Satellite images are the most extensive source of data about our environment; they provide essential information about global challenges. Since petabytes of imagery are accessible, researchers can now track changes continuously. To work with big Earth observation data, scientists are developing data-driven and theory-limited methods. However, numbers do not speak for themselves. Data-driven approaches without robust theories can lead to results that will not advance our knowledge. We need sound theories to deal with big data without drowning in it. In this talk, we argue that current ontologies and descriptive schemas used in image analysis cannot capture the complexity of landscape dynamics unveiled by big data. These schemas lack expressive power. Existing ontologies for land classification are object-centered; to work with big data, we need to include occurrents. For continuous monitoring of land change, event recognition needs to replace object identification as the prevailing paradigm. The presentation explains how event semantics can improve data-driven methods to fulfill the potential of big data.