Navigation – Plan du site

Cities in Worldwide Air and Sea Flows: A multiple networks analysis

Villes et flux mondiaux aériens et maritimes : une analyse de réseaux multiples
César Ducruet, Daniele Ietri et Céline Rozenblat


Les villes et les réseaux de transport sont liés à travers l'histoire en dépit de l’impact des télécommunications. Bien que chaque réseau de transport possède sa propre logique en termes d'acteurs, de lois et de régulation, la position des villes dans ces réseaux révèle leurs complémentarités. Cet article propose de comparer la structure et la position des villes au sein de deux réseaux mondiaux, aérien et maritime, choisis pour leur rôle essentiel dans le développement urbain et malgré leurs organisations spatiale et économique différentes. Trois niveaux de délimitation d'entités urbaines sont testés afin d'évaluer les effets de l'agrégation des noeuds sur la structure des réseaux, sur la hiérarchie de la centralité et sur les corrélations entre les réseaux. La mesure de la spécialisation et de la polarisation permet de proposer une typologie où les "gateways indépendantes" occupent une position de pont dans et entre les régions nodales. Cette recherche clarifie l’interdépendance à l’œuvre entre villes et entre réseaux dans les processus de mondialisation, régionalisation, et développement urbain.

Haut de page

Notes de l’auteur

This paper was presented at the S4 International Conference on « Emergence in Geographical Space », Paris, 23-25 November 2009.

Texte intégral

The authors would like to thank the two anonymous reviewers for their very useful comments on earlier versions of this paper.


1Global cities are developing through invisible flows of finance, services and communication. The economies of global cities generate more and more immaterial and material flows, allowing people to live, travel and meet. In the past decade, the total number of worldwide air passengers has more than tripled (from 1.5 billion in 1999 to 4.8 billions in 2009) and is expected to double again by 20301. Although the aviation sector represents only 2% of the European GDP2, air flows dynamics represent an important aspect of globalization. Passenger flights allow face-to-face meetings for business as well as tourism and family connections for migrants. In other transport modes, such as sea transportation, material flows also support the global division of labor as well as import/export of products and energy. The cost of sea transportation represents only 3% of the value of imports for developed countries and 6% for developing countries (UNCTAD, 2005); however, it is essential for the productivity of the global economy. Sea transport carries about 90% of world trade volumes, while containers alone occupy 70% of sea trade value3. Thus, connectivity and centrality in the main transport routes and flows are essential for urban development, in particular through the combination of higher value-added forms of transportation, such as airlines, and lower value added form in large quantities, such as maritime freights.

2Global cities gain a competitive advantage by being situated in the center of global transport networks, that is, at the center of production, consumption, or flows. Air and sea networks are developing according to economic growth (UNCTAD, 2006). There are numerous studies of ports (i.e. shipping networks) and airports (i.e. airline networks) from a transportation point of view, but no attempt has been made to analyze the position of cities in these two networks simultaneously.

3The aim of the paper is to reveal the complementarities between air and sea transport networks in shaping an urban hierarchy. Each network makes similar contributions to the overall urban hierarchy while maintaining its own specificities. The rest of this paper is organized as follows. We define how we conceive the relationship between the growth of cities and transportation in Section 2. Sections 3 and 4 present the methodology and the main results about the structure of the networks and the respective positions of cities. Finally, Section 5 provides the discussion and future research agenda.


4Although each transport mode possesses its own rationale of development through multiple strategies of actors and regulations, its growth remains bound to urban development. Urban population and production are at the basis of exchanges allowed by transport, which permits wealth increase and attracts new residents. The overall structure of transport networks and the role of cities inside each of them can thus be compared in order to evaluate the potential for each city to exploit simultaneously its centrality in one or several transport networks.

Cities and transport development

5The development of cities and transport are linked throughout history (Bairoch, 1985). The main roads traced by colonization and by industrialization are still relevant (Bretagnolle, 1999). Moreover, investments in industry and in transportation during the 19th century are still visible in Europe. The fact that both the North-West center of European Bank and one of the first European airports are located in Frankfurt is not a coincidence. Industries in the Ruhr area needed long-distance exchange of products and thus have invested heavily in transportation. The current global transportation framework shows a rather high concentration along main axes at the continental level or between continents and a multi-polar development of cities.

6A center-periphery model is no longer relevant to qualify a world composed of a hierarchy of global cities driving economic development (Veltz, 1996). Banks and stock exchanges control investments among global cities, while marketing, research and high services organize the world division of labor (Sassen, 1991; 2000), which produces the movement of goods and the daily mobility of people. Maritime freight flows foster the development of a productive system exploiting differences in taxation, availability of raw materials, labor cost and regulations between nations. The integrated economy facilitates the mobility of skilled workers traveling for training or to control production activities. Multinational companies, which invest far from their headquarters, must maintain control and access to their productive plants (Rozenblat, Pumain, 2007). Thus, air and sea flows complement one another in the control of the world economic system by actors who exploit urban centralities and connections in geographical and topological spaces.

Specificity and complementarity of air and sea transport

7Air and sea transport networks have remained largely separated in the literature. This is partly because scholars tend to specialize in only one transport mode (Brocard, 2009). Another possible reason is the scarce access to relevant disaggregated data from different transport modes to be fitted within a unique database at the city level. Numerous studies of air transport networks on a global scale have discussed how cities are positioned in the multi-level hierarchy of flows (Amiel et al., 2005); other studies are more descriptive in scope when ranking places (Derruder et al., 2007). Air transport networks are better compared with rail transport networks (Cattan, 1995). Global maritime networks have been studied empirically, but most scholars tend to focus on specific carriers or regions (see a review by Ducruet and Notteboom, 2011).

8At first glance, seaports appear bound to nautically accessible coastal locations, while airports can be located in nearly every city. Maritime networks are thus more geographically constrained than airline networks (Xu et al., 2007). While air hubs and hub ports have in common the redistribution function, gateway ports may serve very distant hinterlands through intermodal services, thus making their catchment area very dissimilar to airports serving the outlying urban region. The spatial and functional mismatch between air and sea transport should not hide their numerous similarities. As the two only global transport networks, they provide long-distance and round-the-world services. The relative ease of circulation of vehicles provides them with great spatial flexibility, because they do not need track infrastructure (except for port and airport terminals), resulting in vaguely defined and delimitated distributions (Rodrigue et al., 2009). In addition, they show remarkable similarities in terms of economic organization through global alliances and deregulations as well as comparable technological evolution with regard to economies of scale and increasing size of vehicles (Cullinane and Khana, 2000; Horner and O’Kelly, 2001). The establishment of hub-and-spoke systems is another important common aspect, giving privilege to intermediate locations concentrating flows for regional redistribution (Fleming and Hayuth, 1994; O’Kelly, 1998).

9Air and sea transport are also complementary. Their respective core markets are passengers and freight: air transport specializes in higher value products (e.g. computer components), while sea transport embraces no less than 90% of world trade volumes due to the weight of bulk commodities such as crude oil and minerals (Rodrigue et al., 2009). Cruise shipping, roll-on/roll-off, and ferry occupy specific niche markets at certain locations, while air transport concentrates on passenger mobility, competing more with railways at certain distances rather than with maritime transport. Operators are thus different companies; those advertising themselves as “global logistics providers,” such as DHL and UPS, do not provide sea transport services or operate only through their role as integrator or forwarder. Such companies do not replace airlines or shipping lines that maintain their know-how and innovate within their core activity (Frémont, 2009). On a regional level, air and sea transport may overlap with the urban hierarchy such as in East Asia, which possesses numerous multifunctional global cities (Rimmer, 1999; Taillard, 2004; Lee et al., 2008), growth triangles, and urban corridors (Rodrigue, 1996). Several important cities such as Hong Kong, Dubai, Brisbane, and San Pedro have thus developed so-called air-sea terminals where goods are shipped directly from one mode to another.

10Bogart (2009) noted that few studies exist about the complementarity between different transport systems. His study of the co-evolution of roads, canals, and ports during the English industrial revolution shows noticeable interdependencies among the different nodes and networks over time, based on the exploitation of spatial proximity and the respective functions of each network. The growing awareness that transport networks may be interdependent is also a core argument of Zhang et al.’s (2005) study of multi-layer infrastructure networks. They argue that interdependencies between two or more transport modes may vary in intensity according to the geographical level concerned and to their respective topology, function, potential inter-modal shifts, etc. One example includes the recent work of Jin et al. (2010), which applied a transport dominance measure to rail, freeway, and airline networks in China, highlighting the advantageous position of some large urban regions (capital region, Yangtze and Pearl River deltas) as well as the hub function of some Western cities for redistribution. Indeed, most studies have analyzed transport networks separately and on the level of countries (Gastner and Newman, 2003; Blumenfeld-Lieberthal, 2009) as a consequence of limited data availability.

11The interdependence between air and sea transport has so far received very little attention. Based on employment data by sector in European port cities, Ducruet and Lee (2007) highlighted a high correlation between air and rail transport due to their closeness as passenger-oriented activities. Their factor analysis also demonstrated the opposition between port cities specialized in port and maritime activities on the one hand (i.e. freight orientation), and port cities specialized in air and rail transport activities on the other (i.e. passenger orientation). Looking at how different transport activities are integrated in the portfolio of firms located in European port cities, Ducruet and Van der Horst (2009) underlined a similar opposition, with port, river, and maritime activities forming a distinct class having very few links with air and rail transport, the latter being less integrated than other activities in the transport chain, except with road transport. More recently and from a network perspective, Parshani et al. (2010) demonstrated inter-similarity between air and sea transport networks since high-degree airports tend to couple with high-degree ports based on geographic proximity, thus making the combined network significantly more robust to random failure. Although such research answers part of our initial question, we wish to further verify the influence of urban areas’ definition on the results. A look at how ports and airports can be said to form spatially and functionally coherent gateways is discussed before analyzing the two networks.

Defining gateway cities

12In their review of the research on globalization and transportation, Janelle and Beuthe (1997) address the geographical dispersion of transport demand, which occurs in parallel with the concentration of transport infrastructure and flows at “transportation corridors of high density and diverse transport options” (p. 202). Observations of the scattered development of transport terminals have highlighted the increasingly difficult coordination of various players, systems, and flows, resulting in congestion among other problems (Goetz and Rodrigue, 1999; McCalla, 1999; Slack, 1999; Hesse, 2004; Horst and De Langen, 2008).

13Janelle and Beuthe (1997) deplore the “absence of (…) disaggregated data” on detailed flows, which remains a strong barrier to the study of “specific cities and urban regions” (p. 206). One decade later, the review provided by Keeling (2007) about the progress of transport geography confirms that although “transport geographers have embraced the kinds of analytical challenges presented by changing airline networks, shipping systems, commodity flows, and inter-continental flows of people, goods, and information” (p. 220), there is still much to do in defining commonly accepted concepts and methodologies. We propose to focus on the concept of “gateway,” as it better describes the articulation of various spatial scales in transport systems within particular places, notably from a network and city perspective. In the following sections, we insist more on the multi-scalar dimension of gateways as depicted by Rozenblat and Pumain (2007) or by Pain (2007a, 2007b) in a recent synthesis on global cities, gateways and corridors.

14Referring to the seminal work of Burghardt (1971), the gateway functions of cities have been defined by Bird (1983) as “those that link a home region to other regions in the nation state and the nation state to the rest of the world via international transport” (p. 200), thus placing gateways in contrast to central places that serve their adjacent region. Such a strict distinction between gateways and central places have become obsolete due to the rapid development of modern and global transport systems, yet the gateway functions of cities remain a specific category comprising all transport and communication functions, also extending to logistics and related activities (Hesse, 2008). One typical example is Rotterdam, which functions as a global gateway serving Europe’s heartland (Van Klink, 1998). Precise measures of this specialization are often difficult to obtain. Other examples are the many primate cities of the world concentrating most of their national territory’s gateway functions (Dogan, 1988; Keeling, 1995).

15The concept of gateway increasingly extends beyond the spatial boundaries of one single city. The “gateway city” can thus be defined as the command centre of a wider city-region or urban region acting through a set of terminals located in proximity to certain natural sites (e.g. delta, estuary) and/or at the head of a transport corridor. Vance’s (1970) spatial model compares two different coastal urban systems, with the “old” core located inland/upstream and the “new” core located on the coast due to the extraversion of its trade relations. The worldwide comparison of how urban and port functions are interlocked illustrates this duality between the European trend, marked by continental centrality, and the rest of the world, marked by coastal centrality (Ducruet, 2008).

16Gateway cities and gateways in general are also political constructs for promoting cooperation among scattered locations sharing a common area, as seen in the cases of the Canadian Asia-Pacific Gateway centered on Vancouver4 or the Belgian Extended Gateway linking Antwerp to Limburg through multimodal services5. In those cases, the provision of large-scale transport infrastructure permits economies of scale for industries while connecting to inland distribution centers in order to maintain the specialization (Van Klink and Van den Berg, 1997; Notteboom and Rodrigue, 2005). Extensively using the term of “gateway,” Hall and Pain (2006) also refer to European spatial policies that create fashionable polycentric urban regions. A wide set of case studies have thus focused on competition and cooperation issues within broadly defined gateway regions, notably in the case of ports in proximity (Notteboom et al., 2009), port regions and systems of ports (Ducruet, 2009).

17International comparative studies remain relatively scarce in this field. One exception is the work of O’Connor (2010), which ranks global city regions depending on their importance in logistics. He observes that more diversified gateways (i.e. those possessing multiple airports and seaports within a radius of 70 kilometers from the “core”) generate bigger traffic and larger logistics sectors than more specialized gateways (i.e. those handling either air or sea freight). More evidence comes from Jacobs et al. (2010), who empirically identify a specific type of gateway characterized by a high concentration of Advanced Producer Services firms in the maritime sector compared to their size as a city and/or as a port. It is thus hypothesized that such gateway cities (e.g. Houston, Rotterdam, Hamburg) combine international decision and distribution functions while emerging as important industrial concentrations.

18Most of the other works related to gateways (though not often mentioning the concept of gateway itself) address issues tied to a particular city and thus cannot be entirely reviewed here due to their number and diversity. Among recent ones offering a comprehensive view of planning and economic issues in the development of a gateway are the study on New York and the Bost-Wash corridor (Rodrigue, 2004), Busan in South Korea (Frémont and Ducruet, 2005), Hong Kong and Singapore (Lee and Ducruet, 2009), and Hamburg (Grossmann, 2008), among many others.


19The paper proposes a systematic comparison of city systems formed by their respective air and sea transport networks. These two networks were chosen because of their important roles in urban development, although they are quite different. Using recent advancements in network analysis developed by physicists, the issues will be addressed at the global level of the whole network as well as at the local level of each city. The calculation of centrality indexes at the city level is followed by measures of specialization and polarization among cities. This allows us to propose a typology of the roles of cities in the two transport networks underlying the intermediate central positions of what we define as “independent gateway cities,” which occupy bridge positions between different territorial scales.

20Table 1 shows the data provided by Lloyd’s Marine Insurance Unit (LMIU) and Ecole Nationale de l’Aviation Civile (ENAC) for inter-port and inter-airport traffic flows: in 2006 and 2004, respectively. Maritime flows comprise containers only, which represent approximately 12% of the world’s fleet and no less than 49% of the world’s trade value; air flows comprise all worldwide passenger flights.

21A recurrent problem in the international comparison of cities and regions is the lack of spatially harmonized datasets or definitions of cities. One important aspect of data preparation was thus to delineate coherent spatial units for defining cities. The problem of spatial delineation for urban comparison widely differs from general administrative frames, because urban delineations refer to a precise conceptual definition of the city itself (Unwin, 1996; Ratcliffe and McCullagh, 1999). As opposed to administrative units, the aggregation creating comparative objects must be delineated according to the same criteria, independent to the administrative frame. At the international level, as each country has its own administrative frame and its own definition of city (if it does exist), it is essential to build a comparative database of cities defined in a similar fashion. Several concepts can be used to define cities. For the worldwide network, we define three levels of delineation according to three different concepts of cities (Figure 1):

  1. City level: The first level is the city as a restricted functional urban area, which is generally defined according to the daily commuters converging to the city or to its suburbs. For Europe, we used the definition of FUAs defined by ULB-Brussels for the FOCI project (2010) as well as the SMSA in the USA (2009). For other parts of the world, a comparable definition was applied city by city. Here, every airport or port defines a particular city; some large cities have several airports and/or ports, which have been aggregated in one node. In many cases, further verification was necessary, including for Leixoes (Porto), Port Botany (Sydney), Port Sultan Qaboos (Muscat), Haydarpasa (Istanbul), Fremantle (Perth), and Apapa (Lagos).

  2. Urban regional level: The second level is an extended functional urban region. In our case, we considered the influence area of the airports: if a city has a small airport, where do residents take a flight to reach the global air network? In this second level, 12 cities were aggregated in new delineated urban regions (Table 1). Important aggregation criteria include spatial proximity as well as belonging to a common urbanized area (e.g. Tokyo-Yokohama, Los Angeles-Long Beach) or to a natural site such as an estuary mouth or a delta. One example is the urban region of London Thames, which is comprised of five ports (i.e. London, Thamesport, Tilbury, Sheerness, and Felixstowe) and four airports (i.e. Heathrow, Gatwick, Stansted, and London City). It does not include Southampton, which is located less than 200 kilometers from London but constitutes an urban area on its own of no fewer than 300,000 inhabitants, with its own international airport.

  3. Megalopolis level: The third level takes inspiration from the concepts of the megalopolis proposed by Gottman (1967) and the corridor proposed by Whebell (1969), which are both defined by large regions where several urban regions are close. Another influential concept is the port range, which his defined by Vigarié (1964) as a coastal alignment of port cities sharing seaward and landward linkages through spatial proximity and functional interdependency (e.g. the Le Havre–Hamburg range or the North European range). Examples of large gateways include areas extending from Boston to Washington (Northeast American megalopolis), from Tokyo to Osaka (Japanese megalopolis), and from Amsterdam to Brussels (Ruhr-Flanders), all of which are marked by high density corridors. Here, we aggregated significantly the urban regions of level 2, decreasing the number of units from 476 urban regions to 146 megalopolises (Table 1).

Figure 1: Nodes (cities) and links (flows): three levels of aggregation

Figure 1: Nodes (cities) and links (flows): three levels of aggregation

Table 1: Size of the networks at each level of aggregation



Level 1: City level

TOTAL air and sea



COMMON air and sea






Level 2: Urban Region

TOTAL air and sea



COMMON air and sea






Level 3: Megalopolis

TOTAL air and sea



COMMON air and sea






22Considering the networks, for each level, we counted either the addition of the two networks of sea or air transportation (in a non-weighted approach of links) or the interdependence between the two networks (networks based on common nodes). In order to measure the correlation or the similarity between the two networks, which must have the same set of nodes, even with null values, we instead treat them using a weighted approach. For the second case of articulated weighted networks, we keep the linkages inside each aggregated unit (reflexive links).


23Taken as a whole, the networks possess properties among which the hierarchy and scale-free structure are the most important. It is thus useful to compare hierarchies of graphs (section 4.1), before comparing directly graphs themselves (section 4.2) and studying the relative position of each city in the network (section 4.3). These three approaches are only possible using the comparable databases described previously. These macro approaches of the graphs constitute a first step in order to underline common features as well as differences between the two networks.

Hierarchy and scale-free structure of the three levels

24Several explorations of large-scale properties of different kind of networks (e.g. Barabási and Albert, 1999; Barrat et al., 2004; Guimerà et al., 2005) show that the probability that a node in the network is linked with n other nodes decreases following a typical power-law trend. In extent, this means that if we plot the degree distribution of the nodes in our network, we will obtain a function which approximates a power law: this network structure is widely discussed in the literature and referred to as “scale-free networks” (Barabási and Bonabeau, 2003; Jeong et al.2003). As stated before, maritime and airline transportation networks reflect in part the development of urbanization patterns in different areas of the world.

25A close example of another empirically verified scale-free network is the structure of the power grid in the Western USA, presented by Barabási and Albert (1999), which is approximated by a power law distribution. The general idea is quite simple: we are dealing with networks that have a few nodes with a very large number of links (i.e. the main hubs of the network) and a large number of nodes with only a few links. The city size has a similar distribution, which better fits a power law when considering corridors (i.e. aggregations of cities) as proposed by Marchetti (1991). This pattern describing the Zipf law goes further in the scale-free assumption (Newman, 2005). The growth process of the network follows a preferential attachment that characterizes the development of networks: new links preferentially connect nodes already having the largest number of links, since the gain is greater for already privileged places, thereby increasing the hierarchy. In reality, we can find earlier explanations of the Zipf law proposed by Simon (1955), referring to the Yule process (Yule, 1925).

26This peculiar hierarchy characterizes the structure of the airline and maritime networks studied in this paper, when the frequency distribution of the degree approximates a power law function, although it decays exponentially in the tail for the largest degrees. In Figures 2 and 3 represent the frequency distribution of the airline and sea network, respectively, comparing the distribution of the degree for the three levels (city, urban region and megalopolis) described above. In both networks, levels 1 and 2 are quite similar as a consequence of the small number of cities aggregated; both levels approximate the power law distribution with a typical slope.

  • 6 Trend lines and equations are computed with power law fit. Values of the exponent are: 1.1209 (city (...)

Figure 2: Scale-free dimension of air networks for the three levels6

Figure 2: Scale-free dimension of air networks for the three levels6
  • 7 Trend lines and equations are computed with power law fit. Values of the exponent are: 1.0243 (city (...)

Figure 3: Scale free dimension of sea networks for the three levels7

Figure 3: Scale free dimension of sea networks for the three levels7

27This network feature is thus explained by how networks evolve and grow through the continuous addition of new nodes and links. The new nodes are more likely to be connected with the existing nodes, which already have a large number of connections, with reference to the preferential attachment process. In terms of transportation network, the growth of the worldwide flows privileged a few number of ports/airports, which saw their connections growing at a faster pace. This process is shown clearly by the form of the distribution for levels 1 and 2 (Figures 2 and 3).

28The aggregation of cities operating in order to build level 3 of the megalopolis instead reduced the number of nodes with a very small number of links. Here, the distributions plotted in Figures 2 and 3 do not consider nodes with degrees smaller than 10. In fact, when processing the aggregations as described in Section 3, metropolises have been formed by the addition of cities to the largest metropolitan areas, considering spatial proximity as a crucial criterion. The group of nodes with a small degree is largely localized in geographically isolated areas, and their size does not allow them to be classified as metropolises. The function for level 3 has a lower slope, and the right part of the graph is less dense: small degrees are less represented, following an exponential decay trend. In the metropolis level, the scale-free property is not confirmed.

29Another explanation for the different behavior of the distribution on level 3 of the megalopolises might depend on the smaller number of nodes (from 1767 to 167 for airports and from 1076 to 144 for ports), which make the network less close to the scale free properties sketched above (Barabási, 2007). Nevertheless, in order to better outline the features of the level 3 networks, we compare the degree distribution obtained from our geographical aggregation based on proximity of cities with the degree distribution of a randomly defined aggregation from levels 1 and 3. The result is plotted in the charts presented in Figures 4 and 5. It is now evident that the peculiar slope of the distribution for level 3 depends mainly on the aggregation process. The result confirms that our third level aggregation tackles some of the constituting features of the hierarchical structure inherent in the urban network described by air and maritime transportation. For example, a random aggregation would not take into account geographical distance/proximity: when defining the limits of megalopolis, the proximity between smaller cities and larger urban agglomerations has been a determinant in choosing to include them into the same spatial entity. Also in Figures 4 and 5, the nodes with small degree (i.e. those that are geographically isolated as explained above) are not considered because they cannot be considered as megalopolises.

  • 9 Trend lines and equations are computed with power law fit. Values of the exponent are: 0.714 (obser (...)

Figure 5: Observed vs. random aggregations for level 3, sea transportation network9

Figure 5: Observed vs. random aggregations for level 3, sea transportation network9

Correlations between graphs

30Direct comparison between the two networks reveals that cities generally encompass both transportation modalities. The two networks are structured in different ways, because flights develop more on long-range distances, compared to the short distances of other transport modes. Sea transportation develops as well short- and long-distance linkages by circuits. The issue that we want to address is whether cities occupy the same relative position in the two networks and if some important links of one network could influence the strength of the same links in the other network.

31For each of the three levels, we applied three kinds of correlations (Tab.2):

  • Pearson correlation between the weighted degree of the nodes (total traffic) of ports and airports (with or without the internal (“reflexive”) links);

  • Pearson correlation between weight of links of couples of cities (with or without the internal (“reflexive”) links);

  • Multiple Regression Quadratic Assignment Procedure tests between the whole two networks (Hubert and Schultz, 1976; Krackhardt, 1987; Dekker et al., 2007).

Table 2: Correlations between air and sea nodes’ traffics, links flows, and correlations of weighted matrixes



Nodes without reflexive


Links without reflexive

Correlations of Networks (QAP) without reflexive

Level 1: City level












Level 2: Urban Region


















Level 3: Megalopolis


















32The results show a slight similarity between the two networks as revealed by significant correlations and similarities. The higher these indexes are, the more we could consider that the two networks become interdependent in the formation of the system of cities. As a construction effect, correlations and similarities increase with further aggregation. In fact, urban regions and megalopolises exhibit better spatial coherence with aggregated weights by city in both air and sea networks. At the world scale, an aggregated approach at the megalopolis level is useful to catch the mass effect of large dense urban areas. Although such result was expectable, the increase of correlation indices from one level to another confirms the quality of the proposed delineations, notably at the second level (urban region), in which only few cities have been aggregated. Thus, we can conclude that the second level is effective in identifying some cohesive aggregations of cities in the networks.

33Also as expected, correlations of nodes are higher than correlation of links, which are higher than similarity indexes. Of course, node correlation requires less similarity than correlation in links and network similarities. However, significant similarities between networks show in particular that the networks between main cities are similar, whatever the transportation mode considered. In fact, we could have obtained similar results by considering communication or financial networks (this relation remains to be tested). This is a very important result, showing common features of very different networks although they have been created by the same city-based factors of economic development. It confirms our hypothesis that combined transport networks are good reflections of hierarchical tendencies among cities. It is not so basic because with deregulation and new transport nodes created by private companies and governments, one could think that they develop strategies countering the dominant processes. But in fact, these actors follow the main streams of urban development. However, there are also differences in the respective position of cities in both networks, on which we focus in the next section.

Specialization and polarization among cities

34The situation of cities within air and sea networks results in both hierarchy and specialization. One important issue is how to better understand whether cities with a balanced mix of networks perform better than cities specialized in a particular network. A large body of literature exists on the relation between city size and functional diversity in urban geography and economics (Quigley, 1998; Taylor et al. 2002; Florida, 2004; Ottaviano and Peri, 2006; Pumain et al., 2009). Such framework has not been applied to air and sea transport although it is acknowledged that many cities strive for sustaining their position in both international container shipping and airline flights (Jauhiainen, 1995). Nowadays, urban policies may give priority to air transport over sea transport for higher impacts in terms of value-added logistics and raising overall urban attractiveness (Thayer and Whelan, 1989). Other important questions are related to the hierarchical relations among cities: are specialized cities always under the influence of diversified cities, and can we isolate groups of cities of same nature in the network?

35Cities can be compared by calculating their degree centrality, betweenness centrality, and weighted degree. Degree centrality, the number of cities connected, acts as a local measure; betweenness centrality, the number of positions on the shortest possible paths within the entire graph, indicates overall accessibility in the network; the weighted degree, the sum of traffic of all valued edges of a given city, is equivalent to its total traffic. Measured in different metric units (passengers and containers), air and sea traffics have been standardized (z-scores) in order to be added by city and by edge, as suggested by Dumolard (1975, p. 108).

36Based on betweenness centrality, air-sea networks are polarized by few very central cities at each level of aggregation (Figure 6). In these representations, the positions of cities are defined by their relative position in the network thanks to a GEM-Frick calculation of topologic proximity developed in the TULIP software (Auber, 2003). In accordance with our previous results, the distribution of air-sea centrality and traffic becomes less concentrated from the level of cities to the level of megalopolises (see also Table 3). The major hubs are dominantly located in Europe and Asia. We find at level of cities, the airline hubs (e.g. Paris, London, Frankfurt, Madrid, and Amsterdam) and the maritime hubs (e.g. Singapore, Hong Kong, Rotterdam), which correspond to well-known rankings provided by official traffic statistics.

37The benefit of network combination at urban region level becomes clear in the case of port cities such as Shanghai-Yangtze, Miami-Palm Beach, Barcelona-Tarragona, and Athens-Piraeus, and for some non-port cities such as Santiago-Valparaiso, Sao Paulo-Santos, and Brussels-Flanders, which did not rank among the top 20 locations on the level of cities. Some cities, in turn, have seen their position slightly reduced (e.g. Houston, Rotterdam, Istanbul, and Kuala Lumpur), probably due to their strong modal specialization in either network compared with larger and more diversified gateways. London and Paris, despite their limited port function, remain at the top of the hierarchy because of their global air hub function.

38On the level of megalopolises, wider changes in the hierarchy appear. Ruhr-Flanders has the highest centrality, probably due to the inclusion of several main ports (Rotterdam, Antwerp) and airports (Amsterdam, Brussels, Düsseldorf) comprising Europe’s largest freight gateway, as for the Pearl River Delta (i.e. Hong Kong, Shenzhen, and Guangzhou) and the two large U.S. and Japanese megalopolises. London’s drop to ninth place is largely due to its poor port function compared with such multimodal gateways.

Figure 6: Vizualisation of the combined air-sea networks (3 levels)

Figure 6: Vizualisation of the combined air-sea networks (3 levels)

Table 3: Concentration (Gini) of centrality and traffic among cities in the combined air-sea network (3 levels)

Betweenness centrality

Degree centrality

Weighted degree





Urban regions








39In order to better assess the respective role of each network in the ranking of each gateway, we propose to measure successively the specialization of cities in two ways. The first method consists in comparing betweenness centrality scores at single and combined networks. Ratios are calculated for each city, dividing the percentage of centrality in a single network (air or sea) by the percentage of centrality in the combined air-sea network. Cities with a higher share in the combined air-sea network than in a single network (ratio lower than 1) tend to benefit from integration of the two networks, while cities that lose their position after integration (ratio higher than 1) are disadvantaged. A typology can thus be established depending on which network has been beneficial to which city on different aggregation levels (Figure 7).

Figure 7: Typology of cities in the combined air-sea network (Levels 2 and 3)

Figure 7: Typology of cities in the combined air-sea network (Levels 2 and 3)

N.B. the location of each spatial unit is based on the geographical coordinates of the main urban center

40Red cities are those in which air centrality decreases when combined with sea transport (i.e., they specialize in air transportation); blue cities are those in which sea centrality decreases when integrating air transport (i.e., they specialize in sea transportation); purple cities are those which centrality is higher when integrating either air or sea networks; and pale yellow cities are those which centrality is lower when integrating either air or sea networks. Thus, we interpret the gaps as expressions of relative strengths or weaknesses of cities revealed by their positions in multiple networks. Results show that almost all cities gain from inclusion in the combined network at levels of urban regions and of megalopolises.

41The level of urban regions draws a relatively well-known geography of main sea and air corridors, with the Caribbean basin, South and East Asia, and Europe (Northern range, Iberian Peninsula) depending on the sea network to build their combined centrality. Also in this trend are Africa’s Guinea Gulf and New Zealand. Among the very central megalopolises, only Ruhr-Flanders and Portugal-Northwest Spain gain from both networks simultaneously, the first being the most central gateway in the network at this level. This is also the case for less central mega city-regions such as Melbourne-Adelaide-Tasmania, Northwest Adriatic (Venice), Southwest Australia (Perth), Guyanas, and Reunion-Mauritius.

42There are noticeable differences compared with the urban region level. For instance, a majority of African regions (except Tunis) and Latin American regions (except Sao-Paulo and Santiago) now appear more maritime-oriented. The reliance upon the air network clearly highlights the dominance of the Triade (Japanese, U.S. megalopolises) and large urban concentrations (California, Grand Paris, London Southeast) at the expense of a large portion of other gateways (e.g. southern hemisphere), which now appear to be relatively more dependent on sea transport. Megalopolises for which the combination of the two networks resulted in a loss of centrality are often located inland, as a consequence of remoteness and transport costs (e.g. some inner regions of South America, China, U.S., and Europe). Other examples of this type are megalopolises that are well connected to both networks but with a relative weakness, probably due to the proximity to established competitors (e.g. South Korea, Mediterranean France, and Southeast U.S.). The relative position of gateway cities with regard to their specialization can be better assessed by looking at nodal regions.

43The second approach to the specialization of cities is based on the respective share of air and sea traffic within the weighted degree (total traffic). We compare traffic specialization and regional polarization by means of the Nystuen and Dacey (1961) algorithm. This simplifies the network into a tree graph linking each city to its dominant city (dominant flow). This method has been extensively applied to many intercity flows such as airline networks (Cattan, 2004; Grubesic et al., 2008); to our knowledge, however, it is the first time that it is applied to a combined network. This method allows us to define dependent nodal regions in which cities are either independent or subordinated. Independent gateway cities are those in which the largest traffic flow connects another city of smaller size. Subordinate cities have their largest flow link directed toward a bigger city. Nodal regions are composed of one independent gateway city and its subordinates, but there are cases when more than one independent city is included.

44Figure 8 presents the nodal regions obtained from the combined air-sea network. We retain in the figures only the biggest nodal urban regions and megalopolises in which the majority of top cities are included. A number of nodal regions appear, generally characterized by a relatively high geographic coherence. The biggest nodal region centered upon London includes many gateway cities worldwide, but the other nodal urban regions or megalopolises are more circumscribed based on spatial proximity among cities connected by largest flows.

45The large variety of situations found in the position of nodes may be summarized according to three main and recurrent trends:

  • The largest poles in terms of their number of subordinates are often specialized in air traffic. As a consequence, cities specializing in sea traffic tend to be their subordinates. This evidence is not fully in accordance with our initial assumption that diversified cities should dominate specialized cities. London as air hub dominates the largest nodal region of the world at urban regions level as well as megalopolises level, which includes Paris, New York, Madrid, Houston, Miami, the Baleares, and the Canary Islands. Istanbul also exerts its influence upon maritime cities of the Black Sea. Cities such as Nantes, Puerto Rico, Dubai, and Marseilles, among others, are more diversified and act as independent cities within their sub-network. The dominance of air cities is even more visible on megalopolises level, where most independent gateways have their largest traffic share in air transport (London Southeast, U.S. Southeast, Liguria, California-Mexico, and Sydney NSW).

  • Gateway cities of a certain type tend to polarize cities of the same type due to their common specialization. This result is certainly influenced by the aforementioned complementarity of the two networks resulting in distinct geographic coverage. On urban regions’ level, Rotterdam and Hamburg have mostly maritime cities among their subordinates; the same goes for Savannah, Busan, Vladivostok, Salalah, Surabaya, Kaohsiung, and Tauranga. On megalopolises level, we find this trend in the cases of Singapore-Malacca and Shanghai-Yangtze, although their tributary areas are more diversified and include all types of cities.

  • There is a regional variation in the results. The Asian nodal regions, for instance, show an opposite profile to the European ones, with maritime cities dominating air cities such as Hong Kong, Shanghai, Singapore, Kuala Lumpur, and Qingdao. This is also the case for diversified cities such as Tokyo, Osaka, and Taipei as well as for the main gateway cities of a distinct Latin American nodal region polarized by Santiago, Rio de Janeiro, Salvador, Manaus, Recife, and Sao Paulo, where traffics are more balanced. On level 3, the Japanese megalopolis and the Brazil Sudeste regions are both diversified and independent. This confirms that large urban concentrations of the developing world are more diversified by nature than in the developed world.


46Air and maritime networks have a crucial role in globalization processes: global cities benefit from being at the center of global flows and, although each transportation mode has its own rationale, their growth is determined by urban demography and economy. Air and maritime flows have been examined in detail in the literature, but few studies investigated these two networks simultaneously. Despite their differences, the two modes of transportation have in common their global and long distance range and share similarities in how their respective industry is organized. They have also strong complementarities, and as a matter of fact many global cities have developed large air and sea transportation terminals. This paper thus implements a joint analysis of the two networks in order to underline their complementarities in shaping the global urban hierarchy. Our analysis started from a geographical reorganization of the data available on maritime and air networks. We defined three levels of delineation corresponding to three concepts of cities, in order to obtain comparable urban entities (city, urban region, and megalopolis) and to verify their hierarchies and specializations through linkages.

47We first analyzed the overall structures of the three levels for each of the two networks in order to test their behavior vs. the scale free hypothesis and to validate the aggregation at the megalopolis level. The different analyses confirmed that the structure of our aggregation grasps some of the features of the hierarchical structure along with important geographical properties of the networks, which are interdependent with urban development. Starting with hypotheses on the growth and general properties of the networks, the slope of the degree distributions computed for the city and urban region levels confirmed the existence of a preferential attachment process, corresponding to urban hierarchical processes.

48We then analyzed the correlations between air and see networks’ topology at the three levels:  what is distinctive here is that similarities are higher for larger urban units (i.e. urban regions, and megalopolises). This confirmed the necessity to redefine urban entities in order to show that despite deregulation, global players in the transport industry tend to follow the main paths of urban development. Such result would not be visible on the level of single and separated terminals of which many, in fact, belong to the same functional urban area.

49As to specialization issues, we then examined whether cities could be favored by diversification or specialization in their position in the global urban hierarchy. Measures of network properties (betweenness centrality) clarified that at the urban region level the combination of air and maritime networks has a positive impact on the position of a given city in the urban hierarchy. The comparison between centrality rankings allowed a first typology, based on how cities are advantaged or not by the integration of maritime and air networks: almost all cities at the urban region and megalopolis level benefit from a combined network, except for those localized inland or nearby a strong competitor. A dominant flow methodology, implemented on the combined transportation network, allows a second typology of independent gateway cities and subordinate cities according to traffic flows within and between so-called “nodal regions”. Among the nodal regions highlighted, it has been possible to point out some common features. First, the largest nodal regions are those specializing in air traffic. This disputes an initial hypothesis that dominant cities would have a diversified profile, while it highlights that air transport is less geographically constrained than sea transport by attaining distant locations without detours. Second, independent cities tend to be connected to cities with similar specializations as an effect of the distinctiveness between air and sea networks’ geographic coverage. Finally, the results are very sensitive to the localization of nodal regions: cities of the developing world are more diversified by nature, because they act as both air and sea gateways and hubs.

50We expect future results to follow three complementary paths. First, the analysis of the two networks can be refined by going deeper in the study of interdependencies between networks, such as linked design methodologies (Lazega et al., 2008), network dynamics by ERGM methods (Robins et al., 2007, 2009), simulation of the interaction between multiple networks (Parshani et al. 2010), and the analysis of percolation, catastrophy, and cascading failures (Buldyrev et al., 2010; Vespignani, 2010). Secondly, from a more geo-economic point of view, the analysis will surely benefit from the integration of exogenous data on the nodal properties of cities, such as their economic structure and demographic features. Greater efforts shall be put on refining the delineation of cities due to the absence of definitions that could be valid on a world level. Thirdly, the analysis of multiple networks from an urban perspective would benefit from the inclusion of other spatial networks such as multinational firms networks, land transport networks (e.g. roads), and telecommunications networks, among others, to better assess their respective roles in shaping global urban hierarchies and specializations.

Haut de page


Amiel M., Mélançon G., Rozenblat C., 2005, "Réseaux multi-niveaux : l’exemple des échanges aériens mondiaux de passagers", Mappemonde, vol. 79,

Auber D., 2003, "Tulip: A huge graph visualisation framework", in Mutzel P., Jünger M. (eds.), Graph Drawing Softwares, Mathematics and Visualization, Berlin, Heidelberg, Springer-Verlag.

Bairoch P., 1985, De Jéricho à Mexico. Villes et économie dans l’histoire, Paris, Gallimard.

Barabási A.-L., Albert R., 1999, "Emergence of Scaling in Random Networks", Science, vol. 286, 509-512.

Barabási A.-L., Bonabeau E., 2003, "Scale-free Networks", Scientific American, vol. 288, No.5, 50-59.

Barabási A.-L., 2007, "The Architecture of Complexity", IEEE Control Systems Magazine, August 2007, 33-42.

Barrat A., Barthélemy M., Vespignani A., 2004, "Modeling the Evolution of Weighted Networks", Physical Review E, vol. 70, No.066149.

Bird J., 1983, "Gateways: Slow Recognition but Irresistible Rise", Tijdschrift voor Economische en Sociale Geografie, vol.74, No.3, 193-202.

Blumenfeld-Lieberthal E., 2009, "The Topology of Transportation Networks: A Comparison between Different Economies", Networks and Spatial Economics, vol. 9, No.3, 427-458.

Bogart D., 2009, "Inter-modal Network Externalities and Transport Development: Evidence from Roads, Canals, and Ports during the English Industrial Revolution", Networks and Spatial Economics, vol. 9, No.3, 309-338.

Bretagnolle A., 1999, Les systèmes de villes dans l’espace-temps : effets de l’accroissement de la vitesse des déplacements sur la taille et l’espacement des villes, PhD Dissertation in Geography, Paris I Sorbonne University.

Brocard M. (dir.), 2009, Transports et territoires, enjeux et débats, Paris, Ellipses.

Buldyrev S.V., Parshani R., Paul G., Stanley H.E., Havlin S., 2010, "Catastrophic Cascade of Failures in Interdependent Networks", Nature, vol. 464, 1025-1028.

Burghardt A.F., 1971, "A Hypothesis about Gateway Cities", Annals of the Association of American Geographers, vol. 61, No.2, 269-285.

Cattan N., 1995, "Barrier Effects: The Case of Air and Rail Flows", International Political Science Review, vol. 16, No.3, 237-248.

Cattan N., 2004, "Le monde au prisme des réseaux aériens", Flux, vol.4, No.58, 32-43.

Cullinane K.P.B., Khana M., 2000, "Economies of Scale in Large Containerships: Optimal Size and Geographical Implications", Journal of Transport Geography, vol.8, No.3, 181-195.

Dekker D., Krackhardt D. Snijder T.A.B, 2007, "Sensitivity of MRQAP Tests to Collinearity and Autocorrelation Conditions", Psychometrika, vol. 72, No.4, 563-581.

Derruder B., Devriendt L., Witlox F., 2007, "Flying Where You Don't Want to Go: An Empirical Analysis of Hubs in the Global Airline Network", Tijdschrift voor Economische en Sociale Geografie, vol. 98, No.3, 307-324.

Dogan M., 1988, "Giant cities as maritime gateways", in Dogan, M., Kasarda, J.D. (eds.), The Metropolis Era, London, SAGE Publications.

Ducruet C., 2008, "Typologie mondiale des relations ville-port", Cybergeo :, No.417, 27 mars 2008, 17 p.

Ducruet C., 2009, "Port regions and globalization", in Notteboom, T.E., Ducruet, C., Langen de, P.W. (eds.), Ports in Proximity: Competition and Coordination among Adjacent Seaports, Aldershot, Ashgate.

Ducruet C., Lee S.W., 2007, "Measuring Intermodalism at European Port Cities: An Employment-based Study", World Review of Intermodal Transportation Research, vol. 1, No.3, 313-334.

Ducruet C., Notteboom T.E., 2011, "The Worldwide Maritime Network of Container Shipping: Spatial Structure and Regional Dynamics", Global Networks (forthcoming).

Ducruet C., Rozenblat C., Zaidi F., 2010, "Ports in Multi-level Maritime Networks: Evidence from the Atlantic (1996-2006)", Journal of Transport Geography, vol.18, No.4, 508-518.

Ducruet C., Van der Horst M., 2009, "Transport Integration at European Ports: Measuring the Role and Position of Intermediaries", European Journal of Transport and Infrastructure Research, vol.9, No.2, 121-142.

Dumolard P., 1975, "Région et régionalisation, une approche systémique", L’Espace Géographique, vol. 2, 93-111.

Fleming D.K., Hayuth Y., 1994, "Spatial Characteristics of Transportation Hubs: Centrality and Intermediacy", Journal of Transport Geography, vol. 2, No.1, 3-18.

Florida R.L., 2004, The rise of the creative class, New York, Basic Books.

Frémont A., 2009, "Intégration, non-intégration des transports maritimes, des activités portuaires et logistiques : quelques évidences empiriques", OECD Working Paper.

Frémont A., Ducruet C., 2005, "The Emergence of a Mega-port: The Case of Busan, from the Local to the Global", Tijdschrift voor Economische en Sociale Geografie, vol. 96, No.4, 421-432.

Gastner M.T., Newman M.E.J., 2003, "The Spatial Structure of Networks", The European Physical Journal B, vol. 49, 247-252.

Goetz A.R., Rodrigue J.P., 1999, "Transport Terminals: New Perspectives", Journal of Transport Geography, vol. 7, No.4, 237-240.

Gottmann J., 1961, Megalopolis, the urbanized northeastern seaboard of the United States, New York, The Twentieth Century Fund.

Grossmann I., 2008, "Perspectives for Hamburg as a Port City in the Context of a Changing Global Environment", Geoforum, vol. 39, No.6, 2062-2072.

Grubesic T.H., Matisziw T.C., Zook M.A., 2008, "Global Airline Networks and Nodal Regions", Geojournal, vol. 71, No.1, 53-66.

Guimerà R., Mossa S., Turtschi A., Amaral L.A.N., 2005, "The Worldwide Air Transportation Network: Anomalous Centrality, Community Structure, and Cities’ Global Roles", Proceedings of the National Academy of Sciences, vol. 102, No.22, 7794-7799.

Hall P., Pain K., 2006, The polycentric metropolis: learning from mega-city regions in Europe, London, Earthscan.

Hesse M., 2004, "Land for Logistics: Locational Dynamics, Real Estate Markets and Political Regulation of Regional Distribution Complexes", Tijdschrift voor Economische en Sociale Geografie, vol. 95, No.2, 162-173.

Hesse M., 2008, The city as a terminal, Aldershot, Ashgate.

Hesse M., Rodrigue J.P., 2004, "The Transport Geography of Logistics and Freight Distribution", Journal of Transport Geography, vol. 12, No.3, 171-184.

Horner M.W., O’Kelly M.E., 2001, "Embedding Economies of Scale Concepts for Hub Network Design", Journal of Transport Geography, vol. 9, No.4, 255-265.

Hubert L.J., Schultz J.V., 1976, "Quadratic Assignment as a General Data Analysis Strategy", British Journal of Mathematical and Statistical, No.29, 190-241.

Jacobs W., Ducruet C., De Langen P.W., 2010, "Integrating World Cities into Global Production Networks: The Case of Port Cities", Global Networks, vol. 10, No.1, 92-113.

Janelle D.G., Beuthe M., 1997, "Globalization and Research Issues in Transportation", Journal of Transport Geography, vol. 5, No.3, 199-206.

Jauhiainen J.S., 1995, "Waterfront Redevelopment and Urban Policy: The Case of Barcelona, Cardiff and Genoa", European Planning Studies, vol. 3, No.1, 3-23.

Jeong H., Néda Z., Barabási A.-L., 2003, "Measuring Preferential Attachment in Evolving Networks", Europhysics Letters, vol. 61, No.4, 567-572.

Jin F., Wang C., Li X., Wang J., 2010, "China’s Regional Transport Dominance: Density, Proximity, and Accessibility", Journal of Geographical Sciences, vol. 20, No.2, 295-309.

Keeling D.J., 1995, "Transport and the world city paradigm", in Knox P.L., Taylor, P.J. (eds.), World Cities in a World-System, Cambridge, Cambridge University Press.

Keeling D.J., 2007, "Transportation Geography: New Directions on Well-worn Trails", Progress in Human Geography, vol. 31, No.2, 217-225. elly, 200100;  and per of goods and plism at European port cities: an employment-based approach, World Revieansport

Krackhardt D., 1987, "QAP Partialling as a Test of Spuriousness", Social Networks, vol. 9, 171-186.

Lazega E., Jourda M., Mounier L., Stofer R., 2008, "Catching Up with Big Fish in the Big Pond? Multi-level Network Analysis through Linked Design", Social Networks, vol. 30, 157-176.

Lee S.W., Ducruet C., 2009, "Spatial Glocalization in Asia-Pacific Hub Port Cities: A Comparison of Hong Kong and Singapore", Urban Geography, vol. 30, No.2, 162-184.

Lee S.W., Song D.W., Ducruet C., 2008, "A Tale of Asia’s World Ports: The Spatial Evolution in Global Hub Port Cities", Geoforum, vol. 39, No.1, 372-385.

Marchetti C., 1991, "Voyager dans le temps. Considérations pour une meilleure exploitation de la liaison fixe", Futuribles, No.156, 19-29.

McCalla R.J., 1999, "Global Change, Local Pain: Intermodal Seaport Terminals and their Service Areas", Journal of Transport Geography, vol. 7, No.4, 247-254.

Newman M.E.J., 2005, "Power Laws, Pareto Distributions and Zipf’s Law", Contemporary Physics, vol. 46, No.5, 323-351.

Notteboom T.E., Rodrigue J.P., 2005, "Port Regionalization: Towards a New Phase in Port Development", Maritime Policy and Management, vol. 32, No.3, 297-313.

Notteboom T.E., Ducruet C., De Langen P.W., 2009, Ports in proximity: Competition and coordination among adjacent seaports, Aldershot, Ashgate.

Nystuen J.D., Dacey M.F., 1961, "A Graph Theory Interpretation of Nodal Regions", Papers in Regional Science, vol. 7, No.1, 29-42.

O’Connor K., 2010, "Global City-regions and the Location of Logistics Activity", Journal of Transport Geography, vol. 18, No.3, 354-362.

O’Kelly M.E., 1998, "A Geographer's Analysis of Hub-and-spoke Networks", Journal of Transport Geography, vol. 6, No.3, 171-186.

Ottaviano G.I.P., Peri G., 2006, "The Economic Value of Cultural Diversity: Evidence from US Cities", Journal of Economic Geography, vol. 6, 9-44.

Pain K., 2007a, "Gateways and corridors in globalization: planning sustainable infrastructures for transcontinental ‘spaces of flows’", Paper presented at the Roundtables and International Conference of Canada’s Asia-Pacific Gateway and Corridor Initiative, Vancouver, May 2007.

Pain K., 2007b, "Global cities, gateways and corridors: hierarchies, roles and functions", Paper presented at the Roundtables and International Conference of Canada’s Asia-Pacific Gateway and Corridor Initiative, Vancouver, May 2007.

Parshani R., Rozenblat C., Ietri D., Ducruet C., Havlin S., 2010, "Inter-similarity between Coupled Networks", Europhysics Letters, vol. 92, No.68002, doi: 10.1209/0295-5075/92/68002

Pumain D., Paulus F., Vacchiani-Marcuzzo C., 2009, "Innovation cycles and urban dynamics", in Lane D., van der Leeuw S., Pumain D., West G. (eds.), Complexity Perspectives in Innovation and Social Change, Berlin, Springer-Verlag.

Quigley J.M., 1998, "Urban Diversity and Economic Growth", Journal of Economic Perspectives, vol. 12, 127-138.

Ratcliffe J.H., McCullagh M.J., 1999, "Hotbeds of Crime and the Search for Spatial Accuracy", Geographical Systems, vol. 1, No.4, 385-398.

Rimmer P.J., 1999, "The Asia-Pacific Rim's Transport and Telecommunications Systems: Spatial Structure and Corporate Control since the Mid-1980s", Geojournal, vol. 48, No.1, 43-65.

Robins G.L, Pattison P., Wang P., 2009, "Closure, Connectivity and Degrees: New Specifications for Exponential Random Graph (p*) Models for Directed Social Networks", Social Networks, vol. 31, 105-117.

Robins G.L., Snijders T.A.B., Wang P., Handcock M., Pattison P., 2007, "Recent Developments in Exponential Random Graph (p*) Models for Social Networks", Social Networks, vol. 29, 192-215.

Rodrigue J.P., 1996, "Transportation corridors in Pacific Asian urban regions", in Hensher D.A., King J. (eds.), Proceedings of the 7th World Conference on Transport Research, Sydney, Pergamon.

Rodrigue J.P., 2004, "Freight, Gateways and Mega-urban Regions: The Logistical Integration of the Bostwash Corridor", Tijdschrift voor Economische en Sociale Geografie, vol. 95, No.2, 147-161.

Rodrigue J.P., Comtois C., Slack B., 2009, The geography of transport systems, New York, Routledge.

Rozenblat C., Pumain D., 1993, "The Location of Multinational Firms in the European Urban System", Urban Studies, vol. 10, 1691-1709.

Rozenblat C., Pumain D., 2007, "Firm linkages, innovation and the evolution of urban systems", in Taylor P.J., Derudder B., Saey P., Witlox F. (eds.), Cities in globalization: practices, policies and theories, Oxon, Routdledge.

Sassen S., 1991, The global city: New York, London, Tokyo, Princeton, Princeton University Press.

Sassen S., 2000, Cities in a world economy, Thousand Oaks CA, Pine Forge Press.

Simon H., 1955, “A Behavioral Model of Rational Choice”, The Quarterly Journal of Economics, vol. 69, No.1, 99-118.

Slack B., 1999, "Satellite Terminals: A Local Solution to Hub Congestion?" Journal of Transport Geography, vol. 7, No. 4, 241-246.

Taillard C. (dir.), 2004, Intégrations régionales en Asie Orientale, Paris, Les Indes Savantes.

Taylor P.J., Walker D.R.F., Catalano G., Hoyler M., 2002, "Diversity and Power in the World City Network", Cities, vol. 19, 231-241.

Thayer R.E., Whelan R.K., 1989, "Port cities face complex challenges", in Knight R.V., Gappert G. (eds.), Cities in a global society, London, SAGE Publications.

UNCTAD, 2005, World investment report 2005, New York and Geneva, United Nations.

UNCTAD, 2006, World investment report 2006, New York and Geneva, United Nations.

Unwin D.J., 1996, "GIS, Spatial Analysis and Spatial Statistics", Progress in Human Geography, vol. 20, No.4, 540-441.

Van der Horst M.R., De Langen P.W., 2008, "Coordination in Hinterland Transport Chains: A Major Challenge for the Seaport Community", Maritime Economics and Logistics, vol. 10, No.1-2, 108-129.

Van Klink H.A., 1998, "The Port Network as a New Stage in Port Development: The Case of Rotterdam", Environment and Planning A, vol. 30, No.1, 143-160.

Van Klink H.A., Van den Berg G.C., 1998, "Gateways and Intermodalism", Journal of Transport Geography, vol. 6, No.1, 1-9.

Vance J.E., 1970, The merchant's world: The geography of wholesaling, Englewood Cliffs NJ, Prentice-Hall.

Veltz P., 1996, Mondialisation, villes et territoires : une économie d’archipel, Paris, PUF.

Vespignani A., 2010, "Complex Networks: The Fragility of Interdependency", Nature, vol. 464, 984-985.

Vigarié A., 1964, Les grands ports de commerce de la Seine au Rhin, Paris, SABRI.  

Whebell C.F.J., 1969, "Corridors: A Theory of Urban Systems", Annals of the Association of American Geographers, vol. 59, No.1, 1-26.

Xu X., Hu J., Liu F., 2007, "Empirical Analysis of the Ship-transport Network of China", Chaos, vol. 17, No.2, 023129.

Yule G.U., 1925, "A Mathematical Theory of Evolution, Based on the Conclusions of Dr. J.C. Willis, F.R.S.", Philosophical Transactions of the Royal Society of London B, vol. 213, 21-87.

Zhang P.C., Peeta S., Friesz T., 2005, "Dynamic Game Theoretic Model of Multi-layer Infrastructure Networks", Networks and Spatial Economics, vol. 5, 147-178.

Haut de page







6 Trend lines and equations are computed with power law fit. Values of the exponent are: 1.1209 (city), 1.0994 (urban region), and 0.5618 (megalopolis). Values of R2 are: 0.877 (city), 0.8597 (urban region), and 0.2885 (megalopolis). For the megalopolis level, only the nodes with degree greater or equal to 10 are represented: this consists in removing 36 nodes out of 167.

7 Trend lines and equations are computed with power law fit. Values of the exponent are: 1.0243 (city), 0.9563 (urban region), and 0.714 (megalopolis). Values of R2 are: 0.7413 (city), 0.7458 (urban region), and 0.3833 (megalopolis). For the megalopolis level, only the nodes with degree greater or equal to 10 are represented: this consists in removing 26 nodes out of 144.

8 Trend lines and equations are computed with power law fit. Values of the exponent are: 0.5618 (observed) and 1.0652 (random). Values of R2 are: 0.2885 (observed) and 0.7887 (random). Only the nodes with degree greater than 10 are represented. See note 7 for details on the computation of the exponent of the power law fit.

9 Trend lines and equations are computed with power law fit. Values of the exponent are: 0.714 (observed) and 1.1925 (random). Values of R2 are: 0.3833 (observed) and 0.738 (random). Only the nodes with degree greater than 10 are represented. See note 7 for details on the computation of the exponent of the power law fit.

Haut de page

Table des illustrations

Titre Figure 1: Nodes (cities) and links (flows): three levels of aggregation
Fichier image/jpeg, 188k
Titre Figure 2: Scale-free dimension of air networks for the three levels6
Fichier image/jpeg, 192k
Titre Figure 3: Scale free dimension of sea networks for the three levels7
Fichier image/jpeg, 200k
Fichier image/png, 51k
Titre Figure 5: Observed vs. random aggregations for level 3, sea transportation network9
Fichier image/png, 48k
Titre Figure 6: Vizualisation of the combined air-sea networks (3 levels)
Fichier image/jpeg, 588k
Titre Figure 7: Typology of cities in the combined air-sea network (Levels 2 and 3)
Légende N.B. the location of each spatial unit is based on the geographical coordinates of the main urban center
Fichier image/jpeg, 344k
Fichier image/jpeg, 316k
Haut de page

Pour citer cet article

Référence électronique

César Ducruet, Daniele Ietri et Céline Rozenblat, « Cities in Worldwide Air and Sea Flows: A multiple networks analysis », Cybergeo : European Journal of Geography [En ligne], Systèmes, Modélisation, Géostatistiques, document 528, mis en ligne le 31 mars 2011, consulté le 20 janvier 2017. URL : ; DOI : 10.4000/cybergeo.23603

Haut de page


César Ducruet

Chargé de Recherches, Centre National de la Recherche Scientifique (CNRS)
University of Paris-I Panthéon Sorbonne
UMR 8504 Géographie-Cités / P.A.R.I.S.
13 rue du Four
F-75006 Paris
Tel. +33 (0)140-464-007
Fax +33 (0)140-464-009

Articles du même auteur

Daniele Ietri

Research Assistant, University of Lausanne
Faculty of Geosciences and Environment
Institute of Geography (IGUL)
Quartier UNIL-Dorigny, Bâtiment Anthropole
CH-1015 Lausanne

Céline Rozenblat

Professor, University of Lausanne
Faculty of Geosciences and Environment
Institute of Geography (IGUL)
Quartier UNIL-Dorigny, Bâtiment Anthropole
CH-1015 Lausanne

Articles du même auteur

Haut de page

Droits d’auteur

© CNRS-UMR Géographie-cités 8504

Haut de page