Tuesday, May 12, 2009

Global Sea-ice, Deceptive Reporting, and Truthful Lies

"The latest Arctic sea ice data from NASA and the National Snow and Ice Data Center show that the decade-long trend of shrinking sea ice cover is continuing."

That statement, from the JPL, is dated April 2009. The actual data for northern hemisphere sea ice, measured as the deviation from its 1978-2000 mean, are shown below. The source is "The Cryosphere Today," a web site of the Polar Research Group, Department of Atmosphere Sciences, University of Illinois at Urbana-Champaign, not a site devoted to critics of global warming. If your browser doesn't show the complete graph, click on the image to get to the webbed original.

Looking at the graph, the pattern is pretty clear. For about ten years, from 1997 to late 2007, the area of sea ice was decreasing. That trend then reversed, and the area has now been increasing for more than a year. The claim quoted from the JPL is a flat lie.

Reading further in the article, one finds: "that this winter had the fifth lowest maximum ice extent on record. The six lowest maximum events since satellite monitoring began in 1979 have all occurred in the past six years (2004-2009)."

That's a nice example of how to lie while telling the truth. The trend has reversed—whether temporarily or permanently we don't know. But since the area had been trending down for a decade and has only recently started to trend up, the current figure is still low relative to the past.

Except for the recent past, which is all that is relevant in judging whether the current trend is continuing.




10 Comments:

At 11:56 PM, May 12, 2009, Blogger Steve said...

"The claim quoted from the JPL is a flat lie"

Um, no. Just no. If you look at this graph, also from the Polar Research Group, you see the same downward trend without the amount of noise in the graph you cherry-picked.

 
At 12:50 AM, May 13, 2009, Blogger David Friedman said...

Steve offers a different graph from the same source as mine--for some reason he thinks mine rather than his was "cherry picked." He doesn't explain, nor do I know, the differences.

Since his shows much wider seasonal swings than mine, my guess is that my graph, which is showing the anomaly from the 1978-2000 mean, is seasonally adjusted--it shows how much larger or smaller the area was in (say) January of 2007 than in the average January over the reference period.

If so, it's his graph that's noisy, since the enormous variation with the season makes year to year changes harder to see.

But in any case, his graph doesn't support the quote. It shows a decline ending in late 2007, although the trend after that seems to be flat or slightly rising. The claim was that the recent data showed the decline continuing.

 
At 4:12 AM, May 13, 2009, Blogger Tim Lambert said...

A declining trend is not the same as monotonically decreasing. It is wrong to take any uptick in the series as he end of the trend.

The graph you posted is not very useful for judging trends -- it is very noisy and because summer ice has been declining faster than winter ice there are now seasonal fluctuations in the anomaly.

The JPL press release seems to be referring to this graph of multiyear sea ice coverage. There is a very small uptick, but it is wrong to label this as the end of the trend.

The other place it refers to is the NSIDC report. This has a graph of sea ice extent for March. It's slightly above the long term trend line, but it's less than it was last year and more than it was in 2006. Should we say that the declining trend ended in 2006 and restarted in 2009? Surely not. There is a declining trend and tere is some variation around that trend line.

 
At 6:49 AM, May 13, 2009, Blogger William Newman said...

It's true that "It is wrong to take any uptick in the series as he end of the trend" as Tim Lambert intones. But if David Friedman argues that a particular piece of testimony in the OJ Simpson trial is a lie, it's not reasonable to try to refute him by pointing out that he has not proved Simpson is innocent.

Maybe one could try to rescue the JPL statement by combining a suitable definition of "the" trend and a suitable model for statistical properties of the background noise. Under suitable noise assumptions, after this small uptick it might well be statistically reasonable to increase confidence that an underlying nonzero trend exists (as opposed to being random fluctuations) while reducing one's estimate of its rate. But even if so, it'd misleading to call it "the" trend continuing. Most likely there's a different new best hypothesis, the shallower trend. If not, it's probably because the old hypothesis is devious enough that it's misleading to call it "the trend" as though it is a simple commonsense line. And independent of the innards of any statistical approach, eyeballing the graphs I am pretty sure that "the new data show...continuing" must always be an unreasonably strong claim. A year of new data on that graph just isn't going to "show" any conclusion unless the new data happen to falsify some model by lying unusually far away from what it predicts. (E.g., the new data do show that the Sun didn't go nova last year.) IMNSHO people who want to use this year's data to defend an ordinary trend should be making claims no stronger than "the new data support the hypothesis that..." and probably as weak as "the new data are consistent with the hypothesis that..."

So the JPL quote doesn't seem like a suitably honest summary for responsible advocacy where big money is at stake. Given comparable underlying data, I'd be pretty unhappy about seeing such a claim from an expert witness at a trial, or a stock prospectus, or a document describing the benefits and risks of a pharmaceutical.

Even by the usual standards of statistical political advocacy (e.g., on educational outcomes or minimum wage effects or the Laffer curve) I think this is pretty bad. It's not memorably bad compared to random talking head advocacy, but it's memorably bad for JPL. Seeing a data set as noisy as this and hearing someone summarize an uptick on it as "the new data show" that "the" politically charged downward trend is continuing, I mark down the credibility of that person, of that person's institution, and of any people that try to defend that person.

 
At 6:53 AM, May 13, 2009, Blogger VangelV said...

The AGW movement keeps cherry picking data. If we are talking about global climate than we look to look at global ice cover. And when we do, we find that it stands above the average, which means that there is no ice melting crisis.

 
At 9:36 AM, May 13, 2009, Blogger David Friedman said...

Tim writes:

"The JPL press release seems to be referring to this graph of multiyear sea ice coverage."

That, as you can see by the label, is a graph of "Multiyear sea ice coverage." The JPL article does indeed discuss that, but the opening sentence, which I quoted, is about "sea ice cover," which it's clear in the article is not the same thing.

"There is a very small uptick, but it is wrong to label this as the end of the trend."

We don't know if it will turn out to be the end of the trend. But if the latest data show an increase rather than a decrease, then to say that the latest data show that the decreasing trend is continuing is a lie. That would be true even if they were talking about your graph, and it is more strikingly true if you look at the graph which actually shows the figures for what they claim to be describing.

To put it differently, on your reading any value for the latest data, other than perhaps an enormous increase, counts as evidence that the trend is continuing. If the latest value is lower than the one before, that's evidence for a continuation of the trend, and if it's higher, that is just an "uptick" and still evidence for a continuation of the trend.

Is that really your view?

 
At 11:27 AM, May 13, 2009, Anonymous Anonymous said...

Am I the only one who finds it a bit silly to be refuting a **press release**?

Sure press releases are often wrong about a lot of things, but they really have little do with the actual science.

 
At 7:02 AM, May 14, 2009, Blogger VangelV said...

"That, as you can see by the label, is a graph of "Multiyear sea ice coverage." The JPL article does indeed discuss that, but the opening sentence, which I quoted, is about "sea ice cover," which it's clear in the article is not the same thing."

How do we know the extent of 'multiyear sea ice coverage?' There are currents and winds that come into play and ice is pushed around the arctic normally. How do the people doing the reporting know how old the ice is unless they do physical measurements and detailed analysis?

Before you answer let me remind you that the experts' estimate of Arctic ice thickness was about half of what was actually measured. How credible is the data if the people who provide it were off on their thickness estimates by 100% ?

 
At 9:02 AM, May 14, 2009, Blogger Tim Lambert said...

David, multi-year sea ice is sea-ice. Press releases necessarily simplify things.

Saying that there is a decreasing trend does not mean that every observation lies exactly on the trend line. It means that some will be a little above the trend line and some below. If you get one a little below the trend line and the next one a little above it, it is wrong to say that this slight uptick means that the trend has ended. In fact, because the new observation was close to the trend line, the trend has continued.

 
At 10:56 AM, May 14, 2009, Blogger David Friedman said...

Tim writes:

"David, multi-year sea ice is sea-ice. Press releases necessarily simplify things."

If you actually read the JPL statement, you can see that multi-year sea ice is a subset of sea ice. The release talks about both.

Mortality from cancer is mortality. But if you say mortality is increasing when it's decreasing, you don't get to justify it by showing that cancer mortality is increasing.

Tim writes:

"If you get one a little below the trend line and the next one a little above it, it is wrong to say that this slight uptick means that the trend has ended. In fact, because the new observation was close to the trend line, the trend has continued."

That's true, but it doesn't describe the data. The latest evidence, which is what the statement is about, isn't merely above the trend line--that would be (weak) evidence that the trend was decreasing. It's above the previous level. Tim can surely distinguish between increasing the first derivative of a function and increasing the value of the function, and he is treating the latter as if it were the former.

 

Post a Comment

Links to this post:

Create a Link

<< Home