We examine the predictive ability of habitat-species relationship models in island semi-desert environments using as model species the Canary Islands stonechat (Saxicola dacotiae), an endemic bird inhabiting the arid island of Fuerteventura. We investigated nest site selection and the effects of land use on its distribution in a multi-scale approach using nest positions obtained during three consecutive breeding seasons. We identified two environmental predictors, namely Tasseled cap 1 (a range value of land brightness) and slope, and three variables derived from human use (house densities, unpaved roads and fences) as the best predictors of occurrence of the species. Only slope had a positive and significant effect on stonechat occurrence; the rest being negative. Results were not restricted by the scale, indicating that design of special protection areas should be developed considering the landscape scale. Our results provide a robust prediction of the species distribution throughout Fuerteventura, demonstrating that our approach can also cope for the low vegetation signal within small arid regions such as islands. Future land use planning and management for the island should avoid the presence of negative impact elements such as opening new roads and new urbanisations in nearby habitats with the highest probabilities of species occurrence.