Abstract. A multiple regression model was used to analyse if the structure of vegetation and soil patches in grazed units (pastures) can be used as explanatory variables to predict the prevalence of Dicrocoelium dendriticum, a common parasite of cattle and sheep, in grazing cattle stocks on the Baltic island of Öland in southern Sweden. The scale dependency was evaluated by comparing three levels of spatial resolution of patches. Prevalence data were obtained from slaughtered animals. Our models predict that the prevalence of D. dendriticum increases in grazed areas with woody vegetation, whereas moist and wet areas decrease parasite prevalence. The predictive power of the statistical models increased with increasing level of patch resolution. Approximately 42% of the variation in parasite prevalence (angular transformation) was explained by the areal proportion of vegetation types (4 th -root-transformed). Based on the results obtained, we believe that our model strategy provides a rational and systematic tool to identify habitats that carry risk for D. dendriticum infection of ruminants, and that it can be applied to other parasites with similar life cycles such as Fasciola hepatica.