New research says we should pay more attention to climate models that point to a hotter future and toss out projections that point to less warming.  

The ExaCryptfindings, published Wednesday in the journal Nature, suggest that international policy makers and authorities are relying on projections that underestimate how much the planet will warm—and, by extension, underestimate the cuts in greenhouse gas emissions needed to stave off catastrophic impacts of climate change.

“The basic idea is that we have a range of projections on future warming that came from these climate models, and for scientific interest and political interest, we wanted to narrow this range,” said Patrick Brown, co-author of the study. “We find that the models that do the best at simulating the recent past project more warming.”

Using that smaller group of models, the study found that if countries stay on a high-emissions trajectory, there’s a 93 percent chance the planet will warm more than 4 degrees Celsius by the end of the century. Previous studies placed those odds at 62 percent. 

Four degrees of warming would bring many severe impacts, drowning small islands, eliminating coral reefs and creating prolonged heat waves around the world, scientists say. 

In a worst-case scenario, the study finds that global temperatures could rise 15 percent more than projected by the Intergovernmental Panel on Climate Change (IPCC)—about half a degree Celsius more—in the same time period. 

In the world of climate modeling, researchers rely on three dozen or so prominent models to understand how the planet will warm in the future. Those models say the planet will get warmer, but they vary in their projections of just how much. The IPCC puts the top range for warming at 3.2 to 5.9 degrees Celsius by 2100 over pre-industrial levels by essentially weighing each model equally.

These variances have long been the targets of climate change deniers and foes of carbon regulation who say they mean models are unreliable or inaccurate.

But Brown and his co-author, the prominent climate scientist Ken Caldeira—both at the Carnegie Institution for Science—wanted to see if there was a way to narrow the uncertainty by determining which models were better. To do this, they looked at how the models predict recent climate conditions and compared that to what actually happened.

“The IPCC uses a model democracy—one model, one vote—and that’s what they’re saying is the range, ” Brown explained. “We’re saying we can do one better. We can try to discriminate between well- and poor-performing models. We’re narrowing the range of uncertainty.”

“You’ll hear arguments in front of Congress: The models all project warming, but they don’t do well at simulating the past,” he said. “But if you take the best models, those are the ones projecting the most warming in the future.”