The availability of remote sensing data with improved spatial, spectra
l and radiometric resolution is now available to fully exploit their p
otential for a specific application subject to the relative merits and
the limitations of each sensor's data. Presented here is a case study
where Landsat MSS and TM; and SPOT MLA data for part of the Bijapur d
istrict, southern India, which were acquired on the same day, have bee
n evaluated for mapping eroded lands. The approach involves the geomet
ric registration of all three data to a common map grid using tie poin
ts and third order polynomial transform; and resampling the MSS and TM
data to a 20 m by 20 m pixel dimension and radiometric normalization.
Thematic maps showing eroded lands were generated on a micro-VAX-base
d DIPIX system using a maximum likelihood classifier. Accuracy estimat
es were made for the thematic maps following stratified unaligned rand
om sampling technique, and subsequently, computing overall accuracy an
d Kappa coefficient. Spectral separability and classification accuracy
was maximum from SPOT-MLA data followed by a combination of Landsat M
SS band 1, SPOT-MLA band 2 and Landsat TM band 4; Landsat TM, a combin
ation of Landsat MSS, TM and SPOT MLA; and Landsat MSS data.