A study suggests that Arctic precipitation will become dominant in the 2060s, decades earlier than expected. Another says air pollution from wildfires in the western United States could triple by 2100. A third says a mass ocean extinction could occur in just a few centuries.
The three studies, published last year, rely on projections of the future produced by some of the world’s next-generation climate models. But even modellers recognize that many of these models have one glaring problem: predicting a future that gets too hot too fast. Although modellers are adapting to this reality, researchers who use model projections to assess the impacts of climate change have yet to follow suit. This has resulted in a parade of “faster than expected” results that threatens to undermine the credibility of climate science, some researchers fear.
Scientists need to be much more selective in how they use model results, argues a group of climate scientists in a comment published today in Nature. Researchers should no longer simply use the average of all climate model projections, which may result in global temperatures by 2100 up to 0.7°C warmer than an estimate by the Intergovernmental Panel on Climate Change. climate change (IPCC). “We have to use a slightly different approach,” says Zeke Hausfather, head of climate research at payment services company Stripe and lead author of the commentary. “We have to get away from the naive idea of model democracy.” Instead, he and his colleagues call for a meritocracy model, sometimes prioritizing results from models known to have more realistic rates of warming.
Overall, climate models remain incredibly powerful research tools, and nothing about this “too warm” generation invalidates the principles of climate science, says Kate Marvel, a climatologist at the Goddard Institute for Space Studies in NASA and co-author of the commentary. The greenhouse effect continues to warm the planet. Ice is melting, seas are rising and droughts are becoming more frequent in some areas. But the models aren’t perfect, says Marvel. “These are not crystal balls.”
The problem of too hot models was born in 2019 from the Coupled Model Intercomparison Project (CMIP), which combines the results of global models upstream of the major IPCC reports that come out every 7 or 8 years. In previous CMIP cycles, most models projected a “climate sensitivity” – the expected warming when atmospheric carbon dioxide doubles from pre-industrial times – of between 2°C and 4.5°C. But for the 2019 CMIP6 round, 10 of the 55 models had sensitivities above 5°C– a sudden start. The findings were also at odds with a landmark study that eschewed results from global modeling and instead relied on paleoclimate and observational records to identify Earth’s climate sensitivity. He found the value to be somewhere between 2.6°C and 3.9°C. The discrepancy in sensitivity estimates is a “sobering example of the complexity of the climate system”, says Christopher Field, a Stanford University climatologist who focuses on impacts.
Researchers have since researched the causes of the overly hot models, including those produced by the National Center for Atmospheric Research, the US Department of Energy, the UK Met Office and Environment and Climate Change Canada. They are often related to how models render clouds; one of the results has been predicted excessive warming in the tropics.
Still, many of these models make the world a better place than their predecessors, and the centers that produced them were opened to diagnose the problem, Marvel says. “They are to be congratulated.” But it will be years before the centers can produce new projections for wide use.
The IPCC tried to offset this problem last year when it published its first working group report, which covers the physical basis of climate change. The IPCC rated the models on their ability to capture past historical temperatures. Then it used the deft models to produce its official “rated warming” projections for different fossil fuel emissions scenarios. When it came to studying the future changes of the Earth, the IPCC reported the results of all models based on the degree of warming: 1.5°C, 2°C, 3°C. This made it possible to use useful information from hot models, even if they reach these thresholds too quickly.
While the IPCC rose to the challenge, it didn’t do a great job of making everyone aware of the real problem, says Hausfather, himself an IPCC co-author. “A lot of our colleagues had no idea the IPCC had done this,” he says. And since then, dozens of published studies have used projections based on the raw mean of all CMIP6 models. The results, they note, are often “worse” than IPCC projections – and this has caught the attention of those unaware of the models’ underlying problems. “It’s not because anyone is acting in bad faith,” Marvel says. “It’s just because there are no tips.”
Climate impact researchers should emulate the actions taken by the IPCC, say Hausfather and his co-authors. First, they should avoid dubious time scenarios and instead focus on the effects of specific levels of global warming, regardless of when those levels are reached. They should also use the IPCC’s own “rated warming” projections to know when those levels of warming might occur. And for studies where details of the warming trajectory are important, they can use selected models that capture warming with relative accuracy, such as those produced by NASA and the National Oceanic and Atmospheric Administration, among others.
“I agree with almost everything the authors say and suggest,” says Claudia Tebaldi, a climatologist at Lawrence Berkeley National Laboratory and one of the CMIP climate projection scenario leads. However, she says, the recommendations may underestimate policymakers’ desire for temporal information, which in her experience is almost always requested. And some climate impacts, like sea level rise, change based on how long it takes to reach a level of warming, not just the absolute amount of warming.
Researchers should even think about going further and examining whether certain models have, for example, large regional biases, says Reto Knutti, a climatologist at ETH Zürich, who has been calling for “model meritocracy” for more than a decade. . As more urban planners and outside scientists turn to these projections, they should first be sure to consult an expert in climate models. “Given that these results are guiding climate adaptation and billions of dollars of investment, it seems like a worthwhile effort,” Knutti says.