Global warming’s worst-case projections look increasingly likely, according to a new study that tested the predictive power of climate models against observations of how the atmosphere is actually behaving.
The paper, published on Wednesday in Nature, found that global temperatures could rise nearly 5 °C by the end of the century under the the UN Intergovernmental Panel on Climate Change’s steepest prediction for greenhouse-gas concentrations. That’s 15 percent hotter than the previous estimate. The odds that temperatures will increase more than 4 degrees by 2100 in this so-called “business as usual” scenario increased from 62 percent to 93 percent, according to the new analysis.
Climate models are sophisticated software simulations that assess how the climate reacts to various influences. For this study, the scientists collected more than a decade’s worth of satellite observations concerning the amount of sunlight reflected back into space by things like clouds, snow, and ice; how much infrared radiation is escaping from Earth; and the net balance between the amount of energy entering and leaving the atmosphere. Then the researchers compared that “top-of-atmosphere” data with the results of earlier climate models to determine which ones most accurately predicted what the satellites actually observed.
Sign up for Weekend Reads
Our guide to stories in the archives that put technology in perspective.
Thank you — please check your email to complete your sign up.
Incorrect email format
The simulations that turned out to most closely match real-world observations of how energy flows in and out of the climate system were the ones that predicted the most warming this century. In particular, the study found, the models projecting that clouds will allow in more radiation over time, possibly because of decreased coverage or reflectivity, “are the ones that simulate the recent past the best,” says Patrick Brown, a postdoctoral research scientist at the Carnegie Institution and lead author of the study. This cloud feedback phenomenon remains one of the greatest areas of uncertainty in climate modeling.
The UN’s seminal IPCC report relies on an assortment of models from various research institutions to estimate the broad ranges of warming likely to occur under four main emissions scenarios. In another key finding, the scientists found that the second-lowest scenario would be more likely to result in the warming previously predicted under the second-highest by 2100. In fact, the world will have to cut another 800 gigatons of carbon dioxide emissions this century for the earlier warming estimates to hold. (By way of comparison, total greenhouse-gas emissions stood at about 49 gigatons last year.)
Various politicians, fossil-fuel interest groups, and commentators have seized on the uncertainty inherent in climate models as reasons to doubt the dangers of climate change, or to argue against strong policy and mitigation responses.
Recommended for You
“This study undermines that logic,” Brown says. “There are problems with climate models, but the ones that are most accurate are the ones that produce the most warming in the future.”
In fact, the new paper is the latest in a growing series that project larger impacts than previously predicted or conclude that climate change is unfolding faster than once believed.
The goal of the research was to evaluate how well various climate models work, in hopes of “narrowing the range of model uncertainty and to assess whether the upper or low end of the range is more likely,” Brown wrote in an accompanying blog post.
Ken Caldeira, a climate researcher at Carnegie and coauthor of the paper, says the growing body of real-world evidence for climate change is helping to refine climate models while also guiding scientists toward those that increasingly appear more reliable for specific applications.
But an emerging challenge is that the climate is changing faster than the models are improving, as real-world events occur that the models didn’t predict. Notably, Arctic sea ice is melting more rapidly than the models can explain, suggesting that the simulations aren’t fully capturing certain processes.
“We’re increasingly shifting from a mode of predicting what’s going to happen to a mode of trying to explain what happened,” Caldeira says.