Statisticians Comment on Status of Climate Change Science
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
Pages: 1 2
(No Ratings Yet)
Loading ...
[...] The March issue of AMSTAT News, the newsletter for the American Statistical Association, has some comments by members of the Climate Change Policy Advisory Group on recent events in the climate change [...]
I was excited to share this article with my students, colleagues and skeptics, but a minor typo like the fact that perihelion represents the point in the earth’s orbit furthest from the sun can really shed some doubt on the credibility of such an article.
Yes they should fix the reference to perihelion, as it is the closest point to the sun. Not that it matters, it is only a reference point for determining the orbit time.
Thank you for all of your comments. The reference to perihelion has been changed to aphelion. If you have additional questions or comments, visit this page again on March 31 at noon (EST) for a live online conversation.
I think that this article could be relevant for this discussion:
Global and hemispheric temperatures revisited
Abstract To characterize observed global and hemispheric temperatures, previous studies have proposed two types of data-generating processes, namely, random walk and trend-stationary, offering contrasting views regarding how the climate system works. Here we present an analysis of the time series properties of global and hemispheric temperatures using modern econometric techniques. Results show that: The temperature series can be better described as trend-stationary processes with a one-time permanent shock which cannot be interpreted as part of the natural variability; climate change has affected the mean of the processes but not their variability; it has manifested in two stages in global and Northern Hemisphere temperatures during the last century, while a second stage is yet possible in the Southern Hemisphere; in terms of Article 2 of the Framework Convention on Climate Change it can be argued that significant (dangerous) anthropogenic interference with the climate system has already occurred.
http://www.springerlink.com/content/h0tx44h508602755/?p=9507c7c71bd044bf9d0c6beb8aa9c0fc&pi=7
What is not discussed in this article is whether the normalized temperature data assembled by CRU (and which apparently was used to normalized the other temperature databases) is itself trustworthy.
It is my understanding that the normalization from the raw temperature data was not sufficiently peer-reviewed. Furthermore, since CRU’s raw database is no longer available (per my understanding), it is impossible to determine what normalization was even done.
One published comparison of raw vs. normalized temperature for (I believe) the Dalton station showed a very anomalous correction which, in my view, was not justified as it grossly overstated the temperature rise and even created a sort of “hockey stick” effect. This one data point calls into question the entire process of normalization.
Since the computer climate models and satellite data are themselves calibrated to the normalized temperature data, the entire conclusion of AGW theory should be called into question until the normalized database is regenerated under full public peer review. This would require, of course, re-acquiring all the world’s raw temperature data for the last century if it doesn’t already exist in one place.
Before we make multi-trillion dollar political decisions that will adversely affect the western world, it is important to have proof which is based upon trustworthy, and full peer-reviewed normalized data.
There is a relevant discussion of statistics starting here (http://ourchangingclimate.wordpress.com/2010/03/01/global-average-temperature-increase-giss-hadcru-and-ncdc-compared/#comment-1216) by a person called VS that I think it would be extremely valuable for the ASA to comment on. Is VS correct? Is the Beenstock and Reingewertz paper being discussed correct? If not, what are the errors?
This is a pure statistics question, rather than involving physics/climatology, so I believe the AMS will have the relevant expertise to comment.
Thanks very much.
Regarding Fred’s link to a post by “VS”:
A detailed debunking of the statistical claims made by “VS” can be found here:
“Not a Random Walk”
http://tamino.wordpress.com/2010/03/11/not-a-random-walk/
and updated here:
http://tamino.wordpress.com/2010/03/16/still-not/
Hi everybody,
Hmm, I missed this activity/thread.
A question to the ASA, just for the record:
“What’s the ASA’s take on the non-stationarity of the instrumental temperature record?”
You all know what that means, and you all know what it implies.
———-
Those mentioning Tamino might want to take a look at my reply to Tamino’s claims. As a matter of fact, I’m inviting the ASA to look at them as well.
http://ourchangingclimate.wordpress.com/2010/03/01/global-average-temperature-increase-giss-hadcru-and-ncdc-compared/#comment-1643
Hamilton, you might want to take a look at this book:
http://www.amazon.com/Time-Analysis-James-Douglas-Hamilton/dp/0691042896
All the best, VS
Also, I would like to take this opportunity to invite *each and every* statistician/econometrician reading this to take a look at this post, and join the discussion with her or his expert opinion (be that critical or supportive). All intelligent review welcomed!
http://ourchangingclimate.wordpress.com/2010/03/01/global-average-temperature-increase-giss-hadcru-and-ncdc-compared/#comment-2740
The debate in this thread has been going on for over month now. There are plenty of test results, monte carlo simulation results (with code) and specification diagnostics posted under that link. I know for a fact that a lot of people are following the discussion.
Furthermore, I believe that the questions adressed are *important*, and the implications for statistical analysis performed in climate-science *severe*.
Looking forward to everybody’s contributions!
Kind regards, VS
PS. Here’s how Josh the cartoonist ‘translated’ the contents of the discussion on Bart’s blog, in case you are in need of an ‘executive summary’
http://www.cartoonsbyjosh.com/unit-root-presence_scr.jpg
Francisco said:
” The temperature series can be better described as trend-stationary processes with a one-time permanent shock which cannot be interpreted as part of the natural variability; climate change has affected the mean of the processes but not their variability; it has manifested in two stages in global and Northern Hemisphere temperatures during the last century, while a second stage is yet possible in the Southern Hemisphere; ”
Actually, I tested the GISS global-mean series for a unit root, with an endogenously determined structural break (turned out to be 1964) in the Ha. This was performed via the so-called Zivot-Adnrews unit root test.
The presence of a unit root is not rejected. Test results posted under the link given above.
Again, I implore all statisticians/econometricians to join the discussion and share their thoughts.
Regards, VS
Dear VS, please be aware that “the Zivot-Andrews tests often diverge or are not invariant to break parameters (Vogelsang and Perron 1998; Kim and Perron 2007)”. For these reasons, the results of this test were omitted in the paper i’m citing.
“These problems are not present in the Perron (1997) test when the break date is selected by minimizing the sum of the squared residuals. Furthermore, Kim and Perron (2007) provide a testing procedure that allows a break in the trend function under both the null and the alternative hypotheses and that has the same limit distribution as if the break date was known. These tests have greater power, maintain
the correct size and simulation experiments have shown that they offer an improvement over other commonly used methods in small samples.”
Even Eric Zivot has stated that his test should not be used.
Best regards
Francisco
Hi Francisco,
Thank you for your valuable comment!
I know (through MC and Stock (1994)) that the PP test has severe size distortions conditional on a ARIMA(3,1,0) w/o drift or ARIMA(0,1,2) w drift that, respectively, me and Breuch and Vahid found.
ADF with SIC/HQ selection is exact (but obviously doesn’t account for potential breaks, I hence also used ZA).
I’ll try to run the MC’s on Kim and Perron (2007) as well.
Again, thanks for the pointer
Best, VS
I suggest that instead of indulging VS in his pseudo-skeptical campaign, you indulge a more credible site run by a statistician:
http://tamino.wordpress.com/
Hi Steven,
If Grant Foster (a.k.a. Tamino) was actually a trained statistician, instead of a ‘home schooled’ civil engineer, you might actually have a point.
However, as the link in my first post here demonstrates, Foster doesn’t know how to test for stationarity (or a structural break, for that matter). One can infer a lack of formal training in statistics from this very piece of information.
In fact, I have yet to find any evidence of his statistics qualifications online (and no, ‘self-proclamation’ is not evidence).
Best if people stick to what they know, don’t you think?
Kind regards, VS
PS. I did enjoy listening to his Bedlam Boys song while explaining to him why you can’t ‘prove’ that unit root testing ‘doesn’t work’ by performing a spurious regression on 34 observations (i.e. his second ‘debunkation’, entitled ‘Still Not’
.
Here’s the song in question: http://www.youtube.com/watch?v=mQIAT4Hh7Jc
And where are VS’s statistics qualifications online (and no, “self proclamation” is not evidence)?
VS, you don’t seem to have looked very hard. Try a search at google scholar (hint, +grant +foster). His top cited paper has 118 cites, pretty good for someone who is not a statistician.
> 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.
Now that you’ve taken this public stance, I hope someone at the ASA is going to watch the comments and not ignore the people claiming to be statisticians who are making claims as statisticians about climate change.
You’ve taken notice of an open can of worms — which is good.
Now, please keep paying attention to the people claiming statistical expertise about climate change who only appear in blogs and videos.
They’re reaching the public with assertions they claim are good statistics.
Your move.
>>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<<
Human beings have industrialized as the earth has come out of the little ice age. As such, it is the warming that is driving human activity, which is driving the release of CO2.
However, the human released CO2 is such a small portion of total released CO2 each year, it could be that the effect is natural. The sun and clouds have been proposed as just such a mechanism. Increased solar radiation reaching the surface heats the oceans, endig the little ice age, releasing more CO2.
ge0050 says: the human released CO2 is such a small portion of total released CO2 each year, it could be that the effect is natural.
This claim is false.
How do human CO2 emissions compare to natural CO2 emissions?
See also: Last time carbon dioxide levels were this high: 15 million years ago
ge5000, yours is a common misconception.
More details at the link, suffice it to say that there are many lines of investigation including isotope studies that show that humans are responsible for the ~30% increase in atmospheric CO2.
These clowns can’s accurately predict the weather a week ahead of time and yet they confidently predict what will happen 100 years from now in a highly non-linear system based on very imperfect computer models that disagree with one another. This is science?
I’ve been reading with interest the comments of VS on various message boards. Aren’t you curious about the identity of VS? There aren’t that many people in the world with that particular profile of specialist knowledge. Some of you are assuming VS is male. After 10 minutes of detective work, I concluded that VS is a female econometrician working at a very prestigious institution. Keep on posting VS. Global warming needs more scrutiny by bona fide experts in random processes.
Your overview of the “hockey stick” appears to imply that the Wegman Report preceded the NRC report. In fact, the NRC reported in March 2006 and the Wegman Report in July 2006.
The Wegman Report raised no new legitimate criticsms and highlighted no new statistical issues, beyond those already in the previously published NRC Report. In fact, even though their mandate was narrower than the NRC’s, Wegman et al failed to address all of the relevant literature concerning the M&M critique. In particular, Wegman et al did not even mention the two critical comments (Huybers and von Storch) on M&M’s GRL article. And they dismissed, without any substantive discussion, Wahl and Ammann’s thorough examination of M&M, which remains to this day the lone peer-reviewed article on the matter.
Wegman et al purported to identify supposed problems of peer review in the field of paleoclimatology, and in general purveyed much misinformation about the field. As well, serious questions have been raised about the scholarship and the process behind the Wegman Report. It stands as a textbook case of the wrong way to involve statisticians in climate science in every respect.
Re-reading the comments (which include some from some DeepClimate regulars), I have a feeling that few have read page 2.
I read the sidebar article, “The Paper that Convinced Me of the Connection Between Carbon Dioxide and Climate Change”, by Peter Guttorp, in the March 2010 issue of Amstat News, with interest. I have been interested in the greenhouse effect since the early 70′s, at which time I was working on a bachelors degree in physics at Reed College, doing research on solar energy. I decided to write comments on Guttorp’s article and ended up writing 11 pages on the mechanics of climate change. I concluded that there is probably too much noise in the precessions of the equinox and perihelion/aphelion to pick out the signal of the greenhouse effect from the precessions in the short time of 120 years. However, I am not familiar with the technique that Guttorp used. My comments can be found at http://www.globalccs.net/~vanstat/climate.htm. The book Storms of my Grandchildren, by James Hansen, who is a climate researcher and much more knowledgeable and experienced than myself in the area of climate research, is a good overview of the causes of climate change. As a disclaimer, Dr. Hansen’s sister is married to my second cousin, who sent my husband and myself the book.
[...] science? And yet the American Statistical Association endorses the statement that there is "strong scientific evidence that climate change is happening and that human activities are the primar… [...]
Welcome!
Follow Us
ASA HOME
Departments
Archives
QUOTABLE
“Statisticians are special because, deep in our bones, we know about uncertainty.”
Statistical Modeling, Causal Inference, and Social Science
ADVERTISERS
MISC. PRODUCTS AND SERVICES
Brigham Young University
Cambridge University Press
Northwestern University
PROFESSIONAL OPPORTUNITIES
Children's Healthcare of Atlanta
Dartmouth College
NORC
U.S. Census Bureau
The University of Kansas Medical Center
University of Connecticut Health Center
The University of North Carolina at Chapel Hill
Westat
SOFTWARE
Cytel Inc.
JMP, a business unit of SAS
Minitab Inc.
NCSS
Salford Systems
SAS
StataCorp
Statistical Solutions
StatSoft
Editorial Staff
Managing Editor
Megan Murphy
Publications Coordinator
Val Nirala
Graphic Designers / Production Coordinators
Melissa Muko
Kathryn Wright
Advertising Manager
Claudine Donovan
Contributing Staff Members
Steve Pierson
Rebecca Nichols
Eric Sampson
Contact us
Amstat News
American Statistical Association
732 North Washington Street
Alexandria, VA 22314-1904
(703) 684-1221
www.amstat.org
Address Changes
Amstat News Advertising