Predicting of cattle density and location at the frontier of Brazil and Paraguay using remote sensing

In this paper, we explore the potential of remote sensing to map pastures areas and by this way establish models for predicting cattle density and location. First, an object based classification (OB) was made in Landsat 5 images for three different municipalities to provide a land-cover map. Second, on the basis of Brazilian official livestock database, a statistical model to predict number of cattle in function of declared pasture area by the farmers was produced. Finally, this model was applied to the pasture areas detected by remote sensing to predict cattle density. Coefficient of determination of the model was 0.63. The results indicate that the methodology used for estimating cattle density has a potential to be applied in regions where no information about farm location and cattle density exists. (Résumé d'auteur