Cryosphere Today 2013 Arctic SIA daily minimum: July poll

[ChrisReynolds]

Richard,

Agreed, but what about the sub-pixel level?

[Wipneus]

Take 10 pixels of arctic of which 7 are completely ice and 3 completely water.

Both CT and Wipneus will call that 7 pixels at 100% + 3 at 0%.

That is already an approximation. Actually to know the signals of 100% ice and 100% open water is the one big problem that algorithms have to solve. Not having this correct, the grid cells (not pixels please) may be counted as 80% or 120% ice. The 120% you will of course never know as it is truncated. Likewise the open water may have 10% ice.
The signals for 100% water an 100% water are refered to as tie points

Now diverge the ice so that its equally spread out amongst the 10 pixels. I'm pretty sure that of the two methods CT is the most likely to say thats 30% which looks like water and call it 10 pixels at 70%, while and Wipneus is the one most likely to say that it looks like 30% water but 2/3 of it is probably melt ponds, and call it 90%.

No, all algorithms will return 30% if they have the tie points correct.

Resolution plays a minor role at most.

[Wipneus]

BTW, -58k3 in two days.

Still plateauing, Arctic Roos is even showing Area increasing.

[ChrisReynolds]

Richard,

Having spent the last hour looking at various data I'm not convinced it's possible to see which factor is dominant, melt ponds or divergence.

[Richard Rathbone]

Richard,

Having spent the last hour looking at various data I'm not convinced it's possible to see which factor is dominant, melt ponds or divergence.

For short term changes I don't think it is. But there's a huge systematic difference between CT area and Wipneus area which is to do with melt ponds not divergence so when a short term effect in CT area is seen that is explainable by melt pond behaviour, there's no reason to ascribe it to divergence.

If you've got two plausible effects, one of which you know is happening, and one of which might be, stick to the one you know is present. Occam's Razor.

[Richard Rathbone]

Wipneus,

Pixels not being grid cells, I'd need to see the actual concentration profiles to be sure, but the concentration maps look very different to what I'd expect if the discrepancy between Bremen having the CAB average at 95% when CT had it at 70% was down to tie points. There just seems to be too much 90-98% and not enough 100% on the Bremen map for that. I'd have thought the grid cells would need to be hundreds of times smaller than they are before systematic differences in tie points could produce the systematic differences shown on the maps.

What the algorithms call their tie points, might be a whole lot more variable than what I understand as a tie point, and hence capable of more subtle systematic differences than I am assuming they are. I am assuming that the tie point is the same when looking at cell with 100% ice as it is when looming at a cell with 50% ice. If there is some sort of adjustment that adjusts the tie point according to how much ice is seen, thats not what I'd call a tie point, but it could cause the sort of systematic concentration differences I see between the Bremen and CT maps.

[ChrisReynolds]

Richard,

But you have two equally valid things going on. There is divergence and there is a drop in temperature. Both are happening, the issue for me is the relevant importance, which is just another question I don't think can be answered.

[Richard Rathbone]

Richard,

But you have two equally valid things going on. There is divergence and there is a drop in temperature. Both are happening, the issue for me is the relevant importance, which is just another question I don't think can be answered.

Find somewhere that area measurements get distorted to the order of 1e12 m2 by divergence. That's the scale of melt pond distortion in the CAB in the recent past.

If you can come up with a set of reported measurements that show that order of area distortion from divergence, then its reasonable to assume divergence is significant too. Otherwise hold divergence in reserve for when melt ponds are not changing in the right way to explain the observation.

[ChrisReynolds]

But you cannot neglect divergence, the ice edge retreat has slowed.

[ChrisReynolds]

Richard,

Having got the latest PIOMAS data I now agree that divergence is a smaller factor than re-freeze/slowed melt.



The uptick at the end of the anomalies is from 25 July, the date strong cyclonic pattern took hold and temperatures dropped. An uptick in anomalies means a volume loss less than average, for PIOMAS this is a strong uptick at this time of year.

[wili]

Holy Crimminey, another uptick--from about 4.8 to about 4.83! That means essentially no change in area detected in about 10 days. That looks like a pattern we expect at the end of the melt season on the way to refreeze more than what one would expect at the beginning of August.

I'm still expecting another sudden fall, but I've been expecting for a long time now. (Taps fingers on desk.)

[BornFromTheVoid]

An increase of 47.7k, to 4,877,649km2, so no change for 11 days.

We're now 7th lowest on record, and 1.09 million km2 below 2012.

[Pmt111500]

An increase of 47.7k, to 4,877,649km2, so no change for 11 days.

And now we're on record territory, there's been only one longer stall in CT SIA numbers before this, occurring during late summer melt period. That was on 1-17th Aug, 1985, and this started a week later. So this is a unique event this early in the season in the whole record of ~34 2/3 years. One might argue that that's not rare since the maximum frequency of this occurring might be 3 times/century, but really there's no reason to suggest that sort of behavior from this short (albeit the longest available consistently organised) a record.

[wanderer]

Prepare for record August area loss!

[Richard Rathbone]

But you cannot neglect divergence, the ice edge retreat has slowed.

Its is observable that different area methods can report area differences of 1e12 m2 due to different treatment of melt ponded ice and that daily changes of the order of 1e11 m2 in CT area occur simply due to melt pond behaviour.

Who cares what difference the nibbles of a 1e6 mouse is making when they can see a 1e11 elephant stomping around?

As long as the elephant is moving in your direction, you should ignore the mice. Leave them for when the elephant has gone to sleep.

For a single grid cell, divergence might be an elephant, but not averaged over the entire arctic. Its the most fundamental part of the calibration. If its that far out it would show up as the same sort of obvious gross bias that melt ponds do.

[Steven]

Neven, in the latest ASI update you suggest that the freezing of melt ponds is the main reason for the recent plateau in the CT sea ice area graph.  But I'm not sure.  Probably it's one of several reasons.  But is there (anecdotal) evidence for the hypothesis that melt pond freezing is the main reason, rather than drainage and the fact that cold temperatures slowed down the top melt and the formation of new ponds?

Some of the buyo's in the Beaufort Sea suggest a refreeze (as was noticed).  But that's mainly MYI.  Most of the ice is FYI; isn't it so that the temperatures in the FYI region have to be pretty low in order for refreeze to play a major role? 

[Neven]

I don't know, Steven. Either it has something to do with melt ponding, or something causes data to be off, or something else. I don't know what it is, except that CT SIA looks weird to me for many weeks now, somehow very different than last year (but that's subjective).