Multi-scale image analysis for ecological monitoring of heterogeneous, small structured landscapes

The main concept behind this paper is that pattern and processes are linked in a mutual way. In the last decades landscape ecology was dominated by quantitative descriptions (landscape metrics ) of the landscape under concern and its components. Now there is a growing interest in the cause-effect-relationships between these environmental characteristics. Highresolution aerial photography hold an important amount of valuable information, but until recently only a little proportion of the entire information was usually used in scientific analyses due to conceptual and technical limitations. In this paper we present results derived with a multi-scale image segmentation approach and it is demonstrated how this approach allows for an identification of pattern at several scales simultaneously. First results testify that this is a suitable method for the delineation of meaningful landscape elements and subsequently for landscape monitoring, particularly if dealing with complex or small-scaled pattern. It is shown that hierarchically linked objects are more suitable for monitoring than pixels although the necessity for a comprehensive methodology for object-based change detection arises.