RSS feed for comments on this post.

  1. Would I be wrong to think that the West Antarctic ice shelves are under greater stress than previously thought?

    [Response: West Antarctic is a lot colder to begin with than the Peninsula, so we’re not talking about anything catastrophic in the near term. Still, surface melting has increasingly been observed on the West Antarctic ice shelves (see here). If current trends continue (and that is definitely an “if”), then of course we will see more of that. But it is not really significant yet — it is not yet influencing the stability of the ice shelves outside the Peninsula region–eric]

    Comment by Tom G — 21 Jan 2009 @ 1:28 PM

  2. Eric I just spent about 10 minutes searching every possible way and can’t find the article. Can you give the ref?

    [Response: Here. It’s also on the cover. – gavin]

    Comment by Jim Bouldin — 21 Jan 2009 @ 1:33 PM

  3. Any chance the authors can provide a larger version of their figure, properly referenced, that could be used in digital presentations by those of us who talk to the public about climate change?

    [Response: I will put this up on my University of Washington web site early next week and let you know.–eric]

    Comment by Andy Gunther — 21 Jan 2009 @ 2:00 PM

  4. So the Wilkins Ice Shelf is supposedly still held together by a shrinking strip of ice, now about 500 meters wide at its narrowest point. News reports say a break-up expected at any time – 2 weeks to 2 months. (article 1-19-09 )

    Why aren’t there bets on on when this will happen? Catastrophic events have long associated with gambling – and this is one of the rare sudden AGW events. Who is running the pool? I say Feb 14th just for the poetic factor.

    Comment by Richard Pauli — 21 Jan 2009 @ 2:09 PM

  5. Gentlepeople, well done on nipping any controversy in the bud – as usual; though I’m left wondering if the warming trend isn’t related to a subject that i’d like to see Real Climate Address more often; The possible shut-down of The North Atlantic Conveyor – as extreme warming of the Southern Oceans, along with the plunging of Europe into a new Ice Age would be the result of this, as I’m sure you all know.
    I’ve also been wondering, and will now do a little investigating of the issue, whether or not the Level of Oxygen in our Atmosphere has been declining measurably; as it seems to me to be of an issue of Great Concern, now that we’ve cut down so much our Rainforests and are in danger of destroying the ability of Chloroplankton to form the Carbonate shells necessary to their existence, through Oceanic Acidulation.
    Even if you, at Real Climate, are too busy to address these issues; I’ll share what I can find out in a Technorati Blog posting, and – if it seems urgent enough – I’ll send you guys an E-mail as well.
    Love Ya!

    Comment by James Staples — 21 Jan 2009 @ 2:15 PM

  6. “Our paper — by itself — does not address whether Antarctica’s recent warming is part of a longer term trend.”

    Isn’t that because it ignores the ocean component? For example, see this article from Reuters, Feb 2008:

    The paper is here, I believe.

    Satellite altimetry data show a strong increase in sea level in various parts of the Southern Ocean over the 1990s. In this paper, we examine the causes of the observed sea level rise in the region south of Australia, using 13 years of repeat hydrographic data from the WOCE SR3 sections, and the SURVOSTRAL XBT and surface salinity data.

    The hydrographic data show a poleward shift in the position of the Subtropical and the Subantarctic Fronts over the period. In the Antarctic Zone, the Antarctic Surface Water has become warmer and fresher, and the Winter Water tongue has become warmer, fresher, thinner and shallower. Increased freshening south of the Polar Front is linked to increased precipitation over the 1990s. Temperature changes over the upper 500 m account for only part of the altimetric sea level rise. The CTD sections show that the deeper layers are also warmer and slightly saltier and the observed sea level can be explained by steric expansion over the upper 2000 m. ENSO variability impacts on the northern part of the section, and a simple Sverdrup transport model shows how large-scale changes in the wind forcing, related to the Southern Annular Mode, may contribute to the deeper warming to the south.

    Also, isn’t Antarctic sea-ice thickness data incredibly poor? Unlike the Arctic, there is no long-term submarine record of ice thickness – but with the warming Southern Ocean, it seems likely that that has been going on as well.

    Not that it isn’t a good paper that shows the temperature trends, but it’s important (as the postscript admits) that future predictions will depend heavily on getting the ocean component right. It is encouraging that the coupled ocean-atmosphere GCMs give similar results to the data analysis, though.

    Comment by ike solem — 21 Jan 2009 @ 2:28 PM

  7. Tom G (#1)

    It’s hard to make a lot of sense out of “previously thought” because the “old thought” in general was that ice sheets were fairly stagnant on decadal timescales. So it’s true that things are happening that are more interesting than what was “previously thought,” but that isn’t very meaningful. Over the last decade, there has really been a paradigm shift in the understanding and coverage/modeling of the polar areas, and how dynamic they are.

    I recently attended a seminar at UChicago where Michael Oppenhemier gave a talk on ice sheets and sea level rise. A recurring statement was that we have essentially no skill at modeling dynamic processes of accelerated glacier flow, and thus there is low to no predictive ability for either the collapse of ice shelfs or the effects of accelerated flow following ice shelf collapse.

    Comment by Chris Colose — 21 Jan 2009 @ 2:32 PM

  8. Eric — Thank you. Most useful.

    Comment by David B. Benson — 21 Jan 2009 @ 2:44 PM

  9. A beautifully clear contribution.


    Comment by John Gribbin — 21 Jan 2009 @ 2:45 PM

  10. How does the current state per e.g. the uni Bremen map compare to the papers observations. Ross Sea seams to be in a horrible state and big gaps all around the coast lines. Area per CT was a few days ago over a 5 day period above 79-00 mean:

    –2008- 79-00mn Anomaly date, all in km square
    4003000 3716000 287000 jan11
    3917000 3629000 288000 jan 12
    3916000 3552000 364000 jan 13
    3696000 3466000 230000 jan 14
    3624000 3376000 248000 jan 15

    Is the same as for the Arctic, volumetric estimation of the total Sea Ice done, if that at all is of relevance when no longer land fast?

    Surely the Jeff Id’s of this world will be following this thread as he’s come up with these wonderful “play down” equations of global sea ice, which presently is about 600,000 km2 behind 79-00 mean. Already saw an amazing statement of “the current state is right on the 30 year mean” by S. Goddard at WotSuUp, which I interpret as 1979-2008 mean. Ostrich policy of the guys that know the glass was half filled with now vaporised red wine, but when looking from the side the red haze of the former content used, to have the ignorami believe it ain’t so.

    Comment by Sekerob — 21 Jan 2009 @ 2:58 PM

  11. “We have already seen misleading interpretations of our results in the popular press and the blogosphere”

    I saw the NY Times article first
    ‘Warming in Antarctica Now Looks Certain’

    then I went to RC to see if there was a post up. Does the NYT article do a fair job covering the paper?

    [Response: Yes, it’s a pretty good article. Kenneth Chang tends to get these things pretty much right in my experience.–eric]

    Comment by Joseph O'Sullivan — 21 Jan 2009 @ 3:03 PM

  12. This is very interesting.
    Over 31 years, 1969 to 2000, East Antarctica cooled by 0.1 degree-C per decade.
    Over 50 years, 1957 to 2006, East Antarctica warmed by 0.1 degree-C per decade.
    So, must have warmed by an average of 0.2 degrees-C per decade during the
    19 years, (some during the 12 years 1957 to 1969 and some during the 6 years 2000
    to 2006 or all 0.4 degrees-C total during one of those periods).
    What atmospheric hypothesis can account for such behavior?
    Gavin, you are the computer modeler that knows all the scientific laws, theories, and
    hypotheses that have been mathematically entered and weighted in the GCMs.
    Which hypothesis or hypotheses can cause this and what scientific method
    can prove, modify, or disprove any of the individual hypotheses?

    Comment by Gary Plyler — 21 Jan 2009 @ 3:15 PM

  13. Gary Plyler, you’re being too precise in claiming you can isolate 12 year and 6 year trends. Did you read this part of the article above?
    “… Substantial ozone losses did not occur until the late 1970s, and it is only after this period that significant cooling begins in East Antarctica.”

    Comment by Hank Roberts — 21 Jan 2009 @ 3:53 PM

  14. Richard Pauli Said in comment #4:
    “So the Wilkins Ice Shelf is supposedly still held together by a shrinking strip of ice, now about 500 meters wide at its narrowest point. News reports say a break-up expected at any time – 2 weeks to 2 months. (article 1-19-09 )

    Why aren’t there bets on on when this will happen? Catastrophic events have long associated with gambling – and this is one of the rare sudden AGW events. Who is running the pool? I say Feb 14th just for the poetic factor.”

    I thought guys were supposed to prefer breakups before Valentine’s Day for financial reasons.

    Comment by t_p_hamilton — 21 Jan 2009 @ 4:09 PM

  15. Guys,

    A couple of questions:

    1) Wouldn’t it have been more instructive to have shown in Steig et al. (2009) Figure 3b the period, from say, 1969 to present, or 1979 to present, so that we could see the overall behavior in recent decades rather than from a period that ended in 2000 for the sake of comparison to an earlier study? I understand that it is important to compare your new results with previous results to show consistency, but that could have been stated, or perhaps moved into the Supplementary Material. Since, as seen in Monaghan et al. (2008), and as Chapman and Walsh (2007) drew attention to, “Trends computed using these analyses show considerable sensitivity to start and end dates with starting dates before 1965 producing overall warming and starting dates from 1966 to 1982 producing net cooling rates over the region” it would have been interesting to see if such was the case in your analysis. The results of such an analysis would help get a firmer handle on whether or not the predominance of warming in Antarctica took place from the late-1950s to the mid-1960s, or whether it occurred fairly consistently through time. A better understanding of this would help make comparisons to model expectations perhaps a bit clearer.

    2) Your spatial reconstruction procedure does not seem to be pinned down by actual observations. For instance, in your Figure 3a, the observed trend at the South Pole station is given as -0.1ºC/dec while the reconstruction appears to show a positive trend there.

    [Our figure 3a also shows that the observed trend is +0.1C/decade at Vostok, and the reconstruction shows no significant trend at either location.]

    Is this an isolated instance, or are their other observed records with show signs of trends opposite than the reconstruction produces? Your map in Figure 3a (and the cover of Nature) gives the casual reader the perception that all of Antarctica has exhibited a warming trend from 1957-2006, when in fact, it seems that at least one location, direct, local observations indicate otherwise. I understand that your procedure tried to intelligently fill in data where there was none, but in the places that actually do have data, isn’t it best to use that data? It would have been instructive to superimpose an appropriately colored dot at the location of each long-term occupied weather station representing the observed trend (in Figure 3) rather than just the two values included in 3a and 3b.

    3) Why do you need reconstructions of the period of record when you have full satellite coverage? Am I correct to assume that the trend from the satellite data and the trend for the reconstructed temperatures are the same during the satellite data period? And that therefore the slightly higher trend that you all report (as compared with other efforts) comes about from cooler conditions in the pre-satellite era than has been previously determined from the direct observations? Is there an overall trend during the satellite period of record? Maybe that could have been shown in Figure 3 somewhere.

    [Response: Chip, you sure have a lot of suggestions for how I could have or should have done the paper differently. Maybe you should write your own paper. But to answer your question constructively: it is not helpful to splice the raw satellite data with the reconstructed data; this would be very misleading. If you want to look at trends for the raw satellite data alone, see the multiple papers on this in the literature (starting with Comiso, 2000, in Journal of Climate, cited in our paper of course).]

    Clearly Antarctica is supposed to warm in the future as greenhouse levels increase (and ozone loss decreases), but it is unclear just how it should be behaving to date. I am still scratching by head as to whether or not your results clarify that picture.

    The press and “skeptics” alike seem to be quite abuzz by the results (congratulations!), but I can’t quite figure out why as your findings are not that different from what is already out there and since the models are all over the place, it is hard for the observations to support or contradict them. I recognize the hard work that went into your study, the novel methodology, and your contribution to the scientific knowledge base on the issue, but I read your results as more of a tweaking of what is known rather than a large adjustment.

    Thanks for any insights you may have.

    -Chip Knappenberger

    [Response:More thoughts below your later comments, below.–eric]

    Comment by Chip Knappenberger — 21 Jan 2009 @ 4:15 PM

  16. Wilkins ice sheet is an event. Calling it catastrophic may be a stretch.
    I say March 8th.

    Comment by Dave K — 21 Jan 2009 @ 4:34 PM

  17. Geez, you lot are always bearing good news aint ya ;)

    Comment by Alan Neale — 21 Jan 2009 @ 5:37 PM

  18. I looked at that New York Times article, and the one paragraph that stood out was this one:

    Because of the climate record is still short, more work needs to be done to determine how much of the warming results from natural climate swings and how much from the warming effects of carbon dioxide released by the burning of fossil fuels, Dr. Steig said.

    This is a little strange. I do believe that what is being alluded to by Dr. Steig is the role of El Nino cycles in temperature trends across the West Antarctic Peninsula, and how to go about ‘deconvolving’ the relative contributions – is that right?

    The article could have said that we know with certainty that natural climate cycles cannot account for the long-term temperature trends we’ve witnessed, but that these cycles do account for a great deal of year-to-year variability? That would be accurate, right?

    This natural cycle business has been greatly abused in the press, who continue to try and use it as an explanation for global warming, especially in the U.S.

    Take the precipitation situation in the American Southwest, the African Sahel, and in other subtropical regions. Climate models predict expanding Hadley Cells as the climate warms, meaning a poleward expansion of the subtropical desert zones. What I take that to mean is that much of the American southwest will have Sonoran Desert – type climate in the future – when? Not sure. The process seems to be accelerating all across the world, though, and it seems about as robust a prediction as that of rapid polar melting.

    Nevertheless, people are still trying to explain existing trends using combinations of “natural climate cycles.” For another odd example, look at NASA’s JPL press release, Dec 9, 2008, which discusses ENSO and the PDO as causes of local climate change – and doesn’t mention global warming:

    A mechanistic explanation of the PDO, along the lines of the ENSO mechanism, has not been put forward, by the way. Here is the actual PDO chart:

    Now, we know that ocean and atmospheric temperatures have increased since the mid-century – but how has this effected the PDO? Lacking a mechanistic explanation for this time-series analysis-derived cycle, we don’t know – in fact, all the PDO is (so far) is a statistical phenomenon.

    However, JPL says some odd things…

    “This multi-year Pacific Decadal Oscillation ‘cool’ trend can cause La Niña-like impacts around the Pacific basin,” said Bill Patzert, an oceanographer and climatologist at NASA’s Jet Propulsion Laboratory, Pasadena, Calif. “The present cool phase of the Pacific Decadal Oscillation will have significant implications for shifts in marine ecosystems, and for land temperature and rainfall patterns around the Pacific basin.”

    The hot-button topic here is drought across the American Southwest, from California to Texas and even into Georgia. Is it due to “natural climate oscillations”, or are we seeing the predicted outcome of global warming across the region?

    The answer couldn’t be more important for regional water managers – planning for a “return to normal conditions as the cycles progress” could be the most disastrous plan ever – but if you read the press, you won’t find any coverage of it.

    Despite the importance of the issue, there has been a media blackout on the dozens of scientific studies that point to increased drought across the southwest – for example, see the Jan 2 NYT article by Felicity Barringer, which carefully avoids mention of global warming links.

    …“The worry is that La Niña does again what it did last year,” Ms. Lynn said Wednesday, noting that the rainy season, which often lasts through April, ended in February last year. “When we missed March and April, we lost 20 percent of the normal precipitation.”

    I actually called up Ms. Lynn (being a California resident) and asked her if she was familiar with the many studies pointing to a close link between global warming and Southwestern drought – she hung up on me. Her comments have been quoted in about half a dozen stories from the local press, for example Robert Krier in the San Diego Union: “La Nina blamed for more drought.”

    Funny thing, tho… the La Nina prediction is still very questionable, with most models NOT predicting a La Nina, according to the Australians: In fact, we’re still outside the window of reliable ENSO predictability – not that any press articles mentioned that. NOAA says we are in a La Nina cycle, even though the start is unusually late – but their forecast discussion is very if-maybe, not as presented by the press.

    The U.S. press is either woefully ignorant of the state of the science, or is deliberately trying to find explanations for various regional weather and climate changes that don’t involve any mention of “global warming” – and that approach relies on the “natural cycle” argument.

    By the way, the award for least honest website goes to – they’re still promoting the instrumentally flawed Lyman et. al study that showed cooling of the world oceans on their front page.

    Comment by ike solem — 21 Jan 2009 @ 5:55 PM

  19. I thought that, according to the Vostok ice core data,
    the Antarctic has been warming up for the last 12k-yr.

    Comment by Tom in Texas — 21 Jan 2009 @ 5:59 PM

  20. OK Hank,
    I concede the article talked about Ozone. Is that modeled in the GCMs if it is so important?
    If it is included in any of the 19 GCMs, how heavily is that equation weighted in the overall scheme of mathematically represented hypotheses in the GCMs? Is there a way to validate the results and thus use a scientific method to validate individual hypotheses that make up the GCMs?
    I think not.

    [Response: Gary. I have to admit I’m with Hank here. I don’t even understand your questions!–eric]

    Comment by Gary Plyler — 21 Jan 2009 @ 6:24 PM

  21. Though not directly related to this particular thread, I do have reason to write for something very much about Antarctica. I have completed a book, called Our Flooded World, looking at the possible progression of effects that would happen if all ice sheets gradually disappeared through melting. I used extensively for all aspects of the climate science. I did not take a stand on sea level rise rates, other than positing that it is most likely that there will indeed be a one meter rise by 2100. However I did look at what would happen after that in relative, rather than absolute time, by featuring not only changes in geography, but in chapter by chapter fashion looking at trends in human population, agriculture, coastal city infrastructure (all those buried wires!), global transportation infrastructure, the effects of sea level rise on sea level toxic waste dumps (there are many) and the posited future of the further industrialization of China and India, which is to be largely (80% by some estimates) to be coal powered. The assumption is that emissions will not be capped even at 450, but that we hit 1000 ppm sooner or later. The book will be sort of a wet Christmas Carol, by Dickens: the Ghost of Flooding Future is meant to serve as warning, although I am sure that I will be preaching to the choir.

    So – I have surely made errors, and hope that someone might review even a chapter of the book.
    That need is acute, in fact. Today, National Geographic approved a television adaptation of the book as a major special, with a major budget to portray cities gradually being inundated by either storm surge, or eventual sea level rise. It begins with sea level rises of the far and nearer past, something I know something about, and then moves to the current day changes, something I know little about. I told them that the RealClimate honchos would be needed on camera. Here is a chance for this community to have its voice heard – through content. There is still plenty of time – the book is planned for November, and copy editing will continue right up to July. The TV special (or series – a lot to present in 90 minutes) will film this summer as well in all probability.

    Comment by Peter Ward — 21 Jan 2009 @ 6:45 PM

  22. > I think not

    Yes. Best of luck to you.

    “Walking fools”

    Comment by Hank Roberts — 21 Jan 2009 @ 6:46 PM

  23. Eric great work. I look forward to more research on the changing heat balance of the WAIS. I am surprised at how consistent and widespread the warming is across much of WAIS. I am particularly concerned with the warming over the Pine Island and Thwaites Glacier, the week underbelly of the WAIS according to Hughes and Thomas. That is because they are the only main outlet glaciers of WAIS not buttressed by very large ice shelves. They are under the A in West Antarctica in the first figure. Peter send the chapters on ice sheets and melting my way and I will go over them promptly.

    Comment by mauri pelto — 21 Jan 2009 @ 7:27 PM

  24. Congratulations, Peter!

    Comment by Maya — 21 Jan 2009 @ 7:44 PM

  25. > cities gradually being inundated

    Oh dear. Cities change rather fast, what are they going to do to imagine what the cities will look like by then?

    Comment by Hank Roberts — 21 Jan 2009 @ 8:47 PM

  26. I saw this news in an AFP, Agence France-Presse news story in my paper today here in Taiwan. With large photo, too. To me, this all means two things, three things really:

    1. My global warming lawsuit for US$1 billion against all world leaderes today, for manslaughter of all future generations of humankind, must go on, come hell or high water:

    2. My polar cities concept must go forward, if only to pose the question: what if? (with an important AP article on polar cities coming this week)

    3. My “Graduation Speech to the College Class of 2099”, subtitled “Forget Sunscreen, Prepare for Global Warming” must be published in book form ASAP, to wit:

    Comment by Danny Bloom — 21 Jan 2009 @ 8:54 PM

  27. Peter, is your material on-line somewhere?

    Were you able to find much material on the effect of SLR on estuary productivity? Several years ago I saw a paper on this (maybe just an abstract) but wasn’t able to locate it (or any related work) again. IIRC the idea was that productivity is greatly reduced if SLR is fast enough, and that human civilization was held back until sea level stabilized enough for productivity to recover.

    Amd of course thanks for your efforts.

    Comment by Steve Bloom — 21 Jan 2009 @ 9:01 PM

  28. The Sydney Morning Herald, no doubt attempting to reverse The Australian‘s sunny optimism about climate change, is reporting predictions of multi-metre sea level rise by the turn of the century. (Actually slightly more guarded language than this, to be fair.)

    So should I sell my beach house (hypothetically, I don’t own one)?

    Comment by Philip Machanick — 21 Jan 2009 @ 9:02 PM

  29. Interesting theories.

    I just remember winters in McMurdo Station during 1989 1990 and 1991. Nothing like seeing 6″ water lines freeze in a mater of hours. D8 Cat frames break due to cold fatigue, building pylons warp and break… The list goes on.

    40 degrees below zero (C or F, same thing) is still cold enough to freeze human flesh and jet fuel.

    Summers do get balmy there. That is due to our activities of clearing snow, blasting rock, and in general making a mess. But the winters are still brutal, and even worse at Vostock or Amundsen-Scott.

    Palmer always seemed like a planned (but never realized) vacation to the tropics. The Antarctic Peninsula does not strike me as indicative of warming trends.

    Seems odd to think of ancient tropical plants and animals living where nothing can survive today.

    Seems Mother Nature just wants to try to clean up the mess that Humans have made of her planet.

    Lets just see what the next 8 months bring to a continent that we are destroying along with the rest of Earth.

    Comment by Mitch French — 21 Jan 2009 @ 9:11 PM

  30. Eric and Hank,

    Let me please try another way to express why I am not wow’ed by the research (data teasing).

    The average annual temperature at Plateau Station in central East Antarctica is -57 Centigrade (according to my National Geographic Atlas of the World, 7th Edition). Is +0.1 degree Centigrade per decade of any significance? Probably not.

    Comment by Gary Plyler — 21 Jan 2009 @ 9:22 PM

  31. Gary Plyler, it sounds like you’re trying to ask “tough” if not “trick” questions but they come from one of the “skeptic” blog sites, not from your own reading. Just guessing from the phrasing.

    Recommendation: ask questions about what you read, after you’ve read the science as published.

    This paper, for example — it will clear up the reasoning so you can understand why the word “validation” is something of a red herring problem:

    20 January 2009 at 10:40 PM

    Seriously, there’s plenty of science to read, and exercising your ability to ask good questions

    Comment by Hank Roberts — 21 Jan 2009 @ 9:45 PM

  32. [removed at the request of the poster]

    Comment by Anonymous — 21 Jan 2009 @ 9:52 PM

  33. Would be great to see an approximation of the change per decade – obviously it’s a completely different trend from the 70s to the 90s than for the whole five decades, would be interesting to see how different.

    Comment by naught101 — 21 Jan 2009 @ 10:08 PM

  34. A related question. Has anyone tried modeling the pole as a circular disc using the temperature at the center as a measure of average polar temperature, borrowing an idea from harmonic analysis (Average Value Theorem)? Thanks for cite or insight. Daniel

    Comment by daniel — 21 Jan 2009 @ 10:09 PM

  35. You might wish to comment on Horner’s blog response at

    blaming the results reported in Nature on a ” chain of heat-venting sub-sea islands ” , which given the continental trend, seems even more bizarre than invoking volcanic activity in the Arctic ocean

    [Response: err. umm. thanks. Why would I bother? The grip on reality over there is a little tenuous, to put it mildly.–eric]

    Comment by Russell Seitz — 21 Jan 2009 @ 10:26 PM

  36. Would any of your findings put these made on the back of an envelope calculations for Antarctica’s contribution to SLR in doubt?

    Comment by JCH — 21 Jan 2009 @ 10:42 PM

  37. > You might wish ..
    Mud wrestling invitation, eschew. Oh, my eyes.

    > for Gary Plyler

    Wups, sorry — The pointer “20 January 2009 at 10:40 PM” doesn’t work, it should go to the posting of that date by Chris Colose in the ‘FAQ’ thread; see discussion there. I meant to point to this paper, which will help I think:

    Comment by Hank Roberts — 21 Jan 2009 @ 10:46 PM

  38. Russell,

    It’s the Hawaiian Syndrome: The hot spot hypothesis.

    Comment by Mark A. York — 21 Jan 2009 @ 11:29 PM

  39. I’d be interested on your comments on the quality of the AWS records. Take for example the station at Butler Island that you report has a warming trend of 0.45 degrees C per decade. GISS data shows net cooling over the record period (WTF!). But given the paucity of data points any trend appears optimistic. If you examine the station history through the University of Wisconson website there appears to be a long record of malfunctions and changes in site including changes in elevation of over 100m. The station appears to have been buried by snow a number of times. Station Elaine on the Ross Ice shelf is similarly affected.
    How do the problems with the AWS network affect your results? How much time was spent vetting the reliability of the AWS records? Did you accoutn for this in your analysis? Given their unreliability I am wondering why you bothered using AWS at all.

    [Response: The AWS records useful because they provide a totally independent estimate of the spatial covariance pattern in the temperature field (which we primarily get from satellites). I agree with you that in general the data quality at many of the AWS sites is suspect. However, the trends in our results (when we use the AWS) don’t depend significantly on trends in the AWS data (in fact, the result changes little if you detrend all the AWS data before doing the analysis. –eric]

    Comment by MarcH — 22 Jan 2009 @ 12:04 AM

  40. 21 Peter Ward: RE your other book, “Under a Green Sky”, how do you know the sulfur came from the ocean rather than from volcanoes? The oceanographers don’t seem to be impressed. I was impressed/scared. I see that sulfur has 7 isotopes and I guess that that is how you decided.

    5 James Staples: I am concerned about the O2 partial pressure because I am asthmatic. I have heard that in some cases, a small change in O2 partial pressure makes a difference in asthma attacks.

    Dr. Steig: “To read this story in full you will need to login or make a payment (see right).” I can’t read your article without going to a library that subscribes to Nature. Could you post all of it on RealClimate later? Problem with the NASA web site: My computer is too old to run Java.

    Comment by Edward Greisch — 22 Jan 2009 @ 1:11 AM

  41. Any possibility that Solar Dimming of the Eastern Antarctic could account for the temperature differential between the 2 halves?

    Comment by Doc Sief — 22 Jan 2009 @ 1:29 AM

  42. Re #31 response: I see that there is a GISS co-author (Drew Shindell), which means that the extremely efficient staff there has already posted a copy (at the end of the GISS press release here).

    For those wondering about copyright, bear in mind whose copyright law it is and that Nature receives a sweet deal on postage from another division of that same entity. This is something that all U.S. government scientific agencies can do, although many don’t bother at all and only two that I know of (GISS and GFDL) do it systematically.

    Comment by Steve Bloom — 22 Jan 2009 @ 2:17 AM

  43. Hi,

    Just checking – are the trends shown above convergent or divergent?


    Comment by Chris S — 22 Jan 2009 @ 4:00 AM

  44. From RC: 12 February 2008
    “Antarctica is Cold? Yeah, We Knew That”

    “…a cold Antarctica is just what calculations predict…
    and have predicted for the past quarter century…”

    So, when Antarctica “was getting colder” that was exactly what
    the models predicted and no that “it’s been getting warmer” that’s
    according to the models as well?

    [Response: Hmm, the same disingenuous talking point two comments in a row (see comment and response immediately below). I wonder what disinformation outlet is manufacturing this one. We already answered this once below) and, of course, in detail, in the paper itself. So trolls need not bother posting this particular talking point any more. -mike]

    Comment by Sven — 22 Jan 2009 @ 4:22 AM

  45. On many occasions on this site it’s been said that cooling in Antartica is consistent with AGW, as the models show etc…. Now it appears that a warming Antarctica is also consistent with AGW. I am curious to know, is there any kind of change in temperature down there which would invalidate the AGW thesis?

    [Response:Why do the critics think that everything is so simple and binary, for example that we can lump all anthropogenic forcings into a simple “AGW” forcing. Guess what, its not that simple. There are multiple anthropogenic forcings that have quite different impacts (e.g. anthropogenic greenhouse gas increases, aerosols, land-use changes and, yes, stratospheric ozone depletion). Anyone who follows the science is of course aware of this. The temperature trends in Antarctica depend on the time interval and season one looks at, because certain forcings, such as ozone depletion, are particularly important over restricted past time intervals and during particular seasons. The interval over which we expect cooling of the interior is when ozone depletion was accelerating (1960s through late 20th century) and this is precisely when we reproduce the cooling trend both in the reconstruction (primarily during the Austral fall season) and the model simulation experiments discussed in the paper. Over the longer-term, and in the annual mean, greenhouse warming wins out over the more temporary and seasonally-specific impacts of ozone depletion in our simulations, and apparently in the real world. Do you really think that all of the authors and reviewers would have overlooked a basic internal contradiction of logic of the sort you imply, if it actually existed? This is all discussed in detail in the paper. Why not go to your local library and read it and perhaps learn something? -mike]

    Comment by Steve D. — 22 Jan 2009 @ 4:32 AM

  46. Per the map legends, is that 0.25 degrees in 50 years?

    Comment by captdallas2 — 22 Jan 2009 @ 4:41 AM

  47. Dear RC, This looks like it is good science and a trust worthy report and study. The UK Guardian has picked up on the study and posted a good article about it here:

    Beautiful Graphic there.

    GHG or Ozone depletion though I guess is the question or a bit or ratio of both.

    This also beings in the recent change of direction by the skeptics who have used EAIS increased precipitation as a seemingly silly argument along with the new one I seem to find croping up on the notion of water vapour again being a GHG that is a forcing and not just a feedback.

    Water vapour cannot increase itself in the atmosphere can it, hence a feedback ? Only forcing agents can increase water vapour in the atmosphere by essentially adding more energy into it and hence it being able to hold more of it. I doubt that the evaporation rates incrwase, the atmosphere just holds it longer ?

    Comment by Alan Neale — 22 Jan 2009 @ 4:52 AM

  48. Eric
    I note your poit of (justifiable) caution in para 3 of your intro regarding long term trends. But I also wonder why you dont try a little harder to ensure that the popular press dont infer the exactly the opposite, since only bad news is “good news”. Just a cautious scientist’s natural curiosity. Rgds Terry

    [Response: Believe, me, I worked pretty hard on getting reporters to get the right picture. But the popular press doesn’t take the time to get things right. The best reporters try, but they even they are usually (as in this case) on ridiculous deadlines I take weeks to write the first draft of a paper. They take a couple of hours. I kid you not. This is why if you want a reality check, check with RealClimate, not your newspaper.–eric]

    Comment by Terry — 22 Jan 2009 @ 5:15 AM

  49. Fine work. I think you must calculate this extra warmth found into the global whole, or how? How much do your findings rise the global warming during the last 50 years?

    Comment by jorfin — 22 Jan 2009 @ 8:22 AM

  50. daniel — you can’t use the temperature at the pole as representative of Antarctica, because in general, at least in theory, temperatures decline toward the poles — i.e., every other temperature in Antarctica would be higher than at the pole. I’d try integrating the temperature based on the mean at a given latitude band, and the areas of the bands. For comparison, North and Coakley (1979) gives this empirical relation for temperature versus latitude:

    T(theta) = 302.3 K – 45.3 sin^2 (theta)

    North, G.R. and Coakley, J.A. 1979. “Differences Between Seasonal and Mean Annual Energy Balance Model Calculations of Climate and Climate Sensitivity.” J. Atmos. Sci., 36, 1189-1204.

    Comment by Barton Paul Levenson — 22 Jan 2009 @ 8:50 AM

  51. Have noticed that the Ross Ice Shelf seems to have nearly disappeared. Is this normal? If not why hasn’t there been more publicity about it?

    [Response: Huh? Where do you get your information? The Ross Ice Shelf is still very much present. – gavin]

    Comment by dave p — 22 Jan 2009 @ 9:21 AM

  52. Anyone like to comment on this article?

    [Response: Sure. That article is talking about relative cooling of Antarctica compared to the rest of world and the impact of adding dynamic oceans to the models in the 1990s. Spencer’s historical point was very well made – as models got more realistic, the increased ocean thermal inertia simulated meant that the expected trends in Antarctica were much weaker than elsewhere. With weaker expected trends, short term variability will obviously be more important on shorter timescales. This new paper is quantifies that over the 50 year time scale – the trends are positive, but still much smaller than elsewhere. – gavin]

    [Response: Small correction to Gavin’s comment above. Actually not “much smaller than elsewhere” depending on what one is talking about. The mean Antarctic warming over this time period is about the same as the global mean warming. But it is concentrated largely over West Antarctica. East Antarctica is warming considerably less than most other areas. -mike]

    Comment by Calum — 22 Jan 2009 @ 9:35 AM

  53. RE: #15, Chip Knappenberger,

    You wrote: “3) Why do you need reconstructions of the period of record when you have full satellite coverage?”

    But, what satellite data record do you propose to compare with? I suppose you are referring to either the TLT produced by Christy and Spencer at UAH or the similar product from Mears and Wentz at RSS. The TLT from UAH is seriously flawed over the Antarctic, as I demonstrated in a GRL paper some 5 years ago (doi:10.1029/2003GL017938). Here’s a graph of the annual cycle, which is much like that from my paper. Compare that with a similar graph produced from the UAH TMT data. Notice that the TLT has a steep drop in brightness temperature during the early months of the year, a drop not seen in the TMT data. But, the TLT is derived from the the MSU Channel 2, which is also the source of the TMT data. The difference is very curious and was the reason I wrote my paper. In my paper, I also pointed out that data from radiosonds did not exhibit this curious annual cycle. I suggested that the cause of the difference was surface influence, including that from sea-ice.

    Oh, perhaps you want to use the RSS data? But, the folks at RSS understand the problem and do not include any information poleward of 70S. They do this because the MSU channel 2 includes a strong influence from high altitude land areas. RSS also excludes areas over high mountains, such as the Himalayas. One might conclude that there is no other valid “satellite temperature data” available over the Antarctic, so no comparison is possible. But, I’m sure that your side of the “debate” will attempt to do so anyway.

    Comment by Eric Swanson — 22 Jan 2009 @ 9:58 AM

  54. How do these findings square with
    (Sorry, never learnt these Aitch-Tee-Em-El things!)

    [If you look at the figures, you’ll see that we don’t find significant warming at South Pole. This location is not representative of the continent as a whole. Nor is Vostok (where it has been warming, at the same rate South Pole has been cooling). You can’t rely on just one or two sites like this.]

    Seems to show a steady fall of the temperature at the South Pole over a comparable time period.
    And isn’t sea-ice around Antarctica 20% above average?

    [If you look at the paper, you’ll see that we know sea ice is increasing on average, but it has decreased in the Amundsen/Bellinghausen (Pacific) sector where we find the greatest warming]

    Also, why pick 1957 as a start point? Is it simply to select a 50 year period?

    [There are some data but very few before 1957. 1957 was the International Geophysical Year, when most of the weather stations were put in.–eric]

    Is there any good data to extend the study’s period, since man’s been trudging around the pole since the early 1900s?
    I must say, that comment #12 needs answering, if there’s been cooling between 1969 & 2000, (Why has that cut off been set at 2000, why not 2006 as in the study?) but overal a warming between 1957 & 2006, what’s caused it and is the warming been since 2000?
    My PC monitor isn’t good enough to resolve the trend scale, is the maximum about .2 Degrees per decade? If that’s all, then what’s the worry folks? Surely this means it goes from exdeptionally cold, to almost exceptionally cold?

    Comment by Adam Gallon — 22 Jan 2009 @ 10:32 AM

  55. mike

    you say “East Antarctica is warming considerably less than most other areas. -mike]”

    in the paper I read:

    west antarctica: 0.17°C/dec

    east antarctica : 0.10°C/dec

    global antarctica : 0.12°C/dec

    I add: land SH anomaly for the same period (1957-2006) = 0.116°C/dec

    So, If I well understood, the warming trend of east antarctica is a little smaller than elsewhere but not “considerably” smaller.

    [Response: thanks, yes I accept this correction of my correction of Gavin ;) -mike]

    Comment by pascal — 22 Jan 2009 @ 10:37 AM

  56. “dave p”– Gavin asked where you got the picture making you think the Ross ice shelf had almost disappeared. Are you confusing the sea ice with the ice shelf?

    If so this may help, from

    ——excerpt follows——–
    The Ross Ice Shelf

    The Ross Ice Shelf varies in thickness from over 1000 m where it is fed by glacial ice from the ice sheet to less than 100 m at the ice front. On its seaward side it calves icebergs at a rate of about 150 km3 per year, and it also loses volume by bottom melting at a rate of about 2.5 m per year. One of the largest icebergs recorded from the Ross Ice shelf in modern times appeared in the spring of 1987 and measured about 154 km long x 35 km wide.

    Sea Ice

    The sea around the ice shelves is often frozen to form 2m thick sea ice which can extend from 4 x 106 km2 in late summer (February) to 22 x 106 km2 before the thaw begins (September). …

    ——end excerpt——

    Note the sea ice comes and goes:

    Comment by Hank Roberts — 22 Jan 2009 @ 10:42 AM

  57. Re 53:

    Er, Eric, I was referring to the satellite data used by Steig et al. I was interested in the comparison between their reconstruction and their raw data for the period of overlap (~25 years).


    Comment by Chip Knappenberger — 22 Jan 2009 @ 10:53 AM

  58. Erik,

    Quick question regarding the sensitivity of the trend to the starting year: would the trend from 1935-1945 to present still be positive given the large ENSO event in West Antarctica (data permitting, of course)? Did/will you guys release an estimate of average and regional reconstructed Antarctic temperature by month or year that we can play around with?

    Overall, it seems like quite an interesting and important study, especially the results for the West Antarctic region. I’ll withhold complete judgment till I can get through the Nature paywall and read the entire paper, of course!

    On an unrelated note, it looks like the usual suspects in the blogosphere haven’t been entirely dissuaded by this post. Their explanation for West Antarctic warming? Well, apparently it has more volcanoes, and thus must be warming faster independent of any actual trend in vulcanism!

    Comment by Zeke Hausfather — 22 Jan 2009 @ 11:11 AM

  59. I thought the ice shelf was the bit which extended in to the sea. According to the UICC the sea ice has almost gone except for an outer ring. I thought the sea ice at the south of the sea was permanent. Can’t find any comparison maps.

    [Response: There is very little permanent sea ice cover around Antarctica (though there is always some ice to be found). The ice shelf is land ice that has flowed into the ocean and is now floating (and usually hundreds of meters thick). Sea ice forms ‘in situ’ in the ocean is rarely more than a couple of meters thick in the south. – gavin ]

    Comment by dave p — 22 Jan 2009 @ 11:16 AM

  60. Eric & Mike I found the paper very interesting, one thing I found distracting though was the references to the seasons. Sometimes the season name was used sometimes it was prefaced by ‘austral’ with result that I was unsure which season was being referred to.

    Comment by Phil. Felton — 22 Jan 2009 @ 11:21 AM

  61. Mike et al., re: 52 temperature trends

    At the beginning of your paper, you all write “Although the Antarctic Peninsula is one of the most rapidly warming locations on Earth…”—repeating an often used factoid. But later in the paper you all report the warming of the Peninsula as being 0.11+/-0.04ºC/dec, which is just slightly less than the Antarctic average of the same period (0.12+/-0.07ºC/dec), and which you describe (in comment 52) as “The mean Antarctic warming over this time period is about the same as the global mean warming.”

    Thus, it seems that it is not the case that the Antarctic Peninsula is one of the most rapidly warming places on earth (heck, it is warming less than average).

    Perhaps this throwaway line should be thrown away?


    [Response: In our paper we make it clear that we likely underestimate trends, particularly on the Antarctic Peninsula. The weather stations locations on the Antarctic Peninsula definitely do show trends far exceeding those just about anywhere else — as much as 0.5 C/decade (though with large error bars). So, no I don’t think that this line needs to be thrown away. Having said that, it is not clear how representative the weather stations really are of the Peninsula as a whole. Those high numbers come from the coast, where huge declines in sea ice definitely contribute to the warming; but those really high rates warming may not extend very far inland, or very far south.–eric]

    Comment by Chip Knappenberger — 22 Jan 2009 @ 11:25 AM

  62. Alan (47),

    I don’t quite like the whole “skeptical argument refuted” twist on that news article. Pretending that there was a problem before, but now there’s no discrepancies after this new paper provides a misleading perspective.

    The Antarctic trends (with or without the recent Nature paper) were never a contradiction to GW (or the ‘A’ component in GW); certain models (e.g. Shindell and Schmidt 2004) reproduce the observations relatively well as a function of GHG+O3 depletion. Dr. Mann’s comment to #45 is right on-target…you cannot simply lump together all forcings and expect every spot on Earth to warm.

    You are correct about water vapor being a feedback. It’s a feedback because it condenses at Earth-like temperatures and pressures. Similarily, CO2 condenses on Mars, and methane condenses on Titan. The saturation vapor pressure (loosely ‘how much the air holds’) increases almost exponentially with temperature, so if the condensing substance is a greenhouse gas, you can build more up before it rains out.

    Comment by Chris Colose — 22 Jan 2009 @ 12:39 PM

  63. Hank,

    Thank you for the link. I have read the thread on validation. Sorry, but I still don’t see how the GCMs validation process works with accepted scientific method. It would seem to be trying to validate to many hypotheses simultaneously.

    Comment by Gary Plyler — 22 Jan 2009 @ 12:46 PM

  64. dave p, thanks for explaining the source of the confusion.

    The thickness difference between ‘ice shelf’ and ‘sea ice’ is because the ice shelf is glacial (frozen snow) ice, after it’s pushed out to where it floats (beyond the grounding point). The sea ice is what starts to freeze late in February.

    Comment by Hank Roberts — 22 Jan 2009 @ 1:18 PM

  65. What about the last part? How likely is it that if we’re seeing a minimum of sunspots that we’ll get acyclical, lower solar radiation in general?

    Just a couple of things about this. If we’re entering a period of significantly decreased solar activity:

    1) It will be a great way to test climate sensitivity as determined by climate models.

    2) It might buy us some time before we get into some real dangerous climate change.

    Posted by: Miguelito

    Comment by Marion Delgado — 22 Jan 2009 @ 1:31 PM

  66. Comparing both pictures it would look to me as if there was a recent (overall) cooling (1969-2000) and the (overall) warming would have taken place before (1957-1969) Do I get the picture wrong? (thanks in advance)

    *The warming on that period would seem consistent with the sea ice drop (HadISST):

    *You already advised here at RC 4 years ago: “the widely-cited “Antarctic cooling” appears, from the limited data available, to be restricted only to the last two decades, and that averaged over the last 40 years, there has been a slight warming (e.g. Bertler et al. 2004. […]”. :)

    Thank you very much for sharing your knowledge and devoting your time to us.

    Comment by Curious — 22 Jan 2009 @ 2:32 PM

  67. For the Wilkins Ice Shelf bet, I’ll take March 15th, the Ides of March.

    Comment by Maya — 22 Jan 2009 @ 2:35 PM

  68. With the Antarctic temperatures so low I don’t follow why there appears to be so much concern (from others – not by you in this study) about how warming will lead to loss of ice volume or SLR changes. At a degree per century aren’t we looking at a system that is likely very stable in terms of melting?

    [Response: a) Past performance is no guarantee of future returns…. and b) significant ice loss on WAIS is already occurring – though there is an important ocean component to that. – gavin]

    Comment by Joe Hunkins — 22 Jan 2009 @ 2:53 PM

  69. Curious, I think you need to include warming in the 2000-2006 period as well as 1957-1969.

    Comment by Maya — 22 Jan 2009 @ 3:58 PM

  70. Re #58: See #42, Zeke.

    Comment by Steve Bloom — 22 Jan 2009 @ 4:02 PM

  71. While the article is paywalled at Nature, they usually make the Supplementary Information available. It’s here as a PDF

    A chart from the Nature piece appears linked here:
    along with the caption, which I’ll copy and paste:

    a, East Antarctica; b, West Antarctica. Solid black lines show results from reconstruction using infrared satellite data, averaged over all grid points for each region. Dashed lines show the average of reconstructed AWS data in each region. Straight red lines show average trends of the TIR reconstruction. Verification results for the continental mean of the TIR reconstruction are RE = 0.34, CE = 0.31 and r = 0.73. Grey shading, 95% confidence limits.

    Comment by Hank Roberts — 22 Jan 2009 @ 4:16 PM

  72. Gary Plyler,
    In complicated systems with lots of feedbacks it simply is not feasible or informative to separate everything into nice, neat little boxes. Rather, what is done is construct a dynamical model with the best physics (each piece in itself validated) and observe how it does on the trends observed in the real system. This is standard practice for everything from semiconductors to supernovae. I think it is you who needs to update your understanding of the scientific method.

    Comment by Ray Ladbury — 22 Jan 2009 @ 5:31 PM

  73. Gary Plyler, in addition to Ray’s good explanation above, I would add that you are misusing scientific terminology, which may be a reason for your confusion. What is in the GCMs are not hypotheses, but well established laws of physics. The form of verification you are suggesting would bring all attempts to model physical systems to a halt, including such things as launching a rocket. As Ray says, you build the model using sound principles of physics, then see if the result matches what actually happens. The model is never perfect, so you see where you may need to tweak the model, using more sophisticated analysis to better match the real world.

    Comment by Ron Taylor — 22 Jan 2009 @ 6:18 PM

  74. Mike,

    In Spencer Weart’s prior RealClimate article discussing Antarctic cooling, he mentioned a Schneider and Thompson model from 1981 and a Bryan and Manabe model from 1988 that predicted “no warming at the sea surface, and even a slight cooling over the 50-year duration of the experiment”. Both of these came out well before the Thompson and Solomon article on the cooling effect of stratospheric ozone depletion. Weart also stated that “computer models have improved by orders of magnitude, but they continue to show that Antarctica cannot be expected to warm up very significantly until long after the rest of the world’s climate is radically changed”, and implied that this was primarily due to the deep mixing in the Southern Ocean.

    Was Weart only referring to East Antarctica here? If not, how do we reconcile this with your results (that Antarctica as a while is actually warming slightly faster than the southern hemispheric average)? Does this have any implications for the modeled heat transfer rate from ocean mixing in the Southern Ocean, or is that extrapolating things a tad far?

    I understand that there is a lot of nuance here, given all the different forcings and sources of variability at work (GHGs, negative forcing from stratospheric ozone depletion, ENSOs, etc), but I’m genuinely curious what this study implies for past modeling efforts. Were there any models produced post-2002 that included the ozone forcing that closely matched the temperature reconstruction in your article?

    [Response: Zeke. Your query probably deserves a longer answer but a very short answer is that you really should look at the Connolley and Bracegirdle paper summarizing AR4 model results. Some (most?) of those models include ozone. As you’ll see, some of the models match our results rather well. Some don’t.–eric]

    Comment by Zeke Hausfather — 22 Jan 2009 @ 6:51 PM

  75. Chris, 62.

    You’re forgetting a BIG elephant in the room.

    The size of the earth and its systems compared to the energy coming in.

    If it takes 10 years to raise the temperature of the earth 1 degree so that it really IS hotter rather than just a warm surface then a week of over-present GHG before it rains out is practically irrelevant.

    How long it will take could be worked out by figuring out the total mass of earth to, say 10m (so deep that it will stay around long enough even if it’s a clear cold night for months and therefore keep the surface temperature warm enough to matter). Multiply by the specific heat capacity of, say, sandstone.

    Now do a similar thing for the sea (but you have to do probably 100m, or several 10’s of m because light does penetrate ~10m, so you can get “light” *out* from that depth quickly) and multiply by the SHC of water (saline if you like, but we’re going to do a back of the envelope figure here, so why bother).

    Add these together.

    Makes a number X.

    Now how much energy would we get from the sun each day, if 100% were kept in and none radiated?

    Call that number Y.

    Divide X by Y and that gives you how long it would take to warm it up 1 degree in days.

    I don’t know, but I would be suprised if it’s only decades. That’s a LOT of mass to warm up…

    And it’s not going to be 100% capture, is it.

    Comment by Mark — 22 Jan 2009 @ 6:59 PM

  76. I don’t get this 50 year warming but 30 year cooling trend supporting AGW. Surely this must indicate warming in the 50’s and 60’s and cooling therafter which sounds to me like a local effect since it does not correlate with the rapid rise in CO2 since 1970 and has negative correlation with the Northern Hemispheric trends.AmI being stupid.

    [Response: Read the article. Yes, much of this is local. No, you are not being stupid in thinking that it is more complicated that a simple “yes or no” on the question of the relationship with the global picture.–eric]

    Comment by colin Aldridge — 22 Jan 2009 @ 7:32 PM

  77. Any idea when the base data will be made available?

    And regarding the Wilkins ice shelf break-up: it looks like it has already broken off. The latest satellite pic shows the end of the thin ribbon has already been separated from the island as of December 31, 2008. Newer pics show there is new snowfall over the region so it can’t be determined what has happened since the end of the year.

    Comment by John Lang — 22 Jan 2009 @ 9:50 PM

  78. I am somewhat confused by the philosophy behind this method of analysis. I very much appreciate your point that prior analyses which show cooling in Antarctica do not cover a long enough time span to be of much interest to AGW science. 20 years of cooling is nothing to be worried about. This gives scientists a prima facie reason to investigate Antarctic temperature trends via other methods. In particular, your method allows us to see temperature trends over 50 years–which looks like an improvement. However, it strikes me that your analysis still hinges on climate trends from the original ~20 years of data.

    We know that the relationship between temperature and time over the past 20 years does not necessarily reflect a relationship that holds over a longer period of time (nor is the 20 year relationship as important as the longer-term relationship: thus, 20 years of global cooling would not disprove AGW).

    Now, why should a 20 year spatial relationship between Satellite and Surface temperatures hold over a longer period of time? Which in turn implies the question: why should a statistical analysis based on this 20 year relationship allow us to estimate longer-term trends? In order for this analysis to improve upon our understanding of Antarctic cooling/warming, we have to assume that spatial relationships between temperatures in Antarctica are much more reliable over time than are temporal relationships (i.e. are relationships between coastal and interior temperatures, or satellite and surface temperature, more highly correlated over time than are relationships between avg. temperature in year x and avg. temperature in year x+1?–something we can only investigate in the 20 year time span for which we have actual measurements of all relevant data). To make this assumption without rigorous support would implicitly beg the question in favor of both the conclusion in your paper and AGW.

    I understand that the statistical analysis still suggests a statistically relevant 50 year trend, statistical results sometimes overlook the philosophy with which they were put together.

    [Response: You are not thinking about the problem correctly. The shorter dataset is simply being used to provide a means of spatially interpolating the long-term dataset in a ‘smart’ way. The long-term trend is coming from the long-term data. Rather than using parameteric basis functions to perform the spatial interpolation, you are instead using non-parametric basis functions which are eigenfunctions of the covariance matrix of the data. There is a rich literature on this, and you would be well served by reading the paper and the references in the paper that are provided for the underlying statistical methodology, including the various tests that have been done to validate the methodology in the literature. -mike]

    Comment by Colin A — 22 Jan 2009 @ 10:21 PM

  79. Mike – Thank you [edit] I did not immediately see reference to the rich literature on the algorithm used in the paper (other than what appeared at first glance to be other papers on climatology–which though likely very rich, did not immediately suggest that they would contain substantial validation of the method–which tied to my view above, would require a forray beyond just climatology). But it sounds interesting, so I thank you for the opportunity to investigate further.

    [Response: Sometimes you actually have to look at the references cited. If you read the Mann et al ’07 and Schneider ’01 papers cited, you will see that they reference a substantial statistics literature (Tikhonov, Golub, Dempster, Fierro, Little and Rubin, etc.) behind the method used (a regularized version of the classical Expectation-Maximization algorithm). As for validation, there is no such as thing as a generic validation of a method, because the performance of a method is, of course, related to whether or not the assumptions of the method are fulfilled in the context of the type of data being analyzed (are the data essentially Gaussian distributed? Poisson Distributed? Are they discrete, or quasi-continuous? Are they stationary or not? Is the noise ‘white’ or ‘red’ or ‘blue’?). A statistical method that works well for analyzing digital cell phone signals will not necessarily work well in analyzing gappy atmospheric or oceanographic data, and vice versa. In this case, it is climate data that are being analyzed, and the relevant question is, how does the method perform in tests using synthetic data that have the characteristics typically encountered in atmospheric and/or climate data. You will find this addressed not just in the Mann et al ’07 and Schneider ’01 papers, but in many others that are referenced by them. Happy reading! -mike]

    Comment by Colin A — 22 Jan 2009 @ 11:40 PM

  80. Eric, over at the cold weather and astrology blog is a letter to you (text pasted below) from a meteorologist who objects to the paper. His complaints seem to be pretty much either anecdotal evidence or arguments from personal incredulity, with the exception of the final one where he says you allowed the Peninsula weather station readings to contaminate your results for West Anarctica. My impression is otherwise, and indeed it seems a little offensive to accuse a scientist of such a gross error without a specific reference, but your thoughts would be appreciated.

    [Response: Yes, this guy wrote to me. Yet, it is offensive because it is a thinly-veiled accusation of scientific fraud. I’ve spent many seasons in Antarctica over the last 18 years. My personal anecdotal observations don’t bear on the science, and nor should Ross Hays’s. (For the record, last time I was there, it was much warmer than the first time I went, but that’s because I was at lower elevation!)–eric]



    Let me first say that this is my own opinion and does not represent the agency I work for. I feel your study is absolutely wrong.

    There are very few stations in Antarctica to begin with and only a hand full with 50 years of data. Satellite data is just approaching thirty years of available information. In my experience as a day to day forecaster that has to travel and do field work in Antarctica the summer seasons have been getting colder. In the late 1980s helicopters were used to take our personnel to Williams Field from McMurdo Station due to the annual receding of the Ross Ice Shelf, but in the past few years the thaw has been limited and vehicles can continue to make the transition and drive on the ice. One climate note to pass along is December 2006 was the coldest December ever for McMurdo Station. In a synoptic perspective the cooler sea surface temperatures have kept the maritime storms farther offshore in the summer season and the colder more dense air has rolled from the South Pole to the ice shelf.

    There was a paper presented at the AMS Conference in New Orleans last year noting over 70% of the continent was cooling due to the ozone hole. We launch balloons into the stratosphere and the anticyclone that develops over the South Pole has been displaced and slow to establish itself over the past five seasons. The pattern in the troposphere has reflected this trend with more maritime (warmer) air around the Antarctic Peninsula which is also where most of the automated weather stations are located for West Antarctica which will give you the average warmer readings and skew the data for all of West Antarctica.

    With statistics you can make numbers go to almost any conclusion you want. It saddens me to see members of the scientific community do this for media coverage.


    Ross Hays

    [Response: Non-rigorous, muddled rantings from a member of the climate change-denying “Inhofe 400”. – mike]

    Comment by Steve Bloom — 23 Jan 2009 @ 12:18 AM

  81. John Lang wrote in 77:

    And regarding the Wilkins ice shelf break-up: it looks like it has already broken off. The latest satellite pic shows the end of the thin ribbon has already been separated from the island as of December 31, 2008. Newer pics show there is new snowfall over the region so it can’t be determined what has happened since the end of the year.

    You may be right about it having broken off. However, ESA images suggest to my untrained eyes at least that it is still there and in roughly the same condition it was this fall.

    Please see:

    Webcam’ from Space
    Image archive

    In the upper right hand corner you will see “‘Webcam’ from Space: Wilkins Ice Shelf” with a rough hourglass image of the shelf in white, and in the photos for January 19 through January 5 descending you will see the essentially the same image only with the ice shelf in black.

    In any case, this is definitely something to watch. The earlier Larsen B collapsed five degrees to the north at 65 degrees seven years ago in 2002. Five degrees to the south is the end of the West Antarctic Peninsula and the beginning of the mainland.


    Captcha: Reckoning sort

    Comment by Timothy Chase — 23 Jan 2009 @ 12:39 AM

  82. Re #42: Oops, I erred regarding availability of the full text. It’s not on the GISS site yet, but if they follow their usual practice it will be within a few days.

    Comment by Steve Bloom — 23 Jan 2009 @ 12:58 AM

  83. You guys did a very good job on that report.

    5. I agree with you that there should be more discussion on Thermohaline Circulation. A complete collapse of that would be a nightmare.

    Comment by EL — 23 Jan 2009 @ 1:19 AM

  84. I’m following this paper both here and, (Dare one mention them?) wattsupwiththat.
    In view of the statement on watts..

    “In a 2007 press release on Antarctica, NASA’s describes their measurement error at 2-3 degrees, making Steig’s conclusion of .25 degrees Celsius over 25 years statistically meaningless.

    “Instead, the team checked the satellite records against ground-based weather station data to inter-calibrate them and make the 26-year satellite record. The scientists estimate the level of uncertainty in the measurements is between 2-3 degrees Celsius.”

    That is from this 2007 NASA press release, third paragraph.

    Saying you have a .25 deviation over 25 years (based on one-tenth of a degree Celsius per decade per Steig) with a previously established measurement uncertainty of 2-3 degrees means that the “deduced” value Steig obtained is not greater than the error bands previously cited on 2007, which would render it statistically meaningless.”

    Are the findings of this paper statistically significant?

    [Response: The 2-3 deg C uncertainty is an absolute error, it isn’t the same as the uncertainty in the trend (i.e. you can often tell that something has warmed much more accurately than you can tell it’s absolute temperature). The WAIS temperature trend is estimated to be 0.17+/-0.06 deg C/dec (95% confidence) – so yes the changes are statistically significant. – gavin]

    Comment by Adam Gallon — 23 Jan 2009 @ 4:23 AM

  85. The ‘Watts up …’


    site has already commented that this latest ‘warming’ is highly suspect, probably down to observational error, or volcanoes under the ice, or misunderstanding the effects of the Ozone Hole, or scientists being too keen to get out a ‘hot’ story, etc. etc. etc., and certainly ABSOLUTELY NOTHING (of course!) to do with CO2 or Man’s activities.

    Just thought I’d let people know …

    Comment by Nick O. — 23 Jan 2009 @ 6:11 AM

  86. Re #62, the media and science are poles apart (no pun intended)when it come to the nature of AGW. I am just pointing out that every UK broadsheet newspaper(The Times, Daily Telegraph, Guardian and Independent) covered this report and seemingly fairly. Real climate had the DT down as second favourite in their list of supporting deniers (which is in actual fact fairly accurate). Its just that some reports on climate science cannot be twisted although I am sure it will not be long before some people attempt to do so. I will take a look on the web for people attempting to contradict this report and surely it will happen.

    This site was setup to both educate the interested and refute the deniers I guess. Sometimes it does both very well. The contributors names are literally pasted all over the deniers sites and other places so it must be having a great effect.

    Comment by Alan Neale — 23 Jan 2009 @ 8:04 AM

  87. site has already commented that this latest ‘warming’ is highly suspect, probably down to observational error, or volcanoes under the ice, or misunderstanding the effects of the Ozone Hole, or scientists being too keen to get out a ‘hot’ story, etc. etc. etc., and certainly ABSOLUTELY NOTHING (of course!) to do with CO2 or Man’s activities.

    I’ll pay attention when Watts personally dives and takes macro photos of those subsea volcanoes in action …

    After all, his mantra is that a photo is worth 1000 research papers.

    Comment by dhogaza — 23 Jan 2009 @ 8:59 AM

  88. Re (84) and Pielke Jr. here

    …is making some hay from Mr. Steig’s response to Spencer Weart’s earlier post here:

    Comment by bigcitylib — 23 Jan 2009 @ 9:40 AM

  89. dhogaza, do you think that dive will be before or after he looks into the history of paved parking lots at NOAA?

    Comment by Kevin McKinney — 23 Jan 2009 @ 10:27 AM

  90. The paved paradise, put up a parking lot.

    Comment by Mark — 23 Jan 2009 @ 11:08 AM

  91. Re:61,

    Eric, this then brings me back to one of my initial questions, that is, how do the trends from the long-term weather stations compare to your reconstructed trends for the same locations? That is why I thought that perhaps a map with the observed trends superimposed on the reconstructed trends may have been useful.

    Given your response in 61, I take it that you are suggesting that the reconstruction is useful when taken as a whole (as an average), but that for smaller spatial scales (like over the Peninsula or at the South Pole, for example), the reconstructed trends may differ (somewhat sizably in some instances) from the actual trends.

    So, I guess the bottom line is, that your conclusions are that while the reconstruction is not better than the long-term observations at any specific location, that taken as a whole (including many areas with no local, long-term observations) the reconstructed trend better represents what has occurred than do other methods of averaging the long-term records.

    The problem is, from a user’s standpoint, is over what spatial scale (and which locations) is the reconstruction preferable?

    When I suggested that it was preferable over the Peninsula, you suggested I was wrong. But, based on your paper, I presume that you think it *is* preferable over the continent, and I guess even over subcontinental East and West Antarctica. What about smaller scales still? (I am not sure if I am supposed to be able to figure this out from your Figure 1 or not).

    Thanks for your time and guidance.

    -Chip Knappenberger

    [Response:The point of the paper was to get a large-scale averages. Indeed, we essentially ignore local information; any variations that are not correlated on a fairly large spatial scale are treated as noise in the principle component analysis. We would have done a rather different analysis if the point was to get a the very local scale. There are already comparisons in the blogosphere between our results and the South Pole weather station record, and the claim is made that they are “in conflict” (the South Pole record shows cooling over the last 50 years, whereas our record shows no significant trend. [As our paper Figure 3 shows, Vostok, the only long term record in the interior other than South Pole, shows warming; of course, they don’t tell you that on the contrarian web sites]. This kind of comparison is just silly and entirely misses the point. You cannot get at the local scale without local information, which we don’t have (except where we have weather stations of course) prior to 1982. You can of course use the satellite data for post 1982 local information. There is a very nice study that did this — Shuman and Stearns, 2001, in Journal of Climate — in West Antarctica, and demonstrate greater rates of warming that we show (but with larger uncertainties too, and only for a 20-year stretch); they used, incidentally, an independent satellite data set, using microwave rather than infrared frequencies. See Figure 3a in our paper, which shows the locations of their study. The microwave data have a different set of potential biases from the infrared data, so this is very independent support of our results for West Antarctica. Similar local studies such as theirs ought to be done at more locations.
    I hope those comments help you better understand the uses (and potential abuses) of our work.–eric]

    Comment by Chip Knappenberger — 23 Jan 2009 @ 11:19 AM

  92. FAO Eric Steig

    Hi Eric (my first post here),

    I note that you have been sent an email critical of your paper by one Ross Hays which has been reproduced on ‘Watts Up With That?’, in which he references his experience of Antarctic summers as follows:

    “In my experience as a day to day forecaster that has to travel and do field work in Antarctica the summer seasons have been getting colder.”

    In case this should be of any relevance, may I draw your attention to the following blog entry of his, in which he makes clear that he has only experienced one summer visit to Antarctica:

    I realise that this may seem trivial and that, doubtless, you get many emails of such a kind, but given the prominence given to his letter at the “Best Science Blog” I trust that my observation here does not seem inappropriate.

    Comment by Simon Evans — 23 Jan 2009 @ 11:52 AM

  93. Apparently there already is video of one Antarctic submarine volcano:

    ….The research team led by Eugene Domack of Hamilton College in Clinton, N.Y., used a bottom-scanning video recorder, rock dredges and temperature probes to survey the sides and crest of the submarine peak.

    While large areas were colonized by submarine life, none was found on dark rock around the volcano itself, indicating that lava had flowed fairly recently.

    In addition, dredges recovered abundant fresh basalt, a volcanic rock. It normally would be rapidly acted upon and transformed by seawater.

    Highly sensitive temperature probes moving continuously across the bottom of the volcano showed signs of geothermal heating of seawater, according to the agency.

    Domack said the volcano stands 2,300 feet above the seafloor and extends to within roughly 900 feet of the ocean surface.

    The volcano is in an area known as Antarctic Sound, at the northernmost tip of Antarctica. There is no previous scientific record of active volcanoes in the region where the new peak was discovered. The volcano is located on the continental shelf, in the vicinity of a deep trough carved out by glaciers passing across the seafloor….

    [Response: I don’t think anyone is taking exception to the existence of undersea volcanoes or hydrothermal vents. It is more that the idea that they are releasing enough increasing amounts of energy to warm the whole continent of Antarctica is ridiculous. There is first no evidence that their activity has increased over the last 50 years, and more to the point, the energy they release is completely trivial compared to what is required to warm a continent that is thousands of miles across. – gavin]
    [I’ll add another note: There are under-the-ice-sheet volcanoes too of course. But the heat from the volcanoes goes almost entirely into melting ice at the base of ice sheet, a good two miles below the surface. Averaged over the ice sheet are mere milliwatts per meter squared of heat coming up from the ground (even including the volcanoes). That’s 1/1000th of a 60 W light bulb folks…–eric]

    Comment by John Lederer — 23 Jan 2009 @ 12:00 PM

  94. dhogaza Says:
    23 January 2009 at 8:59 AM

    Is that 50mm, fisheye or 500mm telelense for optimal image compression ;?

    He just started part 2, maybe because part 1 has some molto dubbioso in there.

    Momentarily, Antarctic SIA actual is playing fast catch up with 79-00 mean. Would it hit that low this year?

    Comment by Sekerob — 23 Jan 2009 @ 12:10 PM

  95. Is that 50mm

    Oh, I was thinking 50mm, I want him as close to the bubbling vent as possible. And if he falls inside, well, at least he did so proving science wrong, eh?

    Comment by dhogaza — 23 Jan 2009 @ 12:52 PM

  96. Eric Steig (author of this post) is going to be on NPR’s Science Friday today (2009/1/22) in the second hour: 3:00-4:00 U.S. Eastern Time. It’s a call-in show, but you don’t have to use a phone; you can Twitter, e-mail, or visit in Second Life. See the Science Friday site for instructions.

    If you miss it live, you’ll be able to listen to it via the Science Friday web site (or download it as a podcast) by about a day later.

    Comment by Tom Dayton — 23 Jan 2009 @ 1:36 PM

  97. Volcanos notwithstanding, the problems with Mr Watts’ coverage start with the title. ‘ An evolution of viewpoint’, the journalistic hook he has chosen is the ‘shift in scientific opinion’ represented by these results.

    He has two mocked up satellite images of the continent from NASA one showing a predominantly cooling Antarctic from captioned ‘NASA Viewpoint 2004’, another from this year showing an Antarctic dominated by West Antarctic warming, captioned ‘NASA Viewpoint 2009’. the text has … Let’s take a look at how the imagery has changed in 5 years…. the combined effect strongly implying that the new work contradicts the earlier viewpoint and has brought about a shift of opinion from a cooling to a warming Antarctic..

    Incomprehensively the ‘Science Blog of the Year’ fails to mention that: the two images cover different periods, the 2004 picture is of the period 1982-2004 and was lifted from here

    the second is of the 5 decades to present day covered by Steig et al, and included in the NASA press release. Here’s the larger scale version, rather beautiful:….

    So the two images do not illustrate an ‘evolution’ of view about the same thing at all, and the captions inviting direct comparison of the two images are, shall we say, misleading. If only Mr Watts had consulted the author: Our results do not contradict earlier studies suggesting that some regions of Antarctica have cooled. Why? Because those studies were based on shorter records (20-30 years, not 50 years) and because the cooling is limited to the East Antarctic.

    Wouldn’t it be droll if Mr Watts was criticising a paper he has not actually read?

    Comment by John Philip — 23 Jan 2009 @ 2:36 PM

  98. Agh. One of the most disappointing SciFri bits I remember. Blew the intro, corrected it. Ira asked one good question, then cut Eric off in the middle of his answer saying there wasn’t time for it. This was unprepared, no help from whoever they have as the SciFri staff. Not even enough time to do what seemed the plan — have Ira state some old ideas then give Eric time for a few words to say ‘no that’s not right’ or something about what’s new. I think he gave Eric maybe five minutes on the clock, and far less speaking time.

    Dang. I know it’s always disappointing to see what a newspaper does with any story one knows anything about, but SciFri usually seems to do better than this.

    [Response: Hank. Yeah, Ira wasn’t quite on his game today, it seemed. The only really good short answer to “is Antarctica cooling or warming?” is “YES”.–eric]

    Comment by Hank Roberts — 23 Jan 2009 @ 3:36 PM

  99. And SciFri is doing much better with the half hour given to Andy Revkin, going on now.

    Comment by Hank Roberts — 23 Jan 2009 @ 3:57 PM

  100. Note for next time — find out in advance from the staff how many actual seconds of airtime the scientist gets to speak during a typical segment of the planned length. Is there a science fair candidate reading? You could listen to some typical segments of that length with a stopwatch. I’d guess five minutes total on the clock allows maybe 180 seconds to explain the subject. Ow.

    Comment by Hank Roberts — 23 Jan 2009 @ 5:31 PM

  101. Eric, I think you and Hank are being too negative about this. I often listen to SciFri, and I just listened to the podcast of your segment. By comparison with others I have heard, I thought it was actually pretty good. I was concerned when you seemed late in the segment to get diverted into an explanation of the cooling of Eastern Antarctica, but you got back on track near the end, emphasizing the essential point, which is that West Antarctica is warming significantly. This is a general audience with limited time available, so things cannot be explained at a level that would satisfy a scientific conference.

    Comment by Ron Taylor — 23 Jan 2009 @ 9:05 PM

  102. Simon Evans(92), that has been duly noted:

    Comment by bigcitylib — 24 Jan 2009 @ 6:07 AM

  103. [edit] How does the fact that satellite data was not collected prior to 1982 which was during a professed cooling period 1969 – present support that argument that warming was being shown by any thing but the surface stations?

    Further how does this compare with the study:

    Twentieth century Antarctic air temperature and snowfall simulations by IPCC climate models. Andrew Monaghan, David Bromwich, and David Schneider. Geophysical Research Letters, April 5, 2008

    “We can now compare computer simulations with observations of actual climate trends in Antarctica,” says NCAR scientist Andrew Monaghan, the lead author of the study. “This is showing us that, over the past century, most of Antarctica has not undergone the fairly dramatic warming that has affected the rest of the globe. The challenges of studying climate in this remote environment make it difficult to say what the future holds for Antarctica’s climate.”

    The authors compared recently constructed temperature data sets from Antarctica, based on data from ice cores and ground weather stations, to 20th century simulations from computer models used by scientists to simulate global climate. While the observed Antarctic temperatures rose by about 0.4 degrees Fahrenheit (0.2 degrees Celsius) over the past century, the climate models simulated increases in Antarctic temperatures during the same period of 1.4 degrees F (0.75 degrees C).

    The error appeared to be caused by models overestimating the amount of water vapor in the Antarctic atmosphere, the new study concludes. The reason may have to do with the cold Antarctic atmosphere handling moisture differently than the atmosphere over warmer regions.

    That shows that based on physical evidence there was only .2C warming for the century.

    [Response: What the study you are writing about shows is that some models overestimate the amount of warming relative to one data-based estimate of the amount of warming. The data-based estimate is from a paper of mine, showing about 0.2C warming on average. That’s for the entire continent though on average; the West Antarctic warming is greater. It is based on ice cores, which is an inherently conservative estimate because there is generally a seasonal bias due to non-uniform snowfall rates. Borehole thermometry results I’ve seen suggest greater warming than that, in at least two locations in East Antarctica. Watch for publications on this in the next year. I think the assessment in the Monaghan paper that Antarctic warming is overestimated by the models is a premature conclusion.–eric]

    Comment by Vernon — 24 Jan 2009 @ 7:39 AM

  104. I don’t quite get this focus on the last 50 years. Aren’t we most concerned with the latest trend: the one that relates to the steep part of the hockey stick graph, the one that confirms man’s influence on climate, the one that justifies alarm, the one that show consistency with greenhouse gas effect? Shouldn’t we therefore focus on the period for which we have identified a clear manmade global warming signature, hence the period 1975 – present? To attribute the cooling in East Antarctica as a local effect indirectly caused by man induced ozone and hence as atypical for the region ignores the fact that MSU data for the Troposphere of the Polar and ExtraTropic regions of the Southern Hemisphere also show no warming and even some cooling.

    [Response: What you say doesn’t make any sense. If the MSU data show cooling in the recent past over East Antartica, that’s entirely consistent with the surface temperature data and the ozone-related interperetation. For the pre-ozone-hole period, the MSU data agree well with our assessment, as shown in the paper linked to my post (click on ‘troposphere’)–eric]

    Comment by Chris Schoneveld — 24 Jan 2009 @ 9:06 AM

  105. Another episode of Drew Shindell’s “Who should you trust, the models or the data?”

    The winning answer is neither alone but both together, which brings me to my point. Here Eli descends to speculation: As has been emphasized in the paper under discussion there is a warming from global climate change going on in Antarctica and a cooling from ozone depletion. The latter has a HUGE annual cycle which kicked in in the early 80s. Therefore, we can a) trust the modeling of the cooling effect after 1980 when we actually have good, correlated data of the depletion depth and ground level temperatures (well more than anything else) and b) take it as a given that there was only a small cooling effect before say 1975 or so and c) use observed the annual cycle as a measure of the contribution of the two effects.

    This means that we can model what the Antarctic warming would be WITHOUT springtime ozone depletion and will be as the effect of chlorine/bromine loading of the stratosphere decreases.

    Comment by Eli Rabett — 24 Jan 2009 @ 11:06 AM

  106. Ron Taylor, I’d like to think you’re right. I listen to SciFri regularly.

    I’d bet they tried to squeeze in five minutes for this but hadn’t read it. Before the news break Ira said the surprise is it’s cooling not warming); after the news he said he’d gotten it backward. Neither was correct. As Eric said: Cooling or warming? Yes. That much, Ira’s staff should’ve gotten for Ira. Forgivable, likely not much harm done, missed chance. Felt like it dropped on his desk too late.

    Just sayin’ — it’s cautionary. Radio may need preplaning of “elevator talks” — the short phrases that educate.

    [Response: Well said. Exactly: neither of the two ways that Ira put it were right!–eric]

    Comment by Hank Roberts — 24 Jan 2009 @ 12:17 PM

  107. Overall the media coverage of this paper was excellent. I’m not completely sure why, given that lots of papers of equal or greater significance don’t get this degree of attention, but perhaps the difference was that it was the Nature cover story. Interestingly just a day later came the paper on the decline in western North American forests, which got similar attention. That was no surprise given its subject matter, but notably I didn’t see any signs of the two papers crowding each other out.

    Re SciFri, I suspect that the two 20-minute segments were originally planned to be a half-hour and that the two breaking news climate segments (the forestry paper had the following 10-minute chunk) were squeezed in at the last minute. Eric says that Ira wasn’t “on his game” yesterday, but I have to say my impression from listening to many shows is that Ira just isn’t very comfortable with climate stories. I’m not sure why, although part of it may be that when they allow call-ins during climate segments (which they didn’t for either of these) they tend to get lots of the sort of troll with which we are all too familiar in the climate blogosphere.

    Comment by Steve Bloom — 24 Jan 2009 @ 4:39 PM

  108. The paper notes that radiometric surface temperature “Tir” is different from 2-m shelter height air temperature.

    Was any attempt made to correct for this difference ? It’s not clear from the paper that there was and text for Figure 2 simply says the two temperatures actually ‘agree well.’ Physically I don’t see why one would expect them to agree well.

    [Because we are looking at changes through time, there is no need to correct for this difference unless the difference itself changes with time. You can’t easily change the surface temperature without changing the temperature above the surface, due to mixing, so of course they are related. The relationship doesn’t have to be constant, because the amount of mixing can change. But our results indicate that the relationship doesn’t change much — this isn’t a trend in mixing (which would be very remarkable, and a story in itself, if it were true) rather than in actual temperature. If it were, the results would differ between the satellite-based results and the AWS-based results. They don’t differ significantly.–eric]

    Comment by Stuart — 24 Jan 2009 @ 4:52 PM

  109. I agree they should **correlate** over time and trends may be similar, but ‘agree’ meaning ‘approximately equal’ is a stronger condition. As a general rule, of course, surface/skin temperatures don’t equal 2-m air …

    [Response: Eric already explained this once above. We are not working with temperatures but, rather, temperature anomalies. The mean is subtracted off. So a constant offset e.g. such as one might expect between ice surface temperature and a 2m air temperature, would have no influence on the result of the analysis. By showing that we get the same result using both AWS and satellite ice surface temperatures, we have indeed demonstrated that to be the case. –mike]
    [It’s worth adding that this simple point gets overlooked all the time, including by some colleagues of ours that really ought to know better (I won’t name names). It is akin to misunderstanding the difference between precision and accuracy.–eric]

    Comment by Stuart — 24 Jan 2009 @ 5:25 PM

  110. Eric:

    “However, the trends in our results (when we use the AWS) don’t depend significantly on trends in the AWS data (in fact, the result changes little if you detrend all the AWS data before doing the analysis. –eric]”

    What method did you use to detrend the AWS data? I have looked at a number of East Antarctic stations in GISTEMP and all of them show a statistically insignificant warming trend using linear models. I would like to detrend the data to see what impact that has.

    [Response: Errrr… none. Detrending means removing the linear trend, and if it’s small and not significant, then detrending won’t change much. – gavin]

    Comment by Richard Steckis — 24 Jan 2009 @ 10:20 PM

  111. Here again a good piece of work, which gets close to the real meat,
    which is the temperature of the entire troposphere, as opposed to the usual selected from the surface to a few upper levels. Today, High Arctic Canada, Resolute Bay has -30.3 C on the surface with largely clear air, unqualifiedly warm, but 2 C warmer than average. But when you look up, 1100 meters up, its -9 C!

    Incredibly warm for this time of the year. Which makes it obvious, the clear air
    is warming the surface! The tropospheric weighted temperature is 244 K, probably 5 degrees K above a tentative especially small data base.

    Finally if there was a survey of Arctic people throughout the world, there would be near unanimous result, its getting so warm Up Here, the sky is changing…
    Its simply too bad, we dont hear much from the Antarctic long term transient population, from the plumber to the scuba diver, there should be a likewise response just about starting there.

    Comment by wayne davidson — 25 Jan 2009 @ 1:48 AM

  112. Eric,

    I have not read the study but I have read the SI and abstract and I have a couple of questions:

    1. If I read the SI correctly, the satellite data for clouds was removed when it did not within 1 SD of the climate mean. How did you tell the clouds from the ice? Did you develop your model and then remove any cold outliers since they would have to be clouds?

    [The cloud masking is done with *daily* data, and is based on multiple channels to identify clouds. The details of the method are in Comiso, 2001, Journal of Climate. Cloud masking is done *before* anything else. The clouds actually tend to be warmer than the surface, so warm outliers would be removed more often than cold.]

    2. It was not stated in the SI but did test your model on any well documented area to see if the satellite/SST collected the 82-2006 period would then give the correct trend line for the previous 25 years using just the SST data?

    [Not sure what you mean about SST data, which we didn’t use in the analysis. If you are referring to the general circulation modeling, all we did for this paper was look at the already-published results from a 2007 paper in JGR. As for the satellite data, yes of course we compared it with other data in well document areas. If you look at Comiso’s 2001 paper you’ll see the demonstration that it is extremely high fidelity.]

    3. Why is it more important that you start measuring from 1958, a cold point for the century and not from 40′ or 69, 79, 82 which all show cooling trend for the century?

    [Response: If you can find me comprehensive data from Antarctica going back to 1940, I’d be delighted to hear about it. Almost all the data start in 1957. And of course we do show the results for starting in ’69 (Figure 3b), ’79 (Figure 4), and ’82 (Supplementary Information). A very clear point in the paper (if you bother reading it) is that West Antarctica is warming based on any of these starting points. How do you know 1958 is a “cold point” for this century, since there are virtually no data going back prior to 1957? (There are in fact some data going farther back, but just in a few isolated places. The only continuous long record is from the sub-Antarctic island of Orcadas, showing pretty much monotonic warming since 1901. That doesn’t help much though since that is near the Antarctic Peninsula, which everyone already knew was warming. An analysis of the rest of the available data by Jones, 1990 in Journal of Climate shows overall warming since the early 20th century. Those data are simply too discontinuous and sparse for us to have used them in our analysis. None of these data suggest anything special about 1957. Although some people seem bent on suggesting that we “chose” 1957 for some nefarious reason, 1957 was the start of the International Geophysical Year, when most of the weather stations were established; that’s why the weather records generally start then.–eric]

    Thanks for taking the time to look at this.

    Comment by Vernon — 25 Jan 2009 @ 12:26 PM

  113. “[It’s worth adding that this simple point gets overlooked all the time, including by some colleagues of ours that really ought to know better (I won’t name names). It is akin to misunderstanding the difference between precision and accuracy.–eric]”

    That’s worrisome–those are both pretty basic concepts.

    Comment by Jim Bouldin — 25 Jan 2009 @ 1:39 PM

  114. The Sunday Telegraph, a right-wing UK newspaper, has today (25th Jan.) published an article by Christopher Booker about these Antarctic results. It is entitled “Despite the hot air, the Antarctic is not warming up”, see:

    It is riddled with errors and misinformation, in common with all of Booker’s pieces on global warming. Most of the comments following the article are supportive of Booker, with no comments critical of his statements. The reason appears to be that The Telegraph is censoring any such critical comments, which now appear far less frequently than a few months ago.

    I posted a comment that pointed out and corrected eight errors in Booker’s article (see below), but the Telegraph has not published it. Misinformation and errors, followed by censorship: that is the state of one of the main newspapers in the UK at present.

    Errors, briefly:
    1. New evidence contradicts “all previous evidence”.
    2. E. Antarctic cooling “major embarrassment to the warmists”
    3. Antarctica is “source of all the meltwater which will raise sea levels by 20 feet”
    4. Antarctic peninsular “tiny” and only part that is warming.
    5. “The study relied ultimately on pure guesswork”
    6. “Dr Kenneth Trenberth”
    7. “hockey stick” rewrote the scientific evidence.
    8. ““well-established fact that the world was significantly warmer in the Middle Ages than it is now”.

    [Response: This appears to be par for the course for the Telegraph. The last time they messed up they didn’t even allow a correction from the main interviewee (see Ben Goldacre’s column). – gavin]

    [Response: A reminder of why leading environmental journalist George Monbiot (who writes for the Guardian in the UK) has aptly termed Booker the Patron Saint of Charlatans – mike]

    Comment by Slioch — 25 Jan 2009 @ 2:44 PM

  115. There have been a lot of avalanche deaths this year. The snow banks are collapsing at an alarming rate due to the extreme swings in the temps.

    Could this be a prelude to what is in store for the collapse of the ice sheets?

    Freak weather death toll mounts
    Three die in Scottish avalanche, while four children are killed in Barcelona when roof of sports hall collapses in 100mph winds

    [Response: Don’t confuse these two things! Avalanches in Scotland have nothing to do (and I mean nothing) with ice sheets! The scale, the relevant physics, etc. etc. is utterly different.–eric]

    Comment by paulm — 25 Jan 2009 @ 3:12 PM

  116. Correction to my post above. Replace “cooling in latter decades” with “a decline in the warming trend in latter decades”.

    Comment by Richard Steckis — 26 Jan 2009 @ 6:56 AM

  117. > why start …

    Launch of Soviet Sputnik satellite. Cold War concerns support 1957-58 International Geophysical Year, bringing new funding and coordination to climate studies.
    Revelle finds that CO2 produced by humans will not be readily absorbed by the oceans.”

    Weart, AIP History, see first link at right sidebar under Science

    Comment by Hank Roberts — 26 Jan 2009 @ 10:27 AM

  118. Eric, (re: 15, 61, 91)

    I appreciate your time and your explanations.



    [Response: You’re welcome. –eric]

    Comment by Chip Knappenberger — 26 Jan 2009 @ 11:14 AM

  119. Eric,

    Any plans to make the reconstructed data (AVHRR, AWS, and combined) publicly available? I checked your Web site as well as the paper and SI but couldn’t find a link…

    [Response: I’ll have the reconstructed data available shortly on my web site at U. Washington. All of the data that go into this work has always been available on line, through NSIDC (National Snow and Ice Data Center) (for the AVHRR) and through the READER site at the British Antarctic survey (for the weather stations, including the AWS). These raw data were already available before we published the paper. The RegEM code — in the form we used — is (and was) available at T. Schneider’s web site at CalTech. I’ll put up links to all of this when I put our reconstruction on line. We’ll send the reconstruction to NSIDC, a reliable archive, when we have time.–eric]

    Comment by Joe S — 26 Jan 2009 @ 9:10 PM

  120. I apologize for being somewhat off topic (Is there a better place to ask these questions?)- I was reading about a simplified model of ocean circulation which showed that the presence of the Drake Passage increased the magnitude of the coupling between the Antarctic circumpolar winds, Southern Ocean currents, and the Atlantic MOC. Unfortunately, my computer crashed and I can’t find the article again, so my memory of things is probably inaccurate. Anyhoo, my question is, if this coupling does occur, why are we getting more Southern Ocean wind, more energy coupling into ocean circulation, but less MOC? Is the negative drive of fresher North Atlantic water
    ( overriding the positive effects of the increased Southern Ocean wind?

    “Over the past 50 years, wind energy input into the oceans increased 20%, with a large increase in the Southern Ocean and a slight decline in the equatorial band.”

    “At 25 deg N ‘snapshot’ measurements over the past 50 years suggest that the MOC has slowed by 30% and the structure of the overturning circulation has changed so that the southward transport of lower NADW has halved and the southward recirculation of upper waters in the subtropical gyre has doubled.”

    Comment by Brian Dodge — 27 Jan 2009 @ 3:06 AM

  121. For info: there is a letter and comments, published 27th January, in The Scotsman newspaper (Scotland) concerning Ross Hays’s correspondence (ref. #80 above). See:

    [see also Tamino’s amusing take on the famous letter from Hays: here]

    Comment by Slioch — 27 Jan 2009 @ 4:56 AM

  122. Re: 43. Perhaps I was too ambiguous with my question. Huff states that it is worth giving statistical material a “sharp second look before accepting”. The maps above give only trend data, my ‘second look’ (looking primarily at the 1969-2000 map) is to enquire whether West Antarctica is warming towards the mean temp of the East (with the Eastern temp cooling from a higher point) i.e. convergent trends, or whether both sides of the continent were previously closer in their mean temps and are now warming/cooling away from each other (i.e. divergent).

    My feeling is it is the latter but I would welcome clarification (in the full knowledge that there is not likely to be a simple yes/no answer to the question).

    Hope that’s clearer.

    [Response: Chris: West Antarctica is getting warmer, and it is already much warmer than East Antarctica, because it is much lower in elevation. So to the extent that cooling will continue in East Antarctica, I suppose you could say “divergent”. I’m not sure anything is learned by such terminology though: it doesn’t tell you anything about the underlying mechanisms.–eric]

    Comment by Chris S — 27 Jan 2009 @ 7:15 AM

  123. Eric or Mike, some methods questions:

    I don’t understand why any infilling technique is needed, RegEM or otherwise. Can’t you just regress the AVHRR data on the available years of the occupied station data after 1982, apply the relationship (with noise, and accounting for spatial autocorrelation in the AVHRR data) continent-wide to the usable occupied station data, and call it good? (And still test for cloud and inversion biases by incorporating the AWS data as you did) How does the infilling actually generate any necessary information? I’m assuming it’s because the temporal change in the set of usable occupied station data makes the regression process very messy and your method eliminates that?

    Also, it’s not clear to me–is RegEM also used in spatial filling for the masked-out (cloud-covered) areas in the AVHRR data, or only for infilling of the instrumental T data? Or is it using the covariance matrix between AVHRR and occupied stations to fill both simultaneously. What exactly does it do? And the T Schneider paper says it’s appropriate for conditions where the number of variables exceeds the number of records. How is that the case here?

    [Response: Unclear about what you’re suggesting. The ‘infilling’ is nothing other than estimating missing values based on available values, i.e. it is the sort of regression you allude to. However, such multivariate regression problems need to be carefully regularized to avoid overfitting, hence the use of methods such as RegEM. Note that, as shown in Supp Info, we get the same result using the more conventional approach of employing EOFs to infill missing data (this been done to infill gappy instrumental climate records by the UK Met Office, NOAA, and many others for more than a decade). Note also that conditions you cite Schneider for corresponded to his observations using ‘ridge regression’ as the regularization scheme in RegEM. There are a number of papers (including the one we cite in the paper) showing that this does not work well w/ the infilling of sparse data, but that the alternative use of truncated total least squares (TTLS) as a regularization scheme does work quite well, based on independent tests using model simulation data (of course, this what the cross-validation tests in the paper are all about as well). This is what was used, as described in the paper. All infilling was with done w/ the final AVHRR product (i.e. after any cloud masking had already been done). -mike]

    Lastly, what about just using the microwave data, avoiding the AVHRR/cloud masking issue, like Shuman and Stearns, except with the full suite of stations. Unacceptable spatial resolution?

    [Response: I agree with Mike’s points above. Also, we did use the microwave data, but it has the huge problem that it doesn’t see the snow surface — it is effectively seeing about 1 m into the snow, so changes in snow properties create spurious non-temperature noise. Shuman and Stearns had to adjust for this on a case by case basis using the automatic weather station data. The results don’t actually differ much from what we obtained with the AVHRR, but statistics were terrible.–eric]


    Comment by Jim Bouldin — 27 Jan 2009 @ 9:30 AM

  124. Yes I understand that the infilling is for estimating the missing instrumental records. But I don’t see how the process is multivariate in the first place. Because of the multiple bands in the AVHRR data, or the multiple stations in the record? I was assuming the multiple channels get synthesized into surface temperature values via some standard physics equations, in which case you would then just regress the time-averaged AVHRR temperatures against the co-located, non-missing, occupied station data, and apply that relationship (with noise etc) to provide the continental and regional estimates. Or is it the number of occupied stations that makes it multivariate, each considered one variable? But why would one do that? Something basic is not clicking.

    [Response: Have you read the paper? I really think this is very very clear as we wrote it. If it isn’t, read the papers we cite: notably Comiso 2001 and the two or three by Mann and/or Rutherford. Not meaning to be dismissive here; I just don’t understand what you don’t get. –eric]

    Comment by Jim Bouldin — 27 Jan 2009 @ 12:18 PM

  125. Eric, do you think I could ask the type of questions I did without reading the paper and the supplemental information? I’ve also looked at 3 of the references (Schneider, Mann et al, Shuman and Stearns) as much as time would allow me. These are complex, less than familiar techniques that are germane to the results, and most of us don’t have time to trace the background and details of them. If I did, I would. That’s why we ask questions.

    I tried to be as clear as I could. To re-state: (1) what exactly makes the data multivariate, such that a multivariate procedure (a modified RegEM using TTLS instead of ridge regression) is needed, and (2) why is filling in the missing data even necessary in the first place, i.e. why can’t you just regress the co-occurring AVHRR data and the non-missing occupied station data, and then apply that relationship contintent-wide, using the spatial info in the satellite data, to get your large scale T estimates? What purpose does the data filling serve?

    [Response: Jim. Thanks for trying again, especially when I may have seemed dismissive. Now I see what you’re asking. (1) The point of using RegEM, instead of conventional PCA, is that this allows one to account for spatial covariance information both in the predictor data (weather stations) and the predictant data (the satellite data). Any PCA analysis of climate data is just a snapshot of reality. The patterns one gets are hopefully reflecting the climate dynamics, but the longer the time series you use, the more likely that they reflect a meaningful, representative average picture. As we say in the paper, the point of TTLS is to solve the general matrix inverson A = bx where both A and b may be approximations. More typical regressions would assume all the wiggle-room (in the least squares sense) is in b, and that A is perfect. That’s very rarely the case. A = bx is a model. ~A = ~bx is almost always a better model. (2) You could do exactly what you say and the results would be identical. Some temporal infilling would still be necessary in the time series, because they are discontinuous — there are gaps in *all* the records, even the weather stations. But ignoring that, you are right. You don’t really need to do the spatial infilling we did. However, in doing it, it makes the calculation of large scale averages simple arithmetic, and it has the advantage of providing a picture of the spatial weighting of the large-scale averages. I hope that helps more!. –eric]

    [Response: Hmmm. I must confess that I, myself, don’t follow. We’re interested in getting the best estimate we can of the spatiotemporal evolution of Antarctic temperatures over the past 50 years. The belongs to a class of problems in atmospheric science/climatology/oceanography that several different groups have been working on for well over a decade. Namely, given a set of sparse, but long-term climate/atmospheric/oceanic field data and a spatially complete (or at least far more widespread), but short set of data describing the same (or equivalent) field, how do we use the combined information in both datasets to get a best estimate of the full spatiotemporal history of the instrumental field in question? The Mann et al paper cites 18 studies of this type dating back more than a decade by Smith, Reynolds, Kaplan, Rayner, Folland, and other leading researchers interested in this question. All use some variant on the basic principle we’re using, i.e. using eigenvectors of the data covariance matrix (whether it be simple PCA, Regularized Expectation-Maximization, or some other variant). What is it, Jim, that you think you have come up with that solves the problem in a much simpler way and avoids having to work with data covariances and other messy entities? I can promise you there are literally teams of scientists around the world who would like to learn. So please spell it out for us, if you would. Thanks! –mike]

    Comment by Jim Bouldin — 27 Jan 2009 @ 10:56 PM

  126. Thanks Eric, I appreciate your efforts to explain and clarify. I’m interested in these techniques because of their wider possible applicability to other types of ecological/environmental analysis where missing data are common and satellite imagery can help fill in the blanks.

    Mike, [edit] you’re reading things that aren’t there. These are QUESTIONS, not accusations, about the methods used. I am simply trying to understand what was done, not argue that I have a better way, because I’m highly interested in methodological issues in general, including how they can potentially be used in ecological research where spatial variability and missing values are common issues. [edit]

    [Response: Jim, perhaps time to tone this down. It did objectively seem to me that you were implying that the climate research community has somehow missed a trivial solution to a difficult problem, and instead has pursued unnecessarily technical and overly complex approaches to the problem at hand. I feel as if Eric and I provided you with a lot of information, and all of the relevant literature, hence both of us expressed some frustration with your continued questioning along seemingly similar lines. If I am mistaken in my interpretation, then I most definitely apologize. I would still however suggest you consider phrasing things in such a way so as to make such a misinterpretation less likely. We’re quite open here to honest questioning when it doesn’t come across as leading or overly aggressive. –mike]

    Comment by Jim Bouldin — 28 Jan 2009 @ 12:19 AM

  127. Mike, I’m aware that there is a high degree of sophistication in some climate science methods–it’s clear just by reading any random part of the literature–which is why I am even interested in them in the first place. You can be 100% sure there was no ill intent in my questions, however they may have come across to you. I’ve looked back at my original post and cannot honestly see how this could have come across as accusatory, so don’t know where to go with that. You may feel the methods are obvious because you have been working with them for a while–I can relate to that–but they are less than obvious to many of us who have not been, including those with a decent statistical background. [edit]

    Comment by Jim Bouldin — 28 Jan 2009 @ 7:18 PM

  128. From the innumerate gallery, I appreciate y’all working to get past the raised-hackles (very human) responses. And text conveys what, five or maybe ten percent of meaning, the rest we get from body language or when it’s missing, our brains, er, interpolate the missing data (grin).
    Which is hard to do.

    Pray keep talking. I won’t understand the math but I seriously do understand the effort it takes competent people to make progress in an area like this with this medium. And greatly appreciate the effort.

    Comment by Hank Roberts — 29 Jan 2009 @ 12:13 AM

Sorry, the comment form is closed at this time.

Close this window.

0.282 Powered by WordPress