428 
Another combination of different Ratio is 
illustrated in photo (6) part (C), the 
ratio are TM5/TM4,TM3/TM4,TM2 displayed 
in red, green and blue respectively. 
High Ratio is only in the red displayed 
ratio TM5/TM4 and that is why the damaged 
and severly damaged forest appears red. 
To seperate the strip maining, bare soil 
and lignit deposit it is adviseble to use 
TM6 (Themal infra red) instead of TM2. 
TM6 has higher gray value for above men 
tioned classes than in TM2 and allow to 
distinguish between the strip maining and 
the lignit deposit from dead or clear cut 
forest areas. 
The analysis of the ratio image TM5/TM4 
as a density slice image requieres an 
interpretation of the ratio pixel value 
with the help of the interpretation of a 
suitable color composit of the same area. 
The high Ratio represents the damaged and 
severely damaged forest area, and the low 
Ratio represents a healthy and slightly 
damaged forest area. Most, but not all of 
the pixels having a high Ratio represents 
a high damage level of the forest. Some 
of them indicate dry or harvested agri 
culture fields. To avoid such a kind of 
false evaluation and interpretation, it 
is advisable to eliminate the agricul 
tural area from the forest sites through 
artificial masks including only the 
forest site. The mask could be obtained 
from the RVI-TM4/TM3 and with a certain 
streching of the Ratio image and 
following with a manual delination of 
forest area. 
3.3 Results of Supervised Classification 
3.3.1 Test Site St.Blasien/Schluchsee FRG 
Forest damage assessment in this test 
site began since 1985, with the acquisi 
tion of multi spectral scanner data (Ben- 
dix and Daedalus Airborne Scanner) from 
the altitude 1000 and 3000 m above 
ground. Also Landsat-TM images from 1985 
and 1986 have been ordered and evaluated. 
The main evaluation methodology that has 
been used is a supervised classification 
according to the maximum likelihood algo 
rithm. Use of this method requieres from 
the user a number of training areas and 
their statistic corresponding to the 
different damage levels; and of course, 
training areas representing other classes 
such as deciduous, meadow and etc.. The 
determination methodology for the trai 
ning areas is discribed under(3.1). 
Furthermore the supervised classification 
methodology requieres also a number of 
spectral bands that seems to be useful 
for distinguishing different classes. 
The selecting spectral bands can be ob 
tained from the results of the spectral 
signatures investigations. 
The results of damage assessment using 
computer-aided calssificatin after above 
mentioned methodology are shown in 
photo(1,3,4,5,6) . 
3.3.2 Test Site Erzgebirge, CSFR/GDR 
As mentioned in (chapter 2) many Landsat 
images have been evaluated in terms of 
forest damage assessment, mapping and the 
estimation of forest decline development. 
The supervised classification (maximum 
likelihood method) has been also used for 
the landsat data of the Erzgebiige. 
In this paper the results of two images, 
landsat-MSS 1976 and landsat-TM 1985 are 
presented. Photo(3) is IR color composit 
bands TM2, TM3, and TM4 displayed in 
blue,green and red respectively of Land 
sat-TM. Photo (3) shows only a part of 
the Erzgebirge region. The lignit strip 
maining and the plume from lignit power 
plant can be easily identified on the 
landsat-TM from 1985 (down right). The 
plume from lignit power plant lengths 
measured by pixel counting, is about 30 
km. On landsat-MSS image from 1972 was 
also possible to clearly identify the 
plume and the directon of the plume over 
the Erzgebirge region. And the lignit de 
posit on landsat-TM from 1985 was an 
agricultural area on the landsat-MSS from 
1972 . 
The dark red area in photo (3) shows 
healthy and slightly damaged coniferous 
forest area. Since the images are geome 
trically corrected it is easy to make a 
visible comparison between two images 
from 1976 and 1985. The healthy forest 
area is strongly reduced on the image of 
1985, but the damage and severely damaged 
area has strongly increased and has a 
gray/green color diagonal on the 
photo(3). A part of the classified 
results of the region are shown in photo 
4 and 5) of the landsat-MSS 1976 and 
landsat-MSS 1985. The spectral bands that 
are used for supervised classification 
are bands 4,5 and 7 in case of landsat- 
MSS and bands 2,3,4, and 5 in the case of 
landsat-TM data. 
4 Discussion and Conclusion: 
The important task associated with use of 
landsat imagery for the damage assessment 
is the verification of the obtained 
results. This can be done by interpreta 
tion of CIR photographs and/or field as 
sessment of the study test site (Kadro 
and Kuntz 1986,Kadro 1989). 
For the verification of results of the 
study test site at St.Blasien/Schluchsee, 
FRG, many stands have been established 
and delinated on the CIR photographs and 
on the landsat images. 
All the trees in each stand have been in- 
terpretated and counted from CIR photo 
graphs according to different damage 
levels .Also their respective numbers in 
percentage have been calculated. However, 
as far as the case of the classified 
landsat data is concerned the pixels have