Performance of seasonal forecasts of Douro and Port wine production

• Autor(a) principal: Santos, J. A.
• Data de Publicação: 2020
• Outros Autores: Ceglar, Andrej, Toreti, Andrea, Prodhomme, Chloé
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://doi.org/10.1016/j.agrformet.2020.108095; http://hdl.handle.net/10348/10911
• topic: Crop seasonal forecasts; Crop modelling

Wine production is intricately dependant on the evolution of weather conditions in a given year. Therefore, seasonal weather forecasts coupled with empirical wine production models can play a critical role in the short to medium-term management of vineyards and wineries. The implementation of suitable and timely adaptation measures based on predicted wine productions may contribute to risk reduction and improve efficiency. The performance of seasonal forecasts of wine production in the Portuguese Douro & Port wine region (D&P WR) is here assessed for the first time. This application may serve as a case study to be potentially extended to other wine regions. Here, we develop a predictive logistic model of wine production based on monthly mean air temperatures and monthly total precipitation, averaged over the periods of February–March, May–June, and July–September, complemented with an autoregressive component of wine productions. The wine production in the D&P WR during the period 1950–2017 (68 years) is keyed into three classes: low, normal and high production years. The model reveals a correct estimation ratio of approximately 3/4 for the full period, and 2/3 when applied to independent 10%-random subsamples. We then evaluate the performance of the ECMWF 7- month seasonal weather forecasts, issued from February to August, in predicting the meteorological conditions relevant for the wine production in the D&P WR. Overall, the performance is satisfactory for the meteorological predictors. As for the weather forecasts coupled with the wine production model, results reveal that forecasts from May to August are strikingly the best performing, as 1) more observed data is integrated into the empirical model and 2) the skill of seasonal forecasts for summer months is higher. The operational application of these forecasts in the D&P WR is already foreseen. Given the encouraging results, we believe this case study and the established methodology may be tested and adapted to other wine regions worldwide, with obvious benefits for the winemaking sector.