New Computer Fund

Tuesday, November 26, 2013

The Atlantic Multi-decadal Oscillation Misconceptions

Vaughan Pratt, one of the more qualified commentators on the Climate Etc. blog made another one of those comments that just mystify me, that is impact on the AMO is only +/-0.1 C degrees.  The AMO is an oscillation by definition not design that is detrended for easy illustration more than anything else.  The 0.1 +/- C is a result of the use not the phenomenon.

Once you take an anomaly after detrending your mean or average is locked in place then with a little extra smoothing like annual anomaly you pretty much have lost all of the reality of the "thing" being used to create the oscillation index.  This is a rough mask of the North Atlantic from 20N-70N and latitude 20E-90W.  The average anomaly in orange has all of the seasonal signal and the anomaly in blue has the average seasonal cycle for the full period from 1854 to 2012 removed.  The average of the full period is 16.2 C degrees and the median is about a half degree lower at 15.7C degrees.  The comparison seems to indicate that the "AMO" signal drifts about +/- 0.5 C which is about 5 times larger than Dr. Pratt's +/-0.1 C degrees.  Dr. Pratt appears to have underestimated the impact of the "AMO" because of the smoothing methods he is trying to explain and the general confusion over what is and is not a climate "oscillation".

Since Redneck's aren't statisticians or logicians in the formal sense, all I can do is question the "common sense" in assuming something is "normal" and negligible when from the looks of it the formal statisticians and logicians seem to have underestimated the potential impact by a factor of 5.

Dr. Pratt does seem to be more impressed with the Pacific Decadal Oscillations (PDO) because that has a larger swing so by his logic it can have more impact.  The PDO is based on how North Western Pacific fish stocks respond to climate oscillations which varies more in the North Western Pacific than it the whole northern Pacific.  Same basin range of fluctuation just a different region picked out for different fish.  If he looked at the entire northern oceans from 20N to 70N he would find that the whole shebang fluctuates pseudo-cyclicly.

I used both axis to highlight things with this one for the northern oceans.  It has been noted that the northern hemisphere with a larger percentage of land tends to amplify temperature changes which some seem to think indicates that it's "worse than it looks" because they assume they know what "normal" is supposed to be pretty much like they assume that defined oscillations mean more than they are supposed to mean. 

I have to admit though that thanks to Dr. Pratt and Greg Goodman I now know how to smooth the crap out of any time series. 


  1. I would not dismiss AMO so easily.

    In fact I don't like detrending or "anomaly" calculations. Now we know how to apply a good low-pass let's do it. (Please note the arbitrary 0.1 adjustment around 1925)

    This is gaussian filter, non detrended N. Atl SST and and ACE (cyclone energy)

    PDO was _discovered_ because of fishing data, it is not calculated that way. It is derived from SST. Look it up for details.

  2. The PDO and ENSO were discovered because of fish stock changes and then "defined" to relate to the changes. I am not dismissing either, I am just not assuming they are the best form to use related to "global" climate since they aren't causes they are effects.

  3. Hi Capn,

    You're probably thinking of my sentence "There is more to ocean variation than just the 0.1 °C amplitude Atlantic Multidecadal Oscillation" in my Climate Etc. post of 6/11/11, "Harmony of the climate: isolating the oscillations in many climate data sets" Google will find it for you or click on

    That was the only mention of the AMO anywhere in my post, and was not intended as more than an order-of-magnitude side remark about the AMO. Had anyone complained that I should have said 0.2 °C I'd have taken a more careful look at the AMO and then probably cheerfully agreed. I would have been less cheerful about agreeing with your suggestion of 0.5 °C however, at least when AMO is interpreted as "multidecadal" (there are indeed larger subdecadal swings as you say).

    As Bob Tisdale pointed out here

    "the PDO data you’re discussing from JISAO is not presented in deg C. The data has been standardized. There are no units."

    So you can safely ignore my uninformed ramblings from 2011 about comparing the PDO to the AMO. :)

    Looking at ESRL's idea of the AMO at

    (which they've smoothed with a 120-year moving average), the first swing up is 0.1 C but after that it's more like 0.2 C. I've plotted their smoothing here

    along with the smoothings obtained by my F3 filter (using my coefficients, not Goodman's 1.3371) and also as obtained with the equivalent gaussian filter (red curve). The main difference between the latter two is that the gaussian has a couple of slightly sharper bends, consistent with slightly more leakage beyond F3's 120-month cutoff point. Since 1.3371 is considerably sub-optimal for an F3 filter it too produces similar sharp-bend leakage though less extreme than for the gaussian.

  4. Vaughan, " I would have been less cheerful about agreeing with your suggestion of 0.5 °C however, at least when AMO is interpreted as "multidecadal" (there are indeed larger subdecadal swings as you say)."

    As I said I don't consider the AMO as defined to be very good climate index, but a good weather index. For climate though, the north Atlantic covers ~11% of the total ocean surface and has the amplified impact on land of ~1.8 times the actual SST variation. So if you consider ~.2 plausible the ~.38 is just as plausible including the land amplification typically neglected.

    However, since it is the entire northern oceans, not just the AMO region that fluctuates, 0.5 is likely for a properly considered "climate" index.