Models have been developed for blast disease to predict rice blast severity, yield losses due to blast and the best time for treatment, all at the field scale. At a global scale, modelling has been used to predict potential disease incidence using Geographic Information System and and climate data. Our aim in this study was to develop a model at the landscape level to understand the role of cropping practices on the development of rice blast epidemics and the exchanges between fields. Since biodiversity is often considered as a mean to increase durability of plant resistances against diseases, we first focused on including varietal diversity in the model. The model was developed with the Ocelet modeling language (Degenne and Lo Seen, 2016). Ocelet is able to manage maps and to simulate spatial processes at the landscape scale. The major processes simulated are: -the evolution of the leaf area affected by blast in each field, - the evolution of the potential of contamination of each field, -the disease dispersal from one field to another, -the contamination due to spores originating from crop residues from previous years. The simulation is conducted for several years. Several blast populations and cultivars, each cultivar with a different level of susceptibility to each blast population, may be represented. Cropping practices such as straw management and crop rotations were included in the model because they are supposed to play a role in blast primary inoculum production. The specific situation of the Madagascan Highlands was used as a first experimental case. In this region, rice is extensively grown in small farmer fields. Lowland and upland rice are cropped with different cultivars. In upland rice fields, one partially resistant cultivar was recently adopted and now covers more than 80% of the area (due to its rusticity allowing it to grow without fertilizer application). Cultivars in the lowlands are more diverse. Our aim was to use the model to represent the po