HYCOM

[uniquorn]

Re- ridging, looking back through the mosaic panorama images may be useful.

[FishOutofWater]

Over the past few decades the loss of thick multi year sea ice, and the presence of more open water and larger waves has led to more ridging. Because Hycom has been updated and improved it is difficult to use it to analyze the changes in Arctic sea ice thickness over the years. The new version of Hycom is probably a better model than PIOMAS to understand what the ice is doing NOW but it is inferior for inter year comparisons that involve years before the change to the GLB - global model. The new GLB version is a major improvement that is hard to compare to the maps of the old Hycom.

[johnm33]

"Keep up the good work interstitial  " +1 and thanks.

[interstitial]

One of the issues with either model is resolution. Ridging strikes me as something best described on the meter scale and both models only work down to the multiple kilometer scale. Resolution also comes into play when looking at thick ice flows frozen into place. I have included an image of Mclure strait on worldview. Is it as thick as the thick ice flow or more like the fast ice holding it together or an average based on area of each thickness. Hycom’s guiding principle is to indicate the maximum thickness rather than estimating volume. If Hycom made a volume estimate I am confident it would be a lower estimate than Piomas. I made that point before but I have examples of Hycom that are instructive. Hycom shows the Mclure strait at around 2 meters thick  but the fast ice holding it together is probably under a half meter if I had to guess. Not a great image of the Beufort but that ice all appears thicker and does not appear thicker in hycom some of shows thinner. When I looked at the first Chukchi image I was surprised that that boot shape showed up as thick as it did since there is a very low concentration of ice. When I looked above it I was even more surprised as there is almost no ice there at all. Look at the upper left corner. When I look at the Hudson I do not think they factor in concentration at all. In worldview the band of higher concentration on the left that only slightly shows on Hycom. On the other hand looking at worldview there is now indication of a thicker chunk in the middle. That green is around 2 meters thick while the purple around it is closer to half a meter thick. I am not sure why I included Greenland.
The main takeaway here is use a concentration map or satellite image when looking at Hycom thickness. 

[interstitial]

and the rest

[interstitial]

okay maybe one more

[interstitial]

If I can read this right and I am not sure about that. This buoy reads 85 cm of ice on 7/10 for that location hycom shows about 200 cm.

[uniquorn]

dtc's are 2cm apart

https://www.cryosphereinnovation.com/simb3

Bruncin 3.85m DTC

192 dtc values

so maybe 170cm, I didn't check your estimate

[oren]

Image of ridging, from Mosaic:

"Heavily ridged sea ice seen off the bow of the ship during ice observations on 14 June (Photo: Luisa von Albedyll / AWI)."

[interstitial]

Thanks for the correction Uniquorn.
Oren: It appears my understanding of ridging may be flawed. I was picturing a line of ridging here and a line there not a region of ridging. Your insights are helpful as always.


To everyone reading. Though it is time consuming and humbling I learn more when I post things and get corrected than when I just read others posts. I am not here to win arguments mostly I just like to know how things work.

[HapHazard]

To everyone reading. Though it is time consuming and humbling I learn more when I post things and get corrected than when I just read others posts. I am not here to win arguments mostly I just like to know how things work.

That's why your posts are awesome & I'm sure many more readers here than myself also learn a lot right along with you. 

[interstitial]

The forecasts do not show when the final bit of ice melts out. The forecasts seem to show white until every spec of ice melts out. This forecast shows a little more melt in a number of places but not all. More melting appears in the central arctic seas in the forecast than in the next day image. This is shown by less purple turning white in the next day compared to the forecast in the central seas the CAA and Hudson.  A bit of 5+ meter ice north of Greenland is reduced to less than 2 meters in the next day but not the forecast. A bit more green (1.8 meter) ice melts in the next day compared to the forecast.


Overall thinner(purple) ice melted more in the forecast than the next day. Specific chunks of thicker ice melted more in the next day than the forecast.

[interstitial]

Hycom 7/18/2020

[ajouis]

Piomas is getting quite different from hycom, especially on the beaufort and chuchki front

[interstitial]

When comparing Piomas to Hycom for 7/15  I note similar things to what I have seen before. Zoom in a bit to compare.
Piomas resolution does not allow it to show some small but thicker chunks amidst thinner ice.
Piomas ignores ice concentration below 15%
Hycom for operational reasons shows the upper limit of possible ice thickness in an area. Piomas is supposed to estimate volume but Piomas shows thicker ice than Hycom.
Hycom shows ice if even one fine chunk is visible. This is frustrating during melting season. I prefer Piomas in this.
Each Piomas color is 0.25 m thick and by counting shades from yellow which is easy to identify thickness ranges can be determined. Piomas shows the ice about 0.5 m thicker than Hycom in Beufort. In Chukchi and ESS Piomas shows thickest ice in the middle where Hycom shows some thicker ice near the edge and around 80 N latitude. North of Laptev to about 85 N shows mostly thinner than 0.5 m thick ice on Hycom and thicker Piomas shows ice mostly thicker than 1 m. To the west of the Anzhu islands Hycom shows a chunk of 2 meter ice. Piomas shows no more than 0.75 M. North of Svalbard Piomas shows 1.75 – 4.00 m thick ice. Hycom shows the ice is not over 1 m thick. North of Greenland and CAA shows a significant swath of over 3 - 4 m thick ice on Piomas. The same location on Hycom shows most of that ice is less than 2.25 m thick with small thicker sections up to 4.5 dispersed throughout. The southern CAA shows no ice in spots that worldview show full of ice with some cracking. Hycom in the same area shows ice 0.25 – 1.75 m thick.

[Tom]

Piomas also shows ice that isn’t really there.  I noticed in particular around the islands east of Svalbard.  Attaching a mostly cloud free pic from 2 days before.  Hycom seems to match better what can be seen in worldview.

[grixm]

Piomas also shows ice that isn’t really there.  I noticed in particular around the islands east of Svalbard.  Attaching a mostly cloud free pic from 2 days before.  Hycom seems to match better what can be seen in worldview.

PIOMAS does not try to accurately model the ice edges, only the average thickness of an area *if* there is ice there. To calculate the volume it uses external data for area and thus it doesn't count ice in locations that the area data says there is no ice in, even if PIOMAS has modeled a thickness above 0 there.

[Tom]

Thanks for that info.  It’s been bugging me for a while that the images didn’t seem to reflect reality in worldview, especially since I knew it incorporated area data, but that makes much more sense if it’s using the area data as a mask to produce the volume numbers from the model output.

[oren]

Hycom for operational reasons shows the upper limit of possible ice thickness in an area.

I am certain this is not true. The 7m floe example upthread should have been enough to refute this. But in general pressure ridges in all regions exceed the Hycom thickness, which I assume represents an average, not maximum.

[interstitial]

Hycom for operational reasons shows the upper limit of possible ice thickness in an area.

I am certain this is not true. The 7m floe example upthread should have been enough to refute this. But in general pressure ridges in all regions exceed the Hycom thickness, which I assume represents an average, not maximum.

I respectfully disagree. They searched for thick ice for a long time before they found that particular 7 meter thick chunk or the 6 meter thick one nearby. When I saw images I realized that these were not tiny areas like the size of a house but more like a few kilometers.  Most of the rest of them near the polarstern showed much thinner ice. There was some concern they wouldn't be able to find a good spot to begin. It was ridging which seems to be picked up near shores but maybe not elsewhere? Is it much rarer? Maybe it is more localized? In any case I would consider it more likely to be a model failure than indication of anything. In the Greenland sea it starts should melt quickly. Maybe not this year but most of thickness will probably melt before the freeze season.

The phrasing I used was not precise enough. Looking at submarine lidar scans of the underside of ice sheets show long spike like structures from above. I do not think anybody considers that part of thickness. What is less obvious is how to report the thickness of thicker flows frozen in place by thinner ice. If you are modeling volume a weighted average is appropriate. That is what I think Piomas tries to do. I don't know for certain what either model does. A few weeks back I noticed an area in the CAA that showed much thicker ice in Hycom than Piomas by about a meter or so. Almost everywhere Piomas shows thicker ice than Hycom. Correct me if I am wrong but I believe this is why Oren assumes Hycom must be an average. This location stood out to me. I looked at the location in Worldview and it was clearly thicker ice frozen in place by thinner ice. White chunks surrounded by darker ice thinning to see water underneath and cracking. My interpretation  is Piomas shows the average thickness and Hycom shows thickness of the thicker ice.  I have heard it before on this forum but I looked for and never found any such statement on Hycom. It made sense on an intuitive level that marine vessels would benefit more from over rather than under reporting of ice. That is not the only reason I think it is true. I have heard from some on the forum that really thick sea ice appears to melt quickly. That makes more sense if those are thick ice surrounded by thinner ice. Hycom still shows thin ice when at maximum magnification I see one small spec of ice in a much larger area. Any location that currently shows ice will show at least thin ice (0-5 cm thick) in their forecast. The nowcast mostly shows a thin ribbon of this really thin ice but always in small areas. Once you are aware of it it is easy to spot the difference between a nowcast and an 8 day forecast without looking at the dates. Hycom does not report a volume. If they were showing average thickness I would expect them to report a total volume.  These things together support the idea that the model is looking to over rather than under report thickness IMO.

[oren]

If you are modeling volume a weighted average is appropriate. That is what I think Piomas tries to do. I don't know for certain what either model does. A few weeks back I noticed an area in the CAA that showed much thicker ice in Hycom than Piomas by about a meter or so. Almost everywhere Piomas shows thicker ice than Hycom. Correct me if I am wrong but I believe this is why Oren assumes Hycom must be an average. This location stood out to me. I looked at the location in Worldview and it was clearly thicker ice frozen in place by thinner ice. White chunks surrounded by darker ice thinning to see water underneath and cracking. My interpretation  is Piomas shows the average thickness and Hycom shows thickness of the thicker ice.

Yes, PIOMAS models a weighted average (but divides the volume in a grid cell by its full cell area, not by just the ice area in the cell).
I assume Hycom is an average because it would surely have shown higher thickness if it were to show maximum thickness in a grid cell. Weirdly thick floes are found in various regions of the Arctic, and I am sure the 7m floe was not the only one. I can't prove it as I don't have the data or even the knowledge to back this up in detail.
That PIOMAS shows higher volume than Hycom could be a result of PIOMAS modeling error, Hycom modeling error, or both. I believe the main difference would be the PIOMAS modeling of extra-thick floes resulting from pressure ridges.
Some insight into the PIOMAS model can be gained from the following charts produced by Wipneus from time to time. The distribution shown could be wrong or right compared to physical reality of course. But the model calculated roughly 300-400k km2, and 3000 km3, of ice thicker than 4.2m, as of June 30th last year.

As some will remember, ice in each gridcell in the PIOMAS model is specified as a discrete distribution: there exist 12 categories of ice thickness (m):
[0.00, 0.26, 0.71, 1.46, 2.61, 4.23, 6.39, 9.10, 12.39, 16.24, 20.62, 25.49]

Another difference between the models is that as far as I am aware, Hycom does not cut off ice below 15% concentration, while PIOMAS does. Maybe Hycom does this for the reason that it is used by navy ships, so it gives maximum ice extent (but not maximum thickness as has been speculated).

As for the anomaly in the CAA, I am quite sure it was caused by anomalous behavior of NSIDC ice concentration data that at some point showed open water in the location, though in real time full ice cover (of very blue color) could be seen on Worldview. NSIDC is sensitive to melt water over the ice, and when the melt water is deep enough and widespread enough it does not "see" the ice in that grid cell. PIOMAS uses NSIDC concentration data as a constraint, so when NSIDC reports zero ice PIOMAS reports zero volume. Later when the melt water drained NSIDC brought the ice back, and PIOMAS brought only some of the volume back. In this case PIOMAS is clearly wrong.

[ajouis]

As some will remember, ice in each gridcell in the PIOMAS model is specified as a discrete distribution: there exist 12 categories of ice thickness (m):
[0.00, 0.26, 0.71, 1.46, 2.61, 4.23, 6.39, 9.10, 12.39, 16.24, 20.62, 25.49]

I think that piomas could do with a tightening of the scale, because apart from ice cap calving (which should not count towards piomas as it is not sea ice), I do not really see the use of the 10+ meters thickness

[oren]

From Wikipedia, bolding mine:

Pressure ridges are the thickest sea ice features and account for about one-half of the total sea ice volume.

One of the largest pressure ridges on record had a sail extending 12 metres  above the water surface, and a keel depth of 45 metres. The total thickness for a multiyear ridge was reported to be 40 metres. On average, total thickness ranges between 5 metres and 30 metres, with a mean sail height that remains below 2 metres.

I don't know much about the subject, and am unable to evaluate the validity of the quoted numbers, but it seems a volume model must take pressure ridges into account in order to properly estimate total volume.
The end-June >4.2m volume in the chart above was about 25% of total volume in 2019 according to PIOMAS, while in the past it reached 35%-40%, still less than the half cited by Wikipedia.

[ajouis]

Oren, I do not doubt the importance of pressure ridges, but it seems to me that given the difficulty of finding a 7 meter floe, and there already vanishing area on the map, the 10+ ones have just melted away, couldn’t find any source one way or the other thiugh 

[interstitial]

Oren:
[/color]Yeah I am not an expert either I am just trying to dig up answers in doing so I make assumptions. Your experience helps me to identify at least some of the flaws in my assumptions. I based my opinion of the Piomas model on the description I read and the low resolution image Wipneus supplies. These charts are new to me. Clearly the output by Wipneus is not the model output. Looks like I need to learn more about Piomas to make a reasonable assessment.
[/color]
[/color]Chart by thickness indicate about 5% of the area is thicker than 4.2 m I do not think the map shows that. The scale does not indicate ice thicker than 4 m. My flawed assumption was the average thickness of any grid cell was no greater than 4 m. Ice that was 7 m thick would have to be surrounded by enough thinner ice to bring the average down to 4 m thick.
[/color]
[/color]With pressure ridging creating 50% of volume as you site or 40% of volume as the article cited in the wiki changed to a 2000 paper. I am not sure which is newer or likely to be more accurate.
[/color]
[/color]I do not assume all pressure ridges must be thicker than 4.2 m. The citation that says “On average, total thickness ranges between 5 metres and 30 metres” leads to a book with a 2005 publication date. The book is pay walled but it probably cites an earlier article. This observation is at least as old as 2005 but likely before 2003. 
[/color][/size] [/font]
[/color][/size]In September 2003 when around 1.9 million km² was 4 years old or older. As compared to September 2019 when around 50,000 is 4 years old or older and [/font]
[/size]https://svs.gsfc.nasa.gov/4750 [/color]It is a different arctic. [/color]

[oren]

All good points interstitial. The only thing I know for sure is that PIOMAS calculates an average of the various thicknesses inside a grid cell, and the color shown on the map represents the average in that grid cell. The map shows no more than 4m, but this could be comprised of 10% ice that is 13m, and 90% ice that is 3m. Only someone who can analyze the raw PIOMAS output files can elaborate more than this, or show the distribution rather than just the average. I have long desired this technical ability, but I still lack it.
For example, one could produce a map showing concentration of ice at a given thickness or above. This map could be compared to common sense, to the ice age map by NSIDC and to findings "on the ground". Let's say the map said a certain grid cell had 10% of 13m ice. A research vessel or underwater robot could do statistical sampling and test this in real life.
Or one could produce a map showing concentration of ice <0.26m in each cell, and compare this to actual ice area drops a week or two later, assuming 26cm of thickness are lost in that time period. This could be a predictive tool, or help validate/invalidate the model.

It is very much possible that PIOMAS overestimates the amount of super-thick ice, as it was built at a time when such ice was more prevalent in the Arctic. However my limited knowledge both of PIOMAS and the actual state of the ice prevent me from making this judgement properly.

I remind this thread is about Hycom. Two types of questions arise, PIOMAS vs. Hycom, which we have discussed herein to the best of our abilities, but also Hycom itself, which I think we should try to address:
* Does Hycom model pressure ridges creation in winter and their evolution in summer? Is there a way to find out?
* Does Hycom have numerical outputs, besides the map and animation? Is there a chart of total Hycom volume? If so, how long does it go back in time?
* Does Hycom have a regional breakdown of data? (As Wipneus produces for PIOMAS from the gridded output files)

Assuming that Hycom has been greatly improved and is by now the better model, as has been hinted above, these questions could determine its usefulness within the forum context.

[interstitial]

* Does Hycom model pressure ridges creation in winter and their evolution in summer? Is there a way to find out?
* Does Hycom have numerical outputs, besides the map and animation? Is there a chart of total Hycom volume? If so, how long does it go back in time?
* Does Hycom have a regional breakdown of data? (As Wipneus produces for PIOMAS from the gridded output files)

No numerical data I can find of a reasonable size. Your welcome to download the 50+ Gb daily data file and try to extract it from that or convince someone else too. I have a decent internet connection but my internet provider would shut me down if I tried to do that.
Alternatively I have considered writing a script to count pixels for each color however I decided without accounting for map distortion it would be too inacurate. I thought about masking for rings and multiply some scaling factor for each thin ring. I would have to learn more about map types and distortions. I have an inkling of how to proceed but I am not sure I want to know that badly. My coding skills are slow and not great. I would spend more time learning to code than I would coding.

Hycom changes their model occasionally when they do they recalculate some historical data to judge performance. The Last time they went back to 2014.

Yes I get the thickness average idea. The problem is Piomas says they calculate each cell as if conditions were uniform inside the cell. Thats a mess.

[ArcticMelt2]

From Wikipedia, bolding mine:

Pressure ridges are the thickest sea ice features and account for about one-half of the total sea ice volume.

One of the largest pressure ridges on record had a sail extending 12 metres  above the water surface, and a keel depth of 45 metres. The total thickness for a multiyear ridge was reported to be 40 metres. On average, total thickness ranges between 5 metres and 30 metres, with a mean sail height that remains below 2 metres.

The compression ridge in the Hudson Bay is now clearly visible. How thick is the ice in it? 2-3 meters?

[johnm33]

Hycomsss main interest here is the outburst from 79N/Zacheriae

[interstitial]

<snip same gif as below>
Hycomsss main interest here is the outburst from 79N/Zacheriae

Is this salinity? for what days?

[blumenkraft]

Hycomsss = Hycom sea surface salinity i assume.

[johnm33]

Yes current sea surface salinity hind/forecast at time of posting so 03:07-01:08

[be cause]

Hi folks can you load gifs as click and play .. the page won't load for me .. too much going on .. b.c.

[blumenkraft]

Something Freegrass and i found out only recently: Externally hosted GIFs (via [ img ] tag linked files from other websites) will autoplay (automatically download), no matter the canvas size.

If you want your GIF to behave to forum rules (click to play) you have to upload it to the forum server as an attachment to your post.

A GIF should always be >500 pixels (in at least one direction) in order to save data traffic.

Very small GIFs (smaller than 1MB of file size) can be below this canvas size. It doesn't hurt too much traffic-wise if those would auto-play.

[oren]

Discuss Hycom and its reliability here.

[OffTheGrid]

Here are todays concentration and thickness charts.
And the same day in 2014 which is the first year the arctic only model was replaced with the global model with weather and ocean data incorporated for prediction purposes.
Does my memory correctly recall that 2014 was the year piomass was updated for the last time? Along with a quiet media release forgotten or missed by most that they believed the piomass model had understated the 2012 minimum, and they revised it upwards?
I think this may be a little charitable for concentration in the central and coastal Beaufort, judging by worldviews.

[OffTheGrid]

For comparisons, piomass monthly Thickness averages July/ Aug 2014.

Reverend milkbones 29 July 2012, thickness, also from the previous arctic only Hycom model. Still looking for Pio2012

[interstitial]

OffTheGrid
The concentration drops after the recent storm. I suspect that the difference may be in large part due to different timing. I haven’t checked but between time zones and such events labeled the same day could reasonably be almost 48 hours apart. I have included a 7/31 forecast for 8/1 concentration after the storm which seems like a more accurate picture now. The 7/28 concentration forecast for 7/29 does show a significant portion of the Beufort in the 50 to 60% range. To my eyes it shows most of the Beufort less than 85% concentration.

[OffTheGrid]

Unfortunate that Hycom has seemingly decided to censor the higher resolution Beaufort model. Its last run on the 24th of July looks like a darn good effort in advance of the accurate picture SMOS is presenting. Bearing in mind that the Soil Moisture/ Ocean Salinity sensor works on detecting the averages over a 40km square grid between the less salty ice surface and the horizon of slush soaked with seawater below.
Its low frequency radar sees straight through ice and water cloud, including fog. The maximum resolution of the higher frequency radars used by Jaxa, etc for their much more promoted extent and area metrics, is at the expense of being badly compromised by cloud effects. And they are still limited to about 500m resolution, so do not see open water below that breadth. A bigger and bigger problem in this season every year. Hycom is probably suffering from falling behind in the need to update algorithims to account for these rapidly crumbling models Its assimilating data from.
Ice crumbling to fields of dispersed slush is far different to solid sheets with crosshatchings of long leads.
Other than the Chukchi not getting the Beaufort arm shoved as far into it, and more ice surging into mckenzie bay from the CAA mega flow than Hycom predicted before the storm, good predictive skill I think. Backed up by what satellite visuals we've had through the clouds and fog.
Still have to click to animate the absurdly small smos gif.

[interstitial]

The most recent update I see is May 2020 so I think they update whenever they identify something that needs fixing. I really don't think Hycom incorporates data from other models. It is a global oceanic model coupled with atmospheric forcing from their own atmospheric model. it does use a concentration map but it does not seem to suffer from the same anomalies from storms. They may use the same raw data and filter it I really do not know at this time. I am doing a deeper dive into the system. My real frustration is output data runs about 50 gb a day. I have located some smaller subsets of data but the smallest is 380 mb a day sea surface height.

[JNap]

Here is the current August 3rd run from Hycom.  The first pic is the current state of the ice.  The next one is their forecast for the ice on August 11th.  Also attached links for people to navigate there themselves.

https://www7320.nrlssc.navy.mil/GLBhycomcice1-12/navo/arcticict/nowcast/ict2020080312_2020080400_930_arcticict.001.gif

https://www7320.nrlssc.navy.mil/GLBhycomcice1-12/navo/arcticict/nowcast/ict2020080312_2020081100_930_arcticict.001.gif

[glennbuck]

While HYCOM seems accurate around North Greenland for the 11th August. It has 2-2.5 Metre thick ice where it is open sea in the Northern coast of Ellesmere Island in northern Canada, near the Ward Hunt Ice Shelf in the Arctic Ocean, across for 150 miles west on the 6th August!

[Rod]

I really have my doubts about that model.

Look at all the ice that it says is less than 1m thick.  If it is correct, then we just need to average about 3cm per day, top and bottom melt combined, to have a BOE (defined as less than 1 million square km) in about a month.

That seems unlikely.

[interstitial]

Glenbuck:
Please do not confuse concentration and thickness they are only very loosely correlated. Concentration does not care about thickness it measures only what fraction of ice to open water. 1 meter thick 10 meters thick concentration makes no distinction.

[interstitial]

Rod:
It is ok to doubt the model. I do not I doubt the model I doubt your melt assumptions. I have seen melt numbers estimating 5cm a day this season but that seems extremely high to me. At that rate 2 m melts in 40 days. I think we can agree when thick ice gets transported out of the region melt rates are higher than melting in place but we can mostly ignore that.
Looking at piomas model I see ice thinning by 1 m between 4/15 and 7/15. In less central locations the ice thins 1.25 m while in more central locations the ice thins by 0.75m. That is a melt rate of 8mm to 14mm per day. I also looked at Hycom on the same dates and came up with similar numbers. 7/15 includes about 4 weeks after the solstice. We are now almost three weeks beyond that. We are getting close to the time when the CAB normally starts to refreeze. The Beufort will probably melt to the end.
It has been said in past melt seasons that there is enough heat in arctic ocean to melt all of the ice if it mixed thoroughly. The melting along the Greenland CAA coast seems to be from Eckmen pumping of heat from lower down. The right storm could also do it. One of the buoys shows freshwater protection breaking down. This may or may not be happening elsewhere. There was a lot of heat in the ocean in 2019 that did not do much at the end of last season. It could have gone the other way.
I don’t think we are getting a boe this year but looking at that thickness map and saying it will result in a boe is your interpretation and not mine. By the numbers the CAB loses another 400k km^2 of area unless something unusual happens. Even so If all the other seas melts out completely, which has not happened before, the central arctic would still need to lose another 1.6 million km^2 of area in addition to the 400k km^2. The region in which the ice is in makes a huge difference. By the numbers I would expect the Beufort to end with 500k km^2 but due to its location and thickness and concentration 250k km^2 seems likely.
By the way I am using 3.75km area numbers, which I think come from NSIDC, for my estimates. They will be different for other measurements. The more seasons I participate in the more nuances I see in the data. 

[interstitial]

The Hycom model grabs data from all the major cryosphere satellites and buoys and incorporates it daily. I also learned that ice thickness is mission critical because US nuclear subs often hide in the arctic under the ice but their wakes (Is it a wake on a submarine or something else?) are clearly evident to satellites when the ice thins. They do not mention what the critical thickness is but they do indicate the ice is too thin for much of the summer and leaves them vulnerable because the area under thick ice is too small to hide. This next bit is pure speculation on my part. I have previously commented on how the color scale clearly shows what ice is thicker or thinner than 135 cm. If I were a betting man I would suggest that is the critical thickness.

[oren]

Looking at piomas model I see ice thinning by 1 m between 4/15 and 7/15. In less central locations the ice thins 1.25 m while in more central locations the ice thins by 0.75m. That is a melt rate of 8mm to 14mm per day. I also looked at Hycom on the same dates and came up with similar numbers. 7/15 includes about 4 weeks after the solstice. We are now almost three weeks beyond that. We are getting close to the time when the CAB normally starts to refreeze. The Beaufort will probably melt to the end.

By the way I am using 3.75km area numbers, which I think come from NSIDC, for my estimates. They will be different for other measurements. The more seasons I participate in the more nuances I see in the data.

Thanks for your interesting post interstitial. A few notes:
* The CAB continues to accumulate volume until May 5th-10th, at least according to PIOMAS.
* Peak volume loss rate is between mid-June and end-July. Using Wipneus regional numbers, an area of ~4M km2 loses ~4000 km3, so about 1m over 6 weeks, roughly 2.5 cm/day. Of course this mixes up top melt, bottom melt and export/import.
* CAB normally starts actual freezing in early to mid-Sep, so at least a month away. Earlier we can see effects of melt water freezing on existing ice surfaces.
* If you mean 3.125km grid, it comes from UH (University of Hamburg) AMSR2 data. NSIDC uses a 25km grid.

[Frivolousz21]

Glenbuck:
Please do not confuse concentration and thickness they are only very loosely correlated. Concentration does not care about thickness it measures only what fraction of ice to open water. 1 meter thick 10 meters thick concentration makes no distinction.

It looks really good.

If we had a major dipole we would crush 2012.

The ice is very thin all over.

This year is ending just like  2011

[interstitial]

Oren:
That is later than thought for volume but makes sense because the 4/15 date was selected to be prior to any melting so I had the images handy.
Thanks maybe I will try again with different dates but not today.
2.5 cm/day for 6 weeks around the solstice seems more reasonable than 3 cm/day from now until a month from now.

[interstitial]

Oren:
This response may not be well put together but it is informed.
JAXA produces their own 25km sea ice extent numbers
The NSDIC  has a 24 km^2 sea ice area and extent product
The NSDIC also has a 4km and 1km Multi-sensor analyzed sea ice extent (MASIE) products
All NSDIC data is archived at University of Colorado Boulder website
MASIE is produced primarily by NOAA utilizing primarily US satelites.
Japanese, US and EU share raw environmental satellite data with each other and coordinate instruments and coverage to maximize resources.
National Oceanic and Atmospheric Administration (NOAA US) operates a series of Polar and Geostationary Operational Environmental Satellites (POES/GOES US) with visible and infrared data. These are supplemented with visible data from Geostationary Meteorological Satellite (GMS Japan), Multifunction Transport Satellite (MTSAT Japan) and Meteorological Satellite (METEOSAT EU) data.
Microwave products from Defense Meteorological Satellite Program (DMSP US) and Advanced Microwave Scanning Radiometer (ASMR-E Japan). 
Synthetic Aperature Radar imagery (from?), surface observations and National Weather Service (NWS US) weather prediction models and whatever other data they can find stuck in the couch cushions are also used.
The above data is used by NOAA analysts to create the Interactive multisensor snow and ice mapping system (IMS US).
NOAA then combines DMSP, AMSR2 and IMS data utilizing an algorithm developed by University of Hamburg researchers to create a 4km and 1km Multi-sensor Analyzed Sea Ice Extent (MASIE US) product. This data is archived at University of Colorado as an NSIDC product and apparently at University of Hamburg.
I would also note that cooperation among NASA, JAXA and ESA on environmental satellites is considerable and will likely increase going forward from what I could tell so assigning countries or organizations is not entirely accurate either.
IMS is an ongoing process utilizing many NOAA algorithms and daily manual interpretation by NOAA analysts. Predominately US satellites and one Japanese satellite are used for data with some data gaps filled in by ESA and JAXA satellites. Data from Japanese and European satellites is processed from raw data with NOAA algorithms.
Looking at University of Hamburg website
 
Beitsch, A.; Kaleschke, L.; Kern, S. developed an algorithm that combined high resolution visible data with lower resolution microwave data in a way that produced a higher resolution product than the old algorithm. Kudos to them they moved the science forward a bit. They did not write the actual algorithm used by NSDIC to produce either the 4km or 1km product.
University of Hamburg researchers use raw data from AMSR2 and AMSR-E to produce a distinct 3.125 km product that would rightly be referred to as a U of Hamburg product.