Home brew AMSR2 extent & area calculation

[A-Team]

We need the netCDF to open in Panoply the file as a geo2D. If it opens merely as 2D, we cannot make the stereographic or equal area projection maps. So I guess that means lat,lon rather than x,y. These two do not open a geo2D in Panoply, just a 2D:

ftp://ftp.awi.de/sea_ice/product/amsr2/v100/nh/2020/08/nh_20200818_21_SIC_12h.nc.gz
ftp://ftp.awi.de/sea_ice/product/amsr2/v100/nh/2020/08/nh_20200818_21_SIC_72h.nc.gz

Dr Schmunk of NASA GISS, the sole creator of Panoply, told me Panoply will not ever support the conversion of x,y to geo-registered. Since Panoply does implement netCDFdump in its export, this could presumably be done elsewhere in command line.

Earlier, i was able to use an online converter (a GDAL command in web browser) to change a 'latest' tiff file to a netCDF geo2D. This is better done at your end!

Note reducing land/water edge pixels will really improve the products we make here. Also the two new filters that reduce erroneous weather pixels will make  improvements. Our 'darken only' filter is similar but not likely as good as the 'latitude-dependent minimum' and 'time-domain median' filters.

There is no citation describing these filters. Can that be added to the ReadMe?  .

Any chance you could post an example of before-and-after affects of these two filters?

I could not confirm 3125 m grid cell size. Or rather, the tiff image is 5000 m pixels, derived from direct comparison to UHH LARGE which I'm assuming is 3125m.
 
I am just testing the downloads now from the current directory ...

ftp://ftp.awi.de/sea_ice/product/amsr2/v100/nh/2020/08/

Here are the highlights (a few typos fixed) from the upper ReadMe file which people might not see:

AWI AMSR2 ice concentration product v100 25.8.2020
Contact: lars.kaleschke aatt awi.de

The AWI AMSR2 ice concentration product is based on previous developments at the University of Bremen (UB) and the University of Hamburg (UHH) [1,2,3]. The level 1 brightness temperature data are provided by the Japan Aerospace Exploration Agency’s (JAXA) Advanced Microwave Scanning Radiometer 2 (AMSR2) [4]. The AWI AMSR2 product is under continuos development and feedback is appreciated.

Reprocessing of the full AMSR2 time series is planned as soon as the data product and format is finally settled. The earliest at UHH is 01 Aug 2012 to 30 Sept 2020. The new AWI archive currently has Aug 18 through Aug 24.

A) FAST product

The product is generated twice daily for two different median times (filenames with _09_ or _21_)
09:00 +- 9h
21:00 +- 9h

After 18 hours the spatial coverage is almost complete. The twice daily (file name includes _12h) product has data gaps and errors due to weather effects and along the coastline.

B) CLEAN product

This product takes into account the latest 72 hours of data (file name includes _72h). The coastline is extended by about one grid cell and masks the frequently occurring erroneous coastal overestimation. Time-domain median- and latitude-dependent minimum-filtering reduces erroneous weather pixels.

The land mask is derived from https://osmdata.openstreetmap.de/data/

An obvious difference to previous (UB and UHH) products is a correct representation of northeast Greenland, e.g. the Danmark Fjord, which was wrong in the previous land mask derived from The Global Self-consistent, Hierarchical, High-resolution Geography Database (GSHHG).

A very high resolution map 058 x 3641 of the fjord and NE Greenland 3can be found at wikipedia:

https://upload.wikimedia.org/wikipedia/commons/2/29/North_Greenland_section-txu-pclmaps-oclc-8322829_a_1.jpg

NetCDF and GeoTIFF formats provided

Projection and grid
-------------------
https://epsg.io/3411 NSIDC Sea Ice Polar Stereographic North
https://epsg.io/3412 NSIDC Sea Ice Polar Stereographic South
Grid cell size 3125 meter

Data distribution: free without warranty
ftp://ftp.awi.de/sea_ice/product/amsr2/

References:

[1] L. Kaleschke, C. Lüpkes, T. Vihma, J. Haarpaintner, A. Bochert, J. Hartmann & G. Heygster (2001) SSM/I Sea Ice Remote Sensing for Mesoscale Ocean-Atmosphere Interaction Analysis, Canadian Journal of Remote Sensing, 27:5, 526-537, DOI:10.1080/07038992.2001.10854892

[2] Spreen, G., L. Kaleschke, and G. Heygster (2008), Sea ice remote sensing using AMSR-E 89-GHz channels, J. Geophys. Res., 113, C02S03, doi:10.1029/2005JC003384.

[3] Beitsch, A.; Kaleschke, L.; Kern, S. Investigating High-Resolution AMSR2 Sea Ice Concentrations during the February 2013 Fracture Event in the Beaufort Sea. Remote Sens. 2014, 6, 3841-3856. https://doi.org/10.3390/rs6053841

[4] https://gportal.jaxa.jp/gpr/

/*/*/*/*/*/*/*/*

Wiki history: In May 1907 Mylius-Erichsen entered the unknown Danmark Fjord with his three-dogsled exploration team deeming it would be leading him to the Navy Cliff and the Peary Channel. The team, which included cartographer Niels Peter Høeg Hagen and dogsled expert Jørgen Brønlund, traveled southwestwards until the head of the fjord and, becoming aware that it was a dead end, they backtracked to the northeast.

By the end of May Mylius-Erichsen's team was back again at the mouth of the fjord. As they met Johan Peter Koch's northern team at Cape Rigsdagen, already on their way back from Cape Bridgman, Mylius-Erichsen realized that they had wasted precious time and provisions by entering the long unexplored fjord. The delay would eventually lead the three men to their death as they pressed westward along the southern shore of Independence Fjord instead of returning to the ship.

[A-Team]

The images below compare the new and old ('latest') versions of Aug 18th. There is quite a bit of ice movement between them so perhaps the date we assigned to 'latest' is off.

Can you please clarify how the 72 hour how the time domain filter is applied without a model of how the ice has moved? OsiSaf will have purple arrows on the peripheral ice where motion is most pronounced.

Also I did not understand the explanation of how coastal artifacts were minimized. It seems that some issues remain (AK coast shown below).

The bottom image compares the 09, 21 and 72 of the nh_today between Greenland and the pole. The latter two are pixel-perfect identical. The 09 has some subtle differences as expected.

ftp://ftp.awi.de/sea_ice/product/amsr2/v100/nh/2020/08/nh_20200818_21_SIC_72h.tiff

[Jim Hunt]

Thanks for the feedback! I tried to reproduce this. However, with Panoply 4.11.6 I can view the data (attached).

Having downloaded 4.11.6 I can now see the data.

Another question: do you need the latitude, longitude grid included in the NetCDF file? Or are you fine with X,Y? Could the lat, lon grid be provided in a separate file? Including lat, lon would increase the file size.

I'm used to having lat/long included in a single NetCDF. Is there a way to combine that info from a different file with the existing data in Panoply?

[A-Team]

Having downloaded 4.11.6 I can now see the data.

It's not good enough just to 'see the data'. If it is not geo-referenced geo2D, it won't make the maps we want. I just downloaded 4.11.6 of 13 Aug 2020 and it is still showing 2D, not geo2D.

Two possible embeddings of the palette are shown. We would prefer them covering land near the Arctic Ocean rather than at the bottom like on UHH. They could be tightened slightly by managing the last 7 tiles better. The central pole hole gray and land green are not needed in the palette.

Note the forum does not currently allow .tiff files to be uploaded or displayed so i saved it out as png. If we have to use the GDAL command to make proper netCDF geo2D, that will not work off the png as it does not retain the geo-referencing.

Here is where to go for that:

https://mygeodata.cloud/

[seaice.de]

Having downloaded 4.11.6 I can now see the data.

It's not good enough just to 'see the data'. If it is not geo-referenced geo2D, it won't make the maps we want. I just downloaded 4.11.6 of 13 Aug 2020 and it is still showing 2D, not geo2D.

Ok, thanks, I have included the latitude and longitude. Now you can do whatever projection you like. 

[A-Team]

any projection u want

That's great. (Not sure how often I will be using that particular one though.) Panoply does have that lambert conformal conic but I'm not sure about the exact NSIDC EASE grid nor whether there are any options to import that pj. It's easy though to import palettes (such as the one here) but no way to distribute that.

Panoply also supports netCDF animations. There is a way of folding many days worth of data into a single file with scrolling between dates. The end user sets up a map on one date, then selects a data range and exports a gif at a specified playing speed. I don't know how the initial file is made, netCDF java library maybe.

This is so timely, given we're in the last 5-6 weeks of melt season with the Polarstern currently moored out there at 87.8  104.3 on 20-08-25 at 16:00. It's completely overcast with the cloud deck only at 100-200 m but 1-2 km visibility for bear guards though foggy with light snow above old ice with few leads and unspecified drained/undrained melt pond coverages.

The shear they are measuring has been very much in the forum news because of 2020 mooring papers from Polyakov et al that suggest that a tipping point (runaway halocline loss, atlantification) is developing.

increased coupling between AW heat and the sea ice may lead to a positive feedback between reduced sea ice and higher mixing rates as the longer periods and increased areal extent of open water facilitate more energetic wind‐driven inertial oscillations (and, potentially, less damping of baroclinic tidal currents) and associated upper‐ocean shear coinciding with weakening halocline stratification. As sea ice declines, a new Arctic state is emerging which, due to the positive feedback mechanism outlined above, may be pushing the system toward a tipping point.

Weakening of Cold Halocline Layer Exposes Sea Ice to Oceanic Heat in the Eastern Arctic Ocean
https://doi.org/10.1175/JCLI-D-19-0976.1

Intensification of Near‐Surface Currents and Shear in the Eastern Arctic Ocean
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020GL089469

https://www.essoar.org/doi/abs/10.1002/essoar.10502530.1
https://www.nature.com/articles/s41597-020-00578-z

https://mirjamglessmer.com/2019/02/14/measuring-turbulence-with-a-microstructure-sonde/
https://www.sea-sun-tech.com/product/microstructure-probe-mss-90/
ftp://ftp.awi.de/sea_ice/product/amsr2/v100/nh/2020/08/nh_today_21_SIC_72h.tiff

[A-Team]

The image below shows the 15-80% marginal ice zone (MIZ) as defined by NSIDC colored in stages in today's AMSR2_AWI, namely nh_20200826_21_SIC_72h.tiff. This is easily done on a time series after tiling the 9 dates available so far.

Nullschool offers a wave product but one that comes with a huge pole hole and so misses the ice edge in most places currently. The new article below explores using the altimeter on IceSat-2 to actually visualize ocean swells passing through the ice interior (which we know can be completely devastating .. see Jim Hunt's collection of incidents).

Observing waves in sea ice with ICESat-2
C. Horvat, Ed Blanchard-Wrigglesworth, A. Petty
Aug 2020 doi: 10.1029/2020GL087629

"Wave-ice interactions create the marginal ice zone (MIZ), a region critically important for ecology, transportation, and the polar energy budget. Typically, the MIZ is defined using satellite products as those regions where sea ice concentration is,between 15% and 80%.

Ocean waves, whether locally-generated wind waves or long-period swell waves, are increasingly observed in the Arctic and are a persistent feature of Southern Ocean sea ice. In the presence of ocean waves, sea ice floes bend, and if the wave field is sufficiently energetic, will break  impacting the sea ice floe size distribution (FSD) without significantly changing sea ice concentration or thickness.

Fractured sea ice is mobile and susceptible to lateral melting and so a feedback between ocean surface waves, the FSD, and floe melting show clear evidence that IS-2 can detect waves in sea ice.

We begin by studying a case study of a swell event in the Barents sea. Figure 1a plots the trajectory of a descending IceSat-2 profile of the Barents Sea region on March 22, 2019, the significant wave height (SWH, the average height of the 1/3 highest waves on 3/22 from NOAA’s Wavewatch III production hindcast (WW3), and the 15% and 85% ice concentration contours from the NASA Team passive microwave product (PM-SIC). To verify the presence of ice we also compared to the SMOS sea ice thickness product with similar results.

On 3/22 a large cyclonic system drifted toward the Barents and interacted with a developing low pressure system north of Svalbard, producing high surface westerly winds of approximately 60-100 km/h over the Norwegian Sea. This storm resulted in eastward-propagating surface waves with SWHs of up to 10 m."

[igs]

Nullschool offers a wave product but one that comes with a huge pole hole and so misses the ice edge in most places currently. The new article below explores using the altimeter on IceSat-2 to actually visualize ocean swells passing through the ice interior (which we know can be completely devastating .. see Jim Hunt's collection of incidents).

Windy's Pole-Hole is significantly smaller for similar results and then it' seven easier to "travel" in time.

https://www.windy.com/-Waves-waves?waves,66.653,-14.150,3,i:pressure

[A-Team]

windy?

Windy dropped polar-sterographic Arctic coverage quite a while back. Said there was insufficient interest. On satellites in near-polar orbit (wanted for sun-synchronous planetary coverage), the phrase 'pole hole' refers to the area around the north pole where the satellite cannot gather data because its swath is too narrow. Usually it is as small cap, often colored gray, covering say 88-90º as with Sentinel-1AB, the smaller the better.

These swells need a long reach and consistent winds to develop, then hit the ice edge head on to really have an effect. Astonishingly the ice pack can be fractured hundreds of km in.

Below I made a new grayscale palette (LUT) for AMSR2_AWI that is better suited for graphical computation (e subtraction). It is easy to make exactly what you want in ImageJ which has excellent export/import of plain text file LUTs which can be built to order in spreadsheet using its fill-down.

Here i started at 28,28,28 and bumped by 2,2,2 the one hundred steps of 1% concentration in the original AMSR2_AWI had been built to get somethng more consistently spaced than AMSR2_AWI  converted to grayscale. File --> Import will apply any LUT on your desktop to any open 8-bit image. For the second time series, one of the stock lookup tables (5 ramps) was applied because it highlighted the mixed ice zone (MIZ) that is coupled to open water waves and swells.

ImageJ will also save out (list the RGB numbers in a text table) any special purpose color edits made to highlight features in AMSR2_AWI so they can be done again later. The new AMSR2 comes with a very nice palette to begin with but it cannot anticipate what someone wants to emphacize.

[A-Team]

There's been a major quality upgrade today to AMSR2_AWI, causing the files to be placed in a new location and the version 100 archive to vanish. No surprise: continuous improvements were promised already in the initial ReadMe. So far, versioning changes have not been listed (in the manner of Panoply).

ftp://ftp.awi.de/sea_ice/product/amsr2/v101/nh/2020/08/

Only Aug 25-26 are available so far in v101. While the whole AMSR2 satellite record will be re-processed, that massive compute will not plausibly happen until the algorithm stabilizes.

The pole hole still has even x even dimensions (56x56 pixels), meaning no central pixel defines the north pole. This will cause unnecessary error in constructing a lat lon graticule overlay. Mercifully that is not irrevocably embedded on top of precious data the way it is in AMSR2_UHH.

The graticule should be an undithered plain black and white file stored next to the ReadMe and fit right over the uncropped archive images. End users can replace the white with transparency with a click at their end and soften the black before dropping it. (ImageJ implements the alpha channel differently, through Edit -> Paste Control.)

The two dates available for v101 are differenced below. The main action, as expected, is in the Beaufort, Chukchi poleward extension, by SZ, north of Greenland and below the pole. Bulk motion of the ice pack towards the Pacific side is quite noticeable.

Highlighted MIZ areas are shown in the full size animated gif; some shrinkage overnight is apparent, The attached auto-palette (LUT) can instantly modify new images in coming days (ImageJ only).

Nullschool GFS anticipates some significant winds in parts of this marginal ice zone that could hasten fragmentation, over-washing, dispersion and melt in the remnant Beaufort and southern Laptev. Wind direction is hard to see (look for sprite heads) but is dispersive in the former and head-on in the latter.

Hycom is showing a lot of dubious ice still around the periphery. The 4th image cuts this down to the more accurate ice edge boundaries of AMSR2_AWI, giving a thickness over concentration relationship.

After rescaling a nullschool screenshot to fit, 'lighten only' will display the winds over black open water; after selecting land and ice on the AMSR2_AWI of the 26th, those portions of the nullschool layer can be deleted.

https://earth.nullschool.net/#current/wind/isobaric/1000hPa/orthographic=-45.00,92.00,1600/loc=107.200,87.900

[glennbuck]

Hycom is showing a lot of dubious ice still around the periphery. The 4th image cuts this down to the more accurate ice edge boundaries of AMSR2_AWI, giving a thickness over concentration relationship.

After rescaling a nullschool screenshot to fit, 'lighten only' will display the winds over black open water; after selecting land and ice on the AMSR2_AWI of the 26th, those portions of the nullschool layer can be deleted.

https://earth.nullschool.net/#current/wind/isobaric/1000hPa/orthographic=-45.00,92.00,1600/loc=107.200,87.900

Not sure about the HYCOM ice thickness in the 4th map of 2 metres near the Ward Hunt ice shelf,  Worldview on 25th August. The AMSR2 maps seem right for the area.

[A-Team]

I checked v101 to see if anything had changed in the color scheme. Yes, a new color has emerged but is never used in the image itself. The land and pole hole pixels have been displaced to the right, at least as ImageJ interprets the LUT (with pure black squares removed).

The first row is proving too dark to resolve isolated remnant ice. Pure black should not be used for open water but rather be reserved a unique color of its own, say 1,1,1 rather than 0,0,0 (better saved for text overlays).

@zLabe uses a clever 6x7 double ramp color scheme enabling a self-explanatory and effective display of line graphs of satellite data from 1979 to 2020, with earlier years deprecated. However ice concentration data comes more in the form of scattered dots rather than lines so this approach does not work here.

Looking now at what AMSR_AWI is actually resolving close to no ice, there are currently very few pixels in the 1-14% concentration range. They are incoherently located with respect to each other so resolving them into individual or color groups just produces distracting noise. These are shown lumped below.

The coloration suggests the western pink cluster will be the next to melt out in the Beaufort leaving a doughnut hole that may or may not make it to freeze season. The hycom model supports this somewhat but shows most of week's melt ahead over on the atlantification side.

The magic rescaling number is 546.01% to get hycom to fit over AMSR2_AWI. As interstitial notes, excess ice can be done by cropping out to the latter's open water.

[interstitial]

Hycom shows thickness when ice covers 1% of the square. On worldview that looks like no ice. I do like the idea of using AWI concentration data on Hycom to remove peripheral ice.

[Jim Hunt]

Observing waves in sea ice with ICESat-2

Thanks for the heads up A-Team. I hadn't spotted that paper before now.

The swell described in there counts as one of my "incidents":

https://GreatWhiteCon.info/2019/03/facts-about-the-arctic-in-march-2019/

That swell had a peak period of around 15 seconds. I don't see anything in the current forecast of that magnitude, although the Fram Strait gets closest in the not too distant future:

[Jim Hunt]

P.S.

https://GreatWhiteCon.info/2020/08/facts-about-the-arctic-in-august-2020/#comment-355046

"Waves in ice" gets a mention in a long article in Science about declining Arctic sea ice.

So does the "warm blob" (mostly) 80 metres down and assorted other topical research.

[A-Team]

Hycom shows thickness when ice covers 1% of the square. On worldview that looks like no ice. I do like the idea of using AWI concentration data on Hycom to remove peripheral ice.

Right, far too generous in showing thin ice and so the ice edge. There is a method though for correcting hycom which consists of clicking on the lower palette color bins with the colorpicker set to radius 10 (as hycom isn't made correctly as a graphic).

Replacing with 'ocean gray' then removes the dodgy ice throughout the image. That is done in 5 stages for the 04 Sep 2020 forecast below, which brings the minimum thickness shown to 0.25m. Below that is really a job for Smos/Smap (but not this time of year).

Watch the bottom of the palette fill up with ocean gray as the ice edge recedes in the thickness portion of the image. Current sea ice concentration is shown in the second image along with the predicted shrinkage by Sept 4th in the final image. The Beaufort is going but not yet gone, the Laptev is still creeping in, but the main change is continuing shrinkage along the Sv-FJL-SZ line.

[uniquorn]

pixel count of the 12hr 2100 v101 compared to the v1. A one day gap has been left to make the distinction clear.
The increase in very low concentration pixels may be due to dispersion/less cloud cover or a processing change.
edit: could be melt

[A-Team]

The new AMSR2_AWI is up to version 102 today and has the early Sept dates but slightly changed per swath hour as noted in the new ReadMe.

Because the ice pack has been moving quite rapidly from SZ towards Banks for over a week, there are some distinct advantages to the 14 hour assemblies. Thus the 12 12 is probably more accurate than the 12 72 for now.

Hycom foresees ice pack motion continuing quite dramatically through Sept 9th. This is fully attributable to the persistent anti-cyclonic winds coming off central Siberia and the low mass of late season ice.

The pixel count of open water increases by about 10% over this time frame, an undetermined combination of melt, compaction and dispersion to sub-display resolution. There appears to be two fixed pivot points, one in the upper Fram and the other in the ESS at 80º 150º.

It can have some gaps in coverage (swath gray at the pole hole and periphery) just like we see with Ascat. These can be repaired by layering over the 72 hr, adding an alpha channel, selecting gray and deleting to let   the 72 show through but only where needed, then capture as 'new from visible'.

An example of v102 quality is attached below for Sept 2nd. The concentration palette has not changed. A small substitution palette is also included for those who wish to highlight specific ranges. Put a pair in the foreground/background and simply click on squares in the original palette (and thus image data) and fill.

Note individual remnant floes in the Beaufort can clearly be seen; the anomalous ice south of SevZem is finally dispersing/melting out.

ftp://ftp.awi.de/sea_ice/product/amsr2/v102/nh/2020/09/

The FAST product is generated twice daily for two different UTC start times: filenames with _00_ available in the afternoon or _12_ available in the morning

After 14 hours the satellite's spatial coverage is almost complete.  The twice daily (filname includes _12h) product has data gaps and errors due to weather and coastline effects.

The advantage of this twice daily processing is no time differences larger than 14 hours in swath assembly. The 24 hour difference in traditional daily averaged products can cause displacement artifacts due to ice drift. This effect can be observed along a moving ice edge which doubles up in the 24 hours average.

B) The CLEAN product takes into account the latest 72 hours of data indicated in the file name by _72h. The coastline is extended by about one grid cell and masks the frequently erroneous coastal overestimation of sea ice concentration.

Time-domain median- and latitude-dependent minimum-filtering reduces erroneous weather pixels. Moving isolated ice floes in open water are also removed by the filtering, a negative.
 
The grid cell size remains at 3.125 km though the image pixels are at an intermediate scale of 5.0 km

[gandul]

The new AMSR2_AWI is up to version 102 today and has the early Sept dates but slightly changed per swath hour as noted in the new ReadMe.

Because the ice pack has been moving quite rapidly from SZ towards Banks for over a week, there are some distinct advantages to the 14 hour assemblies. Thus the 12 12 is probably more accurate than the 12 72 for now.

Hycom foresees ice pack motion continuing quite dramatically through Sept 9th. This is fully attributable to the persistent anti-cyclonic winds coming off central Siberia and the low mass of late season ice.

The pixel count of open water increases by about 10% over this time frame, an undetermined combination of melt, compaction and dispersion to sub-display resolution. There appears to be two fixed pivot points, one in the upper Fram and the other in the ESS at 80º 150º.

So you demonstrate the HYCOM model to be complete crap in one thread and use it to illustrate the ice pack edge displacement in another, even when:
- it is predicting half the thickness and therefore it is predicting half the inertia of a given ice element, and half the coriolis force.
- it is predicting half the thickness which implies the model may be making disappear ice edge elements that in reality will persist.
- it is predicting half the thickness and therefore mobility internal forces between the ice elements that may be as fictitious.
- it is inferring a drag and resistance that can differ quite significantly from the real floes. But anyway does not matter, the balance of forces is already botched.

<> moderators are discouraging the use of these models, HYCOM and DMI for obvious reasons <>

<Removed antagonistic language. O>

[interstitial]

So you demonstrate the HYCOM model to be complete crap in one thread and use it

He made the claim but the thickest ice flow they could find does not demonstrate a model failure. It represents a stated selection bias by the mosaic team.

[oren]

I discourage the use of DMI thickness/volume because it appears to have very serious inconsistencies. Regarding Hycom thickness it appears the model has been much improved, and the jury is still out.

And I agree with interstitial that the Polarstern floe measurements do not provide typical thickness of surrounding floes, but rather thickness measurements of a selected thicker floe, and thus do not invalidate Hycom per se.

[jdallen]

The new AMSR2_AWI is up to version 102 today and has the early Sept dates but slightly changed per swath hour as noted in the new ReadMe.
<snip>

So you demonstrate the HYCOM model to be complete crap in one thread and use it to illustrate the ice pack edge displacement in another, even when:
<snip>

Thank you for stepping in Oren.

Now, to put some context into this, models can still be useful even if flawed, if they are consistent.  HYCOM I think fits into this category.  Seeing directly how they depart from direct observation is useful, and I think A-Team has been pretty clear about that. It's not an either/or proposition.  One can be critical on one hand about it's overall accuracy and still find information it provides useful.  The two things are not mutually exclusive.

Second, since 2013, the Arctic has pretty much been reduced to a bowl of ice cubes (MYI) surrounded by slush (interstitial lead ice that forms between MYI fragments).  It is utterly unsurprising that Polarstern could find a chunk of relatively "thick" MYI in the middle of this.  This doesn't make the HYCOM model more or less relevant.

When you are dealing with models and sensors that are working in multi-kilometer resolution, it is not just easy, but pretty much *certain* that you will find serious localized exceptions to what a model might be reporting... unless you are talking about 100% open water.

[interstitial]

To be clear I agree with most of what A-Team says about the models.While I have been corrected (mostly by Oren) about what Hycom thickness model says, after investigating, I find the Hycom model is more accurate than  others.
1)The lack of a 15% concentration cutoff confuses many into thinking it is inaccurate. That is annoying but not inaccurate.
2)The lack of volume numbers makes it less usable for comparison but is not indicative of a problem with accuracy.
3)Piomass shows thicker ice than Hycom in many places. The old regional Hycom model showed similar thickness to Piomass. The timing more than the change to global model was probably more important. The new global Hycom model shows much thinner ice. In my opinion that reflects the change in the arctic when ice > 4 years old melted. I also think mosaic and buoy data support Hycom more than other models. Piomass showed 3 meter thick ice for a significant portion of the trip to the north pole. Hycom showed less than 0.8 meters for most of the trip. Given their speed and description of conditions the ice was not 3 meters thick.

[gandul]

To be clear I agree with most of what A-Team says about the models.While I have been corrected (mostly by Oren) about what Hycom thickness model says, after investigating, I find the Hycom model is more accurate than  others.
1)The lack of a 15% concentration cutoff confuses many into thinking it is inaccurate. That is annoying but not inaccurate.
2)The lack of volume numbers makes it less usable for comparison but is not indicative of a problem with accuracy.
3)Piomass shows thicker ice than Hycom in many places. The old regional Hycom model showed similar thickness to Piomass. The timing more than the change to global model was probably more important. The new global Hycom model shows much thinner ice. In my opinion that reflects the change in the arctic when ice > 4 years old melted. I also think mosaic and buoy data support Hycom more than other models. Piomass showed 3 meter thick ice for a significant portion of the trip to the north pole. Hycom showed less than 0.8 meters for most of the trip. Given their speed and description of conditions the ice was not 3 meters thick.

That’s fair. To end this off topic from my side, just compare Hycom to Cryosat2+SMOS measurements of thickness during winter.

[A-Team]

The first image below overlays the thicker portion of SMOS-SMAP ice thinness of Sept 3rd over AMSR2_AWI of the same date. While this is 'too early' for this satellite product to approach its peak accuracy, it does seem that the >0.5m ice position correlates fairly well to where we think residual MYI CAB ice resides. Indeed, the thinner ice categories too are not at odds with AMSR2_AWI concentrations in areas of melting ice.

So SMOS-SMAP, as an accurate observational measurement of thin ice, is something to keep on eye on even in the pre-freeze season. It has one of the largest pole holes of no swath data which unfortunately includes the current location of the Polarstern. Ice models would do well just to ingest its daily thinnesses and focus on extending them.

The ship has drifted an amazing 122.4 km over the last 12 days under the persistent southerlies off SevZem in western Siberia. Those winds have now shifted to more off the New Siberian Islands, missing the PS position and no longer being orthogonal to the eastern Atlantic front.
 
88.7 120.8 20-09-04 10:00
87.7 105.5 20-08-23 04:00

The second image overlays OsiSaf motion on SMOS-SMAP for the same dates to show ice displacement vs thickness. The wind stress is applied quite unevenly and in an uncorrelated manner to ice thickness. OsiSaf is a lagging product that  takes two days of observation to develop a visible display even at 3x arrow exaggeration so it will be another day or two until it catches up with the change in wind.

On the technical side, a new AMSR2 paper has come out that may need assimilation into the AMSR2_AWI work-in-progress to map ice edges more accurately:

Assessment of AMSR2 Ice Extent and Ice Edge in the Arctic Using IMS
by Y Liu, S Helfrich, W Meier, R Dworak
https://www.mdpi.com/2072-4292/12/10/1582/htm#B9-remotesensing-12-01582

This work assesses the AMSR2  ice extent and ice edge in the Arctic using the ice extent products of NOAA’s Interactive Multisensor Snow and Ice Mapping System (IMS) from the period of July 2015 to July 2019. Daily values and monthly means of four statistical scores (hit rate, false alarm ratio, false alarm rate, and Hanssen-Kuiper Skill Score*) over the Arctic Ocean show distinct annual cycles. IMS ice edges often extend further south compared to those from AMSR2, with up to 100 km differences over the Beaufort, Chukchi, and East Siberian Seas in August and September.

* True Skill Statistic or Hanssen-Kuiper skill score of 1965  is defined as “recall minus false alarm rate =TP/P-FP/N (or recall+specificity-1)”. Widely used to test performance of weather forecasts (McBride & Ebert, 2000) in place of the Heidke skill score of 1926.

https://arxiv.org/pdf/1202.5995.pdf

[A-Team]

AMSR2_AWI continues to improve with the release of version 103 (though it's not 'there' yet). The new location for the twice daily files is provided below. The main change is explained in the ReadMe: the very latest sea ice concentration will appear at a fixed hour depending on your relative time zone. Rapid changes in concentration can be better resolved than with once-a-day products that use a full 24 hours of satellite swaths.

The winds have shifted to more off central Siberia and so differencing v103 shows that as the gold band in the upper Laptev. Surprisingly, these winds are not consolidating lower concentration ice as recognizable features persist for a week or more despite all the displacement on both OsiSaf light blue and AMSR2_AWI.

The product is generated twice daily for two different start times (filenames with AM or PM).

    [AM] ~22 UTC the day before until ~14 UTC. The AM product is available in the evening.

    [PM] ~10 UTC the day before until ~2 UTC. The PM product is available in the morning.

After about 16 hours the spatial coverage is almost complete. AM includes data mostly from before noon and PM after noon.

The advantage of this twice daily processing is A) a reduced latency and B) reduced time differences in the resulting ice concentration maps. A 24 hour difference as in traditional daily averaged products can cause displacement artifacts due to the ice drift. This effect can be observed along a moving ice edge which seems to be doubled in the 24 hours average.

The products so far often contain data gaps represented by gray swaths extending out from the pole hole (which never has data). These gaps can be repaired very simply as follows: open a time-ordered stack, switch mode to RGB, add a transparency alpha channel, select the gray, delete to let data from the immediately earlier file to show through, capture as a new layer, and merge down to retain file name. This can be done in one step if tiled. The dotted line in the illustrative image is not present in the actual repaired product. This process 'mixes' AM and PM but only to a very minor extent.

ftp://ftp.awi.de/sea_ice/product/amsr2/v103/nh/2020/09/

[A-Team]

Below is an easy but accurate product to follow the 'end of the melting season' which is ill-posed because the various measures (volume, open water, area, extent) all behave differently according to location. For example, ice loss goes on 365 days a year in the Barents and certain Svalbard areas provided ice has blown in. Volume and extent aren't accurately known to begin with and have poorly known seasonally dependent errors.

The highest resolution product is the 3.125 km AMSR2 concentration product. While intermediate ice concentrations are sometimes not fully correctable for weather artifacts (ie have a white bias), that is not true at the open water end of the palette. AMSR2 has a consistent instrumental record back to 2012.

The best approach would be through the netCDF repository, perhaps making equal area maps and simply counting open water pixels. However not all the archives provide the necessary Geo2D files for this. An easier approach is just to count open water pixels in the polar stereographic projection  which is conformal so not equal area except at 70º.

This introduces small errors because there is hardly any open water pixels at the very highest latitudes that would significantly depart (by 5-6%) from equal area. Further, by comparing consecutive days, the relative error will be much smaller than the absolute error. It is only the former that matters for determining the minimum.

What's new this season is slightly better time resolution, 12-hour from AMSR2_AWI. And with that comes better coastal and ice edge determination relative to 24-hour because the somewhat scattered ice periphery tends to be in motion at the minimum.

However this archive has not quite settled down. For example the 0907 PM is nearly identical to the 0908PM, indicating an error. At color picker radius=0, open water is not quite increasing monotonically. It may be better to pick up all the dark pixels, say out to 16% concentration, to account for wind dispersal and small floes appearing and disappearing in the Beaufort. Next up: check these values in AMSR2_UHH!

[A-Team]

A fairly extensive and detailed new feature of mid-concentration ice has opened in the CAB well above Ellesmere in this morning's AMSR2_AWI. It was too foggy to confirm on WorldView but seems correlated with a rather localized cyclonic weather feature shown on nullschool at the time.

Given weather artifacts, this wouldn't be worth mentioning except for the appearance and unexpected persistence of the mega-crack along the CAA and the opening between Greenland and the North Pole. These showed how little we really know about the current condition of the ice, even within the CAB. If real, this feature will stay visible in future days.

We don't actually have a swath timing map to coordinate better with 3-hr GFS weather maps and other satellites. That conceivably could be derived from the orbital map for GCOM-C if that indeed is the satellite carrying the AMSR2 sensor that Jaxa data is taken from for AMSR2_AWI. The period is 140 minutes but how the final image is actually built may require QGIS.

https://gportal.jaxa.jp/gpr/notice/case/list/

update: uniq has found a clear channel in the visible that confirms the AMSR2_AWI concentration break-up, below enhanced and with land mask and distance scale. It's always a good idea to check all the visibles because they are at different times and the clouds sometimes clear. The nh_20200909PM_SIC has not yet come onto the AWI server but should also confirm; the AM/PM should be posted 12 hrs apart at predictable times but currently are not.

This was not an epic cyclone though it attained low gale force winds that persisted over 42 hrs. The CAB ice is really in a vulnerable state, far worse than anyone had imagined or modeled. Just like the NP at the time of the Polarstern transit.

[uniquorn]

I think it's real. Noaa-20  https://go.nasa.gov/33gEmhs

[jdallen]

I think it's real. Noaa-20  https://go.nasa.gov/33gEmhs

Oh, there isn't any doubt.

It really confirms what A-Team and others have been saying about the ice this season.

It also highlights the disconnect between actual conditions an the models we are relying on.  It illustrates that fundamental changes have taken place in the Arctic as a system, and similar fundamental assumptions made by the models - in particular volume - need to be reevaluated.  In previous years, certainly pre-2012, possibly as late as 2016, a modest cyclonic feature - not even a storm - like that would not have shattered the main pack across such a wide area.

We desperately need more and different data.

[gerontocrat]

The linear trend of NSIDC September average extent is a reduction of about 1.25% per annum of the 1979 value. So about 50% gone.

The linear trend of PIOMAS September average volume is a reduction of about 1.75% per annum of the 1979 value. So about 75% gone.

Simple arithmetic tells us that these two linear trends will, in a few years, be mutually exclusive.
So what will happen? Thick remaining ice compacted into a small area, or the ice area remains large and thins to the point it just falls apart?

Is the above post by A-team an indication that maybe it will be (is?) a case of "things fall apart, the centre cannot hold..." ?

[RoxTheGeologist]

I think we are in the tail of 1/x rather than y=x. The last sea ice close to the pole doesn't receive that many days of insolation and has a lot of FDD days to thicken the ice. Even in a super hot year, like this year, with melt ponds forming early, we aren't going to break the 2012 anomalous minimum.

The Arctic freshwater lens will persist, as much of it is created by input from the rivers. Only in years like 2012 will the ocean be adequately mixed to melt more ice (and lose additional heat).

To melt that last ice there have to be fewer FDD days and more insulation during the cold months.
Or, like in 2012, mixing melts the ice. At least the latter would help in the long run with additional heat vented.

[A-Team]

Disagree. These notions are not consistent with insolation diagrams, air temperatures at high latitude wx stations, geography of river discharge into the Arctic Ocean, measured spread of their freshwater plumes, observed daily surface salinity, 2020 melting degree day temperatures and long-modeled subsurface properties affirmed by moorings.

We are looking on this forum for an explanation of what visible, infrared and radar satellite imagery and drifting buoys have been showing for months above Greenland, Ellesmere and CAA which are nowhere near major rivers or their plumes yet the ice has been melted out to the point of oblivion by insolation and MDD this summer. We have direct confirmation from the captain of the transiting Polarstern.

Indeed the very thickest CAB ice evidently remains highly vulnerable to even a modest four-day cyclone well into September. That ongoing development worsened today and calls for an explanation, perhaps via new metrics along the lines jdallen put forth above in #3529 as the current ones are manifestly inadequate.

Mosaic described the ice encountered en route to the pole as 'porous'. That's not quite the same as permeability of ice in drainable melt ponds requiring freezing snow melt to plug. It sounds like extensive brine channel melt-out, perhaps taking out adjacent Ih crystalline ice and air bubbles along with it and so leaving floes severely structurally weakened with respect to high curl wind stress events.

There's a good account of Russian research on river plumes posted on 10 Sep 2020: these plumes are confined to peripheral seas and remain coastal behind islands. They account for ~75% of all river inflow into the Arctic. There is near-zero inflow for several thousand km (SevZem to SFJL to SV to Nares to western Banks Island. The influence of the McKenzie is limited to off-shore Beaufort as discussed here many times.

"The Ob, Yenisei and Lena rivers flow into the Kara and Laptev seas and account for about half of the total freshwater runoff to the Arctic Ocean. The total annual runoff from these three rivers is estimated at 2,300 cubic kilometers. The majority of this volume is discharged into the sea during the ice-free season, forming the Ob-Yenisei plume and the Lena plume, which are the largest in the Arctic and among the largest in the world ocean.

"River plumes are freshened water masses [river runoff mixing with ambient saltwater] that form near river mouths and spread at sea as a relatively thin surface layer. River plume dynamics are mostly determined by wind forcing and river discharge rate," explained Alexander Osadchiev, a co-author.

"In the absence of strong wind, the Coriolis force and density gradient between plume and  ambient seawater cause along-shore spreading of river plumes. That process induces a large-scale eastward freshwater transport that is observed in the Arctic Ocean along large segments of the Eurasian and AK/CA shores. 

The Ob-Yenisei plume spreads from the Kara Sea to the Laptev Sea through the Vilkitsky Strait, which is located between the Severnaya Zemlya archipelago and the Taymyr Peninsula. The Lena plume spreads into the East Siberian Sea through the Laptev and Sannikov straits.

"Continental runoff from the Ob and Yenisei accumulates in the Kara Sea during the ice-free season. Topographic barriers — the western coast of the Taymyr Peninsula and Severnaya Zemlya archipelago — hinder eastward spreading of the Ob-Yenisei plume to the Laptev Sea. This process occurs only as a result of very specific wind forcing conditions.

"On the contrary, the Lena plume is almost constantly spreading to the western part of the ESS as a large-scale water mass, forming a narrow freshened coastal current in the eastern part of this sea. Known as the Siberian Coastal Current, it is intensified by freshwater runoff from the large Indigirka and Kolyma rivers and flows farther eastward to the Chukchi Sea.

"Freshwater from the rivers flowing into the Arctic Ocean very slowly mixes with seawater so the large river plumes are very stable. Freshwater can spread eastward across hundreds of kilometers, forced by local winds."

https://phys.org/news/2020-09-scientists-freshwater-arctic-ocean.html

Freshwater transport between the Kara, Laptev, and East-Siberian seas
A Osadchiev, M Pisareva, E Spivak, S Shchuka & I Semiletov
Scientific Reports v10 13041 (2020)
https://www.nature.com/articles/s41598-020-70096-w free full

[gerontocrat]

Even in a super hot year, like this year, with melt ponds forming early, we aren't going to break the 2012 anomalous minimum.

As far as sea ice extent in the Central Arctic Sea is concerned - we did.
As of today sea ice extent at 2.57 million km2 is 38k below the 2012 minimum.

Area is not - but the NSIDC pixels at 25 x 25 km are not going to pick up all this porosity and thin leads, instead the NSIDC shows increasing concentration - (not necessarily compaction).

[RoxTheGeologist]

My point (somewhat) is that whenever we fit a line we are assuming a model.

In response to A-Team: From what I read I thought freshwater input into the Arctic is largely from rivers e.g.

https://people.earth.yale.edu/sites/default/files/files/Carmack_etal_Freshwater2016.pdf

[A-Team]

The Beaufort Arm is hard to figure. Going by GFS nullschool surface winds and forecast waves, it should have been long gone by now. But it isn't, at least on the experimental twice daily AMSR2_AWI -- the ice concentration mostly seems to be just rearranging itself. Worldview visible, shows some larger CAB floes in rubble.

 If small debris goes above and below 3.1 x 3.1 km AMSR2 pixels according to wind, wave and current aggregation and dispersion, that could lead to inconsistency in the imagery. Yet on the visible, clouds prevent regular quantitative assessment of the ice left.

[liefde]

In response to A-Team: From what I read I thought freshwater input into the Arctic is largely from rivers e.g.

And if we're talking freshwater incoming from the NW European landmasses, it'll be record hot water coming 2 weeks. Not sure how soon that will impact Arctic ice-mass, but no doubt it will..

[uniquorn]

We are looking on this forum for an explanation of what visible, infrared and radar satellite imagery and drifting buoys have been showing for months above Greenland, Ellesmere and CAA which are nowhere near major rivers or their plumes yet the ice has been melted out to the point of oblivion by insolation and MDD this summer. We have direct confirmation from the captain of the transiting Polarstern.

Hopefully not too far off topic, this long term view of mercator 34m salinity could provide an explanation. There is a noticeable acceleration during the 2020 strong tpd event pushing higher salinity atl water? towards the greenland coast.
sep2018-sep2020

[BornFromTheVoid]

Not "home brew" in any sense, but here's an animation of the Beaufort Sea so far this month, to add to the discussion.

[A-Team]

New hot spot? The northern Laptev has perhaps not been told about the end of melt season, the gold line showing the week's shrinkage. The AMSR2_AWI are just coming in now as an afternoon assembly. It seems that the 14 hr am swaths did not quite add up to complete coverage often enough.

Monday morning: the archive is back to its twice-daily! The Beaufort Arm and northern Laptev remain active areas of intra-day change.

The orbital progression is shown below. It might make sense to make an avi using each consecutive pass as a frame. This might be useful to people wanting the very latest data (or recent time series) on just a specific area who don't need the whole Arctic Ocean build. That area might fall within a consistent repeat swath (or not).

[uniquorn]

It might make sense to make an avi using each consecutive pass as a frame.

That might be a more efficient than the am/pm. A daily avi of swaths which could be compiled by the user, one daily tiff compiled with lut, one compiled png and one compiled .nc
Avi might not be the best option. I think a tiff can contain multiple 'pages' which might retain the lut.

[A-Team]

tiff can contain multiple 'pages' which might retain the lut

Right, worth exploring. Seems like we once had a WorldView url that showed its individual swaths and their timestamp. Given the rapid change in cloud cover, WV often has nasty swath lines and discontinuities when assembled into a full day. Be good synergy if we could get NOAA-20 VIIRS and AMSR2_AWI synched up.

I have never seen a step-by-step list of how to get from the satellite sensor's raw orbital data download to the first raster array (restricted to the Arctic).

There has been a race to the bottom in file formats to see who can provide the most incomprehensibly broad format prescription, for example geotiff and netCDF. They are really more like folders than files. Which is good because related information can't get separated but bad if web browsers and common software can't open them.

The question is inclusivity, how much of it the joe sixpack end user can see. NetCDF had a close call with oblivion -- one person makes Panoply happen, without it 99.999% of the world could not see climate model data. As mentioned, a single netCDF can store image sequences; Panoply is set up to animate and combine them.

The recent webp story is also interesting. It is a big improvement on gif89 in many ways (better compression, better color) but until all the major web browsers provided support, it couldn't really be used. Which is not to say forum software here will catch on soon.

[glennbuck]

NSIDC 25km, 15% Area has 2020 as lowest Sea ice Area on record below 2012, Jaxa/AMSR2 has it second lowest above 2012.

[oren]

No, this compares NSIDC with 2012's AMSR2, apples and oranges. Should compare NSIDC 2020 with NSIDC 2012.

[glennbuck]

No, this compares NSIDC with 2012's AMSR2, apples and oranges. Should compare NSIDC 2020 with NSIDC 2012.

Ok is NSIDC area 2012 25km 15% area, lower than 2020 area do you have the chart to compare?

[oren]

You have published it yourself in the data thread. 2020 is #2, 2012 #1.

]https://forum.arctic-sea-ice.net/index.php/topic,2975.msg286603.html#msg286603[url][/url]

[uniquorn]

I wonder if awi amsr2 sic v103 is using some of the SIC LEADS (experimental) processing.
Yesterdays worldview terra modis with v103 inset.

[A-Team]

using SIC LEADS (experimental) processing?

There's a lot of that going around but that region seems also lit up in the old AMSR2_UHH which does not have that processing so apparently is just a reduced sea ice concentration area (or passing weather artifact since persistence has not yet been established).

Meanwhile has CMEMS has released an amazing interactive product, sort of Mercator Ocean on steroids. Only a fraction of its available layers are shown in the screenshot below. The graphical slider is fully interactive so a particular lat lon can be followed for a whole year for SST, thickness, concentration etc.

As always, a slick display can get ahead of underlying data quality. And it's one thing to study the data at the site but another to port it away somewhere else. It needs its own tinyUrl generator like WorldView. The display cannot be rotated to Greenland down in situ. However this is a very rich site that we may need to use a lot.

https://t.co/JSBMKl00Nm?amp=1
https://twitter.com/CMEMS_EU

SMOS-SMAP ice thickness is shaping up as colder temperatures settle in. It needs to be scaled up 130.07% to fit over AMSR2_AWI. Actually it needs to be redone entirely as an 8-bit color tiffs

Ascat is also looking better though not there yet. The url has been moving around but the link below is working (if you know the day-number(,

https://www.ospo.noaa.gov/data/atmosphere/ascat/MetopB/ICE/msfb-NHe-a-2020260.sir.gif

https://www.epochconverter.com/days/2020

[uniquorn]

v103  daily pixel count shows the return of the possible harmonic at 50% concentration. 0% tending to agree with the jaxa minimum extent.

I wonder if seaice.de might comment on amsr2 representation of the Beaufort Sea recently. Was it perhaps freezing rain showing up more noticably over open water? Air temps of ~-2C on sep17 according to nullschool gfs

Worldview terra modis with amsr2-uhh inset, sep17
Beaufort, v103, sep4-22

[A-Team]

Just a couple of follow-ups on the mixed melt, dispersion and flash freeze of the baffling Beaufort plus apparent wind-driven spread on the Atlantic side.

The Polarstern is currently on a line between the pole and Sv/FJL not quite at the ice edge at 86.0 32.7 on 20-09-26 at 16:00 with moderate winds at 7m/s and temperatures not currently low enough to freeze sea water at -1.8ºC.