Materials in the scene such as minerals and plants have characteristic patterns of absorbtion and reflectance of light of different wavelengths. One way to analyze a scene is to survey samples of various materials to determine what are the "spectral signatures" of material found on the ground and then to identify these signatures in the scene, (ASTER Spectral Library).
For this illustration, a simpler method was used. All ASTER bands, including the visible and near-infrared (3), mid-infrared bands (6) and thermal (5) bands were used to extract summary bands called principal components, one for each group of bands. The first components, containing over 95% of the information (variance) in each group, were used to make a false-color composite. Next the pixels in the scene were classified into 7 clusters using an isocluster routine. The composite was used to start the classification process, but the final classification depends on the three first components, and bands 6, 8, 10, and 14. The isocluster routine makes a map showing the ASTER scene in 7 colors. All the pixels of the same color are more alike in spectral response than pixels of other colors. The selection of bands is based on the spectral response of various minerals that are products of geological weathering, including sand, clay, and carbonates (limestone). (John Jensen, Remote Sensing of the Environment,, p. 483). (Geoimage: See the TM spectral processing flyer (PDF 224K) for examples.)
Ground Truth
The method presented above results in clusters with numbers but no names. To determine what the colors represent the analyst must survey the ground, preferably with large-scale geological maps and descriptions of the formations. The colors on the geological map of Langkawi indicate the chronological divisions rather than the minerals on the surface. Nevertheless the geological map would guide the analyst in ground-truthing. There are other methods of geological analysis using satellite imagery that are beyond the scope of this project.
