Introduction - Methods for Multilevel Analysis and Visualisation of Geographical Networks - page 11

In the first section, Céline Rozenblat and Guy Melançon - a computer scientist
with 5 years of practice and training in geographical issues - describe the “multilevel
and multitheoretical framework for geographical studies” that they developed. This
paper provides the initial formalisation of the process of creating a coherent corpus
for the Spangeo project. Incorporating the concepts of centrality and proximity
developed in physics, information visualisation, sociology, and regional science,
the two authors describe how the concepts might be adapted to fit the geographical
context. Next, Alain L'Hostis describes how the concept of anisotropic space
developed in geography - in particular, by Tobler ( 1963 ) - contributes to this
approach. The vision of networks from the computer scientists' point of view, is
developed by Antoine Lambert, Romain Bourqui and David.
The methods developed by the SPANGEO team - in particular, the TULIP
software 1 designed by the LaBRI team in Bordeaux - are presented in a second part
with three papers. Guy Melançon describes the basic measures for identifying graph
centralities and clustering, and Jean-François Gleyze presents specific approaches
to the topological aggregation of graphs. Finally, Pierre-Yves Koenig explains
the various methods for analysing and visualising networks and their inherent or
exogenous properties.
The papers in a third part describe five applications of these methods, which are
presented at the global scale, the European scale and the city scale. Cesar Ducruet
describes sea networks, and Céline Rozenblat, Romain Bourqui and Guy Melançon
describe air traffic networks. At the global scale, sea and air transportation networks
exhibit different types of clustering based on internal network structures. These
transport networks have the methodological advantage of being quasisymmetrical,
unrestricted by physical limits, and global. The clustering methodologies of these
applications operate at the global scale and at the macro level of urban systems.
The third paper by Charles Bohan and Berengere Gautier presents a micro-level
analysis of multinational firms' behaviour at the global scale and introduces space
as exogenous variable that makes it possible to compare worldwide networks
of corporations through graphs and DAGMaps. Marie-Noelle Comin describes
research networks at the European scale that define classes of cities according to
their specialisations, associations or type of research, alternating between the macro
level of the urban system in which groups are based on topological proximities
and the meso level of processes occurring within cities due to the diffusion
of innovations and specialisations. Finally, Patrice Tissandier, Trung Tien Phan-
Quang and Daniel Archambault couple topological and geographical proximities,
describing the polycentrism that emerged in cities from 1975 to 1999 and describe
commuting networks, which are explicitly positioned at the meso-level of the cities
because they involve people's daily movement but also involve the macro-level
because people increasingly fragment their work and home lives between distant
locations by commuting weekly or tele-commuting.
1 The TULIP Software is free and open access: http://tulip.labri.fr/