Test space

[Wipneus]

Thread used for testing images/animations. In my case an mp4 animation created with ffmpeg.

This image uses a -pix_fmt yuv420p option.

[b_lumenkraft]

Hey Wipneus,

this one is working fine on my Mac! Thanks a lot for doing this.

[Niall Dollard]

Tried it on two browsers. Chrome and believe it or not, IE 11, and it works fine. 

[Sterks]

 does not work in ipad (latest ios) either chrome or safari. Sorry

[Wipneus]

does not work in ipad (latest ios) either chrome or safari. Sorry

Not happy to hear that Sterks. Unfortunately I cannot test that device myself. What I can do is to find another file that works and compare with mine that does not. 

Are you able to see the mp4's that A-team posted, like in this post (and many others) :
https://forum.arctic-sea-ice.net/index.php/topic,1259.msg138898.html#msg138898

[uniquorn]

testing avi. testing mp4
settled on

 ffmpeg -i tst2.gif -crf 29 -pix_fmt yuv420p -vf "scale=trunc(iw/2)*2:trunc(ih/2)*2" tst2e.mp4

-crf 29 the higher the number the lower the quality, smaller file
-pix_fmt yuv420p better chance of working in all browsers (same as Wipneus)
-vf prevents odd number of frames

probably still doesn't work for everyone but tested ok in firefox and edge on win10

[uniquorn]

whoi itp107 test

[sesyf]

Perhaps it’s just me but my ipad / safari does not allow mp4 for some reason... I should try to find if it’s some setting or whatever. I get the ’run video’ icon with a slash over it...

[uniquorn]

No apple devices here to test with either but these specs are quite limiting for this forum. Ipad users may have to download and play (more efficient if played more than once) Please see https://airmore.com/play-mp4-on-ipad.html

[uniquorn]

h.264 test
ffmpeg -i rb2.gif -c:v libx264 -preset slow -c:a copy rb4.mp4
not working for me

[Sleepy]

Testing the above with quicktime and iOS compatibility options (which makes it a bit larger):
ffmpeg -i rb4.mp4 -vcodec libx264 -pix_fmt yuv420p -profile:v baseline -level 3 rb4-MAC.mp4

[uniquorn]

Thanks Sleepy. Will get around to reading the manual sometime.
@b_l you could try ImageJ (or Fiji which already has more plug-ins) though you may need gimp for transparency (alpha)

[uniquorn]

ascat last 50days test for mac
ffmpeg -i asc-mac.gif -vcodec libx264 -pix_fmt yuv420p -profile:v baseline -level 3 -vf "scale=trunc(iw/2)*2:trunc(ih/2)*2" asc-mac2.mp4

[BenB]

Testing

[johnm33]

test gif

test gif2

[A-Team]

Just seeing if ImageJ: 57 gif frames --> avi --> mov --> mp4 works using the free and fast online converter at https://www.freefileconvert.com/mov-mp4

It does works for me on a Mac and accomplishes a huge reduction in file size, from 90 to 3.3 MB making it feasible for the forum.

The movie shows the developing arm of the Beaufort Gyre as seen by Ascat from day 255 of 2018 (Sept 12th) to day 141 of 2019 (May 21st) using 5 day intervals except for the last 3 dates. This spans 251 days.

The question here is where and when does the ice arise. The arm begins to form over Banks Island very early in the freeze season. The ice is whitish, meaning that it is older thicker ice that has largely excluded salt from its initial brine channels meaning minimal surface polarizability sko better return of Ascat's radar beam.

The ice appears sourced from the upper eastern portion of the resilient CAA ice. The winds about the  Beaufort high cause a gyre to form that is however constrained by the fixed islands of the CAA. This appears as a slicing, with large leads developing in the ice too far east of Banks' promontory to participate, to the extent the ice had mechanical cohesion.

Gyre means circle. The Beaufort ice arm has never attained anything approaching a full circle in recent years. Instead, increasingly broken up floes move up the Alaskan coast up to the eastern Chukchi and then bend towards the pole where individual floes either dissipate completely or weakly persist by the end of the melting season. Ice inside the arm has been wrongly said to thicken in place over multiple years.

The' Beaufort Gyre' is an exceedingly vague concept geographically scarcely ever depicted in the same location at the same size. It is probably best defined operationally by the lowered salinity basin shown at depth by Mercator Ocean.

As an off-axial surface rotation on a non-inertial non-spherical reference frame that itself is rotating, the ice experiences centrifugal as well as positionally varying coriolis and buoyancy forces in addition to the wind stress field applied to floe edges and compression ridges.

Since large regions of the Beaufort have melted out in recent years, an additional complexity for figuring upward Ekman transport has set in because the wind forcing is now divided over the course of a year between a rigid complete ice pack, loose floes in open leads, and large patches of strictly open water.

The pack is not moving as a western boundary current in the sense the arm is well off the bathymetry of narrow continental shelf. If there is an operative Alaskan coastal current, it cannot be on the surface as countercurrent surface ice eddies are not seen (unlike in East Greenland). However ice production and tidal effects in the Amundsen Gulf do prominently affect a band of near-shore waters along the Alaskan coast, as does Mackenzie River discharge in season.

The Beaufort Gyre is depicted in scientific journals in wildly varying ways. One of the more extreme versions is hosted by NSIDC (left of 2nd image). This is at odds with the location of a reduced salinity basin at depth provided by Mercator Ocean (right of 2nd image) and observationally confirmed by profilers and gliders.

No fresh water reservoir exists anywhere in the Arctic Ocean (away from river discharge; recent Borneo expeditions have had to melt snow as ice melt was undrinkable. Oceanographers use 'fresh water' as a figure of speech for waters with slightly reduced psu salinity.

Beaufort waters at depth cannot cross over the shallow sills of the main CAA channels and so do not reach Baffin Bay or threaten Gulf Stream overturning in the North Atlantic. Net annual export is roughly 44,000 cubic km per year from the Arctic to Baffin Bay, a small portion of sverdrups coming in with the West Spitsbergen Current or leaving via the East Greenland Current.

[gerontocrat]

Just seeing if ImageJ: 57 gif frames --> avi --> mov --> mp4 works using the free and fast online converter at https://www.freefileconvert.com/mov-mp4

It does works for me on a Mac and accomplishes a huge reduction in file size, from 90 to 3.3 MB making it feasible for the forum.

The movie shows the developing arm of the Beaufort Gyre as seen by Ascat from day 255 of 2018 (Sept 12th) to day 141 of 2019 (May 21st) using 5 day intervals except for the last 3 dates. This spans 251 days.

Works fine - on Windows 10, Google Chrome (on-line and downloaded)

ps: need to find freebie mp4 and gif players that let me slow down the frames per second. My old eyes and brain can't keep up.

[johnm33]

"ps: need to find freebie mp4 and gif players that let me slow down the frames per second. My old eyes and brain can't keep up."
right click/play speed works for me in firefox

• and rightclick/enhancements/play speed in WMPlayer

[A-Team]

Testing the waters again on all-ImageJ Ascat mp4. Good labels can be made and positioned as desired if the files are renamed properly before loading. Duplicating the stack, cutting and filling to text box size, then enlarging canvas to fit final size of the data stack, using 'paste control' set to 'zero fill' which is ImageJ's way of setting pure black as alpha transparency.

I also compared 'equalize' to manual brightness and adaptive contrast as a quick and easy contrast adjustment. If an Ascat land mask is loaded, these tools have a better histogram to work with that does not include noisy weather patterns over open water. The land mask is helpful in defining crop boundaries. If it is included in the stack of Ascats, it will be cropped just like them and can be used at any time up to mp4 production.

The specific task here is optimally depicting movement of the last holdout ice between the North Pole and central CAA (which has poor inherent contrast). Normally it just sits there but over the last 172 days a great unprecedented swath of ice has steadily moved from islands off central Siberia across the pole nearly to and out the Fram. The older ice is been squashed up against the Canadian islands, with some pushed towards the Beaufort, some forced through Nares, some possibly to the Barents, with most going down the Fram. The Ascats are again spaced at five days to reduce file size of product.

Despite wx predictions to the contrary, this pattern has continued through May 22nd. This, if it continues another month, will lead to shocking developments by fall so I am inclined to bury the mp4 here in development where fewer will get unduly alarmed ahead of events. Even with a so-so melt season, there would be very little surface area, volume, or multi-year ice left going into September. However we have seen other years like 2012 where weather simply shifted and the melt season trajectory changed to something less damaging.

It's never been clear what would hold the old ice against the Canadian islands if the outer ice pack is not there to exert pressure; wind patterns could well change with more open water, this ice might drift off, be broken up by waves, with its remnants disappearing.

Some movements of continuously recognizable features are shown in the static image. It is better to circle reference areas than to use vectors of point displacements because of rotation and deformation. The dark ice is mostly first year and so the displacement of its front is feasible to measure. For scale, it is 1050 km from Longyearben, Svalbard to the North Pole.

The attached high precision land mask is from AMSR2UHH. It fits very accurately over the 1170 x 1170 Ascats.

[A-Team]

Here I am testing a few other ImageJ processes while looking at the question of whether the massive ongoing Arctic Ocean icepack rotation is leading to significant export out the Svalbard-FJL opening to the Barents (where it would very likely melt away later in the season. While the answer to that is no for the last 30 days (to 23 May 2019, day 143), more ice will be shaved off into a path between the islands if the current wind forcing continues.

However the amount would be negligible compared to the area/volume exiting the Fram. The ice in the Barents today predominantly derives from left-over ice of the previous melt season and ice leaving the Kara Sea. Last year, the Kara injected a massive tongue of ice north into the Arctic Ocean basin, some 5-6% of the total area. This ice moved zonally west but did not quite make it to the Fram.

On the technical side, the first image shows a developing seasonal problem with Ascat: as more open water develops, image quality degrades and ice features can scarcely be distinguished from increasing dielectric in fo,g low clouds and surfaces enriched in sea salt. This affects boundary areas like the Bering/Arctic, Chukchi/Arctic and Barents/Arctic almost all year.

The aim here is to extend the Ascat season into June by masking both islands which contribute distraction in time series but are fixed in position and open water/regions of low concentration ice as determined by AMSR2 which is not subject to the same artifacts. The former is a fixed mask that is valid all year; the latter varies from day to day.

The first image shows the Ascat download (top) for May 2nd compared a white fringed island mask (bottom) from UHH AMSR2 that stays fixed throughout the animation (2nd image). There are issues in resizing the mask within small channels between islands because 'large' pixels can only do so much. The second frame shows open water and the third what 0-20% ice concentration adds to the mask (red band). The AMSR2 has to be enlarged here by 7% to fit the 2.5x enlarged Ascat. It would have been better to have chosen the original Ascat enlargement properly (228.084%) as in 3rd graphic.

There are no no real additional steps involved in placing daily variation masks over montaged enhanced Ascats. The AMSR2 itself cannot be enlarged by non-integral amounts with interpolation of palette steps which would disrupt precision color picking from it.

Some text to orient non-experts has been included by placing black text in a blank alpha channel over any frame, then duplicating to the total number of frames, removing the alpha channel to white, inverting, and placing it over the ImageJ montage with 'copy control' set to black as transparent. Date labels come from the original file names, as modified, and are captured early on (before ImageJ loses them).

[SteveMDFP]

Here I am testing a few other ImageJ processes while looking at the question of whether the massive ongoing Arctic Ocean icepack rotation is leading to significant export out the Svalbard-FJL opening to the Barents (where it would very likely melt away later in the season. ...

It is so good to have A-Team back on the Forum.  Such combined clarity and sophistication is a rare treat in any realm.

[wili]

Hear! Hear!

[uniquorn]

Good ideas for cleaning up the images but ascat often 'sees' ice that unihamburg amsr2uhh doesn't. Example is the never melting ice in the beaufort/chukchi last september, 20180902,day245 shown here.(apologies for the rough cropping/scaling)
Also amsr2 has it's own 'flashing' 'scorching' and other 'weather' issues.

[A-Team]

Here I am testing another double-masking of Ascat, this time with U Bremen SMOS-SMAP new thinness product instead of UHH AMSR2 low sea ice concentration and open water. This product is provided at a very generous scale of 1173 x 2170 png, a big improvement in resolution (if it is real and not just a rescale). It comes with a satisfactory land mask too.

https://seaice.uni-bremen.de/data/smos_smap/png/north/2019/ png archive back to 31 Mar 2015

Because thinness and concentration are fairly similar in location in that edges of the ice pack have thinner ice as well as lower ice concentrations, it isn't feasible to triple-masking because of the extensive overlap in areas affected.

This combination of two satellites seemed to have extended the season well past the previous SMOS-by-itself terminal date of May 1st. However farther along in the season as melt ponds and liquid clouds become important, both the SMOS-SMAP thinness and Ascat products will likely deteriorate (or the former not even be archived).

The mp4 shows 53 days from April 1st to May 23 of 2019. A larger view is needed because ice thinness in peripheral seas such as Bering, Barents and Kara is more important than in the central Arctic Ocean this time of year. Both SMOS-SMAP and Ascat have a few missing days or partial images which I replaced by adjacent complete ones; this happens as well with UHH AMSR2 but less frequently.

The May 23rd looks like it has a blob of weather in the northern western Beaufort interfering with proper thinness determination. Ice thinness doesn't change as rapidly as weather or melt or and persists from frame to frame. These artifacts could be edited out manually or by AI per the criteria stated.

On scaling, SMOS-SMAP needs an enlargement of 104.414 to fit the smaller AMSR2uhh whereas Ascat needs 228.084 so to match SMOS-SMAP to Ascat requires 228.084/104.414 = 218.442 magnification of Ascat. I measure from the Aleutians to southern Sweden to get the largest pixel lengths to compare as this reduces percent error in the scaling ratio.

The AMSR2 masking shown a couple of posts back covered 30 different days so I am not reposting it here. "30 days of Ascat land and water masked to 24 May 2019.mp4"

Combined SMAP–SMOS thin sea ice thickness retrieval
C. Patilea et al U Bremen AWI 28 Feb 2019
https://www.the-cryosphere.net/13/675/2019/ free full text

Consistent Combination of Brightness Temperatures from SMOS and SMAP over Polar Oceans for Sea Ice Applications
AU Schmitt, L Kaleschke April 2018
https://www.mdpi.com/2072-4292/10/4/553/htm free full text

https://seaice.uni-bremen.de/data/smos_smap/png/north/2019/ where to get the pngs

The passive microwave sensors Soil Moisture Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) provide brightness temperature data at 1.4 GHz. At this low frequency the atmosphere is nearly transparent and in polar regions the thickness of thin sea ice can be determined up to 0.5m for an extended period into the spring.

The thickness of ice partly determines the resistance against the deforming forces of wind and ocean currents. Even a thin layer of sea ice inhibits evaporation, reduces heat and gas exchange between ocean and atmosphere and increases the albedo. Sea ice — even if thin — also provides a solid surface for snow to deposit on, which further reduces heat exchange and increases albedo.

SMOS has been developed for retrieving soil moisture by inferring the surface emissivity which is correlated with the moisture content and sea surface salinity and link measured brightness temperatures with sea salinity through the dielectric constant of the water in the first few centimeters [[to which upwardly extruded brine or melt ponds might contribute]].

Effects of decimetre-scale surface roughness on L-band Brightness Temperature of Sea Ice
M Miernecki, L Kaleschke et al  03 Jun 2019
https://www.the-cryosphere-discuss.net/tc-2019-110/tc-2019-110.pdf free full text

Sea ice thickness measurements with L-band radiometry is a technique which allows daily, weather-independent monitoring of the polar sea ice cover. The sea-ice thickness retrieval algorithms relay on the sensitivity of the L-band brightness temperature to sea-ice thickness. In this work, we investigate the decimeter-scale surface roughness as a factor influencing the L-band emissions from sea ice.

Most affected by surface roughness is the vertical polarization around Brewster's angle, where the decrease in brightness temperature can reach 8 K. The vertical polarization for the same configuration exhibits a 4 K increase. The near-nadir angles are little affected, up to 2.6 K decrease for the most deformed ice. Overall the effects of large-scale surface roughness can be expressed as a superposition of two factors: the change in intensity and the polarization mixing. The first factor depends on surface permittivity, second shows little dependence on it.

[b_lumenkraft]

That's quite an amazing result! Great work A-Team.

[sark]

 1000mb temperature anomaly by month, 2012-2018 vs 1979-2000

tiled in gimp with python

[A-Team]

'Wind roses' are produced by weather stations to succinctly summarize wind direction and strength seasonally or over a year. Typically, the wind direction is binned into the 16 cardinals (eg SSW), wind speed is binned into color, and wind frequency of a given direction displayed continuously as spoke length in these 'directional histograms'.

The first image below collects various depictions of wind roses, including on for the Svalbard weather station. The links show how to roll your own from raw data with code or even entirely within excel, plus a journal paper on actual North Atlantic wind roses and vorticity.

https://www.gigawiz.com/windrose.html
https://www.climate.gov/maps-data/dataset/wind-roses-charts-and-tabular-data#undefined
https://windroseexcel.com/guides/using-excel-make-wind-rose-step-step-guide/
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2016JC012453 free full

It would be feasible to pick any (gridded) point in the Arctic Ocean, collect recent and predicted GFS data in nullschool and make a 2m wind rose. However how accurately are wind speed and direction really known in the un-instrumented situation? (Note buoys drift around and moorings do not monitor surface conditions.)

Now Ascat and QuikSCAT were actually designed to measure surface winds on the open ocean via Bragg backscatter from wind-generated capillary-gravity waves. That might work for open water of the Arctic Ocean in late season though we've never looked. Ascat can be repurposed for ice just like the soil moisture satellite SMOS can be repurposed for ice thinness.

Since surface wind data is not at hand but near-daily ice motion is, I looked at whether 'ice roses' could be made from T Lavergne's OSI SAF gridded sea ice motion vectors.

ftp://osisaf.met.no/prod/ice/drift_lr/merged/ ice motion .nc files
https://tinyurl.com/y4b6qblq daily ice motion map

Now daily data is online in properly geocoded netCDF but I elected to prototype with a hack on the graphical product of the last 90 days. Among other things, I expected ice roses to show (1) ice NOT gyrating in the Beaufort 'Gyre', (2) regular rapid ice motion Chukchi-ward off the Alaskan coast, (3) a broad swath of rapid ice motion towards the Fram, and (4) nearly dormant thick ice above the CAA.

ImageJ, under 'Z project, provides various ways of making ice roses out of an ice motion stack. These are not ideal because of the 3x scale, original arrowheads and magenta coloring for uncertain motions. The 'Volume Viewer' plugin has much fancier ways of looking stacks (3rd illustration; inset shows a 5x5 grid square from the Ak Beaufort compared to southern CAB/

[A-Team]

Here I am testing another double-masking of Ascat, this time with U Bremen SMOS-SMAP new thinness product instead of UHH AMSR2 low sea ice concentration and open water. This product is provided at a very generous scale of 1173 x 2170 png, a big improvement in resolution (if it is real and not just a rescale). It comes with a satisfactory land mask too.

https://seaice.uni-bremen.de/data/smos_smap/png/north/2019/ png archive back to 31 Mar 2015

Because thinness and concentration are fairly similar in location in that edges of the ice pack have thinner ice as well as lower ice concentrations, it isn't feasible to triple-masking because of the extensive overlap in areas affected.

This combination of two satellites seemed to have extended the season well past the previous SMOS-by-itself terminal date of May 1st. However farther along in the season as melt ponds and liquid clouds become important, both the SMOS-SMAP thinness and Ascat products will likely deteriorate (or the former not even be archived).

The mp4 shows 53 days from April 1st to May 23 of 2019. A larger view is needed because ice thinness in peripheral seas such as Bering, Barents and Kara is more important than in the central Arctic Ocean this time of year. Both SMOS-SMAP and Ascat have a few missing days or partial images which I replaced by adjacent complete ones; this happens as well with UHH AMSR2 but less frequently.

The May 23rd looks like it has a blob of weather in the northern western Beaufort interfering with proper thinness determination. Ice thinness doesn't change as rapidly as weather or melt or and persists from frame to frame. These artifacts could be edited out manually or by AI per the criteria stated.

On scaling, SMOS-SMAP needs an enlargement of 104.414 to fit the smaller AMSR2uhh whereas Ascat needs 228.084 so to match SMOS-SMAP to Ascat requires 228.084/104.414 = 218.442 magnification of Ascat. I measure from the Aleutians to southern Sweden to get the largest pixel lengths to compare as this reduces percent error in the scaling ratio.

The AMSR2 masking shown a couple of posts back covered 30 different days so I am not reposting it here. "30 days of Ascat land and water masked to 24 May 2019.mp4"

Combined SMAP–SMOS thin sea ice thickness retrieval
C. Patilea et al U Bremen AWI 28 Feb 2019
https://www.the-cryosphere.net/13/675/2019/ free full text
https://seaice.uni-bremen.de/data/smos_smap/png/north/2019/ where to get the pngs

The passive microwave sensors Soil Moisture Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) provide brightness temperature data at 1.4 GHz. At this low frequency the atmosphere is nearly transparent and in polar regions the thickness of thin sea ice can be determined up to 0.5m for an extended period into the spring.

The thickness of ice partly determines the resistance against the deforming forces of wind and ocean currents. Even a thin layer of sea ice inhibits evaporation, reduces heat and gas exchange between ocean and atmosphere and increases the albedo. Sea ice — even if thin — also provides a solid surface for snow to deposit on, which further reduces heat exchange and increases albedo.

SMOS has been developed for retrieving soil moisture by inferring the surface emissivity which is correlated with the moisture content and sea surface salinity and link measured brightness temperatures with sea salinity through the dielectric constant of the water in the first few centimeters [[to which upwardly extruded brine or melt ponds might contribute]].

[Jim Hunt]

This combination of two satellites seemed to have extended the season well past the previous SMOS terminal date of May 1st. However farther along in the season as melt ponds and liquid clouds become important, both the thinness and Ascat products will deteriorate.

Note that although the SMOS images are readily available for download if you know where to look they are not accessible via the Bremen data browse interface, which states "no data during summer":

https://seaice.uni-bremen.de/databrowser/#p=smos

Use with care?

[Jim Hunt]

Testing whether this animation autoplays or not:



For context see:

https://forum.arctic-sea-ice.net/index.php/topic,2591.msg201758.html#msg201758

[A-Team]

SMOS images are available for download if you know where to look; they are not accessible via the Bremen data browse interface, which states "no data during summer. Use with care.

Right, the interface discourages use of summer SMOS probably because some people wouldn't use it with care and subsequently diss the product. It must be hard to separate information from noise though or they would have done so. However there's no secret link to the summer archive, just the usual: https://seaice.uni-bremen.de/data/smos/png/

SMOS Level 1C v6.20 goes throughout the summer back to 2012. The png though are a meagre 275 x 275 pixels for the Arctic Ocean proper, needing 132.54% enlargement just to overlay Ascat and 302.05% to overlay the smaller AMSR2 ice concentration.

For reasons given in the 2019 article above, SMOS-SMAP probably improves on plain SMOS. However we don’t know at this point whether SMOS-SMAP will continue on into June or through the summer. I expect so because it is easier just to leave the pipeline running than shut it off and then back on.

https://seaice.uni-bremen.de/data/smos_smap/png/north/2019/ png archive for SMOS-SMAP

Another direction taken by L Kaleschke et al to improve SMOS is pairing it with CryoSat-2 swaths to extends its thickness down to 1m. However I believe that shuts down on May 1st because of summer dielectric issues.

So the question is, if the Ascat record goes all year but is overlain with unfixable contrast artifacts in summer, how do we best mask those in order to still visualize sea ice motion (especially in view of MOSAiC floe anchorage and trajectory)?

It helps somewhat to mask land, open water, and low concentration in Ascat but thinness and thickness are still problematic. Very few current journal articles on sea ice thickness even mention Piomas, possibly because it is not observational nor testable in situ. That leaves thinness and IceBird.

https://www.awi.de/en/science/climate-sciences/sea-ice-physics/projects/ice-bird.html
https://epic.awi.de/id/eprint/47841/1/CampaignReport_ASIMBOsummer2018_final.pdf

"The analysis shows that the largest fraction of surveyed sea ice in 2018 originated in the Laptev Sea. It took approximately 2–3 years of drift with the Transpolar Drift until ice reached Fram Strait." (2nd image below)

Below I tested whether we could get a better Ascat image for the single mid-summer date of 01 Aug 2018, with a method that would carry forward to the current season. The final frame asks whether the 0.5m thinness boundary is trackable despite its artifacts.

Once again, the motivation is that ice areal export this year may exceed in situ areal melt — wind-driven ice movement being a critical consideration in addition to seasonal clouds, melt ponds, insolation and radiative balance. Ascat is about the easiest tool for measuring ice areal export and lagrangian trajectories of extended features, important as very little ice resides in distinctive floes outside the Beaufort streamer.

It's quite difficult to continuously track sea ice motion credibly over multiple years because of the summer microwave data hiatus, yet MOSAiC would like to know an accurate locational history of the floe the Polarstern will be anchoring up to for the year. For a back-trajectory approach, see:

Arctic warming interrupts the Transpolar Drift and affects long-range transport of sea ice
T Krumpen et al April 2019
https://www.nature.com/articles/s41598-019-41456-y

SMOS-SMAP may be done for the summer -- or they can't make product on Ascat glitch days. Here are the offerings on their server, version 205 of the algorithm has just recently kicked in:

2019: daily to May 23rd, version 205 [today is May 28th!]

2018: hiatus from June 5th to August 7th and Oct 8th to Dec 31st, version 100
20180604_north_mix_sit_v100.png
20180809_north_mix_sit_v100.png

2017: December only, version 100
20171201_north_mix_sit_v100.png
20171231_north_mix_sit_v100.png   

2016: nothing

2015: full year version 100

[Jim Hunt]

For reasons given in the 2019 article above, SMOS-SMAP probably improves on plain SMOS. However we don’t know at this point whether SMOS-SMAP will continue on into June or through the summer. I expect so because it is easier just to leave the pipeline running than shut it off and then back on.

I have my own pipeline running to acquire images, but currently it simply assumes yesterday's date. I'll need to work on that since the latest SMOS-SMAP image seems to be from the 23rd. Does the apparently < 0.5 m thick area at ~78 degrees north of Utqiaġvik exist in the real world?

Another direction taken by L Kaleschke et al to improve SMOS is pairing it with CryoSat-2 swaths to extends its thickness down to 1m. However I believe that shuts down on May 1st because of summer dielectric issues.

I've been acquiring the NetCDFs for a while. The most recent is for the week beginning April 21st.

[A-Team]

Does the < 0.5 m thick area at ~78 degrees north of Utqiaġvik exist in the real world?

This feature was discussed above along with techniques for manual or AI discrimination between weather artifact vs flash melt vs ice thinness . Another couple of days of SMOS-SMAP data are needed for sandwiching.

The IceBird reports make for interesting reading. One of several observational ice thickness data images from a low elevation August 2018 flight out of Alert is shown below.  Coverage is limited to flight tracks but still very worthwhile for high resolution calibration and reality checks. The March/April 2019 report is not yet posted; we can wonder if it will incorporate Polarstern's serendipitous in situ data from the early March reverse surge of the Fram.

https://www.awi.de/en/science/climate-sciences/sea-ice-physics/projects/ice-bird.html

[Jim Hunt]

ICESat-2 data has dropped. However my alter ego is having difficulty finding the sea ice thickness numbers:

https://twitter.com/GreatWhiteCon/status/1133480350200545281

[A-Team]

The first mp4 below tests whether Ascat is able to display trackable ice feature motion in late summer and fall using 5-day increments. The first frame (red arrows) shows a noxious weather system on 25 Aug 2018 that cannot be removed by land and open water (AMSR2) masking or adaptive contrast enhancement.

However the other 26 days to 02 Jan 2019 show stable ice motion features. Thus by removing late summer dates with transient passing weather from the time series, it may be feasible to track a few features well into June and well back into August, possibly even attaining the holy grail of (multi) year-round tracking of the entire icepack by establishing a sufficient subset of floe trajectories.

For vital observational projects like full-year MOSAiC, this could provide the history and conditions that the anchoring site experienced in the past.

The second mp4 compares raw Ascat (UL) to double-masked and mildly enhanced Ascat (UR), to o double-masked and strongly enhanced Ascat with moderate sea ice concentration in pink (LL)), to land- and open water-masked false color of mildly enhanced Ascat (LR).

These are best viewed over and over in loop mode (right-click mouse contextual menu) because only 27 days of data were used.

[A-Team]

Here I am testing whether ice motion vectors from osisaf really go all that well over Ascat time series motion of persistent features as masked for land and daily open water. Even though the Beaufort streamer has all but disappeared in Ascat except for the Chukchi by 26 May 2019, the arrows go quite well with perceived motion and feature displacement. Thus the colored contours provide the line integral of the daily osisafs.

The movie shows enhanced Ascat, followed by Ascat features, followed by osisaf arrows over Ascat, followed by osisaf arrows over Ascat features over Ascat. It looks more like a 5x4 slide show than a movie because the frame rate is set at one second. It would be feasible to do the whole melt season in this manner.

[Jim Hunt]

ICESat-2 data has dropped.

NSIDC have got back to me:

Since the data were just released yesterday, published papers aren't likely to be available yet, but they'll almost certainly be forthcoming. Here's a link to the NASA ICESat-2 website list of publications: https://icesat-2.gsfc.nasa.gov/publications if you want to check there occasionally for new content.

We don't yet have an idea of when the gridded products will be made available due to the many variables in play considering that Level 3B products comprise Level 0 - Level 3A data sets as inputs. I don't know if you've browsed the NSIDC ICESat-2 project page yet, but here's a link detailing the products available, expected, and the processing workflow: https://nsidc.org/data/icesat-2/products/

The ATL10 (ATLAS/ICESat-2 L3A Sea Ice Freeboard, https://nsidc.org/data/atl10/versions/1) data set outputs total freeboard height (height from top of snow/air interface to sea surface) where it can be calculated from also measuring sea surface height. For a highly detailed explanation of the processing steps and data output of ATL10, you may be interested in reading through the Algorithm Theoretical Basis Document (ATBD) for ATL10 available from this page: https://icesat-2.gsfc.nasa.gov/science/data-products

CPOM have also suggested some light bedtime reading matter:

https://twitter.com/GreatWhiteCon/status/1134091514274037760

A gridded sea ice thickness product still seems to be some way off however. 

[A-Team]

Please, test-forum material here only: graphic experiments.

Testing various combinations of masks and markups for interpretive effectiveness just on one day, 30 May 2019 (that continues the 5-day of the previous post), as it had an especially revealing Ascat for the Beaufort/Chukchi side. The file size is small so posted as a gif rather than mp4.

The first frame AMRS2 of the 'same' date provides land, open water and 50 palette bins intermediate concentration sea ice; Osisaf the sea ice motion from two days of AI. The latter needs a 85.123% reduction to fit over 6.25 km AMSR2, the scale here until the final reduction to 700 pixel width.

The Beaufort streamer is deteriorating so rapidly that the open water mask actually degrades the view in regions of scattered floes which are quite visible already in the enhanced Ascat. The open water mask remains very useful over the Bering and Chukchi seas, in the Fram and around Svalbard-FJL.

These images don't come with no time stamps because they are swath assemblies; sea ice motion is so rapid at ~1 km/hr in the Beaufort and quite heterogenous across the overall icepack so swath offset between different satellites leads to slightly incorrect masking.

The Osisaf overlay shows an approximate clockwise rotation of the entire ice pack, which has been going on fairly steadily for an over six months. The center of rotation is neither the North Pole nor the 'Beaufort Gyre' but better approximated by the Arctic pole of inaccessibility at 85º 48N, 176º 09E.

The maximal possible rotational (without hitting land) has radius 1118 km (excluding the small De Longa islands ) but in late May the radius appears rather larger, giving rise to a semi-circle of sea ice motion that takes ice off Banks Island all the way around the Siberian side to the Fram.

Indeed the Kara Sea has been swept up in this, exporting short-lived ice to the Barents instead of massively to the Arctic Ocean like last year where it was not fully exported out the Fram by the end of melt season.

There does not seem to exist a single optimal combination of overlays in melt season that works for every region. Consequently, it works better to prepare an interpretive packet. The May 30th data substantially clarifies (mis)interpretations arising last week from weather and ice surface disturbances, meaning Ascats will continue to provide good data into June but RT or NRT analysis should be deferred or restricted to a few scenes earlier as the last date lacks necessary flanking context.

The seven frames here for May 30th would best reside in ImageJ as a hyperstack. Each frame could be viewed horizontally in the time direction or vertically in the overlay. However only 0.5% of forum viewers use ImageJ.

We've looked before at automating production. Daily data downloads already present issues such as archive posting delays and omissions, as well as defective but patchable imagery. Patching is separately scriptable at least for Ascat and AMSR2. Once the image files are all in their expected desktop location, it is feasible via an ImageJ macro of macros to upload, rescale, adjust contrast, align to the pole, and assemble into a stack that could be added to the existing hyperstack as a new 'frame'. The question is, how flexible would such an automation pipeline be. Not very.

[A-Team]

Here I am testing a throwaway method that enable me to quickly see if the current day continues the incredible sequence of over 180 days of consecutive anti-cyclonic ice rotation. The example simply jiggles back and forth between May 31 and June 01. Indeed this is effective for this purpose. (We have tried other options over the years such as 'color interferometry' of three consecutive days which works but is harder for non-specialists to understand.

The second animation tests an algo for making a quick daily "executive summary" of the most recent developments. It uses AMSR2uhh for land and open water masks, osisaf for quantitative gridded vectors of ice motion, Ascat, ImageJ enhancements of that Ascat, and composition to final image within Gimp. The only technical part is floating arrows off the osisaf in a non-obtrusive but still visible overlay.

The third mp4 tests whether very fast (80ms) back and forth looped display over the 15 Mar -01 Jun 2019 early melt season is visually effective in understanding cumulative sea ice motion which has been truly extraordinary the last 6-7 months.

Towards the end, it shows (rainy?) weather coming in from Alaska over the Beaufort. The file size is so small at this speed at 1.2 MB that a whole year could be shown under forum size limits, possibly even two if water and land masking had been applied.

[be cause]

 .. cheers for what you are doing here A-team .. I can just @ press play and type .
but   wow ! that shows how big a biter Nares is very clearly .. and how much has left ,stage right . That is/was a lot of the oldest and the coldest ice  in the Arctic . Looks like we are going to find out what 80% plus of 1st year ice looks like in a challenging season . b.c.
 
 

[A-Team]

bc: Nares a lot of the oldest and the coldest ice in the Arctic is leaving. a challenging season.

Right. Am I just documenting an unusual start to a melt  season or writing the Arctic ice obituary? This persistent CW circulation has been devastating for remnant thicker ice. In addition to direct export, a lot of ice is being translocated to areas like the Chukchi where it will melt out even in a so-so summer (along with almost all the first year ice, the dark ice in Ascat). We've seen all these pieces before but not in concert for so long; we've also seen the summer weather change to something else.

Below I am just testing some very basic one-click enhancements of Worldview natural color of the Beaufort Sea on 01 Jun 2019 and looking to see if Ascat can see individual floes. These modifications are available in about any image program on any device including phones. However few forum posts clarify images.

The first 'as is' frame shows WV has deplorable initial contrast: whitish clouds over whitish floes next to snow covered land. If not looking for melt ponds, a linear contrast adjustment (frame 2) followed by local adaptive contrast enhancement (frame 3), followed by some unsharp mask (frame 4), followed by sometimes useful palette inversion (frame 6) show much better details of floe cracking.

If the original image had instead been a time series, those images can be stacked or montaged side by side and the enhancements applied to all of them at once with the same single clicks. This gives much more effective animations.

The first frame's lower inset shows Ascat of the same day at 4x. It can indeed detect individual large floes but not nearly as well as WorldView. The largest flow -- the one with the full-length fracture -- is 61 km in length. It is represented in Ascat by only 15 pixels. The hair-line fracture cannot be seen. Its best resolution in WorldView is shown in frame 5.

[oren]

A-Team, may I ask your permission to occasionally post some of the outstanding stuff here onto the main melting season thread? I think it deserves a wider audience even if by your standards it's just test material.

[be cause]

.. seconded

[A-Team]

Hi Oren, feel free to distribute and add commentary anywhere you see fit. I believe you can just link to the graphic alone, no need to download and re-upload, but have not tested that for all file types.

I still have some more pieces to test here such as more Google Earth Pro files. (They are an export/re-display option for netCDF .nc files out of Panoply.) For now, these are attached as .txt even though they are really .kml. GEP is now free but still prone to inexplicable crashing.

The attached image shows a new boundary polygon for the Arctic Ocean proper. This basin is where ice area, volume, extent and concentration matter. The rest -- like the Barents, Bering, Baffin and so forth -- just contribute background noise to inter-year comparisons. (Export is a ratcheting process.) The problematic relationship of DMI's 80ºN temperature circle to the Arctic basin is indicated.

Yesterday had quite an unusual Osisaf ... just about the entire ice pack was rotating though not quite about the maximal center of rotation (pole of inaccessibility). Some ice gets shaved off by Nord, Greenland and exits the Fram. There has not been consequential ice export out the Svalbard-FJL gap this melt season.

See also:

Barnes, R. (2019). Optimal orientations of discrete global grids and the Poles of Inaccessibility. International Journal of Digital Earth, 1–14. doi:10.1080/17538947.2019.1576786

Arctic Pole Rees et al. (2014) 176.149 85.802 1007.6039 1008
Arctic Pole GSHHG 176.1423 85.8015 1007.6777
Arctic Pole OpenStreetMap 176.2386 85.7911 1008.9112

Global Self-consistent, Hierarchical, High-resolution Geography (GSHHG)coastline data (Wessel and Smith 2016)
 
G Rees et al 2014.
Finding the Arctic Pole of Inaccessibility.”
Polar Record 50 (01): 86–91. doi:10.1017/S003224741300051X.

[A-Team]

Here I am testing the limits of Ascat enlargement and contrast enhancement for depicting ice movement in the core thick CAB ice from Sept 2018 to June 2019. The original starting Ascat is shown in the lower left corner exactly as offered at its online archive, cropped to the same view.

While some areas of ice above and to the west of Ellesmere hardly seem to move over the 256 days, other large and normally immobile areas below the pole are moving rapidly towards and out the Fram. If this pattern of ice pack motion continues, it will devastate the core of thick multi-year ice remaining, even with so-so melt season thermodynamics. The motion is far more extreme than 'trans-polar' drift as that is usually depicted.

On the technical side, the passages of the CAA are masked out as they were just adding distraction. Open water masking is not needed for this area. Ascat histograms are misleading, even not including land and water, because the older ice (whiter colors) is more confined to a specific region than first and second year ice.

After linear contrast adjustment, it benefits from gamma correction prior to subtle adaptive contrast and unsharp masking. It is easy to overdo this, giving an outcome that is too lumpy in appearance.

A single frame from DMI's Lincoln Sea is shown for June 2nd. This too could be animated though it would not be feasible to match dates exactly because of clouds. The same could be done in WorldView, skipping exact date matching to favor less cloudy days close in time. However the Arctic is a very cloudy place.

[A-Team]

Just testing if direct .avi to mp4 out of ImageJ displays after https://www.media.io/convert/mov-to-mp4.html. Seems like it works ok.

The interest here is not so much the Beaufort arm as along the coastline to the east, which is getting engaged these last few days. This, if it continues, would draw even more of the oldest thickest ice into warmer seas where it would melt out.
 
The initial gif was made using the animator at Worldview. Since it is limited to 40 days, I made two animations and concatenated them in ImageJ to get from April 10th (when the drama started) to June 3rd. Optical imagery is not available too much earlier because of winter darkness.

These images were gathered at the 1km resolution setting. They are fairly hard to enhance other than linear contrast squeezing. I also applied a very slight adaptive contrast to sharpen, converted to grayscale and inverted. The dates are floated in from a cropped duplicate as WV provides no control over their placement.

I'll post the key region farther east in a bit as June 4th becomes available, along with the Ascat view. Some of the action can already be seen in the latter. It has been far too cloudy above the Nares funnel to make a time series in the visible wavelengths; DMI's Greenland imagery is cropped too close to shore.

The core of thickest least mobile ice is under attack on the Fram, Nares and west CAA fronts. This will greatly increase the effects of melt season on the overall ice pack.

[b_lumenkraft]

Awesome contrasts A-Team. This is perfect!

[longwalks1]

I am sure there are buttons or drop-downs for gui video players. 

For mpv  -- to slow it down

Code: [Select]

mpv --speed=0.1 78\ days\ fast\ back\ and\ forth\ to\ 01\ June\ 2019.mp4

man mpv

  --speed=<0.01-100>
              Slow down or speed up playback by the factor given as parameter.

              If --audio-pitch-correction (on by default) is used, playing with a speed higher than normal automatically inserts the scaletempo audio filter.

     

I am in awe  - thanks A-Team

[A-Team]

Testing benefits hereto masking open water in late summer and early fall Ascats. It does seem that features visible today can be seen as early as 17 Sep 2018. This allows monthly measurements of displacement over the 260 days that have elapsed since.

The feature in the animation has largely moved along a meridian to the north pole and sped up this spring. However even at this rate, it would not be halfway to the Fram for another 5 months (early November) and not be exported until April of 2020.

In recent days, ice motion on the Atlantic side has changed somewhat in direction, with Kara Sea ice now entering the CAB whereas for the eight previous months it was being exported (inconsequentially for the melt season minimum) to the Barents.

Dire predictions are being made for a strong Arctic dipole for the rest of June. This might well come to pass but it does not suffice to determine local surface winds nor the stress applied to sea ice, nor predict that motion. (Conversely, atmospheric pressure gradients cannot be inferred from ice motion beyond generalities.)

The Arctic dipole (high pressure on NA arctic; low pressure on Eurasia) is an alternative to the NAO and annular AO patterns first mentioned in 2006 and blamed for summer 2007 record melt.
It was also the subject of an informative Dec 2009 post at WeatherUnderground:

https://www.wunderground.com/blog/JeffMasters/the-climate-is-changing-the-arctic-dipole-emerges.html

Today we have direct daily observational sea ice motion provided by Osisaf (though no forward prediction). Clear-sky insolation and advected heat only affects ice motion after a substantial  delay. It is the synergy between those and the 800 lb gorilla of this year's wind-driven ice fragmentation and translocation that poses the cataclysmic threat to Arctic sea ice this season.

How does early June 2019 compare to previous years? That proves not to be instructive. Here Ascat imagery did not appear until 01 Aug 2012 so that year is not available. The June 4th images for 2013-2019 are shown in the second animation. They would no doubt benefit from open water masking but there would be no benefit to understanding the remarkable current drift year. (Recall 2017-18 had anti-TransPolar drift completely orthogonal to conventional depictions near the Pole.)

The ice pack is largely broken up into free-standing floes in the Lincoln Sea and north; the western CAA is also being reduced to rubble by the CW rotation of the ice.

[A-Team]

Here I am testing an AI algo for removing passing weather artifacts from a Smos-Smap ice thinness color overlay on Ascat for 66 days beginning April 1st and ending June 5th 2019. Smos-Smap v205 looks like it will be continuing throughout June and maybe the whole summer.

See #27 for a similar overlay ending two weeks earlier; note its original color key survives intact whereas Ascat processing map of sigma-0 is quite complex though original brightness values can still be read if a copy of the original layer is left floating underneath as the cryptic active layer.

https://seaice.uni-bremen.de/data/smos_smap/png/north/2019/

The basic idea is that passing polarization in the atmosphere confuse the satellites more and more as the melt season progresses. This gives rise to rapidly moving artifacts that only last for a few frames. Colors persist and grow outwards from where the ice is genuinely thin. Oddly AMSR2, while having weather artifacts of its own, has them at different places and so can be used to screen Smos-Smap but here one wonders if it is a swath timing issue.

Another approach that helps later in the season is to discard isolated colors in the central basin. Operationally, the contiguous (connected) color picker in Gimp can select the valid thinness fringing the ice pack and near troublesome island polynyas like Wrangell's. At any rate, the second mp4 below had many of the putative artifacts in the original mp4 removed.

However there is always the risk of removing something unusual and newsworthy like a newly healing lead or thin ice at the Pole so it may be better just to view as-is with the understanding improbable isolated colors are likely just passing weather not representing anything in the ice.

Thinning has really accelerated in the last 4-5 days on the Svalbard-FJL line (gif). The beige in Smos-Smap turns to color at 0.5 m and below. In other words, a lot of ice was very thin already, in the 0.5 - 0.7 m range.