To follow up on my post about Understanding Lewis (2013), I thought I might highlight a recent paper called The dependence of transient climate sensitivity and radiative feedbacks on the spatial pattern of ocean heat uptake, by Rose et al. One of my main issues with Lewis (2013) was that adding 6 years of data (1995 – 2001) changed the ECS estimate from 2.0 – 3.6^{o}C, to 1.2 – 2.2^{o}C. This seems quite an alarming change, given such a small increase in the data being used.

What’s interesting about this new paper (which I discovered through a tweet from Gavin Schmidt) is that it attempts to understand how climate sensitivity depends on regional changes in the deep ocean heat uptake. I’m not sure I’ve quite understood precisely what they’ve done (I’m trying to finish off a paper of my own, so am trying to avoid reading too many climate science papers at the moment), but the key figure is – I think – below. Essentially they seem to have run 3 sets of simulations. One of which is a standard run in which they double CO_{2}, and the other 2 are runs in which the deep ocean heat uptake is either mainly in the tropics, or mainly in the polar regions.

The main result in the paper seems to be that varying the regional ocean heat uptake can significantly change the sea surface temperatures and also changes the radiative forcing from feedbacks. The key figure is the top left one above which shows the SST results from the 3 sets of simulations (each set uses 4 different GCMs). My understanding is that this means that if one wants to estimate climate sensitivity (transient or equilibrium) using transient observations, it will be strongly influenced by the regional distribution of the deep ocean heat uptake. One would expect the regional variation in deep ocean heat uptake to be cyclical so, overall, these variations shouldn’t influence the actual climate sensitivities, but it does mean using transient observations to constrain them can be very difficult, and likely unreliable.

The paper itself actually says,

Transient climate response is governed both by an evolving pattern of sea surface warming activating different local feedbacks and by changes in the local feedbacks themselves as the pattern of OHU slowly evolves. This casts doubt on the possibility of estimating the feedbacks governing transient climate change from equilibrium mixed layer models (as noted by Shell [2013]), and more importantly, of estimating equilibrium climate sensitivity from inherently transient climate observations.

So, the paper would seem to be suggesting that the equilibrium climate sensitivity cannot *be reliably estimated from transient observations*. That Lewis (2013)’s ECS estimate appears to have changed dramatically with the addition of only 6 years worth of data would seem to be consistent with this new paper. Alternatively, I guess, Lewis (2013) could just have made some kind of error. It’s possible, of course, that the Lewis (2013) result is somehow correct, but that would seem quite remarkable if true, even if the results in this paper do not turn out to be robust.

This seems a bit like doing statistical and/or monte carlo tests on systems where the answer is known by construction, (e.g. a linear trend plus noise) to see if you’re approaching it right.

On similar lines I think it would be interesting to have an open competition where people test their methods of estimating ECS on a system (i.e. an earth-like model) or a set of model runs where the answer is known, but they are given only obs with uncertainties etc.

In other words if techniques such as those of Lewis are any good, let’s see how they fare on similar problems where we know the answer first.

I’d also like to know what happens with Lewis’s technique as you introduce additional years of data.

Yes, that may be one way to describe. One could regard the CO

_{2}only as the expected result and then the other two as how regional changes in the OHU influences TCR and ECS estimates.I, too, would very much like to see what happens to Lewis’s result if you add more data.

“I, too, would very much like to see what happens to Lewis’s result if you add more data.”

Wouldn’t it be possible to do that just by withholding the last N years of data and successively adding them back to see how it changes the result?

Frank,

I suspect it would. There’s a Libardoni & Forest (2013) correction to one of their earlier papers that seems to also use the Lewis (2013) method and shows that it is very similar to their method (and never returns an ECS range close to 1.2 – 2.2

^{o}C. It also considers different temperature datasets, but doesn’t explicitly say what time periods are considered.Title typo, “budgest” – unless you really do mean the most budge possible.

jsam, I presume he didn’t mean it and so have fixed it. Hope you don’t mind, AndThen.

jsam and Rachel, thanks. Maybe we should call Lewis’s work “Energy budge constraints” since he doesn’t seem to want to budge even if it does look rather suspicious 🙂

Not really sure what to make of this. The paper is about a set of highly idealised GCM simulations – aquaplanet setup (no land, no sea ice, just dark blue sea covering the entire planet), 10 metre mixed layer ocean, no annual cycle with insolation fixed to Equinox conditions, and not sure about rotation.

Conceptually you can say the results point to something which needs to be taken into account with regards observational sensitivity estimates but they don’t appear to have actually investigated what quantitative, or even qualitative, impact it might have on such estimates.

You might be interested in this paper. The revised version is behind a paywall but the change in the estimate was only a littler higher than in this version.

Heat capacity, time constant, and sensitivity of Earth’s climate system. Schwartz S. E. J.

Geophys. Res., 112, D24S05 (2007). doi:10.1029/2007JD008746

As I understand the paper Dr. Schwartz estimate climate sensitivity to doubling of CO2 as 1.1 ± 0.5 K. He based his estimate on ocean heat content. This early estimate was later adjusted a little upwards,

Chris,

I haven’t had a chance to read it thoroughly, but my current thought is that he’s calculated our equilibrium temperature today (i.e., what it would rise to were we to maintain the current level of CO

_{2}) and not the equilibrium temperature due to a doubling of CO_{2}. I could be wrong about that, but that probably doesn’t change that he’s wrong. Whatever he has actually done, he’s done an Otto et al. like calculation and got a very different, and completely unrealistic, result.Pingback: Sensitivities and things | And Then There's Physics

Pingback: Energy budget estimates explicitly using feedbacks | And Then There's Physics