Information content in COSMO-SkyMed data

We analyze the information content in COSMO-SkyMed data with different acquisition modes and polarizations. A set of discrimination problems ranging from difficult to simple using samples from different land cover types is presented. Several separability measures, i.e. stochastic distances and their derived hypothesis tests, are applied to pairs of samples, and their ability to discriminate is assessed. From the studied modes, class separability of water, pasture, forest and urban is enhanced if the lowest resolution mode is used. Both, ascending and left looking acquisition geometry yield better classification results. Distance measurement tests between samples of the same class give better results for HH polarization than for VV polarization suggesting that the analyzed cover properties are better described by that microwave configuration.