Geoscience Reference
In-Depth Information
many applications in geography). The change of state at time
t
to another at time
t
+
1 is governed by a transition rule relating this change to the state of the neighboring
cells in
t
.
The first applications of CA in geography are very close to the “game of life”,
with two possible states for the cells (built-up and non-built area, for example), and
transitional rules relating to Moore's neighborhood (eight neighbors), in order to, for
example, simulate the spreading of urban growth in space. The counterpart of this
simplicity is the lack of realism that it implies. Therefore, models richer on the
thematic plan have quickly been developed with modifications of the original model
in several directions:
- taking into account a more significant number of states. In order to simulate,
for example, land use dynamics, a variety of categories have been considered in the
applications (residential, industrial, commercial, green spaces, etc);
- taking into account the effect of land use, not only of the immediate
neighborhood but also of more distant neighborhoods, to formalize the cell transition
from one state to another. A weighting system, depending on the distance and the
land use concerned, is often associated with such an approach (the effect of a
commercial activity, for example, impacts a smaller area than that of an industrial
activity);
- taking into account factors other than the effect of the neighborhood: the
accessibility relative to the existing network of transportation, the physical
suitability and the existence of planning rules.
One of the first models developed in geography in this sense is that of White and
Engelen's [WHI 93, WHI 94], initially applied to simulate the development of the
city of Cincinnati from its first urban core. They are interested in the spatial
configuration of the three basic categories of land uses (commercial, industrial and
residential), and consider in the transition rules of a land use to another several
levels of contiguity to assess the effect of the neighborhood and a factor of physical
suitability. The transition rule is represented there in the shape of a function linking
these different factors and to which a random term is added. According to its state at
time
t
, its location, physical suitability and the states of the neighboring cells up to a
degree 7, this function calculates the potential for transformation in each of the
possible states for each cell. The highest potential determines the state of the cell at
time
t
+ 1. Such models are rather more operational for the geographer but the
consequence is nonetheless that the model's formal properties are less controlled.
This modeling framework has given rise to numerous applications and the
development of a modeling platform called Metronamica [VAN 05] has led to a
