Richard L. Smith, University of North Carolina; L. Mark Berliner, The Ohio State University; and Peter Guttorp, University of Washington and Norwegian Computing Center

The authors discussed this article online, live, on March 31, 2010. The discussion can be viewed at the end of the article.

In November 2009, ASA Past-President Sally Morton joined with the leaders of 17 other science organizations to sign a letter (pdf) to all U.S. senators summarizing the consensus of climate change science. In short, the letter cited the strong scientific evidence that climate change is happening and that human activities are the primary driver. It went on to list the many likely consequences, some of which are already starting to occur.

As members of the ASA’s Climate Change Policy Advisory Committee, we commented on early drafts of the letter and, upon reviewing the final version, advised Morton to sign it. We are well aware that some disagree with the statements in the letter. The views of climate change ‘skeptics’ and ‘deniers’ appear in many media, from blogs and videos to op-eds and congressional testimony. We prefer to think of the views of skeptics as part of the scientific spectrum, but nevertheless believe they are a minority who do not represent the mainstream scientific viewpoint.

Some organizations that feature these views in sophisticated advertising campaigns have manipulated the evidence to create the impression that the consensus among climate scientists is quite different from what it is. Here, we comment on some of the most common arguments that climate change is not happening or that humans are not responsible.

Influence of Solar Activity

The idea that solar variability, rather than greenhouse gases, could explain a large part of the observed variability in climate gained considerable publicity due to the 1991 Science paper by Eigil Friis-Christensen and Knud Lassen (see “The Paper That Convinced Me of the Connection Between CO₂ and Climate Change”). This point is still heavily debated. Nicola Scafetta and Bruce West, for example, argue in two recent Physics Today pieces that traditional reconstructions of the solar signal underestimate the influence of solar effects on climate. Their arguments were rebutted in the same publication by Philip Duffy, Benjamin Santer, and Tom Wigley. At the core of the dispute are two reconstructions of the solar intensity signal—each formed by combining data from several satellites—one containing a trend, the other not.

We do not have the expertise to say which reconstruction is to be preferred, but we believe it is important to set such disputes in the broader context of climate research. The relevance of this issue is in detection and
attribution studies in which the climate signal is apportioned over different external forcings, including both greenhouse gases (GHGs) and solar fluctuations. Different constructions of the solar signal could lead to different attributions of GHGs, which, in turn, could affect projections of future climate change in which GHGs are considerably increased.

Nevertheless, there is no credible physical theory that would deny the GHG influence. If the claims made about the solar influence are correct, that could somewhat modify projections of future temperature increases, but there is already plenty of uncertainty about those projections. Therefore, we do not feel controversies about the solar signal should play a major role in the assessment of climate science overall. It is a legitimate area of research to try to quantify the solar signal more accurately, including its uncertainty, and to assess the influence of such uncertainty on future projections.

Connection Between CO₂ and Climate Change

Peter Guttorp, University of Washington, Norwegian Computing CenterFifteen years ago, most scientists had not yet convinced themselves that greenhouse gases led to observable climate change. Indeed, the influence of solar activity was still a viable explanation for the observed increase in average global temperatures, thanks largely to a 1991 Science article by Eigil Friis-Christensen and Knud Lassen that showed temperatures were highly correlated with sunspot numbers.A 1995 Science article by David J. Thomson, titled “The Seasons, Global Temperature, and Precession,” provided the first strong evidence in favor of an observable greenhouse gas effect. Also notable is that the work was based primarily on a careful statistical analysis of the temperature series, rather than on climate models. To me, this paper was the first smoking gun that global warming is connected to the increase of CO₂ concentrations in the atmosphere.There have been many works since to substantiate the connections, but an explanation of Thomson’s paper may be helpful to Amstat News readers.The amount of solar radiation reaching the Earth depends on the angle of Earth’s rotation to the ecliptic (the plane of the orbit) and its distance from the Sun (because the Earth’s orbit is elliptical). The former follows the “tropical” year, the time between two vernal equinoxes, which is 365.2442 days and governs seasons. The distance of the Earth from the sun follows the “anomalistic” year, the time between aphelion (farthest point from the sun) in Earth’s orbit, which is currently 365.2596 days.The interaction between these two cycles so close in time yields long temperature cycles, which result in the quarternary ice ages. The shortest of these cycles is about 26,000 years, very long compared to the instrumental record of temperature.Because these cycles are so close in value, one must use statistical tools to study their influences. Complex demodulation, in effect, removes one of the influencing cycles and looks at the remaining spectrum of a quantity, called the “phase.”
Phase of the Jones-Wigley Northern Hemisphere temperature series (solid) with average phase from 156 northerly land stations (dashed) and the line expected if the anomalistic year frequency dominates the tropical yearWhen one removes the influence of the tropical year, the phase should be flat if the dominant frequency is that of the tropical year. If the anomalistic year is dominant, we would expect to see a linear phase in the residuals with slope equal to the inverse of the difference between the frequencies (57.3 arc seconds per year). The figure above shows the phase of the Jones-Wigley Northern Hemisphere temperature series (solid line), the average phase for 156 stations above 23°N (dashed line) and the dotted line with a slope equal to the precession constant (the rate at which the Earth’s axis rotates).The analysis shows that between 1880 and 1920 the dominant frequency in the temperature series is the anomalistic one. To explain the phase diagram after 1920, a statistician would look at the “residuals,” the difference between the predicted (dotted) line and the estimated phase. The figure below shows the residuals, together with a fit to the logarithm of atmospheric CO₂ levels.
Since the fit is excellent, we have two possible explanations: Either the CO₂ levels are influencing the phase and thus changing the distribution of temperature (i.e., the climate) or there is a common underlying feature driving both the phase change and CO₂ levels. No mechanism has been proposed that can do the latter.Graphics reprinted with permission from The American Association for the Advancement of Science

‘Hockey Stick’ Curve

A considerable amount of controversy has concerned the reconstruction of temperatures over the last millennium, in particular the famous “hockey stick” curve, first published by Michael Mann, Raymond Bradley, and Malcolm Hughes in a 1998 Nature article.

In 2006, Rep. Joe Barton (R-TX), then chair of the House Energy Committee, requested statistician Edward Wegman to re-examine the evidence behind the hockey stick curve. At the same time, the National Research Council (NRC) formed a committee, also containing several statisticians, to report more broadly on the scientific underpinnings of paleoclimatic research.

The report by Wegman (along with David Scott and Yasmin Said) supported 2005 criticisms made by Steven McIntyre and Ross McKitrick in Geophysical Research Letters of the statistical methods used in creating the hockey stick curve and the selection of proxies for the reconstruction. The NRC report was issued in 2006 and affirmed that these questions are relevant, but also pointed out that the evidence for a hockey stick shape was robust and had been reproduced in numerous other studies.

The NRC report also emphasized the inherent uncertainty of such reconstructions, especially concerning the existence of a “medieval warm period” and the estimation of temperature extremes in the early part of the millennium. Since 2006, a number of new studies have appeared proposing alternative statistical techniques for paleoclimatic reconstructions, including assessments of uncertainty and distributions of extremes. Presently, this is a lively area of statistical research on climate change.

In summary, while the Wegman report served to highlight relevant statistical issues, the research community has responded successfully to these challenges.

Temperature Leveling Off?

Another area of controversy has been the apparent leveling off of temperatures post-1998 and the suggestion that greenhouse gas warming has ended. However, this is largely a selection effect arising from the specific choice of 1998 as a starting year. Overall assessments of trend, including the past decade, still show a strong increase in global mean temperatures. Annual temperatures are affected by many sources of variability. For example, it is well known that 1998 was a peak El Niño year. Though the effect of El Niño on temperature varies strongly with season and location, the 1998 El Niño was an extremely warm event.

What are relevant are the long-term trends, not the influence of specific events. From this point of view, the slight decrease of observed temperature post-1998 is of no relevance to assessments of the long-term trend. We also note that reliance on global temperature hides much of the story and effects of climate change. For example, the 2009 average temperature over the southern hemisphere was higher than any observed since at least 1880.

A recent contribution to this topic is that of Susan Solomon and collaborators in Science, who in an article published online in January, showed that increases in stratospheric water vapor may have explained part of the post-1998 cooling.

‘Climategate’

Some have perceived the well-publicized leak of emails associated with researchers at the Climate Research Unit (CRU) of the University of East Anglia in November 2009—dubbed “Climategate”—as weakening the scientific case for climate change. In our view, none of the Climategate revelations change the fundamental science about climate change. To the extent they concern scientific challenges to the theory of anthropogenic global warming (AGW), they involve issues that have already been extensively debated in the scientific literature and elsewhere. Nevertheless, it seems desirable to review some of these issues and indicate our view of them:

• Following Climategate, the view has been widely repeated that climate scientists are secretive about their raw data and suppress data products that are inconsistent with AGW. This is not true. CRU is only one of several centers that have made extensive data archives available to the public; in the United States, the leading such center is the National Climate Data Center (NCDC), a division of NOAA. These data archives contain daily, monthly, and annual data from weather stations around the world, as well as gridded data (e.g., aggregates into 1° and 5° latitude and longitude boxes) that are intended to be directly comparable with the output of climate models. There are also extensive archives of proxy data, such as tree ring and borehole temperature records, that are used in paleoclimatic reconstructions of meteorology before the instrumental record began. Other organizations have created extensive archives of climate model runs, both past data (e.g., 20th century) runs that can be compared directly with the observational record and future model projections that have been at the center of Intergovernmental Panel on Climate Change reports. Click here for numerous sources of data and computer code.
• In one respect, the criticisms made in the Wegman, Scott, and Said report do seem to have been reinforced by the recent Climategate revelations. Wegman analyzed the social network of paleoclimatic researchers, suggesting that their conclusions were suspect because few outsiders had deeply scrutinized the results. The email exchanges have, indeed, shown considerable evidence of a social network in action. However, the more relevant point for the broader picture of climate research is that the conclusions of these researchers have, by now, been scrutinized by many others outside that small community and have largely held up to that scrutiny.
• We do think there are weaknesses in the ways climate data are currently processed. While it would not be reasonable, or even possible, to make every source of raw data available to the public, it is important that raw data sources used in creating public data sets—such as gridded data summaries—be fully documented, as well as the computer algorithms used to create public data sets. In many cases, these algorithms involve subtle statistical judgments (e.g., about outliers, adjustments for missing data, and temporal and/or spatial correlations) that should be open to scrutiny by others, including statisticians who may possibly have ideas for improving them.

Some of the leaked emails suggested that authors tried to keep data secret after publication of the results. Such data secrecy, we believe, is not appropriate in most cases. The Journal of the American Statistical Association (JASA) has a data policy that requires that “Whenever a data set is used, its source should be fully documented. When it is not practical to include the whole of a data set in the paper, the paper should state how the complete data set can be obtained.” Many other scientific journals have similar policies, but it not clear how rigorously any of the policies are enforced in practice. We believe the JASA policy, or something similar, should be the norm in scientific publishing.

The leaked emails of Climategate highlight relevant issues to the assessment of climate research, including data handling and documentation. On broader issues of response to skeptics in such areas as the hockey stick reconstruction, the influence of the solar signal, and the interpretation of the post-1998 leveling off of temperatures, we think climate skeptics have raised legitimate points in each case—several times involving statistical methods or the assessment of associated uncertainties—that should be the focus of further research. However, these points do not change our view that the climate is warming, that anthropogenic influences are likely responsible, and that appropriate mitigation measures need to be considered.

Richard L. Smith is a statistics professor at The University of North Carolina and the Mark L. Reed III Distinguished Professor. Mark Berliner is a statistics professor at The Ohio State University and has held positions at the National Center for Atmospheric Research and National Institute of Statistical Sciences. Peter Guttorp is a statistics professor at the University of Washington, researcher at the Norwegian Computing Center, and director of the Northwest Research Center for Statistics and the Environment.

The following is the transcript of the online question and answer discussion which took place March 31, 2010.

• 12:11 PM: Steve Pierson Welcome to the live Q&A with Richard L. Smith, L. Mark Berliner, and Peter Guttorp, the authors of the March Amstat News article, “Statisticians Comment on Status of Climate Change Science”. Starting at noon EDT, they will answer your questions. ASA Director of Science Policy Steve Pierson will be moderating. Click on the black bar to enter your question. (Note: For carriage returns, hit shift-enter.)The authors will give preference to questions over comments.
• 1:04 PM: Frederick K. Childers [Question 1:] Are you entertaining questions and comments concerning the letter that you recommended Sally Morton sign? If yes, what is the basis for stating climate changes include … greater threats of extreme weather events, increased regional water scarcity, … and western wildfires? The climate change community claims that their computer simulations do not predict regional climate/weather well.
• 1:05 PM: Frederick K. Childers With reference to Susan Solomon’s recent paper: The paper showed a decrease in stratospheric water vapor not an increase. SWith regince water vapor is the most important greenhouse gas (by a significant margin), less water vapor in the stratosphere should result in cooling. The modelers will have to tell us if their calculations agree with the measurements.
• 1:05 PM: Frederick K. Childers I found the discussion of “The Paper That Convinced Me of the Connection Between CO2 and Climate Change” confusing. Whereas the intensity of solar radiation incident on the surface of the earth depends on the angle of Earth’s rotation to the ecliptic, we are concerned with average global temperature which not related to this angle. I found the comment distracting and confusing.
• 1:07 PM: Peter Guttorp @Frederick K. Childers [Question 1 reply:] The issue of regional climate effects has been studied by the US Climate change Science Program. The basis for the statement was one of their synthesis reports
• 1:12 PM: Mark Berliner @Frederick K. Childers : [More on Question 1:]The argument is simple and reasonable over a wide range of quantifications: warming>more precip as rain in fall and spring>faster, sooner snow melt>decreased snow pack>decreased soil moisture>drought>fire
• 1:13 PM: Richard Smith @Frederick K. Childers Comment on Susan Solomon’s paper – yes, thank you for the correction, the paper did refer to a decrease in water vapor after 2000, not an increase.
• 1:14 PM: Guy [Question 2:] You mentioned that the NRC report “pointed out that the evidence for a hockey stick shape was robust and had been reproduced in numerous other studies”, but neglected to mention the Wegman report’s criticisms of those other studies for not being independent. Could you comment on this now?
• 1:19 PM: Richard Smith @Guy [Question 2 reply:] The other studies that we were referring to are studies that have re-analyzed the data essentially taking account of Wegman’s criticisms of the principal components analysis – so the datasets are not independent, but the analyses are new. In particular, Bo Li, Doug Nychka and Caspar Ammann have written two papers using Bayesian hierarchical models – one in Tellus (2007), the other accepted for JASA but currently available (I believe) as a preprint. Other papers in process include one by Tingley and Huybers, and one to appear by Mark Berliner and a student.Richard Smith
• 1:19 PM: petercraigmile [Question 3:] In your article you mention gridded data are “intended to be directly comparable with the output of climate models”. Can you comment further on how you each believe gridded data products are useful in climate analyses.
• 1:21 PM: Peter Guttorp @petercraigmile [Question 3 reply:] The obvious direct issue is to compare the distribution of gridded weather data to the distribution of climate model output (climate, in this context, being the long-term distribution of weather).
• 1:22 PM: Peter Guttorp @petercraigmile [Question 3 reply continued:] The issues are change of support, in addition to weather not being directly comparable to climate model output.
• 1:32 PM: Mark Berliner
• 1:37 PM: PaulM [Question 4:] Climate scientists seem to have invented a new statistic, the “RE” statistic, to replace the “R2” statistic which is used by all other scientists and statisticians. What do you statisticians think of the “RE” statistic? Have you heard of it? 🙂
• 1:38 PM: petercraigmile @Mark Berliner [Question 5 regarding image posted by Mark Berliner:] It is interesting to see how different the tree rings reconstruction differs from the other proxies. Also, why is the borehole reconstruction so smooth?
• 1:39 PM: Fredrick K childers Thank you for the opportunity to “talk” to the authors.
• 1:48 PM: Richard Smith @petercraigmile [Question 5 reply:] This issue is discussed in the forthcoming Li, Nychka, Ammann paper. The basic issue is that it is necessary to smooth over a longer time period to reconstruct temperature from the borehole, especially going further back in time, Therefore, I suspect the smooth shape of the reconstruction is a consequence of the type of data rather thsan a real climatological effect. Bo Li has a list of preprints on her webpage www.stat.purdue.edu though when I tried just now I did not succeed in downloading that paper.
• 1:48 PM: Peter Guttorp @PaulM [Question 4 reply:] Not really. From the description in the paper it seems to be related to the percent variance explained, but the formula given does not make sense to me. I think my pdf file is damaged somehow.
• 1:48 PM: J. Arthur [Question 6:] You say that you are of the view that the climate is warming. What are the main lines of evidence, or statistical analyses, to support that view?
• 1:51 PM: Richard Smith @J. Arthur [Question 6 reply:] The IPCC report (www.ipcc.ch, 2007, working group 1) has a fairly strong statement to this effect in its summary for policymakers (they say evidence for warming is “unequivocal”, whereas the evidence that it is human caused is merely “very likely”). Besides, several of us statisticians have analyzed temperature data from numerous different points of view, and I don’t think any of us doubt that there is an overall warming effect.
• 1:54 PM: J. Arthur @Question 6 Reply [Richard Smith], you say “several of us statisticians have analyzed temperature data from numerous different points of view”; will you give examples of published statistical analyses of temperature data (that support global warming)?
• 1:57 PM: Peter Guttorp @J. Arthur [Question 6 thread:] Here is one
  Trend assessment in a long memory dependence model using the discrete wavelet transform
  Environmetrics
  Volume 15, Issue 4, Date: June 2004, Pages: 313-335
  Peter F. Craigmile, Peter Guttorp, Donald B. Percival
• 1:58 PM: Richard Smith @J. Arthur [Question 6 thread:] One of my own is: R.L. Smith (1993), Long-range dependence and global warming. In Statistics for the Environment, edited by V. Barnett and F. Turkman, John Wiley, Chichester, 141-161. There were a couple of other papers around 1991-1992 by Peter Bloomfield (and one by Bloomfield and Doug Nychka) in the journal Climate Dynamics
• 1:58 PM: Guy [Question 7:] On the IPCC reports: Do you consider the statistical content of the last IPCC report to be adequate? Do you think more statisticians should be involved with the IPCC and other climate science research?
• 1:59 PM: Peter Guttorp [Question 6 thread:] Trend assessment in a long memory dependence model using the discrete wavelet transform
  Environmetrics
  Volume 15, Issue 4, Date: June 2004, Pages: 313-335
  Peter F. Craigmile, Peter Guttorp, Donald B. Percival
• 1:59 PM: J. Arthur @Richard Smith, Thank You!!
• 2:00 PM: Steve Pierson Thank you all for watching and commenting. Thanks also to the authors for their time. Feel free to follow up through Steve Pierson: pierson@amstat.org. We will wrap up shortly.
• 2:00 PM: Guy [Question 8:]On humans causing the warming trend: Do you think the IPCC assessment of this as “very likely” has a good statistical basis? How did they come to this conclusion?
• 2:01 PM: Peter Guttorp [Question 7 reply:] I have spent the last decade or two promoting interest in statistical climatology, and trying to get statisticians involved in IPCC. The history is interesting, but I don’t have time to do it here. However, there is now for the first time a statistician as vice chair of IPCC WG 1 (Francis Zwiers from Canada), More are definitely needed.
• 2:07 PM: Richard Smith @Guy [Question 8 reply:] There is a by now very large field known as “detection and attribution” analysis. See chapter 9 of working group 1 of the 2007 IPCC report (available online through www.ipcc.ch). The essential idea is to decompose the climate signal into (typically) four components, due to greenhouse gases, sulfate aerosols, solar radiation and volcanic effects. There is quite a sophisticated statistical methodology, mostly developed by climate scientists (see e.g. Allen and Stott, Climatic Change 2003) that performs a multiple regression in a high-dimensional space of observations. If the regression coefficient for greenhouse gases is statistically significant, then the effect is said to be “detected”. The practical interpretation of this is that the observed warming cannot be explained by natural forcings alone. When the IPCC says “very likely”, they mean the probability of the effect is more than 90%. They usually interpret this to be the same thing as rejecting a null hypothesis at the 0.1 level of significance. As a statistician, I am not so comfortable with this definition of “very likely”, but it is one very widely used in IPCC and other summary reports.
• 2:09 PM: Steve Pierson Thanks everyone! Now closing!

References

1. Duffy, P.B., Santer, B.D. and Wigley, T.M.L. (2009), Solar variability does not explain late 20th-century warming. Physics Today, January 2009, 48-49.

2. Friis-Christensen, E. and Lassen, K. (1991), Length of the solar cycle: An indicator of solar activity closely associated with climate. Science 254, 698-700.

3. Hegerl, G.C. et al (2007), Understanding and Attributing Climate Change. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (S. Solomon et al, eds), Cambridge University Press, pp. 663-745.

4. Mann, M.E., Bradley, R.S. and Hughes, M.K. (1998), Global-scale temperature patterns and climate forcing over the past six centuries. Nature 392, 779-787.

5. Mann, M.E., Bradley, R.S. and Hughes, M.K. (1999), Northern Hemisphere temperatures during the past millennium: Inferences, uncertainties, and limitations. Geophys. Res. Lett. 26, 759-762.

6. McIntyre, S. and McKitrick, R. (2005), Hockey sticks, principal components, and spurious significance. Geophys. Res. Lett. 32, L03710, doi:10.1029/2004GL021750.

7. National Research Council (2006). Surface Temperature Reconstructions for the Past 2,000 Years. The National Academies Press, Washington, D.C. 2006.

8. Scafetta, N. and West, B.J. (2008), Is climate sensitive to solar variability? Physics Today, March 2008, 50-51.

9. Scafetta, N. and West, B.J. (2009), Interpretations of climate-change data. Physics Today, November 2009, 8-10.

10. Solomon, S. et al. (2010), Contributions of Stratospheric Water Vapor to Decadal Changes in the Rate of Global Warming. Science Express, Published online 28 January 2010 [DOI: 10.1126/science.1182488]

11. Wegman, E.J., Scott, D.W. and Said, Y.H. (2006), Ad hoc committee report on the “Hockey Stick” global climate reconstructions. Committee on Energy and Commerce and Subcommittee on Oversight and Investment, U.S. House of Representatives, Washington, D.C.