Abrupt changes in the global carbon cycle during the last glacial period

Global climate and the concentration of atmospheric carbon dioxide (CO2) are correlated over recent glacial cycles. The combination of processes responsible for a rise in atmospheric CO2 at the last glacial termination (23,000 to 9,000years ago), however, remains uncertain. Establishing the timing and rate of CO2 changes in the past provides critical insight into the mechanisms that influence the carbon cycle and helps put present and future anthropogenic emissions in context. Here we present CO2 and methane (CH4) records of the last deglaciation from a new high-accumulation West Antarctic ice core with unprecedented temporal resolution and precise chronology. We show that although low-frequency CO2 variations parallel changes in Antarctic temperature, abrupt CO2 changes occur that have a clear relationship with abrupt climate changes in the Northern Hemisphere. A significant proportion of the direct radiative forcing associated with the rise in atmospheric CO2 occurred in three sudden steps, each of 10 to 15 parts per million. Every step took place in less than two centuries and was followed by no notable change in atmospheric CO2 for about 1,000 to 1,500years. Slow, millennial-scale ventilation of Southern Ocean CO2-rich, deep-ocean water masses is thought to have been fundamental to the rise in atmospheric CO2 associated with the glacial termination, given the strong covariance of CO2 levels and Antarctic temperatures. Our data establish a contribution from an abrupt, centennial-scale mode of CO2 variability that is not directly related to Antarctic temperature. We suggest that processes operating on centennial timescales, probably involving the Atlantic meridional overturning circulation, seem to be influencing global carbon-cycle dynamics and are at present not widely considered in Earth system models.

Understanding the causes of past atmospheric methane (CH4) variability is important for characterizing the relationship between CH4, global climate and terrestrial biogeochemical cycling. Ice core records of atmospheric CH4 contain rapid variations linked to abrupt climate changes of the last glacial period known as Dansgaard-Oeschger (DO) events and Heinrich events (HE)1,2. The drivers of these CH4 variations remain unknown but can be constrained with ice core measurements of the stable isotopic composition of atmospheric CH4, which is sensitive to the strength of different isotopically distinguishable emission categories (microbial, pyrogenic and geologic)3-5. Here we present multi-decadal-scale measurements of δ13C-CH4 and δD-CH4 from the WAIS Divide and Talos Dome ice cores and identify abrupt 1‰ enrichments in δ13C-CH4 synchronous with HE CH4 pulses and 0.5‰ δ13C-CH4 enrichments synchronous with DO CH4 increases. δD-CH4 varied little across the abrupt CH4 changes. Using box models to interpret these isotopic shifts6 and assuming a constant δ13C-CH4 of microbial emissions, we propose that abrupt shifts in tropical rainfall associated with HEs and DO events enhanced 13C-enriched pyrogenic CH4 emissions, and by extension global wildfire extent, by 90-150%. Carbon cycle box modelling experiments7 suggest that the resulting released terrestrial carbon could have caused fromone-third to all of the abrupt CO2 increases associated with HEs. These findings suggest that fire regimes and the terrestrial carbon cycle varied contemporaneously and substantially with past abrupt climate changes of the last glacial period.

The role of Southern Ocean winds, upwelling and CO2 in glacial abrupt climate change

<p>Atmospheric carbon dioxide (CO<sub>2</sub>) concentrations during the last glacial period (70,000 – 23,000 years ago) fluctuated on millennial timescales closely following variations in Antarctic temperature. This close coupling has suggested that the sources and sinks driving millennial scale CO<sub>2</sub> changes are dominated by processes in the Southern Ocean. However, recent work revealed centennial-scale increases in CO<sub>2</sub> during abrupt climate events of the last deglaciation which may represent a second mechanism of carbon cycle variability. </p><p>Here we analyze a high resolution CO<sub>2</sub> record from the last glacial period from the West Antarctic Ice Sheet (WAIS Divide) that precisely defines the timing of CO<sub>2</sub> changes with respect to Antarctic ice core proxies for temperature, dust delivery, and sea-ice extent down to the centennial-timescale. Although CO<sub>2</sub> closely tracks all these proxies over millennia, peak CO<sub>2</sub> levels most often lag behind all proxies by a few hundred years. This decoupling from Antarctic climate variability is most prominent during the onset of DO interstadial events when CO<sub>2</sub>, CH<sub>4</sub> and Greenland temperature all increase simultaneously. Regression analysis suggests that the CO<sub>2</sub> variations can be explained by a combination of two mechanisms: one operating on the time scale of Antarctic climate variability, and a second responding on the Dansgaard-Oeschger time scale.  </p><p>Recent δ<sup>13</sup>C-CO<sub>2</sub> data from the last glacial period support our finding that CO<sub>2</sub> variability is the sum of multiple mechanisms.  The Antarctic climate variability is likely associated with the release of respired organic carbon from the deep ocean.  Superimposed on these oscillations are two types of centennial-scale changes: CO<sub>2</sub> increases and δ<sup>13</sup>C-CO<sub>2</sub> minima in the middle of Heinrich stadials and ii) CO<sub>2</sub> increases and small changes in δ<sup>13</sup>C-CO<sub>2 </sub>that at the onset of DO interstadial event.</p><p>To provide a comprehensive and quantitative constraint on the mechanisms of CO<sub>2</sub> variability during the last glacial period, we run a large suite of transient box model experiments (n = 500) forced with varying combinations of forcings based on proxy time-series (e.g. AABW formation, NADW formation, ocean temperature, dust delivery, and sea-ice extent).  Using data constraints from the ice core records of CO<sub>2</sub>, δ<sup>13</sup>C-CO<sub>2</sub> and mean ocean temperature, we arrive at an ensemble of scenarios that can explain a large amount of the centennial and millennial-scale variability observed in the ice core record. Parsing this into a series of factorial experiments we find that Southern Hemisphere processes can explain 80% of the observed variability and Northern Hemisphere processes account for the remaining 20%.  A further breakdown on the level of individual mechanisms is marred by the high degree of correlation between carbon cycle forcings likely operating in the Southern Hemisphere.  None-the-less, our results highlight how multiple mechanisms operating over multiple timescales may have interacted during the last glacial period to drive changes in atmospheric CO<sub>2</sub>. </p>

During the last glacial period, the North Atlantic region experienced a series of Dansgaard–Oeschger cycles in which climate abruptly alternated between warm and cold periods. Corresponding variations in Antarctic surface temperature were out of phase with their Northern Hemisphere counterparts. The temperature relationship between the hemispheres is commonly attributed to an interhemispheric redistribution of heat by the ocean overturning circulation. Changes in ocean heat transport should be accompanied by changes in atmospheric circulation to satisfy global energy budget constraints. Although changes in tropical atmospheric circulation linked to abrupt events in the Northern Hemisphere are well documented, evidence for predicted changes in the Southern Hemisphere’s atmospheric circulation during Dansgaard–Oeschger cycles is lacking. Here we use a high-resolution deuterium-excess record from West Antarctica to show that the latitude of the mean moisture source for Antarctic precipitation changed in phase with abrupt shifts in Northern Hemisphere climate, and significantly before Antarctic temperature change. This provides direct evidence that Southern Hemisphere mid-latitude storm tracks shifted within decades of abrupt changes in the North Atlantic, in parallel with meridional migrations of the intertropical convergence zone. We conclude that both oceanic and atmospheric processes, operating on different timescales, link the hemispheres during abrupt climate change. Abrupt glacial climate changes were slowly communicated between hemispheres by oceanic heat transport. Ice core data point to more rapid atmospheric teleconnections linking the North Atlantic, tropics, and southern storm track.

<p>During the last glacial period, abrupt climate events known as Dansgaard-Oeschger (DO) and Heinrich events have been observed in various types of Northern Hemispheric (NH) paleoclimate archives. It has been speculated that volcanism may play a role in the abrupt climate variability of the last glacial period, for example as a trigger of abrupt changes. The investigation of a possible link between abrupt climate events and volcanic eruptions has been hampered by the lack of a global volcanic eruption record from the last glacial period. A recent identification of 80 major bipolar volcanic eruptions in Greenland and Antarctic ice core records within the interval 12-60 ka BP now enables us to investigate this link.</p><p>Using high-resolution ice-core records of climate (δ<sup>18</sup>O), atmospheric circulation changes (calcium) and volcanic eruptions (sulfate and other volcanic proxies) we investigate the timing of abrupt climate events and large volcanic eruptions at decadal resolution. We consider possible links between major volcanic eruptions and DO onsets (NH warming), DO terminations (NH cooling), and Heinrich stadials (strong NH cooling). Heinrich stadials are cold Greenland stadial periods during which Heinrich events occurred; large Hudson Strait iceberg discharge events that are characterized by deposition of significant amounts of ice rafted debris in North Atlantic marine sediments.</p><p>Significant links of volcanic and climatic events are tested in a statistical framework under the null hypothesis of random and memoryless volcanic activity. Our analysis shows that while certainly not all abrupt climate change of the last glacial period is associated with volcanism, we find that volcanism may have induced some abrupt Greenland warming events and perhaps several of the extreme North Atlantic cold Heinrich stadials; no significant link is found between volcanism and DO terminations. We speculate that volcanic cooling can drive such transitions when the coupled system of Atlantic Ocean circulation and North Atlantic sea ice is close to a tipping point.</p>

Reconstructions of ancient atmospheric carbon dioxide (CO2) variations help us better understand how the global carbon cycle and climate are linked. We compared CO2 variations on millennial time scales between 20,000 and 90,000 years ago with an Antarctic temperature proxy and records of abrupt climate change in the Northern Hemisphere. CO2 concentration and Antarctic temperature were positively correlated over millennial-scale climate cycles, implying a strong connection to Southern Ocean processes. Evidence from marine sediment proxies indicates that CO2 concentration rose most rapidly when North Atlantic Deep Water shoaled and stratification in the Southern Ocean was reduced. These increases in CO2 concentration occurred during stadial (cold) periods in the Northern Hemisphere, several thousand years before abrupt warming events in Greenland.

[1] Past variations in the concentration of the greenhouse gas CO2 are thought to have played a major role in controlling Earth's climate on pre-Quaternary and Quaternary timescales. To identify the contribution of CO2 to past climatic change requires accurate quantification of its content in the ancient atmosphere, and a number of proxies have been developed for this purpose (for a review see Royer et al. [2001a]). For the Late Quaternary, there is the unique opportunity to measure directly the composition of fossil air samples trapped in bubbles preserved in the polar ice sheets. Results from Antarctic ice cores reveal that the glacial-interglacial changes characterizing Quaternary climate were accompanied by variations in the atmospheric concentration of CO2 [Petit et al., 1999]. Although detection of phase relations between isotope-derived temperature estimates and trace gas concentration values is hampered by the difference in age of ice and air from the same ice sample, it is believed that CO2 lags glacial-interglacial temperature change and has acted as an amplifier of orbitally forced changes in temperature [Petit et al., 1999; Shackleton, 2000; Mudelsee, 2001]. [2] Detailed ice core data from the last glacial period indicate temperature-CO2 covariation on shorter, millennial, timescales [Stauffer et al., 1998]. However, such detailed studies are rarely feasible because the CO2 concentrations retrieved from the ice cores are inherently smoothed by diffusion during air enclosure, an effect controlled by ice accumulation rate. It is therefore not a surprise that ice cores drilled at low accumulation sites show little evidence of abrupt CO2 change during the last deglaciation [Monnin et al., 2001] and the Holocene [Indermühle et al., 1999]. A temporally detailed understanding of global carbon cycle dynamics over shorter, century to millennial timescales, might be achieved with sufficiently sensitive and reliable CO2 proxies. [3] Of the various CO2 proxies available, the carbon isotope composition of phytoplankton and the boron isotope composition of calcium carbonate shells of foraminfera have temporal resolutions suitable for detecting CO2 changes on 103–104 years [Royer et al., 2001a], but only the stomatal method has the capability to record century scale CO2 dynamics. Moreover, it is also arguably the most direct, with cellular differentiation taking place as the leaves sense CO2 in the environment. The CO2 signal was encoded in fossil leaf cuticles because as atmospheric CO2 concentrations increase, leaves of terrestrial vascular plants develop fewer stomatal pores for regulating gaseous exchange with the atmosphere [Woodward, 1987]. However, the recovery of quantitative information from this unique paleobotanical CO2 archive is still in its infancy and has yet to gain widespread critical acceptance. To achieve this goal, its performance deserves to be critically evaluated against empirical data drawn from field studies, laboratory experiments, ice cores and the plant fossil record. [4] An important criterion for any CO2 proxy is that the fidelity of the CO2 signal remains undiminished by changes in other features of the environment. Paleo-CO2 estimates from leaf fossils which utilize measurements of stomatal index (SI, defined as the percentage of leaf epidermal cells that are stomata) are likely to be secure. This is because observations on leaves of tree and shrub populations growing across natural climatic gradients, and controlled environment experiments, show SI is relatively insensitive to soil water supply, irradiance, atmospheric moisture and temperature [Beerling, 1999; Royer, 2001]. The stability of SI arises because it is a proportional quantity independent of leaf expansion; CO2 actually changes the number of epidermal cells that develop into stomatal pores [Woodward, 1987]. In fact, altering the growth atmospheric CO2 concentration is one of the few means of inducing a marked change in the SI of leaves within a given species. Stomatal density, in contrast, simply expresses the number of stomata per unit area of leaf, and is influenced by the climate-controlled degree of leaf expansion (see Figure 1). [5] If measurements of fossil leaf SI securely reflect past CO2 changes, how do the stomatal-based paleo-CO2 reconstructions compare with benchmark CO2 records reported from ice core studies? A key demonstration of the capacity of the technique to capture and retrieve the past CO2 history of the atmosphere derives from temporally detailed measurements on radiocarbon-dated Holocene fossils, made exclusively with leaves of the dwarf willow (Salix herbacea) [Rundgren and Beerling, 1999]. The resulting reconstruction shows a pattern of CO2 accumulation in the atmospheric reservoir over 7000 yrs that closely tracks the Taylor Dome Antarctic ice core CO2 record, although the stomatal record shows a greater amplitude [Indermühle et al., 1999] (see Figure 1). Both records also reveal minor fluctuations during the so-called Medieval Warm Period and the Little Ice Age climatic oscillations, as well as the more recent exponential rise in CO2 due to fossil fuel burning and deforestation (see Figure 1). [6] For the Holocene, the CO2 patterns compare favourably, but terrestrial high-resolution CO2 reconstructions [Beerling et al., 1995; McElwain et al., 2002; Wagner et al., 2002; Rundgren and Björck, 2003] through millennial-scale climate variations of the last deglaciation have revealed interesting new information that is absent from ice cores. They show that the steady rise in CO2 during deglaciation was apparently interrupted by an abrupt fall in CO2 coinciding with the beginning of the Younger Dryas stadial (see Figure 1). In addition, a temporary CO2 decline is registered at the time of the Preboreal oscillation, an early Holocene cooling event. Measurements on the Dome C Antarctic ice core indicate a more gradual deglacial CO2 increase [Monnin et al., 2001] without the relatively high-amplitude changes suggested by stomatal data. The trends in the two sets of data are, however, almost identical (see Figure 1). This discrepancy can partly be accounted for by the smoothing of ice core CO2 records caused by diffusion. The age distribution of enclosed air at the Dome C site, for example, lies between 200 and 550 yrs [Monnin et al., 2001], which is an order of magnitude larger than in a single sample of fossil leaves (30–40 yrs). If the fossil leaves are accurately portraying Lateglacial global carbon cycle dynamics, they suggest a higher sensitivity to climate than previously realized (see Figure 1) and demand a coherent explanation from the modelling community. [7] Other early Holocene and Lateglacial records [Wagner et al., 1999; McElwain et al., 2002] have reproduced similar CO2 patterns, indicating self consistency in the approach both between species and sites. Some stomatal-based records however have reconstructed atmospheric CO2 values higher (maximum 40 ppmv) than those obtained in ice core studies [Wagner et al., 1999, McElwain et al., 2002; Wagner et al., 2002]. The overestimations in these studies may relate to the use of fossil leaf assemblages containing a mixture of closely related species. Leaf SI responds to CO2 in a strongly species-specific manner [Royer et al., 2001a]; even closely related species capable of hybridising with each other differ in their CO2 responsiveness [Rundgren and Björck, 2003]. Additionally, studies involving fossil Betula leaves may be compromised by developing calibration functions with trees of very restricted genotypic diversity [Birks et al., 1999]. [8] However, how important are such mismatches relative to the performance of the other three leading paleo-CO2 proxies? And should they be allowed undermine stomatal-derived paleo-CO2 estimates? The paleosol CO2 barometer has typical errors of ±300–500 ppmv and is unsuitable for tracking ice core CO2 variations because of the time required for the formation of soil carbonates (103–104 yrs) [Cerling, 1992]. Atmospheric CO2 records derived from the carbon isotope composition of alkenones track the CO2 variations of the last glacial-interglacial seen in ice cores, but significant (c. 20 ppmv or more) mismatches are evident [Jasper et al., 1994]. As recently shown by Pagani et al. [2002], accurate CO2 reconstruction by this approach requires parallel estimates of marine phosphate concentration. CO2 records based on the boron isotopic method are likely to be influenced by changes in upwelling [Palmer and Pearson, 2003], and its accuracy is called into question because it suggests that whole-ocean pH was stable over the last glacial-interglacial cycle [Anderson and Archer, 2002]. [9] An implicit assumption for paleo-CO2 proxies involving the biota is that the growth of an organism in responses to its environment is the same on ecological and evolutionary timescales. For vascular land plants such as S. herbacea it is probably a quite reasonable assumption because atmospheric CO2 values reconstructed with fossil leaves dating back to the last interglacial (the Eemian, 130–115 kyrs BP) [Rundgren and Bennike, 2002] are directly comparable with CO2 data reported from the Vostok Antarctic ice core [Petit et al., 1999]. On a multimillion year timescale, the relationship between SI and CO2 exhibited by modern Ginkgo trees has been validated by analyses of changes in the SI of early Paleogene fossil Ginkgo leaves and independent CO2 estimates from paleosols spanning some 3 Myr [Beerling and Royer, 2002]. Strong arguments exist therefore to expect a similarity between the phenotypic and genotypic responses of plants to past histories of CO2; arguments which are critical to establishing the credibility of stomatal-based paleo-CO2 estimates. [10] The elegance of the stomatal paleo-CO2 proxy is that it rests on a simple inverse correlation between CO2 and stomatal formation which is underpinned by a gene involved in the signal transduction pathway controlling stomatal numbers at elevated CO2 [Gray et al., 2000]. Encoded into the leaf fossil record therefore is a rich archive detailing how the CO2 content of the ancient atmosphere has varied. Over the last decade, surprisingly rapid progress has been made in recognizing and recovering this valuable source of CO2 information. It has, for example, provided clues to the causes of mass extinction events [McElwain et al., 1999] and new constraints on radiative forcing by CO2 in the Tertiary [Royer et al., 2001b]. With the capacity to record millennial and century-scale CO2 changes, the stomatal approach to paleo-CO2 estimation offers the potential to identify new undetected rapid reorganizations of the global carbon cycle, as already suggested for the last deglaciation (see Figure 1).

The evolution of the northern hemispheric climate during the last glacial period was shaped by two prominent signals of glacial climate variability known as Dansgaard-Oeschger cycles and Heinrich events. Dansgaard- Oeschger cycles are characterised by a period of rapid, decadal warming of up to 14°C in the high northern latitudes, followed by a more gradual cooling spanning several centuries. Temperature reconstructions from ice cores indicate a dominant recurrence interval of ∼1,500 years for Dansgaard-Oeschger cycles. Heinrich events are quasi-episodic iceberg discharge events from the Laurentide ice sheet into the North Atlantic. The paleo record places most Heinrich events into the cold phase of the millennial-scale Dansgaard-Oeschger cycles. However, not every Dansgaard-Oeschger cycle is accompanied by a Heinrich event, revealing a complex interplay between the two prominent modes of glacial variability that remains poorly understood to this day. Here, we present simulations with a coupled ice sheet-solid earth model to introduce a new mechanism that explains the synchronicity between Heinrich events and the cooling phase of the Dansgaard-Oeschger cycles. Unlike earlier studies, our proposed mechanism does not require a trigger mechanism during the cooling phase. Instead, the atmospheric warming signal associated with the interstadial phase of the Dansgaard-Oeschger cycle causes enhanced ice stream thickening such that a critical ice thickenss and temperature threshold is reached faster, triggering the Heinrich event during the early cooling phase of the Dansgaard-Oeschger cycle. An advantage of our mechanism in comparison to previous theories is that it is not restricted to marine-terminating ice streams, but applies equally to land-terminating ice streams that only become marine-terminating during the actual Heinrich event. Our simulations demonstrate that this mechanism is able to reproduce the Heinrich event characteristics as known from the paleo record under a wide range of forcing scenarios and provides a simple explanation for the observational evidence of synchronous Heinrich events from different ice streams within the Laurentide ice sheet.

The mid-latitude westerly winds of the Southern Hemisphere play a central role in the global climate system via Southern Ocean upwelling1, carbon exchange with the deep ocean2, Agulhas leakage (transport of Indian Ocean waters into the Atlantic)3 and possibly Antarctic ice-sheet stability4. Meridional shifts of the Southern Hemisphere westerly winds have been hypothesized to occur5,6 in parallel with the well-documented shifts of the intertropical convergence zone7 in response to Dansgaard-Oeschger (DO) events- abrupt North Atlantic climate change events of the last ice age. Shifting moisture pathways to West Antarctica8 are consistent with this view but may represent a Pacific teleconnection pattern forced from the tropics9. The full response of the Southern Hemisphere atmospheric circulation to the DO cycle and its impact on Antarctic temperature remain unclear10. Here we use five ice cores synchronized via volcanic markers to show that the Antarctic temperature response to the DO cycle can be understood as the superposition of two modes: a spatially homogeneous oceanic 'bipolar seesaw' mode that lags behind Northern Hemisphere climate by about 200 years, and a spatially heterogeneous atmospheric mode that is synchronous with abrupt events in the Northern Hemisphere. Temperature anomalies of the atmospheric mode are similar to those associated with present-day Southern Annular Mode variability, rather than the Pacific-South American pattern. Moreover, deuterium-excess records suggest a zonally coherent migration of the Southern Hemisphere westerly winds over all ocean basins in phase with Northern Hemisphere climate. Our work provides a simple conceptual framework for understanding circum-Antarctic temperature variations forced by abrupt Northern Hemisphere climate change. We provide observational evidence of abrupt shifts in the Southern Hemisphere westerly winds, which have previously documented1-3 ramifications for global ocean circulation and atmospheric carbon dioxide. These coupled changes highlight the necessity of a global, rather than a purely North Atlantic, perspective on the DO cycle.

The transition from glacial to interglacial periods has been hypothesized to be linked to millennial-scale changes in oceanic/atmospheric circulation, but the relationships between these phenomena remain poorly constrained. Here we present a speleothem oxygen isotope record from Yongxing Cave, China, spanning 40.9 to 33.1 ka and compare this to existing Antarctic proxy records. We find that decadal-to-centennial rapid shifts in the Asian summer monsoon, Antarctic temperature, atmospheric methane and carbon dioxide are all coupled together during Dansgaard–Oeschger cycles, which may suggest an important role of the Intertropical Convergence Zone and Southern Ocean in driving the global greenhouse gas changes. Analogous to millennial-scale variations in trend, amplitude and internal sub-centennial-scale structures during Dansgaard–Oeschger 8 and Heinrich Stadial 4, the Younger Dryas and Heinrich Stadial 1 during the last ice termination provided critical positive feedbacks to changes in terrestrial vegetation and northern ice volume, and may have contributed to glacial to interglacial transition.

Abrupt Change in Climate and Biotic Systems.

The evolution of the northern hemispheric climate during the last glacial period was beset by quasi-episodic iceberg discharge events from the Laurentide ice sheet, known as Heinrich events (HEs). The paleo record places most HEs into the cold stadial of the Dansgaard-Oeschger cycle. However, not every Dansgaard-Oeschger cycle is associated with a HE, revealing a complex interplay between the two modes of glacial variability. Here, using a coupled ice sheet-solid earth model, we introduce a mechanism that explains the synchronicity of HEs and Dansgaard-Oeschger cycles. Unlike earlier studies, our mechanism does not require a trigger during the stadial. Instead, the atmospheric warming signal during the interstadial of the Dansgaard-Oeschger cycle causes enhanced ice stream thickening that leads to the HE during the late interstadial. We demonstrate that this mechanism reproduces the key HE characteristics and provides an explanation for synchronous HEs from different regions of the Laurentide ice sheet.

Abrupt climatic oscillations around the North Atlantic have been identified recently in Greenland ice cores as well as in North Atlantic marine sediment cores. The good correlation between the ‘Dansgaard-Oeschger events’ in the ice and the ‘Heinrich events’ in the ocean suggests that climate, in the North Atlantic region, underwent several massive reorganizations in the last glacial period. A characteristic feature of these events seems to be their hierarchical structure. Every 7 to 10-thousand years, when the temperature is close to its minimum, the ice-sheet undergoes a massive iceberg discharge. This Heinrich event is then followed by an abrupt warming, then by several other oscillations, each one lasting between one and two thousand years. These secondary oscillations do not have a clear signature in marine sediments but constitute most of the ‘Dansgaard-Oeschger events’ found in the ice. Here we use a simplified model coupling an ice-sheet and an ocean basin, in order to illustrate how the interactions between these two components can lead to such a hierarchical structure. The ice-sheet model exhibits internal oscillations composed of ice-sheet growing phases and basal ice melting phases that induce massive iceberg discharges. These massive fresh water inputs in the ocean stop for a moment the thermohaline circulation, enhancing the temperature contrast between low- and high-latitudes. Just after this event, the thermohaline circulation restarts and an abrupt warming of high-latitude regions is observed. For some parameter values, these warmer temperatures have in turn some influence on the ice-sheet, inducing secondary oscillations similar to those found in paleoclimatic records. Although the mechanism presented here may be too grossly simplified, it nevertheless underlines the potential importance of the coupling between ice-sheet dynamics and oceanic thermohaline circulation on the structure of the climatic records during the last glacial period.

<p>Paleoclimate modelling using simple models, EMICs (Earth System Models of Intermediate Complexity) and GCMs (General Circulation Models) combined with ice sheet models has become a powerful tool for understanding how the long-term climate system with ice sheets responds to external forcings such as Milankovitch forcing. With the aid of supercomputers and advances in climate model development, it is now possible to perform a much larger number of snapshot experiments with fixed forcings as well as transient experiments with evolving forcings. This talk will review the models that simulate the Northern Hemisphere ice sheet change and climate during the ice age cycles and discuss upcoming challenges. The talk will also present recent works on simulating millennial scale climate changes and the link with the ice age cycle. The last termination of the ice age cycles as well as glacial periods were punctuated by abrupt millennial scale climate changes, such as the Bølling-Allerød interstadial, the Younger Dryas and Dansgaard-Oeschger events. Abrupt climate changes have been shown to be strongly linked to changes in the Atlantic Meridional Overturning Circulation (AMOC) and the shift between the (quasi) multiple equilibria of AMOC, but the mechanism behind these abrupt changes and the link to climate change in the orbital scale are not clear. Modelling the stability of AMOC under different climate conditions together with deglacial climate change using fully coupled ocean-atmosphere GCMs has been challenging. Here we present a series of long transient experiments of at least 10,000 years with forcings under different ice sheet sizes, greenhouse gas levels and orbital parameters, as well as deglacial experiments following PMIP4 protocols, using a coupled ocean-atmosphere model, MIROC4m AOGCM. When forcing under glacial condition is applied, even without freshwater perturbation, the climate-ocean system shows self-sustained oscillations within a “sweet spot.” We also see a bipolar seesaw pattern and switching between interstadials and stadials, whose return time ranges from 1,000 years to nearly 10,000 years depending on the background forcing during the ice age cycle. Our transient deglaciation experiment with a gradually changing insolation, greenhouse gas forcing and ice sheet with meltwater from the glacial period to the Holocene is analysed and compared with proxy data as well as with the series of experiments with self-sustained oscillations for a better interpretation. Implications on the role of abrupt climate changes in shaping the longer-term global ice age cycle are further discussed.</p>