« If this is real... | Main | Climate, Ice and Weather Whiplash »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Jai Mitchell

This is an excellent post neven, thanks for putting it together, I think that you are right about the ACNFS forecasts. Now it is showing a hole opening up in multiyear sea ice at 140W 80N in the course of a day. This modeled event shows ice going from 3.25M to .5M depth within a day or so. Just doesn't seem plausible. That being said, the historic 1-year chart does show a significant impact from these events over the course of the longer termed melt season and the fact that this is happening at the very beginning of melt on 1 year sea ice may be very significant. its wait and see time. . .

Here is the current projection with the 3M-ice-loss-in-a-day event http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticictn_nowcast_anim30d.gif

Chris Reynolds

To be clear Neven, whereas HYCOM (ACNFS) shows a massively low concentration patch, in reality the concentration doesn't seem to have dropped so much, but this is because the area of low concentration is so much larger. Although not as low in concentration.

I still think HYCOM got the amount of divergence about right, it just pooled all the low concentration in a smaller area than actually happened. Now ice albedo feedback has to toe hold within the pack over a large area, IMO.

In HYCOM the ice seems to be sort of like a single board, pull it and you get an opening at the other end. In reality the board is made of planks joined loosely with string. Pull it and instead of one opening you get a series of gaps between the planks.


I think you're right about that, Chris, and thanks for the explanation. Though imperfect, the ACNFS/HYCOM forecast can still tell us something.

John Christensen


Thank you for the new entry and reference on the new high resolution SIC map!

On resolution and SIA:
Since CT has relatively low resolution it must be constantly over-estimating the SIA, as it does not recognize the smaller cracks and leads.
While this has always been the case, with thinner ice and more cracks as we have seen this winter, it should follow that the over-estimation of SIA at CT is on the increase, right?
Is there another site calculating SIA from higher resolution data, and is this diverging from CT SIA and showing this over-estimation?


Jai, I've examined those images very closely and am fairly certain that such an event is not reported on any day in those ACNFS images.

You really need to grab the screen for the two consecutive dates, paste an arrow pointing to such a pixel where this happened if you want to make your case.

I would guess you are just taking a wild fling at reading their very nuanced palette by eyeball. That doesn't work on a 100-color gradient. Neither of those ice thicknesses you mention has been present in the half hemisphere under the cyclone.

The day to day changes at a fixed pixel are easy to study and have been incremental.

Is there another site calculating SIA from higher resolution data, and is this diverging from CT SIA and showing this over-estimation?

John, on the previous blog post, commenter SATire used image editor Gimp to ascertain the amount of open water for the zone under the cyclone on Wipneus' SIC maps. I wouldn't know how to do it myself, but I'm sure something could be programmed to determine the SIA for the whole Arctic.

Maybe KlimaCampus and Wipneus could arrange for the images to be even further improved (a lot can be cropped for instance), reduced in size, be made visually more appealing, and get some space on a FTP-server. That way the production of daily images would be guaranteed. Determining SIA would be the next step.

I don't want to tell anyone what to do, of course, but suffice to say it'd be pretty cool if we'd have our very own SIA product.

Jai Mitchell


on this link,


during projections of 6/7-6/9 the sea ice at approximately 142W by 79N goes from around 3.5 M (yellow-orange) to between 1 and 2 Meters of depth (dark blue to teal)

I would post pictures but it seems to me that a verbal description and link would be enough to describe it, I don't know how to make and post graphics on this site.

Bob Wallace

Just a suggestion -

There is a section on the forum called the Developer's Corner where discussions of how to use various tools last past the current blog post.



Neven writes "So it's safe to say that the ACNFS forecast didn't turn out to be real".

I'm not so sure about that. Let's not forget ACNFS is unique in making a forecast -- and they successfully predicted the right area and right time for an extraordinary impact (see pinwheel below) that we would otherwise have not anticipated.

We've discussed from time to time compressibility of the ice -- once it is soft, semi-recovered from fractures, pocketed with brine channels and so forth. So how is this to be treated under area, concentration, volume, extent etc? The number of molecules in ice has not changed.

Finally I hope people can find a word than 'divergence'. I'm an native speaker of English yet have no idea what is meant (how different from fracture?). Divergence is very widely used across all branches of science, as in gradient, divergence and curl of ice displacemnt fields. Indeed it would be quite instructive here to calculate the divergence for each of the last ten days (as is done for the Baltic).

Look again at this still-developing central pinwheel of ice with its microwave signature very close to that of open water on the same frame.

I'm seeing a darned good prediction here at the round-up resolution of ACNFS, though we haven't actually computed whether fracture widths sum up to their reported thinning. ACNFS has the rotation about right as well as the gearing off of MYI to the Fram.

Jaxa-Ijis color 38VH 18V microwave 23 May 13 to 03 June 13.
left: unretouched, native resolution; right: 20% fade on equalize, native resolution.

 photo navyEvent1b_zps3608fb1c.gif photo navyEvent2b_zps29e5df40.gif


A-Team, I admit that if it weren't for the forecasted 'hole' by ACNFS and discussion thereof on the ASIF, I wouldn't have been able to announce it in advance. It would have just passed me by, and then looking at MODIS I would've gone: "Well, look at that, holes in the middle of the ice pack."

On the other hand I did write in the first ASI update:

My attention immediately gets drawn to a low pressure area in the middle of the Arctic Ocean. With a pressure of 990 hPa it obviously doesn't come close to the Great Arctic Cyclone of 2012, but it's still pretty low for this time of the year. It's also quite persistent for a cyclone, sticking around for 4-5 days in practically the same spot. It's the cyclones that do the most churning and diverging, so it will be interesting to see what this will have done to the ice pack as soon as the clouds move away.

But I think this was also based on discussions in the Forum. Can't remember. Too long ago.

With regards to 'divergence'. I'm not native, but it makes perfect sense to me. ;-)

I think I started using the word around the time we developed the CAPIE metric, back in 2010. I learned about the word in high school, when applied to light beams that diverge in a prism or water, meaning they move away from each other. That's how I mean it when I use it for Arctic sea ice as well: Ice floes moving away from each other. Like due to cyclonic winds.


Jai, 142W by 79N is off-topic -- it is well outside the cyclone footprint that we're discussing for the ACNFS prediction (upper half hemisphere).

Initially you said 0.5m in a day but that has slipped to 1-2m over two days. Perhaps download the animation and verbally share with us the exact pixel coordinates where you are seeing this change. I'm not seeing it.

There has to be a factual basis for criticism of other people's scientific work -- otherwise it is defamation.

And once we get a precise palette measure of the actual change, then we can ask how you know it didn't happen as ACNF indicated (within their framework of resolution).

 photo notHappening_zps7c960874.gif

Jai Mitchell


This "event" is modeled to occur in 4-6 days. I posted it because Neven mentioned that

"So it's safe to say that the ACNFS forecast didn't turn out to be real, but that doesn't mean that nothing happened"

and the new forecast, which you put up (thank you!)seems to confirm his statement, since this new ice depletion at the new location is even greater than the previous one.

Peter Ellis

A-Team: I have no idea why you're being so combative. It's perfectly well on topic to look at the area of 142W/79N, since that's where the cyclone has moved to! And your own animation shows - or would do if you went back one day further - that at the centre of the storm, the modelled thickness drops from all yellow pixels (~3.6-3.7m) ice in the 06/06 forecast to some cyan pixels (2.2-2.3m) in the 06/08 forecast. That's bang on what Jai said.

Basically, ACNFS is predicting that the cyclone will do the same thing to the MYI north of the Canadian Archipelago as it did to the FYI in the northern Laptev sea. That's incredibly important if it does indeed come to pass, and it'll be interesting to follow it on the various maps and on MODIS.

Will the thicker MYI act more like a single slab, and either resist the diverging forces completely or rip open a few big polynyas? Or will it have the same low tensile stress as the Laptev FYI and open up a whole crazy patchwork of little polynas? Interesting times.

R. Gates


I would highly recommend that we keep the word "divergence" in our discussions here in that it has a very specific meaning related to sea ice dynamics, and using some other word could confuse and muddle things up-- besides being less scientifically consistent. The standards that Neven has set here are high, and if we mean the spreading out (and not just separation through very localized cracking) of sea ice than divergence is the right word. For example, if you apply some paint to a deflated ballon, and then inflate it, the individual molecules of paint under a microscope would be seen to diverge from one another as the balloon gets bigger and bigger. (much like our universe in fact).
So, depending on how closely you are measuring the distance between those moloecules, (Neven's point about higher resolution measurements) you could say that the area or extent of the paint on that balloon is increasing as the balloon inflates, when in fact, the molecules are diverging and the paint is thinning. (as is the rubber skin of the balloon).

In the case of sea ice, you take a region previously defined and see the area increase not by new ice being formed, but by general across the board separation or spreading of the ice, the average thickness of that region must be reduced during divergence.

Peter Ellis

And now two in the spam bucket since R Gates posted while I was writing my second post. Sigh.

[Sorry, didn't check the spam bucket for 15 minutes. N.]



Divergence can have many forms, but for this case think of it this way... In three dimensions a positive divergence is like swelling, and a negative divergence is like shrinking.

In the quasi 2D case, a positive divergence is like thickening, and a negative divergence is like thinning.

What we are seeing is probably better represented as a fractional dimension of about 2.2 - 2.5. And with the fracturing, even that isn't really all that good a representation.


Actually, I'd suggest 'dispersal' to be more apt than divergence.

R. Gates


But if we use the word dispersal, then we have no consisency to actual scientific terminology and we've made up a new word to describe something. Divergence is the accepted term and every sea ice expert knows exactly what it means, as should everyone here eventually.

Artful Dodger

On the proper terminology:

"cyclonic winds induce Ekman transport which causes net divergence and upwelling, or Ekman pumping, while anti-cyclonic winds cause net convergence and downwelling, or Ekman suction."

Care to calculate the values to estimate melt? Start by solving this system of 2 bounded simultaneous differential equations:



P.S. As Neven likes to say, *nothing* is ever 'safe to say' when it come to the Arctic ;^)


With respect, "disperse" and "dispersal" are used all sorts of places in science. Considering the definition, and the behavior of the ice we saw, it seemed the correct way to describe it.

HOWEVER, I have no particular attachment to it. It was simply a suggestion.


Hi Peter,

What do you know about those mass buoys accuracy?
Does the sonars compensate for temperature ? For instance Ice at -20 Cpropagates sound more faster than at 0 C... This would give the impression that ice is getting thicker (if there isn't said compensation application). There also has been a higher softer less denser growing from bottom ice column, this will also give an erroneous sonar reading of thicker ice.


Other than Neven, has anyone considered the fact that it has been cold in the arctic the past month for a legit reason that the extent and area haven't declined much?


Yes, several people have commented on the possibility that colder than normal temperatures might partly explain the apparently slow melt. It's not that simple, tho - divergence is very obviously also happening

John Christensen

Henry and sofouuk,

Yes, it has been cold at least at 80N, as you see on DMI: http://ocean.dmi.dk/arctic/meant80n.uk.php

2004 and 2007 saw similar cold during late spring, and as we know this did not prevent the summer of '07 from becoming a disaster..


Alexander Beitsch informed me that he has changed the plotting routines:

Most of how it has been has
internal or somewhat "historical" reasons internal to our institute.
Changes include:
1) everything is PNGs now;
2) I reduced single file sizes by reducing the image depth;
3) I use a single contrasting color ("coral") for continents;

Alexander and other team members intend to look into possibilities to reduce the graphic sizes even further in the coming weeks.

Repeated message: these data are not officially released and work in-progress.

Indeed the sizes are now 2MB, not much larger than the files I produce. I will continue produce mine for some time longer until things have stabilized on their FTP server:


John Christensen

Looking at Beaufort and CAA, there is no reason why this year we should see the imminent plunge those two areas experienced already early/mid-June in '12. This should be good news for MYI in that area - at least for the time being.
The 'cliff' should probably come from Kara and Laptev, where SIA is high and ice layer is thin/scattered.


To compute SIA from that concentration maps is totally easy with gimp:
* Select the area of interest
* use the histogramm function (Menue: Colors -> information -> histogramm)
* look up the "mean" value and devide it by 255 (=100% ice) to get the ice fraction
* multiply by selected area (e.g. number of pixel from histogramm function times 3.125 km times 3.125 again (pixel area)

It gives a bit more SIA than cryosphere today, probably because of better resolution. Pixel area is only correct at 70° latitude. And as Wipneus allready warned again: "these data are not officially released and work in-progress." Relative measurements are save and interesting anyway.
All of this will not work with jpg or colored pictures - only linear grey scale and lossless bitmaps. If you have the black hole in your selection, choose lower level of histogramm >38 to avoid a wrong mean.

This is a very nice and easy thing to compare SIA in areas of your personal interest from day to day. Just note the position and the size of your selection for comparisons over time and as reference for other poeple interested.


Peter Ellis just made the point that Hudson and Ohkotsk are also behind schedule compared to recent years, and they will make no difference to the final area/extent. We're approaching the cliff alright, the only question is how steep and high it will be


For example I tried to follow the divergence of ice due to the storm in the last days. (Divergence here: Rate of loss of ice concentration in selected area - that means assuming no melt but only flow out).

Selected area: Large circle of 2.761 Mio sq km - about the CIA but away from coasts, because I did not want interference from leads opening in future.
Postion of ellipse: 800 1450 and size: 600 600

date: ice fraction:
20.5. 98.71 % of the 2.761 Mio sq km circle
21.5. 98.71
22.5. 98.55
23.5. 98.27
24.5. 98.55
25.5. 98.43
26.5. 98.35
27.5. 97.65
28.5. 97.14
29.5. 97.88
30.5. 97.02
31.5. 97.06
1.6. 97.25
2.6. 97.25
3.6. 96.86 % of the 2.761 Mio sq km circle

so between 20.5. and 3.6. we have a divergence (loss of SIA probably due to flow out of that circle) of about 50,000 sq km - not much for 2 weeks and I did expect more effect from the hycom concentration forecast.


@ SATire

Since the storm was in the central ice, area losses would be minimal. The issue, at least from my perspective is more that of volume losses, which we see in the controversial HYCOM images.

@ A-Team

I think CICE/HYCOM nailed the thinning prediction, even if they may have over-shot the total impact a little. Note the recent images in the model show slightly less thinning than earlier predictions.

That said, I completely agree with you that they're useful for predictions and that this thinning and fragmenting/divergence has been substantial.

@ Chris Reynolds

A U Washington study last year found that upwelling and churning (Ekman) processes were responsible for 150,000 square kilometers of sea ice lost during the GAC last August. A pretty big chunk for just one storm.

Some (primarily deniers looking for a culprit other than global warming) went on to blame the storm for dramatic sea ice losses during 2012. Though the storm contributed, it was just one of several factors in an overall warming Arctic that knocked the ice down to new lows.

That said, don't count the action of upwelling out. It has an impact. And this particular storm is thinning as well as diverging the ice.

I will agree with A-Team in that the use of the word 'divergence' has tended to blanket the effect of upwelling in discussion. And this is a factor that should be taken into account.

Chris Reynolds



In discussing that I'm going to use the hours of the clock form, but I'll ditch the clock bit. So 6 is the line of latitude straight down, 12 is straight up, 3 is to the right, 9 is to the left.

The thin region is visible within the inner circle between 9 and 12. This is the region thinned by the storm, but the drop in concentration at the same time shows that this thinning is due to a drop in concentration dropping the average thickness of the grid boxes.

1) Look at between 1 and 2 within the centre circle. You can see a dark line (a ridge of thicker ice) parallel to the islands. This is moved upwards and outwards when the thin region forms.

2) Look at about 10 just outside the centre circle, and you'll see a green horizontal line (another ridge of thicker ice). Now you see that after the storm this has moved anti-clockwise (expected with the storm), but it has compressed, it is not as long.

3) Look at the lower left quarter within the centre circle. Here you'll see the bright green of the thicker multi-year ice. Before the storm it is in contact with the 9 (O'clock) line, after the storm it has dropped away and to the right. Part of this is rotation, but the drop requires divergence.

4) Look at the Atlantic ice edge, note the relationship with the islands. There is movement indicative of clockwise rotation, but there is also movement out into the Atlantic.

This is why I think there's a large element of divergence. Without access to model diagnostics however it isn't possible to determine with greater precision the roles of divergence and melt. However before the storm the ice was uniformly around 2m thick, after it is down to 1m thick. That's 1m thick of ice (1m^3 X the area concerned = volume loss). This is the sort of thinning that takes over a month within the pack when driven by insolation. So I don't think it's feasible within a single week.

Chris Reynolds

Oops, should have said 'when driven by insolation after the peak of insolation, July/Aug. Observed from PIOMAS data.


Robert, this concentration prediction by hycom I intended to test: http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticice_nowcast_anim30d.gif

By determing SIA on our own in the very simple way described above we can now test the prediction modell-free - and this test is proving hycoms prediction wrong, doesn't it? Just check on your own by selecting the area of your choice ;-)

Regarding thickness we have to wait for PIOMAS modelled ice thickness to check hycoms values - I am looking forward to see that results.


BTW - the concentration prediction by hycom was allready adjusted. There was much more green some days ago.

And to remind you the place where you can find the Wipneus' useable version of U-Hamburgs concentration maps to do your own SIA-measurements: https://sites.google.com/site/apamsr2/home/pngcby32/

And gimp for your own SIA-measurements you can get here for free: http://www.gimp.org/

Chris Reynolds

Robert Scribbler,

"A U Washington study last year found that upwelling and churning (Ekman) processes were responsible for 150,000 square kilometers of sea ice lost during the GAC last August. A pretty big chunk for just one storm."

It's an interesting paper. The pre-print is here (for anyone else).

That situation is different from the current because the August storm hit ice distributed in water already warmed by the sun. This impact is within the ice pack in late May, the water beneath the ice has not been heated. IIRC that study found the ocean heat source was from the near surface layer (down to ~20m).

If this storm has melted a lot of ice the we'll know in a few days. The impact of the August 2012 storm is visible in the PIOMAS main series.

Eric Orr

Just a lurker here, but if I remember the Washington study right it wasn't only the Ekeman transport but that the Beaufort gyre reversed and shunted the fresh water runoff from the Yenisey and Ob was shunted off towards the Bering and not to the central arctic. This destroyed the fresher lens of water that kept the warmer saltier deeper water from the bottom of the pack. Since it was already significantly later and warmer this had an increased effect than a similar storm would have had earlier in the season.

The problem with comparing eimilar events to previous years is that in such a dynamic system the effects of hysteresis are overwhelming. The current states are so wildly divergent from previous years that long term predictions seem foolhardy. Of course they are also fun and interesting.


PIOMAS update:
Latest value: 2013-5-31 19.087

I have updated my graphics at ArctischePinguin for the latest data.

Monthly DataMonthly data
Daily AnomaliesDaily Anomalies
Daily data Daily data
Daily data with a "prediction" based on exponential trend Daily data with a


@ Chris Reynolds

The U W paper found that sunlight had warmed a layer within 60 feet of the surface and that this was consistent with typical spring/summer conditions. Upwelling from churning and cyclonic action (Ekman) could certainly access this warmer layer. This warming was not atypical for periods when sunlight is present.

More generally, the Cold Halocline Layer has been eroding since the 1980s. This erosion means churning and Ekman type transport is more likely to impact the surface ice to greater and greater degrees. So the subsurface is in a state of dynamic change. One that has already had impacts on the ice.


Sorry, but I'm not convinced this number crunching provides the conclusions you've asserted. We've already seen an impact on the surface ice in the visual and in multiple measures besides CICE. So I think this is jumping the gun a bit.


Eckman transport is not always the whole story in vertical transport of ocean heat to melt the bottom of the ice.

Jiayan Yang has produced a daily map of Ekman transport and upwelling across the whole Arctic Basin using sea-ice drift speed (needed to couple stress from ice to water) for 28 years (1978-2006). A sampler is below.


 photo eckmanSampler2_zps4729243d.jpg


"Sorry, but I'm not convinced this number crunching provides the conclusions you've asserted."
Vice-versa Robert. Hycom concentration prediction is based on number crunching while SIA measurement from concentration maps is theory-free plain observation by satellite.

Nowhere in the storm affected zone in the CAB I can find SIA concentration < 80%. Can you? In that respect I consider hycom concentration calculation wrong. But this finding is premature due to the fact that it is not reproduced by one of you yet. So please: Check on your own.

SIA from 4.6. map in CAB (position 800 1450 ellipse size 600 600): 96.66% of the 2.761 Mio sq km circle. A bit less than the day before but perhaps underestimated due to cloud-features in the map.


Right or wrong, good hands-on effort there, Satire.

RScrib says we're "jumping the gun a bit" overall. That's my thinking too - why the rush to judgment when the predicted event has not even happened yet -- the putative thinness hole at 82.5N 130E is not scheduled to peak (and then plateau) until 09 Jun 13.

Will this unexpected alternative -- to thinness creeping in from the periphery -- create a puddle in the middle of the ice pack? Too soon to say. Meanwhile a double cyclone has developed over the western multi-year ice; massive Fram export of the multi-year ice may prove more significant than localized thinning.

The slide show below overlays the key thinning area of the 04 Jun 13 event, as envisioned by Navy Hycom, onto the relevant satellite images, all rotated and rescaled to the same configuration. The idea here is to find what type of satellite sensor, if any, produces a visual correlate to the thinning.

In doing science, the data is marshalled first, then secondarily interpreted. Try it some time.

I could have added Hamburg sea ice concentration to the slideshow but that is based entirely on the 89 Ghz passive microwave which is already represented below by the Jaxa polarization ratio. 'Arctic Composite' could also have been included for infrared but the cloud tops are too warm this time of year for it to see down to the ice.

Modis provides by far the highest resolution. For each linear feature of interest on the ice, there are 12.5 pixels at Modis 250 m resolution for every pixel provided by Hamburg 3.215 km -- the area ratio is 156:1.

Modis can't see through clouds (the last few weeks have been a mess), so the ice patches showing through neither support nor refute fracture spreading ("divergence"). Floes separated by grayish ice have been around for months -- here the control pre-event Modis imagery will never be available because of cloud cover.

We've seen some confusion here on the satellites: Modis uses reflected sunlight so is not operative at night, in winter, or on cloudy days; AVHRR uses passively emitted infrared which is available at night and in winter but is blocked by clouds depending on their temperature; Jaxa uses passive microwave emission for both the 89 Ghz and color 38VH and 18V which are always available and almost entirely cloud penetrating; Ascat uses active ice-penetrating radar at a shorter wavelength (5.3 GHz, vertically polarised antenna) and seems to serve as the basis for Navy Hycom.

Thus Modis and to a lesser extent near-infrared are the only ones directly impacted by changing surface albedo. The Arctic Ocean more or less radiates as a blackbody of 273 K; the other instruments are picking up on that, except for Ascat which is pinging the ice and calculating a classical scattering cross-section (sigma-0) for display.

 photo NavyslideShow_zps34927b05.gif


Nice pictures again, A-Team.

"In doing science, the data is marshalled first, then secondarily interpreted. Try it some time."

You are right - in science I collect data, calculate all possible errors and dicuss every issue silently for a year with collegues prior to publish some original and safe results.
Here I used the "look-and-shout" approach which I considered appropriate for blog comments. Maybe I am over-excited that we can do now our own SIA measurements from the new maps and I hoped the storm-effect would be a nice motivation for someone to try it, too. It is bloody simple and I have explained every little detail how to do it. It is better for me to stop posting SIA measurements now at least until someone considers it interesting enough to try it oneself.


Actually you are fully on-paradigm there trying to get a grip on the initial datum (the Navy prediction).

Hmmm, double cyclone, both counterclockwise -- continuation of the same event or just the new normal? It is poised to potentially move a lot of multi-year ice Framward.

This highlights the vulnerability of ice pack persistence these days -- if not this event, then another.

 photo navyDoubleCyclone_zpsf81a951d.jpg

 photo NavyLower12Jun13_zpsc0575ac1.gif

 photo navywind_zps74b5fd18.png

Meanwhile, down below the Arctic Basin:

 photo AbruptMelt04Jun13_zps0e0955f0.png


The Navy sea ice thickness animation above starts today 05 Jun 13 and runs until 12 Jun 13 -- I should have embedded some dates. The longer pause marks the end date.


@ SATire

It is certainly very interesting. And I thinks it's great work too. I'd like to fiddle with it myself a bit.

Doesn't CICE show thickness? This is why I was having trouble finding a good conclusion from a comparison with area. So I thought we'd mixed up apples and oranges a bit.

That low is now at around 975 mb. June 14 shows it bombing down to 965...


R Gates and Neven, after reading your posts on 'divergence' and not seeing anything resembling a scientific definition, I took your suggestion and looked up journal papers on Arctic sea ice to see how scientists there are using divergence. It turns out to be exactly as I said, in the vector calculus sense. You may be using conventional mathematical divergence in an intuitive sense without realizing it.

Please read this 2012 paper in its entirety to see what Arctic scientists mean operationally by divergence and how it is calculated over sea ice from satellite imagery. In particular, consider pages 477-8 and the lower left of Fig 4.

Divergence means the dot product of the gradient linear differential operator acting on a vector field, here the velocity vector field of the ice as set in rotational motion by inefficient momentum transfer from the cyclone.

The velocity vector field at each point at each time is commonly determined by observing trajectories of computer-recognizable features.

Navy Hycom provides 30-day animations of ice pack velocity and uses standard del dot velocity divergence to calculate from it (and other considerations) their ice thickness animation, here predicted thinning of ice under the eye of the cyclone.

Sea ice concentration by itself cannot be used to determine the divergence -- the entire ice pack can be at 100% concentration before and after an event, yet still exhibit a very substantial scalar divergence field because of its complex viscoelastic rheology (as modelled unsatisfactorily in Piomas), temperature-dependent brittleness to shear, and ordinary compressibility of bottom mush, brine channels, melt ponds, cracks and intercolated open leads and so forth.

I am satisfied that IceBridge could directly determine the divergence attributable to a cyclonic event along its line of flight. However that only provides a transect and IceBridge is not out there even doing that.

Other than that, we have no experimental method to determine whether the Navy's predicted changes in ice thickness (ie divergence of velocity field) is accurate or not. Ice thickness (and so ice volume) are very difficult to directly determine; the time and spatial resolution of complex model calculations like Piomas are too coarse relative to size of this cyclone.

R. Gates

A-Team, Thanks for that link and your usual detailed explanation. I will review both a bit later today. In the meantime, one dynamic that has not been discussed here at length with these late spring and summer cyclones is the amount of sublimation that is going on. Maybe we talked about it last year with GAC-2012?


One could also check the other Navy-prediction, which is a bit more related to the new map on the block. But maybe they do that checking on their own - the 50% concentration in the old storm effected zone becomes smaller from day to day. Instead they have found a new region in the MYI where concentration will drop below 50% in a few days:
Since the first drop has not been confirmed by observation up to now I would bet the next one will not show up, too.


Nevin notes that getting a steady stream of daily KlimaCampus 3.125 km sea ice concentrations could enable an improved in-house sea ice area product here; Wipneus notes substantial progress above in cleaning up their work-in-progress graphics mess but cautions that an independent archive still needs to be maintained.

Yes, indeed. I took a look at their ftp products again last night -- some bugs are still there, other things are better, still others are worse.

The palette is much improved -- it is now 100 shades of blue, which displays much better. Furthermore, it is clean -- when I formed the additive selection of the 100 colors and took inverse, I could wipe their messy land mask clean.

However when I looked at the 04 Jun 13 image -- being ever so careful not to tamper with its underlying color useage with Gimp manipulations -- only a dozen colors were actually utilized anywhere in the Arctic Ocean. Presumably palette use gets a lot more nuanced during melt season but for now it is very limited.

I tried various parabolic and cubic monotonic transformations of the palette -- realtime interactive in Gimp -- but nothing emerged from that because of the sparse use of color.

Fom this image or the animation I made earlier, it seems that cloud artefacts are not fully removed. That can be viewed either as a bug or feature (subtle overlay of cyclone). Clouds are affecting all our imagery these days with the exception of Ascat, in part because of increasing temperatures, increasing vapor and perhaps increasing airborne salinity.

The land mask is not coral as they say but a very complex mixture of thousands colors arising from setting down the lat long grid right on top of the mask and overwriting the data. Again, we do not need this grid in the first place; it can easily be provided as a separate fixed file, along the lines of what Modis does with its optional coast outline. This is a top priority to fix -- overwriting experimental data with cartography trivia.

The new use of png is better than jpg but still suggests a serious misunderstanding of scientific image file formats. With only 100 colors for the percent increments and 2-3 for the mask, they are well under 2exp7 =128, much less the 2exp8 =256 that constitutes the minimal grayscale that a computer monitor displays.

Under these circumstances, they should be using indexed color here. That amounts to a custom palette or gif file. I fixed this, finding file size (with or without cropping the absurd areal coverage down to CT seas) dropped to 381kb, no surprise for an image that uses only a dozen colors. While gif may compress internally, this would be lossless here.

It looks like they might be backing off 3.125 and going back to 6.250. It was never clear to me how they could sharpen the 89 Ghz in the first place. It might prove more useful to dumb it back down at the very end. Ground resolution is determined by the size of the satellite's microwave antenna and is proportional to frequency. Nobody has gone up there to put in a new antenna.


Satire, the MYI feature has to do with the second cyclone and could quite possibly be real -- under-the-eye has a strong scattering cross-section correlate in Ascat, though actual amount of thinning is anybody's guess. The forum has an excellent discussion of the coming weather by Void.

R. Gates, yes sublimation seems way under-covered here on the blog -- could you write that up for us?

I would also like to see bromine treated better. That very much ties in with snow cover -- we need ice for snow to sit on.

And as Aaron noted a while back, many chemical species in the earth's atmosphere do not remotely equilibrate within the seasonal insolational time scale; this causes the Arctic to somewhat have its own air column and so its own private greenhouse effect which unfortunately would be enhanced by bromine, methane release and additional water vapor.

John Christensen

Neilt added a comment over on the PIOMAS thread about an apparent hole opening up near the Pole itself towards CAA.
It appeared somewhat abruptly on June 4, so is this real or does anyone have satellite images to back it up?


Clouds are affecting all our imagery these days with the exception of Ascat, in part because of increasing temperatures, increasing vapor and perhaps increasing airborne salinity.

What a coincidence that you're saying this, A-Team, because it's something I had been looking at yesterday. This comparison of ASCAT images you had posted a couple of days back, made me wonder why 2013 seemed so much whiter than 2012. To me that indicated that 2013's ice pack was overall thicker than that of 2012, because I associate the whiteness with thickness.

You can imagine my WHOA yesterday when I compared 2012 day 155 to 2013 ASCAT images directly with each other (retrieved here):

Continued in next comment...


But then I had a look at the days running up to day 155 and saw something I didn't know was possible with radar images:

Now, lately I've been noticing low altitude clouds/mist over large parts of the Arctic on LANCE-MODIS satellite images. We're really having a bad view of things below, even where there are high pressure areas. And I guess it's having an effect on the thing that makes the radar images as well.

What do you think, A-Team?



At first glance it seems there is an impact on ASCAT.

BTW, for reviewing the daily forecast plots, the ESRL/PSD has the Reforecast2 500 mb pressure maps including anomalies, and is also forecasting precip and SLP for 360 hrs.

See: http://www.esrl.noaa.gov/psd/forecasts/reforecast2/wx_maps/index.html#nh

It appeared somewhat abruptly on June 4, so is this real or does anyone have satellite images to back it up?

John, my rule of thumb with Uni Bremen SIC maps, is that if a large yellow-green region shows up and stays in the exact same spot for three days, it could be real. If it shifts every day, appears and disappears etc., it's a sensor artifact, due to clouds, melt pond or whatever.

If this concentration 'hole' is still in the same spot tomorrow and the day after that, there might be something below the clouds that we're not seeing. I don't believe we will, but you never know, especially in the Arctic.

Account Deleted

The images of last year before May 10 show a lot of "white" in the left part (Beaufort sea and close to Bering Strait) that gradually faded and disappeared before June 1. Maybe that makes a difference w.r.t. this year


Interesting, ulisescervantes, thanks a lot. Maybe this is part of some sort of final transition from cold to warm. It would make sense that it's happening later than last year, as 2012 had a flying start to the melting season, and 2013 obviously hasn't.

Or is it something that has to do with the satellite?


Those clouds that confuse Bremen's concentration are also very prominent visible in the new maps and disturb SIA measurements - those somehow smeared regions may correspond to that orange-pinkish clouds in MODIS-Alpha. Maybe that is heavy rain or hail? We should try to avoid measurents in that "smeared regions".


Day 2 and Uni Bremen's north-pole-opening is still there


One of the things I've noticed is that holes/cracks recently opened in the central ice tend to steam even after the storm has passed. This obscures the cracks/holes/leads in the visuals and it seems to do the same in some of the non-visual monitors.

Let's see if I can find some examples in today's Lance-Modis:


OK, the above link is to a region near the one currently indicated to have thinned so much in CICE/HYCOM today. Completely obscured by dense, multi-level clouds at the center of the storm today. Not a good example of the effect I'm describing. But it is a good example of why we can't verify HYCOM/CICE, in many cases, on day 1.


Now this region was mostly obscured by the storm yesterday. But note how much in the way of cloud-cover remains. It's still quite obscure even though the bulk of the storm has moved on. A band near the middle where the ice seems to have fractured quite a bit, is still almost entirely cloud-covered. But also notice how generally fractured and broken this region is, even though much of it is still obscured.

@ SATire

I've looked over our posts and I want to apologize for a bit of misunderstanding. I can see now that you're comparing the daily concentration measures in HYCOM CICE in order to determine area accuracy.

For my part, I've tended not to look at the daily concentration measures there, Uni Bremen, or elsewhere for validation. There's lots of artifacts in these day-to-day measures and they tend to jump around a bit. And this can be very confusing, especially when there's a storm that seems to fiddle with the surface sensor a bit.

I've been looking at the thickness model a lot lately though because it seems to provide a good comparison with past history. Some thinning events, I've noticed get exaggerated on day 1 or 2, but by day three they seem to be worked out. For example, the central ice thinning for day 1 and 2 of this storm on Friday and Saturday indicated greater thinning than was later recorded. But the thinning was still significant.

You can take a look at the model from Thursday/Friday where I'd saved it here:


Note the difference between the indicated thinning and the current thinning seen here:


Still significant, but not as significant as indicated in the initial animation. That said, the most recent image does show a much broader erosion of the central ice even if the 'hole' initially indicated was not as deep as first projected.

Part of the reason for this is possibly due to the fact that the storm's impact is greatest on the day its circulation center is directly over the affected region. On days after, the ice tends to re-converge a bit, probably indicating the new state better. The other part is probably a bit of an initial error bar in the sensor itself.

This is why the work you're doing is very helpful and useful. It helps us refine what's happened after the fact. But I would caution to wait a few days or weeks (depending on event duration) to let things wash out before making final conclusions.

Now, I don't use the HYCOM concentration model as much, so I don't know the quirks involved with that particular sensor/model as well as I do with the thickness measure.

Hope this is helpful and please excuse any grammar errors as I'm typing faster than I should.



Slightly OT (apologies), but given the weather impacts globally and conversations here on the contribution of SSW and SLP highs ridging, I thought the following would be of interest.

There is a 1 hour live streamed Climate Desk broadcast at 4:30 EDT today featuring Jennifer Frances and Stu Ostro from Weather.com

See the article at: http://www.motherjones.com/environment/2013/06/one-meteorologistss-come-jesus-moment-climate-change

The live stream link is in the box under the lead photo on the right side.


R. Gates

Regarding Arctic warmer season cyclones and sublimation, there are some research & papers that cover this, at least partially, such as:




And certainly the ice portion of the climate models deal with it (but not very well IMO, given their complete failure to model the rapid decline) as the role of sublimation during Arctic cyclones might be expected to change as the Arctic continues to warm.

Few here may realize the sublimation of sea ice directly to water vapor in the presence of strong winds (and of course the proper temperature profile) does occur. Here, where I live away from sea ice, but having plenty of snow in Colorado, we see sublimation occur frequently in the winter with strong downsloping winds off the Rocky Mountains in the winter. These are called "Chinook" winds, which is Native American for "snow eater" and they literally eat the snow away, turning it from the frozen state to water vapor, with no liquid state in between.

When the temperature profile across the Arctic is right, i.e. near the triple point


And with the proper high enough wind speeds during a cyclone and temperatures, the ice doesn't melt, but is turned directly to water vapor. A bit of a longer-post on this may be interesting, as sublimation could be a factor in some of the ice mass loss during summer cyclones not fully discussed or considered here.

Jai Mitchell


I called the WWF and confirmed that the following link will carry the stream of the event you posted on. The link you posted doesn't have a link to the stream page.

This is the correct page to view the presentation.


Remko Kampen

Presently and staying for a week there's a hot air situation over the Hudson region and some. T850 over the Bay nearing and occasionally passing +10° C.
It should look like this on Sunday:
Sea ice area over Hudson Bay meantime keeps it about freezing all the time. Guess where the heat is going :)

Chris Reynolds


ASCAT is indeed about to become useless. The system was originally set up to monitor winds, as it's predecessor QuickScat. It does this by monitoring the radar backscatter signature of waves on the sea surface.

There's something about the Arctic summer that causes a general haze to settle as air temperatures get to zero and above. ASCAT spots differences in ice due to different dielectric properties of FYI and MYI. I suspect that wetting of the ice surface turns the backscatter behaviour to that of water, not ice, hence the ice fades out.

Earlier in the winter we saw a large amount of MYI enter Beaufort, according to the last couple of weeks of ASCAT this has vanished. This is not the case, it's the start of the summer 'fade out'.


Very interesting, Chris!

Like I said a couple of days ago the ice was turning blue on the Pacific side of the Arctic (like it does every year around this time) and that coincides pretty well with the animation of ASCAT images above. So that's probably it, in combination with those hazy clouds.

Learned something, yet again. On to the next zillion things I don't know yet.


Neven, perhaps not so clear from my earlier post, I was referring to large-scale cloud movement and Ascat support for the region under the cyclone eye in the second Navy Hycom 'incident'.

A large coastal band centered on the the Bering Strait is causing interpretive problems for all the lower frequency microwave sensors including 5.3 Ghz, less so for the longest, the 89 Ghz. A kitchen microwave oven, designed to oscillate the dipole of water, runs at 2.45 Ghz -- a factoid likely related to melt season problems we are seeing. The other big issue swinging into view is swath interval -- too much is changing down below between successive passes of the satellite that comprise a composite image.

However we don't want to throw out the baby with the bath water just yet.

A grayscale like Ascat assigns a number (1-tuple) in the [0 255] range to each pixel of the Arctic Ocean. That number is intended to represent scattering cross-section, hopefully that inside the ice rather than snow or melt water on the ice or the cloud water droplets penetrated in bouncing back to the satellite. When all goes well (in winter), the backscatter number can be unambiguously associated ice age/thickness -- brighter being thicker.

Beginning in late May, brightness became regionally ambiguous -- along the CAA and central Arctic Basin, it still does a good job of picking out the multi-year ice but along the Bering Strait (which is way south at 65º vs Greenland's tip at 83º), brightness means something else is kicking in dielectric constant as you and Chris discuss.

RGB color like Jaxa provide a 3-tuple coordinate for the color cube [0 255] x [0 255] x [0 255] whose diagonal is a grayscale. Any recent image like AM2SI20130606RGB_high would show this too is developing a similar regional ambiguity with time-of-day issues as well (swathing). However three color dimensions provide enough 'room' to sort out blueish Strait band from blueish-green MYI. So it's not so much inherently ambiguous as visually confusing and uncertainty of how blueish should be interpreted. Were the colors projected to the diagonal, we could not sort this out.

Since the PR89 Ghz is less affected and Jaxa has 36 and 18 Ghz, we could probably squeeze out the dependency of the blueish band on wavelength and strip it off the Ascat, returning it to unambiguous brightness/thickness, whereas the other half of the decomposition would then isolate this uncharacterized new contributor.

Microwave does stay useful on ice throughout the summer (see earlier color Jaxa animation of Sept 2012) -- but the interpretive difficulty goes through the roof.

The image below shows the 05 Jun13 Ascat, puffed out to more evenly use its grayscale. This brings out a couple dozen blackish dots in the central Arctic Basin. What they represent is a bit of a mystery. There is a lot of detail in this imagery but our interpretation is lagging.

width 730 pixels; 415 showing:
 photo ascat155_zps42ca8377.png


Thanks, A-Team, that's a great explanation. I never really looked at radar images, except in winter to compare years, so this took me a bit by surprise.

Lars Kaleschke


your comments on our AMSR2 maps are very welcome. Just a few remarks:

The new use of png is better than jpg but still suggests a serious misunderstanding of scientific image file formats.

The PNG files were not intended as a container for the data. The ice concentration data are stored in separate NetCDF files for scientific use. When you can't use NetCDF or prefer PNG for other reasons we could perhaps provide a plain PNG file that contains only the data without grid lines if there is the need.

From this image or the animation I made earlier, it seems that cloud artefacts are not fully removed.

The large atmospheric influence is the disadvantage of the 89 GHz channel. These channels have the advantage of high resolution. But we can't remove the atmospheric disturbance over sea ice without additional information from other frequency channels, i.e. like from the 22 GHz water vapor channel (which has only coarse resolution).

It looks like they might be backing off 3.125 and going back to 6.250. It was never clear to me how they could sharpen the 89 Ghz in the first place. It might prove more useful to dumb it back down at the very end. Ground resolution is determined by the size of the satellite's microwave antenna and is proportional to frequency. Nobody has gone up there to put in a new antenna.

AMSR-E had a failure of the 89 GHz A-scan. Now with AMSR2 we have again both scanlines which sample at 5 km distance in nadir. Thus, one could say that we have a new antenna! The instantenous field of view is about 3x5 km. According to the Nyquist-Shannon theorem it makes sense to use a sampling distance of half the original sampling distance which would be even less than 3 km. The 3.125 km grid allows to compare the data with other data in the "historical" NSIDC format. Anyway, we have not tried to sharpen the data which also could be done with the Backus-Gilbert method at the expense of more noise. We just use a finer sampling to better fulfill the sampling theorem.

By the way, nice animations, A-Team, congratulations!


Thanks Lars, that is fantastic to get some really knowledgeable input -- more posts, please!

Remote sensing interpretation is much more advanced over land than sea ice -- things don't melt or move around, meter-resolution base layers exist for roads, land use, soil, geology and veg classes, plus you can just step outside with a laptop to ground-truth imagery.

For example, those mysterious dark lines on google imagery of our Tucson place -- they used an unusual dawn shot, causing saguaros to cast long shadows. One cactus per pixel -- know the height of one, know the height of all.

In the Arctic Ocean, basically nobody is out there six months a year and even in summer, not where and when you need them to groundtruth something. Our best synoptic resolution is Modis 250 m but more typically a few km -- not a good match with the intrinsic scale of ice feature variation, meaning each individual pixel has a lot going on inside -- we can't fully associate ice core and buoy data with single pixel properties (unless within a unifom area).

IceBridge overflights are probably the best bet and certainly under-exploited by us to date, but those flight lines just sample a few dates and their swaths a small fraction of the 7 million sq km Arctic Basin.

The image-serving web sites could provide a weekly shot marked up by some interpretive expert -- but they don't. No explanatory material whatsoever.

Ditto the journals -- something useful like the PI putting lots or arrows on picture features, is unpublishable (too chatty, not cutting edge research, probably done back in the 70's). Everyone cool already knows.

So we are on our own, playing 3 yards and a cloud of dust football. Occasionally I do figure out what some puzzling feature below really represents. Below, I was baffled for a long time by a white pixel in Ascat and color Jaxa imagery. It never seemed to move or change color. How was that possible out in the open ocean?

 photo mystery2LaptevB2_zps4e597a40.gif

After considering drilling platforms, frozen-in ships and methane blowouts, I finally realized from an overlay that it was just Bennett Island, a speck of land 100 km north of the main island group in the Laptev Sea.

At 150 sq km it was just right to dominate a single pixel; with four permanent glaciers on a plateau, its color would give a steady bright on Ascat. And the other speck dancing around it in animations would just be its ever-shifting associated polyna.

In summary, the sea ice will be long gone before I have its interpretation figured out.


Another drop in SIA in the storm affected region - but like yesterday heavily disturbed by "rain-smear" so I would not consider that as comparable to earlier SIA measurements again. But some "real" black spots and lines open there - so something is happening at last.

But if the second region with concentration predicted <50% right in the MYI will really show up, I would like to doubt again for same reasons as stated above.
(Refering to this hycom -CICE concentration forecast as usual - because that is directly observable by the map of this thread, if the weather allows for: http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticice_nowcast_anim30d.gif )

Chris Reynolds

A Team said:

Thanks Lars, that is fantastic to get some really knowledgeable input -- more posts, please!

It's generally the rule that I don't have time to do more than scan comments here. Which is not to say I'm always working, I must make time to do other things (like watching cute-cat videos on youtube ;) )

But I always read comments Lars Kaleschke, Axel Schweiger, Mauri Pelto, and the other scientists who comment on their areas of expertise here.

Comments by scientists working on the Arctic are always welcome.


The animation below shows how the Navy Hycom ice thickness prediction changes (for the same future fixed date) as time goes on. The left side shows the last days of their animation of 07 Jun 13, which goes out to 14 Jun 13; the right side their animation made on 08 Jun 13 which goes out to 15 Jun 13.

On June 5th, the two sides are identical because these are not predictions but instead their final version. On June 6th, the left side is a prediction but the right side 'observed' (though are not real measured ice thicknesses).

From June 7th on, both sides are making predictions. The last day, 15 Jun 13, is only represented by their prediction on the later date as they go out a fixed seven days.

These diverge quite significantly. Presumably the right side is more accurate because they had one more day of actual satellite imagery and a revised weather forecast to feed into their model.

Notably the developing pocket of light blue 2m ice in the earlier prediction is replaced by a tongue of multi-year ice moving down from 82.5N 180E over 08-11 June. So quite different but still both are projecting very significant ice movement.

Because typepad is no longer allowing side-by-side images (which might not enforce synchronicity anyway), I fused the two animations into single frames. It would be easy enough to add tomorrow's predictions to the mix; this would help us determine an appropriate level of confidence in these thickness forecasts (which may very with melt season stage and weather extremes).

 photo NavyComp_zps85f7c3bf.gif


Chris writes that comments here by Lars Kaleschke, Axel Schweiger, Mauri Pelto and other scientists on their areas of Arctic expertise are very welcome here.

I'll second that appreciation -- unleww we have some new twist on a breaking event, we are just standing outside with our noses pressed to the mainstream science glass.

Folks here might enjoy my latest -- and hopefully last -- tedious journal article in this week's PNAS.

Getting back to Neven's 'issues' with seasonally increasing microwave imagery artifacts affecting year-to-year comparisons, I looked into what is causing 'bad hair days' for Ascat. The four week animation below (up to today, day 158) shows that the faux multi-year ice brightness in the coast-hugging arc about the Bering Strait can be attributed to weather systems passing through (as the effects persist unabated onto land), rather than to blue ice, salinity or some degree of melt (though those can't be discarded as contributing factors). The Kara Sea and nearby regions have been equally problematic for months.

Although the area where the scattering cross-section physically corresponds to image brightness is shrinking by the day, the central region of multi-year ice is still working out.

The animation has made a non-linear but monotonic (hence invertible and lossless) transformation of contrast, seeking the sweet spot for discriminatory human rod motion vision while maintaining native satellite resolution and rotation frame. This would be quite difficult to replicate outside of dedicated image enhancement software (ie in numerical array view). The sigma-0 palette was transformed right along with the Arctic Basin but there was no room to include it.

 photo fadeEqualizedAscatB3_zps50450055.gif

The second animation zeroes in on the persistent black spots in the central Arctic Basin. These might well be isolated remnant floes of thinner younger ice as their brightness suggests. Or something else. While their origin and persistence remains unexplained, by 'connecting the dots' the resulting polygons could illuminate not just regional translation and rotations but also ice rheology. That's all to be found in previously published articles, but those don't reference the markedly different ice we have today.

 photo fadeEqualizedAscatB4_zpsaca436cc.gif

R. Gates

"Blacks Dots"= melt ponding? At least that's the guess of one whose nose is pressed up against the mainstream science glass. Their appearance and persistence seems to have some correlation with the passage of weather systems.

Here's a visual close to where one of the recent cyclones was:


Pretty chewed up ice an lots of open water, especially considering how far into the central ice pack this area is.

R. Gates

On further review, those dots in A-Teams visual are way too big for melt ponds.

The comments to this entry are closed.