Multivariate forecasting model to optimize management of grape downy mildew control

Abstract

Aim of this study was to develop a forecasting model for Plasmopara viticola to achieve rational disease management and to reduce the use of copper treatments in organic farming. Starting from meteo-climatic, agronomic and phytopathological data a partial least squares discriminant analysis was developed. Three different strategies were compared: treatments according to the established organic agricultural practice (standard); treatments according to the predictive model and untreated control where no fungicides against downy mildew were applied. The modelling approach was divided into three phases: 1) model calibration; 2) field testing and 3) a posteriori model performance evaluation. The prediction was separately considered and modelled for: i) disease onset and ii) disease progress. The results for phase 1 show a percentage of correct classification equal to 91.8 % for the disease onset with 3 days elapsed between the prediction of first potential attack and disease onset and to 91.23 % for disease progress. In field testing phase the percentage of correct classification was equal to about 81 % for both the analysed years (2009 and 2010). In the phase 3 the percentages were quietly higher for the 2009. The number of fungicide applications on the partial least squares discriminant analysis model was almost half compared with standard schedule both in 2009 and 2010. Finally this approach showed the possibility to reduce fungicidal treatments and to avoid applying copper not essential for disease control representing a first step in the model validation.