The scientific journal CyberGeo is included in the Scopus database. Paper quality checking. After the audit, you will receive a detailed report on your work together with recommendations for its refinement. The data are from the archives of Russian Ministry of Education HAC List (VAC List), Scopus and Web of Science Core Collection. For example, the UK Data Service enables users to deposit data collections and re-share these for research purposes. Publishing a data paper about the dataset, which may be published as a preprint, in a journal, or in a data journal that is dedicated to supporting data papers.
In this paper we present the methodological issues and choices related to the construction of the TRADEVE database, which allows following the population of European urban areas since 1961. Whereas most of the recent academic works related to this issue focus either on time depth (for larger cities) or on the large coverage of urban hierarchy (for a shorter period), one of the main interests of the TRADEVE database is to extend over a relatively long period (from 1961 to 2011) and to cover small and medium sized cities at the same time.
But, above all, it distinguishes by taking into account the spatial expansion of urban areas during a period characterized by a pronounced sprawling process. First insights are provided that allow studying the hierarchical and regional expressions of urban growth slowing down during this period. A cluster analysis performed on the demographic trajectories of cities shows that 22% are decreasing, i.e. 870 out of the 3,930 considered in the database. Along with the paper, the TRADEVE database fully documented with metadata is available online in open access.
1The following of European cities’ population has received an increased interest in the last decades due to various major issues. A first subject of interest deals with the general evolution of urban systems, the analysis of historical cyclic urban development and the way these cycles are spreading differently among hierarchical levels and regions. In that respect, one of the main questions tackled has long been related to the interpretation of the major turnaround of growth slowdown in the 1970s and to the questioning of the counter-urbanization concept (Cattan et al., 1999; Champion, 2001). More recent works participating in the theoretical field of complex systems have focused on the relation between urban system dynamics and the diffusion of innovations, by articulating the urban hierarchy structures on macro-geographical scales and the trajectories of the cities themselves on micro-geographical level (Favaro and Pumain, 2011; Cottineau, 2014; Pumain et al., 2015; Raimbault, 2018).
Beyond this theoretical framework there have been for some years new investigations on the evolution of the rich heritage of small and medium-sized towns in Europe (Pumain, 1999), which comprises a worldwide concern for the shrinking cities issues (Martinez-Fernandez C. Et al., 2016). The interest for this subject is related to major socio-demographic issues like the ageing process and slowing down of demographic growth (Bretagnolle et al., 2018; Rink et al., 2012). Along with this growing field of studies, the observation of more recent trends of a resurgence (regrowth after years of decline) has more recently made it necessary to consolidate comparative knowledge about the demographic future of cities (Rink et al., 2012, Wolff and Wiechmann, 2017).2Beyond their diversity, these issues have reinforced a common interest for the establishment of a longitudinal urban database in Europe. However, such databases are especially difficult to process since they raise at least three problems at the same time: the harmonization of urban definitions from one country to another, the availability of population data for each date at the level of local administrative units (LAU) and the geometrical and statistical sources for constructing urban delineations at each date. The challenging difficulties related to the statistical following of urban objects through time (Mathian and Sanders, 2016) are here exacerbated by the international context of comparison. Different choices have been made by researchers in order to support these studies.
Some of these works have emphasized the length of time series while focusing on large cities (Van den Berg, 1982; Turok et al., 2007). Others have included small and medium sized cities for a shorter period between 1990 and 2010 (Wolff and Wiechmann, 2017). The Europolis programme (Chatel, 2012) allows following cities over a longer period (1850-2010, with evolving perimeter since 1990); however, the data are not available in open access. 3 This procedure usually required a large expertise by hand for the previous census. It seems now toIn case C, an evolving delineation is based on simple selection criteria applied to building blocks. The minimal distance between built-up areas is only assessed for one year (reference year), in general in the latest year. A retropolation method is then processed at each date by checking the urban nature of each building block (minimal threshold of population or density) and if they are spatially concentrated (contiguity of building blocks).
For instance, a constant minimum density applied at different census periods enables following the urban sprawl of a city by aggregating the surrounding villages that exceed this threshold at each date. The perimeter can therefore be different at each date if the city has been expanding or retracting. That method has been chosen for the Geopolis database (Moriconi, 1994) until 1990.
It is also the method chosen for the TRADEVE database.Case D relies on an evolving delineation based on a reassessment (at each date, with a distinct reference year) of the continuous built-up criteria, in general a maximal 200 meters’ distance between buildings or urban spots, based on aerial photographs or satellite images. This is the most rigorous method, but it remains costly and time consuming. Only few countries use it to our knowledge (France, Sweden and Denmark).9Consequently, it can be seen from the above that methodological and technical considerations explain to a large extent the choice of a database model rather than another.
Evolving perimeters are not usually delivered by national statistical institutes. For instance, in France, INSEE (Institut National de la Statistique et des Etudes Economiques) provides the data about functional urban areas but only from 1990 to 2010. They have thus to be reconstructed by researchers.
Furthermore, the longitudinal local data are not easy to obtain. It raises also the issue of the structure of the database itself, since evolving urban perimeters imply retracing a complex genealogy of objects (fusions, scissions, apparitions, disappearances) (Mathian and Sanders, 2016).10The preference for either of these models also depends on the temporal coverage under consideration. The need to choose evolving perimeters or definitions may actually be justified by a medium or a long-term period (for instance, 1960-2010), especially if the extension of cities has been considerable during this period, whereas fixed limits are often preferred for the 1990-2010 period. On long periods, the choice of a constant delineation excluding spatial growth would lead to overestimating the initial population and to underestimating demographic urban growth. For instance, Paulus (2004) has shown for France that updating the delineation of urban areas between 1968 and 1999 included about 1 more million urban inhabitants at each reference year, as compared to the results obtained with evolving delineation. As a result, the rhythm of urbanization is fairly different (1.4% for evolving limits, against 1.1% for constant ones), even though global trends remain the same (a general slowdown of urban growth), apart from the latest period between 1990 and 1999 (a stabilization for evolving limits versus a slowdown trend for constant perimeter). Evolving perimeters appears thus preferable for longer periods, even though it supposes to tackle the sensitive issue of the fusions between cities.11For these reasons, cases C and D stand out as the most complex models to implement.
However, it is worth considering them since on long periods, the measure of the urbanization rhythm is sensitive to the choice of one of these models. 4 In the United Kingdom and Ireland, for instance, the LAU2 are built from the electoral districts. T15The second important source is the local Historical Population Database, which has been constructed by E. Gloersen et al. (2013) as part of a DG Regio project. This longitudinal database contains the local population data for six dates (every decade from 1961 to 2011), in the limits of the 2012 administrative local units. It refers to LAU2, except for Greece, Portugal, Lithuania and Slovenia (at LAU1, the level just higher), and can be downloaded on the Eurostat website.
We used an updated version after data checkings and adaptations of the LAU shapes made for the TRADEVE project (Bretagnolle et al., 2016b).16Despite the importance of such original sources, the definition of evolving perimeters for urban areas from 1960 to 2010 is far from evident. Indeed, National Statistical Institutes seldom provide evolving perimeters’ databases. The most complete approach would assume to apply the same rules to the data at each date, with, for instance, a 200-metre criteria for delineating the built-up areas extension for urban areas. However, the availability of such data on a European scale is not guaranteed for each reference year.
5 Both have already been used: the MUAs (Vandermotten et al., 1999) are based on a 650 inh/km² densit17An alternative approach comprises reconstructing past delineations in an indirect way, on the basis of present-day criteria (case C of Figure 1). This approach implies to set up rules in order to identify the administrative units that are part of an urban area at each date. On the basis of the most recent perimeter (UMZ 2000 delineation), it assumes it is possible to build urban areas’ delineations at each date by selecting only the LAUs that are considered as urban and that are contiguous. In this regard, the most discussed step relies on the identification of what is an urban LAU at each date.
Two criteria can be chosen in order to select urban building blocks at each date between 1961 and 2001: minimal density and minimal population.18We compared both methods and three different thresholds for each one, and finally chose the minimal population method and the 2,000-inhabitant threshold. The reconstruction of urban areas based on the most restrictive thresholds of population or density (10,000 inh. Or 650 inh/km²) leads to a strong fragmentation in the year 2000 as compared to the reference situation (UMZ 2000 perimeter). Consequently, choosing the 2,000 inhabitants or the 150 inh/km 2 seemed to be more appropriate (Table 1). 6 R and Postgis were used in order to implement these methods (see Cybergeo data paper).19We thus applied the same definition of urban areas through time (continuous built-up area defined from CORINE Land Cover image and Urban Morphological Zones 2000), the same delineation (UMZ 2000) and an evolving content based on two different selection criteria applied to LAU-building blocks (contiguity and the 2,000-inhabitant threshold). This method has been applied from 1961 to 2001 (Figure 2).
At this time, we used case B of Figure 1 (constant delineation with 2000 reference year) and the population change between 2001 and 2011 has been registered within the limits of UMZ 2000. 21The method has been implemented to construct 3,982 urban areas evolving from 1961 to 2011.
In terms of urban spatial extension, the results are very different from one country to another, depending on the average size of the LAUs (see the differences between Paris and Madrid, Map 2) and inequality of LAU2 size between large and small cities in Central and Eastern Europe (see Budapest and Prague compared to other cities). Indeed, in this part of Europe, the urban sprawl that began at the end of the 19 th century throughout the construction of suburban railways was followed by successive annexations of surrounding LAU2 by the central municipality.
Consequently, the size of the eponymous LAU2 (for instance, Prague or Budapest) was progressively enlarged, absorbing the new suburbs. However, the TRADEVE database allows to follow the rise of some new urban areas, as illustrated, for instance, in the north and east of Czech Republic. Sources: TRADEVE Database 2015.25Some inherent limitations due to sources’ availability or data geometries should therefore be taken into consideration.
First, regarding the 2011 population results, it has not been possible to use more recent urban perimeters than those from UMZ 2000. The perimeters built from CORINE Land Cover 2012 are not available yet and there is still no update of the population density grid which was used to build the dictionary of correspondence with LAUs.
As a consequence, the 2011 LAU populations have been attributed to the 2000 perimeter for now and we hope that it will be possible to update them later.26Furthermore, the model does not consider cases of spatial retraction: in the data model, when a building block is considered as urban at a particular date, it is conserved later on even if its population decreases under 2,000 inhabitants, for internal coherence reasons (see step 3 of Figure 2).27Lastly, the heterogeneity of LAU sizes partly impacts the measure of the spatial expansion or stability of urban areas in three different ways. First, one can assume that the measure of spatial expansion in time will be quite sensitive to the heterogeneity of LAU sizes between countries at the same date: small morphological variations are more easily registered/detected when the mesh is tight (like in France) than when it is quite loose (like in Nordic countries). For similar reasons, it is harder to register some spatial variations when territorial divisions have been adjusted to urban expansion: in Central and Eastern Europe, for instance, territorial divisions for cities are usually modified and adapted in order to better fit the spatial expansion of the built-up area. Lastly, the differing stability of LAUs that covers from one reference year to another might also influence the measure of urban growth or decline. That being said, this original method constitutes real progress as compared to former longitudinal studies.28The harmonized database TRADEVE is a useful tool to carry out a comparative approach at the European level. Using the same morphological definition of urban areas allows us to perform some reliable multilevel insights. In this section, the first investigations of the TRADEVE database are presented on the basis of different core indicators about urbanization, first at the macro level of countries (urbanization rates) then at the level of the cities (average population variation).
This last sub-section will focus primarily on population decreasing issues through a clustering analysis of demographic trajectories. 8 For instance, according to WUP data sources ( ), these urban.
9 Luxembourg, Slovenia, Cyprus, Malta and Lichtenstein were removed, due to the weak surface of the c29The European urbanization level (54% of the total population is living in cities in 2011) seems to be low, compared to the World Urbanization Prospect (WUP) 2014 (74% for UE28). This is because the WUP is a collection of national urban definitions that are not harmonized and most of the time not even representative of the definitions adopted by national statistical boards. Moreover, it is based on a minimum population size/density of urban areas much smaller than in the TRADEVE database. Figure 5 displays the evolution of urbanization levels computed for 24 countries. The curve representing the average (Europe) is characterized by a slight increase since 1991 (only 0.8%), which may be explained by the definition of cities adopted in the database: as we do not take into account functional urban areas, we do not include the rural areas located in the fringes of the largest cities and settled by commuters, a phenomenon that increased with the diffusion of automobiles and highways in the recent decades. Nevertheless, using a harmonized definition of urban areas reveals the intensity of the contrasts between highly urbanized countries (84% in the Netherlands, 81% in Belgium in 2011) and the less urbanized ones (47% in Slovakia, 48% in Lithuania in 2011).
If we distinguish countries according to geographical criteria (North-Western Europe, Atlantic and Mediterranean periphery and Central and Eastern Europe), results clearly show that in North-Western countries, urbanization has been more or less steady (with a slight increase in Sweden and Finland), whereas it has been increased mainly in the 1960s and 1970s in the Atlantic and Mediterranean periphery, and also in the 1980s in Central Europe. In these latter countries, stagnation occurs from 1991/2001, except for Bulgaria and Croatia (both show a total population decline). 10 The Annual Average Growth Rate (AAGR) is calculated as follows:30The map of average annual growth between 1961 and 2011 (Map 3) reveals that the majority of European urban areas have experienced positive urban growth in the last 50 years (0,941% per year on average), with the exception of a large number of German urban areas in the West and the East of the country, in Northern Italy, Southern Hungary and Northern Great Britain (England and Scotland) but also of some small urban areas settled in different regions of France. Urban areas in Spain, Poland, Slovakia, Romania, Bulgaria and Greece have reached extremely highly annual growth rates (from 4.5% up to 88.5%) between 1961 and 2011. Medium sized and small urban areas have also experienced important annual growth rates as is the case of urban areas in some parts of France, in the Netherlands, in Southern England and Ireland. 31Taking a closer look inside ten-year periods, the average annual variation rate shows a progressive fall of growth between 1961 and 2011 (from 1.903%/year to 0,350%/year), even though the definition of a constant perimeter between 2001 and 2011 (see section 2.2) tends to underestimate population growth.
Beyond this general trend towards slowing growth until 2011, differences are clearly displayed according to city size (Figure 6). The pace of growth decreases steadily according to city size, except for large cities (between 100,000 and 200,000 inhabitants), which grew more rapidly than small and medium-sized ones between 1961 and 1971. Progressively, there is a clear convergence towards lower rates, regardless of the size of the cities. The strong convergence observed during the last period is due to recovery of population growth for the largest cities (from medium-sized to very large cities) whereas population growth is still decelerating in small and over all very small cities.
Very small: 10,000 to 25,000 inh., 2. Small: 25,000 to 50,000 inh., 3. Medium-sized: 50,000 to 100,000 inh., Large: 100,000 to 200,000 inh., Very large: over 200,000 inh.NB: The cities characterized by exceptional rates (less than -15% and over 15% per year) have been removed from the corpus at each period. They concern mainly polders cities in Netherlands, new cities in United Kingdom or cities with large housing estates in Madrid suburbs.
They are 30 between 1961 and 1971, 16 between 1971 and 1981, 8 between 1981 and 1991, 3 between 1991 and 2000.Source: TRADEVE database 2015.32The mapping of this evolution displays the succession of two main periods. From 1961 to 1991, urban growth is especially strong in Southern and in Central and Eastern Europe, whereas decreasing situations are beginning to diffuse from the United Kingdom to other countries in Northern Europe:.In the period 1961-1971 (Map 4-a), the urban annual growth of European urban areas has been mainly positive in the whole of Europe (around 1,8%). However, we can already point out at that time the decline of the population of urban areas in Northern Europe (for example, in large urban areas of the United Kingdom) and a higher urban growth comparing to other urban areas in Spain and Greece and especially in Romania and Bulgaria.In the 1971-1981 period (Map 4-b), the gap in terms of growth tendencies between Northern Europe and other European countries has started to deepen.
Indeed, urban growth rates in the United Kingdom, the Netherlands, Germany or Denmark have fallen comparing to the previous period to less than -0.5%. At the same time, urban growth rates in Central and Eastern Europe and Spain have increased up to 4.5% or even more in the case of smaller urban areas of these regions.The decade between 1981 and 1991 (Map 4-c) reveals an increase of negative urban growth throughout France, Northern Italy up to Rome and in a few large urban areas in Spain and Portugal (Barcelona and Lisbon). In addition, a decrease has occurred in very small urban areas of Southern Hungary and in small sized urban areas in Bulgaria. Nevertheless, in other Central Eastern European urban areas, the average annual growth rates have remained highly positive (especially in Romania).33Since 1991 (Map 4-d), Europe has passed through an equilibration of urban growth rates and the regional tendencies have taken an opposite direction comparing to the previous period. Indeed, Central and Eastern European urban areas have experienced slightly negative growth rates due to the spread of suburbanization, to the sharp decline in the birth rate related to the post-socialist transition and, for the smallest cities, to the collapse of small-town industrial plants that had been established in the centrally planned economy era (Pirisi and Trocsanyi, 2014; Zdanowska, 2016). The growth rates in Western Europe have switched to slightly above zero (between 0% and 1%, and between 1% and 2.5% in the case of London). Moreover, urban areas with the highest average urban growth rates have become exceptions (suburbs of Madrid, Barcelona, Paris, Athens, and Klaipeda in Lithuania).
The beginning of the 21 st century (Map 4-e) is marked by a certain recovery of urban growth in a few regions of Europe presenting average annual growth rates just above zero. This is the case of the whole United Kingdom, Central and Southern France, Belgium, the Netherlands, Denmark, Spain, Italy and Poland.
Nevertheless, many urban areas in Germany, Hungary, Romania, Lithuania and Latvia have been still losing population as their annual growth rates were between -0.5% and -17.5%. An interesting point is that German central-eastern urban areas have never reached positive annual growth rates throughout the whole period 1961-2011, which confirms the results from Figure 6. 36The cluster 4 (in red) is the smaller one (only 330 urban areas) and it shows a strong population growth profile. It is a very specific profile, concerning an important share of small and medium sized urban areas; for instance, those which are settled at the fringe of large cities and which benefit from urban sprawl (see around Madrid, London, Paris, Stockholm, etc.) or which are located on the sunny coastal areas (Spain, France, Italy, etc.).
These small and medium sized cities are more apparent in Map 6, representing urban populations instead of urban patterns. 37Finally, the cluster 3 (in blue) aggregates stagnation and decay profiles and includes some large urban areas in Germany, central Great Britain and Northern Italy, also a large number of small cities in these latter countries, but also in France, Portugal, Scandinavia, Baltic countries, Hungary, Austria, Romania, etc. It concerns 870 urban areas, representing 22% of all the European cities, two thirds of which are located in only three countries (Germany, Italy and France) (Table 3). These results sound very similar to those obtained by Wolff and Wiechmann (2017), who conclude after an in-depth study of growth rates that 20% of urban areas were decreasing in Europe during the period 1990-2010.Table 3.
The editorial policyFirst entirely electronic journal for social sciences in the world, peer reviewed, European, open (free of charge for authors and readers), with a focus on geography and widely open to the diversity of research agendas and methodologies in all countries.Cybergeo is a success story with now more than one million papers downloaded every year. An app to look backThis app builds on 20 years of publication in Cybergeo.You can play with data, drawing geographical networks of authoring,studying and citing through countries, analyzing semantic networks per key words and articles’ content,you can review twenty years of epistemological and thematic trends in a variety of fields of scientific interest.The networks tell who studies what, where and how. Data are regularly updated. User guide DescriptionThis tab allows to explore the disciplinary neighborhood of cybergeo.
It combines citation data with semantic content of articles. The exploration can be at the article level or at an aggregated level with the full semantic network. Methodology Citation DataThe citation network is constituted by articles citing cybergeo articles, cited by cybergeo articles, and citing the same articles than cybergeo ( citing cited). Structure and number of papers is summed up by the following figure:The data is collected from google scholar (allowing to get only citing articles, thus the particular network structure). Semantic DataFor around 200000 articles of the citation network, abstracts were collected via the Mendeley API.
The extraction of significant keywords (n-grams), following Chavalarias and Cointet, 2013, allows the construction of a semantic network by co-occurrence analysis, which nodes are keywords. Community detection (Louvain algorithm), optimized on filtering parameters (hub filtering and low edge filtering), yield disciplines which are given in the following figure:How to useThe tab Network Exploration provides article-level exploration. The user can search an article in cybergeo with the datatable. Once one is selected, different informations are displayed if available note that 600 papers among 900 have citation and semantic information. citation neighborhood of the selected article. semantic content of the selected article as a colored wordcloud (disciplines legend above), and semantic content for the neighborhood.
User guide DescriptionHierarchical clustering of countries based on the aggregated semantic profile of their articles in Cybergeo. LicenseGNU GPLv3 DataSemantic profiles come from the analysis of Cybergeo keywords, Cybergeo texts and citation-neighbours' keywords presented within this app. Parameters Semantic ProfileThe analysis can be performed on three types of semantic profiles: 1. Cybergeo keywords grouped into 10 communities, 2. Cybergeo terms extracted from the full-texts and groupes into 10 themes, 3.
The communities formed by the keywords of the papers neighbouring Cybergeo articles in the citation network. The first set of themes indicates how articles are promoted and classified by their authors. The second set indicates the semantic areas of the words actually used to write the articles.
The third set refers to the inclusion of the article in a social network of research and the semantic areas in which the paper is used, cited or which the paper uses and cites. Set of countriesThe analysis can be performed on two types of countries: 1. The countries to which the authors are affiliated, 2. The countries studied in the article.
Number of clustersThe user can choose the level of disaggregation of the analysis by determining the number of clusters to be extracted from the hierarchical clustering. Hierarchical ClusteringAfter aggregating the semantic profile by country (using the mean frequency), the clustering method is performed on standardised data. It uses a euclidian distance and the Ward criteria (function 'hclust' and method 'ward.D2' in R).