Science, Vol 304, Issue 5669, 400-402 , 16 April 2004
Andrew J. Weaver and Claude Hillaire-Marcel
A popular idea in the media, exemplified by the soon-to-be-released movie The Day After Tomorrow, is that human-induced global warming will cause another ice age. But where did this idea come from? Several recent magazine articles (1-3) report that abrupt climate change was prevalent in the recent geological history of Earth and that there was some early, albeit controversial, evidence from the last interglacial--thought to be slightly warmer than preindustrial times (4)--that abrupt climate change was the norm (5). Consequently, the articles postulate a sequence of events that goes something like this: If global warming were to boost the hydrological cycle, enhanced freshwater discharge into the North Atlantic would shut down the AMO (Atlantic Meridional Overturning), the North Atlantic component of global ocean overturning circulation. This would result in downstream cooling over Europe, leading to the slow growth of glaciers and the onset of the next ice age.
This view prevails in the popular press despite a relatively solid understanding of glacial inception and growth. What glacier formation and growth require is, of course, a change in seasonal incoming solar radiation (warmer winters and colder summers) associated with changes in Earth's axial tilt, its longitude of perihelion, and the precession of its elliptical orbit around the Sun. These small changes must then be amplified by feedback from reflected light associated with enhanced snow/ice cover, vegetation associated with the expansion of tundra, and greenhouse gases associated with the uptake (not release) of carbon dioxide and methane.
Several modeling studies provide outputs to support this progression. These studies show that with elevated levels of carbon dioxide, such as those that exist today, no permanent snow can exist over land in August (as temperatures are too warm), a necessary prerequisite for the growth of glaciers in the Northern Hemisphere [e.g., (6)]. These same models show that if the AMO were to be artificially shut down, there would be regions of substantial cooling in and around the North Atlantic. Berger and Loutre (7) specifically noted that "most CO2 scenarios led to an exceptionally long interglacial from 5,000 years before the present to 50,000 years from now . . . with the next glacial maximum in 100,000 years. Only for CO2 concentrations less than 220 ppmv was an early entrance into glaciation simulated." They further argued that the next glaciation would be unlikely to occur for another 50,000 years.
Although most paleoclimatologists would agree that the past is unlikely to provide true analogs of the future, past climate synopses are valuable for confronting the results of modeling experiments or for illustrating global warming. A reduction of the AMO due to a global warming-induced increase in freshwater supplies to the North Atlantic is often discussed in relation to a short event that occurred some 8,200 years ago (8.2 ka). During this event, one of the largest glacial lakes of the Laurentide Ice Sheet, Lake Ojibway, drained into the North Atlantic through Hudson Strait, quickly releasing enormous quantities of fresh water (8). However, to our knowledge, unequivocal evidence that this event resulted in a substantial reduction of the AMO has not yet been obtained. Notably, the Western Boundary UnderCurrent (WBUC)--which carries North Atlantic Deep Water masses (originating from the Norwegian and Greenland seas) along the continental slopes of Greenland and eastern North America--apparently remained unchanged during this episode [for example, (9)]. Because we cannot possibly foresee increases in freshwater inputs to the North Atlantic that could approach the magnitude of the Lake Ojibway discharge peak (the present Arctic river cumulative discharge rate is about two orders of magnitude lower), and because the effect of this event on the AMO is still unclear, further reference to the 8.2-ka event with respect to a reduction of the AMO in the near future seems irrelevant (also see letter by Broecker, page 388 of this issue).
Unquestionable evidence for a substantial reduction of AMO has been found only for intervals such as the Last Glacial Maximum (LGM) and some short, particularly cold, intervals of the last ice ages (such as those during Heinrich events). During these time periods, vast ice sheets occupied the Northern Hemisphere, providing a large freshwater source to the North Atlantic through either the dispersal of huge quantities of icebergs (Heinrich events) or the direct release of meltwater into the most critical sector associated with the AMO--the northeast Atlantic. On the other hand, the most critical site with respect to sensitivity to enhanced freshwater supplies from the Arctic has been, and would be, the Labrador Sea (10). Indeed, convection could stop there in response to global warming, as demonstrated by recent modeling experiments, apparently without any major effect on the overall rate of AMO (11). Worthy of mention is the fact that the strong east-west salinity gradient of the North Atlantic, with more saline waters eastward, seems a robust and permanent feature that was maintained even during the Last Glacial Maximum, when the rate of AMO was considerably reduced (12).
A clear picture of the North Atlantic under high freshwater supply rates arises from its recent history. High freshwater supplies may indeed impede convection in the Labrador Sea because of their routing along western North Atlantic margins, but this would result in an increased eastward branch of AMO. Further indication for such behavior is found in records of the Last Interglacial Interval. Relatively dilute surface water existed in the Labrador Sea, preventing intermediate water formation. However, a high-velocity WBUC existed throughout the whole period, indicating a high AMO along the "eastern route" (10).
The observed rate of global sea level rise during the 20th century is estimated to be in the range 1.0 to 2.2 mm/year (3). If one makes the clearly incorrect assumption that the entire maximum rate of observed sea level rise is a consequence of fresh water being added to the North Atlantic between 50º and 70ºN, then this equates to a rate of freshwater forcing of 0.022 Sv (2.2 x 104 m3 s-1). This rate in itself is certainly too small to cause a major shutdown of the AMO, although it may be large enough to cause cessation of convection in the Labrador Sea [for example, (6)].
It is certainly true that if the AMO were to become inactive, substantial short-term cooling would result in western Europe, especially during the winter. However, it is important to emphasize that not a single coupled model assessed by the 2001 IPCC Working Group I on Climate Change Science (4) predicted a collapse in the AMO during the 21st century. Even in those models where the AMO was found to weaken during the 21st century, there would still be warming over Europe due to the radiative forcing associated with increased levels of greenhouse gases.
Models that eventually lead to a collapse of the AMO under global warming conditions typically fall into two categories: (i) flux-adjusted coupled general circulation models, and (ii) intermediate-complexity models with zonally averaged ocean components. Both suites of models are known to be more sensitive to freshwater perturbations. In the first class of models, a small perturbation away from the present climate leads to large systematic errors in the salinity fields (as large flux adjustments are applied) that then build up to cause dramatic AMO transitions. In the second class of models, the convection and sinking of water masses are coupled (there is no horizontal structure). In contrast, newer non-flux-adjusted models find a more stable AMO under future conditions of climate change (11, 13, 14).
Even the recent observations of freshening in the North Atlantic (15) (a reduction of salinity due to the addition of freshwater) appear to be consistent with the projections of perhaps the most sophisticated non-flux-adjusted model (11). Ironically, this model suggests that such freshening is associated with an increased AMO (16). This same model proposes that it is only Labrador Sea Water formation that is susceptible to collapse in response to global warming.
In light of the paleoclimate record and our understanding of the contemporary climate system, it is safe to say that global warming will not lead to the onset of a new ice age. These same records suggest that it is highly unlikely that global warming will lead to a widespread collapse of the AMO--despite the appealing possibility raised in two recent studies (18, 19)--although it is possible that deep convection in the Labrador Sea will cease. Such an event would have much more minor consequences on the climate downstream over Europe.
A. J. Weaver is at the School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia V8W 3P6, Canada. E-mail: firstname.lastname@example.org C. Hillaire-Marcel is at GEOTOP, Université du Québec à Montréal, C.P. 8888, Montreal, Québec H3C 3P8, Canada. 10.1126/science.1096503 Include this information when citing this paper.