Application of a non-linear temperature forecast post-processing technique for the optimization of powdery mildew protection on strawberry

Strawberry powdery mildew, caused by the fungus Podosphaera aphanis, is a dangerous disease in warm and dry climates as well as in greenhouses or plasticulture. Plant protection against P. aphanis is mainly based on chemical fungicides. More than ten chemical treatments for each growing cycle are often applied in tunnel strawberry production in northern Italy. SafeBerry is a decision support system, which optimises, and often reduces, the use of chemicals against this disease. The system is based on a correct fungicide application based on the disease risk level in each tunnel and on the specific action mechanism of the fungicides. The level of risk is based on crop and environmental parameters. The temperature assessment and its forecast represent two key points in the system. The decision-making procedure uses day-time temperatures forecasted for the three following days. They were calculated by post-processing of the operational weather model output (Model Output Statistics, MOS). MOS was carried out for three sites with a "machine learning", multivariate, non-linear technique ("Random Forest"), which uses many meteorological predictors. With this system in 2007 we obtained a strong reduction in the number of treatments (up to 60%). Keywords: pesticide, fungicide, integrated pest management, MOS, temperature downscaling