### 243 Responses to The significance of the “pause”

1. I am puzzled by this whole debate. Clearly there is no pause in the increased radiative forcing, as can be directly measured …

Click to access nature14240_v519_Feldman_CO2.pdf

I doubt there is any disagreement on this point. So the Earth system is warming, and the only question is where it is going. If one period it is 93% going into the ocean and the next period 94%, then maybe the atmosphere shows less warming than expected, but overall, the total Joules accumulated is going up as expected.

Internal variability is as expected.

Is the problem that contrarians and Journalists take the term ‘pause’ to mean a levelling off of warming, and this upsets many who want to emphasise that the Earth system is still warming.

Whereas those (maybe those even at Met Office) say that understanding the sub-30yr variability is an interesting question and worthy of investigation.

Who would disagree with that?

Maybe the pause label is problematic because it was gamed and we need different language. “To-be–explained internal variability events” … ain’t that catchy.

2. Richard,
I’m also a bit confused, but I think some of it might relate to a sense that it’s easy to look back and criticise, and partly because a lot of the research did indeed help us to better understand the role of variability. Personally, I think a lot of the issue did indeed relate to the terminology, rather than to the value in trying to understand variability and how it influences surface temperature trends.

3. brigittenerlich says:

That’s why I once called it wiggle in a wiggly (upward) trend line… that’s all it was… By contrast, using a word like ‘pause’ repeatedly even with the best intentions amplifies the meaning that those who ‘game’ with the word have chosen and they then win the meaning/significance game
http://blogs.nottingham.ac.uk/makingsciencepublic/2015/09/28/the-pause/

4. Brigitte,
Yes, I think this is indeed one of the key issues.

By contrast, using a word like ‘pause’ repeatedly even with the best intentions amplifies the meaning that those who ‘game’ with the word have chosen and they then win the meaning/significance game

5. brigittenerlich says:

But of course, my post and your post and…. all contribute to keeping that meaning/significance alive 😦

6. Yes, that is a valid point. I think that’s one of the key balancing acts when engaging publicly about a contentious topic. When does addressing something ultimately provide too much validity to the very thing you’re trying to refute?

7. Willard says:

8. verytallguy says:

If you define a pause as a period where the warning rate was lower than that expected long term, then I think there are two different questions:

(1) what caused this “pause”
(2) would such a “pause” be expected or was it exceptional?

These questions are (deliberately?) conflated.

It’s perfectly possible to explore (1) without answering (2) in the affirmative.

9. JCH says:

Let’s say warming tracks a similar path to that which occurred after 1998. Gee, not Trump will be saying the models are running too hot.

10. Steven Mosher says:

“(1) what caused this “pause”
(2) would such a “pause” be expected or was it exceptional?

These questions are (deliberately?) conflated.

It’s perfectly possible to explore (1) without answering (2) in the affirmative.

What VTG said.

I think that folks get mislead by the notion of statitical significance.

I look at the pause. Of course I can test whether or not there is a “change” in trend.
That test will be subject to a lot of decisions on my part. Arguable decisions.
( dk has some nice tweets on this )
Any way, what I am really after is physical understanding. If my confidence is low
that there is a change in trend ( say I’m only 60% confident) then the prospects for some
new physical understanding are low. I’m not gunna try to explain a “pause” that has a high probability of being mere statistical flucuation. But suppose I’m 80% confident that something
might be going on? worth my time? worth looking into? probably from my perspective
that is. But then I’m a temperature data geek. Its MY Thing. And maybe modelling guys
will say.. hmm, that falls on my space as well, maybe I could find something.

A paleo guy mught look at that and say ‘Not my thing” he is not gunna waste his time looking
for the physical cause of a phenomena ( a slow down) that doesnt rise to very high level of
statistical confidence.

He might look at it and say? hmm not statistically significant ( 95%) not worth my time.
I might look at it and say 75% confident? hmm. I wonder if there is something there.

The notion that there is this bright line “statistically significant” that is somehow objective isnt
very helpful. Practically I just translate the observed confidence into a “bet” . is it
worth my time to look deeper, or is low yeild mining. That ALSO depends on other
Present opportunities. What ELSE is there to look at? I see a 70% confidence level.
got nothing else to do, no other prospects for interesting things to look at.. ok, have a look
dig deeper.

This would be a better question for an actual Working scientist like ATTP.

on a practical level, what level of significance gets you curious? and do even things like
boredom or nothing better to look at play a role in your decision to dig deeper

11. Ken Fabian says:

There were scientists wanting to understand the nature of the variability – ‘pauses’, ‘reversals’, ‘accelerations’ [i]in the surface temperature[/i] versus those wanting the predictions of continuing warming to be wrong – ‘pauses’ as the early signs that global warming has stopped. And perhaps some open minded, but narrowly informed scientists sincerely wanted to know if overall warming really had slowed or stopped. Yes, the terminology matters more than it should, but outside of surface temperatures – looking at other measures and indicators of a warming world – it was clear that warming had not slowed or stopped.

I saw the way it got abused – that there was a prediction of x degrees per year of warming and more than a decade when that amount of warming wasn’t apparent (especially when taking 1998’s all time record global surface temperature as reference point and using the eyeball method) in surface temperatures. Yet just a couple more la Nina years than el Nino over a period that short would skew any average. Take out the ENSO impact (and volcanic aerosols and solar intensity changes) and even the mirage of warming stopped disappears – as Tamino showed.

12. Harry Twinotter says:

I started reading that article from John Kennedy and stopped one third in – it was too full of weasel words and general rhetoric for my liking. Anyway I have explained before why I reject the “pause”; the pause never existed in the first place in the IPCC AR5 report.

From a propaganda point of view the word “pause” itself is interesting. It means “stopped” in some sense, sure. But it also implies a temporary stop. Some did the climate change deniers intend this more-likely meaning – no I do not think so, they wanted people to believe global warming had stopped for good.

13. David B. Benson says:

There is also what statistician Grant Foster aka Tamino has written about the non-pause on his Open Mind blog.

Surely there are more important topics than this tempest in a teapot.

14. JCH says:

If there was no pause, then climate sensitivity is probably less than it was when there was a hiatus. Yes or no? How about when there was a slowdown?

15. dikranmarsupial says:

I normally try to avoid this side of the argument, but… It seems to me that Risbey et al suggest that there is no strong evidence for a pause (with which I would agree), but that there is perhaps something unusual about (presentation of) the research on this topic. The refusal to address the lack of statistical significance for a change in the average rate of warming seems to me to support that. IMHO if you ask a scientist a technical question that is relevant to the discussion, they should give a straight answer to that question (even if they then go on to explain why they don’t think it is important, as Phil Jones correctly did when asked about “no statistically significant warming since…”). Politicians who want to win an argument can evade questions that undermine their position, but if you are a scientist, you should be interested in the truth of the matter, and if something is true but inconvenient to your position, then I’m afraid you need to be able to acknowledge it.

In this case, I think NHSTs can do a useful job. If we perform a test for warming, we get an insignificant result. If we perform a test for a change in the rate of warming, we also get an insignificant result. This is telling us that just from looking at the GMST time series, there isn’t enough data to draw any strong conclusions either way. Self-skepticism is important in science, so if you want to propose some physical mechansim that explains some phenomenon, then you need to at least be open about whether there is even a phenomenon to be explained.

IMHO, that would have made it much more clear what scientists mean by “pause”, and made it much less easy to over-interpret the papers by the media and the polarised public debate. I agree it is easy to be wise in hindsight, but testing hypothesis and being open about uncertainties is part of basic scientific method. As is being clear in your definitions.

16. brigittenerlich says:

that’s the nub of the mater: being honest and “open about whether there is even a phenomenon to be explained”!

17. dikranmarsupial says:

Just to be clear, I agree with ATTP that just because the evidence for a pause is not statistically significant, that doesn’t mean that there shouldn’t be papers analyzing it, they should just be clear about what they mean by a pause, and open about the equivocal nature of the evidence.

In my field (machine learning) this can be a real problem. In my view, if someone has come up with some new machine learning algorithm, they should evaluate it properly on a range of datasets and only claim that their method performs better than existing methods if there is statistically significant evidence for that superiority. ML has gone through a series of hype-disappointment cycles over the years, and a lot of it is caused by a lack of self-skepticism about our work. However, if the difference in performance is not statistically significant, that doesn’t mean the algorithm shouldn’t be published. If it has performance competitive with the state of the art* and has other benefits, e.g. fast, needs less memory, has theoretical innovation or appeal etc. then that should justify publication anyway. However, reviewers are not always very good with this, or consistent. If you don’t have a statistically significant improvement then sometimes the paper is rejected purely because of that (which I think is wrong), on the other hand one of the papers I had most difficulty getting published was significantly better than the state of the art, but the reviewers repeatedly rejected it because the effect sizes were too small. Sometimes you can’t win! ;o)

* should we really expect anything more? I think it is reasonable to make the threshold alpha = 0.1 (say) rather than the traditional 0.05 as we should be surprised by a genuine superiority over the state-of-the-art, and we don’t want to stifle innovation (but still need a minimal hurdle).

18. KiwiGriff says:

Two pauses.

One: the scientifically interesting one with the trend in surface temperatures failing towards the low side of model projections,
It was not unexpected that such will happen at some time but why in this case?
This drove some worth while effort into further understanding ENSO,pdo,solar influences, china’s rapid industrialization and other interesting research questions. Even with a laymans understanding of AGW like mine, those who have a reasonable grasp of the science understood that global warming had not stopped or even slowed there was just factors effecting variability in the surface temperatures coming into play,

Two: The propaganda effort pushing a pause that allowed the idea that global warming was not happening or not preceding as the IPCC expected so we can pause efforts to reign in human emissions.

I dont see how it is possible to reconcile the two ideas,.
Scientists are always going to find excursions away from expected, the first meaning of pause, interesting. The language of science is not rhetoric . Scientists should not have to operate under an imperative to censor their language in an effort to counter their output being used as a propaganda tool. I dont think they could successfully do so even if they tried . There will always be those those willfully distort to push their own goals or are just willfully ignorant, not interested in understanding no matter what is said.

19. izen says:

‘Pause’ or ‘hiatus’ are terms used by activists when dealing with the scientifically informed.

When preaching to the choir, those of the de… legitimising AGW persuasion where much more explicit.
Monckton, Curry, Micheals all used the term ‘stopped’, ‘ceased’ or ‘reversed’ for the changes in land surface temperatures after 1998. Ignoring the unchanging warming evident in SSTs and OHC.
The ambiguous term ‘pause’ was used for acceptability within the scientific field.
Outside of that, the language was MUCH stronger about how AGW was refuted because there had been “NO global warming for 10/12/15 years.”

20. Harry Twinotter says:

JCH.

“If there was no pause, then climate sensitivity is probably less than it was when there was a hiatus.” I cannot tell if you are joking or not. I have read many good comments on this blog about internal variability.

“How about when there was a slowdown?” People on this blog have spent many words explaining there was never any slowdown. So what is it you are trying to say, exactly.

21. Harry,
I think JCH’s point is that the “pause” period was probably influence by variability in a way that lead to us warming less than we otherwise would have done had variability not influenced it in this way. Hence if you estimate climate sensitivity from observations you can get a result that is biased low.

22. angech says:

dikranmarsupial
“Just to be clear, I agree with ATTP that just because the evidence for a pause is not statistically significant, that doesn’t mean that there shouldn’t be papers analyzing it, they should just be clear about what they mean by a pause, and open about the equivocal nature of the evidence.”
Steven Mosher says:
“(1) what caused this “pause 2) would such a “pause” be expected or was it exceptional
I look at the pause. Of course I can test whether or not there is a “change” in trend.
That test will be subject to a lot of decisions on my part. Arguable decisions.
( dk has some nice tweets on this )”
Thank you both, and ATTP and JCH.
There are always pauses, just as there can be falls or sharp rises.
Are they significant?
Was there any significance?
Will there be any significance.?
Those who argue for no pause are already admitting that even the concept of a pause has great significance to them.
Those who say look a pause but who cares get it, it probably, most likely, is not of great significance so they can recognise but dismiss it.
Those who push it to say that AGW is not occurring are committing the same error as those who do not recognise it, they feel or need it to have great significance as well.
In summary most people who look at some of the global anomaly temp graphs will see that a natural pause is there from 1998. Natural because if one has an El Niño at the end of a temperature rise there must by definition be a pause or fall for a short time afterwards. Here the pause extended and then drew attention to itself. It most likely was natural variation that kept it down but it did stay down in some data sets for a quite extended period.
It will happen again and again.
It could given a run of La Nina’s and a prolonged downturn in natural variability even cause the dread pause to reform with seeming higher statistical significance.
Being a contrarian that is almost a hope for me, sometimes it takes a couple of lost battles to properly see and win the war.

23. BBD says:

Being a contrarian that is almost a hope for me, sometimes it takes a couple of lost battles to properly see and win the war.

You can’t win or even have a war against physics, angech. So long as net forcings increase, GAT will continue to rise. Natural variability will impose wiggles in the short term (decadal) and the longer term (multidecadal, centennial) trend will be set by the forcing increase.

24. “… and the longer term (multidecadal, centennial) trend will be set by the forcing increase.”

Complicated by a possible connection to this : “Does an Intrinsic Source Generate a Shared Low-Frequency Signature in Earth’s Climate and Rotation Rate?”
https://journals.ametsoc.org/doi/abs/10.1175/EI-D-15-0014.1

25. JCH says:

that’s the nub of the mater: being honest and “open about whether there is even a phenomenon to be explained”!

I guess it’s honest that the paper that invoked “pause” whatever it was, ~248 times?, is letting people know that the next pause will never happen because nothing probably caused nothing to happen the last time.

“Slowdown” – Jame E. Hansen

angech – you’re so wrong it’s beyond ridiculous.

26. dikranmarsupial says:

@angech had you paid any attention to the errors that have been pointed out in your arguments on this topic in the past, I might be willing to try and discuss it with you again now, but you didn’t and my patience (or gullibility) is not inexhaustible.

27. That whole rambling and unreadable article seems to be just about one of the tests used in the paper: the Monte Carlo bootstrap. Of course, that’s not the only argument presented and even still the criticism seems completely misplaced, IMHO. He is arguing that Risbey et.al. model is misspecified because the actual GMT is not correctly modeled as trend + noise. That’s irrelevant. All models are misspecified (see, for example: https://www.bayesianspectacles.org/a-galton-board-demonstration-of-why-all-statistical-models-are-misspecified/) and, more to the point, their statistical model is the same type of model used by the papers that talk about the pause. They all –either implicitly or explicitly– model GMT as noise + straight line (or possible multiple straight lines). As they say in the piece, a better model would need to be a GCM and there’s a companion piece (that I’m yet to read: http://iopscience.iop.org/article/10.1088/1748-9326/aaf372/meta) that analyses that side of the argument.

28. dikranmarsupial says:

SM wrote “on a practical level, what level of significance gets you curious?”

It depends on the nature of the problem. Fisher wrote in one of his books that the significance level should depend on the nature of the question and the experiment (or words to that effect). The significance level (and the statistical power) basically depend on you prior beliefs about the relative plausibility of the hypotheses. The real error made by the frequentist in the classic xkcd cartoon is that he followed the null ritual and used a 95% significance level, which is obviously nonsense given how unlikely it is that the Sun has gone nova!

Of course this is rather ironic as the frequentist programme was intended to eliminate the abhorrent subjectivity of Bayesian statistics, such as prior probabilities. NHSTs still have that same dependence, but it is hidden/obfuscated in the significance level, or ignored altogether (c.f. The “null ritual” where it is always taken to be 0.05).

29. ATTP, excuse me ahead of time,
but this post “The significance of the “pause” is a perfect example of playing by the Contrarians Script, rather than creating an informative science based ‘sticky’ explanation.

Manmade global warming is caused by increasing our atmospheric insulation.
This increases the warming our global heat and moisture distribution is sustaining 24/7/365.
http://skepticalscience.net/widgets/heat_widget/heat_widget.htm

How that heat gets distributed once its inside of our climate engine is immensely complex and endlessly fascinating – however it’s pointless discussing those complexities until the fundamental underlying physical cause of current global warming is clearly addressed and just as clearly acknowledged from whatever debate opponent. Before moving on to down stream issues.

PS. surface heat ≠ global heat
https://confrontingsciencecontrarians.blogspot.com/2018/12/stop-calling-surface-cooling-global.html ;- )

30. I think I now understand John’s criticism a little better. It seems to be that they essentially tested the hypothesis that the “pause” period is consistent with a continuation of the previous linear trend plus noise. Given that more than 5% of the trends of similar duration during the full interval had trends smaller than during the pause period, indicates that you can’t reject this hypothesis. Strictly speaking, this then doesn’t allow you to claim that there wasn’t a “pause” (which I think falls foul of the prosecutor’s fallacy).

However, the issue is mainly to do with how that period compared to what might have been expected (was it unusual). Therefore it doesn’t seem unreasonable (to me, at least) to use that you can’t reject that it’s a continuation of the previous trend plus noise to argue that there is little for some kind of “pause”.

31. Elio,
I think John’s post has a style that is representative of a fairly common British sense of humour.

32. “Strictly speaking, this then doesn’t allow you to claim that there wasn’t a “pause” (which I think falls foul of the prosecutor’s fallacy). ”

But the claim is not that there was no pause, rather that there is scant if any evidence of there being one.

33. Elio,
Yes, I agree. I was simply clarifying what I think John’s argument is.

34. izen says:

Perhaps it reveals a covert preference in the way AGW is discussed that the ‘pause’ of no significance has generated so much discussion, but the doubling (equally insignificant) of the rate of warming in the last ~8 years has attracted so little comment.

35. ATTP:

I think a great resource for understanding the wider debate about the “pause” is this short piece by Victor: http://variable-variability.blogspot.com/2015/03/how-can-pause-be-both-false-and-caused.html

When Risbey et.al. (and other et.als) reject the “pause” narrative they are saying that there is no evidence that the rate of long term warming has changed. When the “pause” researcher talk about the “pause”, they (at least the responsible ones) talk about the processes that explain the internal variability superimposed to the long term trend. No serious researcher in the field would say that it would be prudent to extrapolate to the future from the 1998-2014 rate of warming and would accept that a reasonable prediction would be that the earth would keep on warming at about 0.16°C per decade.

The problem, as I see it, is that either the public is too uninformed in climate dynamics and statistical analysis to understand this distinction, or some agents actively try to mislead them. And to be fair, the scientific literature is not so clear on the subject either. Risbey et.al. should have had a section making the distinction.

36. JCH says:

the rate of warming in the last ~8 years has attracted so little comment.

To the contrary, the evident resumption of no warming, brought on by the biggest cooling since 1950, the 8-year period to which you are referring is garnering lots of comments.

Is the 18-19 EL Niño cooked?

37. dikranmarsupial says:

ATTP wrote “Strictly speaking, this then doesn’t allow you to claim that there wasn’t a “pause” (which I think falls foul of the prosecutor’s fallacy). “

I don’t think the Risbey et al. paper quite goes that far (although it certainly hints at it), at least not explicitly. However I may have missed something. “While this is true, it is also important to ask what has been lost by the invention of a pause in global warming?” seems the most strident statement.

38. dikranmarsupial says:

As to whether there actually has been a pause, my own view is that it is likely an optical illusion caused by the 1998 El-Nino spike and problems with some of the earlier datasets, as mentioned by Risbey et al.

39. dikranmarsupial says:

rats! I was hoping it would show the subsequent tweet as well, rather than the previous one. Try again:

40. > That whole rambling and unreadable article seems to be just about one of the tests used in the paper: […]

The first paragraph offers two criticisms. First, it’s not a real review. Second, the authors’ target has been hidden in the appendix.

In the second, that the two main sections of the paper are disconnected. The disconnected list of authors may indicate patchwriting. I haven’t checked, but having authors writing different sections usually does that. It’s a practice that should be deprecated, in my opinion.

In the third, there are two connected criticisms: the authors are judging post hoc authors and topics they don’t name.

In the fourth, there is what I believe is the main claim:

It all seems rather a lot of fuss about a word.

The main argument I take away is a pragmatic one – it’s hard to criticize the usage of a word without referring to it. The criticisms of the methodology also backs up John’s position – just look at the contorsions they took not to utter the unutterable! According to my interpretation, the second part of the post, as with the paper itself, serves as a justification for delving into “but teh Paws” issues.

(“Teh Paws” is my own solution to that pragmatic problem.)

As for the rest of the article, if people could be clearer as to what is a trend and a trend of what, that’d be great.

41. No statistics have been harmed in the making of the following tweet:

Those who like John’s style (disclosure: I do) may appreciate David’s tweets. Definitely not the same style. Definitely personal.

42. I don’t think any of those criticisms is about the substance of the article. Moreover, for not wanting to “make a fuss about a word”, they end up making a fuss about the word “review”.

43. Even thought I think I now better understand John’s criticism, I do think that it’s missing the point that the section being criticised was really about selection bias (which is similar to what is known as the look-elsewhere effect). Essentially it was testing to see if there was anything special about the period referred to as the “pause”, given what we know about the longer warming. The answer appears to be that there isn’t really anything to indicate that it was particularly special period of warming, at least when compared to other periods of similar duration.

I think the argument here is that something (a big El Nino in 1998, probably) made people think that something special had happened, but they didn’t properly test to see if this could simply have been a form of selection bias.

44. I don’t think “review” is that of a loaded word, Elio, so you might need to perfect that jab.

John is not stuck with the same pragmatic problem as the authors. He has no qualms about mentioning Da Paws. As long as there are scientific merits in doing so, which I believe is what he tries to do in his second part.

He comes up empty handed, or at least finds little value in this kind of statistical fisking. Hence why he reaches no real conclusion. After complete snarky destruction comes a time to throws away Jacob’s ladder. Speaking of which:

Even when one tries to be constructive, dead ends can still be reached. My own suggestion is that those who’d like to delve into Da Paws should at the very least clarify what it is Da Paws of. The main trick of Da Paws is to omit what it is Da Paws of.

Showing that it’s a slowdown in some warming acceleration ought to be good enough.

45. dikranmarsupial says:

ATTP the second paragraph seems a good summary. The thing I don’t like is continued resistance to addressing the question now, which seems to illustrate the point about there perhaps being something unusual about the scientific discussion of the issue.

46. If John is allowed to make a fuss about an innocuous words such as “review” then I think Risbey is allowed to make a fuss about a big important word such as “global warming pause”.

Researchers are free to study the particular characteristics of the ups and downs of GMT or any other variable that tickles their fancy, but if there is no excuse of using the term “pause in global warming” when there’s no evidence of any “pause in global warming”. This would be true in any setting but it’s particularly important when deniers actively use the concept to mislead de public.

47. The simplest explanation of the pause has always been that temperatures since the end of the LIA are a sawtooth waveform superimposed on a secular rising trend. Trough to peak changes are driven by natural variability and may last as long as two decades. The pause is one of those.

I, along with quite a few others, were describing the pause in this fashion during the pause. What skeptics and we lukewarmers were mostly fussing about were the incredible lengths that the climate concerned were going to to pretend there was no pause. Hansen understood–he said GAT rises had ‘stalled’ for a decade and he didn’t know what part of natural variability had varied. He wasn’t too fussed about it and neither should have been many who advised the world not to believe our lying eyes.

48. dikranmarsupial says:

Tom wrote “Trough to peak changes are driven by natural variability and may last as long as two decades. The pause is one of those.”

which peak and trough, can’t quite make them out in the BEST dataset

http://www.woodfortrees.org/plot/best/from:1970

49. dikranmarsupial says:

“What skeptics and we lukewarmers were mostly fussing about were the incredible lengths that the climate concerned were going to to pretend there was no pause.”

and yet the skeptics and lukewarmers can’t demonstrate statistically significant evidence for its existence. This is rather the point.

50. izen says:

People who avoided the word ‘pause’ but were happy to use ‘stopped’.

Curry –
“‘There is no scientific basis for saying that warming hasn’t stopped,’ she said. ‘To say that there is detracts from the credibility of the data, which is very unfortunate.’”(2011)

Lindzen –
“There has been no warming since 1997 and no statistically significant warming since 1995.”(2008)
“…one can see no warming since 1997.” (2012)

Micheals –
“There has been no net warming for “well over ten years””(2012)
“Why hasn’t the Earth warmed in nearly 15 years?”(2011)

Pielke (snr) –
“…the global average temperature anomalies are cooling! “(2011)
“”There has not been warming significantly, if at all, since 2003, as most everyone on all sides of the climate issue agree. “(2011)

Plimer –
“There is no problem with global warming. It stopped in 1998. The last two years of global cooling have erased nearly 30 years of temperature increase.”(2009)

Salby –
“CO2 after the turn of the century continued to increase. In fact, if anything slightly faster. Global temperature didn’t. If anything, it decreased in the first decade of the 21st century.”(2011)

Singer –
“The atmospheric temperature record between 1978 and 2000 (both from satellites and, independently, from radiosondes) doesn’t show a warming. Neither does the ocean. “(2012)
“I conclude, therefore, that the balance of evidence favors little if any global warming during 1978-1997.”(2011)

Spencer –
“for some reason it stopped warming in the last 10 years, which is one of those dirty little secrets of global warming science”(2012)

Monckton –
“Since late 2001 there has been virtually no “global warming” at all.”(2011)

Lawson –
“so far this century both the UK Met Office and the World Meteorological Office confirm that there has been no further global warming at all.”(2011)
“Certainly, it is curious that, whereas their models predicted an acceleration in global warming this century as the growth in emissions accelerated, so far this century there has been no further warming at all.”(2009)

Rose –
“The world stopped getting warmer almost 16 years ago, according to new data released last week”(2012)
“for the past 15 years, global warming has stopped.”(2010)

Clearly 2011 was the peak year for the ‘Global warming has STOPPED’ meme.

51. The Very Reverend Jebediah Hypotenuse says:

What skeptics and we lukewarmers were mostly fussing about were the incredible lengths that the climate concerned were going to to pretend there was no pause.

Nope.
Nice gas-lighting of the “climate concerned” though, Tom.

Here we go:
Dr. Bob Carter, (James Cook University, Queensland), warming “stopped in 1998″.
Warming paused as of 1998 – Good.

David Whitehouse, 2001: “global warming has ceased.” and “The fact is that the global temperature of 2007 is statistically the same as 2006 as well as every year since 2001.”
OK – Warming paused starting in 2001, then.

Not to be outdone, Christopher Monckton of Brenchley: “warming stopped in 2002”. And he had a graph to ‘prove it’.
2002, then. Fine.

2005. Hottest year on record. Bloggers everywhere celebrated the end of global warming as of 2006.

2007 was pretty warm – But then 2008 was slightly cooler than 2007.
Blog-scientist Michael Asher: “So we saw global warming not just stop, but actually ‘reverse’ itself in 2008.”

But then: 2009 and 2010 were both slightly warmer than 2008. The pause paused.

Pat Michaels 2011: “Why Hasn’t The Earth Warmed In Nearly 15 Years?” – so the warming paused in 1996, then.

But Steve Goddard seemed convinced the pause began in 2002.

Some folks will go to incredible lengths to make a fuss about a non-fuss.

Try harder next time.

52. The Very Reverend Jebediah Hypotenuse says:

Clearly 2011 was the peak year for the ‘Global warming has STOPPED’ meme.

It’ll be back with bells on, son. Some memes never get old. Just tired.

53. izen says:

@-tf
“What skeptics and we lukewarmers were mostly fussing about were the incredible lengths that the climate concerned were going to to pretend there was no pause.”

What the climate concerned(?!) were mostly fussing about were the incredible lengths that skeptics and lukewarmers were going to, to pretend there was no warming.

54. You might have missed a few, Preacher Jeb and Izen…

https://www.nature.com/collections/jtrxxgrmgl

55. Tom,
What conclusions do you draw from those papers and the general research on the period often referred to as the “pause/hiatus”?

56. BBD says:

What the climate concerned(?!) were mostly fussing about were the incredible lengths that skeptics and lukewarmers were going to, to pretend there was no warming.

Or alternatively, that short-term variability was indicative of low climate sensitivity, which it isn’t.

57. JCH says:

Maybe it’s me, but I like mine the best as the Da Paws and the negative phase of the PDO are raising both hands:

58. Joshua says:

Tom –

“What skeptics and we lukewarmers were mostly fussing about were the incredible lengths that the climate concerned were going to to pretend there was no pause.”

Shorter version = “Mommy, they MADE us do it.”

A little birdie just told me that what the “climate concerned” were mostly fussing about were the incredible lengths thar “skeptics” and “lukewarmers” were going to to “pretend” [pretend is an interesting choice of term there, which asserts poor faith and a lack of integrity not a scientific disagreement] there was a pause.

Also interesting that in the thread directly downstairs you were asserting that virtually everyone is concerned enough to consider some amount of mitigation.

If no one is unconcerned, then why do you use the term “climate concerned” to label those who aren’t “skeptics” or “lukewarmers?”

Are you “climate concerned” Tom?

59. MarkR says:

I’m with Izen on this one.

The 10-year trend in GISTEMP now cannot rule out warming at 6 C/century.

I eagerly await all of the papers about the “surge” in global warming, and I’m sure that any day now I’ll be reading the following in an article with quotes from Judith Curry: “There is no scientific basis for saying that warming hasn’t SURGED”, she said. “To say that there is detracts from the credibility of the data, which is very unfortunate.'”

This public debate around this was fanned in a ridiculous and skewed way by people like Curry, Rose, Monckton etc. I hope they do better in future.

60. Hi ATTP, most people showed good common sense, treating the pause as natural variability. I remember Judith Curry positing that this should give us reason to think about the potential power of natural variability going forward. As I wrote earlier, there were a number on both sides that either wanted to pretend it never happened or that it was the beginning of the end of some huge global warming scam. But the experience overall provided me with more evidence that both numbers and common sense lie between the two extremes.

61. izen says:

@-tf
As you linked list of papers published on the issue shows, there were few, if any that wanted to pretend(?!) it never happened, just that, as they say in the US, it was a nothing-berder.

62. Joshua says:

izen –

Stop “pretending” that they weren’t “pretending.”

One might think that you doubt Tom’s mind-reading ability.

63. > If John is allowed [or free] to make a fuss about an innocuous words such as “review” then I think […]

You are free to think whatever you please, including counterfactuals that are both false and irrelevant. This one is false because no fuss over the word “review” has been made in John’s post. It is irrelevant to the points being made, i.e. making a fuss about a word over which we don’t want any fuss is self-defeating, and statistical analyses without any scientific interest are only made for ClimateBall’s sake.

To the last point I am free to add this other point – the paper has little ClimateBall value. Why? Since I am free to say whatever I please, I could simply assert it without argument. But I will add this argument of my own volition – look at the authors’ list. (I won’t mention the names, since I am free not to say them, and because I believe in learning by doing.) Each time I would be citing this paper on a ClimateBall field I would be opening myself to at least three different flanks, having to defend authors I see no reason to defend. Therefore it won’t be a paper I would often cite.

This argument, along my editing experience, leads me to offer the following prediction – this is a paper that will be self-cited first and foremost. I expect that this paper will be cited by inexperienced ClimateBall players too. I hope these ClimateBall players would improve their playbook, but I am not holding my breath.

64. Steven Mosher says:

Great cartoon DK !

some folks look at the temperature chart and expect (wrongly) to see a monotonic increase. When they dont they take that as a sign or something.

65. Steven Mosher says:

“Even when one tries to be constructive, dead ends can still be reached. My own suggestion is that those who’d like to delve into Da Paws should at the very least clarify what it is Da Paws of. The main trick of Da Paws is to omit what it is Da Paws of.”

yup.

66. dpy6629 says:

Here’s another well written post on this subject that looks quite convincing:

https://patricktbrown.org/2013/07/10/unforced-variability-and-the-global-warming-slow-down/

Hasn’t this horse been properly beaten to death at this point?

67. dpy6629 says:

And here we see that literally scores of peer reviewed science papers claimed often contradictory causes for the unmentionable while a couple said i didn’t exist at all. Is it 98% consensus that it existed?

https://patricktbrown.org/the-cause-of-the-pause/

But I forgot peer reviewed papers must be given high credibility. What if they disagree?

68. Willard says:

> What if they disagree?

As of 2014, the rate of global mean temperature increase over the 21st century has been less than that over the last quarter of the 20th century.

When a vehicule slows down to reach a stop, does it mean it stops?

Not if its conductor makes what we call a rolling stop, e.g.:

As of 0:09, the rate of the speed of the truck has been less than after 0:11.

Science communicators ought to beware that rates, like flows, are hard concepts to master.

69. Chubbs says:

dpy – The first Brown article, written in 2013, was interesting. It indicated that the timing of the end of the hiatus would determine whether temperatures were on a short leash, paced mainly by forcing, or on a long-leash, with more natural variability influence due to AMO/PDO. The end of the hiatus within a year or two of the article indicates that temperatures are on a short-leash.

70. izen says:

Following Willards lead, when there is a claim about changes in the trend of ‘Global mean temperatures’ it is wise to ask – “the GMT of what?”

‘The Paws’ is most easily imposed on satellite measured troposphere temps, then on surface temperatures, because the internal variability is related to the specific heat capacity of the system being measured.

If the climate system with the largest heat capacity is considered it is quite difficult to see any pause in warming since ~1985

71. Willard says:

Nice chart. No wonder JOULE ALL THE THINGS is only worth paternity testing these days, something that has been solved last year at AT’s.

Here’s a note by NG, whom as you may know is on my ClimateBall fantasy draft. It’s a bit technical, but he shows good graphs. Here’s one:

Here’s the comment:

The spacing between the lines is a good measure of the impact of El Niño and La Niña. All else being equal, an El Niño year will average about 0.2 C warmer globally than a La Niña year. Each new La Niña year will be about as warm as an El Niño year 13 years prior.

So we see a couple of recent La Niñas have caused the recent global temperature trend to level off. But be honest: doesn’t it seem likely that, barring another major volcanic eruption, the next El Niño will cause global temperatures to break their previous record? Doesn’t it appear that whatever has caused global temperatures to rise over the past four decades is still going strong?

So about that lack of warming: Yes, it’s real. You can thank La Niña.

As for whether this means that Tyndall gases are no longer having an impact: Nice try.

Notice the strong fall. NG bites the bullet, and adds only one technical concept, i.e. La Niña. He ends up by making the contrarian argument explicit. Sunlight is the best remedy to vices of reasoning. From the last two lines alone anyone can understand the short of it – it’s supposed to be colder, yet Da Paws in the global temperatures was only lukewarm. We all know why.

NG’s use of “lack of warming” is suboptimal (I mean, look at the graph) but at least he knows how to flip the contrarian script to communicate AGW matters in a simple manner and in an engaging way.

72. dpy,

And here we see that literally scores of peer reviewed science papers claimed often contradictory causes for the unmentionable while a couple said i didn’t exist at all.

Which is kind of one of the things Risbey et al. are highlighting; there wasn’t even a consistent definition for the time period, or what was actually meant by the term “pause/hiatus”.

73. dikranmarsupial says:

DPY wrote “But I forgot peer reviewed papers must be given high credibility. ”

No, just higher credibility than blog posts or newspaper articles. One of the first things I teach my graduate students is not to assume something is true just because it is published in a peer reviewed journal. IMHO it would be good for both the student and academic publishing if it were a requirement of a PhD that you debunk something published in a peer-reviewed journal.

74. dikranmarsupial says:

ATTP indeed, and using “the pause” to describe the model-observation divergence (not that that was statistically significant either IIRC – the 1998 El-Nino spike was above the spread of the models and nobody used that to claim the models were systematically wrong ;o) Whether warming has paused is a statement of reality and is independent of the models. If GMSTs pause and there are no modellers to model it, does it still pause? I’d say “yes”.

75. dikranmarsupial says:

Elio wrote “That whole rambling and unreadable article seems to be just about one of the tests used in the paper: the Monte Carlo bootstrap.”

FWIW I found it hard to read as well. The problem with that approach is that (i) it obscures the point you are making (c.f. ATTPs comments above) and (ii) if your criticisms are invalid, there is a greater loss of face in admitting so because of the tone with which you presented them. None of us enjoys admitting we are wrong, but in science it is very important that you do it, so it is best not to make it any harder than it really needs to be.

76. verytallguy says:

ThomasFuller

Hi ATTP, most people showed good common sense, treating the pause as natural variability. I remember Judith Curry positing that this should give us reason to think about the potential power of natural variability going forward

Funny that, I remember Judith Curry using politicised newspapers to falsely imply it was signficant.

‘There is no scientific basis for saying that warming hasn’t stopped,’ she said. ‘To say that there is detracts from the credibility of the data, which is very unfortunate.’

https://www.dailymail.co.uk/sciencetech/article-2055191/Scientists-said-climate-change-sceptics-proved-wrong-accused-hiding-truth-colleague.html

I also remember her predicting the pause would continue until 2030:

Our understanding of circulation regimes in the Atlantic and the Pacific Oceans suggest that the pause will continue at least another decade, perhaps into the 2030s.

Click to access Judy-Curry-2015.pdf

Definitions of “good common sense” may differ, I guess.

77. izen says:

@-dk
“FWIW I found it hard to read as well.”

Perhaps it is a cultural thing, I found it quite engaging and that helped understand the points he was making.
Suggesting that a topical review might be something you rub onto dry skin made me chuckle…

78. izen says:

@-vtg
“Funny that, I remember Judith Curry using politicised newspapers to falsely imply it was signficant.”

To give tf the benefit of the doubt(?), Judith Curry is not ‘most people’.
She is part of the crank fringe.

79. izen says:

@-Willard
Just a ‘heads up’ the graph link you posted from NG does not appear for me, I get –

451 Unavailable For Legal Reasons
Sorry, this content is not available in your region.

Ah, the joys of the European General Data Protection Regulation, no wonder some people though Brexit was a good idea…

80. dikranmarsupial says:

izen the problem is that serveral of the points were misrepresentations/misunderstandings (AFAICS) of the paper. It seemed clear to me from the Twitter discussion that he didn’t understand the statistical point made in Risbey et al. (despite being fairly standard frequentist hypothesis testing procedure). Easy to be distracted from considering the value of the criticism by focusing too much on the humour. Much prefer our host’s style.

81. Joshua says:

Izen –

To give benefit of the doubt, Judith’s views on climate change aren’t those of a crank fringe. There’s lots o’ people who share her views, although not a very high % of scientists in the field do.

82. dikranmarsupial says:

Prof Curry said: “There is no scientific basis for saying that warming hasn’t stopped”

Equally there is no scientific basis* that it has stopped, which is rather the point made by Risbey et al. I find it odd that it is apparently so difficult to present a balanced evaluation of the evidence where it is, at best “equivocal”.

*In science generally we require statistical significance when making assertions based on statistical evidence (post-hoc explanations for specific phenomena are, at best, rather weak evidence and circular if the explanation is also the research hypothesis).

83. dikranmarsupial says:

presented without comment ;o)

84. angech says:

Interesting, very hot here in southern Australia at the start of this year but you guys and my son in Crete are freezing. JCH shows some unexpected cooling of the predicted El Niño, guess 7 out of 8 models got it wrong only 6 weeks ago but it is hard to predict the future.
This should mean a little tropical ocean cooling to add to the large northern land mass freezing.
Overall the UAH for January should fall.
I look forward to an overall drop for 2019 and 2020 as the only way of easily resuscitating the pause. Still a fair way off at the moment.

85. angech,

Interesting, very hot here in southern Australia at the start of this year but you guys and my son in Crete are freezing.

Australia does indeed seem to be currently hotter than it would normally be, but the main reason why it’s hot where you are and cold where we are, is that it’s summer where you are, and winter what we are.

86. Joshua says:

angech –

Why do you have such a keen interest in such short term weather trends?

87. JCH says:

UAH is the airport for Feynman’s cargo cult.

88. BBD says:

Why do you have such a keen interest in such short term weather trends?

Because denialist rhetoric.

89. BBD says:

Angech

Climate change is really starting to bite in Oz, isn’t it?

You can’t have, let alone win, a war with physics. Look outside.

90. JCH says:

For 35 years, the Pacific Ocean has largely spared West’s mountain snow from effects of global warming

… A new study has found that a pattern of ocean temperatures and atmospheric circulation has offset most of the impact of global warming on mountain snowpack in the western U.S. since the 1980s. …

I follow the weather in the Pacific because it’s big, and rules so many roosts.

91. Jeffh says:

Actually, aside for a few days in early January, temperatures in Crete have been normal or even slightly above normal. I find Angech’s blithe posts usually mildly amusing, but given the massive ecological consequences occurring because of repeated, unprecedented heatwaves occurring in Australia in recent years, it makes me really wonder if he is on the same planet as the rest of us. For example, it is estimated that last November’s monstrous heatwave in Queensland killed off as many as 75% of the white spectacled flying foxes (bats) found there. Northwestern Australia has experienced a month of days over 40 degrees maximum, and even a day over 50 degrees. The ecological effects of these heatwaves are impossible to fully comprehend, but they are almost certaintly an emerging disaster.

92. Willard says:

> Sorry, this content is not available in your region.

If hotlinking becomes regulated, we’ll have problems reading past articles on the Internet quite soon. Try this:

93. izen says:

@-Willard
“If hotlinking becomes regulated, we’ll have problems reading past articles on the Internet quite soon.”

It is not really the fault of the regulation. Sites could allow access and hotlinks, but are unwilling to enable views without being able to download your personal browsing history for monitiization.

94. Dave_Geologist says:

JCH: If there was no pause, then climate sensitivity is probably less than it was when there was a hiatus. Yes or no? How about when there was a slowdown?
ATTP: I think JCH’s point is that the “pause” period was probably influence by variability in a way that lead to us warming less than we otherwise would have done had variability not influenced it in this way. Hence if you estimate climate sensitivity from observations you can get a result that is biased low.

If this paper. is right (although it appears to be a non-peer reviewed preprint, and the authors themselves call it only a hypothesis), then yes, TCS/ECS estimates which are dominated by or disproportionately influenced by the faux pause will be biased low. The implication would be that albedo is underestimated during cool PDO phases. As this work is new and not (yet) mainstream, authors who pick up a mainstream forcing history will be inputting too much forcing during the “pause”. There were one or two more mainstream papers saying something similar about coupling between cloudiness above the oceans and PDO cycles. That is the sort of “pause”-inspired research which is useful, because it leads to better understanding of feedbacks and internal variability, as well as perhaps correcting the forcing history. It doesn’t affect the long-term warming trend because that encompasses multiple PDO cycles, so if there is a genuine decline in planetary heat uptake during part of the cycle, it will already be incorporated in the 30-year rolling average.

Ironically, it was heavily promoted by the GWPF. That’s the problem with reality denial: while real science is all-joined-up, fake science is inconsistent and contains mutual contradictions. It’s like the people who shout “Little Ice Age”, without realising that a strong insolation response implies a higher ECS, not a lower ECS.

“Look daddy, clouds!”. “Sorry son, the good news may be that some of that heat isn’t hiding in the oceans after all. But the bad news is that the climate is even more sensitive to CO2 emissions than we thought, certainly more than anyone looking at the last decade or two thought.”

95. Dave_Geologist says:

dikran: “which is obviously nonsense given how unlikely it is that the Sun has gone nova!”

That reminds me of the time the Geiger counter started chattering in the X-Ray lab when I was doing my PhD. A frequentist would have run out of the room, perhaps taking the counter and manual with him to see if there was any information on equipment failure rates. We were relaxed because we knew that by far the most likely explanation was that the counter was faulty. All the equipment emitted hard X-Rays in a vacuum, and everything was in metal boxes with an alarm that went off if you powered it up with a hatch open (plus you’d hear the air hissing if a seal had failed!). We nevertheless checked that all the equipment was properly enclosed (it was ). Then took it out to the street, where it was still chattering. It was at the height of the Poland/Afghanistan crises, when the Doomsday Clock was just shy of midnight, and we did genuinely have a conversation about whether the Geiger counter was broken or WWIII had started. We decided that because there were no sirens and everyone was going about their business, the Geiger counter must be broken. So we went back into the lab and took out the batteries. But left it on the shelf in case anyone from HSE dropped by.

96. dikranmarsupial says:

:o)

The problem isn’t really the frequentism, it is just not thinking about the appropriate significance level.

97. JCH says:
98. dikranmarsupial says:

JCH splendid.

angech describing a paper by Lacis, Schmidt, Rind and Reto:

Not surprising seeing Gavin and Andy together.
To think top scientists think they can fool people.
Or even worse fool themselves. Scrub the second option, they are not foolish but certainly incredibly misleading.
As is anyone putting up this parody of science as an argument.

Being reminded of angech’s behaviour elsewhere confirms I was right not to respond to his trolling here.

99. Dave_Geologist says:

True dikran. Although the significance of sticking your hand in the primary beam is quite significant. The radiation safety officer got the attention of newbies by showing the skin graft on the back of his hand. He’d stuck it in for a second (which is like sticking your hand in front of an industrial laser), and pulled it out the instant he felt the pain. I presume he’d disabled some safety interlocks to do repairs or some such. The Geiger counter manual probably quotes a failure rate of once in a hundred thousand hours or some such, so it was unlikely. But absent an extreme event we’d have noticed, like someone opening a hatch they shouldn’t have, the chance of X-rays escaping was zero. That’s why we were told to take our radiation badges off when we went to the dentist or through airports. The only acceptable dose when the badge is examined is zero. Anything else means a primary bean has escaped and the badge has picked up some back-scatter. Your own dose could be orders of magnitude higher elsewhere on your body so you’d be in for a very thorough examination.

OTOH in a radiography, materials testing or similar environment you would be much more attentive to the possibility that it was not a false positive. There you have exposed beams with variable intensity and duration, sometimes portable, so it’s quite plausible that someone’s forgotten to switch something off when they put it down, chosen the wrong setting, or that something that was pointing at a piece of pipework is not pointing at you.

100. Dave_Geologist says:

oops, “now pointing at you” 😦

101. Dave_Geologist says:

If hotlinking becomes regulated, we’ll have problems reading past articles on the Internet quite soon.

And I’ve forgotten my Old Year’s Resolution to always give the full titles of papers I link to, so they can be found on Google Scholar if the link dies. The cloud one: Reinforcement of Climate Hiatus by Decadal Modulation of Daily Cloud Cycle

102. Willard says:

Doc may also like:

Replies I liked includes “I’m sorry Isaac, the market for this book appears to be…Leibniz. Yeah, just him” and “Maybe you could write a more popular book, like that one Hooke did with the pictures!”

That last one would be me.

103. dikranmarsupial says:

:o)

104. JCH says:

I just bought 45 more acres of twaddlewood forest. Because I think it’s important to remove twaddle from the atmosphere.

105. BBD says:

IIRC, the initial recent use of that word was here, by me. I’m going to sue for plagiarism… 🙂

106. dikranmarsupial says:

At least Newton has his pictures in the Royal Society ;o)

107. That reminds me of the time that we were monitoring a perfect Gallium Arsenide single-crystal wafer under UHV conditions. Suddenly the signal started to go through perfect oscillations. We were wondering what that was all about and whether it was due to artefacts of the envrionment — so all we did was change the temperature of the substrate and within a day we could confirm that this was layer-by-layer sublimation of the GaAs surface atoms, as the heat of vaporization calibrated perfectly to the changing period of oscillation. That was the first time anything like that was reported in real-time (we beat some Japanese researchers to this finding by a few months).

All it involved was fitting a sinusoidal signal. Unfortunately that kind of fast turnaround via experiment in a controlled environment is not possible in climate science.

108. dpy6629 says:

Back to the topic of the unmentionable. One must bear in mind that the temperature datasets have continued to be adjusted and modified since 2013. The result was to make the unmentionable less obvious. Here’s an update of the AR5 chart on the CMIP5 runs vs. predictions from Clive Best (which original chart caused the IPCC to predict less warming than the model mean in the medium term).

It appears that the unmentionable problem of divergence has not completely gone away. Of course on many many other measures of skill such as cloud fraction as a function of latitude, models fail rather badly and this impacts energy fluxes in the system too. Is there anyone left who seriously defends the climate model mean as a good forecast? SOD has a number of good recent posts on this.

109. Willard says:

> Here’s an update of the AR5 chart on the CMIP5 runs […]

From the second paragraph of the paper, DavidY:

The form of alleged climate ‘pause’ varies across the literature, but essentially involves calculation of a short-term trend in global mean surface temperature (GMST) over a decade or two, which is then compared with either other periods in observed GMST (Stocker et al 2013), or with trends estimated from coupled climate model projections (Fyfe et al 2013, Risbey et al 2014). This review addresses the former issue (comparison of observed trends), while a companion review (Lewandowsky et al 2018) addresses the comparison with climate model expectations of trends.

http://iopscience.iop.org/article/10.1088/1748-9326/aaf342

Your “back to the topic” might not mean what you make it mean.

110. Steven Mosher says:

“Back to the topic of the unmentionable. One must bear in mind that the temperature datasets have continued to be adjusted and modified since 2013. The result was to make the unmentionable less obvious. ”

huh?

The biggest change has been doing apples to apples comaprisons

SAT + SST versus SAT and SST from the models, something someone recommended years
ago.

I don’t believe clive does that.

what you meant to say is temperature datasets have continued to improve
and comparison metholodoly has matured, and as a result apparent diferences between
models and observations have narrowed.

progress. no ‘real’ science involved unless you want to call data analysis a science. meh

111. izen says:

@-dyp
“It appears that the unmentionable problem of divergence has not completely gone away. Of course on many many other measures of skill such as cloud fraction as a function of latitude, models fail rather badly and this impacts energy fluxes in the system too.”

But the models of Ocean heat content and its rate of uptake and release do show a good level of skill in matching, and predicting, the observations of this metric.

https://www.nature.com/articles/nclimate2915
“We show that the multi-model mean constructed from the current generation of historically forced climate models is consistent with the OHC changes from this diverse collection of observational systems.”

http://science.sciencemag.org/content/308/5727/1431
“Our climate model, driven mainly by increasing human-made greenhouse gases and aerosols, among other forcings, calculates that Earth is now absorbing 0.85 ± 0.15 watts per square meter more energy from the Sun than it is emitting to space. This imbalance is confirmed by precise measurements of increasing ocean heat content over the past 10 years. ”

And as models of turbulent mixing at various scales have improved, the connection between surface temperature variation and variations of ocean heat uptake have become clearer.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5943251/
“It is still unclear whether a hiatus period arises due to a vertical redistribution of ocean heat content (OHC) without changing ocean heat uptake (OHU), or whether the increasing radiative forcing is associated with an increase in OHU when global mean surface temperature (GMST) rise stalls. … we show that in climate models TOA radiation and OHU do anti-correlate with natural variations in GMST, when GMST leads or when they coincide, but the correlation changes sign when OHU leads. Surface latent and sensible heat fluxes always force GMST-variations, whilst net surface longwave and solar radiation fluxes have a damping effect, implying that natural GMST-variations are caused by oceanic heat redistribution. ”

Whatever divergence there is between models and GMST variations, even the ‘unmentionable’ ones, they may turn out to be because of the uncertainties in modelling the variations in the various ocean quasi-cycles like ENSO, PDO and AMOC. These are much larger flux changes than the variations in flux caused by any errors in cloud fraction or distribution.
Some researcher have suggested these ocean cycles are chaotic/stochastic, although I suspect WHUT may have an opinion on that…

112. dikranmarsupial says:

DPY wrote “One must bear in mind that the temperature datasets have continued to be adjusted and modified since 2013. ”

Yes, that is what competent scientists do when they find biases, issues and errors in the raw data and/or homogenisation methods. SM also makes a good point about performing like-for-like comparisons.

The interesting thing is that in the historical section of that diagram, the observations spend a lot of time in the upper half of the model spread, but interestingly climate skeptics don’t want to talk about that. All we can reasonably expect is that the observations will lie in the spread of the model runs, nothing more. And as vv will tell you, there are good reasons to think that the model spread under-represents the true uncertainties (so it is not *that* surprising if they are near the tails sometimes.

113. Chubbs says:

To follow-up on SM’s ggod point. The CMIP5 line in Clive’s figure needs to be shifted up by about 20% to make an apples-to-apples comparison with the observations, which use SST vs the model’s air temperature over the ocean. It wasn’t until 2015 that the climate science community realized the size of the difference between SST and temperature over the ocean in climate models. Before 2015, climate science was underselling model performance. Even today Real Climate and Climate Lab undersell climate model performance in their widely viewed model/observation comparison charts.

The ocean temperature discrepancy also impacts EBM and model comparisons. Roughly 20% of the difference between model and EBM ECS/TCR estimates is due to faulty ocean temperature comparison. Think how many blog arguments could have been avoided by making this simple correction.

114. JCH says:

Think how many blog arguments could have been avoided by making this simple correction. – Chubbs

Well, since we are now well into the next ~19 year Da Paws, and what could be a “little dipper” divergence, we’ll soon know the answer.

115. Chubbs says:

JCH – Yes, a poor choice of words on my part. The arguments themselves have little to do with anything specific. Perhaps that is the true significance of the pause.

116. dpy6629 says:

“The interesting thing is that in the historical section of that diagram, the observations spend a lot of time in the upper half of the model spread, ”

That’s almost certainly due to using a long baselining period. If two curves have significantly different slopes, using a long baselining period will make the first part of the lower slope curve be above the other one and the last part of the lower slope curve be below the other one. That’s exactly what is shown here. This is a standard way to manipulate how the graph “looks”. If you want to emphasis the divergence, you use a very short baselining period at the beginning which makes the final point difference look bigger. If you want to deemphasize the divergence, you use a very long baselining period and hope no-one notices the divergence at the beginning of the plot.

In any case, just because the data lies within the spaghetti range means little because the models involved are not a randomly distributed sample of realizations of the climate. This almost certainly understates the true uncertainty as Zhou et al provides evidence for.

117. dikranmarsupial says:

DPY so what baseline period was used? Can you give an example of a baseline period that you think would be acceptable for looking at the 1990-2000 period?

It doesn’t need to be a *random* sample. If the observations lie within the spread of the models, then they are consistent with what the models predict might happen. I’ve already said the spread of the model runs understates the uncertainty, for a start it doesn’t include parameter uncertainty, but that weakens the evidence for a model-observations disparity, not increases it.

118. dpy6629 says:

Well, Steven, what you state as fact concerning SST vs. SAT is a matter of controversy with conflicting data and interpretations.

119. Everett F Sargent says:

I don’t think either paper presents a proper in context historical perspective. Everything shown is some form of post hoc reanalysis (blended, masked, updated forcings, updated time series, omit the SAT and omit the radiosonde data sets (note reanalysis data sets is always post hoc, thus would not be present in an historical context)).

The in situ record, at that time (and misuses and/or misinterpretations thereof), clearly shows something occurred that generated 200-300 peer reviewed papers in that era, and I don’t consider ‘so called’ seepage as a valid excuse for the entire set of these peer reviewed papers.

120. Everett F Sargent says:

“In any case, just because the data lies within the spaghetti range means little because the models involved are not a randomly distributed sample of realizations of the climate. This almost certainly understates the true uncertainty as Zhou et al provides evidence for.”

You do realize that if the CMIP5 error range underestimates uncertainty then the observational data will fall even closer to that larger uncertainty range? D’oh!

121. Marco says:

“a matter of controversy with conflicting data and interpretations.”

Citations needed!

122. Ken Fabian says:

“It appears that the unmentionable problem of divergence has not completely gone away.”

Most of the divergence within the graph shown is between high emissions pathways and low – not primarily that of uncertainty of modelling and projection but uncertainty about how much effort we make to reduce emissions. A (naive or intentional?) misinterpretation that seems to encourage the view that we don’t know that high emissions will actually produce more warming than low, so why bother reducing emissions.

123. dpy,
What shouldn’t really be all that controversial is that we should be comparing like with like.

124. izen says:

@-dyp
It is much harder to model the atmospheric surface temperatures because the low heat capacity enables large fluctuations when ocean heat uptake varies slightly.
Models, measured [TOA] energy imbalance, and Ocean heat content all match well, and show little sign of divergence, or Paws.

The ‘divergence’ between models and GMST observation is largely a consequence of the inherent variability of the air surface temperature. The ENSO quasi-cycles have a big effect
In the overlaid graph above of air surface temperature, the La Nina years are coloured blue in the record, the El Nino years red.

The Ocean heat content is well matched by models, and follows the measured energy influx measurements.

125. izen says:

[Fixed. -W]

The big volcanic eruptions also show up far more in the GMST

126. verytallguy says:

One must bear in mind that the temperature datasets have continued to be adjusted and modified since 2013…

…This is a standard way to manipulate how the graph “looks”

Are these “standard” ways to imply malfeasance, or novel ones? We deserve answers.

Oh, and dpy – citations. Please, if you want a technical discussion, merely mentioning the name of your source does not suffice. A full citation, ideally a link, is necessary. Thank you so much.

127. Steven Mosher says:

“Well, Steven, what you state as fact concerning SST vs. SAT is a matter of controversy with conflicting data and interpretations.”

Not really.

The GCM predict temperatures at all locations and at various levels depending on how they grid
their oceans and atmosphere.

For example, a GCM will compute an air temperature on a 100km grid, for example, at various hPa. And for oceans it will compute SST at various depths.

Now comes the data analysis brain teaser.

If you have observations at 2meter above the land (SAT) , and if you have SST observations in the first meter of water (SST), Then… do you

A) compare the GCM air tempature at 2m over land AND water to the observations
without regard to the fact that observations are SST and not MAT ( marine air temp)
B) compare the model prediction of air temps (SAT) over land with the observations over land
(SAT)
AND
compare the model prediction of SST with the observations of SST.?

This is not a hard data analysis question. However, you will find skeptics continuing to
do it the wrong way. Most in fact. ( it may be due to the fact that they grab data from KNMI
which is easy to get, as opposed to going the the primary source)

This insnt even a climate science question. david, if you had a model that predicted drag over the wing, and I compared that to measured drag around the vertical tail, would you call that
a good way to evaluate the model? I bet not.

If my model predicted temperature at 2meters and I compare it to observations at 20000 meters
would that be good?

If my model predicted temperatures everywhere, but I only had observations at 50% of all the locations… how would you compare them?

This question isnt a climate science question at ALL. If this were any other field you would have no issue with comparing models in the correct way to observations.

128. dikranmarsupial says:

DPY wrote

That’s almost certainly due to using a long baselining period. If two curves have significantly different slopes, using a long baselining period will make the first part of the lower slope curve be above the other one and the last part of the lower slope curve be below the other one. That’s exactly what is shown here. This is a standard way to manipulate how the graph “looks”. If you want to emphasis the divergence, you use a very short baselining period at the beginning which makes the final point difference look bigger. If you want to deemphasize the divergence, you use a very long baselining period and hope no-one notices the divergence at the beginning of the plot.

Except (having thought a bit more about this) the divergence would be the other way round. If the models over predict the warming, then the observations would be at the bottom of the model spread not the top.

You show YOUR bias by maligning the reason for using a long baseline. The reason for a 30 year baseline is so that the effects of internal variability (primarily ENSO) don’t affect the result, which makes it difficult to cherry pick. That and your challenge about controversy over the like-for-like comparison shows you are just bullshitting and don’t really care whether your arguments are correct or not.

129. izen says:

Just to clarify, I scaled, or roughly normalised global mean surface temperatures to cover the same scale as the increase in Ocean heat content in the animated graph posted above.

This comparing apples (Joules) to oranges, (degC) but helps visulise the difference in variability between the two metrics.

Obviously we live in and are most directly affected by surface air temperature, but it is the wandering dog, constrained on a leash by the underlying energy increase, and ~90% of that is in the oceans.
No paws in the OHC rise.

Given the relative sizes of the energy resoviors, the added energy in the atmosphere is around an order of magnitude smaller than the increase in energy in the oceans. This is why the ocean heat content dominates the evolution of the climate.
Perhaps this is clearer if I scale the GMST to more closely comparable with energy gained in Joules. The oceans have gained ~30×10^22 Joules. The surface temperature rise required about a tenth of that. It may give a better indication of how although what we most care about is the surface temperature, many scientists have emphasised that the much more important determinant is how much energy is accumulating in the oceans as CO2 accumulates in the atmosphere.

130. izen,
What’s the bottom graph? Is it the heat accumulated in the atmosphere? The mass of the atmosphere is about $5 \times 10^{18}$ kg and it has a heat capacity of 1000 J/kg/K. So, that would seem to suggest that a 1K rise would be associated with a $5 \times 10^{21}$ J increase in energy (about 2% of that accumulated in the oceans).

131. izen says:

Oh dear.
I was taking the common trope of the oceans having 90% of the heat and scaling by x10.
So the bottom graph is about 5x larger than it should be…
Back to the drawing board, or at least photoshop.!

132. You could do it like that, but the numbers are more like 93% for the oceans, and 2% for the atmosphere (with continental land mass and sea/land ice taking up the remaining energy).

133. izen says:

Quick attempt to correct above graph…!

134. 🙂

135. izen says:

I shall take this as a warning NOT to take the common talking points and always go and find the best researched numbers, (or RTFM!).
When you do, it is usually WORSE than you thought….

136. Chubbs says:

dpy

The recent Lewandosky et. al. paper shows that with an apples-to-apples comparison model and observation trends are the same. The monthly SST-blended data is available for each model (and analysis code) at this link. So anyone can make their own comparison. As I have said before, I can confirm based on a much cruder analysis.

https://github.com/StephanLewandowsky/ERL-Lewandowsky-et-al.-2018

137. izen says:

When I make a bit of visual rhetoric I usually try and avoid exaggerating the size/impact of AGW.
I have, on some occasions in some forums, intentionally made an error that under-estimates the size of the effect. When that is corrected it has helped emphasise the point. 🙂
In this case however the error was unintentional.

138. Steven Mosher says:

“The recent Lewandosky et. al. paper shows that with an apples-to-apples comparison model and observation trends are the same. ”

It would be very cool ( someone ask gavin?) if the modelling community had a required output.

1. Output of grids with SAT 9over land) and SST in the ocean.

139. Ben McMillan says:

The actual article:
Stephan Lewandowsky et al 2018 Environ. Res. Lett. 13 123007
http://iopscience.iop.org/article/10.1088/1748-9326/aaf372

140. JCH says:

When DPY mentions a scientist named Zhou, he means a scientist, probably Chinese, whose name might begin with a Z.

Steig 2009 or O’Donnell (Nic Lewis) 2010?

141. izen says:

The original data is here –
89009 Amundsen_Scott 90S 0E 2835m

Is this an automated weather station actually ON the pole ?
Dosen’t show much of the 1998 El Nino, and it looks like AGW is finally overtaking the ozone induced SAM cooling

142. Joshua says:

Of topic, but I think it might be of interest to many here:

143. dpy6629 says:

Steven, Of course we should compare the same quantity unless there is a reason to not do so. Your long winded response is not about my point. Nic Lewis had a detailed blog post on this in 2016 claiming that the evidence that TAS and SST had a different trend was weak to nonexistent. Nic also discusses the coverage of HADCRUT issue there.

Dikran, as to baselining, my point is correct. You can keep the data within the model spread by a long baselining period. Your ENSO point seems a red herring to me. The models would presumably have totally different Enso and thus cover the range at any given time so to speak. One could thus choose a short baselining period where ENSO was essentially neutral.

144. dpy,

Nic Lewis had a detailed blog post on this in 2016 claiming that the evidence that TAS and SST had a different trend was weak to nonexistent.

Of course he did.

You can keep the data within the model spread by a long baselining period.

You’ll have to explain this more, because it doesn’t seem correct. You, of course, need a reasonably long baseline to reduce the possibility of out-of-phase internally-driven variability influencing the comparison.

145. Marco says:

Ah, the one single citation is some 2016 blog post of Nic Lewis, which then, of course, makes the whole topic “controversial”.

146. dikranmarsupial says:

“You can keep the data within the model spread by a long baselining period.”

No, that is obviously incorrect, all baselining does is add a constant to each of the timeseries. The longer you make the baseline period, the less the constant depends on the noise. Consider observations and model runs with no noise, just a linear trend, the baseline period won’t make any difference to how soon a divergence becomes visible.

Your ENSO point seems a red herring to me. The models would presumably have totally different Enso and thus cover the range at any given time so to speak. One could thus choose a short baselining period where ENSO was essentially neutral.

That is the problem, you would be creating an artificial divergence between a model that is in an el-nino or la-nina phase and the ENSO neutral observations. The whole point of a long baseline is to minimise any bias by averaging over several ENSO cycles both in the observations and the model runs,

Perhaps you should ask climatologists why they use long baselines, rather than assume they are being dishonest. It is possible they understand the issues better than you do (FWIW IIRC the WMO explain why they recommend 30 years).

147. verytallguy says:

Ah, the one single citation is some 2016 blog post of Nic Lewis, which then, of course, makes the whole topic “controversial”.

Well, it doesn’t actually run to the extent of a “citation” – we don’t even get to find out which blog it’s on(!), merely that it is the Truth.

dpy, for the nth time, please provide a reference which enables your source to be identified. A link is best of all and surely not beyond your technical capabilities!

148. Dave_Geologist says:

the evidence that TAS and SST had a different trend was weak to nonexistent

That wouldn’t be a shell game dpy, would it? A bait-and-switch?

Consider dataset A: 10±3. And dataset B: 15±3. No significant difference between them so we can choose either. And model C: 16±2. Now suppose that model C and dataset B are measuring the same thing (merged data, with the same geographical coverage), but A is measuring something different. I can choose to compare B and C and I have no significant difference, i.e. a model-to-data match. Or I can choose to compare A and C and bingo!, I have a significant difference, i.e. a model-to-data mismatch.

See what I did there?

149. Willard says:

> Quick attempt to correct above graph…!

Not enough data density in the second part to my taste.

150. izen says:

@-W
“Not enough data density in the second part to my taste.”

If I had done it right, to the same scale, it would have been 2.3% of the above graph, a line about 4 pixels thick at the far right. Even lower data density, but it rather makes the point about the relative size of the energy fluxes involved !

Personally I don’t like the circular areas used for relative size in ATTP’s graph. The human vision is bad at assessing area of round shapes, and tends to judge on relative diameter/circumference.

151. Everett F Sargent says:

My current best guess for the Zhou citation …

“Impact of decadal cloud variations on the Earth’s energy budget”
Chen Zhou, Mark D. Zelinka & Stephen A. Klein (2016)
https://www.nature.com/articles/ngeo2828
(paywalled, but you should know how to get it)

Click to access 820541.pdf

(preprint of the final ‘know how to get it’ version above)

This paper has a rather interesting citation history (so far 52 citations) …

Three of which are …

AE Dessler and PM Forster (2018)
“An estimate of equilibrium climate sensitivity from interannual variability”
N Lewis and J Curry (2018)
“The impact of recent forcing and ocean heat uptake data on estimates of climate sensitivity”
Craig Loehle (2018)
“The epistemological status of general circulation models”

I consider the last two to be outliers (the last one must be a really bad joke) relative to the other 49 citations (e. g. normal mainstream climate science papers).

152. Everett F Sargent says:

Oops, the Zhou et. al. (2016) Abstract reads …

Feedbacks of clouds on climate change strongly influence the magnitude of global warming1–3. Cloud feedbacks, in turn, depend on the spatial patterns of surface warming4–9, which vary on decadal timescales. Therefore, the magnitude of the decadal cloud feedback could deviate from the long-term cloud feedback4. Here we present climate model simulations to show that the global mean cloud feedback in response to decadal temperature fluctuations varies dramatically due to time variations in the spatial pattern of sea surface temperature. We find that cloud anomalies associated with these patterns significantly modify the Earth’s energy budget. Specifically, the decadal cloud feedback between the 1980s and 2000s is substantially more negative than the long-term cloud feedback. This is a result of cooling in tropical regions where air descends, relative to warming in tropical ascent regions, which strengthens low-level atmospheric stability. Under these conditions, low-level cloud cover and its reflection of solar radiation increase, despite an increase in global mean surface temperature. These results suggest that sea surface temperature pattern induced low cloud anomalies could have contributed to the period of reduced warming between 1998 and 2013, and offer a physical explanation of why climate sensitivities estimated from recently observed trends are probably biased low4.

Should bold that last sentence. The pattern effect of SST’s. If this is the ‘so called’ Zhou citation then I don’t think it supports what someone here has suggested (and probably the reason for not responding to three direct citation requests). Not a winner. Sad. 😦

153. dpy6629 says:

I can’t seem to get a pdf file to show up here in the comment box. However, its pretty simple. Take 2 lines of lower and higher slope. If you baseline them at the beginning, they diverge increasingly. If you baseline over the entire record, they cross at the middle of the record and the “difference” at the end is half of the preceding case. And the lower slope curve is higher than the higher slope curve at the beginning, just as Dikran observes for the IPCC’s plot. If you baseline them using the first half of the record, the lines cross at 25% through the record and at the end the “divergence” is in-between the first 2 cases.

Further, if internal variability is random in climate models as it seems to be, then the ensemble medium and spread will be pretty constant over time and independent of actual climate state. It appears to me that’s the case for the CMIP5 runs to first order anyway. Volcanoes do have an effect obviously. But if you select a baselining period free of major volcanic eruptions and neutral ENSO, then even if the period is short, there is no bias from ENSO internal variability. That is an argument for a short baselining period near the beginning of the record.

154. dpy6629 says:

There is another Zhou paper here:

https://journals.ametsoc.org/doi/10.1175/JCLI-D-15-0191.1

to which I was referring. I’m surprised this paper hasn’t gotten more publicity as it seems important to me with regard to model tuning.

http://www.realclimate.org/index.php/archives/2005/07/climate-sensitivity-and-aerosol-forcings/

Is also interesting if one takes into account recent estimates of aerosol forcings.

155. dpy6629 says:

To those with a career in model development and validation, this paper is pretty disturbing because it implies structural instability.

https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2017MS001067

Once again I’m surprised this hasn’t gotten more press. Perhaps its too technical.

156. JCH says:

That is Ming Zhao, not Zhou.

157. Everett F Sargent says:

JCH,

What’s in a name? That which we call a Zhou
By any other word would smell as sweet

But hey, the easiest way to find and list these ‘so called’ confounding peer reviewed papers is to go through certain contrarian authors list of references.from their own peer reviewed publications. JC even asks her predominately contrarian readership for references. Smear OHC as wide as possible, heck smear (messily blur the outline of (something such as writing or paint); smudge) anything climate science related as wide as possible, for that matter.

Why put blinders on when you can get others to do that for you?

158. JCH says:

Everett – Zhou 2016 is my favorite paper as it is pretty much what I had already concluded was going on during Da Paws.

I think maybe I made a recent comment here about Zhou that was deleted.

I rest my case.

Well, not yet. This is the second time I have pointed out the name of scientist who wrote the paper in question is Ming Zhao, and not some guy named Zhou.

Anyway, I do not think Ming Zhao, the author, shares DPY’s slants/conclusions on the paper. It was cited in a paper that was wildly praised at Climate Etc. as they took it to have proven Lindzen was right about, yippee, his Iris hypothesis: pretty darned game changing.

159. dikranmarsupial says:

DPY wrote

“I can’t seem to get a pdf file to show up here in the comment box. However, its pretty simple. Take 2 lines of lower and higher slope. If you baseline them at the beginning, they diverge increasingly.

No. They diverge from the center of the baseline period. Thus if you have a long baseline period and a models which have a high slope and the observations have a low slope, then at the end of the baseline period the observations will lie below the spread of the models.

If you baseline over the entire record, they cross at the middle of the record and the “difference” at the end is half of the preceding case.

NOBODY who know what they are doing baselines to a single year’s data, NOBODY.

And the lower slope curve is higher than the higher slope curve at the beginning,

That is true, except that 1998 is not at the beginning of the baseline period, so it is irrelevant. The WMO normal period is 1961-1990, so 1990s are after the usual baseline and hence affected by the OPPOSITE bias to that which you suggest.

Further, if internal variability is random in climate models as it seems to be,

Bullshit. The models exhibit ENSO like behaviour. The variability is chaotic, which is not the same thing as random.

then the ensemble medium and spread will be pretty constant over time and independent of actual climate state.

Citation required. This assumes amongst other things that ENSO is not affected by increasing temperatures.

It appears to me that’s the case for the CMIP5 runs to first order anyway. Volcanoes do have an effect obviously. But if you select a baselining period free of major volcanic eruptions and neutral ENSO, then even if the period is short, there is no bias from ENSO internal variability. That is an argument for a short baselining period near the beginning of the record.

I’ve already explained why the neutral ENSO baseline is biased and you have not engaged with that explanation in any way. So we have mathematical errors, unsupported assertions and evasion of previous criticisms. Not particularly convinced by that sort of thing.

160. Dave_Geologist says:

Zhou/Zhao is more of what I was alluding to above. Real science is all-joined-up. Fake science quotes papers which show a pattern effect exists because it rings a bell about the Iris Effect, which they still think it is real even though observations debunked it as a major factor within a few years of it being proposed. Then champions an ECS paper that’s bunkum if a pattern effect exists, whose author forcefully denies without presenting evidence that peer-reviewed papers demonstrating a pattern effect invalidate his work (but only on blogs, of course). Blithely unaware of the implicit, no explicit, contradiction.

Because of course it never was about proposing a coherent alternative to mainstream climate science. It was always about presenting a mish-mash of sound-bites which sound vaguely convincing if you don’t know the science and live only in the present, so are unfazed by the mutual contradictions between the various contrarian assertions. Known of course in the creation-evolution “debate” as a Gish Gallop.

161. Dave_Geologist says:

Once again I’m surprised this hasn’t gotten more press. Perhaps its too technical.

Perhaps you’ve misunderstood it dpy. As someone who’s had with a career in making billion-dollar bets on the basis of numerical models, who’s been involved in model development and written my own admittedly-simple models, and who’s seen models validated at the drill-bit and the wellhead, this paper is not disturbing at all because I see nowhere that it implies structural instability. Although you probably have to define what you mean by that. Parameterisation in a sense removes instability because some of the unstable (chaotic, if you like) processes occur at a length and time scale smaller than the final model, so cannot be explicitly represented in it. That is both a feature (goodie, faster, more replicable model runs) and a bug (have I properly upscaled the bulk behaviour?). George Box told us a long time ago how to handle that trade-off

Parameterisation (upscaling as we call it in the O&G industry) is non-commutative. Meh, who’da thunk it? Only the first person who built a model, probably before computers were invented. And every modeller since I don’t doubt. I rather think the modellers have got that covered. You just don’t realise it, or don;’t want to realise it.

Just as well you don’t have to do that sort of thing when you’re designing aircraft, tall buildings, bridges and nuclear power stations. What, you do? Omigod, we’re all doomed!

162. Chubbs says:

dpy

The obvious conclusion is that the data we have, used in model parameter development and tuning, is completely consistent with consensus TCR and ECS estimates. A bit ironic that the papers you cite, with very nice results mind you, would be impossible without climate models and climate modelers that were willing to kick the tires.

163. Dave_Geologist says:

Everett: “The epistemological status of general circulation models”. Double buzzword bingo: “wicked problem” in the Conclusions, as well as “epistemological”. Although actually there are no conclusions, he appears to be Just Asking Questions. The acknowledgements feature a proper Rogue’s Gallery of Usual Suspects.

164. dpy6629 says:

Dikran, What you said about baselining agrees with what I said, at least so far as I can tell. The shift is dependent on the baselining period used. Thus longer baselining period makes the difference at the end smaller. It’s a very simple point. Your error in understanding what I said does not constitute an error on my part.

I read your ENSO point and it seems to me to be incorrect. The point is not about a single model but about the ensemble of models. Their ENSO at any given time is not synchronized and should be essentially random. How else can you explain that the confidence interval is essentially a constant height centered on the mean? The same thing would apply to other “internal variability” noise. This is what all energy balance methods in terms of matching average AMO state, volcanic activity, etc. between beginning and final state.

165. dpy6629 says:

Dave, You have this wrong. In models that are advanced and reliable, the sub grid parameterizations are solved simultaneously with the Navier-Stokes equations. In this case there is a much better chance of getting a unique answer. With a climate model that’s very costly and perhaps infeasible. However, commutativity of the sub grid models is highly desirable because there is no obvious way to choose the order of applying them. Thus the range of answers from different orderings implies a large uncertainty.

166. dpy6629 says:

Chubbs, As I said to Dave, the papers offer negative results which are good and point to the fact that climate models suffer from deficiencies that for example CFD models do not generally (even though CFD models have their own problems that are more tractable). The papers strongly indicate that ECS for example can be engineered over a pretty broad range in any model by changes to sub grid parameters that are consistent with data. That indicates that uncertainty is high for these models.

167. dikranmarsupial says:

” The shift is dependent on the baselining period used. Thus longer baselining period makes the difference at the end smaller. ”

It is the sign of the difference that is the issue, it has the opposite bias to that needed to explain why the observations are at the top of the model spread in the decade following the baseline period.

” Their ENSO at any given time is not synchronized and should be essentially random. ”

before you explain why you know more about this than the IPCC, the WMO and the climatologists, whose methods I am trying to explain to you, perhaps you should test your assumptions rather than just make them?

168. dikranmarsupial says:

I don’t think DPYs conclusion can possibly be true in the case without noise, a model with a slope of 0.2 degrees per decade will diverge from observations increasing at 0.1 degrees per decade by 0.1 degrees per decade. This is true whether you have a long baseline or a short baseline. If you put the short baseline and the long baseline are centered on the same year, they will both give *identical* results in the noise free case.

169. Willard says:

> ECS for example can be engineered

Definitions of chaos can be engineered too, e.g.:

And since I’m here, I will point at this:

[DaveY1] I don’t think Ron is invoking chaos to explain anything.

and this:

[DaveY2] Nonlinear feedbacks must be playing a significant role here I suspect.

That is all.

https://judithcurry.com/2019/01/14/ocean-heat-content-surprises/#comment-887961

170. Dave_Geologist says:

This should come as no surprise since the model is tuned to optimize accuracy in the top of the atmosphere radiation with emphasis on the shortwave fluxes. Thus, we would expect a loss in accuracy for shortwave variables when the process order is changed without a proper retuning of the model.

(My bold). IOW if they didn’t tune for TOA flux after the process order was changed, the uncertainty could be high for those models. Fortunately they do tune for TOA flux, which gets round that potential problem. Which is not to say, of course, that further optimisation is not possible and should not be pursued.

171. Everett F Sargent says:

“Thus the range of answers from different orderings implies a large uncertainty.”

The 1st thing that popped into my head was time step. In essence, a small enough time step reduces the sub-grid parameterizations to quasi-linear and thus ever closer to commutative behaviors. That one is pretty much called a no-brainer.

See … 4.2. Time–Step Convergence
And yes, I see their last sentence (in that section) … “Until then, the impact of process ordering in sequentially split climate models remains an important concern”

And from their conclusions section …

“Because convection is one of the more uncertain and crudely handled parameterizations, we conclude that it is important to position macro and microphysics after shallow convection but before radiation. This finding motivates us to revisit the question of whether there is any best-practices guidance for how to order sequentially split processes. We conclude that best-practices process ordering satisfies the following rules:

1. Macrophysics should directly precede microphysics because it produces the condensate that microphysics acts on—they are two parts of a single process.
2. Radiation should see the most accurate cloud state possible. This is probably the state after microphysics or dynamics. Radiation should not act on the state after convection.
3. Turbulence should be ordered as close to the surface schemes as possible since it is the mechanism by which surface fluxes communicate with the atmosphere.
4. It is unclear whether deep or shallow convection should be called first, but this choice makes surprisingly little difference in our simulations (not shown). It makes intuitive sense to keep deep and shallow convection together unless there is compelling reason to do otherwise.

Of the 24 orderings considered here, one quarter (the default ordering and five others) satisfy these best-practices criteria. The fact that orderings not following our best-practices have very different climates illustrates the need to carefully consider process coupling. In this light, it would be desirable to develop even tighter guidance on what constitutes the best process order. What is the optimal placement for dynamics? Where should turbulence and surface coupling go? E3SM code structure prevented us from separately exploring these after-surface-coupling processes in the current study; doing so in future work would be useful.”

It looks as if their default ordering made sense, almost as if they had already considered the default ordering, in at least a qualitative fashion. And I really do appreciate their constructive criticisms.

My own conclusion? Where you see only darkness I see only the light.

172. Chubbs says:

dpy

After reading your comment, I picture a climate modeler with a big dial on her/his desk set to 3 ECS. The dial can be left at 3 for decades and everything falls neatly into place. Even a pause knows when to end. What a gig. Where can I sign up.

173. Dave_Geologist says:

It looks as if their default ordering made sense, almost as if they had already considered the default ordering, in at least a qualitative fashion.

Because, rather appropriately given our host’s ‘nym, it makes physical sense. It’s almost as if, you know, they understand physics and have thought about it. Most impressive for a bunch of innumerate geographers. Bet they’ve never even heard of CFD, even if they knew what the acronym stood for 😉 .

174. Everett F Sargent says:

I must admit that some of this talk of model uncertainty, (seemingly) duncified numerical climate modelers, (seemingly nefarious) model tuning and surprise over a lack of adequate press coverage of publications that (appear to) reinforce one’s own biases (or some such), smacks of borderline conspiracy theorizing.

175. dpy6629 says:

This issue of sub grid models is actually very simple. A proper sub grid process model should be independent of numerical details like grid spacing and time step for the Navier-Stokes equations. The reason is obvious. If the sub grid model depends on numerical details then any “tuning” done will have to be redone anytime any numerical changes are made. That’s unacceptable as grid spacing will surely be changing in the future. And that’s just one example. There are literally hundreds of these numerical choices.

The only alternative that has any hope of consistency and verification is solving a differential algebraic system of equations that are coupled. This requires simultaneous solution or convergence to such a simultaneous solution. It was a big step forward in turbulence modeling when this was realized. It actually gave meaning to the idea of grid convergence and scientific validation of models.

When several sub grid models are involved, they should be solved simultaneously or by iteration to such a solution. Any formulation that doesn’t do this is little better than fudge factors dressed up as real mathematics or science and any skill is the result of cancellation of errors. The history of turbulence modeling shows this clearly. Accuracy improved dramatically when algebraic models were replaced with PDE models for eddy viscosity. Accuracy and repeatability became much better when simultaneous solution was adopted.

176. dpy6629 says:

Dikran, This is very simple. My only assertion is that with a fixed confidence interval, baselining by the whole interval for example can keep the lower slope curve within this interval (higher in the first half and lower in the second half than the mean curve). Baselining at the beginning will make that much less likely as the last point is a lot more likely to lie outside the confidence bounds.

You still haven’t responded to the point that “internal variability” in a large ensemble will be chaotic, different in each model, and therefore pretty random at any given sampling time. The spaghetti plots from the IPCC show this I think. But it is also obvious from any ensemble of turbulent CFD simulations. In this case of a large ensemble the idea of long baselining periods seems to be unnecessary. You must however, choose a baseline point of interval for the actual data that has essentially average internal variability properties.

177. dpy,
I don’t think you’re explaining what baselining does. Baselining is simply removing some constant from the entire timeseries. The question is, what constant should be removed? If you baseline relative to a single datapoint, then the outcome can depend strongly on variability. Hence, what is typically done is to define the baseline as the average over some time period that is long enough to largely remove the effects of variability. If you do that, then the outcome (when you compare to another timeseries baselined in the same way) should be largely unaffected by variability.

178. dikranmarsupial says:

DPY you are ignoring the fact that I showed you that in the noiseless case, the long and short baseline periods give exactly the same result. Any difference between the two therefore depends on the noise. Think about that for a moment.

The purpose of baselining is because there is a constant offset between model runs (and thus with the observations). The purpose of baselining is to estimate that offset. Long baselines are better than short ones because they average out the noise more than short ones. You idea of using a neutral ENSO baseline period doesn’t estimate that offset because while the period may be neutral for ENSO in the observations, it won’t be for the model runs. It therefore artificially inflate any difference.

You still haven’t responded to the point that “internal variability” in a large ensemble will be chaotic, different in each model,

Hilarious, you initially said random and *I* corrected you to point out that it is chaotic, not random, and yet you present it as *your* point and say I haven’t responded to it. Obviously I have responded to it when I pointed out it was chaotic rather than random. That really is very dishonest indeed.

In that comment I wrote

I’ve already explained why the neutral ENSO baseline is biased and you have not engaged with that explanation in any way. So we have mathematical errors, unsupported assertions and evasion of previous criticisms. Not particularly convinced by that sort of thing.

If you look upthread, you can find that explanation here. I’ve explained it AGAIN at the start of this comment.

179. dikranmarsupial says:

Thanks ATTP (or Willard).

Just to be clear. Let us assume that the observations are composed of a forced component f_o(t) and an unforced component u_o(t). Each model run has a forced component f_m(t), which is (for the sake of argument) identical for each model run, but each run has a different realisation of the chaotic unforced variability u_i(t), for the i^th model. Now unfortunately models have a bit of difficulty estimating the absolute temperature of the planet (but are much better at changes in response to forcing), so we can say that each model has a constant offset. So for the observations, we have

T_o(t) = f_o(t) + u_o(t)

and for the i th model run we have

T_i(t) = f_m(t) + u_i(t) + e_i

The purpose of baselining is to remove the effects of e_i so that we can compare the estimates of the change in temperature in response to the forcing in a fair, unbiased manner (i.e. so that any differences due to e_i are eliminated, or at least atentuated).

One way of doing that is to use anomalies over some baseline period. If f_m(t) is the (essentialy) the same for all model runs, then it will make the same contribution to the average value over the baseline period for all models. The anomaly therefore mostly depends on u_i(t), which is chaotic and different for each model. However if we use a long baseline, we will be averaging out the effects of this chaotic component, giving us a better estimate of e_i.

Now of course, if we use anomalies for the models, we need to use anomalies for the observations as well. If we use a short period where ENSO is neutral in the observations, that is fine if you just want to look at the observations, but it is not fine if you want to compare them with the models because it will mean that the estimate of e_i will be unreliable, so the comparison of the predicted climate change will have an artificially high variance because it includes some dependency of the e_i offsets which are not well estimated.

The take home message is that we don’t choose the baseline to fiddle the plot to look the way we do. There is a specific mathematical aim in baselining, we are estimating something.

Appendix: ;o)

It is worth noting that the ensemble mean is an average of the model runs. If the distribution of the chaotic unforced componets is reasonably symmetric, then for a large enough ensemble, they will effectively cancel out, so the ensemble mean is an estimate of f_m(t), i.e. just the forced response of the climate system. This is an important point for those that think the models predict a smoothly increasing response to GHG. They don’t, the ensemble mean is just a prediction of the forced response, but we know even if f_m(t) = f_o(t), the observations have a realisation of the unforced variability superimosed on top, so we KNOW a-priori that it is unlikely to be smoothly increasing, because the unforced response in the models isn’t smooth either.

180. Everett F Sargent says:

“This issue of sub grid models is actually very simple. … ”

Which has absolutely nothing to do with the one paper you cited. Citations required for your hand waving in the direct context of numerical climate science modeling peer reviewed papers. TIA

More from their conclusion section …

“The work shown here has demonstrated that changing process order can have a very large impact on model behavior. Comparison of model solutions against observed data for a set of 24 simulations with different process orderings has shown that the impact of process order can be dramatic. A near doubling of RMSE is found in some variables and a 20% improvement in RMSE is found for others. Among the top performing models in terms of RMSE, four of the top five are the process orderings with only the deep convection moved. This is unsurprising considering that deep convection played the weakest role in determining the differences between process reorderings and because the default ordering has undergone years of optimization. Interestingly, only one of the top five performing models in terms of global mean error fit into this category and the other four are those where surface/turbulence/dynamics follow shallow convection and macro/microphysics but precede radiation. The top performing sequence in terms of global mean, SC-MM-Oth-DC-Ra, is also among the top five performing runs in terms of RMSE. This shows the importance of a
periodic assessment of the impact of changing process ordering. In this case, it is possible that the dynamics solver used by an earlier version of the model did not provide a very realistic state to physics, but recent advances have changed that situation. When model improvements are made, a second look at coupling order may be beneficial. Optimizing process order in conjunction with proper tuning is likely to yield a more skillful model. Exploration of this joint optimization problem is important future work.”

Bold the “the default ordering has undergone years of optimization” part. Again good constructive criticism is of course most welcome from the two authors of this paper.

Those who demand that the Earth be modeled exactly as an aeroplane (or some such) should stick to modeling aeroplanes. Someone who claims SME status in one rather narrow discipline or sliver of knowledge should not immediately assume SME status in another rather narrow discipline or sliver of knowledge (particularly if those respective domains vary by more then O(10^12)). 😦

181. Dave_Geologist says:

Seconded Everett.

dpy, no amount of chest-beating about CFD or Navier-Stokes is going to persuade anyone that you have a relevant point to make if you continue to demonstrate your ignorance of GCMs by linking to papers which don’t say what you claim they say, or mean what you claim they mean. My own experience of reservoir simulation models (which in many ways are better analogues for GCMs than aircraft aerodynamic models) enables me to see that your wrong. A genuine GCM expert would probably say that you’re Not Even Wrong. Claiming expertise in an unrelated field, while demonstrating a profound lack of expertise in the field under discussion, butters no parsnips. It tells the audience that you either lack expertise in the relevant field, or lack the objectivity to apply what expertise you have. And it rather quickly gets old. Not as quickly as electrical engineers trying to argue by analogy from feedback circuits, but quickly nonetheless.And the smug superiority doesn’t help. “Perhaps its too technical” broke my irony meter, given that it came in the context of a paper which you totally failed to understand.

182. Chubbs says:

Below is an apples-to-apples CMIP5/Temp comparison through 2018. Perhaps the signal is so strong that large amounts of model noise can be tolerated.

183. dikranmarsupial says:

Oh no, the observations have gone outside the spread of the model runs TWICE! They must be getting the climate sensitivity wrong! ;o)

184. Chubbs says:

To expand on the point I made above. Adding CO2 changes the climate equilibrium. The issues dpy raises add random error. These errors are going to increase noise making it harder to detect the CO2 signal. So I don’t think the errors support dpy’s overall climate view. More likely the opposite, since the CO2 signal comes through loud and clear, and the emerging constraint studies indicate that better models have higher ECS.

185. dpy6629 says:

Well aside from the ad hominums and vague pseudo-technical word salads from Everett and Dave, the issue here is the same for all sub grid models including turbulence models (used in weather modeling). Rigorous validation and tuning can only be done if the sub grid models are independent of the numerical details such as time step choice and the order of application of the models. It has nothing to do with the application of the CFD code. And weather models are CFD codes (with a lot of sub grid models) as Nick Stokes keeps saying over and over. He is right about this.

sequential application of sub grid models might be justified if the truncation errors were quite small (very small time step and grid spacing) as the authors of the paper point out. In climate models truncation errors are quite large. Simultaneous solution of models is always preferable as the experience in turbulence modeling shows clearly.

186. dpy,
Please explain what you’re implying? Nick Stokes is right about many things, but I don’t think he’s implying that any of these issues suggests that we should strongly doubt the basic overall results we’re getting from GCMs (at least on scales where we’re confident that we’re resolving what is going on).

187. I’ll add that GCMs are really being used to investigate how the system responds to changes, not to try and make strong claims about what the precise state will be at some point in the future.

188. dpy6629 says:

I’m going to give up on the baselining point. Without a picture, I don’t think you understand the point I am making. How do you get a .pdf file to appear in a comment? Does anyone know?

189. If you have a link to a PDF, or to an image, you can simply put the url on a line by itself. If you have a PDF on your own computer, then you have to (I think) find a way to upload it somwhere so that you can then post it here.

190. dpy6629 says:

ATTP, Nick has more confidence in climate modeling than I do. We’ve discussed it in detail at his site. Of course climate models do get some things right. Rossby waves are probably reasonably well modeled. Pole to equator temperature gradient is well modeled because of that.

The question is the level of uncertainty. It’s probably much larger than implied by the IPCC spaghetti plots as the two cited papers show.

The fundamental problem is just that the truncation errors are much larger than the changes in energy flows we care about. That usually implies that skill in the tuned for outputs is achieved by cancellation of errors. This is sometimes called “right for the wrong reasons” in the trade. The problem is that skill on other unrelated output measures is not to be expected in this situation.

191. dpy,
Okay, the spaghetti plots are almost certainly not a representation of the true uncertainty, but let’s say they underestimate it a lot. Let’s also consider climate sensitivity. Well, could it be much lower than they indicate? Possibly, but it seems unlikely given how much we’ve already warmed. So, if you think the spaghetti plots heavily underestimate the uncertainty in GCMs, then you seem to be implying that climate sensitivity is much more likely to be higher than the likely range suggests, rather than lower.

192. dpy6629 says:

Nick makes a correct point that Rossby waves are much easier to model than a wing because the pressure gradients are generally quite mild. Unfortunately climate involves things that are actually much harder than an aircraft wing such as convective storms. Then there is the planetary boundary layer that is dramatically under resolved and free air turbulence that is completely unmodeled.

Global mean temperature anomaly is probably consciously or unconsciously tuned for by most modeling groups. TOA radiative balance is explicitly tune for. If that is correct and ocean heat uptake is roughly right (as it seems to be), then mean temperature will also be essentially right. That’s assuming conservation of energy which is apparently not strictly true in some models.

193. dpy6629 says:

ATTP, I don’t really think one can conclude much about ECS from climate models since ECS can be engineered over a broad range using parameter values consistent with the sparse or in some cases nonexistent data. That leaves one with other lines of evidence. I actually thought the real climate post (which I think I linked to above) on aerosol forcing and ECS was an early and pretty good pre-echo of some of Lewis’ work.

194. dpy,
Except it really can’t be much lower than the lower bound suggested by GCMs. We’ve already warmed by ~1K and have not yet doubled atmospheric CO2. Hence, if you think GCMs are telling us little about climate sensitivity, you’re suggesting that it could be much higher than the top end of the range suggested by GCMs.

195. verytallguy says:

dpy,

GCM ECS covers the range calculated by other methods.

In a world where GCMs didn’t exist, our range of ECS would not be significantly different.

196. Everett F Sargent says:

Irony meter explodes. 😦

The perfect is the enemy of the good. Nirvana fallacy (perfect solution fallacy).
https://en.wikipedia.org/wiki/Nirvana_fallacy

It is my own opinion that someone here does not understand ESM’s very well. Likely due to lack of proper training in that specific field.

197. Everett F Sargent says:

Argument from authority
https://en.wikipedia.org/wiki/Argument_from_authority
“It is well known as a fallacy, though it is used in a cogent form when all sides of a discussion agree on the reliability of the authority in the given context.”

Somehow, don’t ask me how, but somehow, I don’t think we all will find agreement on the reliability of someone here (or elsewhere, e. g. NL) in the current context.

That this has happened many times before, arguing for perfection in ESM’s, is truly a broken (world) record. The same fallacious argument over and over again, ad infinitum, ad nauseam and ad absurdum.

We are now way off course from the main thrust of this thread.

198. Joshua says:

Mostly off topic, but I would imagine of some interest:

199. JCH says:

ATTP, I don’t really think one can conclude much about ECS from climate models since ECS can be engineered over a broad range using parameter values consistent with the sparse or in some cases nonexistent data. …

Where does Ming Zhao indicate any support of the above at any level?

200. Chubbs says:

I would second Everett’s opinion. Why are we arguing about climate models? Its a skeptics shell-game. Climate models can be debated endlessly in purely qualitative terms. The talking points never change. Ask for evidence. None is provided. Strip away the jargon and dpy is giving us the same argument any skeptic uses. The world is too complex to ever be modeled properly. So lets give up and not even try. This despite a track record of prediction success and a large body of literature.

dpy – “I actually thought the real climate post (which I think I linked to above) on aerosol forcing and ECS was an early and pretty good pre-echo of some of Lewis’ work.”

EBM are the one and only one true answer. Our world is so complex and chaotic, that we can’t use a complex model with a long track record; no, we must use a simple model with a pre-1950 baseline so aerosols and limited data fog the result. Makes sense doesn’t it?

201. dpy6629 says:

Chubbs’ graphic illustrates the way pictures can be manipulated using different baselining periods (just as I said earlier here) and also how selection can be used to prove a point. Here is Hausfather’s plot using a baselining period of 1981-2010.

Here’s one Clive Best did using a baselining period of 1961-1990 (same as the Hadley center uses)

You will note that the earlier baselining period causes the observations to get significantly lower wrt the models.

202. dpy6629 says:

OOPs, wrong link. Best’s plot is here

203. dpy6629 says:

Chubbs, This is not about talking points except perhaps for you. There is real science and mathematics behind CFD modeling and real well known issues with all CFD models.

204. Chubbs says:

DPY

Best is using air temperature over the water, while Hausfather is using SST. So Hausfather is making an apples-to-apples comparison and Best is not. The trend difference between SST and temperature over the ocean is roughly 20%. That is the main reason the charts look different.

205. dpy,
And the difference is probably to do with Zeke using SSTs and SATs (i.e., like-for-like) and Clive simply using SATs. It’s not a baselining issue; it’s to do with not quite comparing the same things.

206. Sorry, I see Chubbs has made the same point that I did.

207. dikranmarsupial says:

DPY wrote

I’m going to give up on the baselining point. Without a picture, I don’t think you understand the point I am making. How do you get a .pdf file to appear in a comment? Does anyone know?

You don’t need a picture to understand the purpose of baselining, the basics of which I set out. If you are not able to acknowledge that the purpose of baselining isn’t to make a point in a diagram, but to estimate the offset of a model run relative to the observations, then there is no point discussing this with you.

As I said, the reason that a small baseline leads to a wider apparent spread is because it isn’t fully compensating for the meaningless offset term (and thus is over-fitting the noise in estimating the offset).

208. dikranmarsupial says:

DPY wrote “ATTP, I don’t really think one can conclude much about ECS from climate models since ECS can be engineered over a broad range using parameter values consistent with the sparse or in some cases nonexistent data”

So why has no climate skeptic taken the publicly available source code for a GCM and demonstrated that it can be tuned to have an ECS less than (say) 1C, whilst using physically plausible parameter values (given the observations)?

I suspect it is because (i) it isn’t true and (ii) the skeptics are not skeptical of their own position and can’t be bothered to put in the work to test their assertions.

209. dikranmarsupial says:

DPY LOL you do know that Best is showing model runs for three different scenarios, which is obviously going to expand the spread!

Rather than assume the differences are “manipulation”, it would be better to understand the purpose of baselining and the reasons why the two pictures are different and then work out which is the better representation (whilst trying to ignore what you want the diagram to say).

210. verytallguy says:

For dpy

Cowtan, 2015:

https://www.researchgate.net/publication/280571227_Robust_comparison_of_climate_models_with_observations_using_blended_land_air_and_ocean_sea_surface_temperatures

Over the period 1975–2014 the use of blended rather than air temperatures accounts for 38% of the difference intrend between the models and the observations

Some suggestions for your future contributions:

1) Yet again you don’t provide a link to the analysis Clive Best actually did, although this time you did let us know his name. Please provide a link or sufficient citation so we actually know what your claims are based on. I’ve quite lost count of the number of times you’ve been asked to do this. Without a cite, we are back to argument by assertion.

2) Rather than rely on blogs like Clive’s, try peer reviewed literature, or failing that, work by experts in the field. Hausfather was a co-author on the Cowtan paper and knows what he’s doing.

3 Constrain your comments to technical matters in future rather than speculating on “manipulation” as you have now done on more than one occasion on this thread.

211. Dave_Geologist says:

Just as well you’d already broken my irony meter with “Perhaps its too technical”, dpy.

What with “ad hominums (sic) and vague pseudo-technical word salads” (OK the argument-from-authority is only a cousin of ad-hom, and it’s a rather boring salad that has only three ingredients, CFD, Naviers-Stokes and sub-grid, for use with all meals).

212. Dave_Geologist says:

Global mean temperature anomaly is probably consciously or unconsciously tuned for by most modeling groups. TOA radiative balance is explicitly tune for. If that is correct and ocean heat uptake is roughly right (as it seems to be), then mean temperature will also be essentiallyright.

Really, dpy, really. You can tune for top-of-atmosphere radiative imbalance and, despite those tens of km of messy, turbulent, inadequately modelled atmosphere in between, you somehow get the right answer for surface temperature? Really? Really? Only in a world where you always pick the winning lottery ticket. If that’s where you live, I have a bridge in New York to sell you. Back in the real world, of course, that’s a demonstration of tremendous model skill. They must be doing something right, eh, despite your concerns.

AFAIK ocean heat content is a model output not an input. If they’re getting oceans right, then that’s another demonstration of model skill. Perhaps a more impressive one, given that it’s one more step removed from TOA. Although since that’s where most of the heat goes, you could argue that conservation of energy means it only shows that GCMs are getting the energy balance right. Very right. Which of course means they’re far more trustworthy than short-term, surface-temperature-based EBM models like Nic Lewis’. That probably explains why those low-ball estimates fly in the face not only of GCMs, but of all other lines of evidence including observed TCS. Glad we settled that. I presume you now agree with me that the most likely ECS is greater than 3C?

BTW the allegation in your first sentence really needs to be backed up by evidence. Otherwise you risk being dismissed as a believer in paranoid conspiracy theories.

213. Dave_Geologist says:

Best is using air temperature over the water, while Hausfather is using SST. So Hausfather is making an apples-to-apples comparison and Best is not.
Hmmm. You don’t think Clive might have engaged in a spot of, well, manipulation do you? To get the answer rhetorical tool he wanted?

(Cherry-picking is a form of manipulation. As is making apples-to-oranges comparisons and not telling the audience that’s what you’ve done. It’s manipulating the information given to the audience, in order to deceive them into accepting falsehood as truth.)

214. verytallguy says:

So, on an actual topic of interest, can anyone explain why SAT and SST trends differ in the models?

I get the part about sea-ice coverage affecting this, but over open ocean, it seems very surprising to me that the trends differ (as opposed to the absolute values)?

215. vtg,
According to this post it’s because SSTs tend to warm slightly slower than SATs, even for SATs over the oceans. Not sure why this is the case, though.

216. verytallguy says:

Thanks AT. It’s the “why” I’m interested in. I struggle to see physically how a long term trend could be different for the two.

217. Chubbs says:

Below is a recent paper that covers land/ocean warming ratios, which shows the importance of using the proper metric to compare observations and model results. Also below is the last para from the summary. WR = warming ratio land to ocean

“Our results provide evidence to support the suggestion of the stabilization of the WR in recent years. We suggest that this stabilization is a consequence of: (i) emerging dominance of the WMGHG temperature response over aerosol forcing; (ii) changed aerosol distribution giving greater forcing over ocean compared to land; and (iii) progressive evolution in the size of ΔTO leading to a reduction in the noise of the WR series. The fact that WR estimates from observations and simulations are in very close agreement since the year 2000 suggests that the dynamical processes within the GCMs responsible for the land–sea warming contrast are responding realistically to well-mixed external forcing.”

http://iopscience.iop.org/article/10.1088/1748-9326/aae46f/meta

218. Chubbs says:

Here is another quote from the paper above. Some of the things dpy is most critical of based on his past experience are well-simulated.

“Interest in the warming ratio (hereafter ‘WR’) reflects the importance of the phenomenon in both societal terms (i.e. the implications of mitigating and adapting to global temperature targets over land) and also in terms of the dynamical processes responsible for the WR under both transient climate change and climate equilibrium. Existing studies provide substantial assessments of the coupled ocean–atmosphere processes which appear to be responsible for the WR showing that whilst thermal inertia differences do contribute, the dominating mechanisms are related to boundary layer properties, humidity and surface processes”

219. paulski0 says:

vtg,

My basic understanding is that it has to do with absolute SATs (at 2m) being slightly colder than SSTs and that gradient reducing with warming, due to elevated water vapour IIRC.

220. I responded to vtg on the wrong thread. Paulskio’s explanation may be more correct than mine, though.

221. Hyperactive Hydrologist says:

The Met Office CIMP6 GCM has a ECS of 5.2-5.4K and TCR of 2.6-2.9K. They also say:
There is a strong indication (not yet published) that a number of other CMIP6 models will have ECS values higher than the upper end of the CMIP5 range.

222. dikranmarsupial says:

FAO DPY6629

223. Dave_Geologist says:

Thanks dikran. maybe I was harsh on Best. Although they seem to be quite different graphs, although both from the same site. I suppose if dpy linked to the article rather than hotlinking the picture, I could find out if the text was misleading. I’m not holding my breath though.

224. dikranmarsupial says:

It is a shame that Clive deleted the tweets that contain the error that he is apologising for. I thought it would have been more instructive to have left them. It is rather unusual for people to be able to admit they were wrong, so kudos to Clive for having done so. I suspect the original article is here (DPY seems to be the only commenter).

225. Dave_Geologist says:

Thanks dikran. I still don’t get how the difference is explained by the tweet. The blended and unblended curves both intersect recent observations in a way that the dpy’s graph doesn’t. So either there are more apples and oranges, or something else has changed.

226. dikranmarsupial says:

Not quite sure, but he did also tweet this, which implies that there was some other issue:

227. verytallguy says:

Paulski0,

Thanks, I think that makes sense. Any idea what the difference in absolute temperatures is between SST and SAT?

228. Ken Fabian says:

The question is whether there is enough uncertainty to justify opposing and obstructing rapid decarbonisation – and when you start from a desire to NOT have to do it, the level of uncertainty doesn’t need to be very great.

Continuing the questioning puts off having to make a clear commitment by creatively using the “just making sure before taking precipitous action” excuse, but it takes things further from the green, towards the mauve (in graph above). Whilst the precipitous action – pumping gigatonnes of CO2 (our No.1 waste product) into the atmosphere – gets a continuing green light, despite the cumulative and irreversible nature of the climate problem.

The greatest source of uncertainty in climate projections is the extent of effective commitment to emissions reductions.

229. paulski0 says:

vtg,

You could have a look through yourself at Climate Explorer. TAS vs. TOS.

There seems to be some geographic variance, difference between models and variation over time, but it’s about 1ºC.

230. Chubbs says:

Re: TAS vs TOS. Speculating that this is a case where the global average reflects a wide range of different conditions that need to be examined in more detail. As an example I would expect the biggest TAS vs TOS differential when cold air exits continents in winter. So weaker arctic/polar air formation in winter could contribute. In the tropics expansion of the warm pool leading to a broader area of tropical convection could also contribute. On-the-other-hand small scale phenomena at the ocean/atmosphere interface could also play an important role.

The paper I linked above, which shows that models are getting the land/ocean warming ratio right increases the confidence that the TAS/TOS difference is real. Per the paper, this indicates ocean/atmosphere interface physics is well modeled. Note that this also means that models are getting matching both SST and land warming rates. Nice “tuning” job by the modelers to anticipate Cowtan et. al (2015).

231. verytallguy says:

Thanks again paulski0.

I guess if the difference is only 1C then over time the effect must become smaller (as the correction could never be >1C).

232. dikranmarsupial says:

FWIW, these diagrams show the effect of baselining using short (5 year) and reasonable (30 year) baselines. Bear in mind that we are trying to estimate a constant for each model run. Notice how the spread of the model runs wiggles about a lot as the baseline moves along the data. It shouldn’t if you have estimated the constant reliably because the constant is, … err… well… a constant! ;o) Note that the analysis is much more stable for a more reasonable window length, because the noise affects the estimate of the constant less.

233. dpy6629 says:

Clive Best confirms at his blog that the difference between SST vs. TAS over the oceans combined with land temperatures is about 0.6C (which is about what his version of Zekes plot shows). That’s about the thickness of the red line in Clive’s earlier plot which is graphically insignificant. The difference in the graphs must therefore be due to the baselining periods being different (indeed those two 30 year periods only overlap by 10 years). indeed if you look in the 1960’s it looks like the data is overall above the model mean to me.

It’s an example of my original point. If you choose different baselining periods you are essentially shifting the curve up or down. If you choose a baselining in the “middle” of the period shown, the curve will be more likely to lie within the confidence interval throughout.

I didn’t say anyone was manipulating their plot, merely that it was possible to do so. The bottom line is that these plots are not very useful except for tweeting and “communication”. The AR5 I believe chose Clive’s baselining as does the Hadley center so that is good enough for me. Don’t know why Zeke chose a much later one.

234. dpy,

It’s an example of my original point. If you choose different baselining periods you are essentially shifting the curve up or down. If you choose a baselining in the “middle” of the period shown, the curve will be more likely to lie within the confidence interval throughout.

That’s why you need to use the same baseline for all the data (model plus obs) as anyone trying to do a fair comparison will do.

235. verytallguy says:

I didn’t say anyone was manipulating their plot, merely that it was possible to do so.

I’m not saying you’re being tendentious dpy, but will merely observe that it would be possible to do so with this form of words.

236. dikranmarsupial says:

“I didn’t say anyone was manipulating their plot, merely that it was possible to do so. “

That really is dishonest. You wrote: “This is a standard way to manipulate how the graph “looks”.” It can hardly be a standard way to manipulate graphs is nobody did it!

237. The Realclimate post has a whole section on baselines. It says

Given the internal variability of the system, baselines to short periods (a year or two or three) cause larger spreads away from the calibration period. Picking a period that was anomalously warm in the observations pushes those lines down relative to the models exaggerating the difference later in time. Longer periods (i.e. decadal or longer) have a more even magnitude of internal variability over time and so are preferred for enhancing the impact of forced (or external) trends.

It’s clear that you have to use the same baseline for all the datasets. If someone is using different baselines, then that could well be deceptive, but noone doing an honest comparison would intentionally do that.

238. Everett F Sargent says:

239. Steven Mosher says:

“Thanks, I think that makes sense. Any idea what the difference in absolute temperatures is between SST and SAT?”

SAT average is around 10C
Global average is around 15C
SST is 70% of the globe.
solve
algebra

240. Chubbs says:

Here are estimated SST trends for 1975 through 2018: HADSST – 0.14; RCP6 – 0.14.

241. Chubbs says:

We were sidetracked in this thread.

242. verytallguy says:

Steven,

I was only referring to the difference over the oceans

This site uses Akismet to reduce spam. Learn how your comment data is processed.