004
y tv
— UD
EEE} Contre riallOther
Houstig|Cth
Othe efe
he r[ Commercial
om Conmercial
EE Hous Hg|Commerchl
[I Othe Housing
Com all Housing
Haus ing] Housing —
Figure 2. Cross-classification map of circular | rectangular
neighbourhood type for 250m resolution
When simulation results of large and small neighbourhoods
were assembled, the differences between them were not
identified at the spatial resolutions of 50m and 100m
respectively. The decrease of KAPPA index from 0.82 to 0.76
for both spatial resolutions was obtained. This indicates that
model outcomes i.e. land-use maps obtained are quite similar
when varying two neighborhood sizes. In addition, the graphs of
FD and class area measures do not depict discordances. From
the visual inspection of the cross-classification maps of 50m
and 100m resolutions, new housing land-use areas were
detected when simulations are performed for the small
neighborhood size.
The simulations results point out that with the increase of a cell
size KAPPA index decreases from 0.82 to 0.56 for 50m to
500m cell sizes respectively. When the CA model was
performed on 250 m cell size, visual inspection of the cross-
classification map (Figure 4) showed that large neighborhood
size produces bigger changes in commercial land-use class
areas than those produced for small neighborhood size. The
class area graph (Figure 3a) indicates discordance in both land-
use classes when varying neighborhood size. Discordance starts
already in the radius of 10km for both commercial and housing
land-use. Fractal dimension graph valucs (Figure 5b) for 250m
cell size reveals similar and were stable for all spatial
resolutions. For 500m spatial resolution, class area metrics
graph demonstrated more discordance in commercial then in
housing land-use type.
89
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
70 : 450
| i 400
f 60 i E
x t
: 350 x
$ 50i g
\ 3 —
$ = 300 =
S A 5
346; NUN : 250 B
§ NV 200 £
5 30 im N S
2 i ^ NON 150 $
3 20 | x NEN <
< i X NN 00 $
é 10 | \ NY ö
= i — 9 | 50
o | + +
pe Roe - E D m 0
0 10 20 30 40 50 50 70
Radius (km)
a —e— commercial for circular n = commercial for rect. n
—&-— housing for circular n —7414-- hous:ng for rect. n
1.8 4 14
|
1.7 | 1.35
= | |
ê 16 113 P
€ x. | i
5 1 a eR | 5
: 5 =" He | 12 5
|
5 14 et RÉ | 12 f
£ = <= | 2
g Se i s
2:13 = { z
Ë 1 ii 11155
RE = e. = f à
S | = ges Go —— R^ | =
11 | i 1.05
|
1 T - - rr | 1
0 10 20 30 40 50 60 70
Radius (km)
b —e€—- commercial for circular n = commercial for rect. n.
—a— housing for circular n x — housing for rect. n
Figure 3. Spatial metrics plots for circular and rectangular
neighbourhood type for 250m resolution a. Class area; b. Fractal
dimension
5. CONCLUSION
The results of this study indicate that there are impacts of
changing CA elements on urban growth modelling especially
with respect to changing of neighborhood size and type. The
proposed approach represents an exploratory method of
sensitivity analysis that can contribute to finding of the
appropriate neighborhood size and type for a CA model. The
results reported in this study indicate that KAPPA statistic does
change for different CA elements when varying spatial
resolutions. However, CA model responses are different
depending on the spatial metrics approach for neighborhood
size and type, and indicate that the discordance in generated
land-use classes is related to increase of the spatial resolution.
It is worthwhile the efforts to expand this study to include a
larger number of different CA element configurations and
spatial resolutions as well as spatial metrics.
The selection of proper configurations of CA elements in urban
CA models is important, since they can generate different
model outputs. Therefore, the SA in CA modeling is a
mandatory process to obtain better and more realistic modeling
output scenarios. It is vital to understand the limitations of the
CA model results pending on the impact of the variation of used
CA elements in order to make proper decisions in the land use
management process.
