-

A number of approaches for optimal satellite data band selection for information extraction from Landsat
data have been proposed. In this study, an evaluation has been carried out based on Landsat TM data
concerning different land cover types of the four areas in different deserts in Iran. For selectior. of optimal band combination for different land cover/ land use types, the methods of Correction Matrix, Optimum Index
Factor (OIF), Principal Component Anaysis (PCA) and two dimentional Feature Space analysis (FS) have
been used.
The obtained results show that selection of the most suitable bands not only depends on the climatic conditions and geographical region, but also land use, land cover types and date of satellite data. It may be
concluded that to use Landsat satellite data effectively, the optimal band based on some important features such as, vegetated area and bare lands under the investigation should be selected.
The obtained results show that TM thermal band contains information complementary to the TM reflective bands and it may be concluded that in determining the nature of an extensive material, spatial resolution might
have less imprtance than the spectral resolution. Therefore it is essential to evaluate the information content of the TM thermal band for computer aided digital classification.