Could the sheets of gray clouds that hang low over the ocean disappear suddenly in a warming world? Yes, if you believe a study published yesterday in Nature Geoscience—and the amplifying media coverage of it. If atmospheric carbon dioxide (CO2) levels triple—an unlikely, but not implausible scenario given past rates of human emissions—these stratocumulus clouds could vanish in a frightening feedback loop. Fewer of the cooling clouds would mean a warmer Earth, which in turn would mean fewer clouds, leading to an 8°C jump in warming—a staggering, world-altering change.
But many climate scientists who research clouds are pushing back against the study, arguing that its analysis of one small patch of atmosphere does not apply to the entire globe. It’s a “simple model [that] essentially has a knob with two settings,” says Joel Norris, a cloud scientist at the Scripps Institution of Oceanography in San Diego, California. “But it is very likely that the Earth has more knobs than that.”
As sophisticated as they are, climate models have a hard time dealing with clouds. Condensing moisture and turbulent air form clouds at scales smaller than models can directly simulate, so instead they use approximations for this behavior. To understand clouds better, scientists have instead developed high-resolution eddy simulations, which re-create the life of small parcels of the atmosphere, including key physics of cloud formation that climate models can’t handle directly.
Several years ago, a project comparing six leading eddy simulations looked at how just a 2°C temperature rise influenced low ocean clouds. Two dynamics emerged that caused the clouds to thin, exacerbating warming. First, higher temperatures allowed more dry air to penetrate thin clouds from above, preventing them from thickening and reflecting more of the sun’s energy. Second, increased CO2 levels trapped heat near the cloud tops, preventing their cooling. Because such cooling drives the turbulence that forms clouds, the effect could impede cloud formation, fueling further warming. If emissions continued, it seemed plausible that these low clouds would melt away.
The frustration with how poorly global models handle clouds was a primary reason that Tapio Schneider, a climate dynamicist at the California Institute of Technology (Caltech) in Pasadena and the new study’s lead author, began construction of a new climate model last year. Dubbed the Climate Machine, it would use artificial intelligence to learn from eddy simulations and satellite observations to improve its rendering of clouds. Doing so first meant building, with his team, their own eddy simulation, one that could dynamically interact, or couple, with the ocean, allowing the simulated clouds to spur warming and vice versa.
The new study, which uses this eddy simulation, shows the same feedbacks that others had previously identified. But Schneider ran it for much higher CO2 concentrations than most had done. As levels reached 1200 parts per million—three times what they are today, and a number that could be reached next century if no effort is made to stop climate change—the low cloud decks rapidly withered away.
The model results themselves look solid, if not particularly novel. Several cloud scientists, however, object to the next step Schneider took: extrapolating the results of his eddy simulation, which represents only one spot that seems prone to cloud loss, to every area with similar stratocumulus cloud decks. Doing so resulted in all of these clouds disappearing nearly at once, allowing much more of the sun’s energy to suddenly be absorbed by the dark ocean. It’s a stretch to think the clouds and ocean would link together in such a simple way, says Bjorn Stevens, a climate scientist at the Max Planck Institute for Meteorology in Hamburg, Germany. “This coupling is done in a manner which does not give one confidence in the result.”
There’s no doubt these feedbacks will be in play. Past work has shown it, says Chris Bretherton, a cloud scientist at the University of Washington in Seattle. “But they’d all happen at different times in different concentrations of CO2 in different places. That would smooth it all out.” There wouldn’t be a sudden tipping point where all the clouds disappeared. It would happen gradually, subject to the complex response of the ocean and atmosphere. “That’s where I take issue with this,” Bretherton says. “I think the tipping point is not right.”
Indeed, the new model is so simple, lacking things such as the noise of weather, that it can only simulate rapid transitions, adds Stephen Klein, an atmospheric scientist at Lawrence Livermore National Laboratory in California. “Because of those simplifications, I don’t find the ‘tipping point’ nature of their work to be believable.”
Schneider stands by his interpretation. “I looked for all possible reasons to be wrong but ran out of them,” he says. The main implication, he adds, is that climate models need to be better equipped to handle clouds. “We shouldn’t be complacent about trusting models to predict the future into the 22nd century. There could be other things that models don’t quite capture.”
Bretherton says more cloud-resolving models are on their way. “Within the next few years, we will have global models that will do what [Schneider’s] does in a more defensible way.” Bretherton is the midst of developing such a model himself, which also relies on eddy simulations to power its simulations. To his surprise, he added, initial runs seemed to suppress the warming feedbacks for these clouds more than expected.
The Caltech climate model, meanwhile, will take another few years to come together. But it’s no coincidence that Schneider began to push to develop the model once, 2 years ago, he witnessed his eddy simulation eliminating clouds.
It will be an interesting test to see whether that tendency extends to the Climate Machine he’s developing, adds Matthew Huber, a paleoclimatologist at Purdue University in West Lafayette, Indiana. The global model might catch this type of dynamic—or it could show that the climate system overall somehow buffers such “tippiness” at smaller scales out of its system. “That is indeed the only reason to develop this new model,” he says, “to predict climate surprises.”