RealClimate logo

West Antarctica: still warming

The temperature reconstruction of O’Donnell et al. (2010) confirms that West Antarctica is warming — but underestimates the rate

Eric Steig

At the end of my post last month on the history of Antarctic science I noted that I had an initial, generally favorable opinion of the paper by O’Donnell et al. in the Journal of Climate. O’Donnell et al. is the peer-reviewed outcome of a series of blog posts started two years ago, mostly aimed at criticizing the 2009 paper in Nature, of which I was the lead author. As one would expect of a peer-reviewed paper, those obviously unsupportable claims found in the original blog posts are absent, and in my view O’Donnell et al. is a perfectly acceptable addition to the literature. O’Donnell et al. suggest several improvements to the methodology we used, most of which I agree with in principle. Unfortunately, their actual implementation by O’Donnell et al. leaves something to be desired, and yield a result that is in disagreement with independent evidence for the magnitude of warming, at least in West Antarctica.

In this post, I’ll summarize the key methodological changes suggested by O’Donnell et al., discuss how their results compare with our results, and the implications for our understanding of recent Antarctic climate change. I’ll then try to make sense of how O’Donnell et al. have apparently wound up with an erroneous result.

First off, a reminder for those not familiar with it: the essential innovation in our work was to combine the surface temperature data available from satellites with the ~50 years of data from weather stations. The latter are generally considered more reliable and go back a full 50 years, but are very sparse and incomplete, whereas the satellite data provide complete spatial coverage of the continent, but only since the early 1980s. We combined the two data sets by calibrating the weather station data against the satellite data, and using the calibration to get a complete spatial picture of Antarctic temperature variability and trends for the last 50 years. The key findings were that the overall Antarctic trend was positive (but not necessarily statistically significant), and that in West Antarctica, the trends were both positive and significant, especially in winter and spring. These findings were important enough for Nature to publish them because most researchers thought that significant warming was restricted only to the Antarctic Peninsula region. None of these findings is contradicted by O’Donnell et al.’s results.

O’Donnell et al. have three main criticisms of our work. First, that the reconstruction we reported was not homogenous. That is, the first part of the reconstruction (1957 through 1981) is based entirely on a linear combination of weather station data (since there are no satellite data during that period); while the second part (1982-2006) is derived simply from the satellite data. O’Donnell et al argue that it would be better to use the only weather station data for both periods, since these data are a priori considered more reliable. (There are all sorts of potential problems with the satellite data, the chief one being that there is a ‘clear sky’ bias.) That is, one wants to calibrate the data during 1982-2006, and then use that calibration to model the temperature field for both the early and the later periods, using only the weather stations.

Second, that in doing the analysis, we retain too few (just 3) EOF patterns. These are decompositions of the satellite field into its linearly independent spatial patterns. In general, the problem with retaining too many EOFs in this sort of calculation is that one’s ability to reconstruct high order spatial patterns is limited with a sparse data set, and in general it does not makes sense to retain more than the first few EOFs. O’Donnell et al. show, however, that we could safely have retained at least 5 (and perhaps more) EOFs, and that this is likely to give a more complete picture.

Third, O’Donnell et al. argue that we used too low a truncation parameter when doing the ‘truncated least squares’ regressions. In general, using too low a truncation parameter will overly smooth the results, and tend to smooth both temporal and spatial information. The problem with using too large a truncation parameter is that it creates problems when data are sparse, resulting in numerical noise (overfitting). O’Donnell et al. try to get around this problem by using cross validation — that is, trying a bunch of different truncation parameters, and using the ones that give the maximum r2, RE and CE statistics.

There are a number of other criticisms that O’Donnell et al. make, such as whether it is okay to infill the weather station data at the same time as doing the calibration against the satellite data (as we did) or whether these have to be done separately (as O’Donnell et al. did). These are more technical points that may or may not be generally applicable, but in any case do not make a significant difference to the results at hand (as O’Donnell et al. point out).

Let’s assume, for the moment, that all of these ideas are on the mark, and that the main reconstruction presented by O’Donnell et al. is, in fact, a more accurate picture of Antarctic temperature change in the last 50 years than presented in previous work. What are the implications for Antarctic climate? How would they differ what was concluded in Steig et al. (2009)? The answer is: very little.

The spatial patterns of annual trends, and how they evolve through time, is similar in both papers. In particular, O’Donnell et al. find, as we did, that the entire continent was warming, on average, prior to early 1980s (Figure below from their main “RLS” reconstruction). As we said in our paper, this would tend to support the idea that cooling in East Antarctica is a recent phenomenon at least in part attributable to recent trends in the Southern Annular Mode (SAM), which is itself forced (at least in part) by stratospheric ozone depletion.

O’Donnell et al. also reproduce our finding that the seasons in which the most rapid and significant warming is occurring are winter and spring — in large areas of both East Antarctica and West Antarctica. In spring, warming is significant throughout all of West Antarctica through the entire 50 years of the record, and in winter, it also occurs throughout all of West Antarctica in the last 25 years. In both seasons in this latter period, the locus of greatest warming has been West Antarctica, and particularly the Ross Sea region and Marie Byrd land, not just the Antarctic Peninsula as virtually all studies prior to ours had assumed. This is an important result that we highlighted in our paper because it has implications for our understanding of the dynamics involving Antarctic warming. Specifically, we made a model-data comparison in the paper, in which we said

… both in the reconstruction and in the model results, the rate of warming is greater in continental West Antarctica, particularly in spring and winter, than either on the Peninsula or in East Antarctica…. This is related to SST changes and the location of sea ice anomalies, particularly during the latter period (1979–2003), when they are strongly zonally asymmetric, with significant losses in the WestAntarctic sector but small gains around the rest of the continent.

In other words, during the period where we have good sea ice data, areas with little sea ice are always areas of surface warming in the Antarctic. It was already well established before our work that sea ice anomalies play a major role in the observed waring on the Antarctic Peninsula’s west coast. Our work showed that this is also true in West Antarctica, and is fully confirmed by O’Donnell et al.’s analysis. The only point of disagreement is in winter, in the earlier part of the record only (prior to the satellite era).

Another point of complete agreement between our results and O’Donnell et al. is that the most widespread cooling occurs in fall — not summer as discussed in earlier work (e.g. Thompson and Solomon, 2000). This may be something of a problem for the hypothesis that ozone depletion is a major driver of the observed East Antarctic cooling, because the forcing is occurring in spring (when the ozone hole develops). If there is a link between the spring forcing and fall temperature, it is not a simple one, but likely would include a role for sea ice, which offers an obvious source of persistence from season to season (a paper in review by Arnour and others argues exactly this point).

Finally, O’Donnell et al. agree with us on the most basic result of all: there is statistically significant warming in West Antarctica. In this context, it is worth being very clear on what is meant by “West Antarctica”. Reading what has been said about O’Donnell et al. in various places in the blogosphere, one would get the impression that their paper returns the warming of Antarctica to its ‘rightful’ place, the Antarctic Peninsula alone. If that were true, it would certainly be a significant refutation of our work. But in the actual abstract of O’Donnell et al., it is stated that “we find that statistically significant warming extends at least as far as Marie Byrd Land.” Marie Byrd Land is that part of West Antarctica that extends eastward from the Ross Ice Shelf up past Byrd Station and over the central West Antarctic Ice Divide (see the map above). In O’Donnell’s results, there is significant warming all the way from the Peninsula westward past WAIS Divide site, at 112°W, well within Marie Byrd Land and nowhere near the Antarctic Peninsula. Prior to our work, no one had claimed that any area outside the Peninsula was warming significantly. Borehole thermometry at WAIS Divide (Orsi and Severinghaus, 2010) and at the Rutford Ice Stream (closer to the Peninsula; Barrett et al,. 2009) has since provided completely independent validation of these results. O’Donnell et al. is thus merely the latest of several studies to confirm our original finding*: West Antarctica is warming significantly.

To be sure, there is real disagreement between our results and those of O’Donnell et al. For the full fifty year reconstruction of temperature trends, the main reconstruction they discuss in the paper shows cooling in the winter and fall over the Ross Ice Shelf, which contrasts with our finding of significant warming there. As a consequence, their overall warming trends are smaller, by about half. These are the only important differences between our results and those of O’Donnell. Nevertheless, they are significant differences, and certainly may be important for our understanding of Antarctic climate change. In particular both results would tend to suggest a greater role for natural variability than our findings implied. If O’Donnell et al.’s results are correct, this would suggest that the damped response of Antarctica to global radiative forcing (i.e. CO2 increases) that is commonly seen in models (as discussed previously by Spencer Weart, for example) is perhaps more on the mark than our paper would suggest (though note that even the much larger trends we estimated are still significantly damped compared with the Arctic.)

Let’s return now to the question of whether O’Donnell et al.’s results actually do represent an improvement over ours. The figure below indicates a rather glaring problem: O’Donnell et al. disagree markedly with the raw weather station data from Byrd, which is the only record of any length anywhere in West Antarctica. Shown in the figure, reproduced again below, are the main reconstructions of Steig et al. (2009) (green) and O’Donnell et al. (2010) (blue), compared with the the actual raw data (black) from the Byrd weather station. The simple linear trend on the raw data is nearly four times larger in reality than shown by O’Donnell et al., whereas it is not statistically distinguishable from Steig et al. There are a lot of missing data from Byrd (and annual means in the figure include some missing months), so also shown in the figure (dashed) is an independent infilling of missing data from Byrd station, done by Andy Monaghan (using no satellite data whatsoever, as described in Monaghan et al., 2008, plus new data available through 2009). The updated Monaghan estimate — currently under review — indicates an even higher trend, >0.4°C/decade, when the data are updated through 2009.

The evident failure of O’Donnell et al. to correctly capture what is going on at Byrd (and presumably elsewhere in West Antarctica) is quite surprising, given that one of key differences in their methodology is to use the weather station data — not the satellite data as we did — as the verification target. That is, O’Donnell et al. use weather stations, withheld one at a time from the reconstruction for verification purposes to optimize their calibration. How then, can they be so far off for the location of the most important weather station? (I say ‘most important’ here because the main point of contention is, after all, West Antarctica). There are three likely sources of the problem, each pertaining to O’Donnell et al. implementation of their suggested modifications to the method we used.

First, as I noted above, O’Donnell et al. use a linear combination of weather station data for their reconstruction, both in the reconstruction period (pre-1982) and in the calibration period (the satellite era, post 1981). This is a very reasonable thing to do, resulting in a more homogeneous data set than ours. However, it also means throwing out information that might be important: namely, that there are strong trends in the temperatures in West Antarctica that may not be captured by any weather station data. This is not a very large problem in East Antarctica, where the scale of spatial covariance is large, and the number of weather stations is also large; it is a potentially huge problem in West Antarctica, where the number of stations is small (again, only Byrd goes back beyond the satellite era) and the spatial scale of covariance is also smaller, due to the greater topographic relief. On top of that, O’Donnell et al. do not appear to have used all of the information available from the weather stations. Byrd is actually composed of two different records, the occupied Byrd Station, which stops in 1980, and the Byrd AWS station which has episodically recorded temperatures at Byrd since then. O’Donnell et al. treat these as two independent data sets, and because their calculations (like ours) remove the mean of each record, O’Donnell et al. have removed information that might be rather important. namely, that the average temperatures in the AWS record (post 1980) are warmer — by about 1°C — than the pre-1980 manned weather station record. Note that caution is in order in simply splicing these together, because sensor calibration issues could means that the 1°C difference is an overestimate (or an underestimate).** Since Steig et al. retained the satellite data, we didn’t need to worry about this. O’Donnell et al didn’t have that luxury, and should at the very least have considered the impact of treating Byrd Station and Byrd AWS as entirely independent records.

Second, in their main reconstruction, O’Donnell et al. choose to use a routine from Tapio Schneider’s ‘RegEM’ code known as ‘iridge’ (individual ridge regression). This implementation of RegEM has the advantage of having a built-in cross validation function, which is supposed to provide a datapoint-by-datapoint optimization of the truncation parameters used in the least-squares calibrations. Yet at least two independent groups who have tested the performance of RegEM with iridge have found that it is prone to the underestimation of trends, given sparse and noisy data (e.g. Mann et al, 2007a, Mann et al., 2007b, Smerdon and Kaplan, 2007) and this is precisely why more recent work has favored the use of TTLS, rather than iridge, as the regularization method in RegEM in such situations. It is not surprising that O’Donnell et al (2010), by using iridge, do indeed appear to have dramatically underestimated long-term trends—the Byrd comparison leaves no other possible conclusion.

O’Donnell et al. do not rely entirely on ridge regression. They also present results from a more explicit cross-validation test, using various truncation parameters for a ‘truncated total least squares’ (or ‘truncated singular value decompositon’) regressions, as we did in our work. However, these tests, as implemented, are also problematic. O’Donnell et al. actually use cross validation in two steps: first, by filling in missing data in the weather station records and choosing the truncation value (kgnd) that yields the best overall verification statistics. Second, by reconstructing the entire spatial field with another truncation value, ksat. In both cases, the optimization is done on the basis of the entire data set; that is, the ‘best’ parameter depends on what works best on average both in data poor regions (e.g. West Antarctica) and data rich regions (e.g. East Antarctica and the Peninsula). The obvious risk here is that too high a truncation value will be used for West Antarctica. There is rather good evidence to be found in the Supplementary Material in O’Donnell that this is exactly what has happened. The choice of kgnd that yields the best agreement with the iridge calculations (which, remember, is already known to create problems) happens to be kgnd = 7, and it just so happens that this yields the minimum trends. In fact, O’Donnell et al. show in a table in their Supplementary Material that the mean trend for West Antarctica for smaller values of kgnd is more than twice (~0.2 °C/decade) what it is for their ‘optimum’ estimate of kgnd = 7 (~0.07°C/decade). Indeed, using any value lower than the one they choose to rely on largely erases any difference between their results and Steig et al., 2009. This simple fact — illustrated in the figure above (trends in °C/decade for 1957-2006) — has been notably absent in the commentaries that O’Donnell and coauthors have made about their paper.

Third, the way that O’Donnell et al. actually do the cross-validation to optimize ksat is itself pretty dodgy. Rather than using split calibrations (that is, comparing early period with late period statistics), they one-by-one withhold each weather station time series over the entire length of the record. To see the problem with this, consider what happens if you withhold the South Pole station record, which is complete for the entire time period, and then repeat the regressions to find the best truncation value for South Pole. For the period 1982-2006, when there are satellite data available for (and highly correlated with the station at) South Pole, the optimal number will be much higher (data richness) than during the pre-satellite era (data poor). The number that gets used will be an underfitting for the pre-satellite era and an overfitting for the satellite era. Note that ksat is actually the number of EOFs that get retained; since one needs many more of these to reconstruct the Peninsula properly, it is inevitable that they’ll wind up with more retained EOFs than we did; that doesn’t mean this is the right number for West or East Antarctica. O’Donnell et al. do report split calibration statistics as well, but this is not how they choose their optimal values.

Does all of this mean that I think O’Donnell’s results are all wrong? Certainly not. I think that they are right to have retained more EOF patterns than we did, though the main impact of this is only in capturing the strong Peninsula warming.*** It is also quite likely that O’Donnell et al.’s results are more accurate than ours for the satellite era, during which most of the problems I have discussed above are less likely to arise. Although their results show much smaller trends, they agree well with the spatial patterns in weather forecast reanalysis data products (NCEP2, ERA-40) during the satellite era. This is a nice, largely independent validation of those products, and suggests that it is okay to use those products — which include detailed information on atmospheric circulation changes, for example — for investigating the causes of the temperature trends. This is something that quite a few of us have been working on, but there has always been the nagging problem that we don’t really know how much we can trust NCEP and ERA products at high southern latitudes. O’Donnell et al. should certainly be cited in support of such work.

In summary, even if their results are taken at face value, O’Donnell et al. 2010 doesn’t change any of the conclusions reached in Steig et al. In West Antarctica where there is disagreement, Steig et al, 2009 is in better agreement with independent data, and O’Donnell et al.’s results appear to be adversely affected by using procedures known to underestimate trends. Thus while their results may represent an improved estimate for the trends in data rich regions — East Antarctica and the Peninsula — it is virtually certain that they are an underestimate for West Antarctica. This probably means going back to the drawing board to write up another paper, taking into account those suggestions of O’Donnell et al. that are valid, but hopefully avoiding their mistakes.

*Contrary to what Ryan O’Donnell has claimed, Doran et al. (2006) reported warming in Ellsworth Land (between WAIS Divide and the Peninsula) only in winter, with cooling in the annual mean. It is worth noting that Doran’s work has previously been misrepresented, though in the opposite way!

**There is, however, completely independent data from the WAIS Divide borehole, showing that this site has warmed by the same amount indicated by the Byrd weather station data — about 1°C since 1958. This is unpublished data, but the results were presented in an AGU talk and in the published abstract.

*** Peninsula warming was not the question we were addressing in our paper, as we made very clear in the text. We chose fewer EOFs based on our previous work (Schneider et al., 2004) showing that this sufficiently captures both East and West Antarctica.) Although retaining fewer EOFs reduces the spatial details, it is a conservative choice for estimating large-scale trends in both West and East Antarctica. See also our discussion on overfitting.

134 Responses to “West Antarctica: still warming”

  1. 101
    BPW says:

    Eric @90,

    I am not assuming anything Eric. Which is why I asked rather than accused and why I sprinkled “alleged(ly)” through my post. Though without an answer, I might get the impression from you that you are being deceitful. Not sure why my reasonable, and I believe polite, questions would annoy you. I would think if the allegations are false, you would be anxious to show that is the case. I look forward to hearing your side of this story. Like I said, you owe me nothing, so I appreciate your response.

    [Response: You wonder why ‘polite’ questions about whether I am a duplicitous idiot would annoy me?? Hey, I heard you stopped beating your wife. GET IT?–eric]

    Nick @95,

    I am speaking for myself in this case, but it is clear if one reads enough that there are plenty of people who are of the impression that this is a perfect example of the alleged problems with peer-review. Perhaps I could have written “some lay people” but since I suspect you and others realize I don’t speak for the masses I’m not sure why it would matter. For the record, I don’t yet have an impression because I have read only one side of the story. Can I assume you have formed yours based on the same information, or are you privy to something I am not?

    Marco @92,

    I think this is a good suggestion, but I would suggest to Eric that he actually should read what O’Donnell wrote because there are some serious allegations he might want to, and probably should, address directly. Otherwise there will be the opportunity for people to claim he dodged portions of the issue.

  2. 102
    Ray Ladbury says:

    Methinks there are a whole lot of folks out there who don’t understand peer review. It is not a friendly little informal cheerleading session. You want to be reviewed by your toughest rivals, because if your mistakes aren’t caught in peer review, the entire community will know about them and you will be asked whether you like your crow medium or well done.

    If you are such a wimp that you can’t take criticism, maybe you ought to consider a career in a field other than science.

  3. 103
    BPW says:

    dhogaza @98,

    Fail? Do tell bird man.

    My first comment you chose simply states that “people like myself” i.e. lay people who have interest in the story may base their opinions, right or wrong, having only O’Donnells side to go on. See, this is actually what is happening. Many people HAVE passed judgment. Nowhere does it even intimate that “I” actually feel that way which is why I asked for Eric’s side of the story to begin with.

    You can play BS games with semantics–as you often like to do–but that does not change the fact that without both sides of the story, people will base their opinion on the side they have read. I actually like to read both sides before I pass judgment. Which, again, is why I asked. Too bad you don’t share that way of thinking.

    No dho gaza, the fail is all yours. Tough to skew my words when they are right in front of everyone to read. But it was a nice try.

    [Response: Let me be clear, once more. There are two sides to this story: O’Donnell’s, and the facts. If you want to read the facts, go here.

  4. 104
    BPW says:

    Eric @91,

    Despite dhogaza’s attempts to twist my words, I am assuming nothing about you. Which is why I asked rather than accused.

    I look forward to reading your side of the story. As I mentioned, you owe me nothing. Thanks for the response.

    [Response: Do you know that old joke about the guy who asks “Do you still beat your wife?” –eric]

  5. 105
    Rattus Norvegicus says:


    So from reading that, it appears that Reviewer 1 raised the same issues during the review process that were raised by Steig here, correct? If my interpretation is correct why the vitriol from O’Donnell, Id, et. al.? I don’t see anything two faced or duplicitous at all, assuming Eric was Reviewer 1 (a pretty good guess). The objections raised here were raised in the the review process.

  6. 106
    One Anonymous Bloke says:

    BPW #101: Dr. Steig’s ‘side of the story’ is right here, in the article above and in the responses he’s made. He has signalled his intention to formally respond to the O’Donnel paper: “This probably means going back to the drawing board to write up another paper, taking into account those suggestions of O’Donnell et al. that are valid…”
    How this demonstrates bias is beyond me. He just said they made a valid contribution!
    The accusations being made say something about the accusers, period.

  7. 107
    dhogaza says:


    If you are such a wimp that you can’t take criticism, maybe you ought to consider a career in a field other than science.

    I don’t think that’s the goal here, I think the goal is to do their best to destroy Eric’s reputation and credibility, as they’ve done with Mann and others.

    Between what went on before the paper was put together, and what’s going on now, I simply do not believe that the advancement of knowledge is the real goal here.

  8. 108
    grypo says:

    Rattus Norvegicus

    I also added ODonnell’s response to the third review to the thread over there. There could be a resolution point there on the science, not sure, I’m not familiar with the science to comment there. But I don’t see how we are able to get the science communicated effectively if one side screams foul play when this type of scientific argument happens. I think it’s important to point it out when the facts get gathered.

    “A lie will go round the world while truth is pulling its boots on.”
    Charles Haddon Spurgeon (1834-92)

  9. 109
    MarkB says:

    BPW writes:

    “1) Were you, as he states, one of the referees on his paper?

    2) If so, do you think that the conflict of interest that would seem to come from that being the case is meaningful and if not, why not?”

    Interestingly, Pielke Sr. threw a fit regarding Watts not being a referee on Menne et al..

    “I was quite surprised to learn that despite the central role of Anthony Watt’s analysis in the paper, he was not asked to be a referee of the paper. This is inappropriate and suggests the Editor did not provide a balanced review process. ”

    Of course, Anthony Watts is on record asserting confidently that the surface record is unreliable and there’s a strong warm bias, so it’s doubtful he would give a fair and logical review.

    [Response: Good catch. There is also Steve McIntyre complaining that Climatic Change breached their contract with him by not sending him the final version of the Wahl and Amman (2006) paper. Then you have McKitrick asking specifically that I not be a reviewer on their paper criticising mine. etc… consistency seems to be too hard a think to ask for. – gavin]

  10. 110
    Robin Edwards says:

    This was a very long posting, but I have not yet been able to find a reference to where the primary Byrd temperature data are to be found.

    Have I missed something?

    I would like to look at the original numbers myself. I’ve always found that this is a satisfying and instructive thing to do.

    Thanks in advance for the URL.

    [Response: Byrd and Byrd AWS. – gavin]

  11. 111
    Rob says:

    Between all the mudslinging going on at climateaudit (and wuwt) after Eric’s scientific post here, I thought it would be refreshing to actually attempt to “discuss the science” as these web sites so often claim to want to do.

    So I took Eric’s first point of criticism (regarding the Byrd data), weeded out the science in O’Donnell’s response on that issue, found that he failed to address Eric’s core concern (+0.08 C/decade at Byrd does not match the Byrd station data), and posted on ‘climateaudit’ to get clarity.

    After an initial quick response from Ryan there were some red herrings thrown, and then the thread went dead.

    Here is my post :

    with this being the second request that went unanswered so far :

    In absence of an answer to my question, let me make the issue a bit clearer.
    Byrd is not just “an individual station”, it is the ONLY station on West Antarctica that has a temperature record before 1980.
    Any reconstruction that does not match the Byrd station data at least somewhat realistically is thus useless and should be discarded as statistical unsound at the earliest convenience.

    Ryan mentions in his post above mentions that “the 50-year Byrd trend is 0.25 +/- 0.2 (corrected for serial correlation)” which confirms assessments done by Monaghan. However, the statistical methods and parameters chosen in O10 result in a trend of +0.08 C/decade for the full Byrd record. Why ? Because
    the statistical methods and choice of parameters used by Ryan calculated an ‘offset’ between the manned (before 1980) and the automatic (after 1980) stations. Why an offset ? Because the considered them as ‘individual’ stations. Why not consider the fact that they represent a single location because they are only 2.5 km apart ? Because Ryan above claims that “microclimates matter”. If anyone has ever been to Byrd, or has seen the pictures on Google Earth, it would be very hard to validate ANY offset between two stations being 2.5 km apart on that deserted ice field without a proper explanation.

    Instead of assuming an offset of 0 (the null hypothesis), Ryan calculates the offset based mainly on “regression coefficients” of two stations located more than 1000 miles away, on the East Antarctic coast (Scott Base and McMurdo). And then again, Ryan avoids mentioning HOW MUCH the offset at Byrd actually turned out to be with his statistical methods.

    Now does this offset at Byrd matter for the final conclusions of the paper ? Ryan states “what happens if you add a 0.5 Deg C / decade trend to Byrd? Why, the West Antarctic trend increases 160% from 0.10 to 0.26 Deg C / decade”. In reality, the only trend that his unsubstantiated offset created is a 0.25 – 0.08 = 0.17 C/decade, which would increase the trend from West Antarctica in the O’Donnell from 0.1 to some 0.15 C/decade. With uncertainty margins, this would bring the O’Donnell ’10 conclusions very much in line with the Steig ’09 reconstruction that it is trying to ‘correct’.

    So yes. Byrd matters. The Byrd offset numbers, please…

    Note that this is my own interpretation of Eric’s point on Byrd, and I take sole and full responsibility for any inaccuracies in my text.

    O’Donnell’s failure to bring such a simple (and crucial) factual number (the Byrd ‘offset’ data that O’Donnell used) in the open, and instead dance around the issue with ad hominems is revealing at multiple levels and scientifically suspect. And that is only the FIRST of Eric’s points of criticism.

    At climataudit discussing science is apparently not appreciated, I’m counting on you guys here to help to re-insert the scientific arguments in the discussion (either here or in the deafening noise of the climateaudit and wuwt comment section).

  12. 112
    Ray Ladbury says:

    dhogaza, all they are succeeding in is making themselves look like petulant children!

  13. 113
    BPW says:

    Eric @91,

    Yep. Know the joke. If I had asked, for instance, “are you still lying to hide the truth” you would have a point. But I didn’t.

    The interesting thing for me is that, with all the commenting elsewhere regarding moderation, and since you did not publish my first questions, that you chose to allow the second ones. As I have said more than once, thanks for your responses, even if they were a bit terse. You could have said nothing.

    MarkB @109,

    I was not insinuating that such a process was inherently “bad”, I was actually asking if Eric thought there could be a conflict. Seems from your comments about others, and Gavin’s response, that the belief is that there, in fact, could be conflict issues. If Watts cannot be trusted–according to MarkB’s comment “…it’s doubtful he (Watts) would give a fair and logical review.”–is it not fair to state the same of others regardless of what “side” they are on? With all the personal attacks and bad vibes surrounding this topic, it does not seem to far a stretch to think those emotions would have an effect in a review environment. Or is there a double standard?

    I also still find it puzzling that such review is normal but, at the same time, expected to remain secret. If so normal, why would Eric, or anyone, care that people knew he was a referee and why deny having read the paper on other forums? Why not just own it and avoid the impression that something untoward took place after the fact?

    Eric, I see you have posted your response and perhaps all my questions will be answered. I will take a look and perhaps comment further on that thread.

  14. 114
    Pinko Punko says:


    This is just how the peer review system has evolved. In very small, contentious fields with highly public debates, anonymity means less because there is a culture and necessity of extremely critical dialogue. Other fields are larger, and anonymity allows hopefully legitimate criticism without the fueling of grudges between author and reviewer.

    If the conflict is scientifically based it usually isn’t an issue. If the conflict is based on competition this could be an issue, but up to the editor to decide what is fair and the reviewer should also interrogate their own conscience. The job of the editor is to publish the best paper that advances the science. To do so may require the input of the most knowledgeable person working in the same field to best review the manuscript. It isn’t perfect, but it works.

  15. 115
    Rattus Norvegicus says:


    Byrd station is occupied by a few dozen people and it does have an airstrip, so obviously it is corrupted by UHI and the fact that it is near an airport. You just can’t trust the data :)

  16. 116
    J.C. Moore says:

    I had an exchange of e-mails with O’Donnell about whether he considered his paper a refutation of Steig or an addition. The article is posted at:

    The significant point to me, and they both seem to agree, is that Antarctica, the coldest place on Earth, is warming.

  17. 117
    RomanM says:

    Thought it would be more appropriate to post this on the science thread, not the [self-snip] thread which was started later.

    I spent the afternoon cobbling up an explanation of several issues for some of the less mathematical readers (not intended as a slur to anyone) with something that you might like to address as well, Eric. It is too lengthy to put as a comment and includes some graphs so I put it on my website:

    Perhaps you might respond to some of the issues that I raise in it.

  18. 118
    Didactylos says:

    BPW: Watts isn’t qualified to review anything, so it really isn’t an issue.

  19. 119
    eric says:

    Comments are still on here. Science only please.

  20. 120
    Susan Anderson says:

    A little help over at DotEarth would be appreciated if anyone has the stomach to push the truth out in front of the hounds.

    I deeply appreciated the cogent scenario posted by Majorajam before the -gate gate was closed and should mention I may be “borrowing” it.

    Have to thank all of you who continue to do science in the face of all these difficulties.

  21. 121
    captdallas2 says:

    118 Didactylos

    If I am not mistaken the Watts situation was that his surface station data was used by Menne et al. prior to Watts publishing his own report. It was his data that he had not completely reviewed so he had a legitimate right to review the paper whether or not he was qualified.

    Kinda surprising a moderator did let you know that.

    [Response: Nobody has a ‘right’ to review any paper. Editors choose reviewers as they see fit. I’m surprised no-one has pointed that out to you. – gavin]

  22. 122
    captdallas2 says:

    [Response: Nobody has a ‘right’ to review any paper. Editors choose reviewers as they see fit. I’m surprised no-one has pointed that out to you. – gavin]

    Even if intellectual property is used without prior approval? That is why Watts asked to be allowed to review the paper. Perhaps he should have exercised other rights?

    [Response:No, he should have exercised some common sense. If you don’t want people to use data, don’t put it online. And as above, editors choose reviewers, reviewers do not choose themselves. – gavin]

  23. 123
    caerbannog says:

    9 Feb 2011 at 4:14 PM

    Of course, Anthony Watts is on record asserting confidently that the surface record is unreliable and there’s a strong warm bias, so it’s doubtful he would give a fair and logical review.

    I recently “rolled my own” simple global-average anomaly calculation app in my spare time. It’s a straightforward C++ app that reads in GHCN V2 data and metadata and computes a simple gridded global average of the station data. Here’s a plot of my results vs. NASA’s “Meteorological Stations” global-temperature results (I used GHCN “raw” monthly-mean data):

    This was not a hard project at all — got the basic app up and running in my spare time over just a few days. If Watts and Co are having trouble getting their analysis done, they should post the WMO ID’s of the stations that they’d like to have processed and I could turn around some results for them very quickly (easily same-day service).

    I can slice and dice the temperature data in all kinds of ways — and no matter what I do with the data, I get a global-warming signature very similar to what the NASA/CRU/etc folks get. I can throw out over 90 percent of the stations at random and still get consistent results.

    The global-warming signal is so strong and the temperature record is so robust that you’d have to work *really* hard *not* to show significant global warming.

    Hmmm…maybe *that’s* the problem that Watts and Co. are having…

  24. 124
    dhogaza says:

    caerbannog: you initialize a couple of static data members of type “float”, which isn’t standard C++ (only integral types can be initalized) and not support by the version of GCC I’m using, at least … (4.01, Mac OS/X 10.5.8).

    Easy to fix, just thought I’d give you a heads up.

  25. 125
    caerbannog says:

    caerbannog: you initialize a couple of static data members of type “float”, which isn’t standard C++ (only integral types can be initalized) and not support by the version of GCC I’m using, at least … (4.01, Mac OS/X 10.5.8).

    Easy to fix, just thought I’d give you a heads up.

    Thanks for the heads-up — The app compiles for me without complaints with Linux/g++-4.4 and OS/X-10.6.0/g++-4.2. (I’m basically a “seat of the pants” programmer who just gets by, so every now and then I get tripped up over standards issues — the very latest gcc/g++ releases help keep me out of trouble in that respect).

    The version of the app that I posted over at doesn’t read in all the GHCN data correctly (due to my failure to RTFM properly) — I described the issue in a followup to my skepticalscience post.

    Since then, I’ve added a simple gridded averaging algorithm and some metadata-sniffing code to sort out urban vs rural stations, etc. Not very hard at all. Even with my modest programming skills, I seem to have been able to accomplish more in just a few days’ effort than Anthony Watts and his followers have managed to accomplish in several years. Does that qualify me for any brownie-points?

  26. 126
    caerbannog says:

    One more embarrassing note — I think that there’s an “off-by-one” error in the moving-average routine. Eliminating the first “iyy_trailing++” line should fix it.

  27. 127
    dhogaza says:

    Thanks for the heads-up — The app compiles for me without complaints with Linux/g++-4.4 and OS/X-10.6.0/g++-4.2.

    OT here so my last on this:

    For some reason I can’t get this macbook pro to upgrade to snow leopard, it’s sick in a way that’s not killed it day-to-day. So it’s still got gcc 4.01.

    Since then, I’ve added a simple gridded averaging algorithm and some metadata-sniffing code to sort out urban vs rural stations, etc. Not very hard at all. Even with my modest programming skills, I seem to have been able to accomplish more in just a few days’ effort than Anthony Watts and his followers have managed to accomplish in several years

    Technically, yes. Politically … they’re not such failures :(

    Anyway, ¡¡¡ good job !!!, thanks, and I’ll look over at skeptical climate to get a later version.

  28. 128


    “Does that qualify me for any brownie-points?”

    Not on WUWT, I’m guessing. . .

  29. 129
    Rob says:

    I made a post here on RC (#111) and on CA
    with a summary in my own words of the first point of critisism that Eric made : The underestimated rate of warming at Byrd in the O’Donnell reconstruction.

    The thread was silent for a couple of days, but an entity called ‘troyca’ now posted a response with an argument that very much looks like an attempt to weasel out of this store issue.

    O’Donnell gave no scientific validation for modeling Byrd as separate stations with an offset. But he needed this offset at Byrd, to create an artificially low trend at Byrd location, so he could claim ‘significant’ difference in trends with S09. Since he cannot validate that offset physically (Byrd is an icefield without ‘microclimates’), he got really upset when Eric brought up the issue. Result was the ad hominem’s and smoke and mirrors that we have all been a witness of.

    I have no time now, but I will respond on that thread at CA later.

    Please feel free to comment on CA or here about this subject. Just be aware that entity ‘troyca’ can be one of the co-authors of the O’Donnell paper.

  30. 130
    Rob says:

    Let me also note that entity ‘troyca’ implicitly questions the trend that Monaghan reconstructed at Byrd. For that trend, I relied on Eric’s number (+0.32 C/decade). It would be great if we can get that rate confirmed scientifically.

    And, ‘troyca’ attempts to equalize O10 and S09 and blame RegEM for failure to reconstruct the offset at Byrd.

    It seems that the time is right now to discuss science, and clear up the issue of Byrd in both reconstructions, and cut through the weasel tactics to get to the core which should reveal that O10 and S09 are really not ‘significantly’ different in results for an area as wide as West Antarctica.

  31. 131
    Jay B. says:

    As I am not a scientist I cannot give to this article some exact resolution. But I have found this article very interesting in many points. Especially I am wondering how difficult can be to measure the surface temperature. I also appreciate that you can take it so seriously and that various teams work out the results with so much declinations. Perhaps it is because of the conditions and technology you are using. What I can see a problem here is that we should expect more prevalence of higher temperature anywhere around the Globe. As we are in Holocene epoch (that is in the middle of two Ice age epoch) together with impact of human´s activities we should have to consider that thing will change. And I assume that this is what´s happening in Antarctica too.

  32. 132
    Ohio says:

    All of this discussion is execllent, but it really should be said in a peer-reviewed paper, correct?

  33. 133
    Hank Roberts says:

    > Jay B.
    > how difficult can be to measure the surface temperature.

    Eric answered that earlier in the thread, here: and you can find more here:

    > Ohio
    See the first post; the O’Donnell et al. paper (and others ) originated in a series of blog posts; the journal editors’ job is to help the authors sort out the ‘discussion’ and publish the science.

  34. 134
    Septic Matthew says:

    I reread this post after I read Eric’s response to O’Donnell, and I think that this post is very informative, though quite technical for people who have not used ridge regression and other techniques for ill-posed problems in high-dimensional statistics. I look forward to a peer-reviewed version in a journal.