When did the Greenland ice cap last melt?

The record preserved in cores through the thickest part of the Greenland ice cap goes back only to a little more than 120 thousand years ago, unlike in Antarctica where data are available for 800 ka and potentially further back still. One possible reason for this difference is that a great deal more snow falls on Greenland so the ice builds up more quickly than in Antarctica. Because ice flows under pressure this might imply that older ice on Greenland long flowed to the margins and either melted or calved off as icebergs. So, although it is certain that the Antarctic ice cap has not melted away, at least in the last million years or so, we cannot tell if Greenlandic glaciers did so over the same period of time. Knowing whether or not Greenland might have shed its carapace of ice is important, because if ever does in future the meltwater will add about 7 metres to global sea level: a nightmare scenario for coastal cities, low-lying islands and insurance companies.

Margin of the Greenland ice sheet (view from p...
Edge of the Greenland ice sheet with a large glacier flowing into a fjiord at the East Greenland coas  (Photo credit: Wikipedia)

One means of judging when Greenland was last free of ice, or at least substantially so, is based on more than a ice few metres thick being opaque to cosmic ‘rays’. Minerals, such as quartz, in rocks bared at the surface to ultra-high energy, cosmogenic neutrons accumulate short-lived isotopes of beryllium and aluminium – 10Be and 26Al with half-lives of 1.4 and 0.7 Ma. Once rocks are buried beneath ice or sediment, the two isotopes decay away and it is possible to estimate the duration of burial from the proportions of the remaining isotopes. After about 5 Ma the cosmogenic isotopes will have decreased to amounts that cannot be measured. Conversely, if the ice had melted away at any time in the past 5 Ma and then returned it should be possible to estimate the timing and duration of exposure of the surface to cosmic ‘rays’. Two groups of researchers have applied cosmogenic-isotope analysis to Greenland. One group (Schaefer, J.G. et al. 2016. Greenland was nearly ice-free for extended periods during the Pleistocene. Nature, v. 540, p. 252-255) focused on bedrock, currently buried beneath 3 km of ice, that drilling for the ice core finally penetrated. The other systematically analysed the cosmogenic isotope content of mineral grains at different depths in North Atlantic seafloor sediment cores, largely supplied from East Greenland since 7.5 Ma ago (Bierman, P.R. et al. 2016. A persistent and dynamic East Greenland Ice Sheet over the past 7.5 million years Nature, v. 540, p. 256-260). As their titles suggest, the two studies had conflicting results.

The glacigenic sediment grains contained no more than 1 atom of 10Be per gram compared with the 5000 to 6000 in grains deposited and exposed to cosmic rays along the shores of Greenland since the end of the last ice age. These results challenge the possibility of any significant deglaciation and exposure of bedrock in the source of seafloor sediment since the Pliocene.  The bedrock from the base of Greenland’s existing ice cap, however, contains up to 25 times more cosmogenic isotopes. The conclusion in that case is that there must have been a protracted, >280 ka, exposure of the rock surface in what is now the deepest ice cover at 1.1 Ma ago at most. Allowing for the likelihood of some persistent glacial cover in what would have been mountainous areas in an otherwise substantially deglaciated Greenland, the results are consistent with about 90% melting suggested by glaciological modelling.

Clearly, some head scratching is going to be needed to reconcile the two approaches. Ironically, the ocean-floor cores were cut directly offshore of the most likely places where patches of residual ice cap may have remained. Glaciers there would have transported rock debris that had remained masked from cosmic rays until shortly before calved icebergs or the glacial fronts melted and supplied sediment to the North Atlantic floor. If indeed the bulk of Greenland became ice free around a million years ago, under purely natural climatic fluctuations, the 2° C estimate for global warming by 2100 could well result in a 75% glacial melt and about 5-6 m rise in global sea level.

Read more about glaciation here and here.

New approach to the Milankovitch mystery

Melting pond on the ice sheet
Melting pond on the Greenland ice sheet (credit: Photo by Leif Taurer)

Milutin Milankovitch’s astronomical theory to account for glacial – interglacial cycles is based on 3 gravitational influences on the Earth that change the way it spins and orbits the Sun. Each is cyclic but with different periods: the angle of axial tilt every 41 ka; precession of its rotation axis on a 23 ka pacing; the change in shape of the orbit around the Sun over 100 ka. Each subtly affects the amount of solar energy, their influences combining to produce a seemingly complex, but predictable variation through time of solar heating for any point on the Earth’s surface. Milankovitch’s work was triumphantly confirmed when analysis of oxygen-isotope time series from sea-floor sediments revealed precisely these periods in the record of continental ice cover. Specifically, astronomical pacing of midsummer insolation at 65°N matches the real climatic pattern through time.

Yet the periods between glacial maxima have not stayed constant over the last 2 Ma or so (Figure showing Phanerozoic climate changes). About 0.8 to 1 Ma ago a sequence with roughly 41 ka spacing was replaced by another about every 100 ka, i.e. both overall climate periods matched one of the astronomical forcings. What is a puzzle is that the current periodicity seems to follow the very weakest influence in energy terms; that from orbital eccentricity. The energy shifts from changes in orbit shape are, in fact, far too weak to drive the accumulation and eventual melting of ice sheets on land. Climatologists have suggested a variety of processes that might be paced by eccentricity but which act to amplify is climatic ‘signal’. None have been especially convincing.

In an attempt to resolve the mystery Ayako Abe-Ouchi of the University of Tokyo and Japanese, US and Swiss colleagues linked a climate model driven by Milankovitch insolation and variations in CO2 recorded in an Antarctic ice core with a model of how land-ice forms and interacts with the underlying lithosphere (Abe-Ouchi, A. et  al. 2013. Insolation-driven 100,000-year glacial cycles and hysteresis of ice-sheet volume. Nature, v. 500, p. 190-193).

CLIMAP map of ice sheets, sea temperature chan...
Map of ice sheets, sea temperature changes, and changes in the outline of coastal regions during the last glacial maximum. (credit: Wikipedia)

Their key discovery is that the ice-sheets that repeatedly formed on the Canadian Shield and extended further south than Chicago had such a huge mass that they changed the shape of the land surface beneath them so much it had an effect on climate as a whole. The reason for this is that glacial loading forces the lithosphere down by displacing the more ductile asthenosphere sideways. But when melting begins rebound of the rock surface lags a long time behind the shrinking ice volume – well displayed today in Britain and Scandinavia by continued rise of the land to form raised beaches. In the case of the North American ice sheet, what had become an enormous ice bulge at glacial maxima developed into a huge basin up to 1 km deep as the ice began to melt. Simply by virtue of its low elevation this sub-continental basin would have warmed up more and more rapidly as the ice-surface fell because of this ‘isostatic’ lag.

Another feature to emerge from the model was the interaction between the 100 ka eccentricity ‘signal’ and that of precession at 23 ka. For long periods that kept summer temperature low enough for snow to pile up and become glacial ice, but on a roughly 100 ka time scale both acted together to increase summer temperatures at high northern latitudes. Melting that instantaneously removed some ice load each summer brought into play the sluggish isostatic  response that helped even more warming the following year. As well as convincingly accounting for the 100 ka mystery, the model explains the far more rapid deglaciations in that mode than in the preceding 41 ka cycles, which were almost symmetrical compared with the more recent slow accumulation of continental ice sheets over ~90 ka followed by almost complete melting in a mere 10 ka.

If true, the model seems to imply that before 800 ka the positions, thicknesses and extents of continental ice sheets were different from those in later times. Or perhaps it reflects a steady increase in the overall volume of ice being produced over northern North America, or that glacial erosion thinned the crust until changing isostatic influences could ‘trip’ sufficient additional warming.