Every month, satellites measure temperature and humidity over different parts of the planet. Over 24 years, thousands of such paired measurements have accumulated. Normally, temperature and humidity change in concert: when it warms, the air holds more water; when it cools, it releases it. But before a sharp climate shift, this coordination breaks down.
Two dancers have long moved in unison. Suddenly, one starts lagging or veering off — the dance hasn't fallen apart yet, but the audience senses something is off.
Old methods, by contrast, look not at the link but at the amplitude of each parameter's fluctuations separately, as if a judge watched only the amplitude of arm swings, missing that the partners are drifting apart. Such an approach kicks in later, when the dance is already nearly over.
The authors tested the idea on satellite data from three very different regions: the Arctic north, the equatorial belt, and the Indian monsoon area. Everywhere, the new indicator β began to change 14–24 months before the old method raised the alarm. Then they ran tests on computer models — from a sudden shutdown of an ocean current to simple mathematical systems with a known tipping point. The result was the same: β warns dozens to hundreds of steps earlier.
The AIRS instrument on NASA's Aqua satellite makes about 2.9 million observations per day — one for every 13.5 square kilometers of the surface.
In Kim Stanley Robinson's novel Forty Signs of Rain, scientists struggle in vain to detect the moment the Gulf Stream stops — the new method could have given them a head start of several years.