Information Technology Reference
In-Depth Information
•
a,b,c: all the three elements are different, and there is probably a deeper functional
relation between the class involved.
3.
as far as the number of changes is concerned, all the cities have been facing a
descending trend, that leads them to keep the existing distribution of land uses.
4.
there are some differences among cities concerning the different growth of some
kind of land uses, especially public and private services and discontinuous
residential urban fabric;
5.
some classes are faster than others to spread, and the accuracy to examinate
transition rules must be valuated according to this issue.
The last observation is a consequence of the results obtained through the first
simulations. In fact, as a starting point, we adopted a constant threshold on probability
for skimming the list of rules drawn from the comparison, in order to choose only the
most significant ones; the threshold was 0.1. Evidently, the word “significant” has a
wide range of meanings depending on the class it refers to, because after applying
these rules, results showed that this threshold was good for some classes, but too high
for others, especially the first six classes, that represent human constructions. The
sequence of maps generated through simulations, that should be as close as possible
to the real sequence, is too conservative for Milan, and fails especially in reproducing
the fast growth of the first transition; better results are available for Prague and for
Palermo (except in the 26 years long transition), that are affected by a slower growth.
The error in the maps showed in Table 3, calculated as the percentage of different
land uses assigned by the CA to each cell in respect to the real map, depends directly
on the speed of growth of each city, because transitions lasting the same number of
steps do have different errors.
Table 3.
Summary of errors in each transition (%); in parenthesis, the span of the transition in
number of steps
CITIES
TRANSITION 1
TRANSITION 2
TRANSITION 3
Milan
34,5 (3)
23,8 (5)
15,7 (6)
Palermo
15 (3)
33,5 (9)
2,1 (3)
Grenoble
4,8 (4)
26,6 (7)
9,1 (5)
Prague
8,4 (5)
12,4 (7)
3,2 (3)
5.
Conclusions and Further Research
Fundamental aspects of this method have been showing their capability to reach the
goal, but some modifications have to be made in order to take into account the
problems arisen. First, it is necessary to define precisely the threshold, which must
depends on the city (and, as a consequence, on available data), and on the particular
class considered.
Moreover, it is necessary to investigate quantitatively the role of stochasticity in
determining the difference between the real map and the simulated one. In fact, even
if we have a list of all the rules with their exact probabilities, we could not obtain two
identical maps after simulations, because of the different combinations of
probabilities.
