I wrote a post a while agoe about a paper that was suggesting that the reason for the difference between the observationally-based estimates for climate sensitivity, and other estimates, was that the pattern of sea surface temperatures can produce different system heat uptake rates for the same change in temperature. A recent paper by Zhou, Zelinka & Klein called Impact of decadal cloud variations on the Earth’s energy budget tries to explain why this happens. They argue that
the global mean cloud feedback in response to decadal temperature fluctuations varies dramatically due to time variations in the spatial pattern of sea surface temperature.
The figure below shows the basic result from a set of model run. The left-hand panel shows the net feedback response, while the right-hand panel shows the cloud feedback only. Essentially if you run a model with prescribed sea surface temperatures you get much more negative cloud feedbacks than the mean from long-term warming runs with either uniform, or patterned, sea surface temperatures. The suggestion is that over time we would expect the feedback response to tend back towards the mean, but we just happened to have experienced a period during which it was more negative than the mean.
Thorsten Mauritsen has a nice News and Views about this paper. It includes the figure on the right that illustrates what is thought to be happening. When the sea surface temperatures are relatively cool, there will be an inversion at 1-2km and clouds form below this inversion level. If the warmer regions warm more than the cooler regions, the inversion gets stronger and more low-level clouds form. Low level clouds tend to reflect sunlight that would otherwise have heated the Earth and, therefore, more low level clouds means a more negative cloud feedback.
So, it seems that one reason for the mismatch between observationally-based climate sensitivity estimates and other estimates is that the pattern of sea surface warming that we’ve actually experienced has produced less warming than would be expected from some kind of typical warming pattern. On the other hand, Thorsten Mauritsen points out that models rarely produce the pattern that was observed, which could suggest that they don’t properly represent variability, or don’t fully represent the response to increasing greenhouse gas concentrations.
This may also be related to the issue of whether or not we should regard the models as truth-centred, or not. It would certainly seem that over periods during which the variability can be comparable to the forced response, we shouldn’t assume that the observations should be comparable to the multi-model mean. Anyway, that’s all I was going to say. This seems like an interesting result and it does seem as though we’re starting to get a better understanding of how variability can influence the feedback response and how we then warm on decadal timescales.