Navigation – Plan du site
2019
889

Potentials and limits of (geo)digital footprints in mobility analysis: the example of the data from the BlaBlaCar carpooling platform

Boris Mericskay
Traduction de Alvin Harberts
Cet article est une traduction de :
Potentiels et limites des traces (géo)numériques dans l’analyse des mobilités : l’exemple des données de la plateforme de covoiturage BlaBlaCar

Résumé

Big data is a field of investigation rich in promises but still complex in renewing spatial mobility analysis. In fact, thinking about mobility through the prism of digital footprints raises many questions regarding the data nature, the accessibility procedures and the methods and techniques of treatment. This paper aims to explore the potential of digital footprints in the analysis of mobility through the example of data from the Carpooling platform BlaBlaCar. By analyzing (temporal and spatial) trips to and from city of Rennes over the course of 5 months, the goal is both to draw a portrait of Carpooling in the capital of Britanny and to give an overview of advantages and limitations of these data in understanding this new form of mobility.
Key-words :
exploratory data analysis (eda), spatial analysis, data, mobility, GIS

Haut de page

Texte intégral

Introduction

  • 1 Digital footprints can be defined as footprints left behind when surfing the Web or using a service (...)

1With the advent of digital technology, massive amounts of data are increasingly feeding into reflections on the functioning of contemporary societies and in particular on the ways in which knowledge about them are produced (Bastin and Tubaro, 2018). The variety of digital footprints1 produced by individuals and connected objects offers multiple research perspectives, including the ability to capture and show spatial practices with granularity and digital footprints on spatio-temporal scales that are difficult to access with other methods (Beaude, 2015; Mericskay et al., 2018). Whether it is smart card, mobile telephony or journey booking data, all these (geo)digital footprints provide new approaches to understanding individual and collective mobility.

2Beyond promising perspectives, the mobilization of (geo)digital traces, emanating from public and private actors in the production of knowledge, raises multiple questions, both in terms of the nature of the data, accessibility and the associated tools and processing methods. On the one hand, the accessibility of these new data sources is characterised by a series of constraints limiting their use (technical, legal, ethical, economic, etc.). On the other hand, analytical methods and tools are still in the consolidation phase and require exploring and experimenting with the potentialities of these data of a particular type (voluminous, heterogeneous and in real time).

3This article aims to explore the potential of digital footprints in understanding carpooling through the data spectrum of the BlaBlaCar platform. Through the temporal and spatial analysis of a corpus of 150,000 trips, the objective is both to paint a portrait of carpooling organised within the city of Rennes and to provide an overview of the advantages and limits of this data in understanding this new form of mobility.

4The first part of the article focuses on big data as an object and material for mobility analysis. The main forms of travel data available are presented (mobile telephony, smart card, bicycle sharing system, social networks, etc.) as well as their accessibility requirements (web scraping, APIs). The technical and methodological issues involved in the analysis of these specific data are also explained.

5The second part presents an exploratory study of the BlaBlaCar platform data on trips to and from Rennes over the course of 5 months. It reviews the various stages of the processing chain set up and presents the main results. The methodology for collecting and setting up the database is explained in advance. The main logics of the trips offer from a temporal point of view are then presented according to different characteristics. Finally, the spatial and territorial dimension of the data is addressed at two scales of analysis: firstly on the national scale, which considers the routes according to their origins, destinations and distances; the other one at the local scale, which looks at the spatial configuration of carpooling areas within the city of Rennes.

Digital footprints as spatial markers of mobility

Analyze spatial mobility by digital footprints

6Researchers have long relied on count surveys or retrospective surveys (such as household travel surveys) to attempt to qualify, quantify and model spatial mobility. In addition, the use of data collection devices on individual spatio-temporal practices (data loggers or dedicated mobile applications) has introduced new modalities in the understanding of forms of travel (Depeau and Quesseveur, 2014; Gwiazdzinski and Klein, 2014). Finally, more recently, the arrival of large amounts of data produced by public and private operators, such as smart card, mobile telephony or social network activity data, has opened up an important field of investigation in the study of travel (Chen et al., 2016; Schwanen, 2017). These new data sources, which are increasingly being explored by researchers, are now contributing to a renewal of methods for collecting, processing and analysing individual and collective mobility (Vidal, 2015).

7Mobile phone data, which make it possible to study large volumes of individuals over long periods of time, are interesting markers to explore in order to understand mobility. Several studies clearly highlight the value of these data for understanding and quantifying the spatio-temporal modalities of individual travel within a city or neighbourhood (Toole et al., 2014; Senaratne et al., 2018). From a methodological point of view, the studies mobilizing these data are largely based on the identification and formalization of mobility patterns (Calabres et al., 2012; Jiang et al., 2017), the exploration of origin-destination patterns (Iqbal et al., 2014; Alexander et al., 2015) or the integration of these new sources into existing modelling processes.

8Data from public transport smart cards make it possible to highlight spatio-temporal dynamics of travel (Barry et al., 2002; Zhong et al., 2015; Mohamed et al., 2017), changes in practices over time (Briand et al., 2017) or even intermodality logics (Richer et al., 2018). In the same vein, data on bicycle sharing system are also being studied by researchers to extract information on urban travel dynamics. The information which is compiled, aggregated and processed (schedules, station occupancy, location of departure and arrival stations, journey time, etc.) allows users' behaviour in time and space to be visualised in the form of maps and graphs. Beyond simple visualization, many studies further explore the analytical dimension for management optimization (Vogel et al., 2011; Wergin and Buehler, 2017). The work of Etienne Côme and Latifa Oukhellou (2014) on the use of Velib's, for example, offer an understanding of the practices of this service in complex forms based on a range of analyses (stations activity patterns, origin-destination matrices, station typology).

  • 2 The Uber Movement service, for example, provides free access to anonymous data from more than two b (...)
  • 3 The Waze application (owned by Google) offers public authorities a partnership program called Conne (...)

9Data from social networks such as Twitter or Foursquare are also increasingly being explored to understand and apprehend, in new forms, the dynamics of travel or frequentation, particularly in urban areas. Work on the subject focuses on the analysis of tourist or immigrant mobility at national and international scales (Hawelka et al., 2014; Jurdak et al., 2015; Rashidi et al., 2017). Finally, the growing role of digital giants in collecting and producing a multitude of data on individual mobility opens up many opportunities. On the one hand, platforms specialised in connecting drivers and passengers, such as Uber or BlaBlaCar, constitute voluminous databases with diversified content (user practices, origin-destination matrices, traffic conditions, etc.)2. On the other hand, mobile navigation aid applications, such as Waze or Google Maps, collect, archive and process large amounts of data in real time on the speed of their users, travel times, congestion or accident location3.

Big data accessibility and processing: economic and technical challenges

10Trying to make all these heterogeneous digital footprints speak for themselves means confronting the multiple techniques of transforming data into information and knowledge, by seeking to open the algorithmic black boxes from which they are derived. With the current trend of big data, the proliferation of data does not necessarily imply open availability. Even if big data is developing alongside the open data movement, access to data is increasingly controlled by the public and private actors who produce it. There are three main methods of accessing these new information sources: the use of web scraping, the establishment of partnerships with data producers and the mobilization of APIs (Bastin and Tubaro, 2018).

  • 4 An API (Application Programming Interface) is a programming interface that allows you to connect on (...)

11The first way to build data sets is web scraping, which allows automated retrieval of web page elements, and is increasingly used today. The principle is simple: transform unstructured information within web pages into structured data that is more easily usable. Secondly, the establishment of a partnership with a platform where you can purchase options that allow access to certain databases (smart card, social networks, mobile phones). Finally, the use of APIs4, which make it possible to retrieve certain content in structured forms, is becoming more widespread among professionals and researchers. The increasing use of these Web services is partly due to the fact that more and more companies and public organisations are making some of their data available via APIs. However, much of the potentially recoverable data is, in practice, increasingly locked up by private actors, eager to control, for commercial purposes, this highly strategic data (Quesnot, 2016). The generalization of APIs thus contributes both to the opening of data by making them available in the form of standardized protocols and, at the same time, to their closing through the generalization of use restrictions (user licenses, data monetization, query limitations). The increasing use of APIs also increases the opacity of data collection, aggregation and release processes. In the end, an API is simply an automaton that executes a request from an external service provider without knowing how it organizes and processes information (Masure, 2014).

  • 5 The example of mobile phone data is symptomatic of the locking systems which exist. Beyond their ex (...)

12Once the data has been collected, recovered or purchased, its processing is often a challenge. On the one hand, the content produced is rarely structured to be cross and analysed with other more traditional types of data (public statistics, administrative databases, business data, etc.). Data from big data is often poorly documented and technical (exchange formats) and semantic (associated information) interoperability with existing systems is still limited (Rashidi et al., 2017; Yang et al., 2017). On the other hand, there are not yet any real software solutions dedicated to processing this particular type of data (which is large, heterogeneous and in real time). Some tools allow basic operations, but the functionalities of transformation and spatial analysis for example, are quite limited compared to geographic information systems (GIS). As for new approaches involving data mining, machine learning or predictive analysis, they are still limited and far from operational practices5.

13Even if the question of data analysis is at the heart of the development of big data, it is important to stress that the uses around these data remain exploratory and are essentially part of inductive approaches, or as Plantin and Russo (2016) rightly point out, the multiplication of available databases clearly marks the return to empirical and deterministic concepts. In other words, the formulation of hypotheses and theories is increasingly derived from the exploration of new masses of data, with the leitmotiv "first the data, then the method". It is in this approach that the second part of the article, which proposes to take a comprehensive look at carpooling practices through the data analysis of the BlaBlaCar platform, is part of this approach. The objective is not to affirm or refute existing hypotheses on the subject but, through the exploration of original (geo)digital footprint, to try to highlight major trends, logics and dynamics relating to this new mode of travel.

Understand the carpooling offer through the analysis of BlaBlaCar platform data

Carpooling, a travel mode in development that has not yet been studied

14Carpooling has undergone significant development over the past ten years or so, for economic, political, ideological and technological reasons. Popularized in the 1980s by associations, carpooling is now essentially organized online with the development of public and private platforms for networking (ADEME, 2015). It is important to note that there is no single type of carpooling since the practices around this mode of travel are so diverse (Jacquot, 2018). There are thus several forms of carpooling (organised, dynamic, commuting, occasional, short distance, long distance, etc.) that refer to different ways of connecting, sharing costs as well as frequency or types of journey (CEREMA, 2016; ADEME, 2017).

15Despite a growing interest in this new mode of travel, carpooling is still poorly qualified because it has not been studied very much. Apart from a few essentially qualitative surveys on the profiles and motivations of carpoolers, few studies really qualify and, above all, quantify the practices of this form of mobility in development (ADEME, 2015; Castex, 2015; CGDD, 2016). The low interest in this issue can, in part, be explained by the low modal share of carpooling, which accounts for about 2% of long-distance travel (CGDD, 2016), or by the lack of available data on the subject. However, with the development of online services specializing in carpooling such as iDVROOM or BlaBlaCar, new perspectives are emerging in terms of understanding carpooling practices. These platforms, for which data is at the heart of their services, marketing and customer relations, have understood the importance of making some of their data available for the creation of innovative services and the production of new knowledge.

The example of the BlaBlaCar platform data

  • 6 Organized carpooling implies organized planning between the driver and the passenger(s) who use a c (...)

16Launched in 2006 in France, the BlaBlaCar carpooling platform has become the main service for connecting drivers and passengers for organised carpooling trips6. In France, BlaBlaCar holds more than 95% of the organized carpooling market, leaving very little room for its competitors. To give a few figures, the platform brings together more than 40 million members, who offer more than 10 million trips annually throughout Europe, with a majority on the French territory. Although the platform launched a dedicated commuting service called BlaBlaLines several months ago, most of the carpool trips published on BlaBlaCar are occasional medium and long distance trips.

17In 2016, BlaBlaCar opened an API7 allowing free access to certain data relating to journeys published on the platform. It is important to note that this API is a journey discovery API, which is intended to be mobilized by websites offering route comparators integrating train, bus and carpooling. On the request side, it is thus only possible to query the trips that are for sale on the platform, so the trips which have already been completed8 are not available. With regard to the route search criteria, it is possible to select the published trips according to a departure and/or arrival city, date, time, price and number of available places. As for the answers returned by the API, in other words the attributes that characterize the trips sought, the service offers about thirty fields relating to the location of the departure and arrival points (geographical coordinates, address, name of the city, country, etc.), time (date and time of departure, travel time), price, cars or even booking methods.

  • 9 During the data collection phase, it was still possible to retrieve the trip history over three mon (...)

18The data given via the BlaBlaCar API is available in structured formats (JSON and XML) and can be quickly reused for short-term analyses. Indeed, it is not possible to recover corpuses of long-duration trips (weeks, months or years) since the API imposes limits and quotas in the interrogation and extraction of data. In response to this lock, a series of daily queries collected published trips to and from Rennes over a five-month period9. The processing chain for data extraction and structuring is organized around four main steps (Figure 1):

Figure 1: Steps of dataset preparation (source: author)

Figure 1: Steps of dataset preparation (source: author)
  • Collection of daily trips to and from Rennes in the form of individual data sets.

  • Aggregation and structuring of daily datasets in the form of two databases: origin-trips and destination-trips.

  • Conversion of JSON formalism to CSV for easier use within statistical processing, data visualization and GIS software.

  • Spatialization of data through geo-referencing of geographical coordinates of trips (departure and arrival points) using GIS software.

19The dataset compiled for the study thus lists 151,125 trips, including 75,390 departures and 75,735 journeys to Rennes. With regard to the variables mobilized, the choice was made for certain attributes that appear to be the most relevant for the analysis of trips according to temporal, spatial and territorial inputs (Figure 2).

Figure 2: Summary of the attributes retrieved for each trip studied (source: author)

Figure 2: Summary of the attributes retrieved for each trip studied (source: author)

Temporal analysis of journeys to and from Rennes

20By analysing the daily frequency of the supply of trips to and from Rennes over the five months of the study, a cyclical logic emerges that is strongly linked to relative weekend trips (Figure 3). Fridays are clearly the days with the highest number of journeys, both to and from Rennes. Beyond the weekend peaks, some tourist periods are also the subject of a significant offer of trips such as the extended weekend in May, the week of August 15 or the last week of the summer holidays. More generally, the number of departure trips appears to be strongly correlated with both the number of arrival trips and especially with the day of the week (Figure 4). This form of carpooling with specific time limits is mainly practiced by young working people, students, non-motorized people, employees returning home on weekends or even holidaymakers (ADEME, 2015).

Figure 3: Number of daily trips over the five months of the analysis (source: based on BlaBlaCar data)

Figure 3: Number of daily trips over the five months of the analysis (source: based on BlaBlaCar data)

Figure 4. Distribution of trips by number of departures, number of arrivals and day of the week (source: based on BlaBlaCar data)

Figure 4. Distribution of trips by number of departures, number of arrivals and day of the week (source: based on BlaBlaCar data)

21The monthly travel frequencies, which range from 25,000 trips (in July) to 34,000 trips (in April), highlight a certain seasonal use of the platform. The supply of journeys decreases with the arrival of summer, particularly in July, and resumes in August with the end of the summer holidays (Figure 5).

Figure 5: Number of trips to and from Rennes by day of the week (source: based on BlaBlaCar data)

Figure 5: Number of trips to and from Rennes by day of the week (source: based on BlaBlaCar data)

22Regarding the weekly rhythm of the carpool offer, the aggregation of trips by day of the week shows a cyclical use of the platform over a typical week (Figure 6). As mentioned above, the supply of journeys is mainly concentrated on weekends, with a main peak on Fridays (24% of journeys) and two other smaller peaks on Sundays (15%) and Mondays (14%). On the other days of the week, the carpooling offer is more modest but there is a real use of the platform for trips in the middle of the week (35% of trips for Tuesdays, Wednesdays and Thursdays). Concerning the origin-destination dynamic, it is interesting to observe a certain pattern of departures and arrivals from and to Rennes on weekends. There is a greater offer of trips leaving from Rennes on Fridays, and more return trips on Sundays and Mondays. Regarding the relationship between day of the week and distance of trips, weekend trips have higher average distances (220 km on Friday versus 177 km on Tuesday).

Figure 6. Number of trips to and from Rennes by day of the week (source: based on BlaBlaCar data)

Figure 6. Number of trips to and from Rennes by day of the week (source: based on BlaBlaCar data)

23In terms of timetables, there is a certain difference in the daily rhythms between departing and arriving trips to Rennes (Figure 7). On the one hand, the frequency of departure times for journeys from Rennes follows a supply logic that starts early in the morning and stabilises in the middle of the afternoon before reaching a peak from 4pm to 7pm, which covers more than 40% of journeys. On the other hand, the daily rhythm of departure times for journeys to Rennes is characterised by two daily peaks, one in the morning from 6am to 10am (32% of journeys) and another in the late afternoon from 4pm to 8pm (28% of journeys). Concerning the relationship between departure time and distance, it appears that the longest journeys generally leave early in the morning or late in the evening.

Figure 7. Number of trips to and from Rennes by time of day (source: based on BlaBlaCar data)

Figure 7. Number of trips to and from Rennes by time of day (source: based on BlaBlaCar data)

Spatial analysis of trips to and from Rennes

24The structuring of BlaBlaCar data, which is based on the principle of an origin-destination matrix, makes it possible to quantify the paths between sectors of analysis. The information of the geographical coordinates of the departure and destination points of each journey is an interesting avenue to explore in a search for visibility of possible territorial carpooling dynamics. In order to show the heuristic potentialities of these (geo)digital footprints, the spatial dimension of the paths is considered at two scales of analysis, on the national and local levels.

Analysis of trips to and from Rennes on a national scale

  • 10 The flow map creates a line between the departure and arrival point of each trip.

25To understand the origins and destinations of the available trips, a first analysis at national scale highlights the general spatial dynamics of carpooling in Rennes. After spatialization of the departure and arrival points (by geo-referencing the associated geographical coordinates), the creation of flow maps10 makes it possible to visualize the attraction points in a generic way and to understand Rennes' area of influence in terms of occasional carpooling (Figure 8).

Figure 8. Travelling to and from Rennes (source: based on BlaBlaCar and OpenStreetMap data)

Figure 8. Travelling to and from Rennes (source: based on BlaBlaCar and OpenStreetMap data)

26Upon reading these maps, it appears that the carpooling offer to and from Rennes covers a large part of the national territory, but with varying degrees of concentration. It can be seen that the majority of trips leave and arrive within the same area of influence, namely in Brittany, the Pays de la Loire, Normandy and Ile-de-France. It can thus be argued that proximity to cities plays a major role in the supply of trips.

27This regional polarity is largely highlighted by the analysis of trip distances (Figure 9). The average distance of trips from Rennes is 161 km and the median is 111 km. With regard to the distances of journeys to Rennes, the values are higher with an average of 185 km and a median of 135 km. In terms of the proportion of distances travelled, a few indicators highlight that the platform's uses are mainly intended for medium-distance travel. For example, the proportion of journeys between 10 and 100 km is 28%, that of journeys between 100 and 300 km rises to 53%, while for journeys over 500 km, their share is only 5%.

Figure 9. Distances of trips between 10 km and 800 km (source: based on BlaBlaCar data)

Figure 9. Distances of trips between 10 km and 800 km (source: based on BlaBlaCar data)

28In order to qualify and quantify long-distance relationships, the spatial aggregation of departure and arrival locations at the regional level reveals strong territorial logics (Figure 10). For example, Brittany alone accounts for more than 46% of departure cities and 47% of arrival cities, demonstrating a very strong regional dynamic in carpooling practices. By looking at regions directly neighboring Brittany, we see that 80% of journeys to and from Rennes take place between Brittany, the Pays de la Loire and Normandy. Then there are the journeys to the Ile-de-France region and, interestingly enough, the further away the other regions are, the lower the carpooling offer.

Figure 10. Distribution of trips to and from Rennes by region of origin and destination (source: based on BlaBlaCar data)

Figure 10. Distribution of trips to and from Rennes by region of origin and destination (source: based on BlaBlaCar data)
  • 11 The choice of the intermunicipal (EPCI) as the scale of analysis makes it possible to take into acc (...)

29Furthermore, the counting of trips to and from Rennes on the scale of the EPCIs (intermunicipal level) makes it possible to highlight more detailed territorial dynamics11. It appears that carpooling is mainly concentrated in the main regional urban centres around Rennes (Figure 11). The only exception is the Ile-de-France region, and in particular Paris, which polarises a large number of journeys due to its attractiveness. Interestingly, there is a clear balance in the dynamics of origin and destination of journeys.

30By focusing on the fifteen main EPCIs in connection with Rennes (which account for 50% of the published journeys), we find very localized dynamics (Figure 12). With more than 8% of all journeys transiting through Rennes, the Nantes metropolitan area is the area with the highest concentration of journeys on offer. Then there is the Grand Paris area followed by Saint-Brieuc, Vannes, Saint-Malo and Lorient. In addition to this offer to the major cities of Western France and the Paris region, there is a very limited offer of trips to major urban centres such as Bordeaux, Lyon or Lille.

Figure 11. Number of trips per EPCI originating from Rennes - number of trips per EPCI with Rennes as their destination (source: based on BlaBlaCar and OpenStreetMap data)

Figure 11. Number of trips per EPCI originating from Rennes - number of trips per EPCI with Rennes as their destination (source: based on BlaBlaCar and OpenStreetMap data)

Figure 12. Number of trips to and from the fifteen main EPCIs "connected" to Rennes (source: based on BlaBlaCar data)

Figure 12. Number of trips to and from the fifteen main EPCIs "connected" to Rennes (source: based on BlaBlaCar data)

31These results clearly illustrate that most of the routes proposed on the BlaBlaCar platform reinforce the existing dynamics around urban areas and at the same time offer only limited access to rural areas. Thus, a person who wishes to carpool via the platform will have a better chance of finding a travel opportunity if he or she wants to travel to a large or medium-sized city than if he or she wants to travel to destinations located in sparsely populated areas (Castex, 2015). For example, more than 90% of journeys arrive in or leave from a municipality in an urban area (INSEE zoning). Another interesting figure is that 57% of journeys leave or arrive within 5 km of a municipality with more than 20,000 inhabitants and this figure rises to 72% for cities with more than 10,000 inhabitants.

32This polarization of carpooling supply to medium-sized and large cities is also largely due to the spatial proximity between these cities. According to the principles of a gravity model, exchanges between two cities will be all the more important if the weight of the cities is large and if the distance between them is small. The examples of Nantes, Saint-Brieuc or Vannes illustrate this logic very well. In view of the range of journeys (frequencies, destinations and distances), occasional carpooling organised via the BlaBlaCar platform, both to and from Rennes, is thus mainly intended for journeys within an area of regional influence with one exception, Paris.

  • 12 A carpool between Rennes and Nantes costs on average seven euros and lasts about an hour, while a j (...)

33This trend towards a "regional" carpooling offer should be seen in the context of the question of complementarity with other modes of transport, particularly rail. For a passenger departing from Rennes and wishing to travel to Nantes, for example, there is certainly a train offer, but it seems uncompetitive in terms of time, cost and frequency compared to the trips offered on BlaBlaCar12. This logic is similar in the case of Saint-Brieuc, Saint-Malo, Vannes or Lorient where organised carpooling is positioned for certain categories (students, non-motorised people) as an alternative to public transport (train, bus).

Analysis of carpooling locations at the scale of the city of Rennes

34The nature of BlaBlaCar's data, through the systematic provision of geographical coordinates of departure and arrival points, also allows carpooling practices to be considered at a more detailed level. The analysis of carpooling locations at the local level is a relevant avenue to explore in order to understand carpooling practices at the scale of the city of Rennes. However, the simple spatialization of carpooling locations within a map repository is not sufficient to understand the spatial configuration of these strategic locations as illustrated in Figure 13. Indeed, at the city level, the spatial accuracy of the data is quite heterogeneous. Several journeys may have the same geographical coordinates of origin or destination because their information is provided by drivers in different ways (direct entry of the address, sharing of geolocation, choice from a list of predefined places). Visually, a large part of the carpooling areas are thus superimposed, limiting the analysis of their spatial distribution.

Figure 13. Location of carpooling places in Rennes over the course of five months (source: based on BlaBlaCar and OpenStreetMap data)

Figure 13. Location of carpooling places in Rennes over the course of five months (source: based on BlaBlaCar and OpenStreetMap data)
  • 13 Raster grid for which each pixel is assigned a measurement value corresponding to the density of po (...)

35In response to this bias, two modes of representation based on the transformation of point data are presented here. First, the conversion to a raster model through a heat map13 makes it possible to process and, above all, to visualize the concentration of carpooling areas in a simplified form. This form of representation gives these multiple data a visual aspect that is easier to capture and makes it possible to highlight the "hot spots" of carpooling in Rennes (Figure 14). The main carpooling areas highlighted by the heat map are concentrated around intermodal (metro stations, train stations) and higher education locations (universities).

Figure 14. Concentration of carpooling places in Rennes over the course of five months (source: based on data from BlaBlaCar and OpenStreetMap)

Figure 14. Concentration of carpooling places in Rennes over the course of five months (source: based on data from BlaBlaCar and OpenStreetMap)
  • 14 In a similar way to the heat map, the choice of 300 m x 300 m vectorial meshes offers a relevant ag (...)

36Another method is the spatial aggregation of the departure and destination locations within a vector grid that allows a finer reading of the issue through the spatial discretization of the study area. The counting of carpooling locations within a regular grid (300 m hexagons14) makes it possible to highlight the spatial distribution of carpooling locations. In addition to identifying areas of concentration, this approach allows for the quantification of the number of trips transiting through these areas over the study period (Figure 15).

Figure 15. Number of carpools that passed through 300 m hexagonal meshes between April 1 and August 31, 2018 (source: based on BlaBlaCar and OpenStreetMap data)

Figure 15. Number of carpools that passed through 300 m hexagonal meshes between April 1 and August 31, 2018 (source: based on BlaBlaCar and OpenStreetMap data)
  • 15 For example, 50% of car-sharing locations are located less than 300 m from a metro station and this (...)

37Two main areas of concentration of carpooling areas emerge from this analysis: “La Poterie” in the south-east of Rennes (final stop on the metro line), which accounts for more than 20% of the areas, and Villejean to the northeast (university campus), which accounts for more than 15% of the areas. Then there is the Henri Fréville shopping area, the Rennes Stadium, the station and the northern entrance to Rennes. The meeting areas favoured by BlaBlaCar users share certain common characteristics, namely the presence of public transport infrastructure (metro stations, train station, bus stops), the presence of travel generating areas (university, shopping centre, etc.) and proximity to the ring road15. The criterion of proximity to public transport infrastructure appears to be particularly important since a large proportion of passengers travel to and from carpooling areas using public transport. For example, a 2015 ADEME study on long-distance carpooling highlights that 44% of passengers reach their carpooling start location using public transport.

38It is important to note that the choice of these places by drivers who publish a trip is not made in an insignificant way. Even if they can specify any place in the municipality to pick up or drop off passengers, in most cases the selection is made via a predefined list of places considered "suitable" by the platform. This list includes the places highlighted in previous analyses, namely the train stations, universities and metro stations. It is interesting to note that the list does not suggest existing carpooling areas to drivers publishing a route. In practice, this equipment set up by the public authorities is rarely used by the platform's carpoolers (Figure 16). Located on the outskirts of cities and poorly served by public transport, these facilities, which are mainly intended for commuting carpooling, are not suitable for occasional carpooling. They essentially allow workers to park their cars for the day but are not designed to encourage load breakage between drivers and passengers, who often do not have a car.

Figure 16. Location of BlaBlaCar carpooling areas and carpooling facilities (source: based on BlaBlaCar and OpenStreetMap data)

Figure 16. Location of BlaBlaCar carpooling areas and carpooling facilities (source: based on BlaBlaCar and OpenStreetMap data)
  • 16 The parking lot of the La Poterie metro station can accommodate more than 700 trips on some Fridays (...)

39In practice, members of the BlaBlaCar platform in Rennes meet mainly in informal areas (public and private parking lots, drop-off points, bus lanes, etc.) where there are not necessarily dedicated carpooling facilities. This can lead to parking or congestion problems in areas already heavily used by pedestrians, cars or buses16. Several places are thus the scene of daily conflicts of use in Rennes; such as the multimodal hubs of La Poterie or Villejean University where cohabitation between buses and carpoolers is complicated at certain times of the day (parking within bus lanes, blocking access lanes, occupying drop-off points, etc.).

40Faced with this situation, Rennes Métropole has been engaged in discussions and actions in recent months to promote occasional carpooling. For example, the city has recently developed two carpooling areas for short-term parking near metro stations. Another initiative is the installation of a sensor at the multimodal hub of La Poterie to measure the use of the carpool area. The consideration of occasional carpooling by the public authorities is still very recent and the first specific arrangements are only just beginning to emerge. It is very likely that with the development of this form of mobility, other actions will take shape within cities to both promote the meeting between passengers and drivers and minimize traffic and parking problems related to carpooling near strategic intermodal locations.

Prospects for the future

41This exploratory analysis of the BlaBlaCar platform data allowed us to take a comprehensive look at the organised and occasional carpooling offer while highlighting a series of limitations inherent in the nature and structure of this data produced and made available by a private actor. Even if the data studied are not exhaustive (no consideration of complete trips), not necessarily very well informed (particularly with regard to journey completion rates) and, finally, not very accessible (API limits), they nevertheless open up many research perspectives in understanding carpooling.

42The study of a corpus of more than 150,000 trips over a five-month period clearly shows the potential of these digital footprints to quantitatively highlight the carpooling dynamics of a French metropolis. The different temporal and spatial modalities of the journeys highlighted in the study help to qualify and quantify the occasional organised carpooling. In addition, the data provide the possibility of changing the scale for detailed analyses of more localized phenomena, such as the spatial organization of carpooling areas throughout the city of Rennes.

43While the overview of carpooling proposed in the article only reflects the practices of BlaBlaCar members, the mobilization of this non-institutional data broadens the traditional frameworks for carpooling analysis through the accuracy and quantity of data available. Similar analyses within other metropolitan areas would make it possible to see if there are similar logics and perhaps local specializations. In addition, the implementation of studies taking into consideration a set of cities at regional or even national levels would help to highlight potential territorial connections relating to carpooling, as shown in Figure 16.

Figure 17. Travelling to and from Nantes, Paris, Bordeaux, Toulouse, Montpellier, Grenoble and Lyon on the 24th of August 2018 (source: based on BlaBlaCar and OpenStreetMap data)

Figure 17. Travelling to and from Nantes, Paris, Bordeaux, Toulouse, Montpellier, Grenoble and Lyon on the 24th of August 2018 (source: based on BlaBlaCar and OpenStreetMap data)

44As we have demonstrated throughout the article, the mobilization of these new data of a particular type requires both upstream preparation and, above all, the implementation of specific methods to make them intelligible and provide them with real analytical added value. The multiplication of new data sources from (geo)digital footprints is accompanied by new questions about their usability in research contexts. Questions of reliability, representativeness, exhaustiveness, accessibility or even data durability are at the heart of new questions for researchers. The transition from raw data to knowledge requires more than ever the use of adapted and ever more sophisticated methods and techniques, which must be explored, renewed but above all constantly questioned.

45Finally, despite their real analytical potential, these data should not be seen as substitutes for more traditional methods, such as surveys, but rather as complementary sources. It should not be forgotten that big data, from public and private actors, are produced in automated ways and exist above all to detect and formalize patterns of behaviours and categories and not to highlight specificities. Mobility as a social fact cannot only be approached by processing large volumes of standardized data, which do not make it possible to enrich analyses with more qualitative information such as the reasons for travel or the characteristics of individuals. New and more mixed methods, combining quantitative and qualitative approaches and data analysis, are thus to be explored and built in order to understand and analyse today's and tomorrow's mobility in a different way.

Haut de page

Bibliographie

ADEME, 2015, "Enquête auprès des utilisateurs du covoiturage longue distance", Étude réalisée pour le compte de l’ADEME par 6t-bureau de recherche.

ADEME, 2017, "Développement du covoiturage régulier de courte et moyenne distance : Guide méthodologique et Synthèse", Collection Clés pour agir.

Alexander L., Jiang S., Murga M., & González M. C., 2015, "Origin–destination trips by purpose and time of day inferred from mobile phone data", Transportation research part c: emerging technologies, Vol.58, 240-250.

Barry J.J., Newhouser R., Rhabee A., Sayeda S., 2002, "Origin and destination estimation in New York City with automated fare system data", Transportation Research Record, No.1817, 183-187.

Bastin G. & Tubaro P., 2018, "Le moment big data des sciences sociales", Revue française de sociologie, Vol.59, No.3, 375-394.

Beaude B., 2015, "Spatialités algorithmiques", In : Severo M. et Romele A. (dir.), Traces numériques et territoires, Paris, Presses des Mines, 135-162.

Briand A. S., Côme E., Trépanier M., & Oukhellou L., 2017, "Analyzing year-to-year changes in public transport passenger behaviour using smart card data", Transportation Research Part C: Emerging Technologies, Vol.79, 274-289.

Bonnel P., Hombourger É., Olteanu-Raimond A.M., & Smoreda Z., 2017, "Apports et limites des données passives de la téléphonie mobile pour la construction de matrices origine-destination", Revue d’Économie Régionale & Urbaine, No.4, 647-672.

Calabrese F., Diao M., Di Lorenzo G., Ferreira J., & Ratti C., 2012, "Understanding individual mobility patterns from urban sensing data: A mobile phone trace example", Transportation research part C: Emerging Technologies, Vol.26, 301-313.

Castex E., 2015, "Organisation des pratiques de covoiturage entre inconnus dans les territoires", Netcom, Vol.29, No.1/2, 153-176. URL : https://journals.openedition.org/netcom/1905

CGDD, 2016, " Covoiturage longue distance : état des lieux et potentiel de croissance", Collection Études et documents du Commissariat Général au Développement Durable (CGDD), No.146.

Chen C., Ma J., Susilo Y., Liu Y., & Wang M., 2016, "The promises of big data and small data for travel behavior (aka human mobility) analysis", Transportation research part C: Emerging Technologies, Vol.68, 285-299.

CEREMA, 2016, "Analyse économique exploratoire du covoiturage longue distance", Rapport d’étude.

Côme E., Oukhellou L., 2014, "Model-based count series clustering for bike sharing system usage mining: a case study with the Vélib’system of Paris", ACM Transactions on Intelligent Systems and Technology, Vol.5, No.3, 39-66.

Courmont A., 2018, "Plateforme, big data et recomposition du gouvernement urbain", Revue française de sociologie, Vol.59, No.3, 423-449.

Depeau S., Quesseveur E., "A la recherche d’espaces invisibles de la mobilité", Netcom, Vol.28, No. 1-2, 35-54. URL : https://journals.openedition.org/netcom/1539

Gwiazdzinski L., 2014, "Du suivi GPS des individus à une approche chronotopique", Netcom, Vol.28, No.1-2, 77-106. URL: https://journals.openedition.org/netcom/1604

Hawelka B., Sitko I., Beinat E., Sobolevsky S., Kazakopoulos P., Ratti C., 2014, "Geo-located Twitter as the proxy for global mobility patterns", Cartography and Geographic Information Science, Vol.41, No.3, 260-271.

Iqbal M. S., Choudhury C. F., Wang P., & González M.C., 2014, "Development of origin–destination matrices using mobile phone call data", Transportation Research Part C: Emerging Technologies, Vol.40, 63-74.

Jacquot M., 2018, "Quel futur pour le covoiturage ? Comment surmonter les obstacles ?", Annales des Mines-Réalités industrielles, No.2, 52-55.

Jiang S., Ferreira J., & González M. C., 2017, "Activity-based human mobility patterns inferred from mobile phone data: A case study of Singapore", IEEE Transactions on Big Data, Vol.3, No.2, 208-219.

Jurdak, R., Zhao, K., Liu, J., AbouJaoude, M., Cameron, M., & Newth, D. (2015). Understanding human mobility from Twitter, PloS one, Vol.10, No.7, URL: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.013146 9

Masure A., 2014, "Le design des programmes : des façons de faire du numérique", Thèse de doctorat, Université Paris 1, URL : http://www.softphd.com

Mericskay B., Noucher M., & Roche S., 2018, "Usages des traces numériques en géographie : potentiels heuristiques et enjeux de recherche", L'Information Géographique, Vol.82, No.2, 39-61.

Mohamed K., Côme E., Oukhellou L., & Verleysen M., 2017, "Clustering smart card data for urban mobility analysis", IEEE Transactions on Intelligent Transportation Systems, Vol.18, No.3, 712-728.

Plantin J.C., Russo F., 2016, "D’abord les données, ensuite la méthode ?", Socio, No.6, 97-115, URL : https://journals.openedition.org/socio/2328

Quesnot T., 2016, "L’involution géographique : des données géosociales aux algorithmes", Netcom, Vol.30, No.3/4, URL : http://netcom.revues.org/2545

Rashidi T. H., Abbasi A., Maghrebi M., Hasan S. & Waller T.S., 2017, "Exploring the capacity of social media data for modelling travel behaviour: Opportunities and challenges", Transportation Research Part C: Emerging Technologies, Vol.75, 197-211.

Richer C., Côme E., El Mahrsi M. K., & Oukhellou L., 2018, "La mobilité intermodale par les données billettiques. Analyses spatio-temporelles du réseau bus-métro de Rennes Métropole", Cybergeo : European Journal of Geography, Systèmes, Modélisation, Géostatistiques, article No.854 URL : https://journals.openedition.org/cybergeo/29132

Schwanen T., 2017, "Geographies of transport II: Reconciling the general and the particular", Progress in Human Geography, Vol.41, No.3, 355-364.

Senaratne H., Mueller M., Behrisch M., Lalanne, F., Bustos-Jiménez, J., Schneidewind, J., Schreck, T. 2018, "Urban Mobility Analysis With Mobile Network Data: A Visual Analytics Approach", IEEE Transactions on Intelligent Transportation Systems, Vol.19, No.5, 1537-1546.

Severo M., Romele A., 2015, Traces numériques et territoires, Paris, Presses des Mines, Territoires numériques.

Toole, J. L., Colak, S., Sturt, B., Alexander, L. P., Evsukoff, A., & González, M. C., 2015, "The path most traveled: Travel demand estimation using big data resources", Transportation Research Part C: Emerging Technologies, Vol.58, No.1, 62-177.

Vidal P., 2015, "Tracer sa route, en toute intransparence numérique ? ", Netcom, Vol.29, No.1/2, 5-12. URL: http://journals.openedition.org/netcom/1841

Vogel P., Greiser T., & Mattfeld D. C., 2011, "Understanding bike-sharing systems using data mining: Exploring activity patterns", Procedia-Social and Behavioral Sciences, Vol.20, 514-523.

Yang C., Huang Q., Li Z., Liu K., & Hu F., 2017, "Big Data and cloud computing: innovation opportunities and challenges", International Journal of Digital Earth, Vol.10, No.1, 13-53.

Wergin J., & Buehler R., 2017, "Where Do Bikeshare Bikes Actually Go? Analysis of Capital Bikeshare Trips with GPS Data", Transportation Research Record: Journal of the Transportation Research Board, No.2662, 12-21.

Zhong C., Manley E., Arisona S. M., Batty M., & Schmitt G., 2015, "Measuring variability of mobility patterns from multiday smart-card data", Journal of Computational Science, Vol.9, 125-130.

Haut de page

Notes

1 Digital footprints can be defined as footprints left behind when surfing the Web or using a service from a digital device (Severo and Romele, 2015). They are called (geo)digital, if they are associated with geographical data (or metadata) allowing their spatialization. These can be geographical coordinates, a toponym, a Wifi terminal number, a mobile phone network cell or a place name (Mericskay et al., 2018).

2 The Uber Movement service, for example, provides free access to anonymous data from more than two billion trips worldwide in the form of origin-destination time matrices.

3 The Waze application (owned by Google) offers public authorities a partnership program called Connected Citizens that allows them to access traffic data from the platform in return for information about work, road modifications or upcoming events (Courmont, 2018).

4 An API (Application Programming Interface) is a programming interface that allows you to connect one program to another. These tools allow data and functions from different sites to be reused in another project (Masure, 2014).

5 The example of mobile phone data is symptomatic of the locking systems which exist. Beyond their expensive nature, these data are also characterized by numerous constraints and limitations of use, such as their availability, their enrichment difficulties (modes, motives, characteristics of individuals), their processing difficulties (poor accuracy, not interoperable with existing systems) or the geographical edge effects related to antenna coverage (Bonnel et al., 2017).

6 Organized carpooling implies organized planning between the driver and the passenger(s) who use a connection service (free or paid). In the case of BlaBlaCar, the platform provides users with services and tools to facilitate exchanges and the organisation of journeys (search engines, messaging, secure payments, evaluation systems, etc.) in exchange for a commission on each booking.

7 API URL: https://dev.blablacar.com/hc/en-us

8 The number of complete trips to and/or from a city is displayed in each response returned by the API. The daily full trip rate varies from 20% to 40% depending on the day.

9 During the data collection phase, it was still possible to retrieve the trip history over three months. Since October 2018, the API no longer offers access to past trips.

10 The flow map creates a line between the departure and arrival point of each trip.

11 The choice of the intermunicipal (EPCI) as the scale of analysis makes it possible to take into account the areas of influence of metropolitan areas and medium-sized cities, unlike the municipal level, which would introduce many biases, such as car-sharing areas on the outskirts of cities. Within the Nantes metropolitan area, for example, 40% of journeys are made to and from Orvault, a municipality north of Nantes with access to the ring road and a tram stop. The same applies to Paris, where a large part of the journeys leave and arrive in municipalities in the inner and outer suburbs.

12 A carpool between Rennes and Nantes costs on average seven euros and lasts about an hour, while a journey by regional train costs about twenty euros and lasts between 1h15 and 2h. The average daily frequency is about fifteen direct trains per day compared to about a hundred carpooling trips with departures that can take place every ten minutes.

13 Raster grid for which each pixel is assigned a measurement value corresponding to the density of points in and around the tile. The map presented is based on a 300 m point density radius for each carpool location in a 10 m by 10 m raster grid. At city level and in view of the spatial distribution of car-sharing areas, a density radius of 300 m makes it possible to clearly highlight very localized dynamics and at the same time to maintain a synthetic global vision.

14 In a similar way to the heat map, the choice of 300 m x 300 m vectorial meshes offers a relevant aggregation scale to understand the spatial distribution dynamics of car-sharing areas.

15 For example, 50% of car-sharing locations are located less than 300 m from a metro station and this figure rises to 70% for a distance of 500 m. As for the proximity to the Rennes ring road, more than 50% of car-sharing locations are less than 500 m from the ring road and 70% less than one kilometre away.

16 The parking lot of the La Poterie metro station can accommodate more than 700 trips on some Fridays with peaks of about 100 vehicles per hour in the late afternoon.

Haut de page

Table des illustrations

Titre Figure 1: Steps of dataset preparation (source: author)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-1.png
Fichier image/png, 41k
Titre Figure 2: Summary of the attributes retrieved for each trip studied (source: author)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-2.png
Fichier image/png, 197k
Titre Figure 3: Number of daily trips over the five months of the analysis (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-3.png
Fichier image/png, 99k
Titre Figure 4. Distribution of trips by number of departures, number of arrivals and day of the week (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-4.png
Fichier image/png, 154k
Titre Figure 5: Number of trips to and from Rennes by day of the week (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-5.png
Fichier image/png, 58k
Titre Figure 6. Number of trips to and from Rennes by day of the week (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-6.png
Fichier image/png, 45k
Titre Figure 7. Number of trips to and from Rennes by time of day (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-7.png
Fichier image/png, 98k
Titre Figure 8. Travelling to and from Rennes (source: based on BlaBlaCar and OpenStreetMap data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-8.png
Fichier image/png, 1,3M
Titre Figure 9. Distances of trips between 10 km and 800 km (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-9.png
Fichier image/png, 41k
Titre Figure 10. Distribution of trips to and from Rennes by region of origin and destination (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-10.png
Fichier image/png, 86k
Titre Figure 11. Number of trips per EPCI originating from Rennes - number of trips per EPCI with Rennes as their destination (source: based on BlaBlaCar and OpenStreetMap data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-11.png
Fichier image/png, 695k
Titre Figure 12. Number of trips to and from the fifteen main EPCIs "connected" to Rennes (source: based on BlaBlaCar data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-12.png
Fichier image/png, 107k
Titre Figure 13. Location of carpooling places in Rennes over the course of five months (source: based on BlaBlaCar and OpenStreetMap data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-13.png
Fichier image/png, 1,0M
Titre Figure 14. Concentration of carpooling places in Rennes over the course of five months (source: based on data from BlaBlaCar and OpenStreetMap)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-14.png
Fichier image/png, 1,5M
Titre Figure 15. Number of carpools that passed through 300 m hexagonal meshes between April 1 and August 31, 2018 (source: based on BlaBlaCar and OpenStreetMap data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-15.png
Fichier image/png, 972k
Titre Figure 16. Location of BlaBlaCar carpooling areas and carpooling facilities (source: based on BlaBlaCar and OpenStreetMap data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-16.png
Fichier image/png, 850k
Titre Figure 17. Travelling to and from Nantes, Paris, Bordeaux, Toulouse, Montpellier, Grenoble and Lyon on the 24th of August 2018 (source: based on BlaBlaCar and OpenStreetMap data)
URL http://journals.openedition.org/cybergeo/docannexe/image/33225/img-17.jpg
Fichier image/jpeg, 250k
Haut de page

Pour citer cet article

Référence électronique

Boris Mericskay, « Potentials and limits of (geo)digital footprints in mobility analysis: the example of the data from the BlaBlaCar carpooling platform », Cybergeo : European Journal of Geography [En ligne], Cartographie, Imagerie, SIG, document 889, mis en ligne le 31 octobre 2019, consulté le 19 novembre 2019. URL : http://journals.openedition.org/cybergeo/33225 ; DOI : 10.4000/cybergeo.33225

Haut de page

Auteur

Boris Mericskay

Université Rennes 2 – UMR Espaces et Sociétés (ESO-Rennes), France
Maître de conférences en géographie
Boris.mericskay@univ-rennes2.fr
https://perso.univ-rennes2.fr/boris.mericskay

Articles du même auteur

Haut de page

Droits d’auteur

Licence Creative Commons
La revue Cybergeo est mise à disposition selon les termes de la Licence Creative Commons Attribution - Pas d'Utilisation Commerciale - Pas de Modification 3.0 non transposé.

Haut de page