No, climate change is not experiencing a hiatus. No, there is not currently a “pause” in global warming.
Despite widespread such claims in contrarian circles, human-caused warming of the globe proceeds unabated. Indeed, the most recent year (2014) was likely the warmest year on record.
It is true that Earth’s surface warmed a bit less than models predicted it to over the past decade-and-a-half or so. This doesn’t mean that the models are flawed. Instead, it points to a discrepancy that likely arose from a combination of three main factors (see the discussion my piece last year in Scientific American). These factors include the likely underestimation of the actual warming that has occurred, due to gaps in the observational data. Secondly, scientists have failed to include in model simulations some natural factors (low-level but persistent volcanic eruptions and a small dip in solar output) that had a slight cooling influence on Earth’s climate. Finally, there is the possibility that internal, natural oscillations in temperature may have masked some surface warming in recent decades, much as an outbreak of Arctic air can mask the seasonal warming of spring during a late season cold snap. One could call it a global warming “speed bump”. In fact, I have.
Some have argued that these oscillations contributed substantially to the warming of the globe in recent decades. In an article my colleagues Byron Steinman, Sonya Miller and I have in the latest issue of Science magazine, we show that internal climate variability instead partially offset global warming.
We focused on the Northern Hemisphere and the role played by two climate oscillations known as the Atlantic Multidecadal Oscillation or “AMO” (a term I coined back in 2000, as recounted in my book The Hockey Stick and the Climate Wars) and the so-called Pacific Decadal Oscillation or “PDO” (we a use a slightly different term–Pacific Multidecadal Oscillation or “PMO” to refer to the longer-term features of this apparent oscillation). The oscillation in Northern Hemisphere average temperatures (which we term the Northern Hemisphere Multidecadal Oscillation or “NMO”) is found to result from a combination of the AMO and PMO.
In numerous previous studies, these oscillations have been linked to everything from global warming, to drought in the Sahel region of Africa, to increased Atlantic hurricane activity. In our article, we show that the methods used in most if not all of these previous studies have been flawed. They fail to give the correct answer when applied to a situation (a climate model simulation) where the true answer is known.
We propose and test an alternative method for identifying these oscillations, which makes use of the climate simulations used in the most recent IPCC report (the so-called “CMIP5” simulations). These simulations are used to estimate the component of temperature changes due to increasing greenhouse gas concentrations and other human impacts plus the effects of volcanic eruptions and observed changes in solar output. When all those influences are removed, the only thing remaining should be internal oscillations. We show that our method gives the correct answer when tested with climate model simulations.
Estimated history of the “AMO” (blue), the “PMO (green) and the “NMO” (black). Uncertainties are indicated by shading. Note how the AMO (blue) has reached a shallow peak recently, while the PMO is plummeting quite dramatically. The latter accounts for the precipitous recent drop in the NMO.
Applying our method to the actual climate observations (see figure above) we find that the NMO is currently trending downward. In other words, the internal oscillatory component is currently offsetting some of the Northern Hemisphere warming that we would otherwise be experiencing. This finding expands upon our previous work coming to a similar conclusion, but in the current study we better pinpoint the source of the downturn. The much-vaunted AMO appears to have made relatively little contribution to large-scale temperature changes over the past couple decades. Its amplitude has been small, and it is currently relatively flat, approaching the crest of a very shallow upward peak. That contrasts with the PMO, which is trending sharply downward. It is that decline in the PMO (which is tied to the predominance of cold La Niña-like conditions in the tropical Pacific over the past decade) that appears responsible for the declining NMO, i.e. the slowdown in warming or “faux pause” as some have termed it.
Our conclusion that natural cooling in the Pacific is a principal contributor to the recent slowdown in large-scale warming is consistent with some other recent studies, including a study I commented on previously showing that stronger-than-normal winds in the tropical Pacific during the past decade have lead to increased upwelling of cold deep water in the eastern equatorial Pacific. Other work by Kevin Trenberth and John Fasullo of the National Center for Atmospheric Research (NCAR) shows that the there has been increased sub-surface heat burial in the Pacific ocean over this time frame, while yet another study by James Risbey and colleagues demonstrates that model simulations that most closely follow the observed sequence of El Niño and La Niña events over the past decade tend to reproduce the warming slowdown.
It is possible that the downturn in the PMO itself reflects a “dynamical response” of the climate to global warming. Indeed, I have suggested this possibility before. But the state-of-the-art climate model simulations analyzed in our current study suggest that this phenomenon is a manifestation of purely random, internal oscillations in the climate system.
This finding has potential ramifications for the climate changes we will see in the decades ahead. As we note in the last line of our article,
Given the pattern of past historical variation, this trend will likely reverse with internal variability, instead adding to anthropogenic warming in the coming decades.
That is perhaps the most worrying implication of our study, for it implies that the “false pause” may simply have been a cause for false complacency, when it comes to averting dangerous climate change.
Where can I find a good time series for the PMO?
[Response: Should be able to find all the data (observations and models) at the Supp Info link provided in the article :-) –mike]
Secondly, scientists have failed to include in model simulations some natural factors (low-level but persistent volcanic eruptions and a small dip in solar output) that had a slight cooling influence on Earth’s climate. Finally, there is the possibility that internal, natural oscillations in temperature may have masked some surface warming in recent decades, much as an outbreak of Arctic air can mask the seasonal warming of spring during a late season cold snap. One could call it a global warming “speed bump”.
How are those not flaws in the models? I respect the models and the modelers, but it seems to me that the models need improvement, and you just listed areas that prominently need improvement.
[Response: You need to distinguish between the model itself, and the inputs that go into a particular simulation does with a model. The model could be perfect (meaning all the physics is exactly right) but if the climate forcings that are used are wrong (e.g. if the amount of dust in the atmosphere is wrong because it cannot be perfectly known), then a prediction made with that model will be wrong too. Our imperfect knowledge of things like the history aerosols makes some lack of correspondence between model hind casts and the actual temperature history of the planet inevitable. This is not a “model flaw”. I’m not just being semantic here. The point is that future projections with models are subject to *less* such uncertainty, because greenhouse gases are increasingly dominant, and this part of the forcing (as opposed e.g. to aerosols) is extremely well understood. –eric]
It’s paywalled. Maybe I can find it at the library.
So, what this analysis shows is that the flip side of the small slowdown in the rate of surface temperature warming is the small increase of warming in the ocean?
Steve
Two of the curves in the figure are trending downward, and the third appears to have peaked. Appears. Does this combination portend that the “faux pause” will continue (for at least a decade)?
“…we show that internal climate variability instead partially offset global warming.”
You probably want to rephrase that for the general public. You open the door to that quote getting picked up and abused to claim that the Earth has natural mechanisms to modulate global warming, so there’s no need for alarm since the planet will just sort itself out.
It’d be better to much more explicitly tie your results to the subsurface heat burial in the Pacific with a note that due to thermodynamics that Heat won’t just magically go away and is likely to show up again in the future. Global warming hasn’t been offset. Atmospheric global warming has been offset by warming in the deep ocean due to natural climate oscillations that will certainly reverse in the future.
“Slowdown”
What slowdown? Please not another zombie! The pause zombie is still walking! (Or was that a Lom Borg recently spotted in the Wall Street Journal?).
If I take a surface temp graph and draw in the linear trend since 1970 and look at the wiggles around it – then I cannot see any slowdown, just noise. Dito Tamino, who looked several times during the last years, last time here: https://tamino.wordpress.com/2015/01/20/its-the-trend-stupid-3/ animation here: http://www.skepticalscience.com//pics/there-is-no-pause.gif – and that’s NASA data, not even Foster/Rahmstorf.
Matthew Marler wrote: “How are those not flaws in the models?
All models are imperfect, but lets think about the three “flaws” Mike noted.
That low-level but persistent volcanic eruptions may be a more significant negative forcing than previously thought comes from new science only recently published in the journals, so how could it have been incorporated in the models before now? In any case, how are modelers supposed to predict the location, timing and magnitude of volcanic eruptions in advance, beyond including a certain arbitrary level of volcanic aerosols in the models?
Similarly, the slight reduction in solar output over the last solar cycle was not predictable, so how could solar variation beyond the historic mean over recent solar cycles have been incorporated in the models before it happened?
Understanding of the interaction of internal, natural oscillations as described in this post is emerging science. How can it be incorporated in the models before it is understood?
That these factors are not included in the models to date for sure makes them imperfect. As new knowledge is gained it can be incorporated in the models going forward, but some of them (timing and magnitude of eruptions and changes in solar output) simply can not be known in advance and can never be included with exact precision. To expect that they can is unrealistic.
@Matthew Marler (#2). That’s just the thing, the models are continuously being tested, improved, and updated. Hence the investigation into and identification of the causes of mismatch.
If the recent “slowing” of surface warming is the result of oscillations, then it stands to reason that on the back-end of the current oscillation periods we will experience much more rapid surface warming than usual. Do we expect that this will happen in the coming decades? If so, won’t it conceivably cause a very sudden increase in flooding, heat waves, drought, etc, beyond even what is predicted?
“How are those not flaws in the models?”
Because the model inputs for external factors (volcanoes, solar insolation) had to be set before the models were run. My understanding is that in most cases the models use solar & volcanic forcings based on observed data up to 2005, and even the forcings to 2005 for volcanoes have been updated after this was done.
-MMM
Add in the fine detail as contributed by ENSO and volcanic eruptions and the variability is well accounted for. This variability has the characteristic of reverting-to-the-mean and that mean is zero, so the resulting residual is close to a monotonically increasing AGW signal.
#5 Matthew R Marler – Does this combination portend that the “faux pause” will continue (for at least a decade)?
Note that the PDO/PMO has dramatically flipped over the past year or so: http://research.jisao.washington.edu/pdo/PDO.latest
Amongst calendar year averages 2014 was the first positive PDO year since 2006, but also the highest index value since 1997.
#2–“How are those not flaws in the models?”
Did anyone say that they weren’t flaws?
Of course, the non-inclusion isn’t a flaw when it’s remedied–but that can only really be done in an ex-post-facto study–as, inconveniently, there is low forecastability (if that’s a word) for volcanic eruptions and solar variability.
This paper oddly enough doesn’t seem to point out the earlier works which were first to show that linear detrended was a flawed method for calculating the AMO index using a similar approach.
[Response: Oh, that is odd. You mean you know of a paper that demonstrated this earlier than the referenced Mann and Emanuel (2006) Eos article? Please provide the reference. Thank you! -Mike]
It further doesn’t explain much in terms of the synoptic mechanism
[Response: I have no idea what you mean by “synoptic mechanism”. Actually, the phrase doesn’t have any clear meaning. ‘synoptic’ refers to simultaneous measurement of the atmospheric state in time (usually referring to such measurement on the 3-7 day dominant weather timescale). “mechanism” refers to a process by which something occurs. So I’m failing to understand what you mean. Sorry :-( -Mike]
– does the results agree with the OHC data for the North Atlantic which was part of the devastating critique of Booth et al (2012) by Zhang et al (2014). Does it agree with regional temperature series – e.g. does it make physical sense?
[Response: I think Booth et al stands on its own merit, but our analysis has nothing to do with Booth et al per se. As for OHC data, yes: sustained La Nina-like conditions are consistent, obviously, with increased heat burial in the tropical Pacific noted by a number of recent studies. -Mike]
Secondly, and more importantly, without updated forcings it is not possible to completely rectify whether the post-2005 series are the major contributors
[Response: One works with what is available. If you have a more up-to-date multimodel mean (say CMIP6) kindly provide the link. Thank you! -Mike]
– I surmise that the PDO being negative does contribute but how its relative contribution exceeds that of volcanism (see recent Santer paper for instance) is not yet well-understood.
[Response: Again, I have no idea what you mean. Volcanic cooling has a completely different spatial footprint from increased tropical Pacific heat burial. They are neither consistent or inconsistent. Both are viable mechanisms to explain aspects of the slow-down in warming. Which is of course what we say. Early on in the introduction. -Mike]
The regional response to volcanism in the North Atlantic for instance is particularly strong so how can we conclude that the AMO is currently only weakly positive when the large ensembles of climate models do not account for the updated forcings.
[Response: Same as above. This makes no sense to me at all. Sorry. -Mike]
Accounting for the uncertainty by using a full range of models does not mean that the target series being produced is correct – particularly given that the variability between what is included in the models exists. The only way this approach would be applicable to today would be if you had an ensemble of models re-run with updated forcings… A reduction in the projected trend for the North Atlantic increases the AMO contribution and may affect the PDO related variability. Whether the method works or not synthetically does not address the above issues.
[Response: You appear to be repeating yourself, and some of this I’m still failing to find a way to parse meaningfully. We didn’t address the uncertainty “by using a full range of models”. We *attempted* to address the uncertainty in radiative forcing estimates by stratifying models relative to what types of indirect aerosol effects were used in the simulations, and by showing that the results are robust even using single models (with adequate ensemble sizes to get a meaningful ensemble mean). That is all show in the Supplementary Information, which I trust you’ve read? -Mike]
Also with my comment above – it is odd that there are not spatial correlation maps presented with the proposed method – does the spatio-temporal evolution make physical sense?
[Response: We sought to keep the analysis simple for the time being, using representative spatial means rather getting into the (considerable) additional complexity of representations of modes in spatial fields (something that was investigated in: Steinman, B.A., Mann, M.E., Miller, S.K., Emanuel, K., Assessment of forced and internal variability in the AMO through analyses of SST data from CMIP5 historical simulations and observations, AGU Fall meeting, San Francisco, CA, December 2013). Much of my Ph.D. research was about that latter topic, see e.g. the review paper by Mann and Park (1999). Or Delworth & Mann (2000) on the AMO specifically, for that matter. This is the 2nd article in a series. There will be more to appear. Stay tuned! -Mike]
The mode of variability known as AMO doesn’t seem to be restricted to the North Atlantic SST only. It’s rather global and can be observed anywhere on the Earth’s surface. Here for example, the AMO (detrended North Atlantic SST anomaly) compared to a SH temperature index (HADCRUT4 SH, detrended).
http://www.woodfortrees.org/plot/esrl-amo/plot/hadcrut4sh/detrend:0.745/plot/hadcrut4sh/trend/detrend:0.745
[Response: Surely you are aware, from having read our article (or to be fair, just from reading this blog post alone), that a substantial focus of the study was the demonstration that the detrending approach is fundamentally flawed. -Mike]
8, 9, 11, 14, This was in the text: This doesn’t mean that the models are flawed.
Jim Eager prefers “imperfect” to “flawed”.
13, Paul S, thank you.
Thanks Paul. I was about to comment that the highly smoothed graphic of PDO AMO etc. suggests that the turnaround could take a decade or two. But looking at your monthly numbers it seems this slowness is an artifact of lowpass filtering, not of the natural system.
[Response: If you’ve got an alternative way to identify multidecadal timescale variability, I’d love to hear it. I’ve spent a couple decades working on that problem. Perhaps I’ve missed something really simple? -Mike]
This piece is referenced in a related recent piece by robertscribbler: https://robertscribbler.wordpress.com/2015/02/26/bad-climate-outcomes-atmospheric-warming-to-ramp-up-as-pdo-swings-strongly-positive/#comment-34070
“Bad Climate Outcomes — Atmospheric Warming to Ramp up as PDO Swings Strongly Positive?”
“[Response: Oh, that is odd. You know of paper that demonstrated this earlier than the referenced Mann and Emanuel (2006) Eos article? Please provide the reference. Thank you! -Mike]”
I’m aware that was the earliest reference. But Ting et al provides a series of AMO reconstructions using the CMIP3 archive for detrending and I believe that the analysis provided in that study was very systematic in doing so considering they analyzed the spatial fingerprint as well.
“[Response: I think Booth et al stands on its own merit, but our analysis has nothing to do with Booth et al.”
Unfortunately Zhang et al (2014) showed that Booth et al (2012) does not stand on their own merit. But yes it has nothing to do with Booth et al (2012) except in saying that the strong aerosol forcings in that particular model coupled with the mis-match with OHC data for the North Atlantic make the aerosol component contributions to the AMO requiring much more detailed analysis.
“Secondly, and more importantly, without updated forcings it is not possible to completely rectify whether the post-2005 series are the major contributors
[Response: One works with what is available. If you have a more up-to-date multimodel mean (say CMIP6) kindly provide the link. Thank you! -Mike]”
I guess my point was that if the models are not available using the forcings necessary to discriminate the true signal then sometimes you have to wait until you have those available. For over a year I have spoken to people about doing virtually the same thing as in your study and came to the decision that without updated forcings in the historical runs we wouldn’t trust the results over the last decade. That was the decision we made and you made a different one and that’s your right. My opinion is that if the data is not available with the updated forcings then the conclusions have to be discussed as being very uncertain over the recent period rather than being discussed conclusively. Particularly given that the volcanic contribution is real and detectable.
“[Mike… Volcanic cooling has a completely different spatial footprint from increased tropical Pacific heat burial. They are neither consistent or inconsistent.”
I think you’re missing the point here. The point is that with updated forcings the entire multi-model mean would be shifted downwards. Since climate models are used for removing the forced component if you use a dataset that you know is overestimating this component it impacts the entire procedure. This is particularly the case when its a volcanic-derived forcing reduction because its spatial character varies substantially but can be imprinted particularly on high latitude northern hemisphere regions. It even elicits a prolonged response in the Arctic Oscillation for the North Atlantic for instance. This contributes why several of the PDO related publications were considered to be inconclusive by Santer et al (2015) as being the primary contributor.
[Response: In all fairness Robert, no, I’m afraid you’re missing the point. There is no amount of shifting of global mean temperature or interpolation of Arctic temperatures that is going to explain away why we have seen anomalous increase in tropical Pacific trade wind strength, sustained tendency for La Nina-like conditions, and enhanced tropical Pacific ocean heat burial, over the past decade+. There is a healthy body of research, as you know (and as is cited in our article) showing that these factors have contributed to a slowing of global-mean warming over the past decade+. That is what our article (and several others by England et al, Risbey et al, etc) are investigating. I think we explain this pretty clearly in the article and in the blog post, so I’m rather perplexed that the point is being lost. There is an excellent discussion of all of this by Stephan Lewandkowsy here. I’m honestly perplexed at the impasse in this discussion. –Mike]
If you can’t represent both the spatial character of the forced response and the magnitude accurately over the past decade then how can it be used for disentangling with confidence?
17 Mike, inline: that a substantial focus of the study was the demonstration that the detrending approach is fundamentally flawed. -Mike]
Here is what you have in the supporting online material:
Regression Method
To calculate the AMO, PMO, and NMO we 1) regressed the observed mean temperature series onto the model derived estimate of the forced component, 2) estimated the forced component of observed variability using the linear model from step 1, then 3) subtracted the forced component from the observations to isolate the internal variability component.
I think you have shown that an alternative to detrending can yield alternative estimates of the AMO, PMO and NMO such that Model + Background is much closer to the observed data. And you have shown that the approach has serious consequences for interpretation/expectation. That is good work, imo, but to conclude that it improves upon the flawed smoothing I think will require other evidence, not based on your model, that your estimates of AMO, PMO, and NMO are in fact improvements. I expect you have already thought of this (really!) and I shall Stay Tuned, as you advised another commenter.
[Response: Sorry, not quite following. We show that linear detrending aliases forced variability into apparent internal variability. That has nothing to do with smoothing. It has to do with the simple fact that the (anthropogenic + natural) radiative forcing is not linear in time. In the real world, this is not precisely known, but in the model world, it is–and we show that linear detrending fails to separate forced and internal variability in that laboratory, while the method we propose succeeds into doing that. The smoothing is simply to isolate the multidecadal component of the internal variability, but it is not in any way involved in the separation of forced and internal variability. –Mike]
MMM:
The landmark 2013 paper by Kosaka and Xie showed that in a GCM, when sea-surface temperatures in the tropical Pacific were constrained to the observed history rather than allowed to vary randomly over time, it accounted for nearly all the departure of the observed GMST from previous model projections:
Of course, pseudo-skeptics would say that proves the models are tuned to give the desired results, but genuine skeptics would say it’s a significant advance in understanding.
Do tell us, Mathew, how you would propose to “improve” the “flaw” of not being able to predict the location, time and magnitude of volcanic eruptions years and decades in the future?
Or how how you would propose to “improve” the “flaw” of not being able to predict unusual changes in solar output in advance.
Neither of those factors are within the purview of any climate model to predict, so sorry, but “imperfect” is as good as it is going to get in their case.
Watch that phrase “natural oscillations in temperature.” It is sure to be picked up and repeated as “RealClimate admits GW is nothing more than natural oscillations.”
Yes, Mike did explain it well. The denialists are very good at failing to understand.
“So there remains a justified general bias toward the simpler of two competing explanations. To understand why, consider that for each accepted explanation of a phenomenon, there is always an infinite number of possible, more complex, and ultimately incorrect, alternatives. This is so because one can always burden failing explanations with ad hoc hypothesis. Ad hoc hypotheses are justifications that prevent theories from being falsified.”
http://en.wikipedia.org/wiki/Occam%27s_razor#Ockham
this has been picked up by the mainstream press, for obvious reasons. personally, i’m glad it’s being addressed by the msm, as i’m tired of the hand-having.
i just want to point out that, for pr reasons, the term “natural process” is problematic. people tend to interpret that as meaning it does not have a cause, and cannot be altered – like as though it’s the work of god, or something. “natural” is also associated with “pure” and “perfect” by the dominant hippie culture. it’s the opposite of “unnatural”. if you ask around, you’ll realize there are deep roots to this in human culture [it’s a pythagorean cosmic opposite]. to a scientist, it’s absurd: sure. but i think this is something that requires a little thought, and maybe some language modification to account for how people interpret it.
in this case, “random” would be better than “natural”. generally, “physical processes” would get the point across better. but, i wish to point out a problem rather than to provide a solution.
This kind of approach teaches much more than claiming the features in the temperature record are not distinguishable owing to noise as was done recently. https://www.realclimate.org/index.php/archives/2014/12/recent-global-warming-trends-significant-or-paused-or-what/
25 Jim Eagler: but “imperfect” is as good as it is going to get in their case.
I agree. I didn’t say different.
I think that this is the wrong way of looking at whether the pause is “faux”. There are many factors, many probably unknown or unmeasurable or badly measured, that affect short-term climate changes. One cannot expect models to include them all, and one can always suggest factors not in the model to explain departures from the model.
Rather one should determine whether any departure from a trend is at odds with there being an underlying long term trend. If there is an underlying trend, the difference between the temperature line and a trend will not grow in the long run. The temperature curve will snap back to the trend, and snap back harder the greater the distance. If that has happened in the past when temperature appears to depart from the trend, it will do it again for any current departure, absent a change in regime. That is, one should test for trend stationary (or cointegration with CO2, if one thinks that drives temperature).
[Response: Time series analysis (a topic on which I’ve published dozens of times alone provides no physical insight. Using climate models allows us to introduce some actual physics into the problem. As someone who has worked in the area of time series analysis for two decades now, I’m painfully aware of the shortcomings in relying entirely on such approaches when it comes to establishing causal relationships. There are paper that
have used precisely the approach you’re suggesting, and they’ve been found wanting. –Mike]
23 Mike inline: We show that linear detrending aliases forced variability into apparent internal variability. and more
Thank you for your replies.
Re Mike inline in comment #1:
The Data and Output diectories at the location linked from the Science article appear to be empty. However, there are Matlab programmes in the Code directory so perhaps the permissions for the files in the other two directories may not have been set properly.
[Response: Thanks for the heads up Roman. Byron is working on this as we speak. Should be up later. –Mike]
Michael Mann’s study reinforces what was already known with high confidence from many previous studies; namely, that nothing in any of the global temperature metrics indicates a significant reduction in the planetary energy imbalance, which is the driver of global warming. Therefore, the climate will continue to accumulate heat for the foreseeable future. Claims that “global warming has stopped” are not just demonstrably wrong, they’re also unphysical. Attempts to argue for a lower climate sensitivity based on a short-term reduction in surface warming are also invalid.
I’ve been trying to post comments but I keep getting flagged as spam – is there something I can do to change that?
I agee with this statement but I don’t see how you can be sure the ‘NMO’ oscillation did not conributed to the observed warming in the previoius 3 decades1970-1999. Is that not because you are assuming that the models alone explain all the observed temperature increase during that period?
Finally where does this leave the AR5 attribution statement?
Fig 10.5 assumes natural variability 1951-2010 to be 0.0 ± 0.1 C. Your analysis implies a positive NAT temperature anomaly during that period with larger error bars.
I think where Tsonis and Curry and Edim went off the rails is the point where the Surface Air Temperature makes a pronounced departure from the PDO in roughly 1983. The AMO was in a positive phase then, and they foolishly concluded the PDO was no longer driving the SAT, and that the AMO had replaced it as the driver.
What the PDO was doing was progressively offsetting warming from 1983 until around 2012. I have been saying this at Climate Etc since 2012, and would often tell that crowd the PDO was about to go positive. I personally don’t think the AMO does much at all. I view it as an aimless wanderer that occasionally sticks its thumb out and takes a ride on the PDO. The PDO is likely in a positive phase, or about to be in a positive phase, so the 60-year pattern of the AMO is going to be up 30, a slight flat spot, and up for 30 more.
Children today may never know the negative phase of the AMO!
#33 Inline.
Thanks Mike.
Not to veer completely off-topic, but (@deathtokoalas): “the dominant hippie culture”? Huh?
2, eric in line. That was a good response, and I accept most of it. Mike also wrote: Finally, there is the possibility that internal, natural oscillations in temperature may have masked some surface warming in recent decades, much as an outbreak of Arctic air can mask the seasonal warming of spring during a late season cold snap.
I would think of an unmodeled natural oscillation as a “flaw in the model”, but I can see how you would think that the model is pretty good in the main; and the point of the study was to explore the implications of the unmodeled natural oscillations for model validity, and they showed that the natural oscillations can be estimated post hoc.
Thank you for your reply.
Would it be possible to scale the AMO, PMO, and NMO signals down for their contribution to the global trend and then subtract that from a surface temperature dataset to get a better approximation of the underlying warming signal?
> flagged as spam … anything
The popup for that warns you of words that contain strings that aren’t tolerated. Words related to gam bling and pres cription are iffy.
sp ec ialist
One way to manage it is to type the response in a text editor then post one paragraph at a time — and break up each likely word.
It’s a puzzlement.
Robert Way
I had that problem on my PC, which could not handle Recaptcha. My tablet is more competent, though you have to follow the procedure faultlessly!
I notice that the Recaptcha format has changed since I last posted and now shows a separate window with a word superimposed on a photo, underneath a “Type word here” panel
I think there is a bit of over-confidence in conclusions here. If all the answers to the model discrepancies can now be explained, I look forward to updated model runs which will now be free of these flaws that can then be verified with future observations.
Should we now know when the faux pause will end and temperatures will get back in line with the original model predictions? When will this be?
Assuming that we might not actually have the correct answer to internal variability yet, what parameters should we be looking at to verify these new calculations/forcing adjustments are correct when looking at future observations?
Retrospective analysis is fine and necessary, but the real proof of model efficacy is prediction skill against a reasonable null model. It is reasonable and justifiable to claim this standard has not been met yet. We all would like to see 30 years more data to come to better conclusions.
There are mixed messages to claim that the runs are too short to demonstrate major model problems due to the observational discrepancies so far but also claim they are long enough to show model efficacy as is apparently claimed here.
[Response: The natural variability is inherently unpredictable on timescales of more than a few years (people are working on decadal predictability, but I don’t expect much success). But we can constrain it’s variance from models and observations, quite well. That’s why the prediction skill — against a reasonable null model based on those observations and models — has indeed been demonstrated, time and time again. See e.g. https://www.realclimate.org/index.php/archives/2007/05/hansens-1988-projections/, https://www.realclimate.org/index.php/archives/2012/02/2011-updates-to-model-data-comparisons/, and https://tamino.wordpress.com/2015/01/20/its-the-trend-stupid-3/, just as a few examples. –eric]
How does this paper fit in with previous ones that identified the “pause” as being due to heat moving into the deep oceans, volcanic aerosols, inadequate analysis of polar temperatures etc, etc etc?
[Response: Thanks for question Adam. It is discussed briefly in the piece above (see hyperlinks too), and in more detail in the article. Our mechanism is consistent with increased heat burial in the tropical Pacific as noted by Trenberth & Fasullo etc. And we’re only looking at one contributor. Doesn’t mean that other contributors (i.e. volcanism, etc) aren’t playing a role. –Mike]
Clive, If you are concerned about isolating the NMO’s contributions and not having it contaminate the regression, simply use other non-temperature metrics that have reversion-to-the-mean properties. Two of these are the ENSO SOI time series and the delta LOD anomaly. This combo (plus volcanos and TSI) works just as well and it answers the nit-pickers.
“…the dominant hippie culture…”
OT, but LOL.
Re: my post 31 and mike’s answer. The bedrock of time series analysis is now stationarity and cointegration. As far as I can tell, it is little used in climate studies. If it’s been discredited there, I’d like to know about it.
Basically time-series analysis assumes that the forces driving long term temperature change in the past are largely continuing. The climate models assume that the inputs and putative interactions are reasonably correct. I, like almost everybody else, have no way to judge that.
The time-series analysis is easy to understand – if many departures from the trend in the past have gone away, then they are very likely to in the future. It’s not hard to understand that the leveling off of temperature growth is a common occurrence and does not mean a long-term slow down.
Temperature is “trend stationary”, which means it keeps moving back to a trend line. It is “cointegrated” with CO2, which means that its trend does not depart much from the CO2 trend.
These procedures in themselves do not allow for forecasting, but they mean that the current pause suggests large temperature increases in the next few years.
Another point – I think that studies of climate trends should concentrate on SST. Land temperatures are erratic, and depend on things that cannot be forecasted, such as albedo changes. Ocean temperatures represent a much bigger reservoir of heat and determine land temperature, and not the other way around. Trends in land temperature are really irrelevant when looking at global warming.
@ t marvel feb27 3:23. I think you are touching on the logical limitation in climate modeling that is the basis for a rational skepticism. Time series analysis doesn’t work for climate modeling because a non-linear dynamical system is formally unpredictable. The math is intractable. That leaves an academic climate scientist looking for his keys under the streetlamp; using conventional quantitative mathematical models. The problem is not just that relevant “factors” are unknown, unmeasurable or badly measured. It’s that the “factors” landscape is itself necessarily a chosen post hoc construction at a chosen point in time, from which it then deforms going forward in time. This means that progressive refinements of a model to fit the latest data is not mathematically/logically the same thing as improving the predictive power of the model. But this is how progress in climate science is depicted to laypersons—progressively refined models. This is why explaining away discrepancies in prediction versus outcome as previously unrecognized “factors” in an otherwise accurate model is actually a logical faux pas.
[Response: Thanks for stopping by and brightening our day with your wisdom! –Mike]
Great post, and pretty on target. Multiple studies have shown the TOA energy imbalance to be something around 0.6 to 0.9 w/m2. This energy has been mainly going into the ocean, but natural variability has an impact on tropospheric warming mainly due to the ocean to atmosphere rate of latent and sensible heat flux. The “pause” was a great chance for pseudoscience to create uncertainty about anthropogenic global warming.
[Response: Thanks for the comment. –Mike]