RealClimate logo


Hiatus or Bye-atus?

Guest commentary by Stephan Lewandowsky, James Risbey and Naomi Oreskes

The idea that global warming has “stopped” has long been a contrarian talking point. This framing has found entry into the scientific literature and there are now numerous articles that address a presumed recent “pause” or “hiatus” in global warming. Moreover, the “hiatus” also featured as an accepted fact in the latest IPCC report (AR5). Notwithstanding its widespread use in public and apparent acceptance in the scientific community, there are reasons to be skeptical of the existence of a “hiatus” or “pause” in global warming [Ed: see also this previous post]. We have examined this issue in a series of three recent papers, which have converged on the conclusion that there is not now, and there never has been, a hiatus or pause in global warming.

We are not alone in coming to this conclusion; evidence for this has also been reported by Cahill and colleagues in a recent statistical change point analysis, which failed to identify a slowing in warming at any point in time during the last four decades.

But because this conclusion is potentially controversial, it requires a careful analysis of the conceptual landscape of research on temperature variation over the recent period.
To date, research on the “pause” has addressed at least 4 distinct questions:

  1. Is there a “pause” or “hiatus” in warming?
  2. Has warming slowed compared to the long-term warming trend?
  3. Has warming lagged behind model-derived expectations?
  4. What physical mechanisms underlie the “hiatus”?

Those questions are not only conceptually distinct, they also involve different aspects of the data and entail different statistical hypotheses. Nonetheless, those questions have frequently been conflated in the literature, and by using a single blanket term such as “pause” or “hiatus” for distinctly different phenomena and research questions, unnecessary confusion has resulted.

To reduce this confusion, our recent work has been exclusively concerned with the first question: Is there, or has there recently been, a “pause” or “hiatus” in warming? It is this question—and only this question—that we answer with “no”, based on multiple lines of evidence:

In an article published in BAMS this year (Lewandowsky et al, 2015), we reported on the results of a blind expert test by professional economists. Blind tests are the methodological gold standard in many fields of enquiry, from pharmaceutical research to cognitive science. The economists in our sample were shown the global temperature data (NASA’s GISTEMP) but it was labeled as “world agricultural output” as shown in the figure below.



The experts had to evaluate a statement accompanying the graph which read: “A prominent Australian critic of conventional economics, Mr. X., publicly stated in 2006, that ‘There IS a problem with the growth in world agricultural output—it stopped in 1998.’ A few months ago, Mr. X. reiterated that ‘…. there’s no trend, 2010 is not significantly more productive in any way than 1998.’ ” This statement was an exact translation, into economic terms, of a series of public statements by an Australian contrarian (Bob Carter) claiming that global warming had stopped.

The experts in our sample clearly disagreed with the notion of a pause or hiatus: Experts rejected the idea that the data confirmed the statement and they instead found the data to contradict the accompanying statement. The experts also found the statement to be misleading and ill-informed. And nearly two thirds of our experts endorsed the possibility that the claim of a pause or hiatus might be fraudulent in light of the data.

In a second article, just published in Scientific Reports (Lewandowsky et al, 2015b), we followed up on this experimental test with a formal statistical analysis that buttressed the conclusions of the blind expert test. We began by considering a corpus of 40 peer-reviewed articles that have addressed the pause and inferred from those publications what those authors considered to be the onset year of the pause. Did the pause commence in 1998? 2001? Or some other year? We found that there was considerable variation, shown by the blue histogram in the figure below, which shows the distribution of presumed onset times together with global temperatures during the modern record of global warming.



The histogram shows that the presumed onset of the pause spanned an entire decade (1993-2003). The mean presumed duration of the pause across the 40 articles was 13.5 years. We next took the onset and duration of the pause reported by each article and compared the associated decadal temperature trends against the distribution of all possible trends of equal duration during the last few decades. If there were a pause, we would expect the distribution of trends in the literature to differ considerably from the distribution of all trends of equivalent duration.

Because there is some disagreement about the onset of modern global warming, we used three reference dates for our comparison involving all possible trends: the year 1951, used by the IPCC in AR5, and 1964 and 1976, which are 2 standard deviations below and above and below the mean estimate of the year of onset of modern global warming derived by the Cahill et al change-point analysis.

The results are shown in the 3 histograms below, with onset times 1951, 1964, and 1976 from left to right. The vertical red lines in each panel represent the long-term trend (1951–2012) used by the IPCC in the AR5. The solid line is for the GISS dataset analyzed here, and the dashed line is the long-term trend for the same period (.12K/decade) in the UK Met Office’s HadCRUT4 data set.



This analysis shows that the distribution of warming trends labeled as the pause by the literature is indistinguishable from the overall distribution of trends that have been observed from the middle of the 20th century onward. When 1964 or 1976 are instead used as onset of warming, then the distribution of trends labeled as a pause does sit at the lower end of the overall distribution, but it is still by no means consistently extreme or unusual.

Additionally, virtually all articles on the pause referred to a time period during which the decadal warming trend exceeded zero (the black vertical line). This is incompatible with standard dictionary definitions of a pause or hiatus, which cite a process that has been suspended or stopped. Periods in which warming continued (>0 K/decade), by definition, cannot be a pause or a hiatus.

For the notion of a hiatus in warming to be scientifically well-founded, there must either be a demonstrable and statistically-relevant absence of any trend in global temperatures or, minimally, the observed trend must differ in a statistically identifiable way from the historical record. With conventional frequentist statistics, the absence of a trend is difficult to establish, and results can be affected by vantage point. That is, different results may arise depending on when one chooses to look backward in time, not because anything notable has happened but because natural variability can override a long-term trend if only brief periods of time are considered and if the end points of the trend are varied.

We therefore compared the alleged pause against all possible historical trends once more, but this time across all possible historical vantage points during the last three decades. From each vantage point, we look backward in time a varying number of years and determine the magnitude and significance of the trend. The results are shown below using GISTEMP:



The top panel (A) shows the warming trends that were observable at any vantage point between 1984 and 2014 (horizontal axis). For each vantage point, between 3 and 25 years were included in the trend calculation (vertical axis). So for example, looking backward from 2014, no matter how many years were included in the trend, all trends were positive. When 15 years are included, then the trend becomes statistically significant (hence the dot in the cell at latitude 15 for the last column). By contrast, in 2000, the most recent 3 or 4 years exhibited cooling, but by the time 15 years were included the trend was again significant. The bottom panel of the figure (B) presents the same data using a ternary classification of p-values for the linear trend into non-informative (p>.10; beige), partially informative but not conventionally significant (.10 > p >.05; gray), and significant (p < .05; terracotta). This panel also includes three diagonal lines that identify the earliest calendar year included in the analysis. Any observation to the below and to the right of the line labeled “1975” only includes observations later than that, and so on for the other two lines. The observations above and to the left of the 1965 line go back to 1960 (top-left corner; looking back 25 years from 1984 inclusive).

The figure shows that at every year during the past 30 years of modern global warming, the immediately preceding warming trend was always significant when 17 years (or more) were included in the calculation. In a number of cases—including in 2014—fewer years were required to reach significance, but never more than 17. This result should not be surprising: Significance requires statistical power to be detected, and the more observations are considered the greater the power of the analysis. The fact that a trend fails to reach significance with, say, only 5 or 10 years of data is therefore non-informative: no matter how robust the warming trend, once natural variability is superimposed on the trend, it will escape statistical detection with a small sample. To conclude from that that global warming has “stopped” is unwarranted. Nevertheless, a number of papers in the peer-reviewed literature have done just that. That is, they concluded that there was a pause in warming using a time period that was too short to achieve conventional statistical significance.

To illustrate, we used the definitions of the pause found in our corpus of articles (mean duration 13.5 years), and asked how often the null hypothesis of no warming would fail to be rejected during the last 30 years. It turns out that during those 3 decades, the 14-year trend escaped significance 10 times and the 13-year trend 13 times, suggesting either that global surface warming “paused” between 30% and 43% of a time period during which the Earth warmed 0.6K overall, or that global surface warming never paused and what we have been observing are routine fluctuations superimposed on a warming trend.

Taken together, the statistical and behavioral evidence demonstrate that the notion of a pause or hiatus—as commonly understood—is incorrect. The evidence from the blind expert test suggests that it is also misleading.

The question of a recent pause in warming is distinct from the matter of fluctuations in warming rates over any particular time period of interest. It is uncontroversial that the climate system is highly variable and, as our analysis shows, there have been many fluctuations, both positive and negative, as compared to the longer term warming trend. It is a valuable research endeavor to examine why these fluctuations occur, and to what extent models and observations may on occasion diverge. Such research has the potential to improve our overall understanding of the climate system.

This brings us to our final issue: If there is no pause and there was no pause, why did the recent period attract so much research attention? We can suggest a number of reasons. One is a matter of semantics. Many articles on the pause addressed not the absence of warming but were concerned with a presumed discrepancy between models and observations. We do not believe that those articles should have been framed in the language of a “pause,” but that does not mean their method or findings are compromised.

A second reason is that owing to the incessant challenge of climate science by highly-vocal contrarians and well-organized ‘Merchants of Doubt’, scientists may have become not only reticent in reporting the full spectrum of risk they are concerned about for the future (see for example Brysse et al, 2013), but have also subtly changed the way in which they approach their science. We explored the possible underlying mechanisms for this in an article earlier this year (Lewandowsky et al, 2015c). In a nutshell, we argue that scientists have unwittingly been influenced by a linguistic frame that demonstrably originated outside the scientific community, and that by accepting the word “pause,” scientists have subtly framed their research in ways that our statistical and behavioral analysis has revealed to be inappropriate.

When scientists use the terms “pause” or “hiatus” they may indeed know—and their colleagues may understand—that they do not mean to imply that global warming has stopped. The problem is that words such as “pause” or “hiatus” have vernacular meanings, and when scientists use a term from the public vernacular to describe a feature of science, confusion results when the vernacular term is an inappropriate description of that feature. Scientists might tacitly understand that global warming continues notwithstanding the “pause,” or they may intend “pause” to refer to differences between observed temperatures and model-derived expectations, but the public is not privy to that tacit understanding.

References

  1. N. Cahill, S. Rahmstorf, and A.C. Parnell, "Change points of global temperature", Environmental Research Letters, vol. 10, pp. 084002, 2015. http://dx.doi.org/10.1088/1748-9326/10/8/084002
  2. S. Lewandowsky, J.S. Risbey, and N. Oreskes, "The “Pause” in Global Warming: Turning a Routine Fluctuation into a Problem for Science", Bulletin of the American Meteorological Society, vol. 97, pp. 723-733, 2016. http://dx.doi.org/10.1175/BAMS-D-14-00106.1
  3. S. Lewandowsky, J.S. Risbey, and N. Oreskes, "On the definition and identifiability of the alleged “hiatus” in global warming", Scientific Reports, vol. 5, 2015. http://dx.doi.org/10.1038/srep16784
  4. K. Brysse, N. Oreskes, J. O’Reilly, and M. Oppenheimer, "Climate change prediction: Erring on the side of least drama?", Global Environmental Change, vol. 23, pp. 327-337, 2013. http://dx.doi.org/10.1016/j.gloenvcha.2012.10.008
  5. S. Lewandowsky, N. Oreskes, J.S. Risbey, B.R. Newell, and M. Smithson, "Seepage: Climate change denial and its effect on the scientific community", Global Environmental Change, vol. 33, pp. 1-13, 2015. http://dx.doi.org/10.1016/j.gloenvcha.2015.02.013

92 Responses to “Hiatus or Bye-atus?”

  1. 51
    Hank Roberts says:

    Op. cit., also featuring Dan H’s oft-rebunked claim, worth remembering: Berkeley and the Long-Term Trend

    Note the sizeable departure of the data from the linear trend over the last several decades….

  2. 52

    Dan H asks:

    “Does the multi-year increase in Arctic sea ice indicate that the long-term reduction has halted, and the sea ice extent will return to the large expanses of 1979?”

    I answer:

    Following the fourth-lowest minimum ever last September, a value (coincidentally) pretty close to right on the long-term (declining 13.4% per decade) trend line–NO!!!

    http://nsidc.org/arcticseaicenews/2015/10/

    Sorry to ‘shout’, but frankly, it’s a dumbass question.

  3. 53
    t marvell says:

    Temperature and C02 trends are cointegrated, which means that they can depart in the short term only. In the long term they have to move together. Any departure has to lead to a “snapping back.” CO2 has been consistently moving upward. I don’t know if you gave the economists the actual data. If you did, they should have seen that temperature is cointegrated with a trend. If not, they should have suspected it. Better yet, you should have given them the CO2 data,

  4. 54
    Richard Caldwell says:

    Kevin McKinney: I think you are on your own island on that, RC…

    RC: You’re probably right. Just call me Gilligan 2

    —-
    BPL: Please read: http://bartonlevenson.com/30Years.html

    RC: Um, your paper says nothing about volcanoes. Your paper ignores ENSO, even though it is by far the largest variable!! Your paper says nada about deep ocean temps. Your paper says squat about anything except a tiny slice of data. Yep, if we insist we know nothing except that one data series, you are right. But conclusions based on deliberate ignorance are unsupportable.

    Look at Foster and Rahmstorf (2011). They took the data series and added other data to explain most of the variation. Once those differences were quantified and explained to a degree, the time series completely changed, resulting in a much smoother climb.(and no pause!)

    And F&R only looked at solar activity, ENSO, and volcanoes. If they had included everything (and we had accurate measurements), then the results would have been an unwavering line, much like the CO2 curve once seasonality is removed. Variations in temperature increase measured are NOT variations in the temperature increase of the system, but simply deficiencies in our accuracy in measuring the lockstep increase that is actually occurring to the global system. Physics guarantees this. Energy cannot be created or destroyed.

    So, given that you are aware of F&R2011, why didn’t you calculate standard deviation based on all available data? By calling ENSO “noise”, you vastly overestimated the necessary time to determine the warming signal.

  5. 55

    RC @54: Your paper ignores ENSO, even though it is by far the largest variable!!

    BPL: Except that it is not “by far the largest variable,” if that phrase even means anything. I take it you’ve never had a stats course. You might want to google “analysis of variance.”

    Uncertainty is not your friend. Yes, ENSO and volcanoes and so on might delay the crash. Or they might bring it earlier. Assuming that ignorance is bliss is a bad idea.

  6. 56
    Dan H. says:

    Kevin,
    Then you agree that making long-term predictions based on short-term trends is ridiculous? Yet, posters here are doing just that with 2015 temperature data, which is not even finalized yet.

  7. 57
    Mal Adapted says:

    Hank Roberts:

    Op. cit., also featuring Dan H’s oft-rebunked claim, worth remembering: Berkeley and the Long-Term Trend

    Hank’s link is to a post Tamino wrote in response to Dan H.’s assertion of the same AGW-denier meme four years ago:

    On a recent thread which was not about the temperature trend, but about Judith Curry’s mischaracterization of it, “Dan H.” stated that what mattered was the long-term trend, which was a steady increase at a rate between about 0.006 and 0.0075 deg.C/yr, and that the Berkeley data reinforced this idea. He later said that it was a steady increase plus a cyclic variation with period about 60 years. Let’s examine those ideas closely, shall we?

    At the end of his very clear and concise analysis, Tamino concluded:

    The long-term linear model is nonsense. The long-term linear-plus-cyclic model is nonsense.

    This is proof of Dan H.’s resolute AGW-denial, and his virtually anosognostic inability to accept that he’s wrong.

  8. 58
    Hank Roberts says:

    the long-term linear model isn’t very good. In fact it isn’t right, which is easily confirmed statistically. The biggest difference between reality and the Dan H. model is the rapid upward trend over the last 30 years. Could it be … global warming?

    Tamino, op. cit.

    Tamino also debunks the “cycles” notion there, and replies directly to Dan H.’s rebunking of his old, old claims.

    I’d say why bother responding to them yet again, but as our hosts approve his posts so they continue to appear here, I am guessing our hosts want people to think about them — and respond to them.

    Sigh. Life is short, to be doing this over and over.

  9. 59
    Richard Caldwell says:

    BPL: Except that it is not “by far the largest variable,” if that phrase even means anything… …Yes, ENSO and volcanoes and so on might delay the crash.

    RC: I agree that the warming signal is larger than ENSO, but we’re talking about things which cause the record to jiggle up and down. Since the goal is to minimize the time period needed “resolve” the warming, then our first task is to tame the largest jiggles we can. That’s what F&R did.

    OK Mr. StatsCourse, when trying to resolve the warming trend in as short a time as possible, what variable(s) are more important than ENSO? Given the goal, do you think defining ENSO as noise or data gives a superior result? Why do you think F&R chose to treat ENSO as data? Do you feel your choice to treat ENSO as noise is superior to their choice?

    Not sure why you directed that last bit my way. ENSO can’t significantly help or hinder. We breezed by 1998’s spike with la ninas in around a decade. And the number of standard deviations needed for volcanoes to matter… It’d have to be massive and ongoing.

  10. 60
    Dan H. says:

    Mal,
    You may want to read what Gavin has to say about the recent pause and the long-term trend. “There is no evidence that the long term trend is really much different to what it has been.” – Gavin.

    “Recent research has implicated long term cycles in the oceans but there is no agreed mechanism with some papers attributing the pause to Pacific Ocean cycles, other research pointing to changes in the Atlantic and one recent paper saying that all the oceans are involved.”

    http://www.reportingclimatescience.com/news-stories/article/pause-over-within-10-years-says-nasas-schmidt.html

  11. 61
    Ric Merritt says:

    Dan H, are you making any predictions markedly different from mainstream climate science? Say, Arctic ice getting back to 1979-ish, or multiple decades of surface temps at < +1K / century? (They would have to be *way* less to keep the long-term trend you are so devoted to at that, considering the last 30+ years.)

    If so, I would love to work on terms for a substantial, public bet to top off my retirement fund. I'll treat you respectfully if you put your money where your mouth is. We could go on TV as 2 ordinary folks who care about this stuff, and make a little tour of it. (Nowadays, we can do that virtually, without needing airplanes.)

  12. 62
    SecularAnimist says:

    Hank Roberts wrote: “Life is short, to be doing this over and over.”

    Some folks seem to enjoy endlessly rehashing the same old “debates” with the same old deniers, and refuting the same old denialist talking points, over and over again. A great deal of the commentary on this site is devoted to that.

    I suppose it is somewhat gratifying, in that arguing with a global warming denier about the scientific reality of anthropogenic global warming is an argument you will always win.

    And affirming the reality of the problem is a lot easier than dealing with that reality.

  13. 63
    Silk says:

    #58 Hank, don’t do it. Seriously. There was a chap around here a while back, the last time I took to posting regularly (ish) here. Septic Mathew I think he was called. He’d raise the usual points. I’d answer them (or someone else would). Nothing difficult (I’m not a climate scientist). He would NEVER acknowledge that he was wrong, he’d merely pretend the entire discussion hadn’t happened then regurgitate some new (old) crap he’d read somewhere.

    I stopped engaging.

    I see he’s gone. Doubt that’s entirely due to me, but he was obviously only interested in being a time-waster.

    If you don’t engage they’ll get bored and go somewhere else (though I agree it would save everyone time and energy if the mods just removed posts that had been addressed 100 times already)

  14. 64
    Mal Adapted says:

    Dan H.:

    You may want to read what Gavin has to say about the recent pause and the long-term trend. “There is no evidence that the long term trend is really much different to what it has been.” – Gavin.

    You might want to actually read what Gavin has to say. By “long term”, he means since the mid-1970s, when the mid-century cooling ended following the Clean Air Act. Since you’ve apparently forgotten, you were talking about the trend since the Industrial Revolution got going:

    The long term trend based on CRU from 1880 is ~0.61C/century, while the GISS data is slightly higher at ~0.67/century. BEST appears to be ~0.75 (eyeballing since I do not have the actual data) since 1800.

    Look — just go to Kevin Cowtan’s temperature trend calculator and select the Berkeley global dataset, then experiment with different starting and ending dates. If you confirm what you already think you know, you’re doing it wrong!

  15. 65
    Hank Roberts says:

    affirming the reality of the problem is a lot easier than dealing with that reality.

    From my perspective that’s backwards, at least on an individual basis. But I’ve been dealing with this since the 1960s — poorly, approximately, often enough ineffectively, but I’ve known it was happening on my watch. Doing what seemed best was a lot easier than “affirming” it to people who found it incredible anyone could think this could be happening.

    I recommend John Brunner — Shockwave Rider and The Sheep Look Up — from that era. People I pushed to read those still tell me I was dammit right and why the hell wasn’t I better at convincing them back then.

    Alas.

  16. 66

    RC: OK Mr. StatsCourse, when trying to resolve the warming trend in as short a time as possible, what variable(s) are more important than ENSO?

    BPL: Carbon dioxide.

    Seriously, Google “analysis of variance.” Learn a bit about it. Your idea that stuff which “jiggles” on a shorter time scale is qualitatively different from stuff which does so on a longer time scale is part of your perceptual problem here.

  17. 67
    MA Rodger says:

    Silk @63.
    Dan H is (for some reason I know not) a troll of immense longevity here at RealClimate. There have been many times when commenters have requested his input be banned. If you look in the Borehole, you will find over one in eight of the comments are from Dan H.
    Dan H actually presents an argument of his own for AGW so he isn’t a troll in the true sense of the term. He argues that AGW is only responsible for a global temperature rise of 0.6ºC/century and all the rest is down to natural wobbles. Most of his comments are pushing that idea in some manner or other. The reason for this long-term linearity in AGW, apparently, is because CO2 is rising exponentially and the forcing from CO2 is logarithmic, so the temperature response will be linear. Assessing the trend over the instrument record yields that linear AGW signal to be 0.6ºC/century.
    Mind, that doesn’t explain a whole lot of things, not least why the last quarter of the instrument record contains three-quarters of the warming and thus has an average temperature increase ten-times bigger than the increase averaged over the first three-quarters of the instrument record. But, as we have learnt over the last five or six years, Dan H is incapable of seeing things any differently.

  18. 68
    Dan H. says:

    Ric,
    No, quite the opposite. I was simply pointing out the foolishness of making such predictions based on short-term trends. Mainstream climate science estimates that Arctic sea ice will dip below one million sq. km. in about 40 years. I take no exception to that estimate, but if you are interested in a placing a bet much earlier than that, I would be game.

  19. 69
    Sobieski says:

    In a real science, the gold standard is a comparison of out of sample test predictions vs. actuals, not putting a fake label on a chart.

    By that standard there has been a hiatus, if not failure of the models.

  20. 70

    Then you agree that making long-term predictions based on short-term trends is ridiculous? Yet, posters here are doing just that with 2015 temperature data, which is not even finalized yet.

    Dan, go back and find me one–just one!–long-term prediction based on ‘2015 temperatures.’

  21. 71
    D says:

    kevin,
    see posts #12 and #21 for starters.

  22. 72
    Ian George says:

    Tom O’Reilly@4
    It’s interesting that all these records have been broken since the ACORN system was introduced some 3/4 years ago. I will leave it to others to conclude if the new homogenisation/shading techniques have distorted the data.
    But the records that haven’t yet been broken are the actual highest temperatures for an individual site. For instance, Oodnadatta’s 50.8C set in 1960 is still Aust’s highest temp.
    According to the BoM, the highest individual temps for each month were never broken during these past three ‘hottest’ years. In March 2007, Canarvon AP tied with Roebourne for the equal highest temp (1998). Only 2002 and 2003 show top monthly temps for this century.
    All data here if you wish to check.
    http://www.bom.gov.au/cgi-bin/climate/extremes/daily_extremes.cgi?period=%2Fcgi-bin%2Fclimate%2Fextremes%2Fdaily_extremes.cgi&climtab=tmax_high&area=aus&year=2014&mon=3&day=27

  23. 73
    Wharfplank says:

    Not only stable, but declining…

  24. 74
    Lennard says:

    I don’t know how you selected the economists but I wouldn’t put much weight on the observations of this group of “experts” if they didn’t notice that the chart indicated that the World Agricultural Output totaled over 50 Trillion Australian $ or 50% of the total World GDP of roughly $107 Trillion AD. (Actual contribution is only about 6% so it should have been apparent with even the most cursory review of the chart)

  25. 75

    Dan, #71:

    kevin,
    see posts #12 and #21 for starters.

    #12 is talking about a 20-year comparison:

    “Temperature in the last 20 years haven’t really been “pretty flat” have they? NASA Giss has 1994 temperature anomaly of 0.32 and a 2014 anomaly of 0.74. That’s 0.4 oC rise in surface temperature in 20 years. 2015 is pretty certain to break the record….”

    (True, a straight comparison of anomalies in two separate years isn’t really a valid way to derive a temperature trend. But it’s a much more accurate illustrative statement than ‘the trend is flat’: according to WFT, the actual linear trend for GISS for ’94-’14 is ~0.15 C/decade, or a 0.3 C warming.)

    So is #21:

    “Finally, you say ask “So, is it going to carry on rising or start a decline?”, concluding that “observations show ‘pause’”, again with no evidence. I have no idea how you can hold on the the idea that observations show a pause, especially in light of the fact that 2014 was a record high, and that 2015 is pretty much guaranteed to exceed 2014 by a notable margin.”

    Nor do I see a long-term prediction made by either of those two comments. Both are simply disputing an (erroneous) assertion that the 20-year trend is flat. In doing so, they use notable facts–record temps in 2014 and (almost certainly) again in 2015 to illustrate or exemplify.

    Care to try again? Or did you really mean that you think that there is a distasteful overemphasis on the actual 2014 record and the pending 2015 one?

  26. 76
    Dan H. says:

    Kevin,
    Yes, choosing two separate years to make a comparison is not a valid way to derive a trend. While the linear trend over those two decades is as you calculate, that is 25% decrease from 1984-2004, and the lowest 20-year trend since 1962-1982. Selectively choosing a twenty-year trend will not show a flat trend. The pause has occurred for less time. In the GISS data, the pause has only lasted about 12 years. In the UAH data, the pause has lasted for 17 years (where neither 2014 nor 2015 are records highs). Even these analyses are simplifications of the overall data. Choosing a particular timeframe in order to make a certain point, does not accurately reflect the overall results. Regardless of the time frame chosen, recent temperatures have not experienced the same rapid rate of increase that occurred in the 1980s and 90s.

    http://www.coyoteblog.com/wp-content/uploads/2013/09/model2.gif

  27. 77

    DH @76: Regardless of the time frame chosen, recent temperatures have not experienced the same rapid rate of increase that occurred in the 1980s and 90s.

    BPL: Sample size matters. You need 30 years to tell a climate trend. Stop using “trend” to mean “temporary direction.” They are not the same.

  28. 78
    Dan H. says:

    BPL,
    Do not be ridiculous. There is nothing special about choosing a 30-year interval. A “trend” can be determined from much shorter intervals. Cherry-picking a 30-year trend, which gives a higher rate than either a longer or shorter time interval, just shows bias.

  29. 79

    DH @78: Do not be ridiculous. There is nothing special about choosing a 30-year interval. A “trend” can be determined from much shorter intervals. Cherry-picking a 30-year trend, which gives a higher rate than either a longer or shorter time interval, just shows bias.

    BPL: I am not being ridiculous, you are being militantly ignorant. The 30 year period was not “cherry-picked,” it was determined from the time needed to find the inflection in the standard deviation curve. I don’t believe you understand what a “trend” is. In fact, I don’t believe you understand any basic statistical analysis at all. Please read a textbook in the subject and work the examples, or take a course, before you post about this again. You are embarrassing yourself.

  30. 80
    Michael Wells says:

    Dan H. at #78: BPL has multiple times in this very thread posted a link to a page he put together explaining in detail exactly why 30 years is the time frame used. All you’re doing is sticking your fingers in your ears and yelling “La la la, I can’t hear you!” As usual.

  31. 81
    Richard Caldwell says:

    BPL: Carbon dioxide.

    RC: Dude, reading comprehension! I started with, “I agree that the warming signal is larger than ENSO,” Since you aren’t as stupid as most who would make such a mistake, I must assume you’re not-even-reading. That means you’re unwise. So, AGAIN:

    Why did you treat ENSO as noise instead of explainable data?

    ——-

    BPL: Your idea that stuff which “jiggles” on a shorter time scale is qualitatively different from stuff which does so on a longer time scale is part of your perceptual problem here.

    RC: Hmmm, “qualitatively different”? Where did you get that nonsense?? It doesn’t matter whether it’s a volcano which cools things for a few years or the PDO which might last longer or even a cloud which pops up for a week. They are all qualitatively the same, though they do differ in that there’s more or less only one ENSO while clouds are numerous and so average.

    We know that a proper summary graph of global warming would be rather smooth, with every formerly negligible drop due to a hurricane explained and smoothed. Since the question being asked is exactly answered by such a graph, our goal is to come as close to reproducing it as possible.

    So, logic dictates that we explain the wiggles on all time scales which obscure the pure smooth line we know exists underneath. With sufficient data and understanding of the system one could determine the climate trend even during a pause. A decade sounds reasonable. A year sounds sci-fi plausible.

  32. 82
    Steve Fish says:

    Comment by Dan H. — 17 Dec 2015 @ 11:12 AM

    Come on Dan, your comment- “A “trend” can be determined from much shorter intervals. Cherry-picking a 30-year trend, which gives a higher rate than either a longer or shorter time interval, just shows bias” -is disingenuous. You are just having a little fun stirring the pot.

    A trend line can be drawn on any data, but it may not indicate a real trend. The period over which a real trend can be detected depends on the signal to noise ratio. The 30 year number is just the period that commonly is required to define a trend in noisy climate data as statistically significant. Drawing short trend lines on noisy data is just silly.

    For a good visualization, see here: http://www.skepticalscience.com/graphics.php?g=47
    For a good explanation, see here: https://tamino.wordpress.com/2015/08/08/russian-roulette/

    In your previous post you talk about pauses. Pause means stopped for a short period. Look at the bottom graph in the Tamino link and see if you can find where warming has stopped for 12 or 17 years.

    Steve

  33. 83
    Richard Caldwell says:

    Dan H: There is nothing special about choosing a 30-year interval. A “trend” can be determined from much shorter intervals. Cherry-picking a 30-year trend, which gives a higher rate than either a longer or shorter time interval, just shows bias.

    RC: 30 years was the old standard given the state of the science at the time. Nowadays we can largely explain solar, volcanos, and ENSO, and the decadal oscillations are giving up their secrets too. What’s best-science today? Dunno, but 30-years can only be derived by ignoring all scientific advancement since the day the standard came to be.

    Be that as it may, choosing a 30-year trend is the opposite of cherry-picking. It’s the Standard.

  34. 84
    Hank Roberts says:

    Oh, good grief, Dan H is rebunking _that_ again?
    It’s arithmetic. Kids, Statistics 101 has the math, it’s a simple calculation, based on how much variability there is from year to year, if it’s annual data. The more it wiggles, the more years you need to look at to have a reasonable chance of saying there is or isn’t a trend. Stat 101. It’s specific to each individual collection of data. You can work this out for yourself, it’s high school arithmetic — see the 2 links below.

    You don’t need to take anyone’s word for it, beyond the arithmetic part.

    Dan H. has been refuted thus, repeatedly:

    More Grumbine Science: How to decide climate trends

    Dec 15, 2008 – How to decide climate trends …. look to see how long a period is needed for the result …

    More Grumbine Science: Results on deciding trends

    ——– but that’s just whack-a-mole exercise by now —-

    There’s plenty of fascinating material at the AGU website.
    Are y’all reading, those who want to learn?

    ——
    E.g.:
    PP32A: Emiliani Lecture
    Scientific Discipline: Paleooceanography and Paleoclimatology

    Searching for Tipping Points in Pleistocene Climate: Are They Real? Are They Portents for the Future?
    Speaker: Alan C. Mix

    Abstract
    Paleoscientists seek fundamental knowledge of Earth systems, and have unique views of system dynamics on the timescales longer than recent observations. This is important, because some processes that may control the trajectory of change in the not-so-distant future are not well illustrated or constrained in the short history of instrumental observations; paleo data offer our only observational window on these longer scales. But paleo vision is also a bit blurry, and sometimes biased. It matters how we approach the records to focus on the right things, and it matters how we communicate our insights so that we can teach models to approximate real Earth system behaviors. In the six decades since Cesare Emiliani set us on the path of quantitative observational paleoscience, we’ve observed more and more. Each year brings more “proxies,” more sites, more resolution, and the more we look the more we find interpretable signals. This data complexity makes the paleo literature daunting to outsiders; it is hard to get past the stories. Most of our narratives are grounded in linear systems, focusing on forcing and response, and we data generators struggle to infer even modest nonlinearity. But the concept of massive nonlinearity, tipping points, has recently entered the public consciousness, and the search is on anew for such events in paleo data. The reality of tipping points as well as the processes that may drive such events are not yet fully clear, but the search is essential because if humanity does indeed push the Earth system across an irreversible tipping point into a new state, we need to know how this might play out and over what time frames. Can we really learn about such occurrences, and the processes that control them, in paleo records? Do we need to think about the past differently? Will this change how we think about the future? Will it inform what we do about the future? Emiliani thought it should. This lecture will explore how.
    —–
    —– that’s on the Planetary Discovery channel, “Watch on Demand” since the presentation appeared.

  35. 85
    Hank Roberts says:

    Guys, don’t rebut Dan H. with opinion, that’s prolonging the exchange.
    Rebut him with citations, and remind him he’s been asked repeatedly to cite sources for his notions, although he never does. He’s boring.

  36. 86
    Ray Ladbury says:

    Dan H.,
    To paraphrase Inigo Montoya, “You keep using that word “trend”. I don’t think it means what you think it means.”

  37. 87
    chick keller says:

    Can anyone explain why satellite and surface temps seem to disagree on recent warming?

  38. 88
    Dan H. says:

    RC,
    Yes, thirty years was chosen as an average value, based on the best scientific measurements of its time. Variations were compared to this average value to determine deviation. However, it was not used to determine trends, as it was recognized that trends can occur over shorter or longer time periods. Using the 30-year averages to mean something other than original intended is disingenuous. A shorter term trend may indeed exist, but not cause a deviation from an intermediate (30-yr) term trend. Likewise, the thirty-year trend may be just a deviation from a longer term trend (i.e. a century or more). Selectively choosing a 30-year trend today is convenient for those wishing to show greater warming, because it includes the rapidly warming decades of the 80s and 90s. Those that confirm themselves to rigid criteria, end up being boxed in by their own rigidity.

  39. 89
    Dan H. says:

    BPL,
    Your condescending nature is not only inappropriate, but wrong. Just because we disagree, does not in any way enhance your opinion over mine. I know plenty about statistics, and how several people here and elsewhere use them selectively to show greater or lesser warming. Do not miss the forest for the trees.

  40. 90

    RC 81: With sufficient data and understanding of the system one could determine the climate trend even during a pause. A decade sounds reasonable.

    BPL: RC, what does “trend” mean in statistics? Can you define it?

  41. 91

    RC 83: 30-years can only be derived by ignoring all scientific advancement since the day the standard came to be.

    BPL: Except that I get 45 years by redoing the analysis. 2/3 of that for practical use. How far did you go in math, Richard? Do you understand how they choose a “long enough” sample in a given phenomenon?

  42. 92

    DH 89,

    It’s not condescension. You’re ignorant about statistics. Asserting you know it while repeatedly demonstrating that you’re confused does nothing to help you. You’re embarrassing yourself.

    Don’t you want to learn?