11 
'-ESSES 
> 
z 
H 
i 
ID 
O 
"O 
ESSES 
H 
H 
ecosystems research 
complement each 
discusses the 
f remote sensing 
es of thematic 
erpretations are 
by the level of 
rpreter. If the 
gh, there is 
gainst. For the 
oned questions, 
terpretation is 
particular this 
heoretical back- 
eal contributes 
fication, and so 
estigation. It 
e influence of 
omputer-assisted 
e amount of 
ise. The latter 
ccount for the 
like with the 
by means of the 
it is formally 
ATA SETS 
Austrian Unesco 
iosphere (MaB)" 
been transformed 
fstem, using the 
Institute for 
iter Graphics in 
to compare the 
itents of the 
ster on a pixel 
3.1 Basic data 
The following data were available: 
Airborne multispectral scanner data of 
the 11 band Bendix Scanner from the 
German Flugzeugmeßprogramm (FMP), 
covering an area of 512 x 700 pixels of 
2 x 2 to 3 x 3 m2 size; 
Landsat MSS data (Buchroithner 1982); 
- Landsat TM data; 
two colour infrared aerial photographs 
(film diapositives); 
a topographic map 1 : 5 000 (Pillewizer 
1982) ; 
a vegetation map 1 : 2 500, only 
covering part of the study area (Karrer 
1980) ; 
a soil map 1 : 5 000 (Müller 1982); 
a geological map 1 : 5 000, generated by 
photographic enlargement of an original 
map 1 : 25 000 (Cornelius & Clar, 1935); 
an ecotopic map 1 : 5 000 (Neuwinger in 
prep . ) . 
3.2 Data processing 
3.2.1 Digitizing of colour infrared aerial 
photographs 
The two CIR aerial photographs available for 
the study area with a format of 23 x 23 
cm2 were digitized with an optoelectronic 
filmwriter. The measured intensities were 
quantized in 8 bits (0 - 255) and stored on 
a magnetic tape. 
3.2.2 Geometric rectification of digital 
imagery 
In order to create a reference system for 
the map data available, the CIR images and 
the airborne scanner data had to be brought 
into the system of the map data which in our 
case was the Gauss Krueger System. For this 
the different geometries of the various data 
sets had to be taken into consideration. 
Whereas the aerial photographs display a 
central projection, the airborne scanner 
data were generated through scanning of the 
earth surface. The geometric rectification 
was done by means of the program system 
GAMSAD (Kaufmann 1985). The applied 
interpolation algorithm was the nearest 
neighbour pixel asignment which prevented a 
falsification of the original grey values. 
The identification of the ground control 
points in the images was rather difficult as 
the clearly identifiable points were almost 
exclusively located along a mountain road 
which more or less runs close to the 
vertical axis of the imagery. Through this 
very unfavourable distribution of control 
points extrapolations along the margins of 
the images were unavoidable. This led to 
slight deviations in the peripheral parts of 
the rectified images. 
3.2.3 Digitizing of map information 
The digitizing of the various maps mentioned 
in section 3.1. was performed using the 
geoinformation system DESBOD (Ranzinger , 
Kainz & HUtter 1985, HUtter 1986). The 
transformation of map data into the digital 
form was carried out at a digitizing light 
table by means of a curser. A special 
problem for the digital storing was the 
geological map which had been produced by 
photographic enlargement of the original map 
1 : 25 000. Through the generalization and 
the thickening of the lines, the correspon 
dence with the other maps was restricted. 
Especially along the road the deviations are 
rather strong (cf. section 5). In order to 
reach geometric correspondence with the 
other maps, the road was not digitized and 
the thematic information was transformed by 
an affine transformation using some well 
identifiable points. 
After the manual digitization the map 
information was transformed into the image 
data formats to allow the analysis of the 
remote sensing data, i.e. a vector- 
to-raster-convertion was performed. After 
that digital map information was displayed 
on a raster screen, and colours were asigned 
to the various classes. 
3.2.4 Generation of a digital terrain 
model 
The basic data for the generation of the 
digital terrain model (DTM) by means of the 
program system GTM (Leberl & Olson 1982, 
Hafner & Raetzsch 1985) were the contoyr and 
drainage lines of the topographic map. The 
respective film sheets were automatically 
scanned and the resulting raster data 
converted to vectors. After that, the data 
were transformed into the Gauss Krueger 
System and discretized into a grid of 
2x2 m2. This raster contains all points 
which are located on a control line of the 
respective height of this line. Eventually 
the height values for all not yet defined 
raster points were determined by 
interpolation between the known points. The 
resulting raster can then be treated like a 
digital image. 
3.2.5 Information derived from the digital 
terrain model 
As an additional information for the 
pixelwise interpretation of remote sensing 
data for each picture element of the 
airborne scanner image slope gradient and 
aspect were derived by applying the first 
derivation of the DTM data. In order to 
prevent a so-called terrace generation which 
might result from the integration of a too 
small area and to better adapt those data to 
the actual terrain, gradients and aspects 
were calculated using an algorithm which 
takes all eight neigboured pixels into 
account. Thus, the desired smoothing effect 
was reached but, on the other hand, a 
significantly higher effort in computation 
had to be accepted. 
In addition, the influence of the sun 
elevation and azimuth has been derived from 
the digital terrain model by applying the 
method of the horizon. For the moment of 
the scanner data acquisition, azimuth and 
sun elevation of those pixels, which were 
situated in the shadow, were marked. This 
resulted in a binary image.