422 
3. The third one concerns the repartition of the vegetation. 
Studying the significance of the limits, we were led to 
distinguish the limits which are related to the superficial for 
mations, among the others limits which reflect the vegetal 
competition between species, or the way the ground has been 
used by man. The two figures below show the vegetal limits 
related to structural limit of superficial formations (overmarked 
figure 2) and the spatial coincidence between these limits 
(overmarked figure 3), and the structural limits (figure 1). 
.1 ___ 
s 
Figure 2 : Soil and vegetation sequences 
Therefore it seems possible to express in the aerial 
photography, the proper information on the structure of the 
whole volume of superficial formations, by the intermediary of 
the surface différenciation and the vegetal cover repartition. 
1.3. The methodology defined for image processing 
The first group of treatments applied to the images, were seg 
mentation on the plane of the reflectances, by establishing 
different classes of vegetation, in one dimension (primary 
channels, channels combined by vegetation index) and in two 
dimensions (two first axes of the P.C.A.). The results of these 
treatments were confronted with the ground reference and 
thereby we conclude that this kind of treatments do not permit 
to distinguish the most important vegetal domains, for they are 
formed by the same associations of vegetal classes. 
This first observation lead us to perform a sequence of 
treatments based on textural features. Textural images of 
gradient (Sobel) were calculated, a local density of gradient was 
extracted by the convolution with a 7x7 scanning window, and 
this textural information was mixed with the reflectance one , 
on the same level, in a Bayesian classification. The choice of 
that sequence of treaments was conducted by the reflection that 
natural photography is interpreted by human vision, putting a 
hierarchy upon limits, the strong one upon reflectance limits, 
and at an inferior level upon the textural ones (reflectance ar 
rangement). 
Now, we saw that the main limits from the point of view of 
structural study of superficial formations were separating 
domains with the same associations of vegetal classes, giving 
the same visual impression of colour, but with different arran 
gement of classes which give a visual impression of different 
textures. The way of expressing textural information at the 
same level that reflectance one, makes images where the im 
portant limits are very visible, and which are very near to the 
map of vegetation repartition we obtained both by 
interpretation and by field s.udy. The two figures below are 
showing the interpretation of the natural photography (which 
can be compared with the map of vegetation, figure 3), and the 
image resulting of the sequence of treatments described. We 
can see that this image is nearer from the map than is the 
interpretation of natural photography. 
Figure 4 : Photo-interpretation of the natural photography 
Figure 5 : Image resulting of ¡he studied sequence of treatment 
2 FRENCH GUYANA 
Fig. 1 shows the site under study. There, the granitic shield is 
covered with a clay-sandy material ('Sérié Détritique de Base' 
for geologists). Tropical forest covers the whole shield (more 
than 97% of this country area.) 
Studies made in Guyana by BOULET and al., starting from 
1976 show that the initial ferrallitic cover, developped from the 
weathered shield, is now out of equilibrium and is therefore 
being transformed. The initial ferrallitic soil is tranformed here 
into podzol, and in geographical space, this transformation is 
more or less rapid; these transformations act at the scale of the 
existing topography, that is here about 1 km; and the transition 
between the ferrallitic pole and the podzolic one may be very 
short (100 m.) or may cover the whole watershed.