The PETM doesn’t only provide a past example of CO₂-driven climate change; scientists say it also points to an unknown factor that has an outsize influence on Earth’s climate. When the planet got hot, it got really hot. Ancient warming episodes like the PETM were always far more extreme than theoretical models of the climate suggest they should have been. Even after accounting for differences in geography, ocean currents and vegetation during these past episodes, paleoclimatologists find that something big appears to be missing from their models — an X-factor whose wild swings leave no trace in the fossil record.

Evidence is mounting in favor of the answer that experts have long suspected but have only recently been capable of exploring in detail. “It’s quite clear at this point that the answer is clouds,” said Matt Huber, a paleoclimate modeler at Purdue University.

Clouds currently cover about two-thirds of the planet at any moment. But computer simulations of clouds have begun to suggest that as the Earth warms, clouds become scarcer. With fewer white surfaces reflecting sunlight back to space, the Earth gets even warmer, leading to more cloud loss. This feedback loop causes warming to spiral out of control.

For decades, rough calculations have suggested that cloud loss could significantly impact climate, but this concern remained speculative until the last few years, when observations and simulations of clouds improved to the point where researchers could amass convincing evidence.

Now, new findings reported today in the journal Nature Geoscience make the case that the effects of cloud loss are dramatic enough to explain ancient warming episodes like the PETM — and to precipitate future disaster. Climate physicists at the California Institute of Technology performed a state-of-the-art simulation of stratocumulus clouds, the low-lying, blankety kind that have by far the largest cooling effect on the planet. The simulation revealed a tipping point: a level of warming at which stratocumulus clouds break up altogether. The disappearance occurs when the concentration of CO₂ in the simulated atmosphere reaches 1,200 parts per million — a level that fossil fuel burning could push us past in about a century, under “business-as-usual” emissions scenarios. In the simulation, when the tipping point is breached, Earth’s temperature soars 8 degrees Celsius, in addition to the 4 degrees of warming or more caused by the CO₂ directly.

Once clouds go away, the simulated climate “goes over a cliff,” said Kerry Emanuel, a climate scientist at the Massachusetts Institute of Technology. A leading authority on atmospheric physics, Emanuel called the new findings “very plausible,” though, as he noted, scientists must now make an effort to independently replicate the work.

To imagine 12 degrees of warming, think of crocodiles swimming in the Arctic and of the scorched, mostly lifeless equatorial regions during the PETM. If carbon emissions aren’t curbed quickly enough and the tipping point is breached, “that would be truly devastating climate change,” said Caltech’s Tapio Schneider, who performed the new simulation with Colleen Kaul and Kyle Pressel.

Huber said the stratocumulus tipping point helps explain the volatility that’s evident in the paleoclimate record. He thinks it might be one of many unknown instabilities in Earth’s climate. “Schneider and co-authors have cracked open Pandora’s box of potential climate surprises,” he said, adding that, as the mechanisms behind vanishing clouds become clear, “all of a sudden this enormous sensitivity that is apparent from past climates isn’t something that’s just in the past. It becomes a vision of the future.”

The Cloud Question

Clouds come in diverse shapes — sky-filling stratus, popcorn-puff cumulus, wispy cirrus, anvil-shaped nimbus and hybrids thereof — and span many physical scales. Made of microscopic droplets, they measure miles across and, collectively, cover most of the Earth’s surface. By blocking sunlight from reaching the surface, clouds cool the planet by several crucial degrees. And yet, they are insubstantial, woven into greatness by complicated physics. If the planet’s patchy white veil of clouds descended to the ground, it would make a watery sheen no thicker than a hair.

Clouds seem simple at first: They form when warm, humid air rises and cools. The water vapor in the air condenses around dust grains, sea salt or other particles, forming droplets of liquid water or ice — “cloud droplets.” But the picture grows increasingly complicated as heat, evaporation, turbulence, radiation, wind, geography and myriad other factors come into play.

Physicists have struggled since the 1960s to understand how global warming will affect the many different kinds of clouds, and how that will influence global warming in turn. For decades, clouds have been seen as by far the biggest source of uncertainty over how severe global warming will be — other than what society will do to reduce carbon emissions.

Kate Marvel contemplates the cloud question at the NASA Goddard Institute for Space Studies in New York City. Last spring, in her office several floors above Tom’s Restaurant on the Upper West Side, Marvel, wearing a cloud-patterned scarf, pointed to a plot showing the range of predictions made by different global climate models. The 30 or so models, run by climate research centers around the world, program in all the known factors to predict how much Earth’s temperature will increase as the CO₂ level ticks up.

Each climate model solves a set of equations on a spherical grid representing Earth’s atmosphere. A supercomputer is used to evolve the grid of solutions forward in time, indicating how air and heat flow through each of the grid cells and circulate around the planet. By adding carbon dioxide and other heat-trapping greenhouse gases to the simulated atmosphere and seeing what happens, scientists can predict Earth’s climate response. All the climate models include Earth’s ocean and wind currents and incorporate most of the important climate feedback loops, like the melting of the polar ice caps and the rise in humidity, which both exacerbate global warming. The models agree about most factors but differ greatly in how they try to represent clouds.

The least sensitive climate models, which predict the mildest reaction to increasing CO₂, find that Earth will warm 2 degrees Celsius if the atmospheric CO₂ concentration doubles relative to preindustrial times, which is currently on track to happen by about 2050. (The CO₂ concentration was 280 parts per million before fossil fuel burning began, and it’s above 410 ppm now. So far, the average global temperature has risen 1 degree Celsius.) But the 2-degree prediction is the best-case scenario. “The thing that really freaks people out is this upper end here,” Marvel said, indicating projections of 4 or 5 degrees of warming in response to the doubling of CO₂. “To put that in context, the difference between now and the last ice age was 4.5 degrees.”

The huge range in the models’ predictions chiefly comes down to whether they see clouds blocking more or less sunlight in the future. As Marvel put it, “You can fairly confidently say that the model spread in climate sensitivity is basically just a model spread in what clouds are going to do.”

The problem is that, in computer simulations of the global climate, today’s supercomputers cannot resolve grid cells that are smaller than about 100 kilometers by 100 kilometers in area. But clouds are often no more than a few kilometers across. Physicists therefore have to simplify or “parameterize” clouds in their global models, assigning an overall level of cloudiness to each grid cell based on other properties, like temperature and humidity.

But clouds involve the interplay of so many mechanisms that it’s not obvious how best to parameterize them. The warming of the Earth and sky strengthens some mechanisms involved in cloud formation, while also fueling other forces that break clouds up. Global climate models that predict 2 degrees of warming in response to doubling CO₂ generally also see little or no change in cloudiness. Models that project a rise of 4 or more degrees forecast fewer clouds in the coming decades.

The climatologist Michael Mann, director of the Earth System Science Center at Pennsylvania State University, said that even 2 degrees of warming will cause “considerable loss of life and suffering.” He said it will kill coral reefs whose fish feed millions, while also elevating the risk of damaging floods, wildfires, droughts, heat waves, and hurricanes and causing “several feet of sea-level rise and threats to the world’s low-lying island nations and coastal cities.”

At the 4-degree end of the range, we would see not only “the destruction of the world’s coral reefs, massive loss of animal species, and catastrophic extreme weather events,” Mann said, but also “meters of sea-level rise that would challenge our capacity for adaptation. It would mean the end of human civilization in its current form.”

It is difficult to imagine what might happen if, a century or more from now, stratocumulus clouds were to suddenly disappear altogether, initiating something like an 8-degree jump on top of the warming that will already have occurred. “I hope we’ll never get there,” Tapio Schneider said in his Pasadena office last year.

The Simulated Sky

In the last decade, advances in supercomputing power and new observations of actual clouds have attracted dozens of researchers like Schneider to the problem of global warming’s X-factor. Researchers are now able to model cloud dynamics at high resolution, generating patches of simulated clouds that closely match real ones. This has allowed them to see what happens when they crank up the CO₂.

First, physicists came to grips with high clouds — the icy, wispy ones like cirrus clouds that are miles high. By 2010, work by Mark Zelinka of Lawrence Livermore National Laboratory and others convincingly showed that as Earth warms, high clouds will move higher in the sky and also shift toward higher latitudes, where they won’t block as much direct sunlight as they do nearer the equator. This is expected to slightly exacerbate warming, and all global climate models have integrated this effect.

But vastly more important and more challenging than high clouds are the low, thick, turbulent ones — especially the stratocumulus variety. Bright-white sheets of stratocumulus cover a quarter of the ocean, reflecting 30 to 70 percent of the sunlight that would otherwise be absorbed by the dark waves below. Simulating stratocumulus clouds requires immense computing power because they contain turbulent eddies of all sizes.