I thought I might revert back to a bit of physics and discuss a recent paper that aims to try and test the Iris hypothesis. The Iris hypothesis originates with Lindzen &Chou (2001) and proposes that as we warmed, there would be a reduction in cloud cover in the tropics, which would allow for more infrared radiation to escape to space and would – consequently – act as a strong negative feedback.The recent paper that tests this is Missing iris effect as a possible cause of muted hydrological change and high climate sensitivity in models by Thorsten Mauritsen & Bjorn Stevens. There are, I think, two motivations for this recent paper. One is that recent observations suggest that climate models might be over-estimating climate sensitivity and (which I had not realised) under-estimating changes to the hydrological cycle. There is also one model-data mismatch that may be consistent with an Iris effect. The figure on the right shows the short- (horizontal axis) and long-wavelength (vertical axis) sensitivity in the tropics. Climate models produce the same kind of short-wavelength sensitivity to that observed, but tend to underestimate the long-wavelength sensitivity. This suggests that the increase in outgoing long-wavelength flux with temperature is greater than climate models suggest, and is at least consistent with an Iris-like effect.
Since the Iris effect is essentially a reduction in cloud cover, and an increase in the size of the infrared window, in the tropics, this was implemented in the models by simply introducing a term, , that allowed them to adjust the rate, , at which cloud water was converted to rain. Essentially
where is the default rate, is the surface temperature, and is a typical temperature in the tropics.
The basic result is shown in the figure below. The grey dots in the left-hand panel show the equilibrium climate sensitivities (ECS) for a range of different climate models. The red dot is for the ECHAM6 model, which has an ECS of 2.8K. The yellow, light-green, and dark-green symbols show the impact of the Iris effect for and , but considering the long-wavelength effect only. This brings the ECS down into the range suggested by Lindzen & Chou (2001). The blue symbols, however, show the ECS when the short-wavelength and other feebacks are also included. The net effect is relatively small, with the ECS reduced from 2.8K, to between 2.2 and 2.5K, depending on the value of . The right hand-panel illustrates why. The long-wavelength effect is quite large, changing the feedback from around +0.5Wm-2K-1, to between -0.4 to -0.8Wm-2K-1. However, there are also changes to the water vapour feedback, lapse rate feedback, and short-wavelength cloud feedback that results in a relatively small net change in overall feedback.
So, this seems like an interesting paper that is really just testing what would be the consequences of there being an Iris effect, without actually demonstrating that it does exist. There are, however, some interesting hints. The results suggest that the Iris effect would reduce the ECS to bring the models more in line with what recent observations suggest. The change is, however, not big enough to bring it down to the kind of values suggested by Lindzen & Chou and, for the model considered here (ECHAM6), the change is from 2.8K to 2.2K at most. The Iris effect would also increase the hydological sensitivity which, again would make models more consistent with observation. Additionally, models currently underestimate the long-wavelength sensitivity in the tropics which is, again, consistent with a possible Iris effect.
Ultimately what this seems to be showing is that if observations do suggest that models are over-estimating climate sensitivity and under-estimating hydrological sensitivity, this could due to an Iris effect that is not being properly represented in the models. It’s almost certainly too early to know if models really are missing something like an Iris effect and consequently over-estimating climate sensitivity, but it’s an interesting paper that certainly presents some intriguing hints.