' the
able
ach
test
ori
ent
d by
hite
le
of the
nito_
reas
ver
ely
species
he
Figures (06),(07) and (08) illustrate the phenological
sequences of the main crops in the zones being examined. These
calendars were used in both the choice of Landsat scenes and
the identification of the content of these scenes. Dates of
purchase of Landsat scenes are given in figures (06),(07) and
(08); these also represent all the scenes available in Spring
and Summer up to 1977 when the research got under way; Autumn
and Winter scenes were not chosen as they held little signifi
cance; there was a disappointing number of summer scenes because
of lack of purchases and cloud-cover.
al) ,a2) ,a3) and a4) groups were established on the basis
of available data through a multi spectral and multitemporal
analysisof the four channels. The groups which do not show, or
show only slight variation in signature c) and d) (urban
settlements and unproductive surfaces) were identified solely
by multispectral analysis and checked subsequently.
A different procedure was used in the identification of
a5) and b) as an analysis of Landsat scenes alone was not
absolutely reliable. Tree crops in the area under examination
(orchards,vineyards) do not cover the soil completely and give
intermediate signatures somewhere between the underlying land
and the vegetation signature. The signatures vary then from
case to case (not only where the land, but also where the pixel
layout is concerned).
The same can be said for woods: thick woods can be identified
from Landsat data whereas sparse woods can be mistaken for
underlying land . The two groups were checked against conventio
nal black and white aerial photographs to ensure the maximum
degree of accuracy.