The Castro Verde SPA is the main pseudo-steppe area in Portugal, holding populations of several threatened steppe bird species of national and international importance. A specific agri-environmental scheme has been put in place in the area to support these populations. While the local steppe bird community is relatively well studied, existing knowledge about habitat selection is mostly based on limited datasets or restricted statistical methods, which could hinder the of recommended management measures. In this study, we used data from several different remote sensing sources (e.g. SPOT VGT, Landsat TM, LiDAR) to characterise landscape features at different spatial scales, while multi-date imagery captured agricultural crop dynamics. A large dataset of steppe bird occurrence was collected through a combined stratified random field sampling design. The species-habitat associations were quantified using a non-linear regression approach (MARS) within a robust methodological framework. The methodology showed consistent results with what is known, confirming most of the existing knowledge on habitat use by the local steppe bird community, while adding further information. The approach is thus reliable and could potentially be used to investigate less well known communities. It is further suggested that the findings of this study be incorporated into appropriate management prescriptions.