8,400 small satellite images are used
to spot the most vibrant neighborhoods in six Italian cities
What makes for a good city street? Some urban planners would say the “fabric”: the collection of streets, blocks and buildings. In Great Streets, the urbanist Alan Jacobs compared the layout of more than 40 world cities, and found that good streets tend to have narrow lanes (making them safe from moving cars), small blocks (making them comfortable), and architecturally-rich buildings (making them interesting). Intuitively, walking down a narrow, shop-lined street is a far safer, more comfortable, and more interesting experience than walking down an arterial road between parking lots.
In The Death and Life of Great American Cities, Jane Jacobs showed that streets, like communities, thrive on vitality.
With modern technologies, high pedestrian activity is, nowadays, captured from multiple sources - from mobile phones to surveillance cameras to even social media. Unfortunately, all these sources are virtually unavailable as they are held by private companies.
What can be accessed by everyone? The sky!
Back in the 1970s, activist Jane Jacobs theorized urban vitality and found that there are four conditions required for the promotion of life in cities: diversity of land use, small block sizes, concentration of people, and mix of economic activities. Without them, a city will die. With them, it will flourish.
To build proxies for those four conditions and ultimately test Jane Jacobs’s theory at scale, researchers used to collect both private and public data from a variety of sources, and that took decades. We investigated whether it was possible to predict vitality from one, openly available source: Sentinel-2 satellite imagery.
In 2014, the European Space Agency (ESA) launched the first of the Sentinel satellites as a part of the Copernicus program, which aims at democratizing access to Earth Observation data. Every 5-7 days, satellites photograph our cities and capture changes in land use.
Roofs give away urban activity. Depending on which materials are used in different regions, residential buildings come with red roofs of terracotta tiles in Northern Bologna and Florence, or with white roof terraces of volcanic lava stone in Southern Palermo.
By contrast, industrial buildings are boring: an infinite collection of similar-looking flat metal roofs. The good news is that, because of their spectral characteristics, they are really easy to spot by satellites.
A deep-learning model extracted visual features from satellite imagery to produce each city’s vitality continuum. To see what that is, take Bologna.
One massive Sentinel-2 satellite image of Bologna was downloaded.
This massive image was divided into 620 imagelets – small images, 64×64 pixels each.
Low-vitality imagelets are green (in forests or fields) or grey (in industrial zones, transportation hubs, stadiums).
High-vitality imagelets are red - Bologna La Rossa.
The deep-learning model was then trained with a dataset combining these imagelets with vitality statistics (extracted from mobile phone data).
Finally, the model was able to accurately predict vitality for unseen imagelets, and show them in a “vitality continuum”.
Each of the six cities had a distinct and unique continuum, making it its DNA. Interestingly, across cities, forests were not always at the beginning of the continuum. By clicking on the buttons, you can see that each urban element from industrial zones to residential areas occupied different positions across the six DNAs.
Imagelets can be placed not only in a continuum but also on a map to spot the vibrant parts of the city. On this pointillist map, each dot comes with the dominant color of its area’s satellite imagelet.
City name
Jane Jacobs in the Sky: Predicting Urban Vitality with Open Satellite Data
CSCW 2021 | Paper | Talk
London Data Visualisation 2023 | Talk
Team: Sanja Šćepanović, Sagar Joglekar, Stephen Law, Daniele Quercia
Data visualization: Edyta Bogucka
