« Pinpointing the minimum | Main | Freezing season 2013/2014 open thread 2 »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Rob Dekker

Kevin, I know that Li is a lead author of the IPCC.
This is what makes this all very cumbersome.

In fact, I would truly appreciate your feedback on this.
Tell me that I'm not crazy.
But I've seen three papers now that seem to make the same mistake, what I call, "the AMO trap".

In summary, "the AMO trap" is a method to attribute (intentionally or not) non-linearity in the 100 year global temperature record (mostly due to AGW) to local natural variability (the AMO).

Here is how it works :

Pick an area of the Northern Atlantic Ocean, and plot it's SSTs over the past 100 years.
Obviously, these SSTs are a combination of the global warming curve, with local variability superimposed on them.

Now, you de-trend these SSTs over 100 years. That will give you a linear increasing temperature slope, and remaining is the non-linear global warming signal plus the local variability. Now, you call that remainder the "AMO".
That's what Li et al did.

Then you work with that "AMO" as if it represents a 'natural' and 'local' pattern, and you can then find a patterns that kind of matches the ups and downs of that "AMO", and you may even find that the NAO index, with a 16 year delay kind of nicely matches with that. Then you can start claiming all kind of causal relationships and predictions.

But remember, you are still testing against a composite signal, which includes BOTH the 'natural' and 'local' Atlantic oscillation signal that you claim AND the 'non-linear part of the global warming signal.

To know for sure if you are looking at a true AMO signal or the non-linear part of global warming over the past 100 years, you can subtract the AMO from the global warming signal.

After all, if you do that, you get the linear increasing trend (the result of your de-trending) plus the local variability and the 'non-linear global warming' signal is eliminated.

So if the multi-decadal pattern is still there, then you know it is the local variability that you were after. If the multi-decadal pattern disappears, then you have been matching against non-linear part of global warming.

Now, lo and behold, Li et al does exactly that. They state :

after subtracting the AMO signal from the NHT, the residual shows a nearly monotonic increase, which is very much like the linear trend in the NHT (Supplementary Figure 4), and the multidecadal variability disappears.


Li et al 2013 has been trying to explain the non-linear part of global warming over the past 100 years with the NAO index.

Classical case of cause-effect confusion.

Now. Please tell me where I've got this wrong.
Because if I'm right, Li et al 2013 has (intentionally or unintentionally) ran straight into the "AMO trap".

Either way : Tamino warned two years ago that this trap exists, and Trenberth and Shea warned about it in 2006.

Rob Dekker

Kevin, regarding Li et al's remark :

"..detrending the NHT time-series using the best-guess radiative forcing due to CO2 and other well-mixed greenhouse gases gives qualitatively similar results (Supplementary Figure 3), implying this significant lead–lag correlation does not depend on the method of detrending."

Note that the "best guess forcing from GHGs" does take away much of the non-linearity of the NHT. However, I can imagine that there will still be some non-linearity left over, since forcing does not immediately result in increase of temperatures.

So the long term AGW non-linear pattern will still be in the NHT, albeit with reduced amplitude.

Then note that Supplemantary Figure 3 only shows correlation plots of the NAO index 'explaining' the non-linear part of NHT.

However, no 'magnitude' of influence is mentioned. Why not ?

John Christensen

With all the noise going on, I have been mostly lurking for a while, but it seems we may now be back to focusing on the ice, which is great.

And on that, I would need to join the recovery team for these reasons:
- Area and extent have increased very well since minimum, with CT SIA now joining the 2005/06 lines
- SST was kept low due to the cyclones/weather of the summer months, allowing the quick refreeze
- DMI 80N temps have been moderate to low, with relatively strong freeze at the most northerly latitudes

What I will expect to see from the next PIOMAS update would be:
- Volume reaching 2009 level, as e.g. the 80N temps in 2009 were higher, and area/extent lower
- Thickness should go up slightly and passing 2012 level due to strong refreeze of the central pack

On the factors:
- We saw the SSW event early in the year causing strong late winter volume build-up, and then we had the much-discussed summer cyclones - all just weather events
- Greenland: While Greenland had a 'heat-dome' in the summer of 2012, we saw the opposite this year, with significantly reduced melting and much lower temperatures for most of the summer. It would be very interesting to compare Greenland climate of 2012 vs. 2013, and also to what extent the weather in Greenland impacted overall Arctic weather in these two summers. From the DMI arctic temperature maps it has seemed as if the lowest temperatures in late September/early October came from the northern coast of Greenland and benefitted from the cold in this area.

In summary, it appears to me that weather events are still able to produce some 'counter-current' effects, and it would also therefore only be a question of time, before weather turns less favorable, and we will see ice area and volume being reduced at high rates once again..

John Christensen

Great link on Greenland accumulated Snow Mass Balance (SMB), where you see this early freezing season faring better - so far - than 1990-2011 average:



'it would also therefore only be a question of time, before weather turns less favorable, and we will see ice area and volume being reduced at high rates once again'

... then you haven't really joined the recovery team, have you? :)

John Christensen

Good point sofouuk; if 'recovery' is to reverse impact from AGW I am not on the recovery team - just saying the ice is having a small rebound at the moment.

Jim Hunt

At the risk of noisily drifting off in the direction of "psychology" rather than "sea ice", the video Neven kindly showed in his original post has now garnered over 1000 views on YouTube along with a few comments, one of which states that it's:

Unscientific, and very troubling that folks like Neven as the Sea Ice Blog would direct us to this unscientific propaganda, tripe even.

Meanwhile the Managing Editor of the Mail on Sunday has finally put virtual pen to paper to say that:

We deny that the article was significantly inaccurate apart from the original headline figure which we have already corrected.

I felt compelled to beg to differ:


Returning from fantasy fiction land to the science of sea ice and John's "small rebound", I am also eagerly waiting to discover what PIOMAS makes of this area of "unbroken ice sheet" (in David Rose speak), pictured at the beginning of September:

Do you suppose the PIOMAS model has a good handle on the volume of what the NSIDC called:

An unusually large expanse of low-concentration sea ice within our extent outline spanning much of the Russian side of the Arctic and extending to within a few degrees of the North Pole.

Do you suppose that the US Navy's new NAVGEM forced version of ACNFS now has a good handle on the thickness of the ice in the area for that matter?


Neven "Although global sea ice is still in decline, certain areas of Antarctica have experienced an increase in sea ice thickness. According to Zhang’s research, this trend is due to an increase in winds in the southern oceans that decrease surface air temperature, causing the sea ice to ridge more often."

An increase in winds may mean greater wind chill factors, which do cause a net loss of heat. The same ridging occurs in the Arctic quite a lot. So there may be sudden jumps of sea ice age by mistake. if thickness is a measure of ice age.

Jim, it can be only one mean super computer which could differentiate broken ice from a basic extent calculation. It should go "pixel by pixel" . Calculating average thickness on total Volume divided by area may be a little suspicious now a days. There are more complexities related to a mix floe at minima or freeze up, compared to a solid pack of ice. The solid compaction offers a much more powerful anti-cyclonic punch or barrier as opposed to loose pack which allows more cyclones to penetrate the Pole. Another factor perhaps in synergy with cloud extent and even likely more extent by an apparent El-Nino contributing to a very warm coming winter.


Al Gore's 24 hours of climate reality kicks off shortly; focussing on the carbon bubble and disvestment.



Meanwhile, Down Under, they are talking about non-linearity, terming what is going on a staircase pattern:


Kevin O'Neill

Given that deniers are always latching on to the latest theory involving cycles (AMO, PDO, celestial orbits, etc), I understand the reticence to discuss them -- but to understand climate we can't ignore them. Obviously Northern Hemisphere temperatures affect the arctic and the sea ice.

I don't know if there is a definitive paper on the NAO/AO/AMO/AMOC; there have been hundreds - if not thousands - written. Many of these (like Wyatt & Curry or Li et al) have noted the pseudo-periodic 50 - 80 year oscillation in the AMO. For instance, in Observed and Simulated Multidecadal Variability In the Northern Hemisphere, Climate Dynamics, 2000, Delworth and Mann write:

In the context of this evidence from the extended instrumental record, previous analyses by Folland et al. (1986), Deser and Blackmon (1993), Kushnir (1994), Mann and Park (1994), and Mann and Park (1996), might be best interpreted as having described varying combinations of two orthogonal standing patterns which are in quadrature. The combination of these patterns describes a true multidecadal oscillatory signal with a 50-80 year time scale.

One of the main complaints about this oscillation is the ability (limited) to infer a relatively long cycle from such a short instrumental record, but reconstructions of the AMO from proxy data find the same varying oscillation, among them Knudsen et al. 2011, Tracking the Atlantic Multidecadal Oscillation through the last 8,000 years. Nat. Commun. 2:178 doi: 10.1038/ncomms1186.

One of the major goals today is to achieve better decadal-to-interdecadal variability in climate models. This means understanding the teleconnections and physical relationships between the various periodic and quasi-periodic climate indices that we’ve already defined. We really could care less that the exploration of these relationships will be misused by climate change deniers. Here’s a newsflash: lying liars lie.

If Li et al have found a tool to help predict Northern Hemisphere temperatures a decade and a half in advance, rather than worrying whether Watts, Morano, Goddard, The Heartland Institute, et al will misuse the information (they will) we need to use it to inform our knowledge.

Rob Dekker

Your conservative response to my criticism of Li et al 2013 is understandable.

After all, we all tend to trust peer-reviewed scientific papers, especially when written by well-respected scientists. In a world where climate sceptic bloggers, "reports" by fossil-fuel funded "think-tanks" and political dominate media, peer-reviewed scientific papers seem to be the last stronghold that separates opinion from fact.
And to a great extend that remains to be the case.

But, that does not mean that we should uncritically accept every peer-reviewed paper blindly. In fact, there have been bad and even horrible papers in the past, and the only way to separate the good from the bad is by actually checking the science.

Which brings us back to Li et al 2013.

The physical effect they propose is that the NAO index (an atmospheric pressure difference) drives (with a 16 year delay) the temperatures of an area of the Northern Atlantic (the AMO), which then in turn affects the Northern Hemisphere Temperatures (NHT).

Now, imagine for a moment that the SSTs over the area of the Northern Atlantic are EXACTLY following NHT. In other words, the NAO nor anything else has ANY influence over the SSTs in the Northern Atlantic, nor the AMO, nor the NHTs.

If you do that, and re-read the paper with that "knowledge", you fill find that their conclusions still stand, and they still pretend to be able to predict NHT 16 years in advance.

That alone should be enough evidence that they have messed up cause and effect, don't you think ?

Rob Dekker

In other words, re-read the paper, and assume that their AMO==DNHT.
And surprise, surprise, all their conclusions still stand.

Why ? Because a graph of the NAO index with 16 year delay is pretty similar to the graph of the linearly de-trended global temperatures over the past 100 years.

And that's it.
Neven, I apologize for an off-topic subject.
I now yield the floor for discussions about the freezing season.

Kevin O'Neill

Rob, Li et al *use* Trenberth's revised AMO index:

The Atlantic Multidecadal Oscillation (AMO) index during 1900–2011 was calculated by averaging detrended annual mean SST anomalies over the extratropical North Atlantic region (30°–65°N, 75°–7.5°W) with the base period 1961−1990 [Trenberth et al., 2007].

The AMO does not enter into their calculations - it is only brought up as the physical explanation of the NAO > DNHT connection.

Equation 2: NHT(t) = aNAO(t−16) + bt + c

where t is time in years and the coefficients a, b and c are determined empirically by linear regression based on the data over the historical period, so that the regression error of Equation (2) is minimized.


Just wanted to note the incredibly high temp anomalies over the southern Canadian archipelago, and to provide a link to a very nice site from the University of Maine (h/t Ed Hummel, meteorologist):


Rob Dekker


Trenberth's AMO index is obtained by SUBTRACTING the global SSTs from the Atlantic SSTs, and he explicitly advices AGAINST linear detrending. Here is Trenberth's paper on the AMO index :


However, previous AMO indices are conflated with linear trends
About 0.45C of the SST anomaly is common to global SST and is thus linked to global warming...
The main difficulty with the traditional AMO index is that it is not possible to discriminate between variations arising from the THC and other phenomena with North Atlantic origins, and global anthropogenic changes. In particular, the recent warming of North Atlantic SSTs is known to be part of a global (taken here to be 60N to 60S) mean SST increase (Figure 2). While detrending [Knight et al., 2005] the AMO series helps remove part of this signal, the SST changes are not simply linear and a linear trend has no physical meaning.
Accordingly, the global mean SST has been subtracted to derive a revised AMO index (Figure 3).

Note that the reference they give ("Trenberth 2007") does not show up in the list of References, so if you don't already know that Trenberth adviced exactly against what they are doing, then you would never know.

So Li et al explicitly IGNORED Trenberth's advice, and then still attribute their choice of AMO definition to him !

Ain't that something ?

Rob Dekker

When you look at Trenberth's revised AMO index (figure 3 in Trenberth's paper) then it also becomes clear that the AMO is no longer a good physical explanation of the 16 year delayed NAO index. For example, Trenberth's AMO index suggests that Atlantic SSTs starter to rise rapidly since '95.

So if Li et al's claim that NAO affects AMO with a 16 year delay which then affects global temperatures is still true, then we should see global temperatures go up rapidly starting about 2011. Which is OPPOSITE from what Li et al is claiming.

It may still be that NAO index controls global temperatures with a 16 year delay, but whatever the physical mechanism may be that would accomplish that, it is not via Atlantic SSTs. So the physical explanation they propose is broken.

What remains is a curve-fitting exercise of the NAO index over the non-linear trend of global temperatures, with no evidence that Atlantic ocean temperature patterns are involved.


Thank you very much for that wonderful link Tenney. It will certainly prove very helpful.

Linda Serena

Neven wrote:

:PIOMASter Jinlun Zhang has a new paper out offering an explanation for Antarctic sea ice increase (link):
...According to Zhang’s research, this trend is due to an increase in winds in the southern oceans that decrease surface air temperature.."


It is not only falling temperatures (see here as well: http://notalotofpeopleknowthat.wordpress.com/2013/08/17/study-finds-antarctic-sea-ice-increases-when-it-gets-colder/ ),

sea surface temperatures also decreased since the late 1970s:


Bottom line: Antarctic sea ice increased, because it got colder !

Sea surface temperatures had increased before between 1946-1976, again just the opposite of northern hemisphere trends.

That looks very much like cyclic transportation of heat, and is almost perfectly aligned with PDO, which is assumed to be a low frequency tail of ENSO.


Watch the Gulf stream to weaken next (may have already started http://s3.amazonaws.com/dk-production/images/17808/large/FS_Transport_cable.jpg?1359771066 ),
then AMO to go down and Arctic sea ice to increase in huge steps.

Kevin O'Neill

Rob, http://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch3s3-6-6.html

You're now going to deny the physical link - one that's been established by both theory and GCMs?

Rob Dekker

No Kevin, I don't.
I do claim however that the correlation between their NAO pattern and the AMO definition is the result of linear detrending. If they use a physical AMO definition, like Trenberth's, then the correlation breaks down. Which they even admit to themselves in the Supplemental material.

I understand that you are not convinced, and I am sorry that I could not explain myself better.

Other papers fall in the same AMO trap, and the deniers are having a party with such papers.
Take Zhou and Tung 2012 for example, which triggered this post :
That publication went all through the sceptic blogosphere all through to Marc Morano.
See my remarks in the comment section of that blog post, where I first attempt to point out the issue with the detrended AMO.

A follow-up of that paper in PNAS (Tung and Zhou 2013) was finally taken on by skepticalscience :
Which pretty much nails it too.

So for the future, please be aware of this "AMO trap". It's a way to attribute the non-linear global warming signal to the "natural
" AMO.

It WILL show up again and again, until we deal with it.

Now I yield the floor for sure..


Linda, hi,

I don’t want to be an irritationg know-it-all. So I’m not going to debit any one-liner out of my concern.
I’ve actually checked some parameters from GISS and NCEP/NCAR.

And, as obvious, the real World is not as simple as your generalized remarks seem to exclamate.

Here’s GISS annual temperature anomaly, I chose the interesting past ’07 period ’08-’12 against a baseline 1981-2010. That’s fair, because it is the usual standard base now and it honestly reflects the ‘hiatus’ (whether that is a credible discriminator or not):

 photo GISSannualtempano08to12base1981to2010vsmall_zps102106a0.jpg

Within the small anomalies, there is some cooling, regionally, over the Amundsen-/Bellingshausen and the King Haakon Sea. But there’s warming too, especially around the Ross Ice Shelf.

Then there’s the NCEP/NCAR record for the same periods:

 photo Temp1000Mbano08to12base1981to2010vsmall_zps6b4b3678.jpg

It would be interesting to check what different approaches both analyses have. This one shows obvious warming anomalies.

 photo SST1000Mbano08to12base1981to2010vsmall_zps2fb8d158.jpg

Last but not least, here’s SST (or maybe better, surface skin). To an extent, cool anomalies correspond with GISS in the Amundsen-, Cosmonauts-, Davis- and Ross Seas. Most of these lie a bit further out into the Southern Ocean, near 65dS.
Against the Antarctic coast, small bands to large swaths indicate warming anomalies. Especially near the Ross-, Wlkins-, Larsen Iceshelves and the d’Urville Coast.

This could be subject of some very interesting analysis. I’m sure our friend AbruptSLR on the Forum could do that better than me.
On this thread, I hope the illustrations help to put your claims in perspective.


Open Thread #2 is now open for business.

The comments to this entry are closed.