Climate Sensitivity Single Study Syndrome, Nic Lewis Edition
Posted on 18 April 2013 by dana1981
Nic Lewis has written a paper on the subject of the Earth's climate sensitivity (how much surface temperatures will warm in response to the increased greenhouse effect from a doubling of atmospheric CO2, including amplifying and dampening feedbacks) which has been accepted by the Journal of Climate. First of all, we would like to offer kudos to Lewis for subjecting his analysis to the peer review process, which is something few climate contrarians are willing to do.
The paper is an outlier, finding a lower climate sensitivity than most other studies, and outside the likely range cited in the Intergovernmental Panel on Climate Change (IPCC) report. It's most important not to fall into the trap of thinking that any single study will overturn a vast body of scientific evidence, derived from many different sources of data (or as Andrew Revkin calls this, single-study syndrome). This was also recently an issue with regards to a similar and unpublished Norwegian study.
Lewis' is just one paper using one of many possible methods to estimate climate sensitivity. The overall body of evidence indicates that the Earth's surface temperatures will warm 2–4.5°C in response to a doubling of atmospheric CO2.
Lewis took the approach of revising an analysis by Forest et al. (2006), applying Bayes' Theorem to a combination of an intermediate complexity climate model and recent instrumental data of surface and ocean temperatures. This Bayesian approach involves making use of prior knowledge of climate changes to establish a probability distribution function for climate sensitivity.
Lewis describes his approach here. Forest et al. applied Bayes' Theorem to three climate model parameters – climate sensitivity, effective ocean diffusivity, and the aerosol forcing. Lewis applied it to the data rather than the model parameters, and also added six more years of data to the analysis. The resulting climate sensitivity estimate in the Forest approach was 2.1–8.9°C surface warming in response to doubled CO2, with a most likely value of 2.9°C. Using an 'expert prior' reduced the 90% confidence interval to 1.9–4.7°C. Using his approach, Lewis estimated the 90% confidence interval at 1.0–3.0°C, with a most likely value of 1.6°C.
The Climate Variability Question Mark in Lewis' Approach
Another new paper by Olson et al. (2013) attempts to determine the amount of uncertainty that the short-term internal variability of the climate system brings into this sort of analysis. In fact, the study mentions Forest et al. (2006) as an example of the sort of approach that's vulnerable to large climate variability uncertainty – those which estimate climate sensitivity with intermediate complexity climate models in conjunction with recent historical observations. Olson et al. investigate three main sources of what they call "unresolved climate noise":
(i) climate model error;
(ii) unresolved internal climate variability; and
(iii) observational error.
The study concludes,
"each realization of internal climate variability can result in a considerable discrepancy between the best CS [climate sensitivity] estimate and the true value ... average discrepancy due to the unresolved internal variability is 0.84°C"
"These results open the possibility that, recent CS estimates from intermediate complexity models using global mean warming observations are systematically higher or lower than the true CS, since they typically rely on the same realization of the climate variability. For this methodology, the unresolved internal variability represents a critical roadblock."
In short, one should be very cautious about putting too much weight on any single study using the methodology of Forest and Lewis, because short-term natural internal variability can considerably bias the result on either the high or low side.
Effective vs. Equilibrium Climate Sensitivity
Even though Lewis refers specifically to "equilibrium climate sensitivity," The methodology used by Lewis is also not even necessarily an estimate of equilibrium sensitivity, but rather of effective climate sensitivity, which is a somewhat different parameter. The two may hypothetically be the same if all energy changes in the global climate system are accounted for (and to their credit, Forest and Lewis do include estimates of ocean heat content, including for the deep oceans), and if climate feedbacks remain constant. However, recent research by Armour et al. (2012) suggests that the latter may not be the case.
"Time-variation of the global climate feedback arises naturally when the pattern of surface warming evolves, actuating regional feedbacks of different strengths. This result has substantial implications for our ability to constrain future climate changes from observations of past and present climate states."
Considering the Full Body of Evidence
Lewis' approach is of course just one way to estimate the climate sensitivity. There are many others, as illustrated in Figure 1 from Knutti and Hegerl (2008).
Figure 1: Distributions and ranges for climate sensitivity from different lines of evidence. The circle indicates the most likely value. The thin colored bars indicate very likely value (more than 90% probability). The thicker colored bars indicate likely values (more than 66% probability). Dashed lines indicate no robust constraint on an upper bound. The IPCC likely range (2 to 4.5°C) is indicated by the vertical light blue bar. Adapted from Knutti and Hegerl (2008).
For example, one common way to estimate climate sensitivity is to examine past climate changes (paleoclimate). A project called the PALAEOSENS workshop put together paleoclimate estimates of climate sensitivity from nearly two dozen investigations of many different geological eras, and published their results in the journal Nature. They found that all of these paleoclimate studies resulted in a likely equilibrium climate sensitivity estimate of 2.2–4.8°C surface warming in response to a doubling of atmospheric CO2 (Figure 2).
Figure 2: Various paleoclimate-based equilibrium climate sensitivity estimates from a range of geologic eras. Adapted from PALEOSENS (2012) Figure 3a by John Cook.
Note that Lewis' best climate sensitivity estimate of 1.6°C is incompatible with the likely range (68% confidence interval) from these many paleoclimate-based estimates, though it is within the 95% confidence interval (1.1–7.0°C).
In Figure 3, the PALEOSENS team also illustrates the amount of warming we can expect to see at various atmospheric CO2 levels, based on these paleoclimate studies, using several different approaches. Doubling of CO2 from 280 to 560 parts per million results in close to 3°C global surface warming at equilibrium, when accounting for relatively fast feedbacks. The paper also discusses various estimates of 'Earth System Sensitivity', which includes slower feedbacks that operate over thousands of years. They estimate this longer-term warming in response to doubled CO2 would be closer to 7°C.
Figure 3: Equilibrium response of the global temperature as a function of CO2 concentrations, based on three different approaches. a) from the PALEOSENS workshop, using data from the late Pleistocene of the past 800 kyr; b) Using data of the past 20 Myr from RW_11; c) Based on JH_12 using similar data of the past 800 kyr as in a); and d) Combination of all three approaches. Plotted areas include uncertainty estimates of one standard deviation.
Misrepresenting Aldrin et al. (2012)
One significant issue in Lewis' paper (in his abstract, in fact) is that in trying to show that his result is not an outlier, he claims that Aldrin et al. (2012) arrived at the same most likely climate sensitivity estimate of 1.6°C, calling his result "identical to those from Aldrin et al. (2012)." However, this is not an accurate representation of their results.
The authors of Aldrin et al. report a mean climate sensitivity value of 2.0°C under certain assumptions that they caution are not directly comparable to climate model-based estimates. When Aldrin et al. include a term for the influences of indirect aerosols and clouds, which they consider to be a more appropriate comparison to estimates such as the IPCC's model-based estimate of ~3°C, they report a sensitivity that increases up to 3.3°C. Their reported value is thus in good agreement with the full body of evidence as detailed in the IPCC report.
Lewis's claimed value of 1.6°C appears nowhere in the paper itself. Rather, Lewis apparently ignored the authors' reported findings in favor of the mode he estimated from graphs in the paper. This misrepresentation both gives a false sense of agreements between the reported senstivity estimates as well as hides the mainstream values reported by the authors of Aldrin et al. These issues are discussed in detail at The Way Things Break along with the relevant figures from the paper.
Most climate sensitivity analyses report the average value rather than the mode, including Alrdin et al. By instead reporting the mode, Lewis is not allowing for an apples-to-apples comparison with most previous climate sensitivity studies. However, this is less of an issue than presenting just one of several climate sensitivity estimates from the Aldrin paper, and one which excludes cloud and indirect aerosol effects.
Beware of Single Study Syndrome
It can be tempting to treat a new study as the be-all and end-all last word on a subject, but that's generally not how science works. Each paper is incorporated into the body of scientific literature and given due weight. It's particularly important not to fall victim to single study syndrome for this type of study, which Olson et al. (2013) documented is subject to large possible biases due to noise from short-term internal variability. As Michael Mann and I noted in a recent article, many people made a similar mistake in giving undue weight to the results of a very similar (not yet published) paper from Aldrin's Norwegian group.
There remains a very large body of evidence consistent with an equilibrium sensitivity most likely close to 3°C surface warming for a doubling of CO2, likely between 2 and 4.5°C. Nic Lewis' result is an outlier, and is inconsistent with past climate changes during many different geologic eras. Another new paper (in press) from Hansen et al. finds that based on paleoclimate data, equilibrium climate sensitivity is likely to be at least 3°C and potentially toward the upper end of the IPCC range, whereas Masters (2013), which is based on recent observational ocean heat content data, suggests the value is toward the lower end of the IPCC range.
It is certainly laudable that Lewis was willing to subject his results to the peer review process, but it's still just one out of many published studies on the subject of climate sensitivity. Every study has its own shortcomings. Paleoclimate estimates are based on past climate conditions which are not identical to those today or in the future, while estimates based on models and recent observations are subject to biases from natural variability and uncertainties such as those associated with aerosols, for example. Ultimately we still can't pin climate sensitivity down better than the IPCC 2–4.5°C range, and we have to be careful not to put too much weight on any single paper.