ferous forest,
gories were
25.000 topo-
ur Landsat
e discrepan-
t over or
somewhat sim-
ere inclusion
| would cause
n Belgium
on living in
ted to a large
s in the Dinant
roximately
d the Famenne,
edmont formed
y the loess
ite: mean
99:0.
'astern part
» central part,
has been cut
| very humid
1e to the imper-
soils are found.
this central
remaining
gher lying
tic nature of
] by steep val-
)hic oak-horn-
xed copsewood
conifers have
yo small in
symbolically.
ir general
much over the
ositioning
of course more
atter may be
- 1499 -
tôtisidered as an important purpose of our work. On the soils of the
central depression which are rather difficult to cultivate there is evi-
dence of a shift towardscattle farming and some 80% of the non forested
area is now under grassland.
2. Data products
The two scenes studied, both recorded by Landsat I, mav be specified as
follows:
Scene ID Origin Recording date
17242-10071 NASA Corrected 22/3/1973
1-213/25-cycle 78 TELESPAZIO Corrected 17/5/1976
Of the March scene a multispectral colour composite (bands 4,5 and 7) was
made by superposition of diazocopies (3). Its study by optical means was
important for the choice of the test site representative of the Famenne
natural region.
The main study was performed on the computer compatible tapes (CCT). Due
to some recording error the southern 20% of the test site was not accessi-
ble on the May scene.
Computations were performed on the IBM 370/158 of the Catholic University
of Leuven by means of the ORSER software package (Pennsylvania State
University) (1). For the transcription of the Telespazio tape format
into ORSER format, the COPTRAN programme written by W.Mehl (J.R.C.Euratom)
was used. The EXIO programme (Extended ratio mapping) a variant of the
ORSER NMAPW programme, was arranged by H.Gulinck.
3. Software description
NMAPW This programme classifies radiation levels in each spectral band,
assigns. symbols to each class, and produces maplike outprints
where each symbol corresponding to a distinct pixel can be identi-
fied by its scanline and elément number. This operation is essen-
tial for the positioning of pixels on the map and on the terrain.
STATS This programme starts from "training areas" delimitated on the
NMAPW print and recognized as homogeneous as far as landuse is
concerned. For each homogeneous polygon the programme calculates
the vector of the mean, the vector of the standard deviation and
the variance-covariance matrix, using the information contained
in one or several spectral bands. Optionally it can compute
correlation matrices, numerical data, frequency histograms for
specified channels as well as eigenvalues and eigenvectors for
the variance - covariance or correlation matrices.
CANAL This is a supervised classification programme, based on canonical
analysis. Inputs are the STATS results obtained on chosen traininc
areas. and: more especially the mean vectors and variance-covariance