Hierarchical feature representation through multi-scale segmentation offers new possibilities in mapping complex systems. We lay out that the recognition of natural features is more difficult than the recognition of anthropogenic features such as houses or roads. For the latter group spectral and spatial characteristics can be anticipated and rules can be defined. Consequently, the automated extraction of roads or the 3D extraction of buildings is very advanced. In contrast, the automatic extraction of natural features like habitats is still far from being operational.