bathymetry --> ice thinning spot?
Thousands of meters deep, tectonically quiet, nothing down there for a heat or turbulence source?
Whenever something is measured in physics, there's an obligation to estimate the error. In the Cryo2Smos netCDFs, there is indeed an included Geo2D file that generates the error map. Subtracting two close dates will lead to more combined error. If this error is too high relative to bona fide data differences, interpretation becomes problematic.
The error map seems to have a hot spot at the blue anomaly above. Here it might be worth clicking over to array view and looking at the patch of numbers for the nature of the oddity. It is more likely to originate from the Cryo than from the SMOS.
The ice thickness time series here only will only get to day 4 this afternoon. So the opportunities for limiting subtraction error will improve, from day2-day1, day3-day1 (above), day4-day1 etc and at some point will become relatively negligible in comparison to data differences.
As weeks and months of data become available, sufficiently separated date pairs can determine a ice thickening rate map across the lower Arctic. Note reducing the map to a single volume number loses all these important regional differences.
On the open water side, still 40% of the Arctic surface, the counterpart to the ice thickness netCDF is a high quality daily netCDF for 'foundational' sea surface temperature, the mixed layer at 10m. That archive goes back for many months but here the interest is how fast the water column temperature has been cooling during the freeze season to estimate when it will become cold enough to freeze over. Again a nuanced regional (map) basis is preferable to a single extent number.
There is
no single reason that explains the late freeze-up this year. Instead, the observed heat built up in the water column over time has to be apportioned among the various sources and balanced against mechanisms of heat loss. Inputs include global climatic trends and tele-connections, Arctic Amplification, creeping Atlantification, clear skies and summer insolation of open water, Siberian heat waves, mid-latitude moisture intrusions, cyclones, wave mixing, cooling buoyancy induced turbulence and so on. Outputs, mechanisms by which the water column loses heat, have been discussed earlier. Right now, the air is too warm and wind mixing to depth too high but those are acting on a heat state attained via earlier effects.
"The Arctic sea ice is growing very slowly this year. The ocean temperature in the entire mixed layer has to drop below the freezing point because the sea water is salty: when it cools down, the density increases and this leads to [buoyancy-driven] convection as it sinks. The lack of new ice growth is [quite anomalous]. SMOS shows an exceptionally small [advance] area covered with thin ice." Fig 3 Fig 4
https://twitter.com/seaice_deThe main difficulty in using Panoply to make time or difference series is in setting the range to accommodate the earliest and latest dates. If too broad, data will be squeezed into a narrow portion of the color range, losing resolution; too narrow, outlying but differing data will lumped into the end points.
Although Panoply offers over a hundred palettes, it may be best just to use the 255-grayscale where subtraction of pixels means what you want it to mean, subtraction of data values. (Images are spreadsheets.) Then colorize the final product later using one of the thousands of LUTs available in ImageJ, with custom adjustment wherever colors are not visually distinguishable.
The key is looking at product histograms of data value scatter and their asymmetry. The latter can be seen by superimposing a horizontal flip on a fast rocker.