Geography Reference
In-Depth Information
calibration with linear values as gain and offset coefficients as unnecessary. This
technique had the robust advantage of the ability to create a scene comparison
even though values were not available or wrong.
The international hierarchical classification scheme (LCCS) of FAO was fol-
lowed as guide in the classification processes. This approach defined and deter-
mined the LULC-classes to be included in the classification/s. These classes were
defined before starting each automated supervised classification procedure.
The classification of the remotely sensed data was based on the traditional
pixel-based classification method. The results of classifications were always pre-
sented as thematic maps. The results of the various tested approaches and algo-
rithms
of
classification
on
the
various
obtained
remote
sensing
data
were
interpreted based on the accuracy assessment method.
In this study, several automated classification approaches (i.e., one-step, and
multi-stage classification) and several algorithms (i.e., MLC, NN, and SVM) were
tested on several remote sensing data (LANDSAT: MSS, and TM; TERRA:
ASTER fused with corrected LANDSAT-ETM+/SLC-OFF/), to find the optimized
approach and algorithm. The multi stage classification approach and the MLC-
algorithm harvested the best results (see
Chap. 5.7
).
The classification of coarse resolution (spatial and spectral) data like LAND-
SAT-MSS in relation to its geographical location of ERB, was suitable to produce
thematic maps of the five wide general classes for the whole large area of the ERB
and to represent the spatial distribution of the one irrigated areas class. These data
had not the ability to classify any more detailed classification level (e.g., agri-
culture). LANDSAT-TM data were more suitable for classifying the general
classes and the irrigated areas. However, it was less suitable for classifying the
detailed agricultural classes. Finally, the low spectral resolution ASTER-data of
only three bands was less suitable in comparison to TM-data, although they had a
higher spatial resolution, i.e., 15 m. However, after fusing them with the
LANDSAT-ETM+/SLC-OFF/corrected data to increase the spectral bands to six
bands, these data harvested the best results. In general, the classification of the
agricultural features using TM and ASTER- ETM+ data was very good over the
State achieved irrigation projects (e.g., the 21,000 ha project, Maskana-East, etc.),
where the individual planted fields were relatively large and thus classifiable in
regard to the used remote sensing data, and the diversity in LULC-features was
small. These were changed starting from the TM-data of 2007, where the fields
became smaller and the diversity of planted agricultural types became more
widespread. The diversity was acceptable over the State and farmer-cultivated
irrigation areas (e.g., Maskana-West), where the private holdings were varied from
small fields to very large fields. However, the classification results were unac-
ceptable over the very old cultivated areas located on the Euphrates River banks,
where the holdings were very small with great diversity in cultivated agricultural
features. Therefore, this area requires remote sensing data with higher spatial and
spectral resolution (e.g., IKONOS).
Search WWH ::
Custom Search