9 
? Evaluation by SPOT 
Site. In : P.Rei- 
)pe 82, 15-42. 
note sensing. 
la compréhension 
action : essai de 
tointerprétation 
Simp.Intern.de la 
Photogr. et Teled., 
Interprétation des 
ison avec les peti- 
c.Française de 
Comparaison de 
images à haute 
Nat.Agronomique 
Interprétation du 
tir de clichés de 
approche botanique 
nr.99-3, 41-51. 
terpretation techni- 
acy. Photogramm. 
.46, nr.3, 359-367. 
hyto-geomorphic 
e. Geoforum, vol. 2, 
.nalysis for Nominal 
sis of Qualitative 
on an Information 
-45. 
ig methods for esti- 
lournal ,2, 113-120. 
îemote Sensing and 
., Wiley. 
;pts of Land Clas- 
I.) Land evaluation, 
ilia, 11-28. 
îulation Europe 82 - 
,E.C., SA/I.04.E2.84. 
Analysis and Remote 
î. 
)sclassificatie toege- 
inalyse van Landsat- 
:.thesis, 102. pp. 
ngebruik in België 
34 volgens de kada- 
.S., Statistische 
1978. Remote Sensing : 
ic Map Accuracy. 
, 7, 3-14 
ion pour 1'approche 
ieu bocager. Applica- 
nousin-France). 
errain classification 
ography : a review 
togramm.Engin., 
Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986 
Interpretation of classification results of a multiple data set 
Helmut Beissmann 
Department for Satellite Image Cartography, Institute for Cartography, Austrian Academy of Sciences, Vienna 
Manfred F.Buchroithner 
Institute for Image Processing and Computer Graphics, Graz Research Center, Austria 
ABSTRACT: Based on experiences from a test area in a high mountain region in Austria, fundamental problems 
of combining various remote sensing data sets with geoscientific maps are discussed. The attempt was made to 
use a computer-aided supervised approach to reclassify the "ground truth" given in thematic maps. Possible 
explanations of the classification results are given. 
RESUME: A la base d'expériences sur un terrain d'essai dans une région de haute montagne en Autriche sont 
discutés des problèmes fondamentales concernant la combination de différentes dates de remote sensing avec 
des cartes gêoscientifiques. On a essayé d'utiliser un approche corrigé à l'aide d'un ordinateur électronique 
pour reclassifier la "ground truth" qui est constatée dans des cartes thématiques. Les explications possibles 
des résultants de classification sont indiquées. 
1 INTRODUCTION 
In c 
orrespond 
enee 
with Gi 
er lo 
ff-Emd 
en 
(1982) 
in 
this s 
tudy 
empha 
sis 
is p 
ut 
on a 
comp 
rehensive 
th 
ematic 
inte 
rprêta 
ti 
on of 
remo 
te sens 
ing 
data 
by 
combi 
ni 
ng and 
comp 
aring i 
nf or 
mation 
of 
con 
ve 
ntional 
geos 
cientific 
di 
stributi 
on maps wi 
th 
remote 
sens 
ing info 
rmat 
ion by 
mea 
ns of 
digital 
meth 
ods. It 
is 
an obje 
ctiv 
e of thi 
s paper 
to p 
oint out 
the 
fundame 
ntal 
geos 
ci 
entific 
inte 
rpretatio 
n pr 
oblems i 
n th 
is metho 
d. For 
thei 
r combin 
ed 
evaluati 
on 
remote 
sensing 
data 
as well 
as 
thematic 
map 
s , whi 
eh 
can be 
take 
n as a re 
flee 
ted mode 
1 of 
reali 
ty 
, need 
a s 
pecific 
cr it 
ic of th 
e ir 
source 
s 
to cor- 
ree t 
ly esti 
mate 
their 
P 
otenti 
al 
f or 
geos 
cientific 
in 
terpreta 
tion 
. Fo 
r 
a study 
area 
close to 
the 
Grossglock 
ner ( 
3 
797 m), 
Aust 
ria's highes 
t peak, 
sit 
uated 
ab 
ove the 
tree 
-line (Figure 
1 ) , the 
att 
empt 
ha 
s been 
made 
to r ec 
lass 
ify the 
obje 
ct c la 
ss 
es of 4 
thematic lay 
er s 
(lithology, 
soil 
types, 
Figure 1. Location map of study area. 
veget 
ation, 
ecotopic 
structure) by means of 
discr 
iminant 
analysis 
of différé 
nt remote 
sensi 
ng data 
and a DTM. The main 
difference 
to th 
e commo 
n land co 
ver classifi 
cation is 
that 
vario 
us categories which 
were to 
class 
ify do 
not alway 
s represent the terrain 
surfa 
ce ( e . 
g. bed 
rock below 
soil and 
veget 
ation c 
over) and 
that in the 
sense of 
the e 
mpiric- 
analytica 
1 philosophy 
of science 
they 
have 
to be c 
onsidered as 
complex 
indie 
es or 
even as 
theoretical 
constructs 
and 
not as 
units 
directly der 
ived from 
empiric evidence. 
2 BACKGROUND THEORY 
Ba 
sed on 
general 
th 
eoretical 
stru 
cture 
mo 
dels 
o 
f the lan 
ds 
cape at 1 
east 
four 
di 
fferen 
t 
types of fu 
nc 
tional rei 
a 
tion 
ships 
can be 
P 
ostulated. 
A 
s indicated 
in F 
igure 
2, 
these 
a 
re the fol 
lowing: the 
ver 
tica 1 
in 
terrel 
at 
ionship betwe 
en the geo 
c 
ompo 
nents 
in 
a spa 
ti 
al referenc 
e 
unit, the 
late 
rally 
oc 
curing 
processes 
between th 
e 
sp 
atial 
en 
titles 
> 
the endogen 
ou 
s, solar an 
d 
an 
thro- 
po 
genous 
processes, 
which are 
f 
r equ 
ent ly 
CO 
nsider 
ed 
as margi 
na 
1 conditi 
ons, 
and 
ev 
entual 
iy 
the temp 
or 
al aspect 
o 
f al 
1 the 
be 
fore 
mentioned 
relationship 
s. 
In 
pr 
actice 
» 
however, 
th 
e problem a 
r 
ises 
that 
fi 
eld me 
as 
urements ca 
n 
hardly be c 
a 
rr ie 
d out 
in 
spa 
ti 
ally and 
temporally 
adequate 
in 
terval 
s . 
Hence, th 
e 
attempt is 
mad 
e to 
in 
direct 
iy 
conclude f 
ro 
m a few pro 
P 
erti 
es of 
ge 
o-comp 
on 
ents to 
th 
e above 
ment 
ioned 
pr 
ocesse 
S, 
which cau 
se 
their prop 
e 
rtie 
s and 
pr 
operti 
es 
of other g 
eo 
components 
( 
e.g. 
by 
me 
ans of 
the concept 
of 
indicator 
P 
lant 
3 ) . 
Fo 
r this 
P 
urpose an e 
xc 
lusive sup 
e 
r pos 
ition 
of 
thema 
ti 
c maps seem 
s 
not to be s 
u 
ff ic 
lent, 
as 
they 
do 
not reflec 
t 
a "geocom 
P 
onen 
t in 
se 
" but 
c 
omplexe typ 
es 
, which are 
the 
basis 
of 
se 1 
ec 
tive feat 
ur 
es which 
can 
be 
CO 
nstruc 
te 
d on poss 
ib 
ly concurre 
n 
t pr 
inai 
pl 
es. 
He 
nee, the 
th 
ematic and 
sp 
atisi 
limitati 
on 
of the 
se 
types i 
s 
, m 
ainly 
dependen 
t 
on the und 
er 
standing o 
f 
th 
eory, 
th 
e resp 
ec 
tive proble 
ms 
and the in 
t 
eres 
t and 
knowledg 
e 
of the auth 
or 
of the map 
. 
Wi 
thout 
an 
y doubt, 
these unit 
s 
are not only 
of 
great