Table 1:Overview of the database on vessel movements,1996-2006 1996 2006 2006/1996 No.Ports 975 1,240 1.27 No.Vessel movements 176,439 390,740 2.21 No.Vessels 1,759 3,973 2.26 No.Operators 497 720 1.46 Total slot capacity (TEUs) 3,352,849 9,590,309 2.86 Share world fleet(%TEUs) 92.15 97.91 +7.75 Source:own elaboration based on LMIU data The global network was modelled based on vessel characteristics,ports of call,and vessel movements.The first result is a global network composed of weighted and non-directed links between ports,which can be analyzed in two different ways.On the one hand,vessel circulations create a graph of direct linkages(GDL)based on the successive ports of calls(i.e. 点 from port A to port B and from port B to port C).On the other hand,it can be argued that two 三 ports are also connected if they belong to the same liner service or loop,although they are not adjacent calls;a graph of all linkages (GAL)thus adds indirect linkages (i.e.from port A to N port C).In the GDL,Le Havre and Tokyo are never connected by a direct link,whereas,in 西 the GAL,this connection might occur inside a pendulum or round-the-world service.The GAL is the overlap of all individual complete graphs created by the circulation of each vessel. These two dimensions of the same reality(GDL and GAL)may exhibit distinct features in terms of network structure and port hierarchy.In order to reveal the structural properties of 岩 the two graphs for each year of observation,we apply conventional measures derived from graph theory,which were originally applied to transport networks by Kansky(1963)and from complex systems theory,referring to the works of Barabasi and Albert(1999)and Watts and Strogatz(1998).This set of measures provides clear evidence about the nature of the network based on topological properties (see Ducruet and Rodrigue,2011 for a review of network measures). One limitation of the data is that it ignores how many full or empty containers were truly handled by ships and ports.In reality,some vessels may not be fully loaded,since their passage in a port does not always include stevedoring activities (e.g.a port visit in the framework of bunkering activities).However,with reference to the observation made by Joly (1999),the linear correlation in our data between vessel traffic and port throughput is very 6Source:Containerisation Intemational 1010 Table 1: Overview of the database on vessel movements, 1996-2006 1996 2006 2006/1996 No. Ports 975 1,240 1.27 No. Vessel movements 176,439 390,740 2.21 No. Vessels 1,759 3,973 2.26 No. Operators 497 720 1.46 Total slot capacity (TEUs) 3,352,849 9,590,309 2.86 Share world fleet (% TEUs) 92.15 97.91 +7.75 Source: own elaboration based on LMIU data The global network was modelled based on vessel characteristics, ports of call, and vessel movements. The first result is a global network composed of weighted and non-directed links between ports, which can be analyzed in two different ways. On the one hand, vessel circulations create a graph of direct linkages (GDL) based on the successive ports of calls (i.e. from port A to port B and from port B to port C). On the other hand, it can be argued that two ports are also connected if they belong to the same liner service or loop, although they are not adjacent calls; a graph of all linkages (GAL) thus adds indirect linkages (i.e. from port A to port C). In the GDL, Le Havre and Tokyo are never connected by a direct link, whereas, in the GAL, this connection might occur inside a pendulum or round-the-world service. The GAL is the overlap of all individual complete graphs created by the circulation of each vessel. These two dimensions of the same reality (GDL and GAL) may exhibit distinct features in terms of network structure and port hierarchy. In order to reveal the structural properties of the two graphs for each year of observation, we apply conventional measures derived from graph theory, which were originally applied to transport networks by Kansky (1963) and from complex systems theory, referring to the works of Barabasi and Albert (1999) and Watts and Strogatz (1998). This set of measures provides clear evidence about the nature of the network based on topological properties (see Ducruet and Rodrigue, 2011 for a review of network measures). One limitation of the data is that it ignores how many full or empty containers were truly handled by ships and ports. In reality, some vessels may not be fully loaded, since their passage in a port does not always include stevedoring activities (e.g. a port visit in the framework of bunkering activities). However, with reference to the observation made by Joly (1999), the linear correlation in our data between vessel traffic and port throughput6 is very 6 Source: Containerisation International halshs-00538051, version 2 - 15 Jul 2012