Defining Landscape Units Through Integrated Morphometric Characteristics

Landscape analysis and landscape planning need sophisticated spatio-temporal image analysis complemented with context-sensitive methodology. The latter means that raster processing has to evolve beyond per-cell analysis. We have integrated two approaches for the delineation of homogenous areas, or “mapping units”. First, a multiscale image analysis technique is used to extract objects from earth observation images. Using multiscale segmentation and object aggregation, homogeneous ‘patches’ are created comparable to those produced via manual interpretation. However, a benefit of the approach is that constituent lower level (i.e. finer resolution) objects are available plus the patch delineation process is explicitly parameterized. These rules (including fuzzy operators) are stored in the project, and can readily be queried and modified when a new understanding of the mapped scene is introduced. We assume that mapped boundaries are based on ontological objects, corresponding to real world entities of scientific and practical significance. Second, we apply segmentation methods to terrain information and add this information to the patch definition algorithms. This enables us to combine boundary characteristics (length, relative proportions to neighbours, spectral gradients to neighbours etc.) with object characteristics (mean spectral as well as elevation and morphometric values, standard deviation, range) and sub-object information (e.g. heterogeneity based on sub-objects, spatial distribution of subobjects). Morphographic landform information is essential for the modeling and the understanding of many physical processes. Analysing geomorphological processes, surface runoff and erosion as well as mass movement risk assessment and microclimatology require consideration of topographic features. We demonstrate that the combination of these methodologies allows for better detection and mapping of complex topographic landuse units and, by incorporating patch generation rules, for improved landscape monitoring and scenario modeling.