Thursday, July 26, 2012

Loss of Arctic sea ice '70% man-made'

Inevitably, this news article is illustrated with a poor polar bear standing on an iceberg. But the study wasn't really anything to do with polar bears:
The radical decline in sea ice around the Arctic is at least 70% due to human-induced climate change, according to a new study, and may even be up to 95% down to humans – rather higher than scientists had previously thought.
You wouldn't have guessed it from the write up, but this is work that Jonny Day did last year while he was working with us here at RIGC (he did revise it at Reading University, so it's entirely reasonable that they get some credit too). Hopefully our next recruit (advert) will be similarly successful! Jonny was only here for about 8 months (filling in at the end of the last 5 year project) so a high-profile paper in ERL seems like a pretty good return. The paper investigated how much of the decline in sea ice was plausibly due to various natural phenomena like the Atlantic Multidecadal Oscillation and Arctic Oscillation. The answer...maybe a little, but not a lot.

It was nearly published in GRL some months ago, but after several increasingly positive review cycles (by which time all the original reviewers had recommended acceptance), the editor then found yet another new one (by this time we were on reviewer #6!!) who asked for yet more analyses, at which point it was rejected...and a few days later, another paper on the same topic appeared in that journal.

It doesn't seem to be up on line quite yet but is probably due later today. Rumour has it that Jonny is also making his screen début - I'll be sure to put a link up when it appears :-)

Ok, it's up now. Enjoy!

11 comments:

Carrick said...

I noticed that the Guardian article had the usual claim of "accelerating rate of ice loss".

My question, is it really accelerating?

I did an analysis here on rate of ice loss. (Linear trend fit using a 10-year window.)

Several interesting things, ice loss seems to be max during La Niña (a connection between ENSO and the polar vortex is a studied problem, I don't claim to fully be up to date on that).

Really there was a big ice loss centered around 2004, I'm not sure how that's attributable to humans directly.

I suppose you could argue there is an acceleration because the baseline the variable part is fluctuating around has a positive slope (secular increase in rate of ice loss=acceleration), but it appears visually to be a slight effect.

I'd have to do some fitting to the various components and look at how much ice loss is attributable to the various signal components.

I'd suspect the biggest influence was the ENSO machine, but whether there's a human impact on that (yet) seems to me to be tenuous, even if I agree in time it seems inevitable that increased equatorial diabatic heating due to anthropogenic forcings will lead to a strengthened ENSO.

Oh yeah, this data is a splice of JAXA and the older GFSC data sets. My (robot daily updated) version can be found here.

James Annan said...

It does look a bit like there is a positive slope in your graph, I'm afraid :-)

William Connolley said...

> I did an analysis here

That's not an analysis, that's a graph.

reasonablemadness said...

Carrick wrote:
"My question, is it really accelerating?"

Yes. If you want to see if the loss of sea ice is accelerating, you could for example compare a linear regression fit with a quadratic fit and look, which one of those fits the data better.

I've done that here:
NSIDC September Arctic Sea Ice Extent

If you compare the residuals between the data and the two fits, its clear that the quadratic fit is a lot better, especially in the last years. You can even see this more clearly by looking at the 5-year average (which eliminates most of the year to year fluctuations). The 5-year avg follows the quadratic fit nearly perfectly. So yes, Arctic sea ice loss is accelerating.

reasonablemadness said...

Carrick wrote:
"My question, is it really accelerating?"

Yes. If you want to see if the loss of sea ice is accelerating, you could for example compare a linear regression fit with a quadratic fit and look, which one of those fits the data better.

I've done that here:
NSIDC September Arctic Sea Ice Extent

If you compare the residuals between the data and the two fits, its clear that the quadratic fit is a lot better, especially in the last years. You can even see this more clearly by looking at the 5-year average (which eliminates most of the year to year fluctuations). The 5-year avg follows the quadratic fit nearly perfectly. So yes, Arctic sea ice loss is accelerating.

reasonablemadness said...

And that's only extent. If you look at sea ice volume, the loss is even more dramatic (for the best we can tell).

If you look at volume, the linear fit is the worst fit of all. So it is here also very clear, that the loss has accelerated:
https://sites.google.com/site/arctischepinguin/home/piomas

EliRabett said...

That envelope looks funny. The minima are pretty linear, the maxima clearly quadratic or beyond. Is that telling us something??

Robert Way said...

I had a paper using a similar method rejected on a related topic - couple of okay constructive reviews then on the fourth reviewer it was rejected.

One thing that's important to note is that your definition of the AMO is problematic and I wonder why you chose to not use Trenberth's definition? Detrended SSTs will result in some anthropogenic trend inclusion whereas if you did a difference AMO similar to Trenberth you could avoid most of that.

James Annan said...

Awkward questions are best put to Jonny :-)

Carrick said...

James I did mention it had a positive slope and that "I'd have to do some fitting to the various components and look at how much ice loss is attributable to the various signal components." The "high frequency" component visually appears to have a much larger net area under it. I'd go back and analyze it but have zero time... we've a field deployment in less than two weeks and I have 60 or so sensors left to get ready.

WC, I said I did an analysis and linked to a result from it. I never said the url pointed to the analysis, but I apologize for any confusion that might have arisen over this. >.>

Reasonable madness, thanks for the comments and the figure. Your quadratic curve is equivalent to the linear curve in my figure (which relates to rate of ice loss.) The trick is getting a reasonable uncertainty interval here.

Carrick said...

Eli: "That envelope looks funny. The minima are pretty linear, the maxima clearly quadratic or beyond. Is that telling us something??"

I said the minima in my graph are "spiky". I think to really understand what it is saying, you'd have to spend some time thinking about the different mechanisms for ice loss (surface melt, Ekman transport, others?).

What struck me as curious was the apparent interrelationship between ice loss and La Niña phases of ENSO. Knowing that there is a link between ENSO and polar vortexes tells you something, but I'm not sure what. (Perhaps the polar vortex is stronger during La Nina phases.)

What isn't clear to me, when you see what is apparently mostly mechanically driven ice loss is how you link that up to a model that includes anthropogenic forcing: It's pretty much self-evident (to everybody except a reviewer maybe ;-)) that decreased thickness leaves the ice vulnerable to greater ice loss, so even if anthropogenic forcings isn't directly causes most of the ice loss, ultimately it's still the culprit.

[It's just how you quantify it, is the rub. When I get down time in a couple of weeks, I plan on sitting down and reading the paper. Sea ice is fascinating stuff.]