Centralities and Hierarchies of Networks and Territories E-Book

Julie Fen-Chong

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: ISTE Invoiced
• Sprache: Englisch

Seitenzahl: 456

Veröffentlichungsjahr: 2025

Beschreibung

Digital data, graph theory and spatial modeling allow us to apply the concepts of centralities and hierarchies to a wide variety of spatial situations.

The authors of this book offer insight into centralities and hierarchies within networks and territories at different scales and temporalities and for different socio-spatial phenomena. The first part of the book explores the contribution of data from cell phones and social networks to understanding the centralities and hierarchies of urban space within a circumscribed temporality. The second part uses network analysis – ecological networks, media networks and scientific knowledge networks – to propose indicators of spatial organization that reveal the centralities and hierarchies of spatial systems. Finally, in the third part, the territorial processes behind the formation of centralities and hierarchies are presented from a long-term perspective.

Leseprobe

SCIENCES

Geography and Demography, Field Director – Denise Pumain

Fractal Logics in Geography – Hierarchies and Centralities, Subject Head – Cécile Tannier

Centralities and Hierarchies of Networks and Territories

Coordinated by

Julie Fen-Chong

Cécile Tannier

First published 2025 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address:

ISTE Ltd27-37 St George’s RoadLondon SW19 4EUUK

www.iste.co.uk

John Wiley & Sons, Inc.111 River StreetHoboken, NJ 07030USA

www.wiley.com

© ISTE Ltd 2025The rights of Julie Fen-Chong and Cécile Tannier to be identified as the authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s), contributor(s) or editor(s) and do not necessarily reflect the views of ISTE Group.

Library of Congress Control Number: 2025935316

British Library Cataloguing-in-Publication DataA CIP record for this book is available from the British LibraryISBN 978-1-78945-218-1

ERC code:SH7 Human Mobility, Environment, and Space SH7_1 Human, economic and social geography SH7_9 Energy, transportation and mobility

Introduction

Julie FEN-CHONG

ThéMA, CNRS, University of Burgundy, Dijon, France

Geographical space is shaped by movements and exchanges between people and activities located across the Earth’s surface. Its heterogeneity has traditionally been analyzed in geography through the lenses of centrality and hierarchy. Over the past three decades, the representations we have of this space have been profoundly transformed. New communication technologies, notably GPS and smartphones using geolocation, have deeply altered our mobility practices, while new social networks have disrupted the flow of information among individuals. The new data generated by these technologies allow for the observation of movements at temporal and spatial scales considerably more precise than was previously possible. This book seeks to explore two broad questions tied to this new context: what knowledge do the data derived from new communication and information technologies bring to our understanding of spatial networks and their associated territories? How, and to what extent, do these new data and technologies reconstruct the previously existing centralities and hierarchies within geographical space?

The concepts of centrality and hierarchy are fundamental to the field of geography. Whether at the scale of a vast territory, a city or a smaller space, differences in potential between locations generate flows of exchanges and movements. Due to their comparative advantages, certain locations attract more than others and thus polarize the space. For geographers, these locations constitute centralities, both morphological and functional, whose varying importance leads to a hierarchical structuring of space.

Central places, the focus of Christaller’s eponymous theory in 1933, are areas where human activities are concentrated. They provide goods and services to their resident populations and the surrounding regions. The level of diversity and rarity of goods and services offered by these places depends on their population size and serves as an indicator of their position within the urban hierarchy. Christaller’s theory explains the spatial distribution of central places based on differences in centrality. Central places of the same level are equidistant from each other. Their zones of attraction are nested according to their position within the hierarchy and polarize populations from varying distances. Central places higher in the urban hierarchy have larger zones of polarization than those that offer fewer goods and services and occupy lower ranks in the hierarchy. The concepts of centrality and hierarchy are thus interlinked, and quantifying centrality helps measure a place’s rank within the hierarchy.

Christaller’s central place theory has been applied in a number of publications. For instance, in the study by Fleury et al. (2012) on the organization of commerce in Paris, functional centrality is expressed relative to the local environment, based on the rarity and diversity of available businesses. The morphological centrality of Parisian businesses, in turn, is defined through the density and diversity of businesses, the intensity of specialization and local differentials. Other researchers view functional centrality as more tied to coordination or relational functions, considering centrality to involve only a small number of places or neighborhoods (Gaschet and Lacour 2002). In urban economics, centrality is linked to job concentration, activities and maximal accessibility. In the monocentric model developed by Alonso (1964), Mills (1967) and Muth (1969), the urban center is the area with the highest land rent due to the advantages of its location. However, the emergence of multiple employment hubs within a single metropolitan area challenges this monocentric model of a city. The monocentric city either no longer exists or has evolved into multiple centralities of varying sizes and specializations (McMillen 2001). As early as the 20th century, observed land values suggested the presence of peripheral centralities, indicating that the monocentric model might be merely a theoretical construct. According to Berroir et al. (2008, our translation), widespread mobility has driven the rise of peripheral centralities: “The links between urban structures and mobility practices contribute to the formation of increasingly polycentric metropolitan territorialities, both morphologically and functionally”. Berroir et al. (2007), analyzing employment hubs and commuter flows, identified 67 hubs structuring the Île-de-France region. The role of employment and commuting in spatial organization is fundamental, and the decentralization of jobs in Île-de-France strengthens the hypothesis of metropolitan polycentrism.

In this book, the concepts of centrality and hierarchy are revisited in light of data from new communication and information technologies. Defining and measuring centrality represents one of the methodological challenges the authors attempt to address. Geolocated data on populations present in a given territory throughout the day enable the construction of centrality indicators suited to the dynamic representation of the urban space. Urban space thus comprises secondary centers that come to life following urban rhythms, as illustrated by Julie Fen-Chong and Françoise Lucchini in Chapters 1 and 2 of this book. To operationally define urban centrality, Mohamed Hilal and Virginie Piguet (Chapter 6) rely on three criteria: population levels, employment levels and the presence of businesses and services.

The concept of centrality, employed in geography to describe and compare locations, is also found in interdisciplinary research drawing on graph theory; here, central places correspond to specific nodes within a graph, defined by their accessibility and connectivity. Physical networks, such as transportation and ecological networks, as well as intangible networks, such as knowledge networks, feature central nodes. Graph theory metrics allow for the measurement of centralities in networks as diverse as global media representations (Chapter 3) or ecological networks (Chapter 5).

In Chapter 7 of this book, Cécile Tannier demonstrates that centralities and hierarchies represent two facets of the same sociospatial phenomenon, applicable at both intra-urban levels and across city or settlement systems. The concentration of individuals and their activities in central places leads to a hierarchical structuring of these locations in relation to each other. Urban hierarchy “refers to a form of societal organization within a given territory, initially identified by engineers and modeled by geographers” (Pumain 2004, our translation). As Juste Raimbault explains (Chapter 8), hierarchy takes on several forms: order hierarchy, inclusion hierarchy, control hierarchy and level hierarchy. These different types of hierarchy are intrinsic to territorial systems and can be observed through scaling laws. Studying hierarchical patterns using various indicators also reveals correlations between different types of hierarchies, such as between population hierarchies and transportation network hierarchies (Chapter 8) or between urban hierarchies and scientific production hierarchies (Chapter 4).

The book consists of three main parts.

In the first part, titled “Spatial Practices as Indicators of Centralities and Hierarchies”, centralities and hierarchies are explored at the territorial level through the study of individual spatial practices. This part presents empirical studies that utilize diverse data sources such as mobile phone data and data from social networks. These data enable the dynamic analysis of spatial practices, providing insights into the changes occurring within centralities and hierarchies in intra- and interurban spaces over short time scales. In the first chapter, Julie Fen-Chong discusses the specificities of mobile phone and Twitter data, which facilitate a fine-grained spatiotemporal understanding of spatial organization, renewing static approaches and enriching previously inadequate datasets for some studies. The author demonstrates how studies utilizing these data have transformed approaches to intra-urban centralities and the hierarchy of places. Centralities are defined as locations that polarize individuals, while hierarchies are conceptualized as the structuring of these centralities based on differences in potential and attractiveness. The second chapter, authored by Françoise Lucchini, emphasizes the value of these materials in measuring ephemeral phenomena. Determining the appropriate spatiotemporal aggregation level to measure recurrent and ephemeral centralities is the key challenge discussed in this chapter. For Françoise Lucchini, phenomena can be perceived through either a surface-based approach or a network-based approach. The surface-based approach observes ephemeral concentrations across territories, identifying specific events. The network-based approach captures flows of movement between territorial grid cells. These interactions between units dynamically reshape existing centralities throughout the day.

The second part of the book, titled “Networks as Creators of Centralities and Hierarchies”, examines networks, both material (e.g. ecological networks) and immaterial, which, though intangible, are inherently spatially embedded. Immaterial flows, such as media and knowledge flows, rely on and are anchored in territories characterized by inherent heterogeneity. Analyzing the circulation of media information, for instance, reveals how immaterial flows contribute to shaping or even reinforcing a polarized geography of space. In the third chapter, Robin Lamarche Perrin, Romain Leconte and Etienne Toureille present several representations of the structure of the media world, drawing on international media flows. The statistical distribution of countries represented in the French media landscape is highly unequal. Using a gravity model and applying graph theory formalism to media co-presences, the chapter explains observed distributions and identifies geopolitical or conflict-related centralities. In the fourth chapter, Marion Maisonobe examines the hierarchy of research activities and its dependence on urban hierarchies. Given its high degree of specialization, can research activities be considered a metropolitan specificity? Using scientific production data, urban data and research workforce data, Marion Maisonobe compares hierarchical distributions obtained between France and the United Kingdom. Beyond the connection between urban hierarchy and scientific hierarchy, the spatial organization of research is shown to be shaped by sectoral logics, historical factors and disciplinary criteria. Meanwhile, Yohan Sahraoui, Céline Clauzel and Jean-Christophe Foltête demonstrate in Chapter 5 how graph theory can address questions in landscape ecology. Animal species’ movements are modeled through networks of habitats and corridors, enabling the creation of synthetic indicators of centralities and hierarchies within territories. Modeling landscape connectivity as a graph facilitates the development of both global and local connectivity metrics as well as centrality indicators that rank species’ habitat patches. This type of modeling provides valuable insights for territorial planning, such as simulating the impact of road infrastructure on landscape connectivity or exploring the benefits of ecological reconnections for the entire network. The constructed metrics leverage network centralities, ecological corridors and hierarchical structures within animal habitat and movement networks.

The third and final part of the book is titled “Hierarchies and Centralities in Territorial Systems”. The chapters in this part aim to explain or measure the processes underlying centralities and hierarchies across different levels of territorial systems. In Chapter 6, Mohamed Hilal and Virginie Piguet provide an overview of centralities in geography and urban economics, proposing to use service and infrastructure levels to identify intermediate centralities. Traditional criteria such as population and employment levels are insufficient to identify small-scale centralities. These centralities can be ranked by their level of fragility, offering public policymakers tools to initiate support policies for intermediary centers. In Chapter 7, Cécile Tannier defines the principles and general laws of human geography that explain the processes of spatial concentration and dispersion observed in the geographical space. Interactions between individual and collective behaviors give rise to social and spatial configurations, some of which emerge as centralities that can be hierarchized. Settlement systems represent cases of strong emergence, where individual-level interactions significantly impact macroscopic patterns. Following this theoretical overview, Cécile Tannier explains how these frameworks account for the emergence of concentrations and dispersions within settlement systems at various levels, ranging from individual settlements to systems of settlements. The author discusses interaction processes, hierarchical diffusion of innovations and the spatial division of activities and knowledge, all of which are crucial for understanding the observed concentrations across various levels of settlement systems. In Chapter 8, Juste Raimbault explores the construction and evolution of hierarchies within city systems through a model examining the co-evolution of cities and transportation networks. These hierarchies manifest across various dimensions: urban population size, transport networks, corporate networks, etc. Juste Raimbault proposes indicators to quantify these hierarchies and introduces a dynamic indicator to capture their evolution. The findings from this analysis highlight that city system hierarchies are shaped by the structure and dynamics of transportation networks while revealing a lack of direct correlation between a city’s hierarchical level and its accessibility.

These recent studies, which leverage cutting-edge technologies and data, not only enrich and revise foundational concepts in the social sciences but also hold promising applications for territorial planning and urban design to inform citizen practices and perceptions. Each chapter of this book contributes to a nuanced yet optimistic conclusion about the potential of new information and communication technologies to reshape our understanding of geographic space in the making.

References

Alonso, W. (1964).

Location and Land Use: Toward a General Theory of Land Rent.

Harvard University Press, Cambridge, MA/London. doi: 10.4159/harvard.9780674730854.

Berroir, S., Mathian, H., Saint-Julien, T., Sanders, L. (2007). Les pôles de l’activité métropolitaine. In

La métropole parisienne. Centralités, inégalités, proximités,

Saint- Julien, T. and Le Goix, R. (eds). Belin, Paris.

Berroir, S., Mathian, H., Saint-Julien, T., Sanders, L. (2008). La mobilité dans la construction du polycentrisme métropolitain. In

Information géographique et dynamiques urbaines,

Thériault, M. and Des Rosiers, F. (eds). Hermès-Lavoisier, Paris, 31–57.

Fleury, A., Mathian, H., Saint-Julien, T. (2012). Définir les centralités commerciales au cœur d’une grande métropole : le cas de Paris intra-muros.

Cybergeo: European Journal of Geography,

588. doi: 10.4000/cybergeo.25107.

Gaschet, F. and Lacour, C. (2002) Métropolisation, centre et centralité.

Revue d’Economie Regionale et Urbaine,

1, 49–72.

McMillen, D.P. (2001). Polycentric urban structure: The case of Milwaukee.

Economic Perspectives,

25, 15–27.

Mills, E.S. (1967). An aggregative model of resource allocation in a metropolitan area.

The American Economic Review,

57, 197–210.

Muth, R.F. (1969).

Cities and Housing: The Spatial Pattern of Urban Residential Land Use.

University of Chicago Press, Chicago.

Pumain, D. (2004) L’invention permanente: la hiérarchie urbaine. In

Mélanges en l’honneur de Giovanni Busino,

Bridel, P. (ed.). Labor et Fides, Geneva, 66–86.

Yan, L., Wang, D., Zhang, S., Ratti, C. (2021). Understanding urban centers in Shanghai with big data: Local and non-local function perspectives.

Cities,

113, 103156. doi: 10.1016/ j.cities.2021.103156.

PART 1Spatial Practices as Indicators of Centralities and Hierarchies

1Urban Hierarchies and Centralities Through the Prism of Tweets and Telephone Calls

Julie FEN-CHONG

ThéMA, CNRS, University of Burgundy, Dijon, France

1.1. Introduction

The diffusion of information and communication technologies and associated mobile devices has made it possible to access data on the location of individuals throughout the day. Geolocated data from the use of mobile telephones and X (formerly Twitter) are examples. They each reveal different realities through the type of locations they make it possible to study and the sample of the population concerned. Mobile telephone data concern a large part of the world’s population, but they are held by different operators. They only enable approximate locating of individuals as they are reliant on the location of the mobile telephone network’s relay antennas. Conversely, data from a socio-digital network such as X (formerly Twitter) may include the precise location of the individual at the moment in time where they are tweeting, but these data can only be collected via X’s API and only concern its users, that is, a small part of the population. Despite their imperfections, these two information sources contribute to knowledge of the spatiotemporal organization of the geographical space. They provide a mass of information on individual mobilities from which it is possible to address the question of urban hierarchies and centralities at fine space and time scales rarely observed before.

Over the course of this chapter, we present the specificities of tweet and telephone call data for the construction of geographical information that can provide certain novel aspects in terms of describing the organization of the geographical space. Drawing on work conducted in recent years by researchers from different disciplines, geographers, computer scientists, geomaticians and physicists, we then show how these sources of information can shed light on urban centralities and hierarchies.

1.2. Tweets and mobile telephony: new sources of data to describe the organization of urban space

1.2.1. The generalization of geolocated data

A great and diverse quantity of geographic data is now available to geographers (Ciuccarelli et al. 2014). In 2017, Barbosa-Filho et al. drew up an inventory of the methods used to work on mobility from different data sources. Alongside traditional sources such as data from censuses or travel surveys, other types of sources can be used to gain an insight into human mobilities: banknotes or coins (Brockmann and Theis 2008), mobile telephone records (Ahas et al. 2007), GPS (Schönfelder et al. 2002) or data from social networks (Lenormand et al. 2014; Manca et al. 2017). These geolocated data from the use of information and communication technologies have become sources of information to understand individual mobilities and urban rhythms. They make it possible to identify and characterize spatiotemporal structures of spatial organization. Girardin et al. (2008a) introduced the notion of “digital footprints” to define these materials. This notion of a trace, or digital fingerprint, refers to the imperfect dimension of this fleeting information source.

However, these considerable volumes of data allow researchers to test hypotheses regarding the organization of space and human mobilities. In 2008, Brockmann and Theis used data provided on a banknote tracking website, https://www.wheresgeorge.com/, offering access to 14 million journeys to track the diffusion of banknotes. In 2008, González et al. studied the traces of 100,000 owners of mobile telephones from a European operator over a period of 6 months. In 2018, Galiana et al. collected 18 million data items from a French operator for the cities of Paris, Marseille and Lyon. By way of comparison, the sample from the French National Transport and Travel survey (Enquête Nationale Transports et Déplacements [ENTD]) in 2018–2019 looked at 20,000 households, which resulted in 12,000 respondents; the National Transport Survey (NTS), the British equivalent of France’s ENTD, drew on a sample of 6,192 households in 2019. “Conventional” surveys, for their part, concern a smaller number of individuals and allow little monitoring of practices over time.

Data from mobile telephony and social networks provide a source of information that is updated in real time. They bridge the gaps created by the length of time that elapses between different surveys on individual travel, time periods that can reach around 10 years for the ENTD, as well as household travel surveys (Enquêtes Ménages Déplacements [EMD]) conducted in French cities, which often focus on a given day of travel. The NTS is renewed every year and covers 1 week of travel.

The study of the temporal dynamics of the organization of geographical space demonstrates the importance of having data with a fine temporal resolution. Based on individual mobility data from French household travel surveys, Le Roux et al. (2017) show that the sociospatial organization of urban space changes over the course of a day. The different social groups travel according to their daily activities, thus modifying the segregation indices usually observed. The working day induces a greater social diversity in central neighborhoods. Conversely, some disadvantaged neighborhoods experience a process of “impoverishment” when their better-off inhabitants leave their homes to go to their workplace. Taking into account the travel and progressive locations of individuals over the course of the day’s activities makes it possible to understand the evolutions of the organization of urban space at a fine temporal scale. This provides useful complements to analyses of space based on the residential location of individuals. In this perspective, new data from telephony and social networks make it possible to analyze the dynamics of urban space at a fine temporal resolution.

The now widespread access to this type of data is concomitant with the development of digital tools and applications. Mobile telephone subscriptions are now commonplace throughout the world. In 2018, the International Bank for Reconstruction and Development counted 106 mobile telephone subscriptions per 100 inhabitants worldwide. This generalization affects all the countries of the world to unequal degrees. Certain countries in Sub-Saharan Africa (82 subscriptions/ 100 inhab.), South Asia (87 subscriptions/100 inhab.) and the Pacific (86 subscriptions/100 inhab.) do not achieve full population coverage. In contrast, OECD countries present very high figures: 121 subscriptions per 100 inhabitants on average. Mobile-telephone subscriptions are more widely adopted than other urban services in the majority of countries. In Senegal, 75% of the population have mobile telephones, while only 24% are connected to the electricity grid (Salat et al. 2020). Building on the widespread use of mobile telephones, socio-digital network applications have multiplied and are widely used by the world’s population. In 2020, networks such as X concerned 186 million active daily users worldwide, including 36 million in the United States.

Mobile telephones are tools carried almost permanently by their owners. They enable the collection of geolocated data. These have been the subject of numerous studies in the field of human mobilities regarding the analysis of geographical structures and their possible use for land-use planning and to respond to development issues (Blondel et al. 2015).

However, despite their considerable volume, data such as the location of mobile telephones or tweets are merely partial indicators of mobility and human activity. They require external validation as well as the implementation of methodologies adapted to their specificities. The imperfections of these data are described in the following paragraphs.

1.2.2. Imperfect or incomplete geographic locations

Data from mobile telephony and socio-digital networks are not based on survey or data-collection protocols established to respond to spatial organization issues. They rely on everyday tools and presents imperfections related to this dimension.

1.2.2.1. Data from mobile telephony

In the case of data from mobile telephony, it is not the exact location of the individual that is recorded, but that of their mobile telephone. The telephone is located based on the relay antenna of the mobile telephone network to which it connects.

The mobile telephone network is composed of relay antennas located on Base Transmitter Stations (BTSs). A BTS comprises several relay antennas. When a mobile telephone is activated, it connects to the mobile telephone network via the nearest relay antenna from which it can pick up a signal. These data are collected at the BTSs for maintenance operations on the mobile telephone network. They are marred by significant noise. For example, the rapid oscillation of a mobile telephone between two relay antennas can result in travel speeds of more than 1,000 kph. Moreover, the location of the telephones is positioned at the spatial resolution of the area of influence of a relay antenna. This varies according to the propagation conditions of telephone radio waves, but this information is confidential and researchers rarely have access to it. The area of influence of a BTS is often modeled using Voronoi polygons. Yet, as the distribution and density of BTSs present on the territory are themselves dependent on the spatial organization of the territory, BTSs are more numerous in dense intraurban environments, which gives a fine grid of Voronoi polygons (Figure 1.1). In this urban context, a municipality may comprise several BTSs. In less densely populated peri-urban environments, meanwhile, the grid is looser and BTSs can cover the area of several municipalities (Figure 1.2). The BTS spatial distribution therefore depends on the population density.

Figure 1.1.Voronoi polygons around Orange mobile telephone stations in Paris (Orange Labs data; Fen-Chong 2012)

Figure 1.2.Voronoi polygons around Orange mobile telephone stations in the peri-urban area of the Paris region (Orange Labs data; Fen-Chong 2012)

The events that give rise to a recording on the mobile telephone network vary in nature. Table 1.1 lists the events that were recorded by the technical network for 2G in France in 2007.

Table 1.1.List of events giving rise to a location recording on the 2G mobile telephone technical network (Orange data 2007; Fen-Chong 2012)

Event type

Status

Outgoing call

Outgoing SMS

Communication practices of the user at their own initiative

Incoming call

Incoming SMS

Communication practices of the user at the initiative of others

Handover (change of cell during a communication)

“Hybrid” either communication mobility or a “system effect”

International Mobile Subscriber Identity (IMSI) attach (terminal turns on)

Neither communication nor mobility, but user action

Change of location area

“Hybrid” either communication mobility or a “system effect”

Location area update (network update around 6 h)

User inactive and not necessarily mobile

It can be seen that different types of events are recorded, related to communication practices or mobility, but also that automatic recordings are made. These data from the technical network do not include socioeconomic characteristics of individuals and are rarely available for research. They serve as the basis for preparing invoicing. However, recent works employ Call Detail Record (CDR) invoicing data, which present the advantage of containing fewer events related solely to the technical network itself.

The data obtained by the mobile telephone network cannot determine the precise location of individuals. These data are imperfect (de Runz 2008) due to the process of observing and collecting the data, their lack of precision, ambiguity over how to classify these data according to the practices they reveal and the unexplained absence of information in certain places. While they provide little information on individual practices, they are nevertheless a significant potential source of knowledge at the territorial level.

1.2.2.2. Data from the X network

X (formerly Twitter) is a micro-blogging social network that enables its users to share messages of up to 280 characters (140 characters prior to 2017). Data from the X network include two possible location identifications: the location of the message and the location declared on the user’s profile. The first possibility requires message geolocation to be enabled. Only a very small proportion, 1%, of tweets are geo-referenced by their authors (Andrienko et al. 2013; Mitchell et al. 2013; Schlosser et al. 2021). The geolocation of the message is therefore derived either from the GPS system of the device used for X or from the location of the user’s IP (Internet Protocol) address (Graham et al. 2014). The second possibility is dependent on the user’s willingness to enter their place of belonging in their profile.

Access to X data depends on the services offered by the platform. It is possible to collect, via an application programming interface (API), messages corresponding to a preselected area, the bounding box. The API allows only a sample of messages posted on X to be collected for free. The information that can be obtained is of several types: the body of the message, the message ID, the author ID, their username, the quote identification (if applicable), the quoted user (if applicable), the retweet ID, the name of the retweeted user, the date and time of the message, the place if geolocation is activated, etc. These data cover several dimensions, that of the discourse and its dissemination, but also that of possible interactions between users. Not all messages posted via X are from real people, with bots frequently being used for advertising purposes. Kocich and Horak (2016) conclude that the REST API, the main tool for retrieving X data, does not enable completely reliable material to be obtained. Reprocessing is required to eliminate outlying data.

Moreover, X network usage is not homogeneous across the world. Cebeillac and Rault (2016) note that as of 2016, X had 316 million users worldwide, with 10 times more users in South America than in India or Africa. Certain groups of individuals, through their age, social class or occupation, also have a greater presence on the social network. Cebeillac and Rault estimate that only 3.7% of India’s population is present on X and that this mainly concerns city dwellers. Ultimately, only the wealthy middle class can be monitored on social networks.

1.2.2.3. Common features of both data sources

Work based on mobile telephone data requires partnerships with private operators. In France, Orange is a partner in research projects based on data from its network. The MobiTIC project (measuring people’s mobility and presence using ICT, 2020–2024), for example, funded by the French National Research Agency (Agence Nationale de la Recherche) is a partnership between several research teams1, the French National Institute of Statistics and Economic Studies (INSEE) and Orange. Louail et al. (2014) meanwhile draw on data from a Spanish mobile telephone operator for their works. Calabrese et al. (2011) use data from AirSage, which is not a mobile telephone operator, but a provider of location data from telecommunications. These data consist of 829 million items of location data collected from 1 million devices. Within the framework of a partnership between Open Lab at Newcastle University, Eurostat, Georgia Institute of Technology and the SENSE department of Orange Labs, Vanhoof et al. (2018) use the operator Orange’s data collected for 154 days in 2007. They concern 18 million users throughout France.

As with mobile telephone data, X data come from private operators and require the establishment of a partnership with the operator, or the use of free or paid services, as proposed by X. The data collection protocols put in place enable significant databases to be obtained: 10 million geolocated tweets across 373 urban areas in the United States for the year 2011 (Mitchell et al. 2013), and 300 million geolocated tweets shared by 3 million users in the United States between January and June 2013 (Luo et al. 2016).

The lack of precise information on individuals is a common feature of both data sources. Mobile telephone location databases do not include information on individuals (Calabrese et al. 2011; Fen-Chong 2012; Olteanu-Raimond et al. 2012; Louail et al. 2014). Similarly, X does not provide information on users other than that provided by their profile (gender and sociodemographic variables). Despite the possibility of using a pseudonym, Luo et al. (2016) believe that a significant number of X users use their first and last names. This information can provide sociodemographic clues. The distribution of surnames corresponds to spatial, social and ethnocultural logics (Mateos et al. 2011). Based on residential segregation mechanisms existing in the United States, Luo et al. (2016) estimate the likelihood of users belonging to an ethnic group based on their surname, likely residential location and census population data characterizing users’ homes. First names offer information on users’ gender and year of birth.

Data (traces) from mobile telephony and Twitter reveal the presence of a telephone or social network user in a territory at a given moment in time. Although collected at individual level, the data are not, however, exploited at this level. Once aggregated, they provide information about the presence of individuals at a moment of time, t, in the space. They then make it possible to analyze the presence of these users on the territory and understand the logics of spatial organization.

Thanks to the new location data provided by mobile telephony and certain social networks such as X, it is possible to identify human activity at a fairly fine spatiotemporal level. From this, many works manage to infer travel and mobility flows between places (Calabrese et al. 2011; Noulas et al. 2011; Frias-Martinez et al. 2012; Ranjan et al. 2012; Bonnel et al. 2015; Barbosa-Filho et al. 2017). The processing and exploitation of the considerable volumes of data employed for these works has been made possible thanks to both the progress made in computer science and the implementation of new methods.

Within these data volumes, many “anomalies” related to the protocol for data collection or recording remain. Researchers use these anomalies to identify “noise” and purify data sets before the processing phase. Among the anomalies identified, the travel speeds calculated between two of a user’s geolocated messages are sometimes too high to be realistic. In the case of mobile telephony, Iovan et al. (2013) propose an algorithm to filter the data pertaining to the “ping-pong” effect that occurs when the telephone signal oscillates between two BTSs within a short time interval, thus creating artificial travel. Similarly, Cebeillac and Rault (2016) propose using outlying speeds to filter X data. Additionally, the data volume can be reduced by eliminating location duplications or outlying location data. Iovan et al. (2013) propose a filter that removes duplicate records for the same user, that is, identical locations within a short time interval. A final source of anomaly lies in the spatial accuracy of geolocated messages (Kocich and Horak 2016). In certain cases, during extraction, the coordinates are reversed, which creates concentrations of tweets in inaccessible places like Antarctica.

1.2.3. Validity of tweet and telephone call data

Due to their lack of representativeness for the total population and their many imperfections, the question of the validity of tweet and telephone call data arises. Cross-referencing these data with more conventional data allows researchers to test their validity.

In 2014, Lenormand et al. compared information extracted from three sources: X, census data and mobile-telephone data for the cities of Barcelona and Madrid. They compared spatial and temporal distributions of human activity and home-to-work distances calculated from these three data sources. Their work shows a strong correlation for the three types of data for the chosen indicators. However, variations could appear, depending on the spatial extent of the cities for which these data were compared. Horanont et al. (2018) analyzed the data of 1.6 million users over the period of May 2014 for a city in South Asia based on three criteria: the operator’s market shares, the urban/rural population ratio and the gender/user ratio. The results of their study show that mobile-telephone data correspond to the distribution between urban and rural populations and gender distribution. Soto et al. (2011) used mobile telephone data to infer the socioeconomic level of BTS coverage areas. Their work shows that with call data from mobile telephones, it is possible to predict to 80% the socioeconomic level of the area covered by a BTS.

In addition to the question of data quality to construct relevant geographic information, Vanhoof et al. (2018) question the suitability of existing methods to extract residential locations from non-continuous location data such as mobile telephone data. Indeed, many users have very few location records. On a given day, a user has an average of four mobile telephone records made in only two different places. Comparing five methods of home detection, Vanhoof et al. observe a variation in the location of the home for 40% of users. For Vanhoof et al., several elements explain this variation: the diversity of local markets, the different uses of mobile telephones, the different definitions of home and probably the technical aspects due to the data collection process. The method of data processing must therefore be chosen according to the urban context concerned.

The question of the suitable level of spatial aggregation (Openshaw 1984) arises as soon as the objective is to couple these data with others. Lenormand et al. (2014), Louail (2014) and Galiana et al. (2018) use grids in order to cross-reference different types of data. In the study by Lenormand et al., the grid consists of cells of 2 km × 2 km and is used to couple mobile telephone data with Twitter and census data. In the study by Galiana et al., the grid is made up of cells of 500 × 500 m and is used to couple mobile telephone data with tax data from INSEE’s Filosofi database. Louail et al., meanwhile, test three different grids to identify hotspots from telephone data. Comparing the results obtained for cells 500 m, 1 km and 2 km wide allows them to evaluate the effect of cell size on the determination of hotspots. Jiang et al. (2021) use the street and street blocks as a basic spatial unit to delineate “natural” cities and observe the distribution of tweet density within these cities. They also stress the importance of the spatial reference unit to process these geolocated data.

We deliberately chose to discuss these data at length before presenting examples of their use for the study of centralities and hierarchies, because their availability leads to a reconsideration of the scientific approach in geography. In their article “Data-driven Geography”, Miller and Goodchild (2015) put forth the idea of a new scientific paradigm in which methods are configured to satisfy data, rather than data being collected and structured according to the methods employed. Data-driven geography allows us to focus on local and daily processes rather than on major forms of spatial organization. For Cavalière et al. (2016), this position involves methodologically exploring the data before formulating hypotheses on the organization of space.

We continue this chapter by drawing on these heterogeneous materials in order to identify centralities and hierarchies within urban spaces. We show how these new data from social networks and telephone calls allow us to understand the spatiotemporal dynamics of territories and are a valuable source for understanding the dynamic organization of territories.

1.3. Using tweet and mobile-telephone data to describe the organization of urban space

Reades et al. (2007) propose the term “cellular census” to describe data from information and communication technologies. Indeed, these data enable an approximation of the presence of individuals on the territory and therefore approach a form of population census. In this section, we see how these new information sources complement existing sources for describing human presence at different times of the day within the space. We see that these data sources can be used to investigate essential geographical questions such as centralities within space and related hierarchies.

1.3.1. Daytime densities

In 2000, Matthieu Mille sought to distinguish “resident” densities from what he called “shifting” densities or “daytime densities”. “Inhabitant” densities are calculated from the population census, which is one of the first sources of information on the occupation of geographical space. The census counts individuals at their primary place of residence and is not sufficient to understand the organization of human activities throughout the day. In order to know the daytime densities (Mille 2000), journeys by individuals throughout the day and their successive locations need to be taken into account. This makes it possible to account for variations in spatial load over the course of the day. The coupling of census data with mobility data allows this monitoring and enables the evolution of spatial organization over the course of the day to be studied (Le Roux et al. 2017).

Some countries do not have an official census. The African continent, in particular, witnesses significant variability in the quality and frequency of data collected. For example, Malawi has had a census every 10 years for the last 30 years, but of data collected at an administrative level, covering an area of 10 km2. In the Republic of Congo, the last census took place in 1984 and the data were also collected at an administrative level spanning an area of 10 km2. Jahani et al. (2017) show that, despite the lack of information on users, mobile-telephone data can be used to enrich official statistics in countries lacking such data. For countries that do not have reliable, regularly updated censuses, these data offer real progress. The techniques derived from machine learning make it possible to predict the socioeconomic characteristics of individuals.

For other countries, the possibility of calculating daytime (“shifting”) densities in addition to inhabitant (“resident”) densities is an advantage of these data. Many works employ data from social networks or mobile telephony to seek to describe dynamically the population present on the territory (Reades et al. 2007; Silm and Ahas 2010; Deville et al. 2014; Horanont et al. 2018; Salat et al. 2020). This dynamic knowledge of space is essential for the development and management of spaces in general but also in domains requiring precise knowledge of locations, such as epidemiology, or to manage natural or industrial hazards. The study by Deville et al. (2014)2, which focuses on France and Portugal, aims to compare densities calculated from mobile-telephone call data and from remote-sensing data (Figure 1.3). Remote-sensing data are more reliable than mobile-telephone data but overestimate densities in sparsely populated environments and underestimate densities in densely populated areas.

The calculated densities show seasonal changes in population density related to vacation periods in these two countries (Figure 1.4). The cities present high population densities during conventional periods, while during vacation season, the places that polarize the highest centralities are different. Mountains, coastlines, such as the Costa da Caparica resort, or certain sites, such as the Disneyland Paris amusement park or Roissy Charles de Gaulle airport, experience the greatest increases in population. Thus, in the case of countries that have reliable and regularly updated population data, data from mobile telephony make it possible to have dynamic population estimates and to respond to territorial management issues in real time.

These materials have been used by many researchers to understand population movements during the 2020 health crisis. They made it possible to estimate population travel and the effectiveness of lock-down measures (Galiana et al. 2020; Gao et al. 2020; Lawal and Nwegbu 2020; Nakanishi et al. 2021). People infected with the disease could be identified at their place of residence but it is also of interest to know where they go to in order to estimate the spread of an epidemic (Cebeillac et al. 2017). Knowledge of dynamic densities makes it possible to assess the population present at different times of the day and therefore the population vulnerable to a particular risk. The analysis of data from X enables individual activity spaces to be identified. Cebeillac et al. (2017) propose a two-step approach: characterize the occupation of space based on points of interest present on Google Places or Open Street Map, and then characterize the places of activity of individuals with the DBscan algorithm (Ester et al. 1996). This method of spatiotemporal aggregation from point data is also used by Olteanu-Raimond et al. (2012) to estimate the frequentation of Parisian tourist sites based on mobile telephone data.

Figure 1.3.Comparison of population densities calculated for Portugal. (A) Population density calculated from the census. (B) Population density calculated using Voronoi polygons from mobile-telephone data. (C) Population density calculated from a grid of 100 × 100 m using remote-sensing data (D–F) Close-up view of the city of Lisbon. (Deville et al. 2014, figure reproduced with the authorization of PNAS).

Figure 1.4.Seasonal changes in population distribution in Portugal and France. (A) France and Portugal in Europe (B–E). Relative difference in predicted population density between the summer vacation season (July and August) and the period from September to June at administrative level in (B) mainland Portugal and (C) mainland France. (D) Close-up view of Lisbon, showing Lisbon city center and the Costa da Caparica resort. (E) Close-up view of Paris, showing the airport (Paris Charles de Gaulle) and three frequently visited tourist sites: the Palace of Versailles, Disneyland Paris and Parc Astérix theme park (Deville et al. 2014, figure reproduced with the authorization of PNAS).

1.3.2. Centralities

Thanks to these point data on the presence of individuals in the space, it is possible to infer the intensity of attendance of certain places at different times of the day. The study of centralities thus has access to a new dynamic data source allowing both a greater spatial finesse in the description of places characterized by a strong centrality, and also a temporal perspective, with certain centralities marked by specific temporalities.

At the level of intra-urban space, centrality refers to places that polarize individuals, economic, political and administrative activities and collective representations (Pradel et al. 2014). Merlin and Choay (2015)