Monday 15 September 2014

The August Extent Loss Oscillation

In my July Status blog post I rounded up by mentioning a possible oscillation in August NSIDC Extent losses. Using the Wipneus extent calculation based on NSIDC concentration, and with the benefit of 2014's August figures, I think it is worth looking at the again.

However I don't have a convincing explanation, and at this stage I am not convinced it needs one.

The August Extent losses are shown in the black trace for whole Arctic extent, and after 2004 what looks like a four year period oscillation may have started. Three regions are used, Beaufort, Chukchi and ESS, and Laptev to CAA (Laptev, Kara, Barents, Greenland, Central CAA)

The peaks of the cycles, 2004, 2008, 2012 seem to be associated with a peak in losses from Chukchi and the East Siberian Sea (ESS).

Looking more closely at some of the regions.

It is worth noting how much of a post 2007 year this year has been within the Central Arctic and Canadian Arctic Archipelago (CAA). In those regions August losses in 2007 were high as the ice edge came well within the Central region and the CAA too faced substantial ice retreat. But 2007 was not one of the peak August loss years. 2012 was a peak August loss year, but overall the Central and CAA don't seem to have played much a role.

Losses in August in Laptev, Kara and Barents are on a clear downward slope, this seems to be related to declining extents in July, certainly in Barents extent by the end of July has proven a severe limit on losses in August - you cannot melt what isn't there! But aside from 2008, there isn't the pattern of high losses were looking for. However combining the high loss of 2008 with the high loss of 2012 in the Central Arctic, it is clear that the region of Laptev to the CAA plays a role.

Turning now to Chukchi and the ESS we see the pattern of high losses in 2004, 2008, and 2012, which is what is seen in overall Arctic Ocean August extent loss. Something might be happening in this area to cause a large component of the oscillation. I have not addressed Beaufort as that shows no relevant peaks in August extent loss.

As there is an apparent 4 year cycle it is worth looking at the difference over 2 year periods, to see what is happening to make highs and lows (or lows and highs) different from each other.

Breaking the Arctic into the previously defined three regions, the above graph shows the contribution each region makes to the two year difference. The two year difference is the difference between the stated year on the graph and two years before, so for example, 2010 (Low) is the low point in a 'cycle' and 2008 (High), a year which happens to be in the high point of the cycle. So the major regions having a role in 2010 being lower than 2008 is the ESS and Laptev to CAA, it turns out that ESS and Chukchi is the biggest factor, that is odd because that region is so much smaller than the vast majority of the pack in Laptev to CAA. Note that 'low' and 'high' refer to low loss and high loss Augusts.

Concentrating on the low point years of the cycle it turns out that in each case, 2006, 2010, 2014, despite the small area relative to Laptev to CAA, the ESS and Chukchi account for a large part of the difference between the high points and low points of the supposed four year cycle.

Looking at the high points it seems that in the Laptev to CAA region there is greater loss of extent in 2008 and 2012 than in the preceding low years.

Stepping back there is the suggestion of a repetition of pattern, however with only two complete cycles it is too early to read anything into this.

In 2000 Polyakov and Johnson published "Arctic Decadal and Interdecadal Varibility', PDF, in which they identified a 60 to 80 year periodic cycle in Arctic Climate and sea ice variation, this cycle consisted of a peak in the 1930s during a warming period then, and a peak in the 1990s, both associated with a limited recession of the sea ice. They concluded that "There is evidence that the LFO [Low Frequency Oscillation] has a strong impact on ice and ocean variability. Our results suggest that the decadal and multidecadal LFO drive large amplitude natural variability in the Arctic making detection of possible long-term trends induced by greenhouse gas warming most difficult." At the time this was a reasonable conclusion.

Subsequently it has been seen that the warming and ice loss of the early 21st Century is far greater than that of the 1930s and the idea that an LFO is driving the current period of warming has fallen by the wayside. There may be a LFO, it's just that the current warming and recession of ice (of which the 1990s was the start) is more likely caused by anthropogenic global warming than any natural climate cycle. The reason I mention this paper is that using only a few cycles is always risky. I consider that strongly claiming the August extent losses shown above reveal an oscillation emerging in recent years is probably not wise. It could be a coincidence, then again...

My suspicion about this is that as the ice has thinned the effect of periodic pulses of multi-year ice entering the ESS and Chukchi might be being seen, but that is little more than a guess about a feature that may be random chance. That said, this might, if real, be a feature of some wider auto-oscillatory behaviour, e.g. Dukhovsky 2004.

Anyway I'll conclude by noting that it may be worth following, and I'll update this in September 2016, high August extent losses then with 2015 fitting the pattern of 2011 and 2007 might make this worth the effort of digging into.


Anonymous said...

A nicely considered post. It is an interesting pattern, but I agree there's nothing for it but to wait. I don't see an obvious direct mechanism, and there are too many potential candidates for subtle, indirect causes.

I don't have a convincing explanation, and at this stage I am not convinced it needs one.

One for the quote folder!


Chris Reynolds said...

Now I just have to remember about it in 2016! ;)

iceman said...

I puzzled over this in your July post, and don't have any real insights now either. My sense is there's more to be revealed by isolating each year's interval of greatest extent loss. But that's a much bigger and more exacting analysis task than using the monthly numbers.

Chris Reynolds said...


I don't get what you mean by "My sense is there's more to be revealed by isolating each year's interval of greatest extent loss."

The greatest interval of extent loss is from April/May to September.

I've got all Wipneus's data in a single massive spreadsheet, I can also process gridded concentration from NSIDC. If you can persuade me I might make the time to write a macro to do the job.

However I'm well behind on following up the 'slow transition' stuff.