Figure 6: Clusters of ports within the Atlantic network, 2006
Source: realized by authors based on LMIU data and TULIP software
1 Corresponding author
2 We used TULIP software created by the Laboratoire Bordelais de Recherche en Informatique (LABRI), and which was initially created for biology. It is today extensively used for social network analysis and in transport studies notably on air transport networks, intra and inter-urban commuter flows, and multinational corporations’ networks in the SPANGEO project (http://s4.csregistry.org/SpanGeo). The TULIP software is free, open source and available at: http://tulip.labri.fr/
3 Number of ports (vertices) directly connected to a given port through inbound and outbound direct maritime connections (edges). This paper does not measure a deeper degree (e.g. level 2, 3, or more) but this is used extensively in sociology and communication science for analyzing diffusion processes (Monge and Contractor, 2003). In the case of maritime transport, extending the measure of degree would highlight possible transfer steps in the vicinity of ports.
4 Sum of all possible shortest paths of the graph passing through a given port. This measure can reflect a “potential maritime accessibility” and can be interpreted as a level of intermediacy - or in-betweenness - condensing multiple insertions of ports within the networks of ocean carriers and their ability connecting various scales from the local to the global (Fleming and Hayuth, 1994).
5 We measure the observed connectivity by dividing the number of edges by the number of ports, and compare it with the optimal connectivity (i.e. assuming that all ports should be connected to each other). The Atlantic network is based on 307 ports (1,821 links) in 1996 and 351 ports (3,609 links) in 2006.
6 The methodology is applied on all direct and indirect weighted edges (traffic in TEUs) as the clustering metric.