he following 
s in all spec- 
s. except in 
han those of 
to those of 
ed land in band 
scene recorded 
| forests are 
ng of the tree 
| especially 
r than that of 
lich is consis- 
| striking in 
6 and 7. This 
ly cover their 
: 
> 
[arch and the 
arated with a 
- in bands 6 
nportant. 
listinguishable, 
ated land over- 
the greatest 
rogramme and 
the March scene. 
ly the same scale 
> from band 5 
s. Fig.6 was 
ds from the 
ed by programme 
For the May scene, there is an EXIO 1 printout from band 5 (£1g.7), an 
EXIO 2 printout (with cultivated lands masked) (fig 8) and a CANAL print- 
out (fig.9). 
5.3. Land use distribution 
Percentages of the test site area occupied by each of the four categories 
retained were compared with those of corresponding areas delineated as 
such on the NGI map. 
by electronic image analysis 
Results are given in table I. 
The latter measurements were made by planimetry and 
using a Quantimet 720 on masks drawn by hand. 
Table I. Comparison of surface percentages obtained by computer aided 
analysis of Landsat data and measurements on the map 
  
  
  
  
  
  
  
  
  
  
  
NGI Map EXIO CANAL 
Planimetry Quantimet | Bd 5| Bd 5 | Bd775) 
720 March | May May March May 
Deciduous 46.1 43.8 55 63.3 58 
65 59 
Coniferous 13.4 14.3 6 2.7 7 
Grassland 28.9 30.3 34 27 26.7 22 29 
Cultivated 11.6 11.6 5 7 7.3 5 4 
Unclassified 14 2 
  
These results call for the following comments: 
— It should be remembered that the map is dated 1966, the Landsat data 
were recorded in 1973 and 1976 
The most important inconsistencies are found in the categories coniferous 
forests and cultivated land 
- The difference for coniferous forest can be partly explained by positive 
errors in the drawing of the masks: many areas designated as coniferous 
on the map are very small indeed and have most irregular forms. 
Moreover many such tiny areas, covered by conifers are too small for 
the resolution 6f the Landsat MSS and may therefore not have been recog- 
nized as such by the computer. 
- As for the cultivated land the difference may in part be due to forage 
crops classified as pastures. 
But it also reflects the well known 
transition accelerated during the tast ten: years of mixed agriculture 
towards cattle farming. 
It is true that edaphic conditions inside the 
test site make it more suited for grassland than for grains