The latest generation of climate models is running hotter—here’s why

So what’s behind that behavior, and what does it tell us about the real world?

Climate sensitivity is one of the most-discussed numbers in climate science. Its most common formulation is the amount of warming that occurs when the concentration of CO₂ in the atmosphere is doubled and the planet gets a century or two to come to a new equilibrium. It’s an easy way to get a sense of what our emissions are likely to end up doing.

In climate models, this number is not in advance; it emerges from all the physics and chemistry in the model. That means that as modeled processes are updated to improve their realism, the overall climate sensitivity of the model can change. As results have trickled in from the latest generation of models, there’s been a noticeable increase in their average climate sensitivity. A new study led by Mark Zelinka of the Lawrence Livermore National Laboratory analyzes these new model simulations, comparing their behavior to the previous generation.

The analysis covered simulations from 19 of the 35 modeling groups, as not everything has come in yet. The climate sensitivity of these models spanned 1.8-5.6°C for doubled CO₂, with an average of 3.9°C. The last generation fell into 2.1-4.7°C, with an average of 3.3°C. So there is a slight uptick in the sensitivity in the latest generation of models. (For context, the overall best estimate for real-world climate sensitivity has for decades been 1.5-4.5 °C, centered on 3°C.)

Models work by dividing up the atmosphere into chunks called grid cells and then simulating atmospheric physics in each of these. To find out what caused the higher sensitivities, the researchers used a tool to break down the model responses in each grid cell. Their focus was on feedbacks—processes that either amplify or dampen warming. Differences in the various feedbacks are generally responsible for the range in overall sensitivity numbers.

In this analysis, most of the feedbacks look to have basically the same strength as in the previous generation of model. Low clouds, however, are behaving differently. That’s pretty important because low clouds can have a huge impact if they’re fluffy and reflective, shading the planet. If the amount of cloud shading changes as temperatures rise—which it does—you’ve got a feedback.

Low cloud behavior in models is a particularly lively topic of study. Previous research has found that, for example, models with the most realistic clouds (by some measure) tend to have higher climate sensitivities. And a lot of work goes into making measurements of the physics and chemistry of clouds, with modelers then trying to make sure their simulations match that behavior.

In the new study, researchers found that the low cloud feedback in the new generation of models seems to have changed outside the tropics, particularly the mid-latitude Southern Hemisphere. The average feedback among the models is a little more positive, amplifying warming. That would mean that as the Earth warms, low cloud cover in this region is decreasing a bit and reflecting less sunlight back to space.

The researchers dug a little deeper to see what was behind this shift and found that it’s probably related to some changes in the cloud physics equations. Comparisons to satellite data had shown that models should allow more liquid droplets to remain liquid at cold temperatures, and at least some models have been tweaked accordingly. That could be enough to alter how much the clouds change as temperatures increase.

In short, the newer models don’t seem to be doing anything unrealistic to cause their higher sensitivities. But that doesn’t necessarily mean they’re right. It’s going to take a lot more work to sift through these models and see which one has clouds that best match the real world. And as the Earth’s climate system is so interconnected, it’s even possible that some other factor—like a pattern of ocean temperature—is partly shaping cloud behavior.

Models aren’t the only way scientists estimate the Earth’s climate sensitivity. It’s also done by studying the historical record, for example, and past climate changes recorded by things like ice cores. Evaluating the new generation of climate models against those events might also provide some clarity.

This situation is certainly going to present a challenge for the authors of the upcoming IPCC report, though, as they’ll only be able to include studies published by September 2020. It’s a good bet that we’ll see more research on this topic in the coming months—and hopefully a lot of it, for the IPCC authors’ sake.

Geophysical Research Letters, 2020. DOI: 10.1029/2019GL085782 (About DOIs).