Climate-Driven Changes in Clouds are Likely to Amplify Global Warming
New research, using machine learning, helps project how the buildup of greenhouse gases will change clouds in ways that further heat the planet.
It’s an important question, because clouds have been the main source of uncertainty in projecting just how sensitive the climate is to increasing greenhouse gas concentrations, and because clouds have a huge effect on the climate system. Just a 20 percent change in their extent or reflectivity would have more of an impact than all the greenhouse gases released by human activities.
A new study published today in the Proceedings of the National Academy of Sciences may help find an answer. The researchers analyzed 20 years of cloud data from satellites and found that it was 97.5 percent certain that changes in clouds brought about by climate change will amplify warming.
Since the cloud effect has been uncertain, its accurate measurement also helps affirm other recent projections that a doubling of carbon dioxide in the atmosphere will warm the planet’s surface by about 5.8 degrees Fahrenheit, said said co-author Paulo Ceppi, a climate scientist with the Grantham Research Institute on Climate Change at Imperial College London.
“Most previous cloud studies focused only on certain regions or regimes, so say they look at places where there are low clouds and they look at low clouds only,” he said. “We did this analysis everywhere, at every point regardless of what type of cloud was there, and that allowed us to get a global picture.”
The new research is an important update to the scientific understanding of clouds in the climate system, said Piers Forster, director of the Priestley Centre at Leeds University.
“It is a really good step forward,” said Forster, who was not involved in the new study, but has worked on other recent research assessing the climate system’s response to building greenhouse gas levels.
“It really tells us how clouds respond to changes in local surface temperature, especially the reflectance of low clouds,” he said. “This is then used to make an accurate estimate of the total cloud feedback: the amplifying effect that clouds have on global warming.”
To get a sense of how important clouds are in the global warming equation, Ceppi said their effects can be compared to the warming effect of carbon dioxide.
“We calculate that, on average globally, clouds reflect something like 50 watts per square meter of solar radiation,” he said. “You can compare that to the forcing from a doubling of CO2, which would be about 4 watts per square meter, much smaller than the average effect of clouds on sunlight. So even a very small change in how much sunlight is reflected by clouds would be comparable to the effect of a CO2 doubling.”
In general, the new research confirms what some of those other studies have suggested, he said.
“People have argued that clouds will amplify global warming because of solar impacts, so less reflected sunlight from low clouds, but also because of the greenhouse effect of clouds, where high clouds rise, which makes them have a larger warming effect,” he said. “Our study finds evidence of both. I’m not aware of any other studies that have been able to show that, especially the greenhouse part.”
One recent study, led by University of Oslo researchers, shows global warming will reduce the amount of ice particles in widespread low clouds around Antarctica that currently reflect a huge amount of solar radiation back into space. That would make the clouds less reflective and amplify global warming, said cloud researcher Trude Storelvmo.
Machine Learning
Ceppi said using a machine learning approach is especially suited for complex problems like cloud changes.
“It’s a complex situation because clouds depend on so many factors that all co-vary.
For example, for a certain change in humidity, you get a certain response from clouds,” he said. “The machine learning method we use is smarter about learning these dependencies. It’s a complex statistical problem, and improved statistical methods can really help. There are so many relationships that it’s hard to calculate them manually. The statistical learning step gives us better predictive power.”
Prior studies showed less strong relationships and thus came up with less reliable projections, he added.
“One strength of our study is that we show, with 20 years of data from observations, we can really predict the feedback in model worlds where we know the answers,” he said. “Our results will mean we are more confident in climate projections and we can get a clearer picture of the severity of future climate change. This should help us know our limits and take action to stay within them.”
While the research helps narrow the range of cloud responses and feedback to global warming, some uncertainties remain.
“I would like to see a physical process understanding of how clouds respond,” Forster said. “This would add confidence that they are looking at the right statistics. It’s really about how much low clouds reflect sunlight in relation to both the local surface temperature and how quickly the temperature drops with altitude. Both of these temperatures are affected by global warming.”
“Understanding how clouds respond locally to these temperatures,” he said, “builds up a complete picture of how clouds respond to global warming, and thereby how much global warming we expect from increasing levels of CO2.”