14 
presence of some few indicator species or 
have even been relatively homogenously 
reformed by cattle grazing which led to 
mutual misclassifications. A fusion of the 
24 associations to only 16 classes let the 
classification accurasy rise from 40 % to 
80 %. 
6 CONCLUSION 
Within the present study once more it turned 
out that object classes of geoscientific 
maps can be automatically reclassified using 
remote sensing data with very different 
success. Some of the reasons for these 
stronly diverging results are the following: 
- poor geometrical quality of boundaries 
of object classes; 
- generosity in field mapping; 
cartographic generalization; 
- problematic spatial delimitation of 
object classes and their topographic 
neighbourhood (length and shape of 
borderlines) ; 
- different radiometric quality within one 
image; 
- similarity of object classes in spectral 
characteristics (feature space); 
- changes between date of field mapping 
and date of remote sensing data 
acquisition; 
selection of training areas; 
- selection of classification algorithm; 
- different suitability of remote sensing 
as a measuring method for various object 
classes; 
- different empirical contents of 
predominantly theory-derived categories; 
different validity of indirect 
conclusions from sensor-exposed surface 
classes to subterraneous object classes; 
allochthonous geo-ecosystems. 
Based on our experiences we would like to 
emphasize that the combined interpretation 
of remote sensing data and thematic maps by 
means of digital image processing is an 
indispensable help in geoscientific and 
geoecological studies. It is mainly two 
applications which are of special interest. 
One is the possiblity to relativize the 
information derived from various sources in 
a way, that as a final product more 
plausible, more consistent and hence, more 
reliable maps exist. This causes the 
necessity to disclose the theoretical 
reflections on the problems of thematic and 
spatial delimitation of object classes and 
thus to stimulate the examination and 
development of geoscientific and ecological 
theories. 
stik-Programra-Systeme für die Sozialwissen 
schaften nach N.H. Nie und C.H. Hüll. Eine 
Beschreibung der Programm-Versionen 8 und 
9, 4. Edition. 323 p. Stuttgart - New York, 
Gustav Fischer Verlag. 
Buchroithner, M.F. 1982. Die Satellitenbild 
karte der Hohen Tauern im Maßstab 
1 : 200.000. In: MaB Map Volume "Hohe 
Tauern". Veröffentlichungen des österrei 
chischen MaB-Hochgebirgsprogramms Hohe 
Tauern, Volume 7, p. 7 - 9, 1 map enclosed. 
Innsbruck, Universitätsverlag Wagner. 
Cornelius, H.P. & E. Clar 1939. Geologie des 
Großglocknergebietes. 1. Teil. Abhandlun 
gen der Zweigstelle Wien der Reichsstelle 
für Bodenforschung, Volume 25, Issue 1, 
1 map enclosed, Vienna. 
Gierloff-Emden H.G. 1982. Remote sensing 
for Coastal areas. Proc. IGARSS 1982, 
p . 1.1 - 1 . 8 . , Munich. 
Hafner, K. & H. Rätzsch 1985. Anwendungen 
des Programmpaketes Graz Terrain Model 
für Planung und Simulation, österr. 
Ingenieur- und Architekten-Zeitschrift 
130, 5: 179 - 183. 
Hütter, R. 1986. Mass digitization - The 
old way. Experiences with a manual digi- 
tizing System. Proc. EUR0CART0 IV, Nachr. 
Karten- und Vermessungswesen, Ser. II, 
44, p. 65 - 77 , Frankfurt/Main (IFAG) . 
Karrer, G. 1980. Die Vegetation im Einzugsge 
biet des Grantenbaches südwestlich des 
Hochtores (Hohe Tauern). In: Untersuchungen 
an alpinen Böden in den Hohen Tauern 1974 - 
1978 Stoffdynamik und Wasserhaushalt. Ver 
öffentlichungen des österr. MaB-Hochgebirgs 
programms Hohe Tauern, Volume 3, p. 35 - 67, 
1 map enclosed. Innsbruck, Universitätsver 
lag Wagner. 
Kaufmann, V. 1984. Entwicklung und Erprobung 
eines Verfahrens zur geometrischen Auswer 
tung von digitalen multispektralen Zellen 
abtasteraufnahmen und analog/digital-ge- 
wandelten Luftbildern. Diploma Thesis, 
University of Technology, Graz, 228 p. 
Müller, H.W. 1982. Erläuterungen zu den Bo 
denkarten Pasterzenvorfeld und Umgebung 
Wallackhaus - Hochtor 1 : 5.000. In: MaB- 
Kartenband "Hohe Tauern". Veröffentli 
chungen des österr. MaB-Hochgebirgspro- 
gramms Hohe Tauern, Volume 7, p. 23 - 28, 
1 map enclosed. Innsbruck, Universitäts 
verlag Wagner. 
Neuwinger, I. (in prep.). ökologische Kenn 
zeichnung von Boden-Relief-Serien längs 
eines Höhengradienten im Gebiet des 
Tauernbaches am Großglockner (Hohe Tauern, 
österreichische Alpen). Veröffentlichungen 
des Österreichischen MaB-Programms "H.T.". 
Opp, Ch. 1984. Zur Entwicklung und Systema 
tisierung der Catena-Konzeption. Hercynia 
N.F. , 21 , 2:204 - 217. 
Pillewizer, W. 1982. Die Sonderkarten 
"Pasterzenvorfeld und Umgebung" 1 : 5.000 
und "Wallackhaus - Hochtor" 1 : 5.000 des 
österr. MaB-Hochgebirgsprogramms Hohe 
Tauern. In: MaB-Kartenband "Hohe Tauern". 
Veröffentlichungen des Österreich MaB- 
Programms Hohe Tauern, Volume 7, p. 29 - 
32, 1 map enclosed. Innsbruck, Universi 
tätsverlag Wagner. 
Ranzinger, M., W. Kainz & R. Hütter 1986 . 
DESB0D - A geographical information System. 
Final Report 1981 - 1985. Dibag Report, 
19, 92p., Institute for Image Processing 
and Computer Graphics, Graz Research Center. 
Symposiur 
Digital ] 
for fore: 
Kuo-mu Ch 
National Taiwi 
ABSTRACT: The 
types classif 
the MSS data ■ 
vised/unsuper 
component tra: 
to standard c 
seven best bai 
and classifiei 
types classif: 
time and man-] 
combination wl 
cation accurai 
images and mi: 
1 INTRODUCTIOt 
Resource surve 
an airborne mt 
the satellite 
scanner, which 
resolution, na 
scanning time, 
mapping than t 
vegetation typ 
The goal of 
length band co 
cation in fore 
that when work 
accuracy in cl 
all wavelength 
a marked incre 
frequently des 
utilizing only 
fication proce 
of wavenlength 
in the classif: 
2 MATERIALS AN] 
2.1 Data utili: 
The airborne mi 
study were col] 
with 11 channel 
mental Forest c 
Taiwan on Woven 
tape (CCT) cont 
was used in dat 
In order to p 
cover types cla 
niques were use 
protray spectra 
surface feature 
are principal c 
spatial filteri 
REFERENCES 
2.2 Processing 
Beutel, P. & W. Schubö 1983. SPSS 9 Stati- 
The image proce