Due to year-end paper shuffling needs and an impending in-law encampment, I will posting infrequently this month.
We have not done a great deal here with either voxel view or fourier transforms of daily Arctic sea ice data, in part because forum display formats and data resolution are so limited. Fancier displays such as the 5D hyperstacks of ImageJ used in microscopy (eg to display sections through moving stained cells) are thus off-limits since only 4-5 of our thousand registrants work within that freeware and there’s no practical way to distribute such files over the forum.
Athough the Arctic Ocean winter ice cover progression has almost stalled in recent days, arguably the main interest in winter re-freeze has shifted to growth in ice thickness. Some of that is ice growth along the periphery (marginal ice zone) and some is bottom growth (subsurface water freezing on underneath existing ice).
The image below looks at the rate of growth of peripheral ice using 61 days of SMOS data from U Bremen. The idea here is to track the history of single pixels as they progress from open water to the maximum 0.5 m sensed by SMOS. That amounts to layering up the available data, monte carlo selection of a pixel, cropping the layer set down to one for that single pixel, tiling it up, expanding the size without tweaking the Bremen color palette, repeating for more randomly selected pixels, tiling up the tiles, slicing that up, and sorting on >0.5 m pixel count to put them in order.
That’s done below for 16 pixels that lay outside the mask of pixels that always had thick ice over the Oct-Nov time frame. This could easily be extended to thousands of pixels (or all of them). In certain software, they could be towered semitransparently over the Oct 1st as is done in medical imaging. Here, the resulting thickness progressions give some indication of how fast freezing sea water thickens up to half a meter (at which point SMOS loses track of growth): about 20 days.
However there are a lot of oddities. These are largely attributable to ice movement. However as the U Bremen team publication states, SMOS has various insurmountable issues with accuracy and pixel footprint. Some of this could be addressed using 3x3 blocks of pixels to fine-tune the central pixel and by regional averaging.
The second image repeats a similar procedure for the same dates with AMSR2 sea ice concentration from U Hamburg.
The two overviews show the available areas for sampling, ie not always thicker than 0.5 m for SMOS and not always 100% concentration for AMSR2. Black dots indicate the pixel tower sites. The SMOS color palette does not desaturate to anything suitable for averaging nor does the geotiff so the netCDF would have to be used to generate quantitative grayscales.
SMOS and AMSR2 are shown side by side in the 61 day animation. This requires passing each series briefly into indexed color squeezing each palette into 127 colors before returning to RGB and gif-differencing to reduce file size to keep the 1x2 final product below the 255 colors this format allows.
The rate of bottom growth — which capable people here have calculated many times from a thermodynamic perspective — could possibly be directly monitored via Cryosat using this same technique (not shown). Cryosat uses a precision altimeter to determine the distance between itself and the ice surface. Given sea level and tides, that measures freeboard but not ice thickness unless the density (buoyancy) is known rather than assumed that of freshwater ice.
However snow cover, partial snow melt, incomplete brine exclusion in younger ice, sea water washed onto the floe and refrozen cause various measurement issues especially in the younger thinner ice that accounts for so much of the Arctic Ocean coverage today. It’s just difficult to measure ice thickness.
For the latest on Cryosat, AGU2016 features 42 presentations, eg
https://agu.confex.com/agu/fm16/meetingapp.cgi/Paper/163021 The final image shows the AMSR2 sea ice concentration averaged over the last 62 days. It is slightly out of gamut because of their palette decision but shows very clearly the difference between the Beaufort-Chukchi-ESS and the Barents Sea boundary. (This is computed by adjusting stack transparencies so each day makes an equal contribution to the final image.)
