The thermostat nobody set

Over the three or four billion years that Earth has been continuously habitable, the Sun's output has risen by about 30%, enough to sterilise the planet if nothing had pushed back.¹ Earth's climate is also balanced finely enough that it could slide into a deep freeze within as little as a million years.¹ Something has held it inside a narrow band for a very long time.

The question is whether life is part of that something, and whether a planet that keeps itself fit for life is the same as a planet trying to. The evidence says yes to the first, in a large way, and no to the second, in a way that matters. Life sets much of the thermostat. Nothing in the thermostat is aiming at anything, and the lever life pulls to keep Earth cool has also nearly frozen it.

The best-known stabiliser does not need life. In 1981 James Walker and two colleagues proposed that the weathering of silicate rock, which draws carbon dioxide out of the air and ends up locked in carbonate minerals, depends on surface temperature, while surface temperature depends on carbon dioxide through the greenhouse effect.² A warmer planet removes more of the gas and cools; a cooler one removes less and warms. The authors called the quantitative details speculative and said the mechanism appeared able "partially to stabilize" the surface temperature against the brightening Sun.² Doubts have grown since: Toby Tyrrell notes that weathering rates seem not to have fallen as the climate cooled gradually through the Cenozoic, and that recovery from one intense warming has been attributed to the burial of organic carbon rather than to weathering.¹

Life's hand on the dial

Weathering is also where life does much of its work. Living things speed it up, directly and through the soils they build.³ In 1989 David Schwartzman and Tyler Volk estimated what that is worth: if weathering today runs 10, 100 or 1,000 times faster than it would on a lifeless Earth, then an abiotic Earth would be about 15, 30 or 45 °C warmer than today, and they preferred the higher two figures.³ On that reading, Earth without its biota would be "uninhabitable for nearly all but the most primitive microbes."³ In 2014 Schwartzman and Volk reported a meta-analysis that put today's biological enhancement of weathering at roughly two orders of magnitude.⁴

Not every proposed biological thermostat has survived. In the CLAW hypothesis, sulphur gas from ocean plankton formed particles that seed clouds, and the clouds' effect on temperature and light fed back on how much gas the plankton made.⁵ After two decades of observation, laboratory work and modelling, a 2011 review concluded that such a biological control over cloud seeds "probably does not exist."⁵

A lever that cools can also over-cool. Tyrrell counts among the shocks that life itself has delivered the extreme glaciations "triggered by the advent of oxygenic photosynthesis and the spread of the first forests across the land."¹ The Snowball Earth episodes, he writes, may have come very close to extinguishing all life.¹ Schwartzman and Volk argue that the lever also limits itself: much stronger biological weathering would pull carbon dioxide down toward the lower limit for photosynthesis, and productivity, and with it weathering, would fall back.⁴ The thermostat and the near misses are the same machinery.

No set point

None of this needs a planet that wants anything. In 1983 Andrew Watson and James Lovelock built a deliberately simple imaginary planet, Daisyworld, whose only plants were two species of daisy, one darker than bare ground and one lighter.⁶ As its sun brightened, the mix of daisies shifted, and after a disturbance the daisies restored their local temperatures to fixed values "despite the fact that no physically real reference temperature exists within the system."⁶ Regulation appeared with no target anywhere in it. Schwartzman and Volk said the same of the real Earth: its earliest life "did not optimize conditions for its existence, but was a crucial factor in creating conditions for low-temperature life to emerge and thrive."³

The stronger claim, that life keeps a planet habitable as if on its own behalf, has fared worse. Tyrrell writes that the Gaia hypothesis "lacks a clear mechanistic basis" and is not consistent with the most recent evidence.¹

The runs we cannot see

There is another way to get a long habitable history without anything regulating it well: we could only ever find ourselves on a planet where it happened.¹ Tyrrell tested how much that might matter. He gave 100,000 simulated planets random climate feedbacks and ran each one 100 times for three billion years, with different random shocks each time, standing in for supervolcanoes and asteroid impacts.¹ About 9% of the planets, 8,710, stayed habitable at least once, and one planet stayed habitable on all 100 runs.¹ A planet that stayed habitable on one run had, on average, a 39% chance of doing it again.¹ The model is a sketch of planets in general, not of Earth, and it counts temperature alone.¹

If Earth resembles the planets in it, most reruns of its history would have ended too hot or too cold, with life's feedbacks part of the story either way. Those runs are invisible to us. A planet that left the band has no one left to study it, which is part of why the band looks guarded.