——————— RÀ
me n i T
€
5. Results
5.1. Spectral signatures
For the categories studied at the given times of the year the following
conclusions can be put forward:
5.1.1. Radiance ranges
- Forest radiance is lower than that of the other categories in all spec-
ral bands. Conifers have lower values than deciduous trees. except in
band 7. Pinus svlvestris stands, often more widelv spaced than those of
Picea abies or Pinus corsica,may have reflectances similar to those of
grassland in band 4 and 5, but similar to those of cultivated land in band
6 and 7.
— There is onlv a small difference between the late winter scene recorded
at the end of March and the spring scene of May. Deciduous forests are
more easy to separate on the March scene because the greening of the tree
tops in early spring gives rise to confusion with grassland especially
along border lines.
- In band 4 and 5 the radiance of cultivated lands is higher than that of
grassland whereas in bands 6 and 7 the opposite is true, which is consis-
tent with data from literature (2). This inversion is more striking in
the March scene than in the May scene, especially in bands 6 and 7. This
may be due to the fact that in May the winter grains already cover their
soil completely and therefore appear similar to grasslands.
5.1.2. Discriminatory capability of discrete spectral bands
In this respect band 5 appears to be the best both on the March and the
May scene. Radiance ranges of all categories are well separated with a
minimum of overlap. Band 4 is almost as good as band 5 but in bands 6
and especially in band 7 overlaps between categories are important.
In band 6 e.g. grassland and cultivated land are almost undistinguishable,
and in band 7, which has the widest radiance range, cultivated land over-
laps both grassland and forests, the latter overlap being the greatest
on the May scene.
5.2. Map-like computer printouts
The map-like printout of fig.4 was produced by the NMAPW programme and
represents the radiance level classification in band 5 on the March scene.
A satisfactory correlation with the NGI map at appromimately the same scale
is already apparent.
The printout of fig.5 was generated by the EXIO | programme from band 5
data of the March scene but with preset radiance thresholds. Fig.6 was
generated by CANAL from data belonging to all spectral bands from the
same March scene. All.thresholds were statisticallv derived by programme
STATS.
For the Ma
EXIO 2 pri
out (fig.9
Percentage
retained v
such on ti
by electrc
Résults a1
Table T. (
‘
€
Deciduou:
Conifero
Grasslan
Cultivat
Unclassi
These res
- It shou
were re
The mos
forests
— The dif
errors
on the
Moreove
the res
nized a
- As for
crops c
transit
towards
test si