Today was the all-Union session on Tipping Points, and several people have asked for comments on what went on there. I suppose this session might have been useful for people who had to miss the more detailed discussion in specialized sections, but I don’t have much to say about most of the talks, since they for the most part went over issues like ice sheet dynamics and rapid arctic sea ice loss, which I’ve discussed in earlier dispatches. Myself, I never found the notion of “tipping points” to be a very useful contribution to public discourse. The concept is ill-defined and very prone to be misunderstood — as in: we’ve passed a tipping point so it’s too late to do anything (might as well have a party). In Hansen’s talk, he did try to clarify what he meant by a tipping point. His notion of this has less to do with what mathematicians understand as “bifurcations,” and more to do with a kind of inertia in the climate system. He means things like having passed a threshold of CO2 which, given warming in the pipeline and the lifetime of CO2, commits a certain discrete event — e.g. loss of perennial sea ice or the Amazon rainforest– to occurring even if we were to later reduce emissions to zero. He tried to distinguish between reversible and irreversible tipping points. The talk was good cheerleading, after a fashion, but rather thin on real examples of what might be a tipping point in his definition. Everything he said was true (especially the emphasis on the importance of a rapid phase-out of coal burning) but the talk had much more to do with energy policy and lamentation of the power of entrenched fossil fuel interests than it had to do with climate science.
I skipped out of the session to catch some posters, but I came back in time for an interesting talk by Booth et al, of the Hadley center, showing the robustness of their simulation of Amazon dieback against variations in uncertain atmospheric parameters (it may die sooner, it may die later, but die it does). He showed, though, that whether the Amazon dies back is sensitive to the formulation of the land surface model, with only about half of the randomly-chosen cases done giving a dieback. Is this a tipping point? I’m not sure I care whether it is or not, but it sure is important, especially given how much CO2 gets dumped into the atmosphere if the Amazon dies. A nasty thing is that the part of the Amazon that is most robust is precisely the part where deforestation from economic development is worst.
What I personally found most interesting today, with regard to climate change issues, was contained in three papers or posters by Camp, Tung and a few other collaborators, concerning the surface temperature response to forcing by total solar luminosity changes in the 11 year solar cycle. The first talk was not specifically tied to the luminosity: it was a slight variant on the Camp and Tung paper which appeared recently in GRL, which used the periodicity of the 11 year cycle to detect the pattern and magnitude of the solar cycle in surface temperature data from the NCEP reanalysis. The slight variant was that instead of doing a composite, Camp used a form of linear discriminant analysis. It gives similar results to the compositing method: polar amplification in the pattern, and a global mean temperature amplitude of about 0.18K peak to peak. That’s nearly twice what most other analyses give; e.g. Scafetta’s estimate yields more like 0.1K .
That wasn’t so terribly exciting, given the earlier GRL result, but where things get interesting is where you try to explain a magnitude of signal this big in terms of basic physics. This is important because there is a perception that GCM’s vastly underestimate the amplitude of the response to total solar luminosity, leading to a perception that there is some “missing physics” (whether it be exotic amplification of a stratospheric response, or something like clouds and cosmic rays). In a second talk, Tung and Camp looked at simple surface energy balance models with thermal inertia, to see what they could do. To set the stage, Tung points out that a naive estimate would say that to get a .17K signal from a solar irradiance cycle with amplitude of .18 Watt per square meter you need a climate sensitivity factor of about 1 — that would give you equilibrium warming of 3.7K for doubling of CO2 (which has a radiative forcing of 3.7 Watts per square meter). That’s actually an underestimate, since the response to the 11 year cycle is damped by thermal inertia, so that underestimates equilibrium sensitivity — the thermal inertia in the atmosphere and ocean averages out the bright sun and faint sun periods to some extent. Thus, Camp and Tung’s result points towards a climate sensitivity considerably higher than the mid-range IPCC number.
Now, they go further, using the surface energy balance. They explicitly go about trying to explain the response in terms of standard energy balance amplified by standard feedbacks (water vapor, ice albedo, and cloud response to temperature changes), without anything exotic. They find that they can do so in their surface energy balance model, though they don’t actually attempt to identify the physical feedback mechanism. That’s just left as a generic “feedback factor.” The feedback factor that gives the best fit to data is compatible with an equilibrium warming of around 4K for doubling. One aspect of the model they use, which troubles me, is that Camp and Tung write a time-dependent energy balance equation for the lower atmosphere — using the thermal mass appropriate to the lower atmosphere. This gives a rapid response to solar irradiance changes, with little averaging, and gives a response that is almost in-phase with the solar cycle (as the observations indicate). That would be appropriate if they were holding the surface temperature fixed and driving the atmosphere with just the 20% or so of solar radiation absorbed directly in the atmosphere. That’s not what they do, though. They dump the full solar energy fluctuation right in the atmosphere. That would be appropriate if the ocean had a thermal response time much less than 11 years, but not, say, for a 50 meter mixed layer ocean. They justify their choice by invoking some evidence that the solar cycle only affects the very upper part of the ocean, greatly reducing the ocean’s contribution to thermal inertia. That assumption seems a bit dicey to me, but it does seem to be consistent with what comes next.
The next part is the really interesting and most important part. In poster by Tung, Yau, Li, Shia, Li, Waliser and Yung (GC43A-0935) the authors look at 22 IPCC models from the AR4 archive used in the Fourth Assessment report. 11 of these models include solar cycle forcing by irradiance variations, and the other 11 use a constant solar irradiance. All of these models have a fully dynamic ocean. The latter, as expected, do not show any significant 11 year cycle in surface temperature. However, all of the 11 models with solar variability show a significant solar cycle in temperature. Some models have a weaker response than others, and all are somewhat weaker than the observed cycle. The NCAR model has the highest amplitude cycle. An ensemble of 10 runs gives an amplitude of
about .10K in surface temperature, but one of the individual runs of the ensemble has an amplitude of .14K, only slightly less than the observations. That says that the high amplitude of the observed cycle could be just a matter of natural variability of the response. Even more important, the spatial pattern of the response is similar between models and observations.
Thus, while it is still possible that models have a somewhat weaker than observed solar cycle, Tung’s analysis would indicate that there isn’t anything major missing from the model physics with regard to response to solar variability. Note that none of the models analyzed has ozone chemistry feedbacks. It appears to be a simple matter of response to solar energy fluctuations, amplified by a feedback factor computed in a conventional way in the model physics (clouds responding to temperature and circulation, water vapor increase with temperature, and sea ice).
Now, it still remains to be understood how some of the models produce such a strong response to such a weak forcing. The key is in the thermal inertia (thermal response time) issue, and this is probably why dynamic ocean models can get a big cycle while mixed layer models don’t. The former have enough vertical resolution to allow the penetration of solar cycle variability into the ocean to be shallow, whereas mixed layer models don’t. They have basically only a single response time. Probably, the difference in amplitude of solar cycle amongst the models is partly a matter of different strengths of feedbacks, and partly a matter of different depths of heat burial in the ocean. Models with shallow heat burial have lower thermal inertia, less averaging, and a bigger response.
At first I thought that K-K.’s result pointed in the direction of a high climate sensitivity, and that may still be true, but the issue is tied up with thermal inertia. For a model which buries heat deeply on the 11 year time scale, the ocean averaging is strong, and response is weak; that does not tell us that equilibrium climate sensitivity is weak, though. On the other hand, if K-K is right and the real solar cycle only affects a shallow layer, then the solar cycle response is close to equilibrium (something that goes along with the small phase shift, since a strong thermal inertia would make the response a quarter cycle out of phase with the forcing). In that case, the solar cycle response is measuring equilibrium sensitivity, and a large amplitude indicates a large equilibrium sensitivity, as in K-K’s earlier back-of-the-envelope calculation. Viewed this way, the slightly too-weak NCAR response could mean that it mixes heat too deeply on the 11 year time scale, or it could be that it mixes to the right depth but has insufficiently strong amplifying feedback. The most parsimonius explanation of the amplitude seen by Camp and Tung in the observations is that (a) the ocean burial is shallow on the 11 year time scale, and (b) the equilibrium climate sensitivity is high. These ideas could be tested by more complete diagnostics of heat burial in the NCAR model, and solar-cycle response runs with a two-layer mixed layer model in which the upper layer is shallow. I think I’ll give it a go, if I can find the time.
But — the take-home point is that at this point the study of solar cycle response very strongly supports the notion that there is no need to invoke any mysterious or exotic missing physics (like cosmic ray modulation of clouds) in order to represent the response of climate to solar variability. If some models underestimate the response, this is likely to have more to do with errors in the vertical mixing of heat than any missing fundamental physics.
More grim tidings: