Geography Reference
In-Depth Information
class of artificial surfaces, were classified with low accuracy, using the available
remote sensing data. If this is insufficient for other studies about the ERB, topo-
graphic maps offer the possibility of digitalization, from which the data can be
extracted with very high accuracy. The use of other remotely sensed data with very
high spatial resolution (e.g., IKONOS) is also recommended.
The problem in the separation and classification of the Earth features in the
study area is the classification of the marginal land. The ability to classify these
lands is linked to several factors.
The temporal factor: some lands were covered with temporary natural vegetation
that grew during spring and at the beginning of summer (from March to early May).
There were also very small areas covered with seasonal and permanent vegetation.
Some of these areas could not be spectrally separated or classified from the sur-
rounding bare areas because their spectral and spatial resolutions were insufficient.
Another reason was the dispersion of vegetation that dominated the spectral reflec-
tance of dry soils, particularly those that had light colours. Also, because of the
presence of the natural vegetation during the synchronism season with cultivation of
winter crops, there emerges the problem of spectral integration/mixing of these
vegetation with one or more types of the agricultural crops classes. The presence of
these marginal lands in the sensory data of August, led to the disappearance (or semi-
disappearance) of the spectral correlation problem between the marginal lands that
were covered with the temporarily natural coverage of vegetation during the spring
months. Between the lands with agricultural crops this natural coverage almost
vanished in August because of the absence of precipitation and domination of
drought. But, in contrast, the remotely sensed data taken in August had the problem of
spectral correlation of the marginal lands with the spectral characteristics of the
fallow lands, especially if they were covered with light soil.
The spatial factor: the presence of these lands within the irrigated agricultural
projects increased the problem size since more details were required concerning
the credited classification system levels, where green areas were classified into
several classes/agrarian crops. The presence of marginal land outside the borders
of the irrigated areas was a secondary problem with only slight effects (here, a
general degree of classification was required, i.e., 5 classes, where green areas
were classified into two agricultural lands with all of their crops and classes, and
natural vegetation).
The climatic factor: the rain-element determined prevalence, location and density
of the natural vegetation, and as a result, it controlled the spectral behaviours which
changed permanently according to time, place, kind of soil and amount of precipi-
tation. Consequently, the spectral behaviour of these natural plants might look
similar and correlate with a spectral behaviour of a crop (barley for example). This
behaviour is likely to change across time from one year to another, and perhaps even
in the same season and location/field, since, the natural plants may correlate with
other spectral response of crops other than barley (e.g., wheat).
LANDSAT-MSS-June-1975 data. By classifying the study area using several
scenes included in one mosaic-scene, it was possible to make classifications for
only three classes (i.e., the cultivated areas, uncultivated areas and water areas).
Search WWH ::
Custom Search