403
coniferous forest. Thus, older spruce
stands from different growth districts
showed spectral signatures, which
increased from the Northeast to the
Southwest of the Regensburg map sheet
(Fig. 5). Regional changes could also be
noticed by comparing the signatures of
three growth districts in the Northeast
and Southwest of the Nuremberg map
sheet, and the center of the Augsburg
map sheet.
In the visible bands, in reduced form
also in the near infrared, changes in
the signatures can well be attributed to
changes in atmospheric conditions
(Sagischewski, 1989). The signature
variances in the short wave infrared
(bands TM5 and TM7) seem to be mainly
due to different geological site
conditions which are partly considered
in the growth districts (Keil et al.,
1989).
3.4 Classification and Verification
For classification purposes, the
supervised maximum likelihood method was
applied using bands TM2, TM3, TM4, and
TM5 of the summer data set. In the map
sheets of Nuremberg and Augsburg, the TM
scenes of different pathes (193 and 194)
were classified separately. The
classification result could considerably
be improved by a separate optimization
in different subregions. These
subregions, which had to be derived due
to the regional changes in reflection,
were based on the digitized map of the
growth zones/growth districts. Finally
the results of the subregions and the
different paths were put together to
complete maps and superimposed on the
information from the topographical map
1:200 000.
The accuracy of the classification was
determined in a verification for which
additional control areas were evaluated,
partly along with the training areas. By
using aerial photos which were available
in parts of the map sheets and control
areas recorded on the ground, a coverage
of verification areas, at least in the
important growth districts, should be
reached.
In some growth districts, a major part
of the reference data had to be used for
classification, especially for catching
the various mixture distributions of
spruce and pine. Therefore the combined
training areas and control areas
provided a better coverage of the exten
ded sites for verification. For this
SPRUCE STANDS OF DIFFERENT GROWTH DISTRICTS
JULY
HISTOGRAn BAND: 1 HISTOGRAM BAND: 2
HISTOGRAM
BAND: 3
HISTOGRAM
BAND:
Fig. 5: Regional changes of grey value
histograms, demonstrated
for old spruce stands of
different growth districts.