Correlation of AO and rate of change of CO2

[Ice Cool Kim]

Not sure what this means yet but while looking at something else I stumbled on a remarkable correlation between recent change in atmospheric CO2 and the AO index.

The break between the two lags is 1992. The earlier segment is not as close but still looks significant.

My main interest at this point is the post 1992 segment. Those small variations from 2006-2010 are quite distinctive and it seems obvious this is not a coincidental correlation.

However, seeing AO being that closely related to CO2 at Mauna Loa is quite a surprise.

[Dromicosuchus]

I can't say I'm really comfortable with this, particularly since you're allowed yourself to chop out 2.2 years in order to get a better fit (if I understand what you've done correctly; if not, please let me know).  I'd be inclined to guess, moreover, that if there is some connection it's a connection with temperature, not CO2 concentrations.  Short term variations in global temperature, of course, will affect the rate at which CO2 builds up in the atmosphere, and I can more easily envisage a connection between temperature and the AO than one between ocean acidity/plant growth rates/ice structural strength and the AO.

[Ice Cool Kim]

Yes, out-gassing is primarily determined by deviation of the water temperature from the equilibrium value that would match the current absorbed gas concentration. There is indeed a fairly good correlation between the two [ie SST and d/dt(CO2) ]. That was one of the things I was investigating when I found the AO connection.

However, AO actually matches Mauna Loa better than any of the SST records I've look at so far, be it regional or global, in particular, that pattern of two smaller variations that so clearly identifies the pattern from general ups and downs that could more easily be dismissed a coincidental.

As for "chopping out" 2.2 years: I think what is seen here is the warming period of late 20th c. correlates somewhat less well and displays a different lag. There is a phase shift or "regime change".  Exactly what happens during the changeover will presumably fit neither case. This is just a quick plot to point out, in particular the high degree of correlation since 1992.


I expected to find a correlation between d/dt(CO2) and  SST, I really did not expect such a strong correlation between arctic isobar height and CO2 in central Pacific.

I thought someone here may have relevant knowledge on this, I'm frankly rather amazed. Maybe I'm just missing something obvious.

This could have serious implications for current theories on the residence time of CO2 in the oceans.

[Ice Cool Kim]

Here's SST plot.

since hadISST is an 'anomaly' dataset the 0.3 is relative to an arbitrary mean and has no significance. The ICOADS is real temps so the scaling offset could be taken to mean something physical. Though when talking about global mean temps I doubt it is anything that can be used in a physical equation because global mean temps don't have much physical meaning.

This is all approximate fitting done by hand but serves as a first estimation of sensitivity of outgassing to temperature.  The two datasets give very similar values.

[Ice Cool Kim]

In fact taking the derivative of each quantity once more removes the slope and the strong similarity in variations is clearly seen.

However, that still leaves me rather perplexed by the AO link.

[Ice Cool Kim]

Here's another remarkable thing about d/dt(CO2) : a period of about ten years that shows almost perfect repetition from circa 1973 onto 1998.

In terms of CO2 the 1998 super El Nino  and the following stabilisation period, was an almost perfect replay of the 1974 event that marked a major shift in PDO.


Since we're plotting d/dt CO2 it is interesting to note that the offsets I used to align those events correspond to an acceleration of atmospheric CO2 concentration of about 2ppm/year/century.

This corresponds to a quadratic rather than exponential growth , though the two are similar quadratic mathematically stronger growth rate. One paper about three years ago referred to finding "super exponential growth" in CO2. That caused some cat calls from less mathematically trained climate sceptics, who did not understand that "super" just meant superior growth and thought the authors to be saying something like "mega exponential" growth.

A further look at 2nd derivative of CO2 showed it to be pretty flat ( ie close to being a quadratic) and the average gave a similar result:

2nd diff MLO CO2 conc:  mean = 0.0282617; SD = 0.797774
ie 2.8ppm/a/century.


NB. Both those figures are derived from relatively short periods and I have no idea of uncertainty estimates. so not much point in projecting backwards or forwards based on  arbitrary assumptions like "if the current trend continues".

I actually doubt the big green bump is going to happen on that plot since the correlation of the post El Nino settling seems to be breaking up.

What I do expect to see is the plunge in d/dt(CO2) next year suggested by AO plot at the top followed by a rebound in 2015.

[Ice Cool Kim]

Another thing that seems sensitive to AO index and is relevent here is the Nenana River ice collapse event the "Ice Classic".

This record looks like it is a useful regional temperature proxy.

Note the same 3.2 year shift was also applied here, giving a handy short term prediction tool for AK weather.

It hasn't gone yet ....
http://www.nenanaakiceclassic.com/

... though it's not far off. I've pre-emped this year's collapse in the plot , it will be close the AO "prediction".  It's likely to be 2nd latest break-up date in the near 100 year record.

That may provide some input for the monthly guessing games on ice area going on here.

[Ice Cool Kim]

Does that cover your questions? 

[ChrisReynolds]

Sorry, but I just keep thinking of Wunsch's warning about apparent patterns of linkage in unrelated timeseries with similar power spectra...


Anyway, thought I should mention: Some of what you were doing on melt season length needed to be addressed as it was related to the subject of a recent post of mine.
http://dosbat.blogspot.co.uk/2013/05/autumn-sea-ice-area.html
It's under the first long graph with the red and blue plots (Dates of maximum and minimum CT Area). Forgot to give you a heads up under the earlier thread on that subject, sorry.

[Ice Cool Kim]

Sorry, but I just keep thinking of Wunsch's warning about apparent patterns of linkage in unrelated timeseries with similar power spectra...

That's a fair point but what I showed in the first plot can hardly be dismissed as an accident of similar power spectra. If they're that similar you can either start to look for common cause or look to see how one is causing the other.


I tried posting comment on your site but I when to get a link here and when I switched back it had closed the comment window and lost my text. I won't bother rewriting it.  BTW you while on black kills my eyes, it was a struggle even to read you post.

In short, assuming random is not a good enough excuse for not looking.

You may like to consider this graph from one of my posts about detecting the min/max by zero crossing of derivatives. It apples just as well to your thresholds. If you don't filter you can have huge changes in the dates you find because of sub-annual noise or pseudo cycle variations.
https://forum.arctic-sea-ice.net/index.php?action=dlattach;topic=200.0;attach=983;image

If your aim is to examine longer term variability in the general state of the ice , you would do well to apply some kind of filtering as I did.  (Obviously avoid running mean as filter)

[Ice Cool Kim]

Back to AO vs CO2.

I could imagine that since tropical waters are much more constant in temperature and already seriously out-gassed due to being a lot warmer , much of the rapid change in CO2 could have its origins in cold arctic waters.

That would at least explain a link between arctic SST and d/dt(CO2)

AO will also influence arctic SST thus a linkage is credible.

However, for all that to explain the tight correlation with MLO, it would infer that the vast majority of rapid exchange of CO2 is happening in the Arctic and I'm not sure whether that is a reasonable proposition.

[Dromicosuchus]

Does that cover your questions? 

Quite comprehensively; thank you!  I'm afraid I'm still not quite convinced that what you've got here is real, but you've clearly given quite a bit of thought to the problem, at least.  Supposing this is a real linkage and not just a coincidental match, can you think of any reason why the 3.2 and 5.4 lags would be present?  I'm not familiar with any effects of the AO beyond the short-term (which is, of course, only a statement about my level of ignorance, not about the level of knowledge in the field), and although I understand that it itself can be affected by events occurring at least half a year previously, I don't recall having come across any description of it having a similar influence on future events--much less future events several years away.

[Ice Cool Kim]

I don't recall having come across any description of it having a similar influence on future events--much less future events several years away.

Well it is used in short term meteorological predictions , that much is not controversial, though I am unaware of anyone one using it 3 years ahead.  (Like yourself, that could possibly be taken as a statement of ignorance rather than a statement of fact. )

I think that pattern is far too distinctive and the correlation far too strong for this to be accidental.

However, I only found this last night so I can't offer much by the way of reasoned argument as to why it should be there. other than my hand-waving suggestions posted above.

That's also a decade that saw some dramatic changes in the arctic ice cover.