Integration of Ancillary Information into Object-based Classification for Detection of Forest Structures and Habitats

The detection of forest types and structure parameters is of major importance for the design of forest inventories, for the application of forest management practices as well as for the monitoring of biodiversity in the context of the NATURA 2000 network. For these purposes the use of additional information about the natural behaviour of forest structure within classification processes of satellite data is widely known. Especially parameters concerning potential natural forest locations, such as elevation, aspect, precipitation, wetness or soil acidity, were taken into account. Although natural site conditions strongly influence the forest types and structure, the presented results will additionally improve the classification results by integrating silvicultural knowledge into the classification process. The study was carried out using QuickBird data at test sites, which are located in the pre-alpine area in Bavaria (Southern Germany). Within the test sites, different semi-natural mixed forest types exist. First results of the presented approach show higher classification accuracy than can be reached without usage of additional data. It is recognisable that higher classification accuracy depends on the kind of ancillary data. While the effects of the local variability of total height and the aspect are very limited in prealpine areas, additional soil-data or information of the forestry site map in combination with fuzzy-based rules can significantly improve classification results. In contrast to improved results with ancillary soil data, silvicultural information tends to have less influence on the classification quality. Additionally, for habitats and species with very distinctly defined ecological niches (e. g. alluvial types of forest) a better definition and integration of rules is possible than for habitats with very broad ecological ranges.