MOSAiC news

[A-Team]

On the snow layer over Arctic sea ice, we really don't want snow thickness but rather (1)  its insulating qualities (called R-value in the US) that reduce the coupling between cold air above and bottom freezing below, (2) the snow's water equivalent (SWE) that contributes to melt ponds, floe buoyancy (flooding of negative freeboard) and drain-off flushing and freshening of salinity stratification, and (3) its time-dependent albedo that governs solar energy uptake.

However snow thickness is about all that's directly observable by satellite, though buoy thermistor chains could potentially measure the temperature gradient through the snow to get at local insulating properties as they evolved over time.

The Polarstern observed only five snows during an entire year but importantly documented its redistribution during wind events to the lee of pressure ridges. A 3km resolution snow thickness file is thus going to provide a very misleading average since ridges are very narrow linear features. As a Laptev-Fram TPD drift, the Polarstern could not observe snows away from their helicopter-widened swath. N-Ice2015 got good data on negative freeboard but only for half a winter and only for a few sq degrees north of Svalbard.

The neXtSIM project produces a quite detailed daily snow thickness layer, apparently from assimilation of ECMWF weather products. This is shown below for 60 days from Oct 15h out to forecasted Dec 13th in the rainbow palette at 70% transparency over its ice area fraction layer darkened to show faux leads and fractures whose reality lies within its Maxwell Elastic Brittle calculations (rather than say a WorldView infrared heat scene).

It's not implausible  for these features to be in the Beaufort area as neXtSIM displayes. While it's likely all solid ice, not literally open leads, the modeling still seems to be doing something useful in identifying substructure within the main icepack. The C++ code, seemingly not at github, takes 4 hours on an ordinary laptop but is run massively parallelized on a Norwegian supercomputer center.

Lobelia makes eye-catching animations out of the CMEMS-furnished data; the starting set-ups can be seen at the links below. The mp4 default is already decent but the gif option is easier to crop, fix legends and change speed. It's fairly slow most days but it helps to reduce size of web browser window before starting video generation.

To get a pause at the end of the mp4 (made from the avi made from the Lobelia gif), 10 duplicates copies of the last frame were added using ImageJ's frame sorter tool to duplicate the last frame, followed by concatenation. This gives only a short pause in the mp4 loop, perhaps because mp4 compression compresses consecutive identical frames a lot. This is a bug in the mp4 specification that gif got right.

The second mp4 shows 90 days of TransPolar Drift out the Fram, from Sep 15th to Dec 13th. NeXtSim's modeled area only includes the upper Fram; still, it provides a very effective visualization (that we hope is accurate).
 
“WMS servers are a bit slow today” -- Lobelia

Wiki: A Web Map Service (WMS) is a standard protocol developed by the Open Geospatial Consortium in 1999 for serving georeferenced map images over the Internet. These images are typically produced by a map server from data provided by  a GIS database.

A WMS server usually serves the map in a bitmap format, e.g. PNG, GIF, JPEG, etc. In addition, vector graphics can be included, such as points, lines, curves and text, expressed in SVG or WebCGM format.

WMS specifies a number of different request types, two of which are required by any WMS server:

   • GetCapabilities – returns parameters about the WMS (such as map image format and WMS version compatibility) and the available layers (map bounding box, coordinate reference systems, URI of the data and whether the layer is mostly opaque or not)
   • GetMap – returns a map image. Parameters include: width and height of the map, coordinate reference system, rendering style, image format
Request types that WMS providers may optionally support include:
   • GetFeatureInfo – if a layer is marked as 'queryable' then you can request data about a coordinate of the map image.
   • DescribeLayer – returns the feature types of the specified layer or layers, which can be further described using WFS or WCS requests. This request is dependent on the Styled Layer Descriptor (SLD) Profile of WMS.[12]
   • GetLegendGraphic – return an image of the map's legend image, giving a visual guide to map elements.

https://tinyurl.com/y6mxph99
https://tinyurl.com/y5amk3tv

[A-Team]

hope neXtSIM can do some reanalysis with '93-'19 data

That must be the plan but at this point, CMEMS is only serving neXtSIM imagery back to 01 Nov 2019. None of that seems based on reanalysis. The original preprint only addressed four months in winter; reviewers wanted a full year to see how the model did on first year ice and the MIZ.

They now have 13 months, including all but the first October of the Polarstern expedition. The issue around scaling is whether neXtSIM can really predict the chaotic ice movements during windstorms that were captured so extensively by bow radar (see up-forum). Those observations took in 75 sq km which is 8-9 pixels of neXtSIM product.

neXtSIM is currently providing one hour steps which already involves a massive daily computation. That seems to be getting ahead of time increments of what they are assimilating (daily or 3-hourly).
As with ECMWF or GFS, forecasts by neXtSIM-F are placed by rolling re-initializations so there is a recompute burden there (7 days x 24 hours = 168).

To some extent the algorithms are still undergoing revisions so the notions of going back to 1993 or re-doing with reanalysis would involve a lot of supercomputer time. They can scale to 512 processors  partitioning a mesh with border hand-offs but the process isn't yet fully parallelizable.

The purpose of hourly is presumably to provide more timely marine operational data, ie help ships find leads and avoid compressional ice. However there aren't any ships in central Arctic ten months a year. This would just leave Svalbard fishing and tourism and already-escorted NorthEast Passage bulk carriers. The US Navy product Hycom states right up front that their's IS NOT an operational product.

Looking at 3-4 days of hourly across a 05 Dec 2020 high wind stress event between Ellesmere and Banks islands is watching paint dry. It's easy enough in ImageJ to reduce an hourly stack to 3-, 4-, 6- or 12-hourly. For example, 3-hourly would allow synchronization with GFS nullschool wind stepping. Speeding up frames per second fixes the pace on hourly but the only benefit is cinematic smoothness -- not worth it for tripling file size.

How does neXtSIM compare in quality to other ice model tools such as Hycom and PSL? Is there agreement on ridging and leads? The models are hard to compare as they issue different products in different projections. neXtSIM’s Maxwell Elastic Brittle rheology really only comes into play during high winds (since ocean currents are mostly minimal); otherwise it is just daily assimilated satellites and ice-air-water thermodynamics.

The first gif below shows three overlapping 2-days of OsiSaf ice motion vectors at 6x exaggeration. These are observational AI, not model. neXtSIM also computes these vector displacements but CMEMS/Lobelia cannot currently display more than separate x.y components. To compare would require 48 hours of hourly vector addition which would not be easy for Lobelia to implement. The final gifs shows GFS winds and 2m air temperatures, then various Hycom products for the four days; the same could be done for the many PSL products (2nd link).

https://earth.nullschool.net/#2020/12/04/1500Z/wind/surface/level/overlay=temp/stereographic=-45,90,2100/loc=-132,78
https://psl.noaa.gov/forecasts/seaice/

[oren]

The second mp4 shows 90 days of TransPolar Drift out the Fram, from Sep 15th to Dec 13th. NeXtSim's modeled area only includes the upper Fram; still, it provides a very effective visualization (that we hope is accurate).

The eye candy is amazing and the science level is way over my head (kudos A-Team for keeping up and summing up so much relevant research for forum ignorants such as myself). And this is probably irrelevant nitpicking. However, surely the model is not accurate at the Lincoln Sea entrance of Nares Strait. Compare to any animation posted by uniquorn recently, the model seems to be missing the flow down the strait and the familiar oval Lincoln region where ice crumbles towards the exit (unless and until a stable northern "arch" forms). Maybe it's below their resolution level but it feels like a serious omission.

[A-Team]

Oren: neXtSIM model is not accurate at Lincoln Sea entrance of Nares Strait. Compared to uniq AMSR2 animations, the model is missing the flow down the strait and the Lincoln region where ice crumbles.

That's correct. neXtSIM needs a slab ocean so given narrow CAA channel currents with an admix of small islands, boundary conditions are hard to specify for the main momentum differential equation. The region of the Laptev south of the NSI is also in a world of its own due to massive freshwater input from the Lena and early formation of landfast ice.

It's not clear how well neXtSIM works with increasingly dominant FYI, especially late-forming ice this year in the Laptev. Performance may decline too when the wind has a strong turning angle (tight cyclonic events) or sharp shearing reversals in direction over short distances.

I have not yet come to an opinion on the merits of neXtSIM in the central Arctic -- the papers  provide minimal validation in the sense we understand it in forums. Its five layers comprise a small part of the many climate resources hosted at CMEMS (which is invitation-only). Lobelia's incredible display software will make anything with 3km resolution look good.

One of the main purposes of Mosaic was to provide on-the-ground refinement of satellite imagery and ice thickness models. Drifting snow undercut microwave sensor experiments; rapid TPD and quarantines reduced transect synchronized airplane overflights.

However the Polarstern's storm weather data may provide validation opportunities for neXtSIM -- high wind stress gives rise to its lead and ridge features as it drives ice motion. Sorting 8,094 hours of AWI wind speed data in Excel for events exceeding gale force winds (>half day of >14m/s winds at the 10m weather mast) identified 16 storm events lasting for 421 hours total (attached as csv).

Intersecting high wind stress event dates with the coverage windows of neXtSIM, Polarstern bow radar,  high latitude ship position and availability of Sentinel 1AB coverage (csv up-forum) picks out the better validation candidates on the scatter diagram of storm duration in hours vs mean wind speed (Fig.2).

Looking briefly at the storm of 13 May 2020 at the time of highest wind speed 23 m/s (Fig.1) is underwhelming: neXtSIM isn't showing overt action in 'area fraction' layer at the lat lon of the ship. This neither validates nor invalidates the product. Quite a bit more could be done using mp4, other dates and other neXtSIM layers.

lat  lon    date    time   m/s dir  ºC   #E  dur  ave  mag
83.4 11.5 13.05.20  03:00  23  50   -3.5  5  33   17.5 576

A 2019 control sequence for the strong cyclone this summer in the Beaufort-Chukchi region around 29 July 2020 is shown in the third gif below.

[A-Team]

Here is neXtSIM's view of the late July 2020 cyclone itself, twenty days shown to pick up some flanking. Seems like it does quite a good job in providing specific detail, if perhaps a bit exaggerated. There is all manner of earlier satellite documentation of this event up-forum for those dates, including dramatic WorldView visible imagery of the cyclone spiral (with AMSR2_UHH perhaps more revealing), for which matching hourly neXtSIM would be of interest. On the melt forum, near:

https://forum.arctic-sea-ice.net/index.php/topic,3017.msg277907.html#msg277907

It is feasible to do an entire 3 month summer of neXtSIM at decent resolution for a file size of 1.6MB as a grayscale mp4 done through an initial gif, depending on how slow/popular the Map Server is at the time of request. The process is completely independent of home internet speed.

Thus an entire year could be shown by concatenating these at 6.4MB. June 15-Sept 15 for 2019 leading up to the Mosaic year is attached; summer and fall 2020 are in the works. It is also feasible to switch into hourly mode to slow events of interest and clip those in.

What exactly neXtSIM is showing remains to be established; whatever, it's quite interesting.

To see how much the 7-day forecast differs from that which will be finalized in the archive 7 days from now, I saved out from Dec 1st on out to Dec 14th (though I am -7 hrs from UTC and today is Dec 9th here). It will simply create multiples of the last date it has, here Dec 14th, if you put in say out to Dec 21st (which can be useful in the mp4 mode for creating a pause at the end of a cycle. The palette squeeze on 'sea ice area fraction' necessary to get enough contrast is 0.91 to 1.00.

When the forecast is replaced next week, the gifs can be subtracted to see how much things changed between the two initializations. This can be done with 'grain merge' in gimp on the layered tile-up, followed by 'equalization' and the 'union jack' LUT in ImageJ.

Note Lobelia's browser url is enormous, 2187 characters most of which are completely unintelligible. However the zoom is given, here at 11.228 but relative to what, the 3x3 km resolution? There's no purpose to an equal area projection if we don't know how to multiply ROI pixel counts to get sq km areas.

EPSG 3408 in the url is the equal area NSIDC grid discussed a few posts back. Replacing that in the url with EPSG 3413 (polar stereographic) did not bomb but nor did it produce a map in that projection. Nor can the zoom be manually changed in the url as it can in nullschool. The time is probably unix, seconds since 00:00:00 UTC of 1 January 1970; we see this sometimes in buoy databases.

https://tinyurl.com/y2eoawv7 will reproduce the starting set-up.

https://cmems.lobelia.earth/data?view=viewer&crs=epsg%3A3408&t=1608163200000&z=0&center=120.4395%2C82.23817
&zoom=11.22&layers=W3siaWQiOiJjMCIsImx...

[Glen Koehler]

     Are the neXtSIM animations going to be available near real-time to track the 2021 melt season?  Even if the view is biased above absolute values, as long as the bias is consistent those animations would provide a useful view of melt season progress.

[A-Team]

will neXtSIM animations be available near real-time to track the 2021 melt season?  as long as the bias is consistent, useful view of melt season progress.

Presumably, though we don't know at this point whether the algorithm, satellite assimilations and defining preprint will be the same or improved. Note neXtSIM goes 7days ahead so it always has current date. Evaluation of the area fraction forecast is in progress as described in the previous posts above. (The other four forecast layers need separate analysis.)

NERSC is currently posting daily 7 day forecasts at 05:00 utc such as ice drift, not currently hosted at CMEMS (for want of vector arrows). The mp4 below (bottom) is of special interest because of the developing anti-cyclone this week. This represents an advance over OsiSaf which cannot make predictions. They made a detailed comparison to OsiSaf in 2016:

OSISAF drift product provides a 2-day displacement vectors on a 62 km grid by averaging the drift of spatially and temporally higher resolved drift measurements derived from feature tracking. To compare with OSISAF we derive the 62 km 2-day displacement by applying the same algorithm to the node displacements of the Lagrangian mesh. The OSISAF data comes with a quality flag for each observation on each grid point. We only use observations with the highest quality, i.e., which contain no spatial interpolation from neighboring points.

ftp://ftp.nersc.no/nextsimf/arctic_forecast_cmems.ma10km/
https://tinyurl.com/yyjgp36z poster

Spring 2019 is shown below. Here the scale compression had to be changed from 0.9-1.0 to 0.7-1.0 to keep the later melt season images from getting too dark. It would be feasible to overlay the Polarstern's position on the gif prior to making the mp4.

Spring 2020 will also have AMSR2_AWI for the first time (and maybe back to 2012). It is not very informative in mid-winter because sea ice concentration is close to 100% everywhere, with Smos-Smap providing better details on the exceptions. AMSR2_AWI provides its netCDF which CMEMS is not (or has it buried). This is key to getting all the data into the same projections as well as making vector arrows out of ice velocity.

There are a fair number of easily fixed bugs in both Lobelia and CMEMS. However neither provides a contact email or reporting mechanism! Four of the bugs are shown in Fig.2. They also need a fixed width smaller font for the dates block-dropped without the comma and choice for placement; UTC can be assumed in scientific graphics and so omitted. Along with the tacky 'My Ocean' overlay. The text boxes for palette squeeze are very buggy, swapping in numbers without end user permission.

These animation have mostly been of the 'sea ice area fraction' which mostly seems to show leads and ridges. Another one of great interest to us is the 'sea ice thickness' layer. That is provided below for 124 days of the past summer. Is the thickness scale accurate on either a relative or absolute basis? That really would require some input from Mosaic (note neXtSIM averages over its 3km pixels).

These animations are easy to produce using Lobelia defaults but require quite a few small tricks to get them optimized for the forum. I can write these up if use of the CMEMS site becomes more widespread.

[A-Team]

The two animations below show the growth of ice thickness this fall. The Cryo2Smos product has colored rings at 0.5, 1.0 and 2.0 meters of thickness that move radially out from the CAA allowing the 'front' to be tracked across the available dates Oct 15 to Dec 09th. (The radial motion will reverse direction shrink the perimeter in melt season.)

To actually measure growth rate between two rolling dates say ten days apart at a given position, the netCDFs can be subtracted in Panoply to generate that map.

The second mp4 shows Smos-Smap ice thinness (less than a half meter) progression. It would be somewhat problematic to measure the earliest rates of ice growth because of the complexities of bottom ice formation (which involved platelets in one published Mosaic underwater observation). Mosaic otherwise has not released any ice growth data though some of its buoys have.

neXtSIM apparently assimilates Smos ice thickness but the Bremen archives do not provide this at during summer months nor would Smos provide any data on the main thicker ice pack. Topaz4 may be the real source of ice thickness so its source (observation? model?) has to be chased down. If fresh updates are put in every morning, then neXtSIM is only calculating the hourly interpolations though neXtSIM-F is taking ice thickness seven days forward with its forecast.

Note mp4's provide first frame stills before playing and last frame stills if looping is turned off. There is no good way to provide end pauses in loops in contrast to gifs where this can be specified. Neither gimp nor ImageJ make, open, play or edit mp4; QuickTime can play and join clips; ezgif.com offers various free mp4 editing tools such as crop, speed, rotation, reverse, re-size, conversion back to gif etc.

[A-Team]

One approach to determining rates of sea ice thickening across the Arctic Ocean is subtracting Cryo2Smos products from different dates. This is somewhat problematic since a week or so is necessary to collect enough swaths. Below, a 20 day interval is used (Nov 21 to Dec 10) to allow for this. (A 30-day window gets into fall freeze-up pattern differences; a rolling 20-day window would better show consistency and rate trend, not shown.)

Panoply can subtract the netCDF arrays ok but adjustments have to be made in using symmetric divergent palettes (split around 0), ie make sure the scale range is symmetric about 0, here -0.70m to +0.70m was chosen which pushes outliers onto the bounding colors. (Panoply will default to the asymmetric ‘fit to data' setting which gives too much emphasis to even one pixel outliers.)

The rate of thickening over the twenty days, as shown, below is quite uneven. There are even interior areas where the ice has slightly thinned. Peripheral areas such as the southern Chukchi, Amundson Gulf and FJL are not treated meaningfully as shown in the AMSR2 comparison of ice edge for the two dates.

Cryo2Smos does not take ice motion into account so in places the 'same' ice is not being compared. Nonetheless, the upper bound on ice thickening appears to be 70cm/10days but more typically it might be 3-4 cm/day. Cryo2Smos provides an error map which needs consideration.

Below, a 20 day interval is used (Nov 21 to Dec 10) to allow for this. (A 30-day window gets into fall freeze-up pattern differences; a rolling 20-day window would better show consistency and rate trend, not shown.)

Mosaic made year-long measurements of the 'same ice' for one particular floe but that data has not been released. Some buoys set out also do this as tracked by uniq.

neXtSIM seems not to offer netCDF but the grayscale thicknesses for 12:00Z on the two days can still be subtracted in gimp ('grain extract') and displayed in the divergent red-blue color table of ImageJ. The comparison to Cryo2Smos does not really offer validation to either because both are physically uncertain. neXtSIM shows ice motion quite well, suggesting that averaging is all that makes sense in comparing distant (or even close) dates.

/=/=/=/=/=/=/=/

neXtSIM-related papers (reverse chronological order, update of #1299, +uniq): note two new articles on waves hitting the ice edge, as well as two reviews of neXtSIM-type models.

Wave–sea-ice interactions in a brittle rheological framework
G Boutin, T Williams, P Rampal, E Olason C Lique  Jan 2020
https://doi.org/10.5194/tc-2020-19  manuscript accepted;  3 reviews)
https://www.the-cryosphere-discuss.net/tc-2020-19/tc-2020-19-AC1-supplement.zip
https://www.the-cryosphere-discuss.net/tc-2020-19/tc-2020-19-AC2-supplement.zip
https://www.the-cryosphere-discuss.net/tc-2020-19/tc-2020-19-AC3-supplement.zip

Towards a coupled model to investigate wave–sea ice interactions in the Arctic marginal ice zone
G Boutin C Lique … F Girard-Ardhuin  Mar 2020
The Cryosphere, 14, 709–735,
https://doi.org/10.5194/tc-14-709-2020

Should Sea-Ice Modeling Tools Designed for Climate Research Be Used for Short-Term Forecasting?
E Hunke et al  review Sept 2020
https://link.springer.com/article/10.1007/s40641-020-00162-y very readable review

Presentation and evaluation of the Arctic sea ice forecasting system neXtSIM-F
T Williams A Korosov P Rampal E Olason    25 Jun 2019 submission
https://tc.copernicus.org/preprints/tc-2019-154/  EGU meeting
https://tc.copernicus.org/preprints/tc-2019-154/tc-2019-154-AC1-supplement.pdf  reviewer #1
https://tc.copernicus.org/preprints/tc-2019-154/tc-2019-154-AC2-supplement.pdf  reviewer #2

Probabilistic forecasts of sea ice trajectories in the Arctic: impact of uncertainties in surface wind and ice cohesion
S Cheng, A Aydoğdu, P Rampal A Carrassi L Bertino   17 Nov 2020
https://arxiv.org/pdf/2009.04881.pdf

On the statistical properties of sea ice lead fraction and heat fluxes in the Arctic
E Olason P Rampal V Dansereau January 2020 preprint
https://tc.copernicus.org/preprints/tc-2020-13/tc-2020-13.pdf

On the multi-fractal scaling properties of sea ice deformation
P Rampall V Dansereau et al 2019
https://tc.copernicus.org/articles/13/2457/2019/

Impact of rheology on probabilistic forecasts of sea ice trajectories: application for search and rescue operations in the Arctic
M Rabatel P Rampal et al   2018
https://tc.copernicus.org/articles/12/935/2018/

Parallel implementation of a Lagrangian-based model on an adaptive mesh in C++: Application to sea-ice
A Samaké P Rampal S Bouillon E Olason 2017
https://www.sciencedirect.com/science/article/pii/S0021999117306368

Wave–ice interactions in the neXtSIM sea-ice model
T Williams P Rampal S Bouillon   2017
https://d-nb.info/1142799980/34

Probabilistic forecast using a Lagrangian sea ice model: application for search and rescue operations
M Rabatel P Rampal et al 2017
https://pdfs.semanticscholar.org/f7fc/b5551b8a352e271999ff736c5674b2d3afa9.pdf

Ice bridges and ridges in the Maxwell-EB sea ice rheology
V Dansereau J Weiss P Saramito et al   2017
https://tc.copernicus.org/articles/11/2033/2017/tc-11-2033-2017.pdf

neXtSIM: a new Lagrangian sea ice model
P Rampal S Bouillon E Olason M Morlighem  2016
https://tc.copernicus.org/articles/10/1055/2016/tc-10-1055-2016.pdf
https://tinyurl.com/yyz2xgwj large meeting poster

Sea ice diffusion in the Arctic ice pack: a comparison between observed buoy trajectories and the neXtSIM and TOPAZ-CICE sea ice models
P Rampal S Bouillon J Bergh E Olason 2016
https://tinyurl.com/y3vx5bmg

Arctic sea-ice diffusion from observed and simulated Lagrangian trajectories
P Rampal S Bouillon J Bergh E Olason    2016
https://tc.copernicus.org/articles/10/1513/2016/

A Maxwell elasto-brittle rheology for sea ice modelling
V Dansereau J Weiss et al 2016
https://tc.copernicus.org/articles/10/1339/2016/tc-10-1339-2016.pdf

Presentation of the dynamical core of neXtSIM, a new sea ice model
S Bouillon P Rampal 2015
https://www.sciencedirect.com/science/article/abs/pii/S1463500315000694

Status and future of global and regional ocean prediction systems
Marina Tonani et al  review Oct 2015
https://www.tandfonline.com/doi/full/10.1080/1755876X.2015.1049892

Scaling properties of sea ice deformation from buoy dispersion analysis
P Rampal J Weiss et al    2008
https://doi.org/10.1029/2007JC004143

[A-Team]

A strong anti-cyclone arising on the Beaufort/Chukchi side on Dec 8th is projected to continue at least through Dec 20th, enveloping the whole Arctic Ocean. High pressure reached 1053 hPa late on Dec 13th with winds attaining 106 km/hr, shown in the mp4 below of eleven days of 3-hour GFS weather.

https://www.nature.com/articles/s41467-019-13299-8   cyclone tracker
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2017GL076446 Beaufort Gyre collapse

The Polarstern did not encounter sustained winds this strong during its entire year, winds peaking at 23 m/s vs 29.6 here. However lesser storms did cause dramatic ice fracture as shown on 6-hour bow radar.

Because so little open water remains now, wind fetch is short and no swells can develop. Ice has blown into the Chukchi (rather than forming there) where it has mostly melted so far.

Initially the wind pattern supported ice co-rotation as a classical Beaufort Gyre though the storm quickly broadened so much that ice would only continue in a partial gyre along the Alaskan coast but never return to the western CAA.

Only a handful of days in recent years have supported the Gyre pattern according to OsiSaf ice motion (https://tinyurl.com/yd5fz5ww). Prior to OsiSaf's start-up in Oct 2015, sea ice motion was not quantitatively tracked ocean-wide with any gridded accuracy. However buoys completed Beaufort Gyre circuits in the 1980's (though they don't today) and low resolution ice age movies are available for 1984-2019.



Because wind stress on the ice pack is been very high and unequally applied, many new leads and pressure ridges are expected. With high pressure, clouds are scarce and visibility is excellent on Suomi band 15 infrared as displayed at WorldView. Leads, even refrozen, are heat leaks and show up as bright lines.

This storm provides a validation opportunity: comparison of satellite-observed leads with neXtSim calculated and forecast leads, per its CMEMS displays of diminished ice area fraction and ice thickness. However current product glitches and today's implementation of a new rheological model suggest a few days wait is advisable.

However 11 days of neXtSIM as currently offered mostly shows healing of existing fracture features during the storm along with older ones in the CAB persisting. The extent with which neXtSIM agrees with Suomi band-15 still needs to be determined.

WHITE OCEAN
New product Arctic Ocean Wave Hindcast
Updated of the stand-alone ice model for the Arctic Ocean Physics Analysis and Forecast (new rheological model used)
Upgrade of the Baltic Sea ice thickness automated products – by including Sentinel-1 IW VV/VH mode
Additional variables in Arctic Ocean ice chart product (total ice concentration, ice type, ice state of development, melting states) –format shapefile
Arctic sea ice displacement at low and medium resolution (sensors ASCAT/MetOp-C & AMSRE/AMSR2)
New Product based on Sentinel-1 satellite images for sea-ice applications covering a part of the Arctic Ocean (in Geotiff format)
https://marine.copernicus.eu/services-portfolio/product-improvements/#dec2020

[A-Team]

The anti-cyclone is forecast to continue out to at least Dec 21st with a high remaining in the 1040's hPa for the 14th day, though the center has shifted closer to the FJL side with shear-off at Nord facilitating Fram export. The persistent high has led to cloud-free conditions over much of the Arctic Ocean and a rare opportunity to track heat leads and ice motion at spectacular resolution (Fig.2).

The westward advection of continues off the Alaskan coast and could be confused with Beaufort Gyre conditions though the 1500km diameter of the high is far too wide for ice to complete the narrow gyre circuit. Days of smaller diameter gyre motion over the Beaufort are very rare on OsiSaf.

Not a single buoy (or multi-buoy succession) over the last 40 years has ever completed the oft-pictured full gyre (per uniq's ongoing displays).

CMEMS has resumed production of neXtSIM-F under an updated but as-yet undescribed rheology model that has removed previously noted glitches. The lead-like features do not correspond well with heat leads observed by Suomi band 15 infrared but both exhibit new features and CCW translation under the anti-cyclonic winds. Predicted neXtSIM features thus cannot serve as the basis for quantitative modelling of winter heat loss.

CMEMS help desk has been helpful though proposals for fixes have to pass through a long chain before entering the priority queue.

[Glen Koehler]

<snip> 
    Not a single buoy (or multi-buoy succession) over the last 40 years has ever completed the oft-pictured full gyre (per uniq's ongoing displays). 

    I'm glad you pointed that out.  Watching those buoy drifts I figured that my confusion at not seeing  a big sweep around the famed Beaufort Gyre as expected was yet another manifestation of my ignorance about Arctic complexities.  If there had only been buoy tracks from the last few years I might have assumed that tracks from the 1980s would have displayed the expected majestic journeys around the Arctic.  In which case I might have cried out to the heavens that the sky is falling because the recent buoy tracks indicate that the epic Beaufort Gyre has collapsed!   But those buoy tracks from the early 1980s don't show much gyre action either, while they do show Fram export. 

    Is the Beaufort Gyre more hype than reality?  Or do the buoy tracks not accurately represent the full degree of drift at the temporal and spatial scales in the animations?  Or am I completely missing the point about what the buoy track animations do, don't, and should show?

[Tor Bejnar]

I watched a GIF or movie A-Team posted a few days ago showing ice movements over many years (20 or 40).  In nearly all years some ice survived the summer in the Beaufort and as it rounded (clockwise) into the 'Atlantic' side of the mid-Arctic, about 3/4^ths of the identifiable 'once in Beaufort' floes headed towards Fram (or other parts South) and about 1/4^th headed back into the Beaufort.  Some 'years' none of the ice made the circuit but one or two 'years' had lots of ice make it all the way around.  (I use the term 'years' loosely; I did not notice how long it took a bit of ice to complete the circuit [in those instances when it did].)  It is likely some ice made the circuit that wasn't obviously 'one in Beaufort' ice, but it was.  The Beaufort Sea isn't the only 'starting place' in the circle, of course.

It was 'obvious' from that movie that Beaufort ice that survived the summer generally moved westward along the Alaskan coat (if in the SE Beaufort) then westward toward Siberia then westward toward Europe then toward Canada.  A "gyre" that takes surface ice 3/4ths around the circle (and some ice all the way around) is a gyre in my book.

[uniquorn]

Will probably try an overlay comparison of buoys on to (modelled) sea ice age once I've worked through all the data.
A comparison of Eulerian and Lagrangian methods for dating in numerical ice-sheet models

A new tracking algorithm for sea ice age distribution estimation

The new algorithm is driven by the new sea ice drift products from OSI SAF, which is void of potential artifacts due to inclusion of autonomous ice drifter buoys. This leads to a more homogeneous distribution of ice age fractions over the Arctic Ocean. The algorithm is also constrained by the observed sea ice concentration from OSI SAF, which reduces fractions of old ice and, consequently, ice age by 20–30 %. It  was  applied  to  generate  time  series  of  daily  sea  ice  age fraction product from October 2012 to October 2017. Comparisons with the NSIDC SIA time series indicate that the fractions of MYI in the new product melt faster during the year and after a spin-up time of 3 years the area of older ice in the SICCI product is almost 20 % lower than in the NSIDC product.Data availability.The   data   generated   with   the   algorithm   are openly  available  at  FTP  (for  bulk  download;  ftp://ftp.nersc.no/ArcticData/esa_cci_sea_ice_age/)  and  at  THREDDS  (subsetting and online visualization; http://thredds.nersc.no/thredds/arcticData/esa-cci-sea-ice-age.html) in netCDF format following CF conventions containing values of sea ice age fractions concentrations, MYI concentration,  and  sea  ice  age  computed  using  weighted  average(Korosov  amd  Rampal,  2018).  Detailed  information  on  datasets used in this paper can be found in Sect. 2

my bold
bulk  download;  ftp://ftp.nersc.no/ArcticData/esa_cci_sea_ice_age/

Is there a Beaufort Gyre thread? I think there should also be some distinction between an ocean gyre and an ice drift gyre.

colour version of ice age 2000-2020 here. It tends to support the case for a small gyre  in the Beaufort, but not so much the large one that extends to the Lomonosov ridge. Similarly for ice age 84-99(not shown) imho.

[A-Team]

gyre doesn't have to mean gyre

'I don't know what you mean by "gyre",' Alice said. Humpty Dumpty: 'Of course you don't — till I tell you. I meant 'there's a nice little drift for you!' 'But "gyre" doesn't mean "a nice little drift",' Alice objected, 'it's meant circular for several thousand years.'  'When I use a word,' Humpty Dumpty said, 'it means just what I choose it to mean — neither more nor less.'

I saw a partial full ice gyre once on an oblique low resolution 40-year youtube

A pity you didn't take note of the date. 

there's a distinction between an ocean gyre and an ice drift gyre.

Right. Many oceanographic writings avoid ice motion, or the lack of it, that we see in buoy tracks and satellite time series. They are talking about inferred deep water currents and a large pool of slightly reduced salinity (confusingly called "fresh water"). There is not a single drop of fresh water anywhere in or under the central Beaufort Sea. T

he 28-frame gif below collects some of the silliness surrounding 'Beaufort Gyre' depictions -- it's whatever and wherever you want it to be! Publishing garbage diagrams undermines public trust in overall climate science -- we cannot afford that.

https://sci-hub.se/https://doi.org/10.1029/2018JC014379 rare serious treatment
https://cpom.org.uk/glimpsing-under-the-ice-measuring-the-ice-covered-arctic-ocean-from-space/

OsiSaf/GFS have shown over and over that when the wind does not blow, the ice does not move. That means no surface currents of any significance (Nares, Fram, Yermak, FJL, Bering excepted). Ice motion in the central Arctic Ocean is entirely wind-driven.

That's a problem for the 'Beaufort Ice Gyre' because the 'Beaufort High' is rarely where it needs to be. Often the high is bounded on its south by land, causing strong winds up the Alaskan coast that make ice drift rapidly west there (before breaking apart melting out in the Chukchi as 'Big Block' did in 2016). The problem is the opposite side of the 'Beaufort High' is often elsewhere, failing to move the pack ice in a gyre (Fig.2).

The Polarstern measured currents, tides, turbulence and eddies under the ice during their long drift but has not released any data. The TransPolar Drift, while highly variable and largely unpredictable, is not as frothy a subject as non-existent Beaufort Ice Gyre but might better be called the CircumPolar Drift because  in the last decades, ice has rarely drifted over the north pole ('trans' means across) as it did during the Mosaic year.

CMEMS, our main online source for oceanographic time series, provides very detained views of Arctic Ocean ice motion from 01 Nov 2018 into late Dec 2020 with its daily neXtSIM product. At some point, that product will be extended much farther back in the satellite record (that's used for daily assimilations).

For now, it allows complete coverage of the Mosaic year, as well as regions elsewhere in the Arctic. However file size is an issue for the full 781 days, especially for the 2.1 GB gif. A small preliminary version of the western sea ice motion is attached below.

Although neXtSIM overlaps (ie extends) Polarstern 6-hourly bow radar scale, the latter is confined to its drift track. The current persistent high associated with the anti-cyclone has allowed an extraordinary and continuing basin-wide view of heat-leaking leads via Suomi Band 15 infrared.

The attached gif shows the full range of scale zooms for the available resolution. While ice fractures aren't really fractal in the mathematical sense, new leads become visible the larger the scale -- the Beaufort ice shown has had a surprisingly extensive history of fracture. These continue to leak heat more rapidly than their matrix until such time as the thicknesses more or less equalize.

[Glen Koehler]

    Dear A-Team. It seems like the Beaufort Gyre is a bit of a sore toe for you, and this thread just stepped on it!  I agree that there is a need for precision in scientific terminology and that it is important to not create entities out of spurious patterns.  I have similar pet peeves about sloppy terminology in my own field (anytime I read or hear "herbicides and pesticides" a part of my reptilian brain stem is fired causing me to view anything else the author says through a lens of hostile suspicion as being based on polemics and ignorance and probably not worth my precious ever-dwindling time on Earth. )

    But there is in fact a regular circular pattern of ASI travel, for which the center of rotation is usually located in or near the Beaufort Sea.  To discuss this phenomenon, a name is more efficient than calling it "that regular circular pattern of ice travel with a center of rotation usually near the Beaufort Sea."  I suppose the acronym "TRCPOITWACORUNTBS" could suffice, but that's too long and reminds me of one of those compound German words for which I pity German school kids having to spell. 

     Gyre refers to circular motion.  And Beaufort refers to the Beaufort Sea.  Voila, we have a name: the Beaufort Gyre!  As long as people don't over-concretize it as a specific fixed thing instead of a regular but variable pattern, and don't confuse wind-driven ice movement with ocean currents, I don't see the harm. If we want to discuss it we gotta call it something.   Later in your post you call this regular pattern "the non-existent Beaufort Ice Gyre".  But everybody else seems to agree that it occurs on a regular-enough basis and is of sufficient importance that they need to discuss it in order to understand the system.  So how should they refer to that regular pattern?

     You include two links.

https://sci-hub.se/https://doi.org/10.1029/2018JC014379 rare serious treatment
https://cpom.org.uk/glimpsing-under-the-ice-measuring-the-ice-covered-arctic-ocean-from-space/

     For the first article by Regan, Lique and Armitage, you credit it for being a "serious treatment".  But placing it right after a condemnation of "Publishing garbage diagrams undermines public trust..." is not exactly a rousing endorsement, and feels more like an insult.  Given two contrary indicators, I don't know how to interpret your tone.  I'm not heavily invested in understanding what Regan et al. call the "BG" (not to be confused Roald Dahl's BFG for Big 'Friendly' Giant, the greatest subversively dirty acronym of all time), but that paper looked pretty good and useful to me.

    But what really got me to write was to defend the second link.  I found that brief blog post (4 pages including pictures and graphs, so even less to read) by Tom Armitage provides some very useful perspectives on how the Arctic Ocean works as a system and how it interacts with the rest of the planet.  If you were upset because Dr. Tom (like the Regan et al. article) commits a cardinal sin in calling lower salinity water "freshwater", I think you'll get better results by politely suggesting that the term should be "fresher" or "lower salinity" water.  While I agree with you that water with so much salt in it that it would probably kill you if you tried to survive on it should not be called "freshwater", Regan et al. did at least define "freshwater" as water with salinity below a reference value of 34.8.

    Apart from fights about science terminology, I encourage those of you hanging out on ASIF, which means that you must have an interest in the topic, to read the Armitage blog post.  It's a quick read and has a high "Oh really, I never knew that or thought of it that way" moments per page ratio.  And it has a pretty picture that I was tempted to re-post on the "caa-greenland mega crack" page for the all those crackheads who hang out there.  

[oren]

he 28-frame gif below collects some of the silliness surrounding 'Beaufort Gyre' depictions -- it's whatever and wherever you want it to be!

As usual, very interesting discussions in this thread.
I thought that gif was brilliant. Terminology should be clearer. And though I am but an amateur, I "know" an ocean gyre is a persistent circulatory current. So the Beaufort Gyre by right should be called what it is, a sometimes-circular sometimes Beaufort-centered statistical ice motion in the Arctic ocean. As A-Team explains, no current underneath, no gyre. If a shortened name is desired, it shouldn't include gyre.
(Really out of my depth here though).

[uniquorn]

update on mosaic tbuoy temperature profiles. Near surface temps lower than -40C on feb6.

[uniquorn]

Beneath T85

Beneath T81, thickening and SST change. Could really do with some salinity measurements there.

The same T81 data with Loess smoothing, y~x, span 0.1  Data points every 6hrs looking pretty random apart from overall temp change over the 3m depth.

[uniquorn]

Forgot about 2020O10 that is on the same mosaic floe2(I think). Note that salinity rises at 50m when temperature drops.
Maybe FOoW will comment on this one.

[uniquorn]

Well we have another chance to develop our thickness estimate algorithm soon. Core ice temperature is dropping to levels where we thought we had some reasonable estimates last year. I have doubts though.
Looking at T78 data the deployment report (1) states the ice was 1.52m thick with 2cm of fresh snow on aug23. Thermistor buoy sensors are 2cm apart so we can't really detect snow at that point but we are also given the sensor id of the air/snow interface as 33. Great. So the bottom of the ice on that day was at sensor 110.
The temperature profile then was almost flat by winter standards but a more detailed look shows the ice thickness quite well. (2)
Thermistors 33 and 110 have been marked roughly at the beginning of the animation(3) and again on nov22 when the temperature profile shows them clearly (with a short pause both times). There is not much additional thickening since then.
So my doubt is this. Did the ice melt quickly during end of august/september or has it been 1.5m thick and possibly porous all the time but took 3 months to cool down again?

added T78 drift path(4)

It's nice to have the actual thermistor air-snow and ice-water numbers at deployment.
Looking at the animation, it appears like there was massive bottom melt, and then re-thickening.
This is what I would expect in general with the given temperature profile.
At first, ice top is warmer than at the bottom. This surely brings about bottom melt, as heat trickles from above and the salt water eats away from below, and I think would also depend on the rate of drag of the ice over the water.
The image from Aug 23 clearly shows ice-water interface at thermistor 110.
Next, the temp profile is flat. My understanding is that the bottom would still be melting, albeit at a much slower pace.
The image from Sep 13 seems to be the last day of top melt, as well as the ice having a cold core again (cold gradient from the middle to the bottom).
Eventually, the ice top is colder than the ice bottom, with a gradient appearing through the ice. From this point on the ice starts bottom freezing and it's easier to spot where the ice ends.
The image from Sep 21 in my opinion shows the ice-water interface at around thermistor 55.
The image from Oct 15 seems to show the ice-water interface at thermistor 70.
The end of the animation again shows the ice to exceed thermistor 110.

Admittedly these are just generalizations, since I can't put any quantitative expectations as to the rates of melting and freezing. The rates do depend on the steepness of the warm or cold gradient though. Therefore I would not expect a whole lot of bottom melting, also given the rather late date of deployment. However, eyeing the animation, it would appear as if the bottom melted all the way to thermistor 55. Some possible explanations:
* Bottom was higher than thermistor 110 to begin with, with 110 being some protruding edge. However, the animation data does support this initial placement for the ice-water interface.
* Melting did not actually reach thermistor 55, though I can't see any other explanation for the cold gradient ending where it ends on Sep 23, except that it was the ice-water edge.
* The ice was extremely mobile which enhanced bottom melt way beyond what the temp profile would suggest. This can be checked by looking at T78 drift.
* The ice has not cooled enough for the gradient to reach the ice-water interface.
* My intuition for rates of bottom melt is disconnected from reality (very probable).

All in all, a very interesting mystery, with wider implications. If indeed ice can bottom-melt so quickly at the end of the season, we are not so far away from blue ocean as the extent extrapolations seem to suggest.

Following up on Tbuoy thickness estimates. Here is an update of T78 temperature profiles, the blue line, with the Heat120 temperatures overlaid in black. A bit confusing in this format so some explanation:

Heat120 records the change in temperature after heating each thermistor for 120secs. The idea being that air, snow, ice and water will heat differing amounts, allowing us to see which is which. Some further interpretation is required.

Again thermistors 33 and 110 have been marked roughly at the beginning of the animation, marking the thickness of the ice the day after deployment date when the first heat120 cycle ran on aug24.

30 days later on sep22 the heat cycle shows a similar thickness but I think by now this is mostly rotten ice from bottom 'not quite' melt. Could perhaps be described as phase changing. This ice would probably be no match for an icebreaker or a heavy storm but just about maintains its integrity to the end of the season.

Skip to day 60 and we see a similar thickness but we can see the ice state changing from the surface downwards as the weather cools during October. Surface temperature on oct21 was about -20C.

By day 88, nov20 the ice is happy again, phase change back at depth almost complete and the heat cycle perhaps shows some signs of bottom thickening, also marked, optimistically, at thermistor160.

After that even a fertile imagination struggles to interpret small changes in temperature though I made a guess at day126, dec28 at thermistor170. Perhaps SimonF32's analysis will show more.
So 170-33 time 2cm is 2.74m, a possible increase of 1.2m.

I'd be interested in other interpretations.

Apologies for removing the 2 temperature scales. It was necessary for the overlay.

The warmer Heat cycle temps in the final frame are interesting....

Added t78 location, not far from the shear line
and amsr2 awi v103 for more detail.

[oren]

Thanks for trying to solve this mystery uniquorn.
Looking at my earlier comments on this issue, it seems my original interpretation was wildly off while these caveats were right:

* Melting did not actually reach thermistor 55, though I can't see any other explanation for the cold gradient ending where it ends on Sep 23, except that it was the ice-water edge.
* The ice has not cooled enough for the gradient to reach the ice-water interface.

It appears to me now the explanation lies in situations where the lower part of the ice is at equilibrium temperature of -1.8C, but has not turned into water as the phase change itself takes time and energy. This happens during bottom melt (e.g. to Sep 23) and during freezing. Unfortunately, this prevents us from simplistic estimates of where the ice-ocean interface is located.
It can be seen that when the top temperatures is negative enough and a cooling gradient is created through the ice, initially the lowest "corner" of the gradient where it tapers into -1.8C, the "naive interface" location, moves rapidly downward. After a while, this movement slows. I am proposing a new interpretation where during the slow movement part the "naive interface" does represent the ice-ocean interface, while during the fast movement part it represents the deepest point where the cooling signal has managed to penetrate at a given time.
I just don't see how such a sharp "corner" as seen at the end of December can be physically maintained if this was not the ice-ocean interface.
How does this fit with the Heat120 data? Not well at all. However, has the ice really thickened to 2.74cm? Looking at CS2SMOS from your post, it seems very plausible.
I therefore give up...

Edit: After staring at the larger animation for a while longer, my eyes insist that the thickness did not pass thermistor 110 before mid-Nov, looking both at the temp gradient and the Heat120 data. Maybe the initial cooling is expended towards reducing the core temp of the ice, and only later can bottom freezing begin? I am weak on the theory of all this.

[uniquorn]

T84 also has additional deployment data but unfortunately a photo is not available. Thermistor 26 at surface, ice thickness 1.14m, Snow depth 0.03 m, so bottom of the ice at thermistor84. Similar analysis below with a closer look at temperatures near to -1.8C.

It's possible that drilling the deployment hole affects the results, allowing faster local melt that refreezes over time.

Similar story for T84 so I'm starting to be convinced by the rotten ice idea. A cautious estimate of ice bottom thermistor 90 gives us 90-84 times 2cm, only 12cm thickening, optimistic would be the barest trace of ice bottom of 115(red dot).

[uniquorn]

Drift path for T84 for february. The TS file now contains surface(ice) temp, barometric pressure, tilt and compass heading. (no cairo to keep file size down)

T84 thermistor1 was 54cm above ice at deployment

[uniquorn]

Mosaic 2020T81, deployed near T78 and T84 but this time in a 26cm melt pond above 1.14m thick ice. Thermistor data hasn't been updated since feb10 but TS data is still coming in. Maybe the processing is slower.

The melt pond appears to have slowed the phase change beneath it as it is still not complete by feb10.
2020R21 maybe waiting for daylight

[uniquorn]

<>How does this fit with the Heat120 data? Not well at all. However, has the ice really thickened to 2.74cm? Looking at CS2SMOS from your post, it seems very plausible.<>

Based on the SIMB measured data showing thickening of only ~65cm and the T84 and T81 data above then 1.2m thickening for T78 is very likely an over optimistic interpretation but the data suggests that we should looking a little deeper than the temperature gradient change to flat line.
So what does a closer look tell us?

T78 Temperature almost flatlines at Thermistor122-126 over the last 4 days.
126-110=16 times 2cm is 32cm thickening

T78 Heat120 looks like it changes to water at Thermistor 139
139-110=29 times 2cm is 48cm thickening 

So the amended ice thickness estimate at T78 is 1.84-2.00m thick

None of these temperature/heat comparisons appear to agree with the 2 papers below.

Interfaces and Ocean Heat Flux derived from SIMBA_2015a and SIMBA_2015f data during N-ICE campaign in winter 2015.
https://www.seanoe.org/data/00485/59709/

The ice/ocean interface is estimated from temperature profiles alone since the winter sea-ice remains colder than the ocean. The ocean just below the ice is at or just above the freezing temperature (estimated from a near surface conductivity-temperature-depth (CTD) sensor see Koenig et al. [2016]). The method detects (1) the first sensor, downward of the snow/ice interface, with a temperature above the ocean freezing temperature and (2) the last sensor in the ice with a temperature below the mean ocean temperature by at least twice the ocean temperature standard deviation in that profile. The ice/ocean interface is then defined as half way between the last sensor in the ice and the first sensor in the ocean.

edit: added buoy data for this paper N-ICE2015 SIMBA quality controlled and derived data

Discrimination Algorithm and Procedure of Snow Depth and Sea Ice Thickness Determination Using Measurements of the Vertical Ice Temperature Profile by the Ice-Tethered Buoys
https://www.mdpi.com/1424-8220/18/12/4162

3.1.2. Interface between Ice and Ocean To identify the lower ice interface (ice-ocean interface), temperature profiles for the lower ice layer were obtained from some thermistors near the bottom of the sea ice. The seawater temperature was determined using the lower five thermistors, which generally had a negligible temperature gradient from the bottom of the sea ice.   The points where the temperature profile of the lower ice layer intersected the ocean temperature were regarded as the ice-ocean interface (Figure 3b). The ice-ocean interface determined by the method of seeking described above had a good accuracy in winter or sea ice growth period. This method became unreliable in summer, especially in ice melting period when the temperature gradient across the lower interface weakened. In summer, the temperature profile of sea ice became non linear with a C-shaped curve. Then the lower ice interface was determined from the obvious inflection point in the vertical C-shaped ice temperature profile (Figure 3d).  In winter,temperature profile of sea ice remained linear and temperature of the basal ice layer was colder than the upper ocean.  There will be a sharp inflection point occurring at the interface.  Thus, ice-ocean interface can be estimated from the vertical gradients of sea ice temperature profile.

[oren]

It's weird that all these buoys are in place but in essence there is no good reliable way to find out where the ice ends and the water starts. There must be a better way than just thermistors with ambiguous results. An interesting engineering challenge, which I have no clue how to solve.

[uniquorn]

Following on from T85 analysis here on dec4.

Here we have 4m thick ice (or thicker) where thermistors 1 to 10 are clearly above surface so all or nearly all of the remaining sensors are in ice.
It's possible that thermistors 210-240 are in the ocean but look at the temps. Flat line from thermistor 140 onwards. The thick ice is still cooling.

That cooling is still ongoing. There is an interesting small gradient change from mid january onwards between thermistors 150 and 210. Previously I would have identified that as thickening, now I think that is also phase change as the winter cold finally reaches the bottom of the ice.
Unfortunately T85 Heat is not working or turned off.

It's weird that all these buoys are in place but in essence there is no good reliable way to find out where the ice ends and the water starts. There must be a better way than just thermistors with ambiguous results. An interesting engineering challenge, which I have no clue how to solve.

Hopefully with all the equipment installed during the mosaic expedition someone is writing a paper on this. Otherwise we wait till 2023 for more detailed data....

T85 temps 1-239
T85 temps 130-239 close up.
Latest date static charts.

[Jim Hunt]

It's weird that all these buoys are in place but in essence there is no good reliable way to find out where the ice ends and the water starts. There must be a better way than just thermistors with ambiguous results.

The good ol' buoys from CRREL have "bottom sounders" as well as thermistors, but they are also less than perfect.

[oren]

Following on from T85 analysis

That cooling is still ongoing. There is an interesting small gradient change from mid january onwards between thermistors 150 and 210. Previously I would have identified that as thickening, now I think that is also phase change as the winter cold finally reaches the bottom of the ice.
Unfortunately T85 Heat is not working or turned off.

Very interesting behavior. A steep gradient fast moving up to thermistor 150, then a shallower gradient slowly progressing. Must be some physics there, but weird.

[uniquorn]

The same treatment of T58 from the beginning of the mosaic expedition on oct9 2019. Ice is 1.57m thick at deployment, snow is 18cm. Highly likely that the 9 thermistor wide peak is snow and the ice surface is roughly at thermistor33, making ice bottom at thermistor111. We see some agreement between Heat and ice temps at ~thermistor74 around nov11 2019 but the ice temperature gradient change doesn't reach thermistor111 until mar13 2020.
'Phase change' during june/july2020 reaches down to ~thermistor83 before failure. Not sure how to interpret thermistors beyond 125 during that period.

[uniquorn]

Lastly, a comparison with the SIMB3 441910 deployed in the Beaufort in 2.1m ice. Here the ice surface at deployment is roughly at dtc50 so the ice bottom is at roughly dtc155. The gradient change is just about reaching dtc155 today.
digital temperature chain

[oren]

So this is what normal freeze-up is like, as temps stay low throughout the winter. It's quite impressive how much damage is caused to the freezing process by spikes in air temperature, at least as evidenced by the very quick rise of the core temp of the ice, and its relatively slow drop afterward (shown in the previous animations upthread).

[uniquorn]

A rough comparison of mosaic T78 which was 1.52m at deployment in the CAB and simb3 442920 which was 2.1m at deployment in the Beaufort. Both are shown up thread. The simb3 profile has been shifted left to align the two ice surfaces.
edit:no temperatures are shown. This is just for comparing rate of change.
sep20-feb18

Thicker ice is said to insulate more and thicken more slowly but that doesn't really explain the slower cool down from surface. Perhaps it is older ice. But it also cools down faster later in the season. That could be due to colder air temperatures or perhaps that warm layer at 30m makes a difference. simb3 442920 is co-located with whoi itp121 providing us with the ocean temperatures directly below.(white is warmer than 1.8C) There aren't enough fully functional buoys to test that idea this year.

There is a mosaic CTD Obuoy, 2020O10 deployed near to T78 measuring temperatures from 10m to 100m depth. 100m is the warmest at around -1.3C.

[uniquorn]

An interesting question posed by Niall Dollard here required investigation of yesterday's polarview north of Greenland. On the same image is the mosaic floe2 identified using the mosaic 2020T85 location.

2021-03-03T12:30:15,86.144000,-28.754640

This can be located using the iwsviewer here (for a while)
This viewer doesn't work for me using Firefox but works fine using Chrome

Here is a close up from the jp2 image.

Floe2 could be the lighter blob, centre of the 3rd image.

[uniquorn]

Update on the remaining Mosaic Pbuoys in the Arctic Ocean, jan1-mar6.
The animation struggling a little with the P115 data.
Static image for low volume users.

[uniquorn]

T84 also has additional deployment data but unfortunately a photo is not available. Thermistor 26 at surface, ice thickness 1.14m, Snow depth 0.03 m, so bottom of the ice at thermistor84. Similar analysis below with a closer look at temperatures near to -1.8C.

It's possible that drilling the deployment hole affects the results, allowing faster local melt that refreezes over time.

Similar story for T84 so I'm starting to be convinced by the rotten ice idea. A cautious estimate of ice bottom thermistor 90 gives us 90-84 times 2cm, only 12cm thickening, optimistic would be the barest trace of ice bottom of 115(red dot).

Update on T84 thickening. A static look at the temperature chart tends to confirm an ice/ocean interface at thermistor96, possibly stretching to therm103.
96-84 times2cm is now 24cm thicker to 140cm

I've added a greyscale higher contrast heat120 to the standard T84 daily temperature chart to highlight the possible 'phase change' turning into slower thickening.  Interesting that the ice/ocean interface shows up more clearly on the Heat30 chart.

[uniquorn]

<>How does this fit with the Heat120 data? Not well at all. However, has the ice really thickened to 2.74cm? Looking at CS2SMOS from your post, it seems very plausible.<>

Based on the SIMB measured data showing thickening of only ~65cm and the T84 and T81 data above then 1.2m thickening for T78 is very likely an over optimistic interpretation but the data suggests that we should looking a little deeper than the temperature gradient change to flat line.
So what does a closer look tell us?

T78 Temperature almost flatlines at Thermistor122-126 over the last 4 days.
126-110=16 times 2cm is 32cm thickening

T78 Heat120 looks like it changes to water at Thermistor 139
139-110=29 times 2cm is 48cm thickening 

So the amended ice thickness estimate at T78 is 1.84-2.00m thick

update on T78. Both of those estimates look overoptimistic using a static analysis. Eyeballing the Heating mode images gives a rough ice/ocean interface at thermistor120. Recent temperatures show a change in gradient at therm118.
118-33=85 times 2cm is 1.70m, so  only 18cm thicker.

[SimonF92]

I ran the python script on T78 yesterday out of interest and it didnt bug entirely

I dont think its accurate but I was surprised it worked!

PS, taking from this graph, 1.2m thickening is what we estimate with our code too- I just dont think the 30cm starting thickness is correct

[uniquorn]

Interesting that the script picked up the temp anomaly on feb11-20. Clearly not melt and refreeze though.
edit: A shame that T78 data hasn't updated since feb26. Only T84 and T85 left.

20 more Tbuoy deployment and temperature images (click)

[OffTheGrid]

Very interesting variability in the thickness and freeboard measures at deployment uniquorn. Other than T78 which has more than 10% freeboard, suggesting to me that it's full of air voids from brine drainage or it has snow with a refrozen crust in that measurement (maybe more likely if the hole was full of slush, all the others show much less than 10%, and must be waterlogged, therefore not really a good measure of how much energy needed to melt them, more cold required to start thickening and seriously throwing a spanner in the works in regards to volume models.

[uniquorn]

https://acp.copernicus.org/preprints/acp-2021-117/acp-2021-117.pdf

Siberian fire smoke in the High-Arctic winter stratosphere observed during MOSAiC 2019-2020
Kevin Ohneiser1 , Albert Ansmann1 , Ronny Engelmann1 , Christoph Ritter2 , Alexandra Chudnovsky3 , Igor Veselovskii4 , Holger Baars1 , Henriette Gebauer1 , Hannes Griesche1 , Martin Radenz1 , Julian Hofer1 , Dietrich Althausen1 , Sandro Dahlke2 , and Marion Maturilli2 1Leibniz Institute for Tropospheric Research, Leipzig, Germany 2Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany 3Tel Aviv University, Porter School of Earth Sciences and Environment, Tel Aviv, Israel 4Prokhorov General Physics Institute of the Russian Academy of Sciences, Moscow, Russia Correspondence: K. Ohneiser (ohneiser@tropos.de)

Abstract.
During the one-year MOSAiC (Multidisciplinary drifting Observatory for the Study of Arctic Climate) expedition the German icebreaker Polarstern drifted through the Arctic Ocean ice from October 2019 to May 2020, mainly at latitudes between 85◦N and 88.5◦N. A multiwavelength polarization Raman lidar was operated aboard the research vessel and continuously monitored 5 aerosol and cloud layers up to 30 km height. The highlight of the lidar measurements was the detection of a persistent, 10 km deep wildfire smoke layer in the upper troposphere and lower stratosphere (UTLS) from about 7-8 km to 17-18 km height. The smoke layer was present throughout the winter half year until the polar vortex, the strongest of the last 40 years, collapsed in late April 2020. The smoke originated from major fire events, especially from extraordinarily intense and long-lasting Siberian fires in July and August 2019. In this article, we summarize the main findings of our seven-month smoke observations and char10 acterize the aerosol properties and decay of the stratospheric perturbation in terms of geometrical, optical, and microphysical properties. The UTLS aerosol optical thickness (AOT) at 532 nm ranged from 0.05-0.12 in October-November 2019 and was of the order of 0.03-0.06 during the central winter months (December-February). As an unambiguous sign of the dominance of smoke, the particle extinction-to-backscatter ratio (lidar ratio) at 355 nm was found to be much lower than the respective 532 nm lidar ratio. Mean values were 55 sr (355 nm) and 85 sr (532 nm). We further present a review of previous height15 resolved Arctic aerosol observations (remote sensing) in our study. For the first time, a coherent and representative view on the aerosol layering features in the Central Arctic from the surface up to 27 km height during the winter half year is presented. Finally, a potential impact of the wildfire smoke aerosol on the record-breaking ozone depletion over the Arctic in the spring of 2020 is discussed based on smoke, ozone, and polar stratospheric cloud observations.

[uniquorn]

Drift update on the Mosaic buoys on Floe2 showing the recent acceleration towards the Fram Strait. Click for movement.
Also a closer look at Mosaic CTD buoy 2020O10 currently around 83.8N -18.2. The processed file has stopped updating but the linked raw file is ok, providing temperature to 4dp and salinity to 5dp every 5 minutes at 10m-100m. Might be interesting as it enters the strait.

Location is roughly centre of the S1 image from polarview, a little to the right of the bright white dot. That doesn't appear on the other S1 looked at today.

[Jim Hunt]

Drift update on the Mosaic buoys on Floe2 showing the recent acceleration towards the Fram Strait.

My Arctic alter ego has been taking your name (and animation) in vain on Twitter:

https://twitter.com/GreatWhiteCon/status/1381548138218061825

I hope that's OK with you? Please advise ASAP if not!

[oren]

Well it certainly deserves wider circulation, I hope (and assume) uniquorn agrees.

[uniquorn]

you're welcome Jim.

Update on the same buoy 2020O10. Here looking at the difference between 10m and 20m temps and a shorter time scale to focus on the recent anomalies

[Jim Hunt]

you're welcome Jim.

Thanks. Whilst perusing my Twitter feed I also came across this:

"The MOSAiC Drift: Ice conditions from space and comparison with previous years"

We find that the MOSAiC drift was around 25 % faster than the climatological mean drift, as a consequence of large-scale low-pressure anomalies prevailing around the Barents-Kara-Laptev Sea region between January and March. In winter (October–April), satellite observations show that the sea-ice in the vicinity of the Central Observatory (CO) was rather thin compared to the previous years along the same trajectory. Unlike ice thickness, satellite-derived sea-ice concentration, lead frequency, and snow thickness during winter month were close to the long-term mean with little variability.

[RoxTheGeologist]

Uniquorn - that is a wonderful graphic. Wow.

On the buoy temperature profile - there isn't the same amplitude of noise in the 10 and 20m traces compared to the 20,50 and 100m. Is there an explanation besides lack of mixing?

[uniquorn]

@Jim   I thought Marcus Rex proposed a different reason last year but perhaps they've had time to look into it further. It surely was rapid drift.

@Rox  If you mean there is more mixing at 10-20m I think that is the reason for the more constant temp, though I might have it all the wrong way round. As to 'noise' I was seeing it as potentially useful data when viewed over bathy. There might not be any correlation though. Here looking at the difference between 50m and 10m temps with the drift path as a starting point. (click)
edit: animation updated below

There are a were 8 Obuoys deployed during the Mosaic expedition in this directory
https://data.meereisportal.de/download/buoys/
search for 2019O to find them. The proc file has friendly names but I think the data is only every 10m. The unprocessed one is every 5m with possibility of uncorrected data I suppose.

https://data.meereisportal.de/download/buoys/2019O1_300234068066320_proc.csv
https://data.meereisportal.de/download/buoys/2019O1_300234068066320.csv

[uniquorn]

I managed to add some bathy contour lines. No labels. (click)
So the 50m temps drop to close to 10m temps as the floe drifts along the basin edge. 10-50m temps are virtually the same as it crosses the 3300m deep bay. A lot to think about.
gmrt bathy of similar area and 36deg rotated worldview of clearish apr10 over blue marble bathy, https://go.nasa.gov/3wPDqyM