Kernel based re-classification of Earth observation data for fine scale habitat mapping

State-of-the-art Earth observation systems that provide very high spatial resolution imagery have given ecologists a powerful tool to directly identify species, habitats and other ecological units. At the same time, there is an urgent need for harmonised tools and methods to evaluate status and trends in European habitats. Towards that goal, the current work explores the applicability and transferability of an advanced pixel window classifier applied on very high spatial resolution satellite imagery for fine scale habitat mapping. The algorithm is tested on images of varying spatial resolutions acquired over test sites designated for the NATURA 2000 list located in different biogeographical zones. Algorithm application to Quickbird and IKONOS images gives encouraging results, regarding both the overall accuracies and the level of class hierarchy (habitats) identified.