Help settle a debate on how concentration data is produced and whether or not we can measure concentration based on features smaller than the concentration data resolution. Here is their argument for why we
can't:As my second reply pointed out, my understanding is that concentration maps - like the DMI (Or was it NOAA - I believe they actually use the same data source but are slightly different models...) one I believe you posted - are derived from plugging in satellite telemetry to a model and based on that, assigning a concentration value based on the signal return strength in that data. That is resolved to a binary "On (covered with ice)/Off (no ice)" toggle.
The resolution of that data is pretty coarse. In the case of the NOAA data, the grid cell used is 25km by 25km - basically the size of a large city.
By necessity that means the sensitivity of the measurement is regional rather than local - like seen in the image I posted from Polarstern. My core argument is,that if you took the ice quality you see around Polarstern in that image, and covered a 25x25km2 block with it, the return *and* the model output would show that section of sea surface as covered with ice. It would not see the 10-15% of that surface which is actually open water between the mish-mash of small floes that caused the model to return a positive signal.
And my argument for why we
can is that the sensors used for concentration data do not resolve individual pixels to a binary on/off. The microwave imagining sensors functionally work much like normal camera imaging sensors work in that
each pixel gets a value on a scale based on the strength of the signal it receives, and this value can be processed into a percentage of ice concentration. Just like how a normal camera gets a value on a scale for each color channel based on the brightness, not just a binary on/off.
And what happens if you have features smaller than the resolution? Well
if you for example draw a white and black checkerboard where each square is 1 mm wide, and take a picture from such a distance so that the resolution of one pixel is 10mm, how does the checkerboard pattern look? It doesn't look black, or white, but gray, in other words a 50% brightness value. That's because the sensor naturally averages out the signal from all the colors found inside the pixel. And if we have more black than white squares, the pixel turns into a darker shade of gray. Meaning that we have successfully measured the concentration of black vs white even though the patterns of black and white is smaller than our pixel resolution.
In the same way, if you have tiny cracks or melt ponds on the ice, they affect the total signal brightness that the microwave sensor picks up even if those features are smaller than the pixel size, and thus it can successfully measure the concentration anyway.
NB: Remember that we are talking about concentration/area data here, not extent. I know that extent maps do have a binary ice/no ice value for each pixel. (But this is still derived from the same analog sensor tech though)
We've gone back and forth with the same reworded arguments and are making no headway so I ask you guys which of us are correct. The other guy can chime in as well if he wants, I didn't post his username just in case he wants to stay private.