ot 
see 
how 
the 
ent 
in 
he 
be 
areas, 
Plotting 
spec- 
ing 
rst 
the 
tered 
n 
» did 
in 
ion 
n the 
uld 
inal 
factor analysis uncovers hidden information which also in the correlation 
matrix are not easily recognizable. 
Example 3: Lechfeld pastures (LANDSAT-1) (Fig. 3) 
A comparison has been maintained for the changes in spectral behaviour for 
a large area pasture in the Lechfeld area for a scene taken in May 1973, 
August 1972 and October 1975. It has been made sure that nothing else than 
natural changes occured during these three years. The only landuse on this 
large area pasture was the grasing by sheep. There exists again the clustering 
of spectral channels MSS4 and MSS5 on one side and MSS6 and MSS7 on the other 
side. The clustering remains constant over the year, but the two clusters 
changed their position in the factor pattern space. This application of spec- 
tral investigation shows more significantly changes in spectral behaviour 
with season than it would be possible by plotting just the spectral intensity 
changes for different periods of time. 
Example 4: Decideous Forest (FMP) (Fig. 4) 
A forest which uniformly consisted of decideous trees has been investigated 
concerning the spectral changes for day 160 (May) and day 182 (July) of the 
year 1976. The absolute position of the factor loadings for channels 8, 9 and 
10 remains the same for these two days. The infrared channel 11 changed its 
position in the factor loading pattern space significantly. This could be due 
to different weather conditions on these two days. The absolute positioning 
of the factor loadings of the spectral channels changed significantly, however, 
by keeping the form of clustering. There is the concentration of spectral 
bands 3, 5, 6 and 7 in both factor pattern spaces and the additional assembly 
of spectral channel 4 for the day 182-scene. The spectral channels | and 2 
which represent the UV and blue region remain significantly separated from 
clustering and relative to each other. 
Example 5: (Coniferous Forest (FMP) (Fig. 5) 
If the factor loadings of the decideous forest is compared to the factor 
loading of the coniferous forest, an agreement can be observed for the channels 
8, 9 and 10. It seems that these channels play an important roll for the