Modelo de Concentração de Açúcar de Videira (GSR): Uma Ferramenta de Apoio à Decisão para a Região Vitinícola do Alto Douro
Climate-smart agriculture involves practices and crop modelling techniques aiming to provide practical answers to meet growers’ demands. For viticulturists, early prediction of harvest dates is critical for the success of cultural practices, which should be based on accurate planning of the annual growing cycle. We developed a modelling tool to assess the sugar concentration levels in the Douro Superior sub-region of the Douro wine region, Portugal. Two main cultivars (cv. Touriga-Nacional and Touriga-Francesa) grown in five locations across this sub-region were studied. Grape berry sugar data, with concentrations between 170 and 230 g L1, were analyzed for the growing season campaigns, from 2014–2020, as an indicator of grape ripeness conditioned by temperature factors. Field data were collected by ADVID (“Associação Desenvolvimento Da Viticultura Duriense”), a regional winemaker association, and by Sogrape, the leading wine company from Portugal. The “Phenology Modeling Platform” was used for calibrating the model with sigmoid functions. Subsequently, model optimizations were performed to achieve a harmonized model, suitable for all estates. Model performance was assessed through two metrics: root mean square error (RMSE) and the Nash–Sutcliffe coefficient of efficiency (EFF). Both a leave-one-out cross-validation and a validation with an independent dataset (for 1991–2013) were carried out. Overall, our findings demonstrate that the model calibration achieved an average EFF of 0.7 for all estates and sugar levels, with an average RMSE < 6 days. Model validation, at one estate for 15 years, achieved an R2 of 0.93 and an RMSE < 5. These models demonstrate that air temperature has a high predictive potential of sugar ripeness, and ultimately of the harvest dates. These models were then used to build a standalone easy-to-use computer application (GSCM—Grapevine Sugar Concentration Model), which will allow growers to better plan and manage their seasonal activities, thus being a potentially valuable decision support tool in viticulture and oenology.