Research Methodology - Historical Land Use/Land Cover Classification Using Remote Sensing - page 128

Different Classification-Data, AlgorithmsApproaches, and Levels/Accuracy,

120

100

80

60

40

20

0

MLC ASTER-15m-May-2005 MLC Landsat-TM-30m-May-2007 MLC Landsat-MSS-60m-June-1975

NN ASTER-15m-May-2005 NN Landsat-TM-30m-May-2007 NN Landsat-MSS-60m-June-1975

SVM ASTER-15m-May-2005 SVM Landsat-TM-30m-May-2007 SVM Landsat-MSS-60m-June-1975

Average/Level based

Fig. 5.46 Illustrate the accuracy assessment values presented in Table 5.6 visually

MSS-data had insufficient spatial resolution), the multi stage classification

approach and the MLC-classification algorithm.

5.8 General Classes Classification

Drought is one of the main characteristics of large areas of the ERB, since vari-
ation in lands and natural coverage is partially measured by average precipitation.
Geological and geomorphologic characteristics and soil types change depending
on the availability of water or the climatic risks affecting the area. One of the more
satisfying results of setting general classification controls representing the natural
coverage and land uses of basin areas falling outside the borders of water agrarian
projects, is the appearance of planted lands with trees in bare, uncultivated lands as
shown in satellite images, particularly, in relatively dry areas with dominated light
color soils. This may occur because of the ratio between dimensions of pixel in
TM-30 m data, and the distance between planted trees within one field. In the
region of Aleppo, pertaining to the widely spread Aleppo-pistachio and olive trees,
the distance between every two trees is estimated with 8-10 m (Fig. 5.47 ), i.e.,
there will be an approximately 16 trees in each pixel of the TM-data. Because the