Eli’s recent post about Black swans, and black cats, motivated me to look into what the whole Black Swan idea was all about. As I understand it, a Black Swan event is simply an unexpected event, that has a significant impact, and that – in retrospect – we regard as something that could have been predicted. Unless there’s some subtlety that I’m missing, this essentially seems to be equivalent to the uncertainty isn’t our friend and high risk, low probability events arguments that have been made in relation to climate change before.
In Eli’s post he touched on Nic Lewis’s Energy Balance approach, and I thought I might expand on this a bit here. Many people seem to use the basic energy balance calculations to argue that climate sensitivity is probably low and that climate models are over-estimating climate sensitivity. The energy balance approach is fairly simple: the transient climate response (TCR) can be estimated using
and the equilibrium climate sensitivity (ECS) can be estimated using
If you carry out an energy balance-type calculation (see, for example Otto et al. 2013 and Lewis & Curry 2014) you do indeed find that the best estimates for climate sensitivity are lower than many other method would suggest, and the range is also shifted to lower values. A number of people are therefore using these results to argue that climate sensitivity is probably lower than previously thought, and that climate models are over-estimating climate sensitivity.
However, there is something that should be borne in mind when making such claims; energy balance models have a number of assumptions which – in my experience – are rarely acknowledged.
- Feedbacks are linear: There is a fundamental assumptions that the feedback response is linear; the feedback response in the future will be the same as it has been over the interval considered by the energy balance calculation.
- Polar amplification is negligible: A number of the temperature datasets suffer from coverage bias and may be underestimating the temperature change through not including sufficient coverage of the Arctic where warming may have been faster than the global average. One can compensate for this assumption by using a temperature dataset that tries to account for this coverage bias (Cowtan & Way, for example) but I don’t think any published estimates have done so.
- Internal variability is negligible: Given that energy balance models assume that the observed temperature change is all externally forced, they’re essentially assuming that internal variability has had no effect.
- Forcings are homgeneous: The forcings are assumed to be globally homogeneous. Given that there is more land mass in the northern hemisphere than the southern hemisphere, the north should warm faster than the south. Any inhomogeneity in the forcings, could therefore influence the global estimates.
So, given these assumptions, I don’t think one can really argue that energy balance calculations suggest that climate models are over-estimating climate sensitivity. They might be, but all I think one can say is that if the above assumptions are true, then energy balance calculations suggest that climate sensitivity might be lower than other estimates suggest. In fact, one might argue that all the above assumptions are probably wrong to some degree or another and that some could change the estimates by 5-10%. If so, it’s hard to really argue that energy balance estimates are evidence against the higher climate sensitivities that other methods suggest might be likely. I think it is this kind of issue that leads Nasim Taleb to say
Skepticism about climate models should lead to more precautionary policies in the presence of ruin. It is incoherent to doubt the mean while reducing the variance.
Maybe I’ve misunderstood the whole Black Swan idea, but it does seem that climate change could be ripe for a Black Swan event. Given how hard and how fast we’re pushing our climate, it wouldn’t be particularly surprising if something unexpected were to happen. Could it be something good? Possibly, but I think the parameter space for unexpected events having a damaging impact is significantly greater than the parameter space that would allow for positive impacts. Of course, if something were to happen, it would probably be followed by some people saying “we didn’t see that coming”, immediately followed by others responding with “we did!”