Landscape composition and cultivar resistance : Bayesian modelling of the wheat leaf rust epidemiology on large scales
Résumé
Structure and composition changes in our agricultural landscapes have made the domestication of pathogens easier. It has also made these agricultural landscapes more sensitive to epidemic risks. To regulate these phenomena in a natural way, functional diversity must be introduced. In plant epidemiology, diversity may be represented by the resistance of each cultivar to a particular disease. These resistances are limited in number, so that management strategies must be defined to avoid their erosion. In this study, assuming a homogeneous dispersion, we tried to link the cultivar composition of the landscape to the symptoms exhibited by the most grown wheat cultivars. In order to integrate three independent datasets describing the pathosystem of the wheat leaf rust - a foliar disease caused by the fungus Puccinia triticina on its strict host Triticum aestivum - we developed a hierarchical model in a Bayesian framework. We proved that, in production conditions and for large scales of time and space, the cultivar frequencies in the French wheat landscape play a leading role on the composition of Puccinia triticina populations. As a consequence we brought to light a maladaptation of several Puccinia triticina populations to their host, which is reflected by a decrease in the wheat leaf rust symptoms. Finally this work shows an investigation method to link heterogeneous datasets collected on macroscopic scales and it gives encouragement to acquire such datasets.