The chicken or egg problem

The elegant glacier-centred hypothesis, regarded as a boon to the field of glaciology, was called into question, however, by the results of efforts to test two of its corollaries. If the hypothesis were correct, the IRD increases in both series should occur at the same time as, or slightly lead, the ocean-climate response, and synchronicity of discharges from different glaciers would be unlikely owing to the vagaries of internal glacier dynamics. Instead, it was found in some high-resolution records from the North Atlantic that the onset of ocean surface coolings in both series actually preceded, by at least several hundred years, the IRD increases (e.g. Bond & Lotti, 1995; Bond et al., 1999). Even more troubling was evidence from records of petrological and geo-chemical tracers that the IRD in D-O cycles was discharged from different glaciers at the same time (Bond et al., 1999; Grousset et al, 2001; Figs 24.2 & 24.3).

Probably the most surprising discovery, though, was that each Heinrich event appeared to have been immediately preceded by an increase in IRD with a different composition to that of the IRD in the overlying Heinrich layers. High-resolution petrological and geochemical analyses focused on the precursory layers demonstrated that the IRD recorded the same simultaneous discharges of icebergs from different sources that were found in D-O cycles (Figs 24.2 & 24.3). The sequence of events was documented in records from the same cores, thereby removing any question about dating error. It appeared then that Heinrich events were immediately preceded by the onset of IRD build-up in a D-O cycle.

Those findings raised the spectre of the chicken or egg problem. The lead of ocean surface coolings and the synchronous IRD discharges from different glaciers during D-O events seemed best explained by the effects of a climate-ocean mechanism that operated upon glaciers flowing into the North Atlantic. The observation of D-O-like IRD discharge immediately preceding each Heinrich event implied that even Heinrich events were triggered by the same climate mechanism.

If true, then the glacier-centred hypothesis for iceberg discharges during D-O cycles and Heinrich events is incorrect and must be replaced by a climate-centred hypothesis in which climate triggers the glacial instabilities. It should be emphasized that the impact of the iceberg discharges on formation of NADW and the subsequent large and geographically widespread climate responses would be the same for either hypothesis. In the case of the climate-centred hypothesis, though, the iceberg discharges and reduction in NADW production operate as non-linear feedbacks that amplify the climate mechanism and transmit large, abrupt coolings into and well beyond the North Atlantic region.

Figure 24.2 Records of climate, IRD and petrological components in IRD: (a) oxgen isotope composition in GISP2 ice core at Summit Greenland; (b-f) climate and IRD records from DSDP site 609. H refers to Heinrich event. Note that in (b) cold peaks point down; in (c-f) cold peaks point up. Dashed lines mark start of Heinrich events as defined by increases in detrital carbonate. Precursory events can be seen in both haematite-stained grains and Icelandic glass. See text and Bond et al. (1999) for details.

Calendar age in kyrs

Figure 24.2 Records of climate, IRD and petrological components in IRD: (a) oxgen isotope composition in GISP2 ice core at Summit Greenland; (b-f) climate and IRD records from DSDP site 609. H refers to Heinrich event. Note that in (b) cold peaks point down; in (c-f) cold peaks point up. Dashed lines mark start of Heinrich events as defined by increases in detrital carbonate. Precursory events can be seen in both haematite-stained grains and Icelandic glass. See text and Bond et al. (1999) for details.

87Sr/86Sr 0.715 0.725 0.735

44 52 60 6S 76 S4 92

100-

cm 0

Precursor

Precursor

Figure 24.3 Sr and Nd isotope composition of IRD in core SU90-09 as fingerprints of IRD sources just before and during Heinrich events 1 and 2. Compositions of layers just below each Heinrich event are precursors and consistent with European sources, including Iceland. Compositions of Heinrich layers are consistent with a Laurentide source. See Grousset et al. (2001) for details.

Attempts to identify the climate triggering mechanism have thus far failed to produce a consensus. Broecker et al. (1990) proposed a salt oscillator in which a climate-driven salt build-up in northward flowing strands of the Gulf Stream strengthened NADW formation, warming the ocean and melting glacial ice. The flow of meltwater and icebergs into the ocean weakened NADW formation and cooled the ocean, leading to glacier regrowth and reduced freshwater discharge. That in turn strengthened NADW formation and initiated the next cycle. The difficulty with that concept is that the relationship of peak IRD (freshwater) discharge to ocean temperature is the opposite of what is observed. Alley et al. (2001) argued that a weak periodic forcing of the climate system combined with noise from ice-related events could, through stochastic resonance, produce the observed increased iceberg discharges and mode switches of D-O cycles. The model required specifying the period of the weak forcing, which they argued was the presumed 1500-yr pacing of the D-O events. What would cause a pacing with that period, and indeed whether the D-O pacing is really periodic are open questions. Hulbe et al. (2004) suggested that Heinrich events were preceded by growth of a large, unstable ice shelf in the Labrador Sea, which, in response to brief atmospheric warmings, broke up and triggered collapse of massive Laurentide ice in Hudson Strait. Evidence for abundant planktonic foraminifera in the Labrador Sea during Heinrich events (Andrews et al., 1998b), however, argues against a large ice shelf in the Labrador Sea, and there is no evidence that atmospheric warmings preceded Heinrich events.

The tropics also have been invoked as possible drivers of the millennial variability in the glacial North Atlantic. Clement et al. (2001) suggested that orbital variations coupled with El Niño oscillations lock the seasonal cycle in the tropical Pacific into a La Niña state that persists for several centuries, thereby cooling the Earth and triggering the cold phases of D-O cycles and associated iceberg discharges. Stott et al. (2002) and Visser et al. (2003) argue that oscillations of a 'super El Niño' caused large, recurring ocean surface salinity changes in the tropical Pacific that appear to correlate with the D-O variations in the North Atlantic. They suggested that the tropical salinity changes altered the greenhouse gas concentration of the atmosphere, thereby changing the Earth's temperature and triggering the D-O variations in the North Atlantic. The critical piece of evidence that would place the tropics in the driving role, a lead of the tropical signal relative to that of the D-O variations, however, is lacking (see also Broecker, 2003).

Some have even questioned the basic tenant of the climate-centred hypothesis, arguing that there is no simple relation between climate, glacial advance and discharge of icebergs from tidewater glaciers. Discharges of icebergs from a glacier in an outlet are known to occur simultaneously with decreases of iceberg discharge from an adjacent outlet, for example. Even if climate drives glacial advances and retreats, the timing of the response at the terminus may vary with the size of the glacier or ice sheet. Ice-rafted detritus may reflect changing debris loads in the basal layers of the ice and whether increases in IRD flux are due to ice-margin advances or retreats is not clear (Clark et al., 2000). In fiords, sea ice slows iceberg transit and IRD in basal layers of the ice may melt before icebergs reach the ocean (Reeh et al., 1999).

In an interesting new approach to the problem, however, Kaspi et al. (2004) suggested from modelling results that all of the observations pointing to a climate trigger are consistent with an elaborate version of the binge-purge model. They argue that it is not reductions in NADW formation that force the large climate responses. Rather, they found that the strong climate response could be due mainly to the albedo effects of sea ice that would build up in the northern North Atlantic in response to even small reductions of NADW formation. The sea-ice cover created a strong atmospheric cooling which then cooled the ocean surface enough that icebergs drifted farther from their sources and spread IRD far into the North Atlantic, thereby explaining the synchronous changes in IRD from different sources during D-O cycles without invoking climate forcing of glaciers and ice sheets.

If more than one ice sheet or glacier was undergoing a binge-purge type of oscillation, each with a different frequency, the Kaspi model predicted that after a few binge-purge cycles the sea ice-induced atmospheric coupling of the different glaciers caused them to become phase locked, thereby explaining the syn-chronicity of iceberg discharge observed in the sediment record. (Fig. 24.4a). Their model also predicted that smaller glaciers or ice sheets could have produced precursor events because under certain conditions, even though the phase locking causes the smaller and larger ice sheets to oscillate with the same frequency, the smaller ones will tend to reach the collapse phase first by several hundred years (Fig. 24.4b).

The Kaspi version of the binge-purge model is compelling because potentially it explains all of the observational evidence in hand surrounding D-O cycles and Heinrich events. The massive build-up of sea ice that is fundamental to the model, however, is difficult to test owing to the lack of reliable proxies for sea ice in

Figure 24.4 Results from the Kaspi et al. (2004) model suggesting that, owing to an atmospheric coupling of ice sheets as sea ice cools the North Atlantic, millennial oscillations eventually will become phase locked, producing (a) synchronous ice-sheet growth and decay. For certain model parameters (b) ice sheet growth and decay have the same periodicity, but smaller ice sheets lead larger ones, thereby producing precursory-like events seen in the North Atlantic marine records.

Figure 24.4 Results from the Kaspi et al. (2004) model suggesting that, owing to an atmospheric coupling of ice sheets as sea ice cools the North Atlantic, millennial oscillations eventually will become phase locked, producing (a) synchronous ice-sheet growth and decay. For certain model parameters (b) ice sheet growth and decay have the same periodicity, but smaller ice sheets lead larger ones, thereby producing precursory-like events seen in the North Atlantic marine records.

25 30 Calendar

35 age in

40 kyrs

441 294 147 0

c

\A'

ro e a. s

I -

1

Red noise model Raw record

25 30 Calendar

35 age in

40 kyrs

5 10 15 20 25 Cycle number

0.0 1.0 2.0 Frequency (cycles/kyr)

Figure 24.5 (a) The precise 1470-yr timing of D-O events from Rahmstorf (2003). Dashed lines mark out exact 1470-yr intervals. Markers with solid dots denote onset of D-O cycles as defined by warming above a threshold defined by amplitude and rate. (b) Deviation in years of onset of D-O cycles from the exact 1470-yr pacing. (c) Spectral analyses (multi-tapered (MTM) method) of haematite-stained grain cycles in Fig. 24.2. The heavy line in raw data is the average spectrum; light lines are the upper and lower 90% confidence limits. The heavy dashed line is the mean AR1 (red noise) spectrum; the light dashed line is the upper 90% confidence limit. The prominent peak centred on 1800 yr indicates a broad-band process (not periodic) that is different from red noise. The process could be periodic, however, and is distorted in the record by errors in the marine age model.

palaeoclimate analyses. Such proxies are being developed and a rigorous testing of the model should be possible in the near future.

0 0

Post a comment