The amount of global warming we can expect in the future has been a tough question to pin down. A new study that I led with colleagues in France has enabled us to come up with a more accurate analysis based on a better understanding of how clouds behave. It suggests that if fossil fuel use continues unabated we can expect warming of at least 4C by 2100. This is at the high end of the warming range suggested by many existing studies.
Our study ultimately concerns “climate sensitivity” - the amount the planet warms for a given amount of greenhouse gas. Think of it this way: to plan for a pizza party, you would start with an estimate of how much pizza an average person eats, and then multiply by the number of guests you plan to have. By analogy, global warming predictions begin by working out the climate sensitivity, and then scaling it to the amount of carbon dioxide and other warming gases we expect to emit.
Uncertainty abounds
The problem is that estimates of climate sensitivity have long ranged from 1.5-4.5C for a doubling of atmospheric carbon dioxide levels. Imagine not knowing whether the average guest will eat two or five slices of pizza. Determining this number more accurately is arguably the biggest question in climate science.
Unrestricted fossil fuel use could bring the carbon dioxide concentration to roughly three times pre-industrial levels by 2100. If climate sensitivity were near the 1.5C level, this would lead to 2-3C of further global warming by the end of the century. But if climate sensitivity were closer to 4.5C, we could expect 5-6C of warming by 2100. And by 2200, the warming in the high-sensitivity case could exceed 10C.
We could probably adapt to 2C of warming, although at the least it would be disruptive to delicate ecosystems like tropical coral reefs. But 6C would be catastrophic. Even 4C would probably be enough to displace tropical human populations, destroy natural ecosystems, force changes to agriculture and our way of life, and lead to eventual loss of the Greenland ice sheet and many metres of sea-level rise. So the amount matters.
Cloud control
This is where the clouds come in. Clouds, along with other factors such as sea ice, are affected by warming in ways that then go on to influence climate sensitivity. The question is whether they increase or decrease it.
In most calculations, warming reduces the amount of clouds near the ocean surface, increasing climate sensitivity because these clouds reflect less sunlight and thus allow more warming. This effect is much stronger in some models than others.
Our study breaks new ground in two ways. First, it identifies what is causing much of the discrepancy between different models, and why most of them show clouds near the surface thinning out as the climate warms. Second, it strongly discounts the lowest estimates of future warming. I am very happy about the first result, not so happy about the second one.
The key process we have identified is mixing of air between levels near the ocean surface where these clouds form, and levels a few kilometres higher up. If this mixing is strong, it tends to thin out the cloud layer as the atmosphere warms.
On the other hand, if the mixing is very weak, the water lost from low clouds is more than matched by increased ocean evaporation. Thus the clouds are thickened and warming is reduced.
By looking at the present-day atmosphere, we can therefore work out which models will offer the best prediction of the future.
We found that the mixing is strong in the real atmosphere, implying a high climate sensitivity. Of the 43 climate models we examined, every one with realistically strong mixing had sensitivity of more than 3C.
This narrows the expected range of warming to between 3C and 4.5C for a doubling of carbon dioxide. This is a momentous and sobering result, but how reliable is it?
The big picture
Some people have interpreted the recent modest warming rates as evidence of lower climate sensitivity. It would be great if those lower values were definitive, but working them out requires us to understand all the influences on climate in recent decades. For instance, pollution particles called aerosols have changed greatly in recent decades, particularly as a result of Asia’s economic growth, with uncertain effects on climate.
In contrast, researchers looking at prehistoric warming episodes have often found evidence of higher sensitivity.
As with any single study, our work does not settle the issue of climate sensitivity. But it does open a new door in understanding the role of clouds, and it raises the likelihood that sensitivity is high. If we are lucky, future research may find something that brings our best estimates of warming back down again. However, our new study shows that such a lucky finding would have to involve some process that is currently missing from every climate model - a much taller order than before.
Some media and blog reports have claimed our work shows that climate models are wrong. This misses the point. All models have flaws, but usually these flaws tell us nothing about whether — or in what direction — their future predictions might be off. What we have found is one particular flaw that systematically causes many (but not all) models to underestimate warming.
Our result is a sobering one, but in a logical world it would not alter policy all that much. After all, if you didn’t know whether the average guest would eat two or five pizza slices, you would probably order enough pizza to cover yourself in any case. Given the appalling stakes for being wrong on carbon dioxide, it is crazy not to take the same conservative approach. But so far, the world has evidently been crazy.
Perhaps our result can serve as a reminder that not knowing everything does not justify complacency. Uncertainty may mean the problem is worse than you thought.