Forest/Non-Forest Classification from TanDEM-X Interferometric Data by means of Multiple Fuzzy Clustering

In this paper we introduce a method to derive forest/non-forest maps from TanDEM-X interferometric synthetic aperture radar (InSAR) data, acquired at global scale in stripmap single polarization (HH) mode. Among the several observables systematically provided by the TanDEM-X system, volume decorrelation, derived from the interferometric coherence, shows to be consistently sensitive to the particular land cover type, and is therefore used as an input data set for applying a classification method based on a fuzzy clustering algorithm. Since the considered InSAR quantity strongly depends on the geometric acquisition configuration, namely the incidence angle and the interferometric baseline, a multi-clustering classification approach is used. Algorithms for the identification of additional information layers such as urban and water areas are discussed as well, and the mosaicking of multiple acquisitions to improve the resulting accuracy is shortly introduced. The preliminary classification results shown in this paper are very promising for the generation of a global land classification map from TanDEM-X interferometric quicklook data as a next step.