3 : slaignt 
at the level 
tending 
1 by industri- 
xistence of 
the western 
he "dark" 
y the layer 
ea on still 
y haze. 
e that, in 
aks and 
s a higher 
crease seems 
, we can 
the signifi- 
dying : 
gally dis- 
flectances 
5 steps : 
these 
|j, using 
mixed, use 
in eventual 
= 803 = 
The first results lead to the following discussion: 
1. The method which has been worked out for the classification of urban areas 
yields a sufficient number of accurate classes as to delineate a correct urban 
perimeter clustering industry and housing. Nevertheless, differenciation of 
industrial estates and built up areas may compel use of supervised or assisted 
classification methods. 
2. This method also enables a good differenciation of suburbs with low densities 
of housing. It also allows, when pattern recognition is combined with the inter- 
pretatiwe analysis of classifications, a good identification of suburban infras- 
tructures such as airfields, highway rings, building sites and shopping centers. 
So it would be possible to measure from successive recordings the losses of rural 
surfaces due to the urban spreading into the countryside (housing and transporta- 
tion infrastructures). This study is in progress for the period 1973-1976. 
We hope to provide the first results in a further contribution. 
The comparison between such successive indexes of urban pressure seems 
to be very useful for regional and urban planning. It also provides an ecologi- 
cal interest. It enables indeed to compare the consequences of ex-urbanization 
process on rural land use elements, in definite homogeneous regions with regard 
to the landscape. 
3. At least, and this seems to be the most peculiar interest of the analysis, 
the identification of areas combining in a constant manner several classes of 
reflectances, points out the existence of urban macrostructural units in relation 
especially with the land use density and the age of built up areas. Such macros- 
tructures would not appear so clearly on any other document but a synthetic clas- 
sification based on multispectral spacecraft scannings.