NSIDC 2021 Arctic SIE September average: August poll

[Juan C. García]

ATTENTION: There are two ASI extent polls on the ASIF. This one is for NSIDC sea ice extent monthly minimum or September average (which is also used for the SIPN sea ice outlook), the other is for JAXA sea ice extent daily minimum [provided by JAXA (ADS-NIPR-VISHOP)].

Make sure you are aware of the difference before voting.

These are the September averages on 2000-2020 (in million km²):

Year      Extent
2000        6.25
2001        6.73
2002        5.83
2003        6.12
2004        5.98
2005        5.50
2006        5.86
2007        4.27
2008        4.69
2009        5.26
2010        4.87
2011        4.56
2012        3.57
2013        5.21
2014        5.22
2015        4.62
2016        4.53
2017        4.82
2018        4.79
2019        4.36
2020 (*)   4.00

* Corrected by NSIDC. According to NSIDC October 2020 post, it was 3.92 million km2.
http://nsidc.org/arcticseaicenews/2020/10/
From lowest to highest:

[icefisher]

2015's average looks good to me right now.  4.62 it is.

[Brigantine]

links to poll threads:
Month - PIOMAS / NSIDC / JAXA
June ___ link ____ link ____ link
July ____ link ____ link ____ link
August _ link ____ N/A ____ link

[Brigantine]

I take my June prediction as an anchor, slide up 1 bin, and squash the spread into half the width.

4.25 - 4.75

Assumed bin chances: (>5.25) 0%; 2%; 10%; 34%; 35%; 14%; 3%; 2%; 0%; 0%; 0%; 0%; 0% (<2.50)

[Stephan]

Unlike the years before I changed my bet one bin higher to 4-4.5 M km².

[The Walrus]

Both 2011 and 2015 look like the best fits.  Somewhere around 4.6.

[Juan C. García]

Moving to the range 4.25 - 4.75 km².

19 hours to vote or change your vote...   

[Brigantine]

I take my June prediction as an anchor, slide up 1 bin, and squash the spread into half the width.

4.25 - 4.75

Assumed bin chances: (>5.25) 0%; 2%; 10%; 34%; 35%; 14%; 3%; 2%; 0%; 0%; 0%; 0%; 0% (<2.50)

I'll go out on a limb and change my vote to 80%*[4.65 +/- 0.25Z]+20%*[4.25 +/- 0.5Z]

4.50 - 5.00

Assumed bin chances: (>5.25) 1%; 7%; 23%; 34%; 21%; 8%; 3%; 2%; 1%; 0%; 0%; 0%; 0% (<2.50)

[gerontocrat]

With just 35-40 days to minimum (probably) the chances of a major uptick or major slowdown in sea ice extent loss diminish every day.

The September monthly average for NSIDC sea ice extent, assuming remaining eea ice extent loss to be at the average of the last 10 years, comes in at 4.64 million km2, some 306k km2 above the long-term linear trend, and 644k km2 above 2020. (see attched)

A "percentages golfer" would therefore go for a range of 4.5 to 5 million km2. But I cling to the faint belief that there is a bit of life left in the season (with no real evidence), so am going for one bin lower at 4.25 to 4.75 million km2.

Perhaps this year there is enough global heating to burn down large parts of the Northern Hemisphere but not enough left over to boil the Arctic Ocean.

click image to enlarge

[The Walrus]

I agree with you gerontocrat.  I hedged my bet by going with the lower bin, fully realizing that if the final number comes in at 4.64, either bin is satisfied.  It seemed more likely that it would be lower than higher, but that just might be my own biases.

[oren]

Upping by one bin to 4.25-4.75, I can't see 2020 (or 2012) performance being repeated.

[Brigantine]

A summary of June-July-August votes:

June - 39 votes total

• 4.21 Upper Quartile (interpolated within bin) - 0.30 above JAXA
• 3.96 Median - 0.28 above JAXA
• 3.69 Lower Quartile - 0.35 above JAXA
• 19 or 49% voters at most can win (sum of 2 adjacent bins)
• 6 poll options required to include 90% of votes
• 0.52 Inter-quartile range

July - 36 votes total

• 4.21 (+/- 0) Upper Quartile - 0.22 above JAXA
• 4.02 (+0.06) Median - 0.24 above JAXA
• 3.77 (+0.08) Lower Quartile - 0.24 above JAXA
• 23 (+4) or 64% (+15) voters at most can win (sum of 2 adjacent bins)
• 5 (-1) poll options required to include 90% of votes
• 0.44 (-0.08) Inter-quartile range

August - 30 votes total

• 4.54 (+0.33) Upper Quartile - 0.24 above JAXA
• 4.43 (+0.47) Median - 0.24 above JAXA
• 4.21 (+0.52) Lower Quartile - 0.21 above JAXA
• 22 (+3) or 73% (+24) voters at most can win (sum of 2 adjacent bins)
• 4 (-2) poll options required to include 90% of votes
• 0.33 (-0.19) Inter-quartile range

[oren]

Big thanks for the analysis and for posting the links to previous polls.

[Killian]

Pretty big change from my first and second: 4.25 ~ 4.75. Will not at all be surprised to see 5.0. Best match continues to be 2011 to a near-perfect match, but I think this year will bottom out higher.

[Tealight]

According to my SIPN forecast from 13th August the 4.75 to 5.25 bin would have been the safest option. 4.5 - 5.0 is also a good option.

[Richard Rathbone]

https://www.arcus.org/sipn/sea-ice-outlook/2021/august

SIPN report now out. The median was 4.39. Tealights was the second highest.

University of Washington (PIOMAS in prediction mode) is in the pack this time (4.32).

[Freegrass]

According to my SIPN forecast from 13th August the 4.75 to 5.25 bin would have been the safest option. 4.5 - 5.0 is also a good option.

I'm one of the two who picked the between 4.5 - 5.0 option. So it's looking good for me...

[The Walrus]

According to my SIPN forecast from 13th August the 4.75 to 5.25 bin would have been the safest option. 4.5 - 5.0 is also a good option.

I'm one of the two who picked the between 4.5 - 5.0 option. So it's looking good for me...

Yes.  I was not as optimistic as you, as I went one bin lower.  Still, it is entirely possible that your bin may not be good enough, if the slow melt continues.

[Freegrass]

According to my SIPN forecast from 13th August the 4.75 to 5.25 bin would have been the safest option. 4.5 - 5.0 is also a good option.

I'm one of the two who picked the between 4.5 - 5.0 option. So it's looking good for me...

Yes.  I was not as optimistic as you, as I went one bin lower.  Still, it is entirely possible that your bin may not be good enough, if the slow melt continues.

It looks like you were still good with your pick. As was I. I was thinking somewhere around 4.75, and it looks like we're a little below that. So it's all good.

[Juan C. García]

Sep 1-15 average is 4.87M km².
Sep 15^th daily value is 4.741M km².
I voted to 4.25-4.75M km², so I am below too.

[The Walrus]

Sep 1-15 average is 4.87M km².
Sep 15^th daily value is 4.741M km².
I voted to 4.25-4.75M km², so I am below too.

Yes, not sure the Sept. average will dip below 4.75.  I suspect the extent will start rising very soon.

[gerontocrat]

Not that I want to rain on anyone's parade (liar, liar, pants on fire) but... a cautionary note.

IFF sea ice extent gains for the remainder of September are at the previous 10 years' average then the September average of NSIDC 5-day sea ice extent will be 4.915 million km2, 12th lowest in the satellite record, 0.92 million km2 above Sept 2020 and 0.55 million km2 above the long-term linear trend.

It is therefore in the realms of possibility that if a cool / cold pattern strengthens, remaining sea ice extent gains this month could produce a September average that just about breaches the 5 million km2 mark.
___________________________________________
Note:- Although this is the 5 day average, the difference in the one-day average will be very very small.

[Freegrass]

OK... Time for a stupid question... What's the September average? I thought this was about the minimum. What's the difference with the average?

Edit: I had some time to think about it, and I guess the average means adding all days of september together, and than taking an average?

So my new question is; Why is that important? What do we learn from that?

[gerontocrat]

At and close to the minimum, there is usually a lot of very low sea ice concentration ice. So a few days of sea ice drift can either greatly compact the ice or greatly spread out ice with a large impact on sea ice extent

So a very low or surprisingly high one day minimum may just be a function of sea ice drift - a flash in the pan. The September average smooths out and tends to remove these short-term changes and is therefore more significant.

Using just one day's measurement out of 365 to make statements about trends in the Arctic sea ice is, to say the least, problematical. (or maybe a bit like Senator James Inhofe who in February 2015 threw a snowball to the Chair of the Senate - It's snowing in Washington D.C. - Global Warming? Hogwash!!)



https://time.com/3725994/inhofe-snowball-climate/

[Juan C. García]

Yes, not sure the Sept. average will dip below 4.75.  I suspect the extent will start rising very soon.

The extent is already rising. As Gero wrote, we can expect an average around 4.9M km2.

So my new question is; Why is that important? What do we learn from that?

I like to follow both, the daily minimum and the monthly average. Sometimes the year was very bad for the ice and we can see it on the daily minimum, but the year had a fast september refreeze, so we don't see it on the average.

Also, the worst years (especially 2012 and 2020), the ocean stayed hot warm, so we have a late refreeze, that can go until mid-october. The september average is very low then.

And it is important to follow the area and volume also.

P.S. I think that 2007 late refreeze made the 2007 September average difficult to match by other years, fact that has been used by deniers to affirm that 2012 was an outlier and other years were not as bad as 2007. That is not true, if you follow the ADI-JAXA daily minimum and especially, if you follow the volume. 2007 is 3rd lowest on NSIDC monthly average now (it was second until 2019), but 5th lowest on ADI-JAXA daily minimum and 11th lowest on PIOMAS volume monthly average.

[Freegrass]

Ok, I understand now. Thanks for the explanation all! I guess the moral of the story is that extent is such an arbitrary number. The more important numbers are area and volume. Especially volume. But that's also the most difficult number to figure out due to the fact that it's not as easy to figure out ice thickness compared to the 2D numbers of area and and extent...

Too bad HYCOM wasn't really good with extent and area this year... I assume thickness is what they're better at - due to the fact that submarines depend on it to surface? What if we use NSIDC extent and area, combined with HYCOM thickness? What would that give us?

[KenB]

Ok, I understand now. Thanks for the explanation all! I guess the moral of the story is that extent is such an arbitrary number. The more important numbers are area and volume. Especially volume. But that's also the most difficult number to figure out due to the fact that it's not as easy to figure out ice thickness compared to the 2D numbers of area and and extent...

Too bad HYCOM wasn't really good with extent and area this year... I assume thickness is what they're better at - due to the fact that submarines depend on it to surface? What if we use NSIDC extent and area, combined with HYCOM thickness? What would that give us?

For a look at NSIDC Extents in September, see https://forum.arctic-sea-ice.net/index.php/topic,3183.msg275410.html#msg275410 where they're plotted in a couple different ways.  Generally speaking, September tends to be pretty flat relative to the rest of the year's changes, it seems. 

WRT combining various measurements, certainly building 'ensemble' models is a very well-studied technique in machine learning, but in almost all those cases, there's some subset of the inputs for which the correct answer is known, which isn't really our situation, unfortunately.  Also, I'm strongly reminded of Juan's signature: 

"Which is the best answer to Sep-2012 ASI lost (compared to 1979-2000)?
50% [NSIDC Extent] or
73% [PIOMAS Volume]

Volume is harder to measure than extent, but 3-dimensional space is real, 2D's hide ~50% thickness gone.
-> IPCC/NSIDC trends [based on extent] underestimate the real speed of ASI lost."

[gerontocrat]

I attach the NSIDC (5-day average data) September average sea ice extent graph.

At 4.916 million km2 it is less than 1,000 km2 different from the NSIDC one day sea ice extent average of 4.915 million km2 (i.e. a difference of a bit more than one NSIDC sensor pixel)

This September average is 0.56 million km2 above the long-term linear trend, 0.92 million km2 above the Sept 2020 average, and 12th lowest in the 43 year satellite record (2020 was 2nd lowest).

click image to enlarge

[The Walrus]

Congratulations to the two posters who chose 4.5 - 5.0.

[Brigantine]

4.92 it is then.

I'll go out on a limb and change my vote to 80%*[4.65 +/- 0.25Z]+20%*[4.25 +/- 0.5Z]

4.50 - 5.00

Assumed bin chances: (>5.25) 1%; 7%; 23%; 34%; 21%; 8%; 3%; 2%; 1%; 0%; 0%; 0%; 0% (<2.50)

my [4.65 +/- 0.25Z] main guess for August worked out pretty well.
my [4.23 +/- 0.60Z] main guess for June/July was also not so bad

In terms of the 4.50-5.00 vote - it's partly a case of being right for the wrong reasons. I thought the minimum would be lower, and further to the pacific side, and that early re-freeze near Greenland would put NSIDC sept avg. more than 1 bin above JAXA minimum. Nope. Pretty much exactly 1 bin higher. Most of you got that bit right! My JAXA guesses were further off on the low side, so the 2 misses cancelled each other out.
(but I'll claim it anyway thanks Walrus!)

Kara and Chukchi were still melting, while Greenland area dropped a bit of concentration, extent held out, so there was no low hanging fruit for an early rebound in extent.