Denial isn't just a river in Egypt, it runs deep in the icecap too

This whine is funny enough to be worth reading. It's McLean et al's attempt at a response to our Comment.

Behind all the hot air (which I'll return to) the basic scientific premise behind their reply (pdf here) is a straightforward bait-and-switch. Remember that the main bones of contention in their original paper were their claims that

"Change in SOI accounts for 72% of the variance in GTTA for the 29-year-long MSU record and 68% of the variance in GTTA for the longer 50-year RATPAC record."

(where SOI = Southern Oscillation Index: GTTA = Global Tropospheric Temperature Anomaly, and the other acronyms are the names of two data sets), and subsequently that their analysis

"shows the potential of natural forcing mechanisms to account for most of the temperature variation" [over the last 50 years].

Of course, our main complaint was that the statistics were based on differenced data, from which any long-term variability had been filtered, and thus their analysis provides no support for the latter claim. In their paper, they specifically claimed that their differencing operation was performed to remove noise:

"To remove the noise, the absolute values were replaced with derivative values based on variations. Here the derivative is the 12-month running average subtracted from the same average for data 12 months later"

which we have shown is not true - instead, it removes long-term variation including any trends.

Their reply first claims that the differencing was only used to establish the lag:

"we used derivatives only to ascertain the existence of the relatively consistent time-lag that exists between changes in the El Niño-Southern Oscillation (ENSO) and later changes in the global average lower tropospheric temperature anomaly (GTTA)"

But then they immediately contradict themselves and admit the obvious, that the statistics they presented were in fact based on the differenced data:

"Our comments about the change in Southern Oscillation Index (SOI) accounting for 72% of the variance in satellite (MSU) GTAA, 68% of variance in the radiosonde (RATPAC-A) GTAA and 81% of variance in the tropospheric temperature in the tropics were made in the context of the discussion of our derivatives based on differentials between 12 month averages, and we stand by them. Contrary to Fea10 claims, those figures do not refer to long-term variations but only to the derivatives that were used."

Obviously I'm pleased to see them acknowledge the obvious, that their statistics have no relevance to the analysis of any long-term trends in the data. But that leaves their original claim concerning "potential of natural forcing mechanisms to account for most of the temperature variation" to be devoid of any support whatsoever.

In summary, it seems we all agree that ENSO/SOI accounts for a lot of the short-term variability in global temperature (which has been well established for decades) and McLean et al now appear to explicitly agree that their analysis has no bearing on long-term trends. If only they'd said that at the outset then there would have been no need for all of this.

Oh, and their fluff and bluster...will have to wait for another post. I'm off to bed.

[jules' pics] 3/23/2010 08:32:00 PM

our camelia, originally uploaded by julesberry2001.

Previously I wrote that camelia were a bit dull, but at that point I had forgotten about *our* camelia. It is all different when they are your own babies... We have three bush-ettes. Just one flower so far.

--
Posted by jules to jules' pics at 3/23/2010 08:32:00 PM

Sea level rise prediction cage fight

RC spends much of its time defending the IPCC (to the hilt). And broadly speaking, this is fair enough, especially in the context of the public discourse where the alternative is not a slightly nuanced and updated version, but rather a torrent of wingnut denialist nonsense.

So it was a bit of a surprise to see Stefan Rahmstorf's broadside about "Sealevelgate":

In its latest report, the IPCC has predicted up to 59 cm of sea level rise by the end of this century. But realclimate soon revealed a few problems.

[...]

Some scientists within IPCC warned early that all this could lead to a credibility problem, but the IPCC decided to go ahead anyway.

Nobody cared about this.

Reading between the lines a little, I somehow don't think this is really just about the AR4. Scientifically speaking, that is old news, indeed it was old news by the time it was published. Rather, this looks like more like an opening salvo over the forthcoming AR5. Alert readers may have noted that the current Chair of WG1, Thomas Stocker, was a co-author on the Siddall et al sea level rise paper that Rahmstorf (with Vermeer) forced the retraction of, and was also one of the Coordinating Lead Authors on the relevant chapter (10) of the AR4. It may not be too great a stretch to imagine that there may be some strained relations there.

I don't have any particular stake in the sea level rise predictions. I do think it was pretty spineless and inappropriate of the IPCC to duck the issue in the way that they did (by basically ignoring dynamical ice sheet response - if they really thought this was not going to happen, they should have said so clearly). OTOH the conceptual time series models presented by Rahmstorf and others may be a bit too simplistic to have much credibility for predictive purposes. I suppose I won't know what I really think until I look into it myself...which may be some time away. In the meantime, pass the popcorn, as Eli would say.

[jules' pics] 3/22/2010 07:11:00 PM

shizen! Near Okutama, originally uploaded by julesberry2001.

The Japanese go to huge effort to make sure their countryside stays beautiful and natural looking.

--
Posted By jules to jules' pics at 3/22/2010 07:11:00 PM

Is all statistical analysis wrong?

Was pointed to this by a commenter here (thanks PB2).

It's another article along the well-worn lines of p-values being misinterpreted (as the probability of the null hypothesis being true), which I have blogged about ad nauseam (it's not that narrow though and talks about multiple comparisons too). It also extensively quotes Steve Goodman, and you can find me doing the same here. Therefore, it should hardly be a surprise to hear that I basically agree with the article, although perhaps the message is perhaps a bit overstated and hyperbolic in some places: "in practice, widespread misuse of statistical methods makes science more like a crapshoot." For some other more nuanced and generally well-informed views on it, see here (and the links in the comments there too).

Here's another article of relevance which I came across recently.

One point that the opponents of p-values sometimes make is to claim that confidence intervals solve the problem that p-values have (by providing an interval within which the unknown parameter probably lies). However, this is not actually true, and they are also equally open to misinterpretation. As it happens, I've recently written an article which touches on this point, which can be found here.

Having said that, I still use p-values when it suits me. I do try to make sure I don't misrepresent what they mean though...

Statistically improbable phrases

"it is time for some humility, concludes Roger Pielke Jr."

Oh, he means *other people*. Almost fooled me there, Roger!

[jules' pics] 3/18/2010 09:57:00 PM

Zuisenji ume, originally uploaded by julesberry2001.

Plum blossom season is well and truly over, but I couldn't not share this lovely picktur wot I took with my 6.5 foot bipod, with the blossom looking like falling snow or summat. Especially appropriate as it did snow several times during the plum blossom season this year.

--
Posted By jules to jules' pics at 3/18/2010 09:57:00 PM

Mclean debunked (at last)

Better late than never, our Comment on McLean has been formally accepted. It has spent most of the last 6 months on the Editor's desk, awaiting the decision. The peer review was painless apart from the time taken - the three reviewers were all positive and made a small number of helpful suggestions which were easily incorporated into the text. However, the time taken (and it could have been months longer) is something that the AGU might like to think about.

Amusingly, the comment will be published alone, without the customary Reply. Why? Because...McLean et al couldn't muster a reply that was publishable (and not for want of trying, either - it was simply rejected). No doubt this will further strengthen their belief in a widespread climate science conspiracy to suppress their groundbreaking research. Hopefully, disinterested observers will conclude instead that there is simply no valid defence of what they did. I have no idea what they tried to say, although the press release at the time of the original publication shows that they were aware of the obvious criticisms and had some wriggling already well prepared. I have emailed them to ask if they will share their response. If their response had been accepted, it would have potentially taken another two months for our (sequential) revisions, and another month of reviewing. All for a short paper which points out glaring errors that should never have made it into print.

Just to go over the science again...it has been well-known for decades that ENSO/SOI is related to interannual temperature variability. It has also been repeatedly shown that this cannot represent more than a modest part of the total temperature variations and a negligible part of the long-term trend. Evidence for all of this is referenced in our Comment. The primary error of McLean et al is that although they filter out all long term changes, they still claim that the resulting high correlation between SOI and global mean temperature (in the filtered series) has relevance for long term trends. As shown in the toy examples in the comment, this is simply not true - the correlations calculated by their analysis method are completely unrelated to any long-term trends in the underlying data. A secondary error is that they splice together two data sets which have different baselines, which artificially reduces the warming by about 0.2C. A third error is that they claim to identify two flat periods with a breakpoint in the middle (for both SOI and temperature), but their statistical analysis provides no support for this.

Update: Sceptical science has a fuller write-up of it here.

Update2: Someone purporting to be McLean replies in the comments, confirming that he received my email but is not prepared to share his reply, which seems curiously at odds with his bullish attitude regarding his research. What odds it will be published in that septic journal of last resort bastion of quality, the non-ISI-listed E&E (eds: Boehmer-Cristiansen and Peiser) or not at all?

[jules' pics] 3/16/2010 11:02:00 PM

Egaratenjin, originally uploaded by julesberry2001.

Wikipedia says, "International Klein Blue is outside the gamut of computer displays, and can therefore not be accurately portrayed on webpages." I wonder if it is true.

Anyway, here is some Japanese Shinto Orange.

--
Posted By jules to jules' pics at 3/16/2010 11:02:00 PM

[jules' pics] 3/15/2010 10:18:00 PM

Broken trees, originally uploaded by julesberry2001.

The ginkgo of Hachimangu wasn't the only tree to fall down in last Tuesday night's snowstorm. On Saturday we climbed Hinode, a little hill near Tokyo, and there were contorted dead bodies and severed limbs scattered all over the place. Poor trees!

Actually, wind ripping up trees is pretty cool even when (or maybe especially when) it is an old tree; it is man ripping them up that is upsetting, and the latter is a whole lot more common than the former.

--
Posted By jules to jules' pics at 3/15/2010 10:18:00 PM