Landscape ecology is seldom used in epidemiology. The aim of this study is to assess the possible improvements that can be derived from the use of landscape approaches on several scales when exploring local differences in disease distribution, using bluetongue (BT) in Corsica as an example. The environment of BT-free and BT-infected sheep farms is described on a fine scale, using high resolution satellite images and a digital elevation model. Land-coverage is characterised by classifying the satellite image. Landscape metrics are calculated to quantify the number, diversity, length of edge and connectance of vegetation patches. The environment is described for three sizes of buffers around the farms. The models are tested with and without landscape metrics to see if such metrics improve the models. Internal and external validation of the models is performed and the relative impact of scale versus variables on the discriminatory ability of the models is explored. Results show that for all scales and irrespective of the number of parameters included, models with landscape metrics perform better than those without. The 1-km buffer model combines both the best scale of application and the best set of variables. It has a good discriminating ability and good sensitivity and specificity.