The 
; and even 
ind 
ation is 
for example 
supervised 
> specific 
nique 
Is a 
na. lt can 
s, the 
number 
be 
al 
| values, 
> the 
vith similar 
r a cluster 
In this 
Jo an 
e 
lar to a 
nformation 
iot 
y become 
j. the 
xploration 
tonics 
s and 
  
  
Fig. 4 Channel 5, linear stretch Fig. 5 Unsupervised classification 
LAND SAT data 
black younger lava flows 
middle grey older lava flows 
light grey basement 
white quaternary sediments 
vulcanic rocks sufficiently, we interpreted LANDSAT images to select areas which meet 
our requirements, Then we used our digital image processing system to display the 
selected areas and to enhance the image for further interpretation. Simple contrast 
stretching (fig. 4) separated very effectively various vulcanic units displaying a complex 
sequence of lava flows. Their spatial relationship and brightness indicate an age 
sequence, whereas colour differences (brown in certain flows, blue in others) may 
indicate differences in chemical composition. An unsupervised classification confirmed 
these results (fig. 5). Therefore this area among others was selected for sampling. 
Black and white aerial photographs at scale 1 : 62 000 were then used to verify the 
results of the LANDSAT evaluation , to refine the differentiation of the lava flows and 
to plan the field work. No supervised classification was attempted since ground informa- 
tion was not available. 
POTENTIALS AND PROBLEMS OF VARIOUS TECHNIQUES 
  
These two examples and many other case histories yield the following conclusions: 
1. Photointerpretation is an extremely flexible and versatile technique. It goes beyond 
simple identification and systematic analysis and includes deduction of new information, 
not directly containéd in the data. Results are dependant on the experience of the 
interpreter. Detailed mapping and digitizing can be very time consuming. Simple 
M ———À— u— 
I n REPRE RUE IN REN NAR CE is Lo CC M RE 
  
NNI mE ——