Maps of global land cover derived from satellite-based earth observation have existed for almost two decades and represent one of the most important sources of baseline terrestrial information for a wide variety of users, e.g. the Convention on Biological Diversity. More importantly, land cover maps provide critical input data for global models of land use and land use changes. Urgent questions have arisen that depend upon an accurate global land cover dataset, e.g. how much land is available for agricultural use or how high will competition for land be between food and bioenergy, considering increasing needs in the future. Some of these questions could be answered if a global baseline map of land cover would exist. However, at present, a unified and satisfactory solution has not surfaced, owing in part to large disagreements among existing global land cover datasets. This paper compares the three most recent global land cover products, namely GLC-2000, GlobCover, and MODIS. Moreover, it presents a methodology for comparing global land cover maps that allows for differences in legend definitions as well as different spatial resolution between products to be taken into account.