17 Comments

Thanks, Zeke. That's not far from a straight line. Dial back on the "f" setting of the LOWESS and take another look.

Alternatively, post up a link to your winter minimum data (or just put the data in a comment here) and I'll do a proper CEEMD analysis.

Sorry for the harsh tone of my initial comment, totally unwarranted. My apologies.

w.

Expand full comment

Transpolar commercial navigation is feeling very optimistic seeing the summer decline.

Expand full comment

OK, I got bored waiting for Zeke to either post his data or a link to the source. So I digitized it. Here's the result of the linear analysis from 2007 on:

===

Residuals:

Min 1Q Median 3Q Max

-1.11505 -0.24525 0.06595 0.29785 0.58394

Coefficients:

Estimate Std. Error t value Pr(>|t|)

(Intercept) 17.889706 46.055950 0.388 0.703

time(tser) -0.006674 0.022856 -0.292 0.774

Residual standard error: 0.4617 on 15 degrees of freedom

Multiple R-squared: 0.005652, Adjusted R-squared: -0.06064

F-statistic: 0.08527 on 1 and 15 DF, p-value: 0.7743

===

Note that there is an extremely slight downward trend to the regression line (-0.006 MKm^2/year), but that it is ridiculously far from significant (p-value = 0.77).

So I was wrong that the trend might be "flat or slightly increasing" … however, this analysis clearly shows that there has been no significant trend in the minimum Arctic sea ice area for the last 17 years.

And that is what is hidden or obscured by Zeke simply drawing a straight line through the whole dataset.

Finally, I've uploaded an image containing the data and a CEEMD smooth of the data, to show what Zeke's straight line is hiding. It's here: https://imgur.com/57kVqaE

My best to all,

w.

My best to all,

w.

Expand full comment

This is quite a silly cherry pick. Why 17 years? why not 12 or 19 or 18.329654? Obviously, reducing the data to the single point of the annual minimum and truncating the horizon is a great way to inflate the variance and diminish the significance of ... well anything. More importantly, it's a great way to ignore the elephant in the room, the giant mode shift around 2007.

Reporting nonsense to 8 sig figs is just the icing on the cake.

Expand full comment

Tom, it is a PITA to round that block of data that comes from the computer. I have to go in and round each of the individual figures. What, you can't round the numbers in your head?

Don't like It? Don't care.

As to why I picked that time period, if you start from the end and work back, that's the longest period with no significant trend. It is the answer to the question "how long has there been no significant trend in the Arctic sea ice.

And as you can see in the graphic linked in my post, it's a very visible change in what up to that point had been regularly dropping minimum ice areas. I understand that you want to minimize that to keep your Arctic emergency alive, sorry that reality isn't cooperating.

w.

Expand full comment

"What's the longest period that makes my claim look good?" is a silly test. Having to throw out more than half the data to get there is, as you put it, "cherry picking of the highest order."

Setting that aside, your results disprove your own claim. You say that "this analysis clearly shows that there has been no significant trend" but in fact your confidence bounds on the time term are fairly wide, admitting significant trends.

The linear regression probably understates the uncertainty, due to autocorrelated disturbances and dynamics that are likely to have the system hopping from one locally stable state to another rather than progressing smoothly.

Expand full comment

"As to why I picked that time period, if you start from the end and work back, that's the longest period with no significant trend. "

That is pretty much an explicit admission of a "cherry pick". You are invalidating the test by performing that search (essentially you are performing multiple significance tests without making the proper adjustments).

Also without an assessment of statistical power you can't draw conclusions about the meaning of a lack of significance, it definitely doesn't mean there is no trend. If the period is too short you will get a non-significant result not because there isn't a trend but because there isn't enough data to "rule out" the absence of a trend. If you argue FOR the trend, that eliminates the self-skepticism that statistical tests are intended to encourage. This has been a very common statistical error on skeptic climate blogs, for an explanation, see https://skepticalscience.com/statisticalsignificance.html

Expand full comment

Thank you for your report. What are the impacts of having ice-free Arctic conditions, even if they occur only in the summer?

Expand full comment

I saw a segment that stressed how much less dense the ice was as well. So even if it's as large or nearly as large (surface area) the total mass wasn't the same. Any truth to that?

On a side note I just saw Sabine's video on the high ECS models. Something tells me the imminent apocalypse crowd is going to use that as gospel.

Expand full comment

Per your graph, winter minimum Arctic sea ice has been flat or slightly increasing since 2007, a period of seventeen years.

You putting a trend line to totally obscure that fact is cherry picking of the highest order. At least have the decency to use a LOWESS smooth or the equivalent.

Bad scientist, no cookies.

w.

Expand full comment
author

Here is lowess-smoothed winter sea ice. I don't see "flat or slightly increasing since 2007". Maybe since 2017 if you squint... https://i.imgur.com/pvgk5om.png

Expand full comment

Can anybody here explain why "lowess-smoothed" sea ice extent should in any sense be a "better fit" than "a straight line", given the physics of the sea ice annual melt/freeze cycle?

https://nsidc.org/learn/parts-cryosphere/sea-ice/science-sea-ice

Expand full comment

Jim, take a look at my CEEMD smooth of the Arctic sea ice extent. Nature rarely moves in straight lines. A smooth, whether Gaussian, LOWESS, or CEEMD, better represents how the ice area has changed.

https://imgur.com/57kVqaE

w.

Expand full comment

I saw that Willis, but it makes no reference to "the physics of the sea ice annual melt/freeze cycle". Hence my question!

See also this "statistical" analysis by "Tamino", which involves three straight lines instead of one:

https://GreatWhiteCon.info/2021/10/facts-about-the-arctic-in-october-2021/

"The annual averages show much less fluctuation than the annual minima, so we can estimate things like rates of change with greater precision. I find that there is statistical evidence that the rate changed over time. One model of such changes uses three straight-line segments with their changes chosen to best-fit the data, like this:"

https://GreatWhiteCon.info/wp-content/uploads/2021/10/Tamino-3segments.jpg

Expand full comment

Willis: "Per your graph, winter minimum Arctic sea ice has been flat or slightly increasing since 2007, a period of seventeen years." You made the claim, you back it up with domain-appropriate statistics. You must know your eyeball guesstimate has no scientific standing!

Expand full comment