refrozen water instead of compressed glacial ice?
The image below shows an approximate 1 km vertical cross-section of the calving front, minus the crevassed top and whatever is missing from very bottom. Click to see at full resolution. We don't know the exact scale of the photo since the logical embedded control, the height function of the calving front, wasn't measured then or now.
While this is not ideal for purposes of spectroscopy, we do know from sharp shadows that this was a sunny day in mid-May at 69ºN and so the angle and energy spectrum of incident light. We can only guess at the camera, its spectral sensitivity and contrast post-processing. However the video does quite well in reproducing known colors of rocks and open water much as the human eye would see them.
Colors in the vertical transect are quite complicated, there's much more going on than just blues and whites. In terms of transition sharpness in meters needed for onset of bubble collapse, we would want to consult with an optical scan of say the NEEM ice cores. These are done at sub-millimeter resolution right at the drill site, again as backscatter rather than transmission.
The originals should be archived online (since the scan is referenced in several peer-reviewed publications) but given the unique custom and culture of data-hoarding in glaciology, I'm not sure this is actually the case.
However in celebratory photographs, the ice appears bluish for the last 900 meters above bedrock. Blue is the natural color of thick pure ice for the same reason the sky and ocean are blue, Rayleigh scattering. H
2O is itself clear (transparent) -- there are no spectral lines in the visible (ie does not absorb there).
The ice in the image, in terms of light
transmission is very clear, like an Andy Lee Robinson melt cube. In fact, it's so clear that a large yellow volume appears to be embedded deep within this otherwise transparent ice.
Meltwater, either moulin or autochthonously produced from friction or pressure variation, could refreeze on the bottom under some phase diagram circumstances (such as suggested by P-Maker). However those conditions are narrow and not necessarily met. It seems a stretch to refreeze 800 m of bottom ice in that way in the short time this ice has been in rapid transit. At 15 km a year, not that long ago this calved ice was far past the ablation zone, sourced from 100 kyr of mundanely stratified ice sheet.
It is not rocket science to experimentally determined the temperature profile from top to bottom. There's been blue ice conveniently stranded on the terminal moraine all summer. According to the heat equation, the true temperature will be retained quite well for weeks after calving. This data will greatly reduce the theoretical sand castles that still need to be considered.
In terms of remelting/refreezing, the re-equilibrated temperatures of steam-drilled cores in and adjacent to the ice stream are on the cold side, -25ºC in mid- nterior. However the very lowest section would be warmer and we have no real grip on how geothermal heating or basal friction vary along the south branch. The latter would vary markedly by season and with velocity trends. However if the lower interior ice were anywhere near melting, Jakobshavn would be thinning vastly more rapidly than observed. Which it may very well do in the fairly near-term future.
The bizarre bottom refreeze or buoyancy upheaval structures seen upglacier from Petermann and Zachariae are largely restricted, for reasons unknown, to northern Greenland. The farthest south these features get is an isolated upwelling at Epiq for which there are hand-waving theories but no convincing explanation. (An explanation must not just explain an occurrence but also explain all the absences of occurrences.) Needless to say, I've looked but not found any in the Jakobshavn area.
I've examined all 22 years of ice penetrating radar data too for layering -- Jakobshavn is densely sampled almost every year, more than anywhere else in Greenland. Even though the west central Greenland ice sheet is entirely unperturbed and exactly as expected in its stratifications over 99% of its area, no structures can be seen in the ice stream itself, even though easily identified age layers reach the coast at Swiss Camp and Kanger.
Thus the ice feeding the south branch comes in well-stratified top to bottom but something then happens to it causing that to be entirely lost from the whole column, not just lower ice. These radar-reflective layers are associated with huge paleo volcanic events, not so much with tephra shards but rather deposited ionic polarity. Usually flow is modeled in 2D vertical sections that would retain stratifications as there is no mechanism for enhanced diffusivity in the upper layers.
The puzzling loss of stratification has never been addressed in a journal publication (which is equally puzzling). The actual transition to loss onset is not so clear because track density is sparse above the ablation zone. (It looks dense in the second image but radar specs were not the best in some years.) That's been attributed by on-board scientists to the acute boredom of flying hour after hour over unchanging regional stratification.
We know for certain that the blue ice did not form post-event because it lasted less than 11 minutes; the berg and specific features were in view essentially the whole time and constant in appearance (after discounting for water sloshing over).