From Fractal Urban Pattern Analysis to Fractal Urban Planning Concepts - Computational Approaches for Urban Environments - page 26

Fig. 2.7 ( a ) A town section of Basel and ( b ) the curve of scaling behavior of the correlation
analysis (Source: digitized topographic maps)
This algorithm avoids the artifact of the k-means algorithm, which is based on
the centroid of experimental data, which is often not one of the experimental data,
whereas in the algorithm used, the representative of each cluster is necessarily one
of the initial data forming the cluster. In this project, we analyzed a set of 49 town
sections from the data set for nine European cities: Besançon, Cergy, Lille, Lyon,
and Montbéliard in France, Brussels and Charleroi in Belgium, and Stuttgart and
the Ruhr area in Germany.
Finally, five different morphological classes were distinguished. The first class
may be associated with districts with ribbon-like semidetached housing typical of
not too densely urbanized centers or pericentral zones as in, say, cities belonging
to the Flemish part of Belgium. Whereas for these curves the “root-like” shape
(Fig. 2.7 b) of the curve is very significant, it is less pronounced for a second group
of patterns which are typical for dense historical city centers. The town sections
with the big “Corbusier” buildings again form their own morphological class, and
the new towns also exhibit peculiar shapes of their curves of scaling behavior.
This is due to a very typical mix of individual detached housing and apartment
blocks, within large green areas and public places. Finally, very particular shapes are
observed for commercial and industrial zones dominated by huge buildings, where
intra-building distances are considerable, erasing the usually observed deviations
for small distances.
While we find globally similar information to that given by fractal dimension
ranges, closer analysis of the curves gives details about distance ranges for which
substantial changes in spatial organization occur, or alternatively, for which the
parameters are stable or not. Hence, the information contained in these curves turns
out to be complementary to that of the fractal dimension D which remains a useful,
albeit rather synthetic indicator.