Earth Observation

Buildings and roads absorb the sun's heat and give it back more than lawns and woods do. Where there is a lot of concrete and little greenery, the city becomes an island hotter than what surrounds it, by day and by night, in every season.

The stored heat escapes slowly, which is why the island is felt more after sunset. The shape of the streets matters too, since it blocks the wind, along with the heat from traffic and air conditioners. Islands exist within the same city too: a concrete neighbourhood is hotter than one with trees.

The satellite does not measure the air. It measures the surface temperature, that is the heat that roofs, asphalt, lawns and tree canopies give off upwards. It is called LST, land surface temperature.

The two temperatures can be very far apart: on a summer day a roof or asphalt in the sun can be 28 to 50 °C hotter than the air, while shaded or wet surfaces stay close to air temperature. At night the two get closer. This is why satellite data is combined with data from ground sensors.

Copernicus is the European Union's Earth observation programme, with the ESA: its data is free and open. The Sentinel-3 satellites carry SLSTR, a radiometer that measures the temperature of the land and sea surface, with two views to correct for the effect of the atmosphere.

The limit is resolution: a Sentinel-3 thermal pixel is one kilometre. That works very well for comparing a city with the countryside and for following day and night; to tell one neighbourhood from another you need Landsat's thermal channels, at 100 metres. In the project we used data from the Sentinel and Landsat missions.

The team was born in March 2025. First the theory, with weekly meetings: data analysis, machine learning and deep learning, up to foundation models for Earth observation. Then the first exercises in Python with OpenStreetMap data, for example the buildings closest to waterways in the Marassi neighbourhood.

Then the professional tools, GIS systems and Google Earth Engine, and the project on Genoa's heat islands: a platform for finding, tracking and showing the hottest areas, comparing surface temperature between parts of the city. The next planned step was models to predict heat and simulate interventions, such as more greenery and materials that reflect the sun.