Thursday, March 15, 2012

Bayesian estimation of climate sensitivity based on a simple climate model fitted to observations of hemispheric temperatures and global ocean heat content

Mentioned previously, but I've just got a copy of it from the author (we don't subscribe to this slightly obscure journal):

Bayesian estimation of climate sensitivity based on a simple climate model fitted to observations of hemispheric temperatures and global ocean heat content

The punchline is this plot which seems similar (but not identical) to that in the Isaac Newton seminar:

The indicated ranges denoted by triangles and line are 90% and 95%, ie a 5% probability of S exceeding 3.5 and 2.5% chance of exceeding 4.3. At a quick glance, the main improvements over previous work are a more detailed treatment of aerosol forcing (one the authors is an aerosol specialist) and the statistical treatment is certainly a bit more sophisticated than some have used before. Interestingly, this result is based on a uniform prior for S, but still cuts off the tail of high values rather effectively. The high tail returns if they add another (pseudo-)forcing term relating to the aerosol effect on cloud lifetime, which is ignored in the main analysis.

Of course this is just one study and I wouldn't expect the IPCC to base their whole report on it, but neither would I expect them to ignore it.

OK, since the commenters seem interested, here are two sensitivity tests where an additional forcing effect (effect of aerosols on cloud lifetime) is included:

You can see that the long right tail reappears when a reasonably strong effect is added. Note, however, that this is still contingent on a prior for S that is actually uniform on [0,20] (eg 70% probability that S is greater than 6) and a more reasonable starting point would not generate such a result.


PeteB said...

It looks like the lower bound and mid sensitivities are significantly lower than the current consensus estimate as well - have I understood this correctly ? ,if so any particular reason for the differences in this paper ?

Mark said...

From the Conclusion:

"The resulting estimate of the climate sensitivity is slightly smaller than the best estimate given in IPCC (2007) and could be compared with
other estimates as well. However, we underscore that our results are sensitive to the indirect aerosol effects, which have a large uncertainty.
In this study, the cloud-albedo effect is treated as a radiative forcing mechanism in the main part of the study, whereas other indirect aerosol
effects will be parts of the climate feedbacks. Therefore, the estimate of S presented here is likely to be underestimated because the net
forcing of the other indirect effects are likely to be negative (Forster et al., 2007). In one of the sensitivity cases, this assumption is further

James Annan said...

Yes, one of the differences is in the way they have handled aerosol forcing. There are both direct and indirect effects. Many people doing similar analyses have simply used a rather vague prior for the total effect. This lot have used a more detailed knowledge of the direct forcing, but then the indirect effects should also be accounted for somehow. I've added plots of their sensitivity tests.

Another nicety is in the use of SOI index to account for part of the natural variability - taking this off means the forced response is slightly more tightly constrained.

EliRabett said...

Looking at the following post, a question occurs, is there hysteresis in the earth system at least for some sorts of forcings?

James Annan said...

The big hysteresis that everyone knows about is the snowball earth due to ice albedo feedback. And theories for getting out of this state are a bit flaky (AIUI).

However that's not a matter of forcing type, but rather magnitude. To first order, type doesn't matter, and small enough magnitude forcings are always reversible.