Research Methodology - Historical Land Use/Land Cover Classification Using Remote Sensing - page 113

The 4-region studyarea PrimarilyNon-VegetatedArea (s)

PrimarilyVegetatedArea (s)

LANDSAT-TM-May-07

Natural and Semi-Natural
TerrestrialVegetation

Cultivated and Managed
Terrestrial Area (s)

Fig. 5.34 Illustration the application of multi stage classification approach (chain-steps), using
MLC-algorithm, on the LANDSAT-TM-data of May 2007 with spatial resolution of 30 m

from those areas covered by the same mask. This results from using the SVM-
algorithm in the classification process. After finishing the multi stage hierarchal
classification, various classes resulting from each stage are collected and fused in
one scene that represents the LULC in the study area using ENVI-program (band-
math), or the ArcGIS-software (Fig. 5.36 ).
Obtaining 100, 90 % or other percentages reflecting accuracy and quality of
automated classification, does not necessarily mean that the percentage completely
represents LU distribution or the prevalence of LC on real ground. The accuracy
percentage of 100 % obtained from the classification of primarily vegetated areas
and non-primarily vegetated areas does not mean that the total area contains the
same percentage of classification. Of the primarily vegetated areas (e.g.,
100,000 ha) perhaps 1,000 ha are primarily non-vegetated areas. This error/s in
classification would then be repeated in each step or stage of the multi stage
classification approach. This means that the primarily vegetated areas class might
appear under classification of the components of the second level (i.e., the second
terrestrial and aquatic level underlying under the first primarily non-vegetated
areas level) within the used classification scheme (i.e., LCCS), although it should
have been classified and separated into the first level. So, the LCCS-principle of
classification should be strictly adhered to, that is separation between classes in
every level and every stage in classification system. Here the primarily vegetated
areas, (for example the 1,000 ha that had been classified incorrectly under the