Medium spatial resolution satellite images have been recurrently used for automatic land cover mapping of large geographical areas. Despite the innovative developments in temporal and spectral resolutions of these sensors, their spatial resolution still hampers the effective discrimination between detailed cover types within fragmented landscapes. In this study we exploit the ability of MERIS and MODIS imagery for the discrimination between land cover classes in Portugal. We aim at identifying the most suitable subset of classes, from a recently recommended land cover nomenclature, to be used in the regular production of medium spatial resolution land cover products at national scale. Final results do not show differences on spectral separability ability of MERIS and MODIS.