Global warming underestimated

Given that Rachel has published three posts today, I feel that it would be okay to reblog one of them. The video in this post is very interesting and suggests that the supposed “hiatus” (which refers to surface warming only) has been rather over-emphasised. Including the polar regions (which are typically not included in surface temperature datasets and which typically warm faster than other regions) indicates that there may have been no “hiatus” in surface warming. A video well worth watching even if what it suggests is not particularly comforting.

This entry was posted in Climate change, Global warming and tagged , , , , , , . Bookmark the permalink.

23 Responses to Global warming underestimated

  1. Rachel says:

    Thank you! I wish I’d put a bit more effort into writing this post now.

  2. Seemed perfectly fine to me, plus how much time can you spend on one post when you’re publishing three a day 🙂

  3. Rachel says:

    Oops, forgot to add the link to The Disappearing Hiatus.

  4. Yes, I noticed that – although it is probably still worth explaining why there has been a slowdown in the surface temperature rise in the 84% of the surface that is covered.

  5. BBD says:

    For some time now I have cautioned enthusiastic wavers-about of the “hiatus” that it will not last. Laws of physics and all that. I have warned them that when, not if, a strong warming trend resumes, the mast they have nailed their colours to so very firmly will drag them down as the entire ship capsizes.

    They laugh, of course, but it is still true.

    As the authors of this study say, this will not be the final word. But if I were a fake sceptic, I would be frightened by the sounds of splitting timber and spraying water from belowdecks. Won’t be long now.

  6. Tom Curtis says:

    Rachel, there is still a “pause”. More specifically, the HadCRUT4/UAH hybrid temperature index merely brings the recent trend of HadCRUT4 into line with that for GISTEMP, and NOAA. The trend is still only two thirds of the thirty year trend – and that still needs some explanation (of which there is an over abundance of candidates).

    As to explanations:
    1) The arctic has been warming faster than the rest of the planet, such that including it will cause a small increase in total trend. Small because the Arctic occupies only a small total area (despite the massive distortion in common maps).

    2) In particular, since 2007 warm open waters in the Arctic where previously there has been ice may have distorted Arctic weather patterns, in particular making Arctic winters warmer and European, Alaskan and Siberian winters much colder. This is in addition to effect (1) and helps explain the additional reduction in trend since 2007. It should be noted that the explanation of the pattern is still controversial, although the existence of it is not.

    3) By excluding much of Africa (in particular), HadCRUT4 is biased in such a way as to make it more sensitive to ENSO fluctuations. This is particularly noteworthy in 2010, when 5 of 19 nations setting all time maximum temperature records came from Africa, with a further 6 coming from the Middle East or Central Asia (which also have restricted coverage). Making the record global reduces the relative impact of ENSO, and in particular the rapid cooling trend resulting from the transition from El Nino dominated to La Nina dominated around 2007-2008.

  7. Very off topic, feel free to remove again.

    Rachel, I tried to comment at your blog today, but the comments did not appear on this page:

    There were no links, no trigger words. I have no idea why.

  8. dana1981 says:

    I’m just going to plug my post on this paper if you don’t mind. The authors provided input as I put it together, and a few quotes.

  9. Rachel says:

    Thanks for letting me know Victor. I have been so busy checking the spam folder here that I forgot to check my own. Your comments have appeared.

  10. Rachel says:

    Great article, Dana. It’s really nice to read one of the authors say:

    No difficult scientific problem is ever solved in a single paper. I don’t expect our paper to be the last word on this, but I hope we have advanced the discussion.

    I wish self-described skeptics would take a similar approach.

    Thanks for the explanation, Tom. So there’s still a ‘pause’ but it’s smaller than it was before 😉 That sounds reasonable to me.

  11. BBD says:

    To repeat some of Dana’s Guardian article, this is further evidence that recent sensitivity estimates somewhat dependent on the last decade are probably uninformative. Hopefully, Cowtan & Way will help improve future attempts to constrain TCR from “observations”.

  12. danolner55347852 says:

    I hope you don’t mind me jumping in here with a discussion point I was wiffling about over at the planet3 coverage of this story (linked to in Rachel’s comment above), I was rather hoping for someone to tell me if I’m talking nonsense! The video explaining this research is excellent and clear and realclimate’s explanation of the difficulties of using satellite data was useful to me. But I’m struggling with the coverage given to it. It’s interesting that Dana says he developed the article in comms with the authors – it’s specifically headlines like “global warming since 1997 more than twice as fast as previously estimated” that concern me.

    It’s a semantic issue, but I think an important one. I could be wrong, if anyone fancies putting me in my place… but we’ve just spent a long time criticising people for making claims about the long-term climate trend based on data from 1998 or thereabouts, usually saying “it’s too short a period to conclude anything statistically significant about that trend and these differences are still within the variability you’d expect.” That’s still true, isn’t it, even if one picks 1997 as your trend starting date?

    Tom says: “The trend is still only two thirds of the thirty year trend – and that still needs some explanation.” Well – the explanation is, as always – too short a period to be saying anything about the longer-term trend. Or more simply perhaps: no, it doesn’t need an explanation, because the range under discussion is too short to give one. To compare, looking at daily temperature data (which I got from York weather station archive data indicentally!) you get trends opposite to the seasonal trend for 3 out of 10 OLSs, using a 29 day range. There may well be some “explanation” for that up-and-down variability in weather patterns, but if the thing you’re wanting to explain is seasons, there’s only one answer: too short a trend period to be reliable. It certainly doesn’t mean the seasons are changing twice as fast or that they’ve stopped, if one particular 29 day trend suggests it is.

    Hence my worry about the way this paper seems to have been presented in the blogosphere. It was absolutely correct to previously say the hiatus talk was premature – that headlines saying “data since 1998 says global warming has slowed or stopped.” So why now lead with headlines saying “data since 1998 says global warming more than twice as fast as we thought”. Aren’t those both equally statistically wrong?

  13. BBD says:

    I think it’s fair to say that the tactics used by “sceptics” do sometimes elicit less-than-ideal science communication. That said, the study presents a synthesis of temperature records that suggests recent warming has been underestimated. So that’s what gets talked about. I do agree with you about the double-edged matter of short term trends but this result has to be discussed in that context.

  14. Dan, I think that you make a perfectly valid point. I suspect there are two issues. One is simply that if the temperature datasets have been ignoring the polar regions (or under-estimating the warming in these regions) then it’s clearly a good thing to try and do that properly (or more correctly). However, as you suggest, the uncertainties still mean that one should be careful about how we present this. I think it’s perfectly fine to present a mean trend and an uncertainty, but one should be clear about what this means. If the mean trend is more than twice as big as previously thought, I have no problem with people presenting this, but they should also present the uncertainties in this trend.

    The other issue, as you too indicate, is that if some have claimed that the period since 1998 has been too short to make any strong claims about the surface warming, then we should be careful about making a big deal of a study just because it suggests faster warming.

  15. danolner55347852 says:

    Thanks both for replying, much appreciated. BBD, I do think perhaps the response has been an over-reaction to what’s been a very difficult few months – watching the success of the “hiatus” meme becoming waay too mainstream (including the BBC, as we were discussing). Which makes it all the more important to be self-aware and not end up caught in the wake of the “skeptic” comms agenda. Any short-term media gain will be lost in the longer term.

    Though we did already know the rate of arctic warming was much higher than average, one can imagine a scenario (perhaps at a different historical point in the overall climate trend) where this research had produced a *less* steep trend. Skeptics would have klaxoned: “global warming since 1997 more than twice as slow as previously estimated”, which would have been just as (in)valid a headline… we do not want to play that game! (Sorry I’m picking on Dana’s headline, there were others…)

  16. Dan, you have to distinguish between “noise” errors due to year to year variability in the global mean temperature and bias errors.

    Let’s take a simplified example. Let’s assume that 50 years ago we only had temperature measurements with Stevenson screens and that nowadays we only have temperature measurements by automatic weather stations. And that the transition was linear, 2% a year. Furthermore, let’s assume that the automatic weather stations are mechanically ventilated and therefore measures 0.5°C less than the Stevenson screens, which are naturally ventilated and may heat up on calm sunny days. (This should still be studied, but I thought that it would make a simple understandable example).

    Now we would have a bias in the temperature trend of 0.1 °C per decade. If we would know the 0.5°C bias with sufficient accuracy, because we have parallel measurements (with Stevenson screens and AWS at the same location) in many climatic regions. Then it would be fine to correct for this trend bias even if the uncertainty in the trend for a specific decade is larger than 0.1°C per decade. Because the parallel measurements are made at the same location and time, you can determine such a bias nearly independently of the variability of the weather (you should only sample a representative number of calm sunny days).

    If the new Arctic temperature paper holds, which we will know in a few years, the Arctic would be a similar bias for the last decade. Then it is fine to correct this bias. Because the deviation in the Arctic itself is quite larger and because you are comparing two methods for the same years and thus the same weather situations, you can compute such biases much more accurately as suggested by the noise in your trend estimates due to the year to year variability.

    If you are just arguing that the correction is very small and thus climatologically not very interesting, your are right. That is probably the main reason why two relative outsiders performed this study. For climatologists it would have been hard to find funding for such study, as the trend in the global mean temperature between 1880 and now is still about 0.8 degree per century.

    Also the year to year variability has a physical cause and can in principle be studied and *have* an explanation. Thus maybe Tom should have better written “have” and not “need”, if taken out of context. However, I think he responded to people claiming that multiple explanations (Arctic, El Nino, oceans in general, volcanoes, etc.) for the atmospheric temperature slowdown would conflict with each other. Seen in that light, I would tend to argue that he said the same thing as you did: there is sufficient uncertainty in such short term trends that you can allow for multiple explanation without any conflict.

    P.S. A review of a paper typically takes half a year and the authors also write on their page that they have worked on the study for several years. Thus this is no reaction to a few month of “hiatus” debate, which, by the way, has also started much earlier. Stefan Rahmstorf wrote it started in 2005 (in German).

    WUWT regularly claims warming is one half of previous estimates. That would be fine, if there was proof for the claim.

  17. Wotts, Rachel, another comment is lost in the spam folder. Sorry.

  18. BBD says:

    Dan Olner

    Yes, of course. Proper scientific caution is preferable to rhetoric. And lying down with dogs gets you fleas; something scientific communication does not want at all right now. Unfortunately, the world runs largely on rhetoric, which keeps the “sceptics” in business and inhibits public understanding of the science. As we all know and regret.

  19. Tom Curtis says:

    Dan, I think you raise a very good point. To my mind, the “rate of global warming” is (first) the rate of the current terminal thirty year (1983-2012) trend +/- 2 s.e. From the various data sources, that is 0.178 +/- 0.058 C/decade (GISS); 0.165 +/- 0.054 C/decade (NOAA); 0.174 +/- 0.056 C/decade (HadCRUT4); and 0.194 +/- 0.063 C/decade (HadCRUT4 hybrid). Ideally we would use a composite of all four indices (plus that produced by the JMA), where we plot the compound PDF and derive from that the mean and median. As a substitute I will take a simple mean of the four above which is 0.178 C/decade. As an alternative, I will happily accept the 15 year (1998 to current) ENSO, Volcanic and Solar adjusted trend, which averages 0.161 C/decade from GISS, NOAA and HadCRUT3. IMO, any argument about current global warming trends that is not valid using either of the above trends is not valid, simpliciter.

    Given that the 15 year trend of the HadCRUT4 hybrid is 0.115 +/- 0.166 C/decade, it is less than the thirty year or EVS adjusted trends and hence does not raise our estimate of the global warming trend, or at most raises it by 0.006 C/decade (the increase in the thirty year mean from including the HadCRUT4 hybrid).

    I disagree with you, however, on what requires explanation. Specifically, when investigating seasonal effects we can take a mean of many past observations by day (or month) in the year and thereby obtain a seasonal signal. We then need only explain the variation in that mean seasonal signal when explaining seasons (excluding variability within seasons). However, we do not have multiple runs of the twentieth and twenty first centuries from which to take mean and distinguish the global warming signal. Therefore, for any variation appreciable at a decadal scale, we need to have an explanation in order to show that it does not represent a change in the global warming signal.

    In this we are different from our opponents, who do not allow science to constrain them. If they did, they would impose on themselves the limits of not claiming falsification based divergences from predicted values with a p value >0.01; and even when such divergences existed, would first check to ensure no short term explanation exists for the divergence. That they do neither, but rather claim falsification on p values >0.05 and actively ignore short term factors shows (as if it had not been shown multiple times before) that whatever game they are playing, it is not science.

  20. Rachel says:

    Sorry about that Victor. Your comment is now alive and well.

Comments are closed.