PRINCIPAL COMPONENT ANALYSIS OF LANDSAT MSS DATA FOR DELINEATION OF TERRAIN FEATURES

The principal component (PC) transformations of LANDSAT's Multi-Spectral Scanner (MSS) data over a part of the Indian desert displaying ample variation in the terrain features were generated to exploit the potential of such transformations for delineating terrain features. A comparison of the image variance contained in the standard false colour composite (FCC) of LANDSAT MSS data and that of the PCs revealed that, in the study area, the former accounts for only 58.33 per cent image variance (information content) while the first two PCs were found to contain 95.13 per cent, indicating the potential of PC transformations to reduce the data dimensionality while retaining most of the significant information. The last two PC transformations could contribute only 4.87 per cent image variance and the rest of it was observed to be mostly random variance (noise). A quantitative evaluation of the FCC print generated from the first three PCs for delineation of various landform categories revealed the overall accuracy of 92.0 per cent.