February 27, 2012

Bigfoot, the Loch Ness Monster, and High Climate Sensitivity

A few months ago, we reported on a paper in the scientific literature (Schmittner et al. 2011) that concluded that there were only “vanishing probabilities” that the value of the earth’s climate sensitivity—the amount of global temperature change resulting from a doubling of the atmospheric carbon dioxide content—was above 3.2°C, and that a climate sensitivity exceeding 6°C was “implausible.” Now, a new paper has been published (Olson et al., 2012) that finds that the 95% confidence range for the value of the earth’s actual climate sensitivity extends only to a value as great as 4.9°C. This is yet another in an expanding list of papers that strongly suggest that that the IPCC entertainment of the possibility that the earth’s climate sensitivity is extremely high (say, greater than 5-6°C, is wrong).

As apocalyptic climate change lurks among high sensitivity values, these new findings virtually eliminate the places where it could be hiding—and relegate talk of apocalyptic climate change to that of Loch Ness monsters, big foot, and woolly mammoths in Siberia.

Roman Olson and colleagues (including Nathan Urban, also a collaborator on the Schmittner et al. project) published their new findings in the Journal of Geophysical Research. They set out to investigate the range of values which most likely contains the earth’s actual climate sensitivity using a combination of observations of the earth’s climate along with an intermediate complexity climate model developed at the University of Victoria in British Columbia, Canada. The researchers varied the parameters of the climate model, including the climate sensitivity, and then used the model to hindcast the observed changes in surface temperature (since 1850) and ocean heat content (since 1950). The model hindcasts were then compared with the actual observations and a probability was assigned to that group of parameters (including the climate sensitivity) which represented the probability that the actual observations could be produced by such a model parameter set. Olson and colleagues employed Bayesian statistics to establish this probability—a technique which employs a prior assumption about the distribution of potential parameter values (including climate sensitivity).

It turns out that the “priors” have a large influence on the final solution. In other words, if you already have some rough idea of the range of potential climate sensitivity, that rough idea can help guide you to a better solution when new, relevant data become available.

Back in 2007, when the Intergovernmental Panel on Climate Change (IPCC) published its Fourth Assessment Report (AR4), the IPCC decided that instead of using an “expert prior” (that is, one that was guided by a rough guess) to help guide its determination of the distribution of possible climate sensitivity values, that it would use a “uninformed prior” (that is, as it sounds, one which adds no previous knowledge). The uninformed prior used by the IPCC was a uniform prior—the IPCC assigned an equal chance that the climate sensitivity could be anywhere in the range from 0°C to 10°C. This choice seems somewhat absurd in light of the fact that ever since the first IPCC report, from back in 1990, the IPCC has issued a rough guess that the climate sensitivity was somewhere in the 1.5° to 4.5°C range. You would think that their own “expert” assessment would be better than one that was “uninformed”—but perhaps that tells you something about how much credence they put in their own expertise!

At the time of the preparation of the IPCC AR4, objection was raised to the IPCC about their use of a uniform prior, but these objections were largely brushed aside. The IPCC, undoubtedly, preferred the use of the uninformed prior rather than the expert prior because in doing so, it produced a solution in which very high climate sensitivity values (say, higher than about 5-6°C) could not be ruled out, and in fact, weren’t even particularly unlikely in some determinations. Figure 1 shows the IPCC version of a collection of climate sensitivity probability distributions that were gathered from the scientific literature. Note that the combined probability that the climate sensitivity value exceeded say 5°C was greater than 10-20% in some cases.

Figure 1. Climate sensitivity distributions retained by the IPCC from their assessment of the literature. Note the “fat tail” towards the right which indicates the possibilities of the climate sensitivity having a very large positive value (that is, a huge degree of global temperature rise for a doubling of the atmospheric carbon dioxide concentration) (source: IPCC AR4).

In doing so, the IPCC furthered the legend that apocalyptic global warming may very well exist. The IPCC figure (reproduced as Figure 1) is sort of akin to the (in)famous Patterson-Gimlin video of bigfoot or the Sturgeon photograph of the Loch Ness monster.

In each case, while a slim few believers still hang on, there is a plethora of evidence that shows that the images are most likely a hoax, or at the very least, a misinterpretation of reality.

The Olson et al. study is yet another addition to pile evidence that exposes the IPCC climate sensitivity figure.

Figure 2 is taken from the Olson et al. paper and shows the distribution of the climate sensitivity value produced from the data generated through their experiments. The blue curve is what they got when using Bayesian statistics assuming an uniform prior (like the IPCC did) and the red curve is their result when using an expert prior that was informed by the climate during the height of the last glacier period and the present day. Notice that that blue “uninformed prior” case has a much higher probability that the climate sensitivity is very high compared to the red “expert prior” case. In fact, in the “expert prior” case, there is a 95% chance that the true climate sensitivity lies between 1.8°C and 4.9°C, while in the “uniform prior” case, that range is 1.6°C to 10.2°C (notice how much the blue curve in Figure 2 looks like the IPCC curves depicted in Figure 1).

Figure 2. The probability distribution for the earth’s climate sensitivity (CS) in degrees Kelvin (K, which is the same as a degree Celsius) as determined by Olson and colleagues using an “expert” prior (red) and using an “uninformed” prior (blue). (Source: Olson et al., 2012).

As with any scientific study, there are a set of caveats that are attached to the findings. In this case, they mainly revolve around the performance of the climate model and the accuracy and natural variability of the observations. But, mindful of these caveats, the authors greatly favor their results using the “expert prior” over those using the “uniform” prior. They note:

The use of uniform priors for climate sensitivity can be problematic as the posterior estimates are sensitive to the upper bound for the prior [Annan and Hargreaves, 2011]. In addition, such priors do not take independently collected evidence from other studies into account. High climate sensitivities become possible in this case because the flat prior assigns them high weight to begin with, while the constraint provided by the observations can be relatively weak. This suggests that it is crucial to use independent prior information during [climate sensitivity] estimation whenever possible. [emphasis added -eds]

This seems directed squarely at the IPCC.

It is worth pointing out that the Olson et al. study does not find that the most likely value for the climate sensitivity (which they peg at 2.8°C) is much lower than the IPCC’s best estimate of “about 3°C.” But the results do suggest that the IPCC’s “likely” (which means greater than 66%) range for the climate sensitivity of between 2.0°C and 4.5°C is too great at the high end.

If the IPCC continues to hold on to its use of an uninformed prior when assessing the climate sensitivity in its Fifth Assessment Report (which is currently under construction), we will no longer have to wonder whether they are perpetuating a myth or just badly misinterpreting realty. The preference for the former will be clear.


Olson, R., et al., 2012. A climate sensitivity estimate using Bayesian fusion of instrumental observations and an Earth System model. Journal of Geophysical Research, 117, D04101, doi:10.1029/2011JD016620, http://www.agu.org/pubs/crossref/2012/2011JD016620.shtml

Schmittner, A., et al., 2011. Climate sensitivity estimated from temperature reconstructions of the Last Glacial Maximum, Science Express, November 24,2011, DOI: 10.1126/science.1203513, http://www.sciencemag.org/content/334/6061/1385.abstract

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress