How many Earths?

This interactive graphic is based on the data for candidate planets identified by NASA's Kepler Space Telescope. Kepler found these planets by recording the slight dimming of the light from a star caused by a planet passing in front of it.

About 10 per cent of the candidate planets will probably turn out to be no such thing – it's possible to mistake the second star in a binary star system for a giant planet, for example. On the other hand, Kepler probably missed around 10 per cent of the planets that passed in front of target stars because the dimming of the star's light was too slight to detect against the natural variability in the stars' light output. These two numbers roughly cancel each another out, so they are not included in our calculations.

The first step in answering "How many Earths?" was to ignore planets twice the Earth's diameter or larger: these are likely to be gas giants like Jupiter, not rocky worlds like ours. However, such planets may possess rocky moons, which could well host life.

Not all of the remaining planets will be hospitable to life. For example, carbon-rich planets could have a graphite crust with layers of diamond below and rivers of oil and tar.

Kepler could not determine a planet's composition, but to calculate how many planets might be friendly to life, we estimated the number in stars' habitable zones – orbits where a planet will be neither too hot nor too cold for water to exist in liquid form.

Defining a star's habitable zone is a complex process, but as a reasonable proxy we used Kepler's estimates of planets' equilibrium temperature. This is the temperature that would be measured at a planet's surface if it were a black body heated by its parent star without any atmospheric greenhouse effect.

The next step – the most uncertain part of our quest – was extrapolating to the total number of roughly Earth-sized planets likely to be orbiting Kepler's 150,000 target stars. Simple geometry tells us that Kepler will have missed most of these planets: the tilts of their orbits mean they never passed between their parent stars and the telescope. And the farther out a planet orbits, the harder it was for Kepler to detect.

Taking everything into account, the best estimate for the average number of roughly Earth-sized planets in each star's habitable zone is 0.15, according to simulations based on Kepler data that Courtney Dressing and David Charbonneau of the Harvard-Smithsonian Center for Astrophysics in Cambridge, Massachusetts, performed. Applying this average to Kepler's 150,000 target stars gave our estimate of 22,500 potentially habitable, roughly Earth-sized planets.

There is an important caveat, though. Dressing and Charbonneau's calculations are for class M stars, which have a reddish hue and account for about three-quarters of the stars in our galaxy. But about 80 per cent of Kepler's target stars are class G stars, like our sun, which are yellowish. Nobody knows for sure whether these different classes of stars have similar populations of planets.

The final step in our quest was to extrapolate to the entire galaxy. Estimates of the number of stars in the Milky Way vary from 100 billion to 200 billion. Applying the same estimate of 0.15 potentially Earth-like planets per star gave our figure of between 15 and 30 billion.

If we had displayed all these potential planets in the final view, the sky would have become a mass of green. To give a meaningful view for someone here on Earth, we selected stars from the European Space Agency's Tycho-2 catalogue with an apparent magnitude of 10.5 or brighter – these stars would be visible on a dark night with a good pair of binoculars. We have displayed a random sample of 15 per cent of these stars, corresponding to Dressing and Charbonneau's estimate of stars with potentially habitable, roughly Earth-sized planets.