1 POTENTIAL OF HYPERSPECTRAL REMOTE SENSING DATA FOR THE AUTOMATED MAPPING AND MONITORING OF URBAN BIOTOPES

Urban biotopes are an important subject matter for ecological urban planning. Area-wide mapping and monitoring of biotopes is based on visual interpretation of color-infrared aerial photographs and field investigations. This combined inventory yields a high level of detail and accuracy but, as a drawback, it is very time- and money-consuming. Thus, many municipalities take the effort to build up an initial biotope map, but not to make regular updates. Hyperspectral data open up new opportunities for solving this problem. They allow a material-oriented identification of urban surfaces and vegetation types and the derivation of additional quantitative parameters (e.g. percentage of area, degree of surface sealing etc.). Since biotope types cannot be classified directly from pixels ’ grey values the idea of this study is to use a material-oriented classification to identify them. Therefore, biotope types are modelled by a set of quantitative parameters derived from classified hyperspectral images, e.g. the percentage of area of their constituent surfaces which could be vegetation types as well as man-made surfaces, open soil etc.. Building up a generally applicable model of biotope types that way, can also include parameters to describe the location and distribution of the constituent surfaces in the biotope. Once the model is build, it can be used to check urban biotopes, taken from existing biotope maps, for changes of their type with hyperspectral data. Only biotopes indicating a change have to be inspected in the field and updated manually if necessary. The study is part of the Helmholtz-EOS project. Concept and methods of the study are presented focusing on the development of distinctive features for biotope types