ABSTRACT:We present a new classification scheme for muddy and sandy sediments on exposed intertidal flats, which is based on synthetic aperture radar (SAR) data, and use ALOS-2 (L-band), Radarsat-2 (C-band) and TerraSAR-X (X-band) fully polarimetric SAR imagery to demonstrate its effectiveness. Four test sites on the German North Sea coast were chosen, which represent typical surface compositions of different sediments, vegetation, and habitats, and of which a large amount of SAR is used for our analyses. Both Freeman-Durden and Cloude-Pottier polarimetric decomposition are utilized, and an additional descriptor called Double-Bounce Eigenvalue Relative Difference (DERD) is introduced into the feature sets instead of the original polarimetric intensity channels. The classification is conducted following Random Forest theory, and the results are verified using ground truth data from field campaigns and an existing classification based on optical imagery. In addition, the use of Kennaugh elements for classification purposes is demonstrated using both fully and dual-polarization multi-frequency and multi-temporal SAR data. Our results show that the proposed classification scheme can be applied for the discrimination of muddy and sandy sediments using L-, C-, and X-band SAR images, while SAR imagery acquired at short wavelengths (C-and X-band) can also be used to detect more detailed features such as bivalve beds on intertidal flats.