Rethinking mobility in the spring of 2020 means having the worldwide crisis triggered by the coronavirus pandemic in front of our eyes. The need for physical distance between individuals is a variable that had never before entered the manuals of transport, architecture, urban planning, or work organization. So far, the variable “occupation of a means of transport” has been taken into account only to dimension the transport offer. Nowadays, the meaning of this constraint is changing radically, encompassing the meaning of the terms “safety” and “security” when it is associated with transport phenomena. The term security now also acquires the meaning of “possibility of minimising the risk of physical contact with potentially infected individuals”. Public health security is a game-changer for mobility ecosystems. Rethinking mobility in the spring of 2020 also means conceiving a “Postcovidic” era where the theme of movement will be entangled with public health security.
Here, we give a first contribution to this topic by introducing the Safety Distancing Index (SDI), an index meant to quantify the risk level while traveling on public transports. Public transportation is presumably one of the urban systems most affected by the safety distancing constraints, and the need to reorganize transport services to face the current crisis is a challenging tasks that institutions are currently tackling, together with the governance of the potentially massive shift in individual habits. To an extent, these new habits might be in contrast with the longstanding aims of reaching sustainable and equitable cities. For example, think aboute the possible re-adoption of private means of transport by citizens, the spread of cities at much lower densities exacerbating urban sprawl, and the attraction of living in small towns rather than large cities. Thus, rethinking mobility ecosystems in the spring 2020 means to be able to control and forecast a variable, that of the occupation density, that will allow for a careful planning of the transport offer.
To define the SDI, we need to shift the attention from average fluxes of people or average occupation numbers to actual occupation numbers. If the maximal capacity of a bus is 100 passengers and in one hour the bus runs 20 times carrying 500 passengers, we can easily conclude that there are, on average, 25 passengers per bus. And 25 passengers can be a safety occupation within a bus. We know very well that fluctuations are always there and we can carry the same number of passengers in a myriad of different configurations on the 20 buses. In this specific case, there is only one safe configuration, i.e., the one where all buses carry 25 or less passengers. In all the other cases (and there are many of them, roughly 10^35 combinations), there will always be at least one bus with more than 25 passengers, leading to a potentially unsafe condition. In other words, to correctly compute the actual occupation number we need to know who is on a specific transport mean at any given time.
This quantification requires either accurate and finely tuned models or the merging of information coming from a lot of different data sources. Being involved in the planning of the reopening phase after the lockdown in the city of Rome (some reports developed in collaboration with Roma Servizi per la Mobilità and Roma 3 University are available here), we decided to leverage on the second approach. Data-Driven approaches are powerful and relatively new ways to develop models that allow us to realize accurate predictions using a minimal amount of arbitrary modelling.
Hence, we used several data source about the Metropolitan Area of Rome:
High-Precision Anonymized Location-Based (HFLB) de-identified data collected via mobile phones app. This data was provided in the context of CUEBIQ’s Data for Good Program. We used this data also in another post, to study mobility patterns during the Italian lockdown. Here, we used this data to have an estimate of the time of the day anonymous individuals move between different parts of the Metropolitan Area. This first-party data is collected from anonymized users who have opted-in to provide access to their location data anonymously, through a GDPR-compliant framework
Census data about population and economic activities in different parts of the Metropolitan Area. This information, together with HFLB data, was used to build an Origin-Destination Matrix, quantifying the fluxes of individuals within the city during the day.
Public transportation data in General Transit Feed Format (published here by Roma Servizi per la Mobilità). This data allowed us to reconstruct the trajectories individuals take on public transport whenever they travel within the city. Fluxes in the town also come from the surrounding areas. We enriched this data with information about buses travelling outside the city. These buses are typically used by commuters from surrounding towns during their commuting to work or study.
Tickets validation at metro stations within Rome. This information unlocks the possibility to calibrate the fluxes on the metro-lines in the city, which represent the main arteries of public transport.
The model obtained by merging all the data mentioned above allows for a quantification of the number of passengers travelling ay the same time on the same mean of transport (bus, metro line, tram, train). We can use this information to define the metrics of safety that we name Safety Distancing Index (SDI). The SDI for a specific mean of transport is defined as the fraction of travelling time an individual on that mean of transport spends in unsafe conditions. The definition of “unsafe” condition depends, in turn, on the transport means capacity. For our case-study of mobility in the metropolitan area of Rome, we adopted the guidelines of the Rome City council, that is:
A bus is unsafe if the number of passengers is higher than 25
A metro train is unsafe if the number of passengers is higher than 150
A tram is unsafe if the number of passengers is higher than 50
A number of passengers higher than the values above for the corresponding type of mean of transport will likely result in the impossibility to ensure physical distancing, leading to unsafe conditions for the passengers.
Given the definition above, we can easily map the SDI for the metro lines of Rome during a generic day, resulting in a simple and informative way to understand what might happen in different scenarios. The map we report below refers to the forecast provided to Roma Servizi per la Mobilità for the 18th of May 2020, the day when many restrictions to economic activities and individual movements were relaxed in Italy. We can see that, accordingly to our forecast, travelling on the metro line was unsafe during peak hours in the morning (8 am - 9 am).
Of course, understanding metro lines criticalities is a task of utmost importance to ensure the safety of citizen. However, another equally important goal is to look at surface transportation means, e.g., buses and tramways, which count around 900 distinct lines in Rome. Given the large number of lines and trips, it is crucial to comprehensively visualize the criticalities of surface transportation on a fairly readable map. The map below shows these criticalities on May the 18th 2020. The high number of different buses, surface trains and tramway lines do not allow for the SDI to be displayed for each path. Here, we show all the bus stops in the metropolitan area. The colour of each stop represents the fraction of unsafe buses/trains/trams travelling towards it. Here, again, an unsafe bus is a one carrying more passengers than the threshold to ensure adequate physical distancing. We call this index Fraction of Incoming Unsafe Connections (IUC).
Total safety during daily travels will be hardly reachable in a nearby future. However, we believe that tools like the ones we are developing at Sony CSL Paris can help to improve the current situation and support the transition towards new, safer mobility. Stay tuned for other updates about our work, and contact us at info@csl.sony.fr for more information.
We would like to thank Stefano Brinchi (President and CEO of Roma Servizi per la Mobilità, RSM), Marco Cianfano (RSM), Umberto Crisalli (“Tor Vergata” University of Rome), Gaetano Fusco (Sapienza University of Rome) and Ernesto Cipriani (Università degli Studi Roma Tre) for sharing data as well as for many interesting discussions in the framework of the Task Force of Roma Servizi per la Mobilità devoted to the COVOD-19 emergency. We would like to thank CUEBIQ for sharing HFLB data in the context of their Data for Good Program (https://www.cuebiq.com/about/data-for-good/)