The ice sheets retreated 10,000 years ago during a peak in solar radiation, but this peak was no larger than previous ones. A modelling study suggests why the ice sheets were unusually vulnerable to melting at that time. See Letter p.190
I first encountered Ice Ages: Solving the Mystery1, the seminal book by John and Katherine Imbrie, as an undergraduate student, and it played no small part in drawing me in to graduate studies on ice-age climate dynamics. Imbrie père et fille describe the various strands of evidence establishing that Earth–Sun orbital variations are the main driver of glacial cycles: the recurring flow and ebb of ice sheets over the continents during an ice age. About 40 such glacial cycles have shaped our planet over the past 2.6 million years (the Quaternary period), representing the most dramatic example of climate variability in Earth's recent history.
But there is one nagging problem: as much as Earth's orbital wobbles seem to pace the advance and retreat of ice sheets, many aspects of ice-age climate dynamics remain a mystery. For one thing, those who model climate and ice sheets have not yet been able to simulate glacial cycles in a realistic way. Glacier advance into mid-latitudes requires severe cooling and increased snowfall compared with present-day conditions, to an extent that far exceeds the predicted response of the Earth system to 'cold' orbital configurations in climate models. It is even more difficult to get rid of continental ice sheets once they gain a foothold on the landscape. The modelling results reported by Abe-Ouchi et al.2 in this issue may provide a solution to these problems.
The crux of the challenge in modelling glacial cycles is that Earth's response to orbital forcing is entirely out of proportion. Changes in Earth's tilt axis and the eccentricity of its orbit around the Sun give rise to geographical and seasonal changes in incoming solar radiation. The global annual impact of these variations is negligible, but what really matters to the ice sheets is the amount of sunlight at high northern latitudes during the summer melt season. Peak radiation in this region varies by up to 100 watts per square metre because of orbital variations (Fig. 1a); this would certainly affect Arctic ice cover. However, integrated summer radiation, which is what counts in ice-sheet melting3, has deviated by less than 10% from present-day values over the most recent glacial cycle (Fig. 1b), and it is not obvious why this has elicited such a large shift in global climate.
In fact, a host of positive feedbacks — cooling influences associated with increases in snow and ice cover — conspire to amplify the orbital signal and send the world careering into glaciations. It is difficult to overcome these cooling influences, and so orbital changes alone are not enough to trigger deglaciation. The most recent glaciation persisted for roughly 100,000 years, and the ice sheets survived several periods of orbital warming before they finally destabilized and withdrew, starting about 20,000 years ago (Fig. 1c). At that time, summer solar radiation in the Northern Hemisphere increased, eventually peaking at about 6% above modern levels 10,000 years ago. But similar peaks occurred earlier during this period of glaciation, so what was different about this one?
Through asynchronous coupling of sophisticated climate and ice-sheet models, Abe-Ouchi and co-authors make a convincing case that the geometry of North America and the long response time of isostatic compensation — the change in height of Earth's surface in response to ice-sheet formation and retreat —are the main agents that transform 19,000-year (19-kyr), 23-kyr and 41-kyr orbital variations into a 100-kyr Earth-system response4. Ice sheets build up and flow southwards in both North America and Eurasia, taking many millennia to thicken and advance to their southern limits. Subglacial bedrock is depressed as underlying mantle material flows slowly outwards. At equilibrium, a 3,000-metre-thick ice sheet undergoes about 1,000 metres of subsidence5, but achieving equilibrium takes thousands of years. Similarly, land that was underneath the glacial ice sheets is still springing back.
Isostatic subsidence is one of the few negative feedbacks associated with glaciation: as an ice sheet slowly sinks, its surface lowers into a warmer climate, increasing the amount of melt and the area of the ice sheet exposed to melting. In Abe-Ouchi and colleagues' simulations, this process becomes most effective late in the glacial cycle, when the North American ice sheets are thick and have advanced far enough south; because this takes a long time, North America is set for a 100-kyr response. By contrast, the geography of the Eurasian ice sheets (which are thinner and less extensive, and occur in a warmer climate) gives them less inertia, and so they are more sensitive to 20- and 41-kyr orbital variations.
This idea is not new — earlier modelling studies5,6 also implicated isostatic rebound as one of the main processes underlying the 100-kyr glacial cycle. However, free-running simulations of the cycle have never before been achieved without invoking 'exotic mechanisms' — such as imposed ocean-circulation changes, dynamic ice-sheet destabilization or 'dusting' of the ice sheets — that force deglaciation at the 'right' time. One innovative technique that helps to capture the glacial cycle in Abe-Ouchi and colleagues' analysis is the use of multiple snapshots from climate models, which provide information about different ice-sheet sizes, carbon dioxide concentrations and orbital configurations. This is necessary because the computational time required to run a sophisticated climate model over tens of millennia is still prohibitively long.
However, some lingering mysteries remain, such as the effects of the oversimplified treatment (or absence) of ice sheet–ocean interactions, basal flow (ice-sheet sliding and subglacial sediment deformation) and ice-stream processes in the authors' simulations. Furthermore, ice-sheet melt rates are estimated only from air temperature, and are not based on energy-balance physics within the atmospheric model used by the authors. As climate and ice-sheet models become more sophisticated, we will see further refinement of these results.
Moreover, Abe-Ouchi and colleagues' findings do not explain the transition that took place 900,000 years ago, when the world moved from 41-kyr to 100-kyr glacial cycles. Isostatic time scales and North American geography did not change across this boundary, so another factor must have been at work. There are some layers yet to be explored in the mysteries of the ice age.