cause the high factor loadings in either one can be explained, they only 
represent a little more than half of the overall variance. The remaining 
variance is almost equally distributed to the factors 3 and 4 which do not 
have such significant high factor load parings. If the data of the Ammersee 
north end are compared to the Starnberger See (south end) which is in its 
vicinity, a high agreement in the behaviour of the factor loadings can be 
observed. In both of the factors 1 and 2 the spectral channels 4 and 6 show 
high factor loadings. This means, as the south end of the Starnberger See 
has no entrance of any river and therefore no sedimentation occurs, that the 
behaviour of pure water bodies can be best described in a reduced from by 
using the spectral channels MSS4 and MSS6. This result is in good agreement 
with the experience of conventional image interpretation of LANDSAT data in 
the prealpine region. The almost equal distributed information content of 
the four spectral bands necessitate for the purposes of classification the 
utilization of all four spectral bands if no detail informations are to be 
lost. 
Example 2: City of Augsburg (LANDSAT-1) (Fig. 2) 
The same district of the city of Augsburg consisting of houses and park areas, 
has been analyzed for the scene of August 31, 1972 and October 6, 1975. Plotting 
the factor loadings over the axes of factor 1 and 2 show the changes in spec- 
tral behaviour of all four spectral channels. For the summertime clustering 
high factor loadings occur for spectral channels MSS6 and MSS7 on the first 
factor and for MSS4 and MSS5 on the second factor. For the autumn scene the 
clustering of these factor loadings is only slightly changed but they altered 
their position in the factor pattern space significantly. The explanation 
could be, that the spectral channels MSS6 and MSS7 which display shadows, did 
not change very much in spatial distribution, in contrary to the changes in 
the behaviour of the spectral bands MSS4 and MSS5, which display alteration 
of vegetation due to the coloring and loss of leaves in fall. In addition the 
leafless trees also generated new shadow patterns on the ground which could 
spatially not be resolved. Factor analysis can therefore be used for the 
detection of phenomena, which otherwise cannot be detected from the original 
intensity values because the position of different randomly distributed 
picture elements in a scene is difficult to detect. Only the application of