SafeCitymap is a data-driven project investigating how individuals' mobility patterns at a metropolitan scale were affected by the Covid-19 pandemic, and especially by the harsh French lockdown conditions enforced from March 17, 2020 to May 11, 2020 (i.e., two weeks before and during the first lockdown). For this, we used spatiotemporal aggregated mobile phone data provided by SFR, a major SFR French telecom operator, covering a geographical region focused on the city of Paris.
An essential property of this data is its fine-grained spatial resolution, which, to the best of our knowledge, is unique in the COVID-related mobility literature. Contrary, to regions or country-wide resolution, the data describes population mobility flows among zones ranging from 0.025 km2 to 5.40 km2, corresponding to 326 aggregated zones over the total area of 93.76 km2 of the city of Paris.
Our goal is to quantify (in space and time) two phenomena: (1) the attendance of and (2) the visiting flows in different urban areas, both before and during the lockdown, so as to quantify the consequences of mobility restrictions and decisions at a urban scale.

Potential Impact and Applicability

The sanitary measures forced drastic changes on population mobility and the dataset we explore in this project allows us to quantify their impact. The mobility data we use are unique in the Covid-19 related literature, to the best of our knowledge, distinguishing themselves through two important properties.

  • The dataset presents a spatial resolution describing population attendance and mobility over small geographical zones and covering overall a significant area in a big European metropolis.
  • The dataset, collected by SFR, a major French mobile operator, temporally covers two different contexts, describing a "usual" and a "radically locked down" urban life. From a human mobility study perspective, this represents a rare opportunity, as a radical change in population mobility habits of an important EU capital can not be planed for research purposes.
Note that the literature on human mobility focused on very big planned events, such as the Olympic Games or the Football World Cup, which induce flow increase in some particular areas of the city (e.g. airports, touristic spots, stadiums, and sports centers), while the large part of the local population pursues a "normal" urban life, with maybe an adaptation of some transit zones. On the other hand, when crisis situations are studied, it is usually too late to retrieve "normal" mobility information about the area of interest.
In our case, the entire population is impacted and the dataset also covers the period immediately before the lockdown, allowing for an original case of study.


  • Conduct a detailed spatio-temporal analysis of human mobility during the lockdown.
  • Automatically infer land use and classify each zone in Paris as either residential, activity, or outlier area, allowing the observation of how the usage of any given area changes once the lockdown is established.
  • Model mobility flows using a time-varying weighted mobility graph and study three different types of graph centrality, which quantify the importance of each zone in the city according to the habits in mobility of the daily habits in mobility of the population::
    • the betweenness centrality captures the paths preferences of people, e.g., shortest or less congested paths
    • the closeness centrality captures the locality of people movement
    • the degree centrality captures topologically central hubs
  • Combine the three centralities and the daily human density of an area into one metric, named impact factor, quantifying the global importance of zones in terms of frequentation and occupation, according to the mobility behavior of the population.
  • Study the correlation between the impact factor and the type of area (residential, activity, other).


  • We associate two signatures to each SFR-IRIS, one signature for the period before the lockdown and a second period for the period during the lockdown.
  • We use the Pearson correlation coefficient two measure the similarity between any two signatures in the dataset.
  • By using an unsupervised machine learning approach, we classify the signatures in the dataset. Each class has a characteristic signature, obtained as the mean of all the signatures in that class.
  • We label each class of signatures obtained in the previous step using information from municipality surveys produced before the lockdown.
  • We create a time-dependent weighted graph to represent the interaction between flows and SFR-IRIS.
  • We use graph centrality measures to quantify the importance of each SFR-IRIS in the dataset in terms of different perspectives related to the way they are visited or frequented by the population.
  • We measure the impact-factor of each SFR-IRIS by combining results of their centrality metrics and attendance, thus defining a specific mobility-related metric for each area in the city of Paris.


We use a privacy-compliant mobile phone dataset provided by the French telecom operator SFR, covering 20% of the entire population of the Paris area and for two time periods in early 2020, before and during the first French lockdown.

What we do

We performed a detailed spatiotemporal analysis of human mobility before and during the French lockdown. For this analysis, we used and developed several methods which are detailed in the following results:


We show statistics for the datasets used in this project

Learn more


We detected the major types of land-use through time series signatures

Learn more


We analyze and capture the interaction between people and the urban scenario during the time through centrality metrics

Learn more


We check whether a certain correlation exists between impact-factor and land-use.

Learn more

Interactive Metrics Visualization

To visualize the different mobility metrics used in this project, we created an interactive page that makes this comparison possible.

Impact-factor Visualization

To visualize the impact-factor, a combination of the metrics (betweenness, closeness, density and degree), we created an interactive page.

Who we are

Partner Institutions


For questions or more information please contact: