Prédiction du risque du mildiou de la pomme de terre à l'échelle de la parcelle en Wallonie : validation et analyse de sensibilité du modèle Guntz-Divoux, test et calibration du modèle français "Milsol", et effet de la résistance variétale de l'hôte sur le cycle du pathogène

Late blight (Phytophthora infestans) is the main disease in potato farming. The extremely powerful and fast destructive potential, as well as the limited number of systemic active ingredients and the emergence of resistance to them in the pathogen, make the use of preventive fungicide treatments essential to control this endemic disease. In Belgium, 15 to 20 fungicide sprays are applied each season to control it, which is more than any other European country. First, a sensitivity analysis with regard to meteorological data was carried out on the Guntz-Divoux model, an epidemiological model similar to that used for late blight warnings in Wallonia by the CARAH (Centre for Agronomy and Agro-industry of Hainaut Province). It was concluded that the future supply of the model with data from the AGROMET system (supply of interpolated meteorological data at a very small spatial and temporal scale) will allow a prediction of the late blight risk more suitable (meteorologically speaking) for a farmer's location than if the model is supplied by data from a "close" (10 to 20km) meteorological station from his field. In addition, the model outputs (and therefore the accuracy of a treatment prediction) are more sensitive to uncertainties in meteorological data during years of low late blight pressure (hot and dry) than during years of high pressure. Secondly, the epidemiological model used in France "Milsol" and CARAH warnings were evaluated retrospectively on observed in-field data monitoring the disease severity, by calculating contingency scores. For susceptible varieties (Bintje), the performance of the Milsol model appears to be slightly better than the warnings provided by the CARAH, although this difference is not statistically significant. For both models, consideration of varietal resistance does not seem appropriate; CARAH warnings skip periods when the disease appears, while Milsol model does not reduce the number of fungicide treatments, discrediting the use of less susceptible varieties. Finally, an attempt was made to calibrate the parameters of the Milsol model to better understand the effect of host varietal resistance on the pathogen cycle. Apart from the number of simulated late blight generations spent before the infection risk is actually present, no particular step of P. infestans's cycle could be identified as inhibited by host varietal resistance. More specific analyses under controlled conditions should be undertaken to identify the critical phases of the pathogen's cycle which are impacted by varietal resistance.