What is there a 97% consensus about?
At the Senate Hearing on “Dogma and Data”, dogma about the 97% consensus went unchallenged. Democratic Senators constantly recited the phrase “97% consensus”, but it is not clear whether they ‚Äì or their Republican opponents ‚Äì had the slightest idea what the phrase meant: 97% of what group support a consensus about exactly what?
The objective of this post is to provide some useful information about this dogma ‚Äì data on the dogma, so to speak.
The most recent study claiming a 97% consensus is Cook et al, Environ. Res. Lett. It was voted the most influential ERL paper of 2013 and was downloaded far more than any other ERL paper. The authors examined the abstracts of 11,944 peer-reviewed journal articles published between 1991 and 2012 to determine the level of scientific consensus for the position “humans are causing global warming”. Unfortunately, the key issue is not whether anthropogenic GHGs have caused any increase in global temperature ‚Äì the issue is how much global warming have they caused.
Quantitative estimates attribution are much more meaningful ‚Äì such as the IPCC AR4 statement that anthropogenic GHGs are “very likely responsible for most global warming”. To understand why, consider the possibility that exactly 50% of global warming has been caused by the rise in anthropogenic GHGs (This is the usual threshold for statements asserting that ‘most’ or ‘more than half’ of the warming is caused by anthropogenic GHGs. ) The remaining 50% of global warming would have been caused by naturally forced variability (solar or volcanic) or unforced (internal) variability. If 50% of warming had been forced by GHGs, climate would be far less sensitive to GHGs than if 100% had been forced.
Otto et al. (2013) studied the relationship between forcing and warming during 1971-2010, a period when this relationship was not significantly perturbed by a change in anthropogenic aerosols, and solar and volcanic forcing was being monitored. (This paper was co-authored by more than a dozen authors of the IPCC chapter dealing with attribution and climate sensitivity.) Assuming that all of the observed warming were attributed to rising GHGs, these authors concluded that the best estimate for equilibrium climate sensitivity was 2.0 degC (with a 95% confidence interval of 1.2-3.9 degC).
If only half of the observed warming were due to rising GHGs (and the rest to unforced variability/chaos), the best estimate for ECS would be 1.0 C. Even under the IPCC’s worst-case scenario ‚Äì a tripling of the pre-industrial CO2 concentration by 2100 the best estimate for total global warming would be only 1.5 C. For all but the most extreme advocates of the precautionary principle, the rational for immediate and massive cuts in CO2 emissions would be fatally weakened. (Extreme concerns could be eliminated by postulating that 40% or 30% of observed warming could be attributed to anthropogenic GHGs.) Consequently, a dramatic difference exists between attributing an unspecified fraction of global warming and attributing most global warming to anthropogenic GHGs.
Recognizing the importance of attributing at least 50% of global warming to humans, Cook et al evaluated abstracts on a scale from 1 to 7 based on how clearly they endorsed or rejected this conclusion. Abstracts in category 1 were supposed to “explicitly state that humans are the primary cause of recent global warming”. Endorsement levels 2 and 3 covered explicit and implicit endorsement of the position that anthropogenic GHGs caused an unspecified amount of warming. Categories 5, 6 and 7 endorsed the opposite position. However, Cook (2013) disclosed only the combined results for categories 1-3 and 5-7.
The ratings by Cook et al can be accessed from the Consensus Project website. Entering “a” as the search term and then choosing a “level of endorsement” provides access to 12,280 rated abstracts in the project database, with the categories 4a (no position) and 4b (uncertain) combined. (I first encountered this strategy for obtaining raw data at Paul Homewood’s blog.) Cook (2013) discusses only 11,944 abstracts, 3% fewer than the project database currently contains. The results in Table 1 therefore represent near the same set of data as used by Cook (2013).
Table 1. Level of endorsement.
After discarding the abstracts that were judged to have taken no position, Cook (2013) reported a 97% consensus that anthropogenic GHGs were causing global warming. However, three-fourths of that consensus was judged to be implied and more than 98% of the agreement was expressed non-quantitatively. Consequently, no widespread consensus exists in these abstracts that humans are responsible for most of the global warming; only that humans are responsible for an unspecified amount of global warming. This consensus on non-quantitative attribution is too vague to be useful in developing policy ‚Äì though it is relentlessly cited by nearly every Democratic policymaker including the President.
The Consensus Project found 65 abstracts endorsing and 10 rejecting quantitative attribution, an “87% consensus” ‚Äì assuming their ratings were accurate. The small number of abstracts involved should not be surprising. Only a few experts are doing the difficult studies needed to quantitatively distinguish between unforced variability and at least four types of forced change: anthropogenic GHGs and aerosols, solar, and volcanic. Quantitative statements about attribution don’t belong in the abstract of papers on other subjects. Abstracts exist to report key findings in context. They are not public opinion surveys.
A tremendous amount of research on climate change has been published during the two decades reviewed by the “Consensus Project”. As can be seen in Table 2, the IPCC’s statements about the role of anthropogenic GHGs in global warming became increasingly quantitative and confident over this period.
Table 2. IPCC statements attributing global warming to anthropogenic GHGs.
|FAR (1990)||“The size of this warming [0 3¬∞C to 0 6¬∞C over the last 100 years] is broadly consistent with predictions of climate models, but it is also of the same magnitude as natural climate variability.”|
|SAR (1995)||“The balance of evidence suggests that there is a discernible human influence on global climate.”|
|TAR (2001)||“Most of the observed warming over the last 50 years is likely to have been due to the increase in greenhouse gas concentrations.”|
|4AR (2007)||“Most of the observed increase in global average temperatures since the mid-20th century is very likely due to the observed increase in anthropogenic greenhouse gas concentrations.”|
|5AR (2013)||It is extremely likely that more than half of the observed increase in global average surface temperature from 1951 to 2010 was caused by the anthropogenic increase in greenhouse gas concentrations and other anthropogenic forcings together. The best estimate of the human-induced contribution to warming is similar to the observed warming over this period.”|
The Consensus Project’s database was analyzed to see if the ratings produced by Cook’s volunteers reflected a similar trend in quantitation and confidence over the two decades considered. The results in Tables 2a and 2b show a constant level of confidence and quantitation. Three hypothesis can be advanced to explain this result: a) The raters may have exhibited a constant bias when categorizing abstracts. b) The authors of the abstracts may disagree with the IPCC’s stronger attribution statements. c) The views of the authors can not be reliably discerned from abstracts. As discussed above, human attribution was usually discussed or implied in the vast majority of these abstracts only to provide context, not to endorse a position. The ratings given to abstracts published by skeptical scientists (discussed below) support this hypothesis.
Table 2a. Change in Endorsement with Time.
|Level of Endorsement||1991-1995||1996-2001||2002-2007||2008-2011||1991-2011|
|(1) Explicit endorsement with quantification||1.2%||1.2%||1.9%||1.7%||1.6%|
|(2) Explicit endorsement without quantification||21%||22%||23%||24%||23%|
|(3) Implicit endorsement||74%||75%||72%||73%||73%|
|(5) Implicit rejection||2.3%||0.9%||3.1%||2.1%||1.3%|
|(6) Explicit rejection without quantification||1.2%||0.6%||0.2%||0.2%||0.4%|
|(7) Explicit rejection with quantification||0.2%||0.0%||0.3%||0.3%||0.2%|
|Total abstracts endorsing a position||431||667||1174||1839||4011|
Table 2b. Change in Percentage of Papers Endorsing a Position.
|Level of Endorsement||1991-1995||1996-2001||2002-2007||2008-2011||1991-2001|
|(4) No Position or Uncertain||56%||64%||70%||68%||67%|
|Endorsing or Rejecting (1)+(2)+(3)+(5)+(6)+(7)||44%||36%||30%||32%||33%|
Any abstract written before AR3 (2001) quantitatively attributing most global warming to humans would have gone far beyond the IPCC’s existing position that the human influence on climate was “consistent with natural variability” or barely “discernable”. (Section 8.6 of the SAR addresses the question: “When Will an Anthropogenic Effect on Climate be Detected? The phrase “a discernible influence” was a late addition.) Before 2007, use of the phrase “very likely” would have indicated more confidence in human attribution than the IPCC believed was warranted. At least two raters in Cook (2013) independently reached the conclusion that five abstracts published between 1991 and 1995 explicitly endorsed the position that anthropogenic GHGs were the primary cause of global warming. These abstracts illustrate ratings misjudgments that might be present elsewhere in the study.
1) Schlesinger, Michael E. and Ramankutty, Navin, “Implications for global warming of intercycle solar irradiance variations”, Nature (1992) 350 (6402), 330-333, [link]
Abstract. Following earlier studies, attention has recently been directed again to the possibility that long-term solar irradiance variations, rather than increased greenhouse gas concentrations, have been the dominant cause of the observed rise in global-mean surface temperature from the mid-nineteenth century to the present. Friis-Christensen and Lassen report a high correlation (0.95) between the variable period of the ’11-year’ sunspot cycle and the mean Northern Hemisphere land surface temperature from 1865 to 1985. The Marshall Institute report concludes that ‘…the sun has been the controlling influence on climate in the last 100 years, with the greenhouse effect playing a smaller role.” Here we explore the implication that such putative solar irradiance variations would have for global warming. Our results provide strong circumstantial evidence that there have been intercycle variations in solar irradiance which have contributed to the observed temperature changes since 1856. However, we find that since the nineteenth century, greenhouse gases, not solar irradiance variations, have been the dominant contributor to the observed temperature changes.
The warming influence of GHGs was found to dominate changes in solar irradiance, but this conclusion certainly doesn’t demonstrate that GHGs were responsible for more than half of global warming. The possibility of unforced variability (and volcanoes) needs to be considered.
2) Roeckner, E., “Past, Present and Future Levels of Greenhouse Gases in the Atmosphere and Model Projections of Related Climatic Changes.” Exp. Bot. (1992) 43 (8): 1097-1109 doi:10.1093/jxb/43.8.1097
Abstract. Ice core analyses of polar ice reveal a high correlation between climatic change and variations in the atmospheric concentrations of greenhouse gases (carbon dioxide and methane) over the last 160 000 years. Although the resolution of the data is not sufficient to determine the phase relationship between the respective variations, it is generally believed that climate change occurred first as a result of the quasi-periodic variations of the Earth’s orbital parameters. However, data and model results are consistent with the hypothesis that climate and atmospheric concentrations of greenhouse gases interact via a positive feedback loop.
The more recent increase in greenhouse gases since pre-industrial times can be related to human activities. Climate models predict a significant global warming of several degrees within the next century if the industrial emissions increase unabated. On the other hand, accelerated policies on emission control will significantly reduce the warming after a response time of a few decades.
This abstract endorses concerns about global warming, but says nothing about quantitative attribution of recent warming.
3) Hansen, J., A. Lacis, R. Ruedy, M. Sato, and H. Wilson, 1993: How sensitive is the world’s climate? Natl. Geog. Soc. Res. Exploration, 9, 142-158.
Abstract. We estimate climate sensitivity from observed climate change on time scales ranging from the 100000-year periods of major ice ages to brief periods of cooling after major volcanic eruptions. The real-world data indicate that climate is very sensitive, equivalent to a warming of 3¬±1¬∞C for doubled atmospheric CO2. Observed global warming of ~0.5¬∞C in the past 140 years is consistent with anthropogenic greenhouse gases being the dominant climate-forcing in that period. But interpretation of current climate change is extraordinarily complex, because of lack of observations of several climate forcings as well as an unpredictable chaotic aspect of climate change. Climate change during the next decade may help confirm knowledge of climate sensitivity, if global climate forcings are accurately observed.
This abstract says that observed global warming is consistent anthropogenic GHGs being the cause, but that attribution was impossible at that time due to lack of information about several climate forcings (presumably aerosols) and the possibility of unforced variability.
4) Tol, R. S. J, ” Greenhouse statistics — time series analysis: Part II”, Theoretical and Applied Climatology (1994), 49 (2), 91-102.
Abstract. The analysis of part I is supplemented, updated and refined, and the resolution bound of simple statistical analysis is tentatively explored. The main conclusion of part I, the hypothesis that the anthropogenically enhanced greenhouse effect is not responsible for the observed global warming during the last century is rejected with a 99% confidence, is reconfirmed for the updated sample period 1870‚Äì1991. The slight decrease in the global mean temperature between 1940 and 1975 is attributed to the influence of El Ni√±o and the volcanic activity. The influence of sunspots, or the length of the solar cycle, is found to be small and unlikely to have caused the observed global temperature rise. The analysis of a number of alternative records lowers the significance of the influence of the enhanced greenhouse effect to 95%. The temperatures on the northern hemisphere rise a little faster than the southern hemisphere temperatures; this distinction is not significant but in line with the larger amount of land at the northern hemisphere. Some indications are found of an unexplained four year cycle in the temperatures of the northern hemisphere. Winter temperatures rise fastest, summer temperatures slowest; this is more profound on the northern than at the southern hemisphere. The difference is not significant; it could be due to the influence of anthropogenic aerosols. The analysis of monthly temperatures confirms the conclusions above, and shows that the models used here are close to being too simple to be used at this resolution.
This abstract does explicitly endorse the position that anthropogenic GHGs are responsible for most global warming. In response to a recent Parliamentary Question, the Chief Scientist of the British Meteorology Office issued several statements explaining why time series analyses (like this one) should not be used alone for attribution, or even detection, of warming: There is no valid way to select a statistical model for the noise that is present in chaotic systems.
5) Mitchell, J. F. B., Johns, T. C., Gregory, M. J., Tett, S. F. B., “Climate response to increasing levels of greenhouse gases and sulphate aerosols”, Nature (1995) 376 (6540), 501-504; doi:10.1038/376501a0
Abstract. Climate models suggest that increases in greenhouse-gas concentrations in the atmosphere should have produced a larger global mean warming than has been observed in recent decades, unless the climate is less sensitive than is predicted by the present generation of coupled general circulation models1,2. After greenhouse gases, sulphate aerosols probably exert the next largest anthropogenic radiative forcing of the atmosphere3, but their influence on global mean warming has not been assessed using such models. Here we use a coupled ocean√°-atmosphere general circulation model to simulate past and future climate since the beginning of the near-global instrumental surface-temperature record4, and include the effects of the scattering of radiation by sulphate aerosols. The inclusion of sulphate aerosols significantly improves the agreement with observed global mean and large-scale patterns of temperature in recent decades, although the improvement in simulations of specific regions is equivocal. We predict a future global mean warming of 0.3 K per decade for greenhouse gases alone, or 0.2 K per decade with sulphate aerosol forcing included. By 2050, all land areas have warmed in our simulations, despite strong negative radiative forcing in some regions. These model results suggest that global warming could accelerate as greenhouse-gas forcing begins to dominate over sulphate aerosol forcing.
This abstract indicates concern about GHGs and future global warming, but doesn’t make any quantifiable or explicit statements about past global warming.
Finally, the Consensus Project Database was queried to determine whether the views of prominent skeptics were reflected in the ratings given to their abstracts. The scientists in Table 4 came from a list of “climate misinformers” posted at John Cook’s website and all have some publications relevant climate change. The results in Table 4 show that opposition to the “consensus” on climate change cannot be clearly detected by rating abstracts. Several explanations for this result are possible: a) Some skepticism arises from an appreciation of uncertainty ‚Äì especially with regard to unforced variability. Abstracts reflecting this position don’t belong in not category 7. b)
The search terms used by Cook (2013) missed many articles by skeptics. Professor Lindzen has written about 20 articles on climate change in the last two decades: The database does not contain his work on the Iris hypotheses nor his two recent articles on low climate sensitivity: Lindzen, R.S. and Y.-S. Choi, “On the determination of climate feedbacks from ERBE data” GRL (2009), 36, L16705 and Lindzen, R.S. and Y.-S. Choi, “On the observational determination of climate sensitivity and its implications” Asian Pacific Journal of Atmospheric Science (2011) 47, 377-390. c) The fact that the latter paper, which rebutted criticisms of the former, appeared in a minor journal is a sign that gate-keeping may be preventing articles from skeptics from appearing in the journals used by Cook. Lindzen’s articles published in Energy and Environment were not in the database either.
Table 4. Abstract ratings for prominent “skeptics” and James Hansen.
Finally, Cook (2013) is often misinterpreted by policymakers and the press as a 97% consensus among climate scientists ‚Äì rather than abstracts. Combining abstracts from Hansen and Lindzen produces an 86% “consensus” (19 to 3 abstracts) from this “community” of two influential climate scientists. In categories 1 and 7 ‚Äì the only ones relevant to policymakers ‚Äì there is a 100% consensus ‚Äì because at least two of Lindzen’s papers weren’t included.
In summary, Cook (2013) contains flaws in conception, implementation and interpretation that invalidate its claim that a meaningful “97% consensus” exists.