Causes of rapid climate shifts

While D-O cycles, IRD and Heinrich events are related phenomena, a full accounting of their linkages has remained elusive. Thinking over the past 15 years has varied widely. One idea is the existence of an internal "salt oscillator" in the glacial climate. Another invokes catastrophic forcing events (halocline catastrophes). A third involves "binge-purge" behavior of the ice sheets as just discussed. These are very different paradigms, but all involve freshening of the northern North Atlantic, forcing a reduction or shutdown of NADW formation and hence widespread cooling of the North Atlantic region, followed by recovery. Other ideas focus on a direct climate forcing and then subsequent climate responses to surface freshening. The literature is vast and ever growing. We can only highlight some of the main arguments here.

The "salt oscillator" concept was first outlined by Rooth (1982) and elaborated on by Broecker et al. (1990). The basis of the theory is that there were periods when NADW production and the thermohaline circulation were relatively strong (albeit weaker than at present). During these periods there was enhanced discharge of meltwater and icebergs (the warm part of a D-O cycle). This resulted in freshening in the GIN seas. The surface freshening then reduced NADW production, rapidly causing widespread cooling of the atmosphere (the transition to the cold part of the D-O cycle). This reduced the meltwater and iceberg discharge. Ocean salinities then rebuilt to a critical point where NADW production recovered, causing warming (the transition from the cold to the warm part of a D-O cycle). The process then repeated itself. The salt oscillator as originally conceived stresses interactions with the Fennoscandian Ice Sheet. However, interactions with the Greenland Ice Sheet and Laurentide Ice Sheet could also be involved. The salt oscillator (associated with "on" and "off" modes of NADW production) finds some support in Lehman and Keigwin's (1992) study of climate variability during the last deglaciation (Section 10.6.2).

As an example of catastrophic forcing, Clark et al. (2001) propose that when the Laurentide Ice Sheet margin was between about 43 and 49° N, fluctuations in the location of the margin allowed rerouting of the main runoff from the Mississippi River drainage into the North Atlantic via the St. Lawrence River or Hudson Strait. The massive freshwater inputs to the northern North Atlantic capped the convection and caused widespread cooling. Another possible forcing is the discharge of water into the North Atlantic from the catastrophic drainage of proglacial lakes (Siegert, 2001). These are large lakes found ahead of the ice sheet margins, such as Lake Agassiz. Drainage of Lake Agassiz is a contender to explain the Younger Dryas cold event (Section 10.6.2).

The above ideas provide some explanation as to why the climate of the Holocene has been relatively stable. As land ice has been much more limited in the Holocene (Greenland and Antarctica excepted), it is harder to get the requisite massive freshwater discharges. NADW production under present conditions, while still somewhat variable, hence stays in the "on" mode. In Chapter 7, we examined the Great Salinity Anomaly of the late 1960s and early 1970s and noted an apparent association with reduced NADW production. The Great Salinity Anomaly could be viewed as a small-scale modern example of a NADW-freshwater link, although this event appears to have been associated with freshwater discharge through Fram Strait.

A problem with the ideas of drainage rererouting or the rapid drainage of proglacial lakes is that, while D-O events are quasi-periodic, catastrophes, by their nature, tend to have a large random component. Regarding the salt oscillator, there is a problem in timing, questioning cause and effect. The combination of the Heinrich events and the smaller intervening IRD events correlate well with warm-cold D-O cycles (Bond and Lotti, 1995). One might consider the freshwater input from icebergs associated with these IRD events (and attendant direct inputs of meltwater from the ice sheets) as drivers of D-O cycles. However, close inspection of the data indicates that the IRD events occurred during the cold phases of D-O events, not before them. This is inconsistent with the idea that freshwater pulses worked to decrease or shut down NADW production. Five of the six Heinrich events (Figure 10.4) occurred not prior to, but at the culmination of progressive cooling and were then followed by sharp warming to almost interglacial conditions (Bond and Lotti, 1995).

Furthermore, IRD from at least two widely separated sources is present at the same time in the same eastern North Atlantic cores examined by Bond and Lotti (1995). This indicates nearly synchronous rates of iceberg calving between these distant areas. One explanation for the synchronous iceberg discharge is that different ice sheets exhibited nearly identical "binge-purge" behavior. However, as argued by Bond and Lotti (1995), a more tenable explanation is that the synchronous calving represents an external climate forcing. In other words, freshwater input did not force the cooling phases of the D-O cycles, but rather followed the cooling. The freshwater input may have subsequently triggered further cooling. They acknowledge, however, that recurring binge-purge processes in another ice sheet may have discharged enough ice each time to alter NADW production, with the resulting ocean coolings triggering the iceberg discharges they identified.

Bond and Lotti (1995) found that each discharge of detrital carbonate-bearing ice from Hudson Strait (associated with the Heinrich events) seems to lag slightly behind the discharge from the other two sources. Farmer et al. (2003) also find that IRD fluxes into the Nordic Seas were nearly coincident with Heinrich events recorded elsewhere in the North Atlantic. It may be that the sudden coolings within the D-O cycles triggered the ice discharges in Hudson Strait at nearly the same time as discharges from the other two sources. An alternative view is that the trigger could have been sea level rises that seem to accompany each 2000-3000 year flood of icebergs to the ocean.

The Bond and Lotti (1995) study raises the obvious question of the possible source of an external climate forcing. Interestingly, modest IRD events have also recurred through the Holocene (Bond et al., 1997), when (excepting Greenland) there was much less continental ice in the Northern Hemisphere. A later effort, using highresolution Holocene data (Bond et al., 2001) found that these IRD events show a rough recurrence on two time scales, the approximate 1500 ± 500 year scale seen during glacial periods, and also around 400-500 years. There is a relationship between these cycles and inferred solar variability. Whether the apparent solar relationship seen in the Holocene records can be extended back into the last glacial cycle is not known. This inferred solar variability is not linked with Milankovitch cycles but to forcings such as those associated with sunspot cycles and altered fluxes of ultraviolet radiation (see Chapter 11 for a brief review). Another emerging view is a link between the D-O events and the tropics via a very slowly varying ENSO-like (EL-Niflo Southern Oscillation) component (G. Bond, personal communication).

Alley et al. (2001) and Rahmsdorf and Alley (2002) propose "stochastic resonance" on a 1500 year time scale. This involves interactions between a weak periodic climate forcing, climate "noise" and a non-linear amplifier. The basic idea is that one has an amplifier, namely the THC, which does not respond to signals below a certain threshold. However, if climate noise (e.g., random fluctuations in freshwater fluxes) combines correctly with a weak periodic forcing, stochastic resonance can occur, seen as a strong response of the THC. The idea finds support in results from a so-called model of intermediate complexity (Ganopolski and Rahmsdorf, 2001). This study showed that in contrast to "on" and "off" modes of the THC, such as associated with the salt oscillator, or catastrophic events such as the draining of proglacial lakes, a more subtle explanation for D-O events is geographical shifts in NADW production. And it seems that such geographical shifts in NADW production can be triggered fairly easily by climate noise. When this model is driven by random noise of realistic amplitude, combined with a weak periodic (1500-year) climate signal, D-O events emerge which are very similar to those seen in the Greenland ice core records. The origin of the weak periodic 1500-year climate forcing has not yet been clearly identified, but the solar variability link postulated by Bond et al. (2001) must be considered as a contender.

A final comment regards inter-hemispheric correlations. The major aspects of the glacial/interglacial cycles as seen in the Greenland ice cores correspond well to those in Antarctic records, pointing to global-scale signals. However, inter-hemispheric relationships between millennial-scale D-O events are not so clear. Through communication via the global ocean circulation, one might expect time lags in their expression in Arctic and Antarctic ice cores. Such phase relationships have proven difficult to demonstrate, however, in part because of difficulties in dating the records precisely.

0 0

Post a comment