Begin by reading Role for Eurasian Arctic shelf sea ice in a secularly varying hemispheric climate signal during the 20th century? That post offers some snarks on the paper, and some indications why you might distrust it, but no really substantive criticism. I’ll try to do that here but I won’t fully succeed because (just like Wyatt and Curry) I don’t really understand MSSA. I’m hoping that someone how does know it will do a proper analysis.
Where’s the meat?
If you look through the paper to find the core substance, you won’t. There are layers of mush and piles of words but precious little hard matter. Figure 12 is about the closest they come to a mechanism, but its just a pile of words arranged in a picture; there’s no maths here. So really, we’re reduced to figure 2 and similar as being the only vaguely convincing bits.
And… well, it looks good, doesn’t it? All those nice smooth wave-like lines showing clear evidence of “propagation”. What more could you want? Weeellll…
First of all, the smoothness is because all the indices have been heavily filtered into a spectral band (update: Eli has a nice post showing quite how heavy the filtering is. Note that this isn’t fatal of itself – the ACW was also heavily filtered – but). Not precisely this, because its been done via MSSA, but effectively so. Unfortunately they don’t show this in spectral space somehow, which I think would have been helpful. However, they do show you the proportion or power of the original signal that remains in this band, and this is really very revealing.
As ever, click on the image for a larger version. Note that I’ve cropped the caption. Ignore the ovals drawn on, instead note the fraction of variance for NINO in the range – about 1%. That makes sense: we all know that NINO is quasi-periodic at ~5 year, so you expect little left over at very slow periods. But this means, in turn, that NINO is irrelevant at their timescales of interest – as, very likely, we could have guessed at the outset. So whatever their MSSA has done, it certainly hasn’t allowed them to filter out variables that contribute little to the pattern.
But then why is the black line in figure 2 of about the same amplitude as all the others? Because they’ve all been normalised. A more representative version of this picture would renormalise the lines by some-version-of-variance; in which case the black NINO line would be essentially flat.
Once you realise that, you can see that EIE, NAOw, and ALPI, all with less than 10% variance in band, are also negligible at this scale. EIE is “East Ice Extent” of the East Eurasian Seas; even W+C note that its negigible at this scale, and its not in fig 2, so we’ll forget it. NAOw is NAO in winter; since they don’t provide a pic for it I’ll have to assume that NAO, too, has negligible variance in-band and it too should be ignored. ALPI is the Aleutian Low Pressure Index.
Then I thought: but hold on, NHT is the hemispheric temperature index; it can’t propagate. Removing this, along with the other three that have essentially no in-band variance, leaves me with this crudely retouched version of their figure 2 a. Its now much less obviously a wave; its just three (four really, but two essentially overlay) different lines filtered to within an inch of their lives into a 60-year-ish band.
I think that’s about it, really. All the stuff about exploding sardines is just fluff and can be ignored. The “mechanisms” is an extended exercise in self-delusion.