The two purposes of this paper are: （1） to study the method of computer-aided interpretation using landsat CCTs data for soil classes recognition and （2） to study the relationship between the spectral soil map and the conventional soil map at a large scale.On the ARIES-Ⅱ image processing system

two scenes of different dates of Landsat images were geometrically corrected and spacially registered. Supervised classification for soil interpretation was made on this image which had seven bands in it. On the basis of analysis of the soil parent materials

landscapes and geomorphological types

the study area was stratified into three strata. Twenty five sample areas were selected by the stratified-random sampling method within the whole area.A correlation of the spectral classes from the supervised classification to the soils from the conventional soil map at a large scale （1:50

000） was achieved by manually overlapping the soil data and the spectral data of every sample areas in each of the strata. The results of the correlation were evaluated. It was found that several spectral classes are

with excellent consistency

highly related to some soils such as sandy soil

clayey meadow soil and drabed meadow soil. Saline meadow soil and clay-loamy meadow soil have good relationships to the spectral classes in spite of lacking a good consistency. Other soils were confused with spectral classes.For the soil mapping investigation of some highly related soils mentioned above

the use of the results of the spectral classification and associated correlation will conceivably aid in soil profile placements and boundary delineations

hence reducing the time and expense of the investigation.