[Correction appended]

There’s been a wrinkle in the global warming fact-checking saga I’ve been following this week.

Just to recap–George Will wrote a column claiming that global warming’s a lot of hype. He made a number of misleading statements, including one that was rejected by the very scientists he claimed as his source.

Will stated, “According to the University of Illinois’ Arctic Climate Research Center, global sea ice levels now equal those of 1979.”

A statement was then posted on the research center’s web site of the Polar Research Group:

We do not know where George Will is getting his information, but our data shows that on February 15, 1979, global sea ice area was 16.79 million sq. km and on February 15, 2009, global sea ice area was 15.45 million sq. km. Therefore, global sea ice levels are 1.34 million sq. km less in February 2009 than in February 1979. This decrease in sea ice area is roughly equal to the area of Texas, California, and Oklahoma combined.

A number of bloggers laid out the problems with the column and sought a response from the Washington Post. The Post announced that they had fact-checked Will’s column, and that it was just fine. I explained why that looks like some mighty poor fact-checking.

Last night in the comment thread, Doug drew my attention to an article on the ice record maintained by the US National Snow and Ice Data Center in Colorado (a different research center). On February 18 (three days after Will’s column appeared), the NSIDC announced that there was a glitch in the satellite sensors measuring ice in the Arctic, and so their record was gradually drifting off. The drift started in January, and gradually increased until they caught it in mid-February. The scientists now say that the latest estimates were off by 500,000 kilometers. They’re working now to compensate for the drift and correct the measurements. Here’s a graph from their web site.

The blue line marks the ice measurements taken by SSMI, the satellite NSIDC uses for the 30-year record of ice extent. The dashed red line is data from AMSR-E, a new  satellite that has also been measuring the ice and has remained accurate. The reason the scientists don’t switch over to the new AMSR-E satellite is that jumping from one data set to another can create the illusion of change that isn’t really there. But AMSR-E is still useful to the researchers, because they can compare its measurements to the ones they get using SSMI satellite to see if everything’s okay.

Some commenters wondered whether this development would cause me to take back my criticism. Let’s just set aside the fact that this news came out after Will had published his column, and thus could not have any real bearing on whether he or the Post bothered to contact the scientists that they cited as their source.

After looking at some of the web sites involved, I thought I ought to get in touch with the scientists who run the University of Illinois’ Arctic Climate Research Center–the center at the Polar Research Group–the scientists on whom Will depended for his claim, and which rejected that claim.

I got a prompt response from Bill Chapman, a University of Illinois climate scientist:

“It’s refreshing to have someone ask about the data before they write about it.”

Just pause to consider that. After all this kerfuffule–involving a nationally syndicated columnist, the assistants to that columnist, the editors at the columnist’s syndication service, the editors at the Washington Post editorial page, and the Post’s ombudsman–Chapman was refreshed that someone bothered to contact him about his research before writing about it. What a concept. For me, this whole affair has been about the value of fact-checking science, and Chapman’s reply shows just how little checking was carried out by the Post and company.

In his reply to me, Chapman explained that the two research centers, NSIDC and ACRC, both use the SSMI satellite readings, but they have different methods for building their time series. Chapman and his colleagues at ACRC use a composite of three sequential days for their ice cover readings. If a swath of data is missing on one particular day, they can go back to the previous day’s concentrations. If there are still missing regions, they go one more day back.

“Missing regions or swaths of data have always occurred from time to time in the SSMI record, which is why we set it up this way,” Champan explained.

Despite the recent trouble with the SSMI satellite, Chapman said the three-day-composites have still been meaningful. “As one check, we have been comparing our time series with those from the independent data source AMSR-E. They are just about identical so we are comfortable that our time series remain solid. Our time series and therefore the statement are unaffected by the recent satellite problems. If the sensor degrades a lot more, our numbers will be affected, but to date, they are not.”

I then asked what he thought about the Washington Post’s support of Will’s claim about ice. (To recap again, their support was decidedly roundabout. A January 1 post on a blog called Daily Tech claimed that global ice cover in late 2008 were unchanged from 1979. In response to that blog post, the Center posted a pdf on their web site explaining that “observed global sea ice area, defined here as a sum of N. Hemisphere and S. Hemisphere sea ice areas, is near or slightly lower than those observed in late 1979.” But then the scientists also explained that climate models predict a decline in Arctic ice, but are less certain about Antarctica, with some even suggesting an increase–making measurements of global sea ice not terribly relevant to the question of climate change. The Post ignored that part.)

Here’s Chapman’s reply:

Since their statements were based on the end of the previous year, and more importantly the end of 1979, the statement ‘global sea ice levels now equal those of 1979′ just didn’t make sense any more. We have received 80-100 emails from confused people who had read George’s column and looked up the graphs on the Cryosphere Today [one of the center's web pages] and said they came to a different conclusion, or, could we point them to the report that said that Feb 1979 and Feb 2009 sea ice area was nearly the same. We had to post the current and corresponding 1979 values to avoid the inconsistency that readers were noting. After doing some googling, it appears that Daily Tech article got repeated on a lot of blogs, so it’s not surprising George Will came across it at some point. Still it was sloppy for them to not double check with the original source and it really points out the danger of making any conclusions on climate change based on any two days in history. I really wish they would have contacted us at some point to avoid this.

Our goal is to present the data in as concise and useful format as possible for interested users. Whether the Washington Post decides to publish a correction is up to them.

Indeed.

Finally, just to illustrate what Chapman’s talking about when he refers to the danger of picking out just two days in history, I thought I’d also include two graphs from Cryosphere Today. The top one shows the extent of Arctic sea ice, as compared to the 1978 to 2000 average. The bottom one is from Antarctica. A number of researchers have found a downward trend in the Arctic ice in recent decades, while there’s a small upward trend in the ice around Antarctica.

Correction: I erroneously called the University of Illinois Polar Research Group the University of Illinois Arctic Climate Research Center. The latter, used by George Will, is a fabrication. Details here.

While recovering from an extracted wisdom tooth this morning, I cheered up when I saw that Talking Points Memo and other blogs have picked up my grousing about George Will’s error-laden global warming column in the Washington Post. When I first became aware of Will’s column on Monday, it seemed to me the perfect example of the general problem with treating op-ed pages as “opinion.” That is, if by opinion, you mean that someone doesn’t have to adhere to the facts. I could state that the Earth is 6000 years old, and no one would dare correct me, because it’s just my opinion. (I guess that’s the rationale that led Forbes and US News to run pieces by young-Earth creationists as “commentary” a couple weeks ago in “honor” of Darwin’s birthday. [Okay. No more air quotes. Promise.])

Now we learn via Andrew Alexander, the Washington Post‘s ombudsman, that the editorial page has a whole team of fact-checkers. Or at least there are personal assistants to George Will, a couple syndication editors, and Post copy editors who have been identified as fact-checkers. Somehow, this army all decided that Will’s piece was just dandy. Even weirder was the post-modern refusal to run a correction from Alan Shearer, the Washington Post Writers Group editorial director: “We have plenty of references that support what George wrote, and we have others that dispute that. So we didn’t have enough to send in a correction.”

It seems as if the Washington Post just doesn’t think this is important. Via Jay Rosen I learned that Alexander’s inaugural ombudsman column today has nary a mention of the affair–even though Alexander himself made inquiries. Maybe Alexander just wanted to say “Hello, World,” in his first piece, without diving straight into any particular complaints. That’s fine. Let’s see what he writes about once the niceties are out of the way. (He invites email: ombudsman@washpost.com )

My own opinion is that this was a serious screw-up, but not an easy one to solve in any systemic way. In an ideal world, editorial pages would employ full-time fact-checkers who felt no fear in pointing out small and large errors of fact. Only after their objections had been satisfied would a column see the light of day. That’s what happens to articles at some magazines today.

In the real world, though, a lot of magazines don’t have fact-checkers on staff, and they expect writers to do the fact-checking themselves. It’s particularly tough for newspapers, which churn out so many stories a day. To fact-check those stories well, they’d have to hire back a fair amount of the people they’ve laid off in recent years. I assume the same probably goes for editorial pages, although I can’t say for sure, never having dealt with them myself.

Still, it remains seriously weird for a national newspaper to run a piece that they claim has been thoroughly fact-checked, which has since been showed to be plainly flawed. It’s also weird for it to then refuse to run a correction based on a bogus sense of balance about the evidence of how much ice there is in the world and what that means for climate change.

A lot of people have left comments here complaining about George Will. And others have then accused them (and me) of being part of a left-wing conspiracy, attacking Will while letting the inaccuracies of others slide by. For me this is not really about Will. It’s about how newspapers and magazines succeed or fail to convey science as accurately as possible. And this case is a textbook example of failure. I hope something is learned from it.

[Update, 2/22: I've added a new post addressing some confusion over some late-breaking news about the satellites that measure ice. And along the way, we are reminded of just how weak the multi-layered fact-checking at the Washington Post editorial page is.]

[Correction appended]

Monday I bemoaned the lack of fact-checking of opinion pieces in newspapers, pointing to a George Will column on global warming in the Washington Post as evidence. Now the Washington Post op-ed folks claim that it was in fact heavily fact-checked. All I can say is that none of them better apply for a fact-checking job here at Discover.

To recap: George Will wrote a column in which he tried to downplay the evidence that global warming has already affected the Earth, and that it will have bigger impacts in the future. Various bloggers have pointed out examples where Will misrepresented scientific studies in this column. The most glaring one was this: “According to the University of Illinois’ Arctic Climate Research Center, global sea ice levels now equal those of 1979.”

The Research Center put a statement on their site explaining that Will was wrong. On February 15, the day Will wrote his column, there was substantially less ice than on February 15, 1979: the area of Texas, California, and Oklahoma combined.

I picked up this story from Talking Points Memo, and it has been bouncing around for a few days now. The folks at TPM and elsewhere have been trying to get a response from the Post about why they haven’t posted a correction. Today, Wonk Room appears to have finally broken through. And, oh, what a response they got. It’s worth quoting at length, because it reveals some intricately baffling behavior:

When contacted by the Wonk Room, the Washington Post’s ombudsman, veteran reporter Andy Alexander, “sought clarification from the editorial page editors”:

Basically, I was told that the Post has a multi-layer editing process and checks facts to the fullest extent possible. In this instance, George Will’s column was checked by people he personally employs, as well as two editors at the Washington Post Writers Group, which syndicates Will; our op-ed page editor; and two copy editors.

Wow. I’d hate to see what Will’s columns look like before the “multi-layer editing process.”

Full email from Andy Alexander (ombudsman@washpost.com):

Dear Mr. Johnson,

Thank you for your e-mail. The Post’s ombudsman typically deals with issues involving the news pages. But I understand the point you and many e-mailers are making, and for that reason I sought clarification from the editorial page editors. Basically, I was told that the Post has a multi-layer editing process and checks facts to the fullest extent possible. In this instance, George Will’s column was checked by people he personally employs, as well as two editors at the Washington Post Writers Group, which syndicates Will; our op-ed page editor; and two copy editors. The University of Illinois center that Will cited has now said it doesn’t agree with his conclusion, but earlier this year it put out a statement (http://arctic.atmos.uiuc.edu/cryosphere/global.sea.ice.area.pdf) that was among several sources for this column and that notes in part that “Observed global sea ice area, defined here as a sum of N. Hemisphere and S. Hemisphere sea ice areas, is near or slightly lower than those observed in late 1979,”

Best wishes,
Andy Alexander
Washington Post Ombudsman

Update: Alan Shearer, the Washington Post Writers Group editorial director, told the Wonk Room that he looked into the accuracy of Will’s claim that “According to the University of Illinois’ Arctic Climate Research Center, global sea ice levels now equal those of 1979″:

We have plenty of references that support what George wrote, and we have others that dispute that. So we didn’t have enough to send in a correction.

There’s a lot of wiggly, lawyerly language here. What does it mean for the editors to check facts “to the fullest extent possible”? As I mentioned in my last post, magazine like Discover and the New Yorker assign a person to check every point in an article. It can become the fact-checker’s Moby Dick. The fact-checker doesn’t rely on press releases or blog posts, but calls scientists up to get the best information.

Did the veritable army of fact-checkers described by the Post fact-check to this degree? We can safely assume the answer was no, because the researchers at the Arctic Climate Research Center were baffled by Will’s claim about the ice. “We don’t know where he is getting his information from,” they wrote in their statement.

If someone from the Post’s crackerjack multi-layer squad of fact-checkers had bothered to pick up the phone, they could have simply asked, “Is it indeed true that global sea ice levels now equal those of 1979?”

And they would have probably gotten an answer like this: “Well, what do you mean by now? Today? And what do you mean by 1979? Exactly thirty years ago today? If that’s what you mean, the answer is no.”

A good fact-checker would then say, “Well, it seems this claim is based on an article that came out January 1.”

To which the scientist would say something along the lines of, “At that point it was near or slightly lower what was observed in late 1979.”

At the very least, that discrepancy would have to be corrected. But a good fact-checker would see a deeper problem, saying, “Whoa, that changed a lot in a month and a half.”

Which would then lead to a discussion of the fact ice cover is such a noisy process that picking out a single day to compare these numbers does not say a lot about how it is affected by climate change. Climatologists look over longer time scales.

A good fact-checker would also learn that almost all climate models project that increasing greenhouse gases will cause a decrease in the Northern Hemisphere sea ice area over the next several decades, but the response of the southern hemisphere is less certain. In fact, evaporation caused by the warming might lead to more snowfall onto the sea ice. If the southern ice expands, it cancels out some of the retreat of the northern ice. And lo and behold, the northern hemisphere ice is almost a million square kilometers smaller than it was in late 1979, and the Southern Hemisphere ice is about half a million square kilometers bigger than in late 1979. So not only is Will wrong on the particulars of his statement, but he’s wrong on what it means about climate change. A good fact-checker would make sure that this was fixed too.

How can I be so confident that a good fact-checker would learn this? Because it is in that same January statement from the Center that the Post cited as “evidence” that Will was correct.

If the Post’s fact-checkers actually looked at the statement before they published Will’s column, they could not have seen the sentence about sea ice coverage without seeing the broader discussion of what climate change does to sea ice as well. And yet, even if they did see it, it did not cause them to make Will change his column.

If that’s indeed what happened, it would be bad fact-checking. But it’s also possible that they only looked at the January statement after this kerfuffle broke out this week, and picked out one line that seems to justify Will’s false statement–even though it was nestled in the discussion of the differences between the two hemispheres. That’s not fact-checking at all. It smacks of quote-mining.

It’s easy to think of fact-checking as a luxury of old-time journalism, akin to three-martini lunches and business class flights. But if fact-checking is done right, it can make newspapers and magazines reliable and trusted–a distinction that may help them survive in these competitive times. Sadly, in this case, we see what happens when the process fails.

[Update, 2/22: I've added a new post addressing some confusion over some late-breaking news about the satellites that measure ice. And along the way, we are reminded of just how weak the multi-layered fact-checking at the Washington Post editorial page is.]

[Correction, 4/7: It turns out that there is no such thing as the Arctic Climate Research Center at the University of Illinois. That is a fabricated name. I should have referred to the Polar Research Group. Details here.]

I had to spend a couple hours this morning footnoting my next column for Discover so that it can be fact-checked. I had to assemble the papers I read, the web sites I visited, and the contact information for scientists who helped me understand the subject. One of Discover‘s intrepid fact-checkers will then spend many hours following in my footsteps and discovering where I tripped. He or she will have no compunction about writing up a detailed report of my mistakes. I’m sure some mistakes will turn up, and I won’t be angry to see them in a fact-checking report. I’ll be grateful that my column won’t inadvertently misrepresent someone’s research. And I’ll be personally glad to have any misunderstanding of mine put straight.

Fact-checking is an underappreciated art. As John McPhee explains in his recent essay, even a legendary fact-monger like himself benefits from the relentless skepticism of the New Yorker‘s fact-checkers. I got my own start as a fact-checker, and it was the best training I could imagine for science writing. Even when I don’t have the luxury of fact-checkers vetting my own writing, I have an inner skeptic that drives me to double-check whatever I feel unsure about. One of the things I like most about blogging is the after-the-fact fact-checking that comes from commenters who catch mistakes. While accuracy is important for any kind of factual writing, it’s particularly important in science writing, because there are just so many ways to get a story wrong. (Trust me, I know.)

This old fact-checker gets rankled fairly often when I read about science on op-ed pages, because there doesn’t seem to be much fact-checking going on there. It’s not necessary for anyone to fact-check subjective statements like, “We must unleash the inventive genius of America,” or “Harry Potter is boring.” But imagine that someone writing a column about global warming (or the supposed lack thereof) wrote the following:

According to the University of Illinois’ Arctic Climate Research Center, global sea ice levels now equal those of 1979.

Surely somebody should actually check with the people at the center to make sure that statement is true. If not the writer, then somebody whose job it is to make sure that paid columnists do not just make stuff up. Wouldn’t the columns be better? Wouldn’t the publication be better for such standards?

Unfortunately, this is not just an imaginary scenario. George Will delivered this bit of information in his most recent column. And, as Talking Points Memo reports, the research center felt the need to post this statement on their own web site:

In an opinion piece by George Will published on February 15, 2009 in the Washington Post, George Will states “According to the University of Illinois’ Arctic Climate Research Center, global sea ice levels now equal those of 1979.”

We do not know where George Will is getting his information, but our data shows that on February 15, 1979, global sea ice area was 16.79 million sq. km and on February 15, 2009, global sea ice area was 15.45 million sq. km. Therefore, global sea ice levels are 1.34 million sq. km less in February 2009 than in February 1979. This decrease in sea ice area is roughly equal to the area of Texas, California, and Oklahoma combined.

It is disturbing that the Washington Post would publish such information without first checking the facts.

This is not a matter of the complex choices between cap-and-trade, carbon taxes, or other responses to global warming. This is a matter of exquisitely simple facts. 16.79 does not equal 15.45.

Of course, this glaring error helps George Will make his case that global warming is nothing to worry about. But it is not true, and two seconds of fact-checking by the Post could have discovered that. It’s fine for op-eds to be a place for opinions. But that doesn’t mean they should be a vacation from facts.

Update, 4/7: In fact, there is no such thing as the University of Illinois Arctic Climate Research Center. Details here.