodology design. The results showed that wine response to stress did not depend on single factors in isolation but on their interactions, and that wines with lower content of free sulfur dioxide (SO2) were markedly more susceptible to chemical and sensory changes. From a chromatic perspective, light exposure promoted an increase in lightness (L*) consistent with bleaching of anthocyanins, whereas the interaction between light and oxygen favored browning (decrease in L*), in agreement with polymeric pigment formation. As for aroma composition, the formation of methanethiol and dimethyl disulfide was confirmed as a marker of the light-struck defect. A key finding was the behavior of norisoprenoids such as TDN and vitispirane, whose degradation followed two distinct routes: a riboflavin-mediated photo-oxidation pathway and a direct photolysis pathway. This indicates that riboflavin-driven chemistry alone does not fully account for varietal-aroma degradation under light exposure, a finding that has implications for the design of light-stress management strategies. A second mechanistic study examined the kinetic interaction between light and dissolved oxygen in a Chardonnay wine. Light exposure accelerated oxygen consumption dramatically, leading to complete depletion in less than 48 hours compared with approximately 7 days in the dark. Under illumination, oxygen consumption followed apparent zero-order kinetics, consistent with photo-sensitised generation of singlet oxygen through a riboflavin mediated Type II photosensitization mechanism. A central result was the “dual” role of oxygen. Its presence strongly suppressed methanethiol formation, reducing its accumulation by about threefold compared with anoxic conditions. At the same time, oxygen promoted the formation of reactive species that triggered alternative degradation routes, as confirmed by untargeted metabolomics, which revealed clusters of metabolites with divergent behaviors under oxic and anoxic conditions. The second research axis addressed the prediction of oxidative behavior under controlled stress. Cyclic voltammetry on disposable screen-printed electrodes was used to characterize 31 white wines and 32 rosé wines. The experiments confirmed that SO2 consumption is not strictly stoichiometric with oxygen but is strongly wine specific: for a given amount of oxygen consumed, the loss of free SO2 varied widely between samples. A non-linear predictive model combining pH, initial free SO2 and selected voltammetric features was able to predict the final free SO2 in white wines with a root mean square error of ≈ 2.9 mg/L. External validation confirmed that the electrochemical fingerprint can be used as a rapid screening tool to optimise antioxidant additions before bottling. In contrast, the same strategy applied to rosé wines showed limited predictive performance, suggesting that their more complex phenolic matrix requires larger calibration datasets and possibly different model structures. The final part of the work investigated the light-strike risk of 36 commercial rosé wines subjected to a standardized light-stress protocol. The results revealed a widespread vulnerability: about half of the wines exceeded the commonly used acceptability threshold for color difference and in all of them methanethiol increased significantly, although to very different extents. Using ridge regression with bootstrap-based variable selection, robust models were developed to predict methanethiol formation. The best-performing model integrated basic compositional parameters with selected FTIR spectral features and allowed estimation of the defect risk without the need for targeted chromatographic analyses of riboflavin or methionine. In parallel, a non-destructive in-bottle VIS-NIR method was validated. Analysis of spectral differences in the blue region enabled correct classification of light-exposed versus dark-stored wines with accuracies ≈ 98 percent. Overall, this thesis shows that the stability of white and rosé wines is governed by wine-specific mechanisms that cannot be inferred from single compositional markers such as free SO2 or color alone. The integrated approach adopted here (i) clarifies that, under the conditions studied, oxygen can strongly reduce light-strike related methanethiol formation while simultaneously accelerating oxidative ageing through distinct pathways, (ii) proposes cyclic voltammetry as a rapid tool to tailor SO2 management to the specific reactivity of white wines, and (iii) demonstrates that FTIR and in-bottle VIS–NIR spectroscopy, combined with chemometrics, provide practical solutions for preventive screening of light-strike risk and non-destructive quality control on the finished product. These results support a transition from empirical shelf-life management towards predictive, data-driven strategies aimed at preserving wine quality throughout distribution chain.