Page 799 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

In the first step of the method, re-classify ordering is done, i.e. the pixels are re-sorted in order of priority from top to
bottom. in the second step, the allocation is done and, in the cells, where the conflict exists, the allocation is made using the
decision line drawing. The spatial allocation operation continues until the intended area is allocated.

Fig 5. MOLA decision

5. Experimental Results

Table 3 shows the matrix of probability changes in SimWeight method. Also, a comparison matrix (Table 4) has been used to
measure the validity and accuracy of the proposed method. This matrix was formed for 4 changeable land uses. Then, general
accuracy criteria and kappa index were used to analyse matrix’s components. These criteria represent the accordance between
reality and the simulated model. One of the advantages of the kappa index is the use of all comparison matrix values to calculate
accuracy. Eqs. (5) and (6) show how to obtain the overall accuracy and the kappa index from comparison matrix [21][22].
Moreover, Table 5 shows the number of changeable pixels per user for the years 2002 to 2008 in SimWeight method. Finally,
Fig. 6 shows the prediction map with SimWeight method.

= ∑ (5)

=1

= ∑ − ∑ ∙ ⁄ 1 − ∑ ∙ (6)

=1 =1 =1

Table 3. The matrix of probability changes in SimWeight method.

Agriculture Residential areas Wasteland Green space
Agriculture 0.9344 0.0091 0.0102 0.0463
Residential areas 0.0300 0.7507 0.0221 0.1972
Wasteland 0.0992 0.0235 0.7684 0.1089
Green space 0.0857 0.0787 0.0074 0.8283

Table 4. Comparison matrix in SimWeight method.

Simulated model
Reality Total
Wastelands Buildings Parks Agricultural
lands
Wastelands p11=0.274 p12=0.032 p13=0.009 p14=0.034 p1t=0.349

Buildings p21=0.027 p22=0.342 p23=0.005 p24=0.018 p2t=0.392

8