Nic wrote: “For the tropics (20S-20N), where solar energy input and ocean temperatures are highest, regressing TOA outgoing LW radiation per CERES data (2001-13) on surface temperature, using detrended and deseasonalised data, gave an increase of 4.05 W/m2/K”.
Thanks for the reply. My problem with detrended and deseasonalized data is that the changes are so small making the confidence interval wide (and potentially subject to systematic error). -4.05 W/m2/K is +/-0.82 for CERES (smaller than I expected), -3.0+/-3.3 for ERBE and -4.0+/-1.5 when analyzed by Lindzen and Choi. When I look at the scatter in the data in M&S(2015) Figure 2a, my training in hard sciences with definitive well-controlled experiments rightly or wrongly makes it hard for me to take the slope seriously. There is much less uncertainty in the seasonal change, because the changes are huge and clearly linear.
It is clear that we want to know about feedbacks during global warming, not during warming in the NH and cooling in the ocean-dominated SH (seasonal warming). It is also clear that the tropical response is underweighted in seasonal warming. However, we need more than 20S to 20N also. The greatest value in looking at the feedbacks in response to seasonal warming is that they prove AOGCMs don’t get these feedbacks right and AOGCMs are mutually inconsistent with each other. In particular, LWR feedback from all and clear skies is equal in observations, but there is positive LWR cloud feedback in models. If I had 1/100 of your data analysis skills, I would look at the feedbacks to seasonal warming in various regions: latitude, ocean vs land, areas where boundary layer clouds are important, and areas with extremely high clouds that cause the most warming. With tight confidence intervals, one might have strong evidence about what processes AOGCMs are getting wrong, and which way they are biased. The scientific method is to attempt to invalidate hypotheses and an AOGCM is a sophisticated hypothesis about how our climate system behaves. Rather than declare AOGCMs invalid, Tsushima and Manabe merely say the information can be used to improve AOGCMs, but I think they need a better idea of what needs improving. Perhaps they have surveyed these possibilities and not published.
The SWR response to seasonal warming is non-linear with temperature and therefore isn’t really a feedback. But it is still diagnostic from how well AOGCMs reproduce observed seasonal changes in SWR. Many others report lagged SWR response, but the surface albedo and cloud albedo response aren’t likely to have the same lag. If about half of the sky is cloudy due to rising air masses and about half clear due to subsiding air masses (and there is no reason for the ratio to change as temperature rises), won’t cloud SWR feedback be near zero? (IIRC, boundary layer clouds are not due to rising air, so they would be in a separate category.) When I look at M&S(2015) Figure 2b, I translate +1, +2, +3 and +4 W/m2/K in SWR feedback into 1%, 2%, 3% and 4%/K, and then extrapolate to 3 degK of AGW or -6 degK at the LGM. Those are massive changes.