30 October 2008

Pielke's poor summary of sea ice

I was amazed to see the following quote from Roger Pielke Sr. in an interview published yesterday in Mother Jones

Roger A. Pielke, Sr.:
In terms of sea ice, if you look at Antarctic sea ice, it actually has been well above average, although in the last couple days it's close to average, but for about a year or longer, it's been well above average, and the Arctic sea ice is not as low as it was last year. So in the global context, the sea ice has been fairly close to average. It doesn't mean it can't happen because we are altering the climate system. But whenever I look at the data, I see a much more complicated picture than what you typically hear about.

It's usually the case that if you look at the data yourself, the picture is more complicated than 'what you typically hear about'. (It also making some difference where you usually listen.) So that's a noncomment.

It's shocking, however, to hear someone who says he is looking at the data arrive at the conclusion that 'sea ice has been fairly close to average'. A brief visit to Cryosphere Today, a site well worth a long visit and run by a fellow (William Chapman) who is a scientist who studies sea ice (which isn't Pielke's area), quickly takes you to the anomaly graphs for the northern hemisphere and the southern hemisphere. The northern hemisphere is far (about a million square km) below the climatology, and has been below that average continually since early 2003. The trend in this curve became apparent years ago (compare the scatter in the first 21 years to the difference between 0 anomaly and where the northern hemisphere has been the last 5 years). The southern hemisphere trend, which is there and positive (towards more ice) only recently emerged from background noise. Compare the current value (eyeball of about +0.3 million km^2 as I write on the 30th of October) to the scatter (eyeball value of about 0.5 million km^2) for the Antarctic and you see why it's taken so long for a trend to emerge from noise).

So on one hand, we have the Arctic ice, which is well (a couple of standard deviations, by eye) below normal, and has been below normal continually for over 5 years. On the other hand, we have the Antarctic, which shows a statistically weak trend and has been bouncing back and forth across normal every year of the record.

However one may describe this for the global net effect, 'fairly close to average' isn't an option.

I'm emailing this to Dr. Pielke once I find an address for him. I'm hoping that he simply was quoted exceedingly badly.


Hank Roberts said...

pielkesr at cires.colorado.edu

3 links deep from his home page to find that, from:


(down near the bottom is a clickable link)

Anonymous said...

I'm hoping that he simply was quoted exceedingly badly.

I wouldn't be taking bets on that...

Hank Roberts said...

I added a comment at Mother Jones.

Robert Grumbine said...

Pielke's reply to my email:

Dear Dr. Grumbine

Thank you for your e-mail alerting me to your comment. With respect to the Mother Jone interview, I was interviewed for this months ago and, at that time, the Arctic sea ice anomaly was not far below zero (e.g. see March and early April 2008 at
http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/current.365.jpg), and the
Antarctica sea ice coverage was well above average at that time (e.g. see http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/current.365.south.jpg). See
also the global ice areal coverage anomaly at http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/global.daily.ice.area.withtrend.jpg
for the same time.

More recently (until the last few weeks), the global anamoly fell well below average, after rising to above average. The use of as linear trend is clearly incomplete when assessing its long term behavior.

With respect to publishing on sea ice issues, I have co-authored several peer reviewed papers on this subject; e.g. see

Pielke Sr., R.A., G.E. Liston, and A. Robock, 2000: Insolation-weighted assessment of Northern Hemisphere snow-cover and sea-ice variability. J. Geophys. Res. Lett., 27, 3061-3064. http://www.climatesci.org/publications/pdf/R-222.pdf

Pielke Sr., R.A., G.E. Liston, W.L. Chapman, and D.A. Robinson, 2004: Actual and insolation-weighted Northern Hemisphere snow cover and sea ice -- 1974-2002. Climate Dynamics, 22, 591-595 DOI10.1007/s00382-004-0401-5.

Pielke, R.A., 2001: Earth system modeling -- An integrated assessment tool for environmental studies. In: Present and Future of Modeling Global Environmental Change: Toward Integrated Modeling, T. Matsuno and H. Kida, Eds., Terra Scientific Publishing Company, Tokyo, Japan, 311-337.

We have another paper in process on arctic and anarctic sea ice and will be blogging on that when ready.

Please post my reply if you would like.


Hank Roberts said...

> look at the data
Global average image, for anyone who didn't find it, is tucked away at the very bottom of the page:

Alastair said...

Pielke Snr.'s comments come as no surprise to me, as I have long regarded him as a climate sceptic.

Wikipedia writes:

Roger A. Pielke claims melting Arctic sea ice is a result of regional warming and not global warming.[139]
“ However, in terms of relating to the global average lower tropospheric temperature changes, in June 2007 (which is the latest data posted), the global average anomaly is +0.22 after being as high recently as +0.51C in January. Thus, it is regional warming, not “global warming” that appears to be the reason for this melting (Indeed, if it were global warming, we should see a similar reduction in Antarctic sea ice coverage, which, however, is not occurring (see[140] and see[141]).

See: http://en.wikipedia.org/wiki/Global_warming_controversy


Cheers, Alastair.

Anonymous said...

The Antarctic trend is not statistically significant, the Arctic trend is. Talking about the "net global average" is not meanignful at all since variability in the seasonal and interannual variability in sea ice extent in the Antarctic is much larger.

Robert Grumbine said...

Thanks Hank and William. I did find the address later. But, as you noted, it's a few levels down.

Hank: good comment. Readers: do take a look at Hank's comment over at the Mother Jones article.

Alastair: Labels like 'denialist' aren't any help. If his science is bad, talk about the science. Since you do start to that, please pursue it. Certainly any bit of ice that melts did so due to the local temperatures. That much is a safe side of Pielke's statement that you quote. On the other hand, we'd want more than just that triviality before making conclusions about the climate system. What would be some, and did he exclude those possibilities or carefully analyze them?

CoRev: (not posted due to absence of science content, and some other points) You complain about my 'abusive' emails (presumably posts as well; I'm no more polite in email than in my posts). Yet Roger wrote his comment after reading my post (I'd emailed it to him), replied as you see -- no complaints of abuse. I sent a response email to him, in which I said:

Thank you for the reply. I have posted it to my blog, and my comments will follow later.

Something I'm hoping to communicate to my readers, which I'm hoping to include jr. high and high school students, is how it is scientists can and do disagree -- without it being the food fight we see all the time on political shows and blogs.

Unlike your interpretation/presumption of 'abuse', Roger replied thanking me for engaging in constructive dialog about the science. As I said in my email to him -- scientists can and do disagree without it being a matter of a food fight.

Robert Grumbine said...

Now to Pielke's response...

Chris Colose grasped an important problem -- that you can't just average Arctic and Antarctic sea ice together and conclude 'normal'. One of the stories on this is two statisticians who went to the rifle range. One hit 1 foot to the left, and the other hit 1 foot to the right. They promptly congratulated each other because 'on average' they'd hit two bullseyes.

Chris: do you have a citation on the significance or lack thereof in Antarctic sea ice area/extent?

In terms of their effects on the climate system, Arctic and Antarctic ice don't compensate for each other. If it were the Great Lakes, on the other hand, and Lake Superior had lighter ice cover than usual and Lake Michigan was heavier, such that the area differences canceled, then I'd accept this sort of argument. Even, perhaps, also for balancing off area excursions in the Bering Sea (near Alaska) against those in the Barents and Kara Seas (near Europe). Arctic vs. Antarctic, no.

Another side is my recurring issue of weather vs. climate. For weather, we make statements today that can be false if shifted slightly in time. Today can be colder than normal. A week from now, it could be warmer than normal. For weather, we don't expect statements to remain true for very long. For climate, on the other hand, they should hold for long enough to get the article published. As of last March-April, Arctic ice was below climatology, as it had been continuously for 5 years by that time. It was only 0.5-1 standard deviations (eyeball) below climatology, as opposed to 0-4, but it was still below. A few days of not as much colder than usual weather does not mean the cold spell is over. Nor do a few weeks of not as much below normal ice cover mean that the 5 year below normal state is reversed. At this point, we need something like 5-6 years of average to well above average ice cover to say that extent and area are -- as a climate matter -- back to normal.

As erratically as the Antarctic ice cover is varying, surely you want more than a few weeks to make any statement about it's climatic state! Every year is showing both above and below normal.

No argument that using a linear trend is a bad idea for a complex system. But in the things you point to, and that I did, linear trends weren't being used. We've all (you, me, Bill Chapman) been working with deviations from climatology.

Perhaps it's just a matter of taste, but in talking about someone's 'area', I don't include every topic they've ever coauthored a paper in. If so, then I will soon be a librarian as I'm coauthoring a note (for their professional literature) with our librarian on institutional repositories. In like vein, I doubt David Robinson would call himself anything other than a 'snow guy', even though he was another of your coauthors on one of those papers.

digress: I really liked the insolation weighting. Partly because it was an idea I'd had myself. But more because it turned out as interesting as I'd hoped.

Anonymous said...

First, the linear trend fit to SH sea ice extent is significant, but just barely. Also the trend rate from linear regression since the beginning of satellite observations is small, only 12,000 km^2/yr, while the trend rate for NH sea ice extent is considerably larger at 51,000 km^2/yr, and yes, the difference is undeniably statistically significant.

Linear trends are certainly not the be-all and end-all of analysis, but neither are they irrelevant. In fact they give one of the best indicators of whether the overall trend (whatever its form) is up or down, and if the result is statistically significant we're able to conclude, with confidence, that the data are *not* just random noise. Furthermore, for the SH the data are indistinguishable from a linear trend plus random noise, so a linear trend rate is absolutely relevant.

For the NH, the trend is *not* linear; the residuals from a linear fit show a distinct departure from random noise. Both a quadratic fit, and a comparison of linear fits to data before and after 2001.5, demonstrate that not only is the NH trend upward and much *faster* than the SH trend, the NH trend is greater *now* than it was before 2001.5. The estimated trend rate since 2001.5 is 134,000 km^2/yr, more than *ten times* the long-term trend indicated for the SH.

Finally, the *global* sea ice extent shows an undeniable statistically significant decline; the estimated trend is a decrease of 39,000 km^2/yr. The fact that it also shows large fluctuations which can bring the *momentary* level above climatology, would only be posited as evidence of constancy by someone who is ignorant of statistical analysis, or is being dishonest.

Anonymous said...

I suppose I stand corrected then. I do remember reading though

//"In contrast,
within the past two decades sea ice extent
and temperatures in the Antarctic have not
been unusual in any season, except along
the Antarctic Peninsula, which experienced
the largest positive temperature anomalies
of anywhere in the Southern Hemisphere,"//

From http://www.cru.uea.ac.uk/~nathan/pdf/5-6-08_Overland.pdf

Just eyeballing http://nsidc.org/data/seaice_index/s_plot.html I can't see a big deal, but I'm ignorant in statistical analysis. I'm an undergrad so I promise I'll be getting better in that regard in the near future, I swear.

Robert Grumbine said...

Chris, it's reasonable, if not always correct, to interpret the figure you linked to as not showing a signficant trend. Rule of thumbishly, we'll call insignificant a trend which is not larger than the standard deviation. The rule of thumb doesn't really work all that well, but it's a starting point you don't need a statistics course for. As the series gets long enough, you can get to a significant trend even for a case like this. As Tamino said, it's significant, but only marginally (or at least that's my recollection and I'm having a hard time remembering which paper I read it in, plus I've done my own trend calculations).

You might also have been caught by the 2 decades part of the comment. We've got 3 decades of data. No doubt Overland had good reason to focus on 2 decades. For the trend, though, not significant for 2 decades can become significant at 3 decades.

Tamino: Was this your calculation, or do you have a cite?

As far as trends go, I agree the important part is to see that we don't simply have a random series of numbers. The Arctic sea ice extent and area is not just going up and down with no particular tendency. There is a clear difference between the last 5 years and any 5 year span of the previous 25 years.

On the other hand, as you illustrate, running a single straight line through the entire period also doesn't represent what's going on very well. If we consider the trend to have started at or before the start of the record, then the last several years show a different, much more drastic, trend.

There's no particular physical reason to think there's a linear trend for any period. The bad news is that there is reason to expect an exponential trend, and reason to expect a series of steps. I say bad news because if it were the simple linear trend, then the changes would be slow and steady. Accelerating collapse, or series of stepwise drops, on the other hand, can get rid of the ice pack much sooner. This'll be a blog post either later this week, or very late next week.

Hank Roberts said...

Good reading anticipating what happened the last 2 years (I've been hoping to find something newer):

Jan 12, 2007 ... Commentary by Cecilia Bitz, U. of Washington

Anonymous said...

I did the calculation myself, correcting for autocorrelation. A note: the trend for the SH is (slightly) positive, but the trends for the NH are distinctly negative.

I wouldn't trust the "significant when the trend exceeds one standard deviation" rule. That rule of thumb really doesn't work all that well.