ALBERT

Description: In this study DAIS 7915 reflective imaging spectrometer (IS) data was analyzed for its potential for evaporite mineral mapping in the area of Mt. Sedom, Israel. The data was analyzed with a new processing strategy, due to the high noise content found in the SWIR-II spectral region, which is of high relevance for a mineralogical interpretation of IS data. The implemented approach avoids analysis on the noisy picture element (pixel) scale through delineation of spectrally homogeneous areas in the data cube based on a supervised neural network approach. Class endmembers are chosen through a manual selection process that is based on ideas taken from multi-spectral image interpretation. The underlying assumption is that color differences found in special false color composites created from the IS data can be directly related to an abundant surface lithology. Results show that based on this procedure, even given a high noise content in the data, lithologically important surface units can be classified, which agree well with an existing geological map. Further spectral analysis of the mean class spectra shows that a lithological interpretation in terms of the class bulk mineralogy is possible. Lithological information extracted from class endmember spectra is in good qualitative agreement with results from XRD/XRF measurements for collected field samples.