NSIDC 2016 Arctic SIE September minimum: August poll

[Sleepy]

Guess the race is on, ADS extent already above 4.8. Amazing.

Edit; sidebreaks can make comments look silly, mine was just a small follow up to previous commenters.

[Tealight]

As of 9/23, if my adding machine input is correct, the average is 4.330 million km2.  The average is likely to stay within the range of 4.25-4.50, but possibly just exceed 4.50 by a hair.

As of 23rd Sep the extent is 4.395million km2, but there are lots of low concentration areas which will flash into existence over the next few days. If I lower the extent threshold from 0.15 to 0.10 the Sep extent is already 4.61 million km2.

[Feeltheburn]

Tealight, can you tell me how you arrived at that number?  Is that the monthly average as of 9/23, or a 5-day average, because the NSDIC website reports the 9/23 ice extent as 4.814 million km2 and 9/24 as 4.899 million km2?

And I'm aware there are other sites that have their own models for estimating sea ice extent, but the poll was for NSDIC.

Thanks in advance.

[Tealight]

Feeltheburn, I calculated the extent like the NSIDC directly from daily concentration data. This way I can exactly replicate monthly extent values. With concentration data its possible to calculate the "opposite" of extent too.

Attached is a map showing all pixels between 5 and 15% average September sea ice concentration. The total extent of these pixels is 0.664 million km².

[Feeltheburn]

Ok Thanks.  NSIDC website is now reporting 5.071 million km2 for 26 Sep 2016.

[Feeltheburn]

Make that 5.174 million km2 as of 28 Sept 2016, which is 1.1 million km2 higher than the reported 7 Sept 2016 min of 4.083 million km2.

And by the way, I'm not a "skeptic" or "denier". I'm an "optimist".

[Tealight]

September is finally over and I calculated the monthly extent as 4.7 million km²

[seaicesailor]

Indeed. The official number is 4.72 M km2 5th lowest on record. A very weird end of season.

[oren]

This is why I don't bother voting in the monthly average poll. The number is quite meaningless, as it loses a lot of resolution around the annual turnaround.

[Tealight]

For me 2016 is a great example of how statistics will fail at future predictions. I calculated NSIDC September extend based on minimum daily area and an average ratio of past years. Due to the exceptionality of this years ratio I ended up too high for area and too low for extent.

Ratio between NSIDC Sep average extent and NSIDC daily minimum area

Year:	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016
Ratio:	1.46	1.54	1.52	1.60	1.59	1.63	1.48	1.48	1.47	1.96

2007-2015 mean: 1.53
2007-2015 standard deviation(sample): 0.064

This puts 2016 at +6.855 standard deviation!

For normally distributed data it falls outside 99.999999999% and for Chebyshev's inequality still outside 97.87%.

[Tor Bejnar]

Thanks, Tealight.  Now I don't feel so bad guessing one bin too low.
And congratulations to the 9 individuals with effective (this time) crystal balls!

[Rob Dekker]

Regarding that 4.72 number for NSIDC-monthly, or 5th place (after 2012, 2007, 2011 and 2015), I did not see that one coming.

Especially since the minimum was 2nd place (only after 2012) and average of ice extent based on daily numbers for both Jaxa and NSIDC turns out to be 3rd place :

Jaxa average extent over September based on daily numbers :

2016: 4.39
2015: 4.51
2011: 4.46
2007: 4.18

which puts 2016 in 3rd place (after 2012 and 2007).

NSIDC average extent over September based on daily numbers :

2016: 4.50507
2007: 4.26727
2011: 4.56133
2015: 4.61577

which also puts 2016 in 3rd place after 2007 and 2016.

The big difference for the NSIDC monthly number of 4.72 seems to originate from the somewhat unorthodox method of ranking each pixel by their average ice content over the month, and counting them as ice 'extent' if that pixel shows 15% average ice over the month.

So when a pixel has 100% ice concentration during only 15% of the time (5 days) of the month, it gets counted in full for that extent number. While it would only add 15% for the daily average over the month.

With the rapid refreeze, there were a lot of pixels that fit that metric and that is why the overall number pumped up to 4.72.

I think that the "average of daily numbers over September" is a better metric (and one that I thought was meant with "NSIDC September minimum") which came out as 4.5, right in line with the most chosen bracket in this poll.

[oren]

Stupid criterion. Even with no refreeze, when the ice is mobile and dispersed you will get many pixels with partial coverage throughout the month.

[Richard Rathbone]

The big difference for the NSIDC monthly number of 4.72 seems to originate from the somewhat unorthodox method of ranking each pixel by their average ice content over the month, and counting them as ice 'extent' if that pixel shows 15% average ice over the month.

It is orthodox. Extent is defined by average, not instantaneous, concentrations.

[budmantis]

The big difference for the NSIDC monthly number of 4.72 seems to originate from the somewhat unorthodox method of ranking each pixel by their average ice content over the month, and counting them as ice 'extent' if that pixel shows 15% average ice over the month.

It is orthodox. Extent is defined by average, not instantaneous, concentrations.

Not necessarily disagreeing with you Richard, but I think there are times when orthodoxy should be questioned. Dogmatic thinking can be found in science as well as religion.

[A-Team]

Stupid criterion. Even with no refreeze, when the ice is mobile and dispersed you will get many pixels with partial coverage throughout the month.

Agree. Also with many similar comments by wayne on the ASI Blog. We're in a rut and need to move on.

These antiquated algorithms remind me of the ghost ships of the Arctic like the Baychimo that just keep sailing on long after the crew has abandoned them. Once a pipeline is scripted, the extent etc web sites will keep churning out these numbers unattended until the govt stops paying the electric bill.

Perhaps we need a new forum next season that considers the 'new Arctic' only, thumbs down for mention of gregorian calendar unit averages, daily whole-ocean measures and century blowtorches that don't consider the actual physical state of the ice. Which is going to drive an abrupt endgame, to which decadal trends have led but can't anticipate.

It could be argued though that between early snow melt on the tundra, albedo drop on Greenland and early extensive open water, that a third of the Arctic refrigeration function is already gone so loss of the other two-thirds, whether gradual or abrupt, is incrementalism.

https://www.researchgate.net/publication/273043683_The_Arctic_shifts_to_a_new_normal

[VeliAlbertKallio]

Many of the features in September 2016 Arctic Ocean resembled the ice floe filled Southern Oceans' in Austral autumns of the Antarctic where pulverized sea ice regenerates very rapidly.

Far Eastern Shipping Company (FESCO) has been commissioned by the regional government to carry out intentional sea ice demolition operations. The 'ice chipping' is carried out during the sunlight-filled spring season by cutting slivers off the sea ice edge whenever the winds blow the loosened sea ice towards warm ocean (waters to melt it). During sunlight filled spring this "sea ice demolition" helps for an earlier onset of open ocean, summer season, and lengthens fishing season. This raises local temparature up to +2C and is popular among local boters and fishermen: http://www.setcorp.ru/main/pressrelease.phtml?news_id=11048&language=russian

Opposite operation to the spring-time "sea ice demolition" ('ice chipping') is the autumn-time "sea ice creation" when in the cooling autumn darkness the broken up and scattering sea ice induces the growth of sea ice once freezing air temperatures are around. Despite all talk about people being worried about Arctic warming, sea ice creation in autumn has not been popular - quite unlike its opposite when ice chippers have been breaking sea ice to warm winds to melt.

Arctic Ocean in September 2016 with its highly pulverised sea ice (at the beginning of month) was much like "Mother Earth's" own ice chipping operation in Antarctic-style with a very rapid rebounce of sea ice area as winds and sea currents spread the myriads of ice floes in frost conditions to facilitate ice growth. However, had these concentration centres (ice floes) been lost, then the re-emergence of sea ice would have been very different: delayed.

The rapid sea ice formation has caused reduction in ocean ventilation in comparision to situation where all sea ice had been lost. This could mean that there might be a tiny bit more energy left for the next melting season. It all depends how thick snow blankets are formed over the sea ice and how thick the sea ice forms. The number of leads in Arctic sea ice over the coming months will determine how much heat is lost during winter from the ocean.  As sea ice has been thinning, and become more seasonal ice rather than multiyear ice, it breaks easier and forms more leads.

This September there certainly were no need for Captain Krasin to pulverize ice to make it grow faster in cooling autum darkness, Mother Earth made our season with +6.855 standard deviation as sea ice returned record fast when these myriads of ice floes spread and refused back together. For normally distributed data it falls outside 99.999999999% it was the tipping of 2016.