Networks of scientific cooperation between cities: a multiscalar analysis

Marion Maisonobe

[email protected]

Toulouse University

NETSCIENCE project

D. Eckert, M. Grossetti, L. Jégou, B. Milard

Mondes Scientifiques - Labex SMS

Networks of scientific cooperation

between cities: a multiscalar analysis

Special Session: From international to global networks ?

mailto:[email protected]



Introduction : background and

objectives

Brief chronology

ANR Geoscience (2010 - 2013)

Netscience (‘Networks of science’) (2013 - 2017) in

the work package ‘Monde Scientifique’ - Labex

Structuration des Mondes Sociaux (SMS)

Spatial scientometrics framework; quantitative

geography of scientific activities

Introduction

Mark Jefferson, 1929

Proposals

The growth of scientific collaborations is perceived as one of

the main feature of the globalization of science.

From the 1970s, bibliometric data retrieved from the Science

Citation Index (SCI) are used to measure this growth.

Usually these data are only processed at the country level.

Other research only relies on limited perimeters : the top

cited cities or European cities.

The spatial scientometrics method developed in Toulouse

University allows us to perform multiscalar analyses from

the urban area level to the world level.

Introduction

Objectives

To process all the publications data retrieved in

the SCI from 1999 to 2008 at the urban area level

To apply a geocoding and a counting method in

order to assign data to urban area perimeters

To use network analyses’ tools in order to unravel

the structure and the growth of the world network

of interurban collaborations

Introduction

Dataset, method and results

‘To use a databasis such as the SCI only enables us to measure the relative

position of countries in the mainstream literature and to include solely the cited

part of science coming from developing countries, forming what we could call

the « world scientific community ».’ (Polanco, 1990, p. 45).

The Web of Science’s world

The Web of Science coverage in 2008. Number of journals per year per

index. Source: ISI Thomson Reuters

The annual Web of Science coverage in 2008 (the 3 main citation index) Number of

journals

Dataset and method

Dataset : the Science Citation Index created in1964 in Philadelphia

A set of bibliographic references issued in selected journals

More than a million publications per year between 1999 and 2010

Two step method

1. Geocode adresses of authors for each publication

2. Assign data at the urban area level

Analyze the spatial distribution of scientific activity and exchanges between places through co-authorship data

Web visualization tool shared among all project participants to control the

results of automatic geocoding processes and delineate urban areas’

perimeters (developed in 2010 with OpenLayers)

Counting method : the case of an article co-signed by 16 authors in Paris and Toulouse

An only link :

Paris-Toulouse

Value = 1

Publication weight :

Paris = 0,5

Toulouse = 0,5

Whole Normalised Counting (Gauffriau et al, 2008)

The geography of contemporary scientific activities

The world growth of scientific publications between 2000 and 2007 Source: SCI Exp (articles, reviews and letters); 3 year moving-average

« The myth of global science » ?

Dataset : SCI Expanded (1999-2001; 2006-2008). 7000 journals, 194 countries, 10730 urban areas

Focus: two normalized time-series to measure global trends:

1. A growing dispersion of scientific activities: toward a deconcentration of the production at several scales

2. The growth of collaboration between places: a growth even more significant inside national boundaries

Visualization: map of the production per urban area and collaboration network between places through co-authorship

Map (LJ) : Evolution of the geographical dispersion of scientific publications

in Europe between 2000 and 2007 (Source: articles, reviews and letters, SCI Exp)

The growth of interurban co-authored scientific publications between 2000 and 2007.

Source: SCI Expanded (articles, reviews, letters), 3-year moving average.

The growth of international collaborations is part of the

overall growth of interurban collaborations, whether or not

they are in the same country.

For countries where

there is a growing

dispersion of scientific

activities,

the growth of interurban

collaborations is more

pronounced within

national boundaries

than between them.

Growing autonomy and

polycentrism for

developing countries !

The partitions of the world territory

Plateau de la Justice, december 2014

Beneath the surface of the world scientific community

Clusters of countries: set of similar countries involved in frequent and intensive scientific partnerships (Frame & Carpenter, 1979; Schubert & Braun, 1990; Okubo et al., 1992; Glänzel & Schubert, 2005).

Objectives : To measure scientific similarity not only at the country level but also at the city level (Louvain, VoS and OSLOM methods)

Co-authorship between cities = from 31% to 35% of the world production between 2000 and 2007 (SCI Exp)

What are the preferential channels of collaboration for scientists according to their location in the world?

Are the macro-regional areas detected in 2000 still structuring in 2007? The European Research area and other macro-regional areas.

Low resolution clustering result

The clustering of the scientific network of collaborations between urban areas in 2007

using the VoS Clustering method (Waltman et van Eck, adapted from the Louvain method)

This interactive map can be explored on the web site Coscimo.net

Developed with Laurent Jégou

Dendrogram plot of a community detection. The Louvain method is applied step by step to the

interurban network of scientific collaborations in 2007.

Multidimensional map of european cities and cities under the influence of Europe according to the

intensity of their scientific collaborations = similarity index (Source : SCI Expanded)

Focus on the European Research Area

(ERA)

Stability despite world dynamics: integration of new production spaces

Multidimensional map of countries derived from the world network of scientific collaborations in

2000 (SCI Expanded)

The evolution of the scientific collaborations’ distribution among macro-regional research areas

(2000-2007). Increase of the share of intra-national and intra-cluster linkages.

European Research Area 2000* 2007* Russian world 2000* 2007*

Intra national links (%) 34,5 35,6 Intra national links (%) 17,2 26,5

Intra european links (%) 26,8 26,4 Intra russian links (%) 2,0 1,9

External links** (%) 38,7 38,0 External links** (%) 80,8 71,6

100 100 100 100

Number of publications 122994 186408 Number of publications 8917 12207

North America 2000* 2007* Australasian world 2000* 2007*


Intro north am. links (%) 4,1 4,1 Intra australasian links (%) 4,7 5,1


100 100 100 100


Asian world 2000* 2007* Arab world 2000* 2007*


Intra asian links (%) 4,5 5,2 Intra arab links (%) 5,8 7,0


100 100 100 100


South America 2000* 2007* Sub-saharan Africa 2000* 2007*


Intra south am. links (%) 5,9 5,2 Intra s.-saharan af. links (%) 3,7 5,4


100 100 100 100


*Whole normalized counting (WNC ), 3-year moving average. Source: SCI Expanded (articles, reviews and letters)

**External l inks: l inks with the rest of the world

Map of the evolution of scientific collaborations between macro-regional collaboration area

(2000-2007)

’’Domestic and South-South

research capacities play an increasing

role in developing country and emerging

economy innovation systems as they

gain momentum”

(Ponomariov & Toivanen, 2014).

Community detection methods

Scientists are working in organized territories. National frameworks did not tend to regress or disappear between 2000 and 2007, the reverse is the case.

The multiplicity of attachment logics and organizational levels difficult to delineate cohesive areas from the distribution of collaborations

Solutions:

Make the resolution parameter varied

Test the evolving integration of spaces using previously known perimeters (for instance the European Research Area perimeter)

Using overlapping community detection methods. E.g. the OSLOM method (Lancichinetti, Radicchi, Ramasco, & Fortunato, 2011).

Applied on the co-authorship matrix of cities, OSLOM enabled us to detect clusters such as : the Netherlands, Scotland, Scandinavia, Eastern Europe….

Conclusion : the concept of

community in the contemporary

scientific world

The concept of scientific world

‘And since the intellectual world, the realm of abstract symbols having universal application, cuts across the material world on which nations are built, it is inevitable that a strong allegiance to the former entails a tendency to ignore the latter.’ Norman Storer, 1971

This plurality of worlds is not external to actors that shape and experience it : there is no external "structure" in which scientific activity is deploying.

Scientific activity does not develop in an universal space, an absolute spatial reference. Each time, scientific activity create its own geography, sets a world, as it is the product of a combination of worlds.

The concept of community

From ‘the scientific community’ (Polanyi, Hagstrom…) = an

unified, pacific and flat world

To scientific communities with a unifying interest or project

depending on :

A territorial membership: the developing of national scientific

communities (ex: Vessuri)

A scientific membership: the developing of epistemic

communities (ex: Roth)

Future lines of research

In a globalization context for science, multiscalar approaches

are necessary

World-wide access to transportation, information, and

communication technologies, as well as collaborative research

policies make necessary to better address the structures and

dynamics of collaborative networks

Our work invites to consider how these scales are organized,

to explore territories, regions and borders that underlie macro-

regional scientific areas