An evaluation of radar texture for land use/cover extraction in varied landscapes

One of the more recent developments in operational spaceborne remote sensing is the availability of radar. This sensor has the ability to penetrate through clouds making it a more easily available data set for some locations. In addition, radar interacts very differently with surface features than optical data providing information more related to shape and structure than composition. A disadvantage of currently available spaceborne radar is that the data are almost entirely single wavelength and single polarization limiting the ability to do traditional digital classification. This study examined the usefulness of radar-derived texture measures for feature identification. Texture measures were compared independently and in combination with the original radar for digital land cover delineation. The primary methodology was standard image processing spectral signature extraction and the application of a statistical decision rule to classify the surface features for several sites in East Africa and one in Nepal. Relative accuracy of the resultant classifications was established by digital integration and comparison to validation information derived from field visitations. Variance texture measures were found to be generally very advantageous over original radar values but quite variable in their delineation accuracies from one cover type to another.