Data Driven Predictive Models Based on Artificial Intelligence to Anticipate the Presence of Plasmopara Viticola and Uncinula Necator in Southern European Winegrowing Regions

Downy and powdery mildews are two of the main diseases threatening grapevine cultivation worldwide caused by phytopathogens Plasmopara viticola Uncinula necator, respectively. These may cause severe damage to grapevines inducing wilting plant organs, including bunches, especially when vines untreated. This fact, together with widespread these pathogens due large extensions land dedicated monoculture, makes necessary develop new predictive modeling tools that allow anticipating disease appearance in vineyard, minimizing losses fruit yield quality, helping farmers defining appropriate more sustainable management strategies (fungicides applied at right time dose). For this purpose, farms located three countries (Portugal, Spain, Italy) were selected study relationship between microclimatic characteristics plots, phenological stage plants throughout annual cycle, presence both using different Machine Deep Learning classification algorithms: Logistic Regression, Decision Trees, Random Forest, Gradient Boosting, K-Nearest Neighbors, Naïve Bayes, Support Vector Machines, Neural Networks. The results showed that, after an entire best performing models Machines for downy mildew Forest mildew, providing a prediction accuracy than 90% infection risk 80% treatment recommendation. will be fine-tuned during additional vegetative seasons ensure their robustness receive short- medium-term climatological forecasts make recommendations. preliminary obtained show promising tool field prevention resource saving.