This January we mentioned Tony Heller’s observation that Arctic sea ice was at its highest level for January 9 in 20 years, and that there has been no downward trend in the minimum since 2006. This comment caused a firestorm that bid fair to accomplish what nature and humans could not and turn the north pole into a sauna. An army of trolls rushed out onto the floes and someone even made an entire video on what fools we were based on their misunderstanding of a trivial and quickly-corrected error in one chart. But as we said at the time, what really matters is what the ice does. And as it turns the annual corner from its seasonal-melt low point, Susan Crockford observes that the latest NSIDC number “now makes 18 yrs of near-zero trend in Sept sea ice, which demolishes claim that more CO2 means less sea ice”. And it’s an important lesson about what does and does not constitute a trend.

It is true that there was at one point a downward line from 1979 that seemed quite dramatic. And because its starting point coincided, by coincidence, with the beginning of modern satellite records they said gotcha… but it wasn’t. In fact every year since 2012 has seen a higher minimum than 2012 and, as Crockford notes, a proper fitting of the line to the fluctuations shows no trend since 2006.

Or does it? The meaning of trend depends non-trivially on how much fluctuation there is. But before we get to that somewhat subtle point, especially by the standards of online polemics, we need to deal with a far cruder kind of misunderstanding about how you fit curves to data points… or at least how you should.

To take an extreme example, if Arctic ice had gone down by 50% from 1979 to 1980, then gone up one percent a year every year since, you could still claim that it was down by six percent for an average annual decline of 0.13 percent. But no sane person would do so.

Or would they? With the ice up considerably since 2012, and with 2024 now around the mid-point of the last five years of this rebound, we still get people haranguing us that a flatline since 2006 is just natural fluctuation but the decrease from 1979 to 2006 is an irreversible and wickedly ominous trend. No really, we do.

We also get people flatly denying that there were meaningful fluctuations prior to 1979. They deny that there was a low point in the 1940s and a rebound to 1979, that there was a minor cyclical low around 1900, or that there was a major cyclical low in the Medieval Warm Period then a massive rebound in the Little Ice Age. That whole thing with the Viking voyages and farming Greenland, then the settlements perishing from the cold, is to them not even noise. It just didn’t happen or if it did it was so marginal and minimally regional that it didn’t even reach into the Arctic. (Where precisely they think Greenland is located is a bit of a mystery.)

No, really. One tireless if tiresome polemicist commented on that troll-magnet video that:

“Reconstructed changes in Arctic sea ice over the past 1,450 years
Christophe Kinnard, Christian M. Zdanowicz, David A. Fisher, Elisabeth Isaksson, Anne de Vernal & Lonnie G. Thompson
Nature volume 479, pages509–512 (2011)
Figure 3 shows that arctic sea ice area had been stable at about 10 million square km from about 600 CE until mid 20th century when there is a rapid drop to less than 8 million square km today.”

Got it? Fourteen hundred years of stability then a plunge. QED.

Actually it shows nothing of the sort, partly because their own chart shows a lot of fluctuation even within the 20^th century, partly because the paper says up-front that “extensive uncertainties remain, especially before the sixteenth century” and especially because of all the dotted and dashed lines where the trend suddenly leaps sideways right where they staple modern observational data onto the proxy stuff, whereas when you mix different data sets, using proxies then switching to satellites, and get a massive discontinuity right at the divide, you should be highly skeptical not dogmatic.

That same person (we said they were tireless) also cited a paper that spoke of dramatic past fluctuations over millions of years then said recent changes observed directly seem unprecedented in the last few thousand years of proxy data (which wouldn’t have captured them had they happened), all while warning darkly that:

“the Arctic Ocean may become seasonally ice-free by the year 2040 or even earlier”.

Well, it better hurry, hadn’t it, because that paper was published in 2010 when 2040 was 30 years away, and we’re now half-way there and the ice is increasing not decreasing. Blips aren’t trends and cherry-picking isn’t good statistical analysis.

So what does it all mean? We repeat, yet again, that it’s hard to say. It’s not a “gotcha” for us either because as Roger Pielke Jr. recently noted, from a serious statistical point of view if a phenomenon varies considerably at random it can take centuries for the signal of a trend to separate itself from the noise.

To take another extreme example, if the height of Mount Everest were to change every year for the next 25 you’d be justified in thinking something important was happening because it is not prone to jiggle up and down. But if you got a lot of Atlantic hurricanes in the next quarter-century you’d be justified in waiting to see because the decadal record is all over the place in terms of number and severity as far back as we have careful observations, meaning you could deduce nothing from two straight decades of a lot of bad ones or of virtually none. A rule many alarmists violated by insisting, when no hurricane stronger than Category 3 hit the US from 2005 through 2016 and they said aha, warming increases wind shear so hurricanes get weaker only to have two Cat 4s hit in 2017 and a Cat 5 in 2018. It was just random.

The same is true of year-on-year fluctuations of Arctic ice and, very possibly, decade-on-decade or even quarter-century-on-quarter-century ones. And only people whose statistical understanding foundered around Grade 10, or who are fanatical, claim otherwise. In large numbers.

P.S. Another confounding feature of climate debates is the number of people putting forward dogmatic explanations of why something that isn’t happening must be. Including an item last month from Arctic News (actually a wildly alarmist site not a news outlet, which recently warned that worldwide “the temperature rise may exceed 18°C from pre-industrial by 2026”) about heat pouring inexorably into the Arctic and blasting the ice while releasing methane in a runaway disaster that, among its other flaws, assumes a linear relationship between potential and actual atmospheric water vapour that appears empirically unsound. It’s those dang facts again.