Classifying and describing the Earth's land use and land cover present unique challenges, leading to a lack of harmonization between existing land classification systems. These problems arise partly because current classification systems take a static view of the world. As such, they aim to support unambiguous definitions of land cover and land use classes at one particular moment. Since the Earth's land cover is continuously changing, such static definitions are limited. This talk considers how large Earth Observation data sets can help to define a new generation of spatio-temporal ontologies of land use and land cover. These ontologies will distinguish terms such as “forest” and “savanna”, associated to continuants, from terms such as “deforestation” and “degradation”, related to occurrents. This novel approach is better suited to describe the land dynamics of our ever-changing planet.