A new dataset of 31 million rooftops finds nearly all of them absorb most of the sun that hits them — on every continent.
More than 4,500 years ago, the Pyramids of Giza didn’t have the sandy color they do today. They were sheathed in polished white Tura limestone, bright enough to be seen for far distances, reflecting the desert sun like massive mirrors. The Egyptians called the Great Pyramid Ikhet — the “Glorious Light.” It must have been a staggering sight.
In a way, they were the first cool roofs. What made them shine is what scientists now call albedo: the amount of sunlight a surface bounces back instead of soaking up. The builders were likrly not thinking about heat, and with millions of tons of stone the pyramids soaked up plenty of it anyway.
Today, most roofs do the opposite. Dark surfaces, rusted iron sheets, asphalt tiles dominate rooftops across much of the Global South, absorbing most of the sun’s energy and radiating it back down as heat. Cities like Ibadan, Nigeria, or Kumasi, Ghana are full of these brown-roofed neighborhoods, and the roofs make the whole area hotter. The health toll is real, especially for older people and infants: broken sleep, no relief from the heat, a higher chance of getting sick.
City planners have known the fix for a long time. Paint a roof white, or coat it with reflective material, and heat absorption drops sharply. Studies have measured roof surface temperatures falling 15°C to 30°C and indoor temperatures 2°C to 5°C. For someone without air conditioning, which is most people in the Global South, that’s the difference between an OK night and a brutal one.
But how do you find the worst roofs at the scale of a city, or a country? What pulled me down this rabbit hole was a paper and dataset from research colleagues and their collaborators that takes on exactly that question. For decades, scientists faced a trade-off: free public satellite imagery is too blurry to pick out the small, old, densely packed buildings in informal settlements — the very places that need cooling most — while the commercial imagery sharp enough to see them is expensive and patchy. The paper’s move is to fuse the two: the complete global coverage of the free data, the detail of the commercial data. It’s a similar super-resolution idea my team used on Open Buildings, an earlier project I was involved with that mapped building and road footprints across Africa but pointed at a new target. Instead of a building’s outline, it recovers the building’s thermal identity, turning blurry 10-meter blocks into rooftop-level albedo estimates as sharp as a drone flyover.
Now the data.
London, one of the wealthiest cities on the planet, in a country that pioneered the industrial revolution, sits at a median roof reflectivity of 0.10 out of 1.0. More than 95% of its rooftops are classified as “very dark”.
According to roof tops sampled in this dataset, London has only 3 cool roofs in total: two industrial buildings, and one backyard storage unit in Eastern London ( yes that tiny white rooftop in the middle below).
(Worth flagging: these are models estimates with margin of errors, a roof could’ve been repainted or sits under the cover of a tree, still just a few vs millions tells a story of its own)
Even within Africa, the range is wide. Surprisingly to me, northern Nigerian cities like Kano have naturally higher reflectivity. Light-colored metal roofing, plus dust that settles in arid climates happens to help them reflect more light. Possibly the one case where dust is actually useful.
Lagos is a coastal megacity with high humidity. Cooler during the rainy season but the rest of the year you feel the heat. It’s a daily physical tax on people who live in the densest, least served parts of the city, who are also least likely to have air conditioning or even constant electricity. Lagos has 6 cool roofs out of 955,543 buildings sampled. Another surprise here, two of them are mega churches in Victoria Island, one of the wealthiest districts in the city.
Kisumu, Kenya, a hot city right on Lake Victoria has 2 out of 147,252 sampled.
Across more than 4.5 million buildings scanned across African cities, over 90% of rooftops are absorbing at least three-quarters of incoming sunlight. Heat magnets, everywhere.
This sounds like a low-hanging fruit for policy in Africa. Ahmedabad, India proves it. Out of the 67 cities across 14 countries in this dataset, Ahmedabad stands out because of deliberate policy: a cool-roofs program launched in 2017 has measurably shifted the city’s reflectivity compared to Mumbai and Kolkata, which have no such program and absorb vastly more heat.
Looking at the whole dataset, every continent falls below the reflectivity threshold needed to meaningfully reduce the urban heat island effect. Nobody is doing well. But the consequences of not doing well aren’t evenly distributed and Ahmedabad shows that a city can decide to change this rather than just live with it.
Since looking into this research, I’ve caught myself looking at rooftops differently on my street, from a plane window, wherever. You start noticing which houses are probably the hottest.
Congrats to everyone who made this significant research possible. (See paper for full author credits.)
Note: Infographics were generated via AntiGravity, Google’s agentic development platform.