The Big Bet on Climate Sensitivity

On the front page of this blog is my candidate for the most important risk indicator in the world: the atmospheric concentration of CO2 (currently at around 398 parts per million, or 42% above the pre-industrial level of approximately 280 ppm).

The degree to which the world’s temperature responds to a rise in atmospheric CO2 is captured in a metric called ‘equilibrium climate sensitivity’. This sensitivity number is an estimate of how much global mean temperature will rise should atmospheric CO2 concentration double (in other worlds rise to 560 ppm) such that the heat going into the earth system is back in balance with the heat emitted from the earth system; i.e, the attainment of a new equilibrium.

To put this number is context, the global community has (somewhat arbitrarily) taken a rise of 2 degree Celsius in global mean surface temperature to be regarded as the threshold beyond which the world will experience dangerous climate change. So the critical question then becomes: at what level of CO2 in the atmosphere will we become committed to 2 degrees Celsius plus of warming? By extension, should our best estimate of equilibrium climate sensitivity be 2, then we will cross the dangerous climate threshold of 2 degrees of warming only if we double atmospheric CO2. If the equilibrium sensitivity number were 3, we would cross the dangerous climate change threshold at a far lower level of atmospheric CO2. In short, a low sensitivity number is good, a high one bad.

In 2007, the Intergovernmental Panel on Climate Change (IPCC) published its latest estimate for equilibrium climate sensitivity in its Fourth Assessment Report (AR4). This report is taken by policy makers to be the consensus view of climate scientists at a particular time. And here is the best estimate as of 2007:

Equilibrium climate sensitivity is likely to be in the range 2°C to 4.5°C with a most likely value of about 3°C, based upon multiple observational and modelling constraints. It is very unlikely to be less than 1.5°C.

Surprisingly, despite a plethora of papers and the advancement of computer modelling, the climate sensitivity number has hardly moved over the years. But we now appear to have the makings of a new consensus that the climate sensitivity number may be somewhat lower than the 3 degree best estimate agreed upon in 2007.

If true, this is certainly good news and should be applauded. The latest paper supporting a slightly lower climate sensitivity number is that of Otto et al., which was published in Nature Geoscience. Unfortunately, the original paper is behind a paywall, but, realising the importance of the paper, Nature has published an open access synopsis (which they term Supplementary Information) that can be found here.

One of the authors of the study is Nic Lewis, who has previously published work suggesting a much lower equilibrium climate sensitivity number than in the IPCC’s 2007 report. Lewis stresses (here) the credentials of the authors in the new Nature Geoscience paper, including the fact that many of them are deeply involved in the creation of the IPCC’s Fifth Assessment Report (AR5), to be published in 2014:

The authors include fourteen climate scientists, well known in their fields, who are lead or coordinating lead authors of IPCC AR5 WG1 chapters that are relevant to estimating climate sensitivity. Two of them, professors Myles Allen and Gabi Hegerl, are lead authors for Chapter 10, which deals with estimates of ECS and TCR constrained by observational evidence. The study was principally carried out by a researcher, Alex Otto, who works in Myles Allen’s group.

In sum, this is a legitimate paper and doesn’t emanate from a closet libertarian fruitcake or some embittered contrarian loon who was passed over for tenure.

Lewis has also helpfully produced a graphic showing the climate sensitivity estimates based on various observational periods (as to why he is putting this onto the climate skeptic blogs Bishop Hill and Watts Up With That I can’t quite fathom):

ECS Estimates jpeg

Helpfully, the graphic also contains box and whisper plots, which Lewis describes thus:

The box-and-whisker plots near the bottom of the charts are perhaps more important than the PDF curves. The vertical whisker-end bars and box-ends show (providing they are within the plot boundaries) respectively 5–95% and 17–83% CIs – ‘very likely’ and ‘likely’ uncertainty ranges in IPCC terminology – whilst the vertical bars inside the boxes show the median (50% probability point).


The Nature Geoscience ECS estimate based on the most recent data (best estimate 2.0°C, with a 5–95% CI of 1.2–3.9°C) is a little different from that per my very similar December study (best estimate 1.6°C, with a 5–95% CI of 1.0–2.9°C, rounding outwards).

Of course, the climate skeptic blogosphere has been wetting itself with delight over the climate sensitivity downgrade. However, from a climate risk perspective,  this reaction appears misplaced.

First, we should remember that climate risk is composed of four main components: 1) the carbon emission trajectory (i.e., how much fossil fuel we burn), 2) the carbon cycle (i.e., how much carbon stays up in the atmosphere, 3) climate sensitivity (how much we warm given a certain increase in atmospheric carbon) and 4) the damage function (how much society is impacted by a given level of warming). Therefore, while we tentatively have some good news with respect to one of these risk factors, it sheds little light on the others.

Second, there are numerous methods of estimating equilibrium climate sensitivity. A good overview of the different approaches is provided by Gavin Schmidt at Real Climate here. Schmidt has this to say:

In the meantime, the ‘meta-uncertainty’ across the methods remains stubbornly high with support for both relatively low numbers around 2ºC and higher ones around 4ºC

Accordingly, the new equilibrium climate sensitivity estimates should not be taken as the only representation of the truth. Different approaches are yielding different numbers. A degree of caution is required when weighting any one particular study—or rather one particular method of study.

Third, we should be very aware of the confidence range. The current study is looking at a 95% confidence interval of 1.2 to 3.9 degrees Celsius of warming. As a result, the probability distribution gives us numerous paths to a greater than 2 degree Celsius warming based on current carbon emissions trends. We should also not neglect the probability tails. Even if we take the Otto et al. study as the final word, there is still a 2.5% probability of a sensitivity above 3.9 degrees.

Indeed, when it comes to the tails of probability distributions, a conceptual gulf exists between the scientific and financial communities in my opinion. For many a scientist, the tail of the distribution is the area you neglect, being a ‘very unlikely’ outcome. For anyone in finance, the tail garners a massive proportion of your time. Risk is ultimately probability times effect. As such, the tail is where your portfolio blows up and you lose your job (or alternatively you reap untold wealth and are feted by your investors) since it contains low probability but high effect outcomes. In few climate discussions (and certainly none in the climate skeptic blogs) do I see much weight given to the climate sensitivity tail. Yet if we do see a climate sensitivity number of 4 degrees or more, the impact on humanity would likely be devastating.

Again, looking through a finance professional lens, I am very nervous over predictions being honed based on our limited observational data. We appear to be at a very early stage in understanding the heat uptake into the oceans and the impact of aerosols on warming, amongst other areas. In the past, such uncertainties (especially with respect to clouds) have been given as a reason by climate skeptics for doing nothing. Personally, I see uncertainty as a source of risk, and therefore as a catalyst for risk management—not as an excuse to dump risk management (such as the ‘burn baby burn’ mantra of the fossil fuel lobby).

Finally, I worry that supposedly stable relationships can suddenly suffer structural breaks , just as we have seen such breaks in recent years in financial markets (together with multi-sigma volatility events that were not expected to happen in thousands of years). Of course, climate change is based on physical relationships not the psychological relationships which are so important for the functioning of economies and markets. Nonetheless, we have just witnessed an extraordinary break in the data with respect to Arctic sea ice extent and volume, supposedly governed by physical relationships that could be modelled.

This example is instructive as it involves one of the most vocal climate prediction scientists, James Annan, who has argued for a more constrained confidence interval for equilibrium climate sensitivity and a lower best estimate. See, for example, his recent commentary on the Otto et al. paper in his blog here. James was also recently quoted with respect to the sensitivity question in The Economist magazine here and The New York Times here, among other places.

I know James from my time in Japan and the fact that we share an interest in how betting/financial markets could shed light on climate outcomes. James was involved in a series of papers with Daniel Bloch, regarding which I made a small contribution (here).

James has also been involved in a quite high profile bet with Joe Romm of Climate Progress with regard to whether the Arctic would be basically ice free by 2020, together with two other prominent bloggers Brian Schmidt and William Connolly. Here is James back in 2007:

Brian has already blogged the details. and so has William. All three of us are joining forces to oppose Joe Romm’s claim that the Arctic will be essentially ice free (down to 10% of the historical mean minimum) by 2020.


I don’t even have a strong view on the short-term fate of Arctic ice. But for that reason, I think it is unreasonable to claim that all the models and research (which suggests ice-free around mid-century and perhaps later) is badly wrong.

Joe Romm’s reporting of the bet back in 2007 can be found here. To win the bet, Arctic sea ice extent minimum must drop well below 1 million square kilometres based on data from the U.S. National Snow and Ice Data Center (NSIDC). Owing to the crash in sea ice extent in 2012, NSIDC recorded a minimum of 3.4 million square kilometres (see here). Accordingly, this bet is going to be a tight run thing.

In others words, James made his bet based on the premise that “it is unreasonable to claim that all the models and research (which suggests ice-free around mid-century and perhaps later) is badly wrong” but unfortunately we have evidence that all the models and research (which suggests ice-free around mid-century and perhaps later) has been badly wrong. I sincerely hope that the climate sensitivity number doesn’t spring such an unpleasant surprise.

2 responses to “The Big Bet on Climate Sensitivity

  1. This is a well-written synopsis of some problems associated with climate sensitivity. I would add to your list the following: while current CO2 concentrations are “only” ~400 ppmv, I see no robust signal that indicates concentrations will peak at 2X pre-industrial levels (560 ppmv). It is of course important to increase the PDF accuracy and reduce the climate sensitivity estimate range. But that gets us only so far. What remains is, from the expectation of continued business-as-usual practices for decades still, just as important. Skeptics tend to focus on the improved sensitivity estimates as evidence that climate scientists are clueless or on-the-take while ignoring the remainder of the message.
    I like your risk-centric view. Continuing on your financial risk theme, it might be instructive to compare high-risk bets that are made in the real world, especially for those of us who aren’t as familiar with the financial world. For example, I assume you could “bet” world GDP or some other metric on a certain climate sensitivity – say 2C (the median of the latest boxplot). If the sensitivity is instead closer to 4C, as you, I, and others acknowledge is just as physically realistic as 2C, what would that mean for the GDP bet? What losses could we incur and what ramifications would those losses have? Or am I totally off-base with my thought experiment?

  2. Weatherdem. I don’t think you are off base at all. The damage function appears one of the the weakest links in climate research. Our current ability to model damage at higher levels of warming looks very deficient and produces some ludicrously low estimates.

    I have blogged on it before here:

    And in more depth on a couple of posts looking at William Nordhaus’ DICE model:

    The two economists who have had a serious look at the tail risk associated with climate change are Martin Weitzman at Harvard and Michael Hanemann at Berkeley.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s