edit: here are Sentinels of 11 and 23 Nov 15 with 128 shades of gray
Interesting question overall, the point of diminishing returns with more bit depth. On Sentinel, my understanding the satellite hardware and processing bins the data so that the end user is looking at pixels taking on one of 256 shades of gray (ie 2
8 or 8-bit), like Landsat 1-7.
Landsat-8 captures 12-bit or 4096 grays but is provided after bicubic interpolation to 16-bit where processing can take place advantageously but ultimately has to drop down to 8-bit, what computer monitors display.
The bit depth isn't the whole story, especially for cryosphere scenes, because the region of interest may only utilize half or less of the available grays. Combined with dropping bit depth (to reduce file size or processing time), this can result in a severely degraded image with posterized histogram and few options for enhancement.
The first animation has cropped out a central fjord scene at the original resolution of #888. Here the the advancing glacier front is displacing individual bergs down-fjord en bloc with no change in relative positions, with the exception of a stationary block in the southern embayment where ice hangs up on a known submerged nunatak.
The left half of the second animation shows the similar use of gray in the two November scenes but with the histogram average shifted towards white by 14 units on the later date. It's not clear what this means in terms of radar reflectivity but the tones could be better matched by adding a constant gray to the lighter image. If hundreds of such scenes were stacked and the sole purpose was to display motion, this would reduce distracting flicker/
The right half of this animation examines the scene to see where the extreme dark and light regions are coming from. Unsurprisingly, exposed rocks on the south contribute most of the dark tones. However some of the bergs are dark too and that could carry significant information about the ice being calved.
Thus an otherwise desirable contrast stretch of the middle tones would have an unfortunate side effect of pushing darker tones towards pure black. The rocks could be masked (protected) from the stretch but it still wouldn't be very beneficial to the mid tones because these have already lost resolution from being posterized to 128 grays = 2
7 or 7-bits. Note rock cracks do not move at all -- the two scenes are perfectly co-registered.
The extremely light regions (right half of 2nd frame of 2nd animation) is mostly contributed by smaller bergs and is physical detail that probably needs to be retained.
In summary, had the original image been kept at 8-bits, greater tonal detail could have been obtained by masking out the rocks and stretching mid-tones over a wider range of grays at the expense of compressing the low end. Visually this would result in a barely perceptible improvement over 7-bit.
However if the scientific objective is measurement of velocities, the improvements brought by 8-bit and contrast optimization are hugely important to the accuracy of automated feature-tracking. It will also be argued that the billions spent on getting the satellite instrument up to 8-bit should not be squandered to save pennies in file processing and storage. (The forum does have issues however with too-large files.)
The flip side is what would have been gained had Sentinel been built to 10-, 12- 16- or 32-bit sensitivity? Probably something at 10-bit, less at 12-bit and not very much after that for the vast majority of end user communities.
Landsat is a different story. The third image shows how little of the grayscale space is actually utilized originally in snowy scenes. However given the higher bit depth, the contrast can be improved without losing tonality by the time it is (inevitably) re-displayed as 8-bit. Contrast is what can be distinguished at a given resolution; resolution is how many sq meters are contributing to an individual pixel.
The final animation shows that the 7-bit definitely has some advantages over the 5-bit posted earlier in #1140. Here the contrast has been adjusted to bring out features in the final km of the calving front, at the expense of fjord image quality. However the latter is still adequate to show there has been no significant calving over the last two weeks, only the glacier advancing, pushing older icebergs down the fjord.