Antarctica Has Been Continuously Glaciated for the

1. Introduction

The Eocene–Oligocene transition (EOT), approximately 34 million years ago (34 Ma), represents one of the major climatic transitions in the past 65 Myr [1]. A steady decline in temperatures from approximately 50 to approximately 34 Ma culminated in the rapid expansion of permanent ice over Antarctica [2–4]. Previous work has generally focused on two main causes of the EOT: the CO₂ hypothesis and the ocean-gateway hypothesis. The first suggests the global decline in CO₂ (itself due to increased biological uptake and carbon cycling, and/or increased silicate weathering, and/or decreased volcanic emissions) caused the drop in global temperature [5] which, owing to polar amplification through ice-albedo and topographical feedbacks, led to the growth of the Antarctic ice sheet (AIS) [6–9]. The second hypothesis is that the opening of the ocean passages around Antarctica—the Drake Passage (DP) between South America, and Tasman Seaway (TS) between Australia—allowed for the initiation of the Antarctic circumpolar current (ACC) which reduced poleward heat transfer and so led to glacial inception [10,11]. Although there is geological and modelling evidence for both hypotheses, recently an increasing body of evidence has given precedence to falling atmospheric CO₂ as being the chief driver of the EOT, with the precise timing influenced by orbital variations [6–8,12–17].

The remainder of this section summarizes proxy records of palaeoclimate through the EOT, the results of a number of recent modelling studies exploring the effects of AIS expansion on global climate, and develops the motivation for our own work.

By way of introduction, the EOT is recorded as a two-stepped, orbitally paced, approximately 1.5‰ increase in deep ocean benthic foraminiferal oxygen isotope (δ ¹⁸O) records [13,18,19]. Benthic and low-latitude planktonic δ ¹⁸O changes are shown in figure 1a. These records reflect both an increase in continental ice volume and a decrease in ocean temperature, owing to concentration of ¹⁶ O in ice sheets by the hydrological cycle and isotopic fractionation during the calcification process, respectively. Attempts to deconvolve the EOT δ ¹⁸O increase into its ice volume and temperature components require the use of palaeotemperature proxies such as benthic foraminiferal Mg/Ca ratios. Further uncertainties arise in using Mg/Ca ratios owing to a simultaneous approximately 1 km deepening of the calcite compensation depth [13,22–25], as well as error in the Mg/Ca value chosen for Eocene–Oligocene seawater and its relationship with foraminiferal Mg/Ca. Uncertainty owing to calcite compensation depth deepening is reduced by using continuously paired δ ¹⁸ O and Mg/Ca records, such as those from Tanzania Drilling Programme (TDP) [4,20]. Additionally, the long residence times of Mg and Ca in the oceans (approx. 10 Myr and approx. 1 Myr, respectively) mean that calculated changes in temperature and δ ¹⁸ O _sw across the transition are more robust than absolute estimates [4].

The first step in the δ ¹⁸O record (figure 1a) predominantly reflects an ocean cooling in low-latitude surface waters with a small component of ice volume growth (approx. 0.2‰ δ ¹⁸O_sw equivalent) [4]. Cooling ranges from approximately 2.5°C at TDP [4] to approximately 3–4°C in a multi-proxy Gulf of Mexico record [26] (note, however, that the second step is absent for this record). High-latitude Mg/Ca records from the Kerguelen Plateau in the Southern Ocean (SO) also suggest a 2.5°C cooling [27]. The second step is dominated by an increase in ice volume (approx. 0.4‰ δ ¹⁸O_sw equivalent), supported by a decrease in the low-latitude TDP benthic–planktonic foraminiferal δ ¹⁸ O gradient, as shown in figure 1b. This is interpreted as reflecting global sea-level fall as ice sheets advanced [4]. There is also evidence for some further deep ocean cooling at the second step (totalling approx. 4°C across the EOT) that is not present at the surface in low-latitude TDP records (assuming minimal change in local surface salinity), as shown in figure 1c [4,13,20]. A multi-site study using the TEX₈₆ and U^K' ₃₇ organic temperature proxies suggests there was a heterogeneous sea surface temperature (SST) response at the EOT [28]. Modelling results, however, are generally consistent with an overall deep-water cooling of approximately 4°C and an overall increase in ice volume of 10–30×10⁶ km³ across the EOT [28].

Modelling is important in fingerprinting the driving mechanisms behind the EOT. Although often studies have chiefly been concerned with the trigger of glacial inception or the ACC, the range of results and simulation set-ups in the literature pose interesting questions about the role of palaeogeography and the presence of the AIS on global climate. Indeed, several modelling studies in the past 2 years have explored the effects of AIS expansion on global climate, and we summarize these here.

Goldner et al. [29] simulated the climate of the modern day and the Eocene, both with and without an AIS, using the NCAR CESM1.0 global climate model with a slab ocean. They find a contrasting response to the formation of the AIS between the modern and Eocene, which they attributed to differing cloud feedbacks. In the Eocene, the feedback is negative, causing warming when the AIS forms, whereas the cloud feedback is weakly positive in the modern. They acknowledge that this result is likely to be model dependent.

Building on their earlier work, Goldner et al. [16] used the same model (NCAR CESM v. 1.0) but with a more realistic three-dimensional fully dynamic ocean and also included simulations with different CO₂ concentrations. They suggest that changes found in the high-latitude ocean core records at the EOT, previously explained only by the gateway hypothesis and not the CO₂ hypothesis, could be an outcome of the formation of the AIS itself. The formation of the AIS increased the meridional pressure gradient around the continent, in turn increasing the surface winds that affect the surface mixing in the SO. This led to enhanced northward transport of Antarctic Intermediate Water and invigorated Antarctic Bottom Water formation. Around the Antarctic coastline, the SST was found to have areas of both increases (in the Indian Ocean sector) and decreases (in the Pacific and Atlantic sectors).

Knorr & Lohmann [30] carried out a similar study with a fully coupled climate model (ECHAM5/MPIOM) with a three-dimensional ocean, but looked at a different time period and hence palaeogeography, that of the Middle Miocene approximately 14 Ma. Their work indicates that the addition of the AIS causes a contrasting response in the SST around Antarctica to that of Goldner et al. [16], with strong warming in the Atlantic and Indian Ocean sectors (particularly in the Weddell Sea) and mild cooling in the Pacific sector. These changes in SST between glaciated and unglaciated simulations they attribute to winds affecting ocean circulation, deep-water formation and sea ice cover.

Hill et al. [31] present results using the same climate model as used in this paper, HadCM3L. They look at several scenarios and palaeogeographies (the Rupelian and Chattian stages, and pre-industrial modern) with differing boundary conditions (AIS presence and an open or closed DP, but not every combination thereof). Although the Rupelian is also used in the current study, the palaeogeography has since been updated (for example in West Antarctica). They suggest that although the ACC starts to initiate with the opening of the DP and the TS, an ACC of present-day strength does not develop until after the Chattian owing to the proximity of the Australian continent to Antarctica prior to this time. They argue this result is consistent with geologic data [32,33].

There are several recurring points that emerge from these studies. First, the response of the climate system to an expansion in Antarctic ice appears to vary significantly between models. However, it is also possible that the results are strongly dependent on the time period (i.e. palaeogeography) examined—[16,30,31] all examine differing time periods and/or palaeogeographies. The significance of palaeogeography cannot be underestimated, particularly when there is such large debate over the timings of key events such as the opening of the DP [34–36]. Additionally, it seems that the effect of changing ocean circulation is a significant driver of the global climate at the EOT, even if it was not the ultimate trigger of Antarctic glaciation [14–16]. Palaeogeography, ocean circulation, CO₂ and climatic change are all therefore interconnected, and each warrants a complete understanding at this major climatic shift.

Here, we model and describe the global atmospheric and oceanic response to the glaciation of Antarctica in the HadCM3L model and compare this with previous studies. This experiment may therefore be indicative of the climatic changes that occurred during the second 'ice growth' step of the EOT. We do this for two successive periods over the EOT: the Priabonian during the Eocene (approx. 38–34 Ma), and the Rupelian during the Oligocene (approx. 34–28 Ma). These stages have subtly different palaeogeographies; for example, the TS is wider in the Rupelian than the Priabonian in our reconstructions (see §2). We do not attempt to model the inception of glaciation in a transient simulation, like the studies of DeConto & Pollard [6] or Ladant et al. [17], but instead, through steady-state simulations, we aim to explore the importance of the palaeogeography on the regional and global response to glaciation, and the associated feedbacks. Our experimental design is briefly outlined in §2, descriptive results in §3, analysis and discussion of the physical mechanisms and implications in §4, and conclusions and future direction of work in §5.

2. Experimental design

To investigate the model sensitivity to AIS expansion and subtle variations in the palaeogeography, the model HadCM3L was run towards a steady state using high-resolution palaeogeographic reconstructions of the stages, with and without reconstructed ice sheets. HadCM3L is a fully coupled atmosphere–ocean general circulation model (GCM) [37]. HadCM3L is a lower-resolution version (3.75°×2.5°) of the HadCM3 model [38], with a few changes in parametrizations (e.g. [37,39]) compared with HadCM3. No ice sheet model was included; the AIS was simply prescribed with an appropriate land surface albedo and topography. The ice sheet topography, as well as global bathymetry and topography were supplied by Getech PLC at a 0.25° resolution. These were based upon tectonic, lithologic and fossil studies and DSDP/ODP deep-sea data [40]. The principal ice sheet used is a reconstruction for the Rupelian and is similar to the full modern extent (approx. 17×10⁶ km³ in size). A smaller ice sheet was also used for cross-examination, similar to only the East AIS (EAIS; approx. 14×10⁶ km³ in size). These reconstructions are shown in figure 2, and experiments listed in table 1.

Figure 2. Palaeogeographic reconstructions of topography and bathymetry (m) for the Priabonian (a) and Rupelian (b) stages shown at model resolution. The thickness (m) and extent of the reconstructed full (c) and East Antarctic ice sheets (d) are also shown.

Table 1. Key details of each model simulation carried out. Those experiments marked with an asterisk are the principal experiments around which the results and discussion are based.

Collapse
simulation name	palaeogeography	ice sheet configuration	spin-up phase with final ice sheet state	CO₂ (ppm)
PriNO*	Priabonian	ice-free	3	560
PriEAIS	Priabonian	East Antarctic ice only	4	560
PriFULL*	Priabonian	full Antarctic ice sheet	4	560
RupNO*	Rupelian	ice-free	3	560
RupEAIS	Rupelian	East Antarctic ice only	4	560
RupFULL*	Rupelian	full Antarctic ice sheet	3	560

Model spin-up was carried out in four phases consistently across the simulations. First, in Phase 1, the model was initialized with pre-industrial CO₂ levels and a zonally symmetric SST and deep ocean distribution, uniform vegetation and the different ice-free palaeogeographies, and run for 50 years. In Phase 2, the vegetation model TRIFFID [41] was activated, CO₂ was increased to four times pre-industrial, the prescribed ozone concentration was adjusted [42], and the model run for 319 years. In Phase 3, prescribed lakes were added to the land surface and the ice sheet added in simulations requiring it, and the model run for 53 years. Finally, in Phase 4, barotropic flow in the ocean is calculated (prior to this point, only the baroclinic component was calculated), the CO₂ is reduced to two times pre-industrial (560 ppm), and for ice sheet sensitivity simulations the ice sheet is adjusted (differences are noted in table 1). This final phase is run for 1000 years. This combined spin-up of 1422 years produces quasi-stable results for atmospheric and surface ocean characteristics as deep as 670 m, as shown in electronic supplementary material, figure S1. There is still, however, a trend in deep ocean temperatures greater than 1000 m, and so these are not the focus of this study. More details of the experimental design can be found in [43].

There are a number of points we note about the methodology for clarity. First, because the land–sea and ice masks do not align exactly between the stages, a small number of grid cells which should be prescribed as 'ice' lie off the continental margin of Antarctica during the Priabonian. It was deemed more important not to influence the bathymetry by changing the land–sea mask, and so these occasional cells of ice sheet were omitted. Second, adjustments to the ice sheet states for sensitivity simulations (PriEAIS, PriFULL and RupEAIS) were made only in the fourth phase of spin-up. These three simulations, therefore, have a slightly shorter spin-up (by 53 years) with their final ice sheet configuration than RupFULL. Third, in our HadCM3L set-up, the standard deviation of topography (used in the gravity-wave drag scheme) is estimated as a constant factor of the topographic height. This is a fair assumption over mountainous regions, but exaggerates the standard deviation over the AIS, potentially altering atmospheric dynamics. Fourth, the solar constant varies fractionally between the stages (1360.86 and 1361.35 kW m⁻² for the Priabonian and Rupelian, respectively). It is not believed that any of these points should have a major impact on the findings of the study.

Finally, the solution of the barotropic component of ocean circulation requires the user to define islands around which the net flow can be non-zero. Any continent and its coastline that is not defined as an island is assigned a barotropic stream function value of 0. For the Rupelian, we define Australia, South America and Antarctica to be islands in this sense, whereas in the Priabonian, with a more constricted TS, we define only Australia and South America as islands. As such, there is zero net flow around the Antarctic coastline in the Priabonian, despite the presence of a (albeit small) TS. By assigning the same value of stream function to Antarctica as the Eurasian and North American continents, we provide a coupling between the SO and tropical Pacific gateways in the Priabonian that is not present in the Rupelian (figure 5, shown later). This should not have a major effect on flow through the TS or the high-latitude response during the Priabonian, but may potentially affect the climate response at low latitude.

3. Results

Here, oceanic and atmospheric response to the expansion of the full AIS (Δ_glac) will be the primary focus (i.e. FULL—NO), discussed in terms of temperature, atmospheric and ocean circulation, and meridional heat flux changes. The discussion of mechanisms follows in §4a.

Annual mean global surface air temperature (SAT) Δ_glac response is dominated by the cooling of up to 30°C over the Antarctic continent in both stages, as shown in figure 3a,b. The annual mean SST Δ_glac response varies substantially between the stages. The Priabonian shows a large area of warming of up to 6°C over the Pacific sector of the SO, whereas there is cooling of up to 1.5°C in the North Atlantic, northwestern Pacific and Bellingshausen Sea (figure 3c). In contrast, the Rupelian shows more subtle variations and is more zonally heterogeneous in the Southern Hemisphere (SH), as shown in figure 3d. Cooling occurs in much of the SH, particularly in the Indian Ocean sector of the SO by approximately 2°C, whereas warming of up to 1.5°C occurs in the South Pacific sector of the SO and some localized coastal areas by greater than 3°C. There is also year round cooling of approximately 6°C in the North Atlantic. In both stages, there is seasonality in the SST Δ_glac response, with warming being stronger during the winter (JJA) and cooling being stronger in the summer (DJF) as shown in electronic supplementary material, figure S2. The only Δ_glac change in average annual sea ice concentration in either stage is a partial retreat of the sea ice in the Ross Sea during the Priabonian (figure not shown).

Figure 3. Annual mean SAT response (°C) to Antarctic glaciation for the Priabonian (a) and Rupelian (b), and annual mean SST response (°C) to glaciation for the Priabonian (c) and Rupelian (d).

Annual mean mid-atmospheric pressure (P; expressed as the 500 hPa geopotential height) Δ_glac response is shown for both stages in figure 4a,b. P decreases significantly over Antarctica and the SO and increases in southern mid-latitudes between 40° and 60° S, increasing the P gradient at 60° S. Changes are more pronounced for the Rupelian than the Priabonian. This response is also seasonally dominated, mostly occurring during the summer with only slight decreases in P over the Ross Sea edge of East Antarctica and some increases over the South Pacific and SO during winter (figure not shown). The changes in P result in Δ_glac changes in zonal winds, as shown in figure 4c (see also electronic supplementary material, figure S3, for annual mean wind vectors before and after glaciation). The largest changes in wind field are off the AIS owing to the initiation of katabatic winds, and these are persistent throughout the year. Elsewhere, changes in winds are generally geostrophic in nature, aligning with changes in the pressure gradient contours. Over the Pacific sector of the SO at 60° S, the increased summer pressure gradient leads to strengthening and a poleward shift in the geostrophic westerlies in both stages.

Annual mean ocean zonal flow through the TS and DP for each stage and ice sheet configuration is listed in table 2, and the depth-integrated stream functions for the principal experiments (PriNO, PriFULL, RupNO, RupFULL) shown in figure 5. There are clear differences in the ocean circulation patterns between the stages, regardless of ice sheet state. The Pacific at this time is a basin with four gateways: the DP, the TS, the Indonesian Seaway (IS) and the Panama Seaway. During the Priabonian, the western gateways (TS and IS) make up the majority (68–77%) of the flow through the Pacific, whereas during the Rupelian, the southern gateways (TS and DP) represent the majority (76–86%) of the total flow. Large gyre systems dominate the Indian, South Pacific and SO during the Priabonian. During PriNO, there is a huge Ross Sea gyre extending to approximately 35° S and along the entire Antarctic coastline between the TS and DP. With the addition of the ice in PriFULL, the Ross Sea gyre shrinks considerably, reaching approximately 45° S and contracting westwards, no longer influencing the Bellingshausen Sea. Simultaneously, flow in the South Pacific and Indian Ocean gyres (and flow between them) strengthens. In the Rupelian, flow through the SO is zonal, dominated by a proto-ACC. Addition of the ice sheet strengthens this proto-ACC, but has limited impact on the rest of the global ocean. Total flow through all Pacific gateways increases by 97% during the Priabonian but only 27% during the Rupelian in response to Δ_glac.

Figure 5. Annual mean barotropic stream function (Sv) for the Priabonian: unglaciated (a) and glaciated (b); and Rupelian: unglaciated (c) and glaciated (d). (Online version in colour.)

Table 2. Annual mean ocean transport through Pacific Ocean gateways for each simulation.

Collapse
simulation	Tasman Seaway through flow (Sv)	Drake Passage through flow (Sv)	total flow through Pacific gateways (Sv)
PriNO	12.0	3.6	31.2
PriEAIS	19.8	9.2	58.0
PriFULL	21.1	9.6	61.5
RupNO	52.6	46.2	130.1
RupEAIS	71.3	64.5	159.0
RupFULL	75.2	68.0	165.8

Mixed layer depth (MLD) in the SO and North Atlantic for the four principal experiments is shown in figure 6. There are increases of 200–1000 m in the Ross Sea for both stages and increases along the Atlantic and Indian Ocean sectors of the East Antarctic coastline by up to 1000 m in the Priabonian and 500 m in the Rupelian. There are decreases in the Weddell Sea of the Priabonian (up to 670 m) but increases in the region during the Rupelian (up to 600 m). There is a decrease in MLD in the North Atlantic in both stages. The increase in the Ross Sea MLD during the Priabonian is important, as PriNO has a very shallow MLD of less than 100 m, which increases up to 1000 m in PriFULL. Additionally, increased surface winds at 60° S result in seasonal Δ_glac changes in the wind-driven mixing in the SO, with increased upwelling to the south and increased downwelling to the north of the peak wind strength and current change. This causes a summer seasonal deepening of the MLD throughout the SO by up to 25 m (figure not shown), but this is small compared with the coastal increases previously mentioned. The global meridional overturning stream function is not shown owing to the insufficient spin-up time for the deep ocean.

The meridional heat flux has the same general form for all ice sheet configurations (and stages), shown by the example of RupNO in figure 7a. The oceanic and total heat fluxes increase southwards at all latitudes for both stages in response to Δ_glac, as shown in figure 7b,c. The largest increases in the total southward heat flux (0.4 PW in the Priabonian; 0.3 PW in the Rupelian) occur in the SO at 65° S, as a result of strong oceanic and atmospheric components at this latitude. The maximum increases in oceanic southward heat flux (0.4 PW in the Priabonian at approx. 40° S; 0.2 PW in the Rupelian at approx. 60° S) are similar in magnitude to those found in previous studies [16,44]. Changes owing to gateway opening (i.e. the RupNO − PriNO; figure 7d) show little net change in the SH total southward heat flux, with the (northwards) atmospheric component mostly balancing the (southwards) oceanic component.

4. Discussion

(a) Mechanisms

Owing to the complexity of GCMs, and in particular the feedbacks that act between many aspects of the system, diagnosing the mechanisms behind changes seen in the model in response to a change in boundary conditions is challenging. The only way to confidently assess mechanisms is to carry out sensitivity studies in which particular aspects of the system are kept constant, and/or feedback loops are cut. However, owing to the computational cost of additional simulations, this is rarely carried out. Here, the mechanisms behind the observed responses will be discussed qualitatively. We aim to provide plausible mechanistic explanations for the changes seen in both stages, but without additional sensitivity studies these mechanisms remain somewhat speculative.

Two fundamental factors of Antarctic glaciation affect climate: (i) the increased surface albedo over the continent, reducing the solar radiation absorbed at the surface and (ii) the increased height of the new ice sheet topography, changing the reference height and modifying atmospheric pressure and circulation. The combined effect of these two factors causes the dramatic SAT cooling over Antarctica (figure 3). The gravitational potential of this cold air initiates katabatic winds off the ice sheet [45]. The outflow of the katabatic winds enhances polar circulation over Antarctica, and so between 60° and 80° S air is rising to replace that flowing off the AIS leading to reduced P (figure 4). It would be expected that cooling air over the AIS would lead to sinking and be associated with increasing P (similar to Goldner et al. [16]). However, the strong temperature seasonality of HadCM3L over Antarctica [46] negates this pressure effect with large decreases in summer P. It should be noted, however, in terms of mean sea-level pressure, P is found to increase over Antarctica (figure not shown; similar to Goldner et al. [16]). This is subject to interpolation that may be unrealistic when adding a thick AIS.

Regardless of the mechanism of P change over the AIS, the Δ_glac decrease over the SO is expected and results in an increased north–south pressure gradient at the polar front, approximately 60° S. This causes increases in westerly winds over the Pacific sector of the SO (figure 4c) that result in slight changes in the wind-driven eastward flow [47]. The change in zonal wind stress over the SO leads to some increased wind-driven upwelling south of 60° S and downwelling north of 60° S (figure not shown). Some of this wind-driven overturning change would be expected to be balanced by an increased eddy field; however, the relatively coarse model resolution means that these eddies are not resolved [48,49].

By comparing the Δ_glac SST response with the change in SST observed between the stages with the same ice sheet states (electronic supplementary material, figure S4), it is possible to identify PriNO as having an anomalously cold Ross Sea and South Pacific sector. This sensitive stratified ocean state is associated with a lack of deep-water formation in the region, indicated by the very shallow MLD (figure 6a). Deep-water formation draws water from lower latitudes, releasing heat as it sinks. The interconnectivity of the causes and changes associated with deep-water formation make untangling these results particularly challenging, but a potential explanation of why this process is absent during PriNO is offered here.

During the Priabonian, constricted ocean gateways do not allow a strong barotropic circumpolar current to form around Antarctica, with eastward flowing water in the Pacific sector being part of larger South Pacific or Ross Sea gyre circulations (figure 5a,b). The sluggish zonal circulation prior to glaciation allows the Ross Sea gyre to expand both northwards and eastwards across the width of the Pacific sector of the SO (similar to [14] with a closed TS). As a result, water is transferred to high latitudes along the eastern boundary of the Pacific and returns westwards along the Antarctic coastline, and is very cold upon reaching the Ross Sea. Combined with the lack of deep-water formation and a sea ice albedo feedback, this causes the region to be anomalously cold compared with the other simulations. Deep-water formation in the polar regions is strongly affected by salinity [50], so despite the low Ross Sea SST in PriNO water density and deep-water formation are suppressed by low salinities (electronic supplementary material, figure S5a). With the addition of the AIS (or expansion of the gateways), the salinity in the Ross Sea increases (electronic supplementary material, figure S5b), initiating sinking. This salinity change could be the result of changes in sea ice formation, changes in net evaporation and precipitation (E−P) and/or changes in ocean circulation.

Changes in sea ice formation are ruled out as being an important driver of the Ross Sea Δ_glac salinity response, because sea ice extent reduces with the inclusion of the ice sheet, which will result in less brine rejection. Sea ice causes a seasonal increase in salinities, but this effect diminishes once the ice sheet forms. The South Pacific and SO experience net precipitation (P>E) in all simulations, contributing to the low salinities found there. In the SO, the highest salinities are close to the TS and decrease with distance east (electronic supplementary material, figure S5). The E−P Δ_glac response shows an increase in evaporation over the Ross Sea of the Priabonian, but there is still net precipitation over the region. This may not be the driver of change but will act as a positive feedback with deep-water formation (whereby enhanced evaporation increases salinity, potentially increasing deep-water formation, which in turn releases heat and further enhances evaporation). Pacific sector salinity shows a strong correlation with flow through the TS from the high salinity Indian Ocean. It is plausible that the low salinity in PriNO is due to the greatly reduced flow through the TS. Once flow is increased, either by opening the gateway or adding the ice sheet, high salinity water penetrates into the Ross Sea, initiating deep-water formation.

A combination of the latter two processes is likely to have brought about the change in the Ross Sea salinity, with advection of water through the TS potentially triggering and enhanced E consolidating the increase [14]. Deep-water formation then enhances meridional transport of water, causing the large Ross Sea gyre to contract westwards away from the Bellingshausen Sea. As a result, the Bellingshausen Sea coast cools (less meridional transport to the region) and the Ross Sea warms. The warming is strong and extensive in contrast to the results of Sijp et al. [14], who compared a fully open and closed TS but with no change in ice sheet. Further feedbacks may have contributed to this strong response: first, with the collapse of sea ice and associated albedo effect enhancing the temperature anomaly in the region; second, with the smaller Ross Sea gyre reducing the transport time of water within the gyre, reducing the freshening effect of the negative net E−P in the region; and third, with steepening of the isopycnals in the SO resulting in a stronger zonal depth-integrated stream function through thermal wind.

The Rupelian SST Δ_glac response is not as strong as the Priabonian, but there are still some clear spatial and seasonal patterns. The warming and cooling responses can be explained by separate mechanisms. The summer SST cooling over the Indian Ocean sector of the SO appears to be a direct result of glaciation through atmospheric cooling and katabatic winds. The proximity of the Indian Ocean sector to the EAIS, where there is Δ_glac cooling of greater than 30°C and katabatic winds to transfer this extremely cold air over the SO, affects the summer SST. The Δ_glac change in southward heat flux (figure 7c) shows a global increase in the oceanic component, suggesting that the cooling has a local source and is not the result of changes in ocean circulation (i.e. reduced southward heat flux). Future work will aim to confirm this by comparing results to simulations with a slab ocean with a fixed ocean heat flux (which we hypothesize to show the same signal). In the winter, the Δ_glac warming response found in the South Pacific sector of the SO is caused by circulation changes. Stronger westerlies and seasonally persistent katabatic winds intensify the Ross Sea gyre (shown by the sharper 'kink' in the depth-integrated stream lines in figure 5d and the surface currents in electronic supplementary material, figure S5d), transferring waters meridionally and bringing heat southwards. The signal is not as clear during the summer, however, because the increased westerly winds at 60° S transfer colder water from the Indian Ocean sector to the Pacific sector of the SO and Ross Sea, overprinting the signal. Some stronger warming occurs along coastlines where there is enhanced deep-water formation, shown by the increases in MLD between figure 6c and d.

It should be noted that the responses and mechanisms outlined here are also consistent for glaciation of only East Antarctica, with the main area of Ross Sea warming still present during the Priabonian but having a slightly lower magnitude (exhibiting maximum warming values of approx. 4.5°C; figure not shown). There are only subtle variations in the Weddell Sea MLD and SST responses to East Antarctic glaciation compared with those discussed here, showing the effect of the West Antarctic ice sheet is small and localized.

(b) Implications

These results would suggest that prior to an established proto-ACC (or sufficient TS opening) there is a negative feedback on AIS growth, in which increases in Antarctic ice volume lead to large increases in SST in the Pacific SO, potentially inhibiting further Antarctic ice growth. If this were the case, it may have contributed to the temporality of ephemeral ice sheets during the Eocene [51]. This feedback wanes once the TS and DP widen and the ACC becomes stronger, with cooler summer SO temperatures, particularly in the Indian Ocean sector. There is also a potential mild positive feedback with winter warming in the coastal SO, which may enhance evaporation and Antarctic ice growth. However, there is evidence for a limited effect of SST changes on the interior of the Antarctic continent within models [44].

Additionally, the unglaciated Priabonian results are inconsistent with some of the (albeit limited) proxy datasets available. Douglas et al. [52] find a zonal gradient in SST proxies between the warmer southwest Pacific and the cooler south Atlantic of 7°C during the middle-to-late Eocene. A similar temperature gradient is found here for the Rupelian, but not for the Priabonian (see electronic supplementary material, figure S6). During the Rupelian, the gradient is greater when Antarctica is unglaciated (4°C), reducing slightly with glaciation (3°C) owing to the cooling and transfer of Indian Ocean sector waters reducing the peak temperatures found in the Pacific sector. It should be noted the absolute values are approximately 15°C cooler than the simulations and proxy reconstructions of Douglas et al. owing to the different atmospheric CO₂ levels used (their simulations used 2400 and 1600 ppm, representative of the mid-to-late Eocene). We suggest, therefore, the key implication of these results is that subtle changes in palaeogeography and model set-up can result in substantially different regional climate responses: potentially of great importance when interpreting model results. Hence, it is not surprising that the results differ from those of Goldner et al. [16] and Knorr & Lohmann [30].

This conclusion is further corroborated by comparing the zonal flow Δ_glac response for the Rupelian with the results of Hill et al. [31]. In our study, adding the AIS in the Rupelian increases zonal flow by greater than 20 Sv (10⁶ m³ s⁻¹), whereas Hill et al. found it reduced flow through the DP and TS by 2 and 10 Sv, respectively. Although their simulations were also carried out using the same model (HadCM3L) and atmospheric CO₂ levels, their spin-up procedure, model set-up and palaeogeographic reconstructions were different, highlighting how experimental design can impact results, even using the same model.

A number of additional experiments were carried out with the same spin-up procedures as those discussed here. One of these was an ice-free Priabonian simulation with four times pre-industrial CO₂ levels (more physically realistic for the late Eocene [7,8]). This experiment also showed a very cold Ross Sea with no deep-water formation, so while we note this result is likely to be model dependent, it is robust to a change in CO₂ concentration. Another set of simulations was carried out for the Chattian stage (approx. 28–23 Ma) that follows the Rupelian (approx. 34–28 Ma). In this stage, there was further widening of the TS and DP by another grid cell in the model, and the same ice sheet and ice-free configurations were used. With this additional opening of the southern gateways, flow approximately doubles compared with the Rupelian, and with the addition of the AIS there is an established ACC and flow through the DP of 103.5 Sv (a similar order of magnitude to present-day measurements of approx. 130 Sv [53]). This would support studies suggesting the ACC strengthened in the mid-to-late Oligocene [32,33]. It is also worth noting the zonal flow through the DP and TS is once again different from the results of Hill et al. [31], being 10–20 Sv higher in our study.

All stages (including the Chattian) show that enhanced zonal ocean circulation owing to Antarctic glaciation and gateway opening causes increased poleward heat transport in the SH, contrary to traditional theory [10,54,55] but similar to the results of Goldner et al. [16]. These results suggest that the enhanced circulation does not act as a barrier to inhibit poleward heat transport. As noted previously, these increases are similar to those found in previous studies [16,44]; however, it remains to be seen if it is great enough to have a significant effect on Antarctic climate [44].

5. Conclusion

This study identified the climatic and oceanic response to glaciation in the HadCM3L model for two geological stages on either side of the EOT. The climatic response to glaciation varied significantly between the two stages, owing to a complex network of feedbacks acting to cool the Ross Sea during the unglaciated Priabonian simulation. The sensitive, cold ocean state in this simulation is easily destabilized with the onset of deep-water formation in response to either Antarctic glaciation or gateway opening. This response is likely to be very model dependent (it does not match with the limited proxy data available for the region for example), but it highlights how, through multiple feedbacks, subtle changes in palaeogeography and model set-up can produce very different modelled climatic responses to Antarctic glaciation.

Future research can expand in multiple directions. Initial work could carry out a slab ocean model run for the Rupelian to affirm the conclusions about Indian Ocean cooling being due to the proximity with the AIS. Running similar experiments with different coupled climate models will be essential in assessing if this sensitive behaviour is a feature of other models, or if this is specific to HadCM3L. Further work will also include more detailed comparison of the modelled response to the geological record. Finally, using these high-resolution palaeogeographic reconstructions, it would be interesting to carry out transient simulations under changing CO₂ forcing with a coupled ice sheet model (similar to those of [6,17]) to model glacial inception and examine if regional sensitivity exists in this case also, and to properly explore the feedbacks between Antarctica and global climate.

Competing interests

We declare we have no competing interests.

Funding

We would like to thank NERC for funding A.T.K. through a GWA+ DTP grant, and Getech for providing the palaeogeography and a CASE award to A.T.K. We also acknowledge NERC grants NE/K014757/1 and NE/I005714/1.

Footnotes

One contribution of 12 to a discussion meeting issue 'Feedbacks on climate in the Earth system'.

References

1
Zachos JC, Dickens GR, Zeebe RE
. 2008 An early Cenozoic perspective on greenhouse warming and carbon cycle dynamics. Nature 551 , 279–283. (doi:10.1038/nature06588) Crossref, ISI, Google Scholar
2
Zachos JC, Shackleton NJ, Revenaugh JS, Pälike H, Flower BP
. 2001 Trends, rhythms, and aberrations in global climate 65 Ma to present. Science 292 , 686–693. (doi:10.1126/science.1059412) Crossref, PubMed, ISI, Google Scholar
3
Katz ME, Miller KG, Wright JD, Wade BS, Browning JV, Cramer BS, Rosenthal Y
. 2008 Stepwise transition from the Eocene greenhouse to the Oligocene icehouse. Nat. Geosci. 1 , 329–334. (doi:10.1038/ngeo179) Crossref, ISI, Google Scholar
4
Lear CH, Bailey TR, Pearson PN, Coxall HK, Rosenthal Y
. 2008 Cooling and ice growth across the Eocene–Oligocene transition. Geology 36 , 251–254. (doi:10.1130/G24584A.1) Crossref, ISI, Google Scholar
5
Zachos JC, Kump LR
. 2005 Carbon cycle feedbacks and the initiation of Antarctic glaciation in the earliest Oligiocene. Glob. Planet. Change 47 , 51–66. (doi:10.1016/j.gloplacha.2005.01.001) Crossref, ISI, Google Scholar
6
DeConto RM, Pollard D
. 2003 Rapid Cenozoic glaciation of Antarctica induced by declining atmospheric CO₂ . Nature 421 , 245–249. (doi:10.1038/nature01290) Crossref, PubMed, ISI, Google Scholar
7
Pearson PN, Foster GL, Wade BS
. 2009 Atmospheric carbon dioxide through the Eocene–Oligocene climate transition. Nature 461 , 1110–1113. (doi:10.1038/nature08447) Crossref, PubMed, ISI, Google Scholar
8
Pagani M, Huber M, Liu Z, Bohaty SM, Henderiks J, Sijp W, Krishnan R, DeConto RM
. 2011 The role of carbon dioxide during the onset of Antarctic glaciation. Science 334 , 1261–1264. (doi:10.1126/science.1203909) Crossref, PubMed, ISI, Google Scholar
9
Foster GL, Rohling EJ
. 2013 Relationship between sea levels and climate forcing by CO₂ on geological timescales. Proc. Natl Acad. Sci. USA 110 , 1209–1214. (doi:10.1073/pnas.1216073110) Crossref, PubMed, ISI, Google Scholar
10
Kennett JP
. 1977 Cenozoic evolution of Antarctic glaciation, the circum-Antarctic ocean, and their impact on global paleoceanography. J. Geophys. Res. 82 , 3843–3860. (doi:10.1029/JC082i027p03843) Crossref, ISI, Google Scholar
11
Sijp WP, England MH
. 2004 Effect of Drake Passage throughflow on global climate. J. Phys. Oceanogr. 34 , 1254–1266. (doi:10.1175/1520-0485(2004)034<1254:EOTDPT>2.0.CO;2) Crossref, ISI, Google Scholar
12
Huber M, Brinkhuis H, Stickley CE, Doos K, Sluijs A, Warnaar J, Schellenberg SA, Williams GL
. 2004 Eocene circulation of the Southern Ocean: was Antarctica kept warm by subtropical waters? Paleoceanography 19 , PA4026. (doi:10.1029/2004PA001014) Crossref, Google Scholar
13
Coxall HK, Wilson PA, Pälike H, Lear CH, Backman J
. 2005 Rapid stepwise onset of Antarctic glaciation and deeper calcite compensation in the Pacific Ocean. Nature 433 , 53–57. (doi:10.1038/nature03135) Crossref, PubMed, ISI, Google Scholar
14
Sijp WP, England MH, Huber M
. 2011 Effect of deepening the Tasman Gateway on the global ocean. Paleoceanography 26 , PA4207. (doi:10.1029/2011PA002143) Crossref, Google Scholar
15
Lefebvre V, Donnadieu Y, Sepulchre P, Swingedouw D, Zhang ZS
. 2012 Deciphering the role of southern gateways and carbon dioxide on the onset of the Antarctic circumpolar current. Paleoceanography 27 , PA4201. (doi:10.1029/2012PA002345) Crossref, Google Scholar
16
Goldner A, Herold N, Huber M
. 2014 Atlantic glaciation caused ocean circulation changes at the Eocene–Oligocene transition. Nature 511 , 574–577. (doi:10.1038/nature13597) Crossref, PubMed, ISI, Google Scholar
17
Ladant JB, Donnadieu Y, Lefebvre V, Dumas C
. 2014 The respective role of atmospheric carbon dioxide and orbital parameters on ice sheet evolution at the Eocene–Oligocene transition. Paleoceanography 29 , 810–823. (doi:10.1002/2013PA002593) Crossref, Google Scholar
18
Kennett JP, Shackleton NJ
. 1976 Oxygen isotopic evidence for the development of the psychrosphere 38 Myr ago. Nature 260 , 513–515. (doi:10.1038/260513a0) Crossref, ISI, Google Scholar
19
Zachos JC, Quinn RM, Salamy K
. 1996 High resolution (104 yr) deep-sea foraminiferal stable isotope records of the Eocene–Oligocene climate transition. Paleoceanography 11 , 251–266. (doi:10.1029/96PA00571) Crossref, Google Scholar
20
Pearson PN, McMillan IK, Wade BS, Dunkley Jones T, Coxall HK, Bown PR, Lear CH
. 2008 Extinction and environmental change across the Eocene–Oligocene boundary in Tanzania. Geology 36 , 179–182. (doi:10.1130/G24308A.1) Crossref, ISI, Google Scholar
21
Marchitto TM, Curry WB, Lynch-Stieglitz J, Bryan SP, Cobb KM, Lund DC
. 2014 Improved oxygen isotope temperature calibrations for cosmopolitan benthic foraminifera. Geochim. Cosmochim. Acta 130 , 1–11. (doi:10.1016/j.gca.2013.12.034) Crossref, ISI, Google Scholar
22
Lear CH, Rosenthal Y, Coxall HK, Wilson PA
. 2004 Late Eocene to early Miocene ice-sheet dynamics and the global carbon cycle. Paleoceanography 19 , PA4015. (doi:10.1029/2004PA001039) Crossref, Google Scholar
23
Lear CH, Mawbey EM, Rosenthal Y
. 2010 Cenozoic benthic foraminiferal Mg/Ca and Li/Ca records: toward unlocking temperatures and saturation states. Paleoceanography 25 , PA4215. (doi:10.1029/2009PA001880) Crossref, Google Scholar
24
Peck VL, Yu J, Kender S, Riesselman CR
. 2010 Shifting ocean carbonate chemistry during the Eocene–Oligocene climate transition: implications for deep-ocean Mg/Ca paleothermometry. Plaeoceanography 25 , PA4219. (doi:10.1029/2009PA001906) Google Scholar
25
Pusz AE, Thunell RC, Miller KG
. 2011 Deep water temperature, carbonate ion, and ice volume changes across the Eocene–Oligocene climate transition. Paleoceanography 26 , PA2205. (doi:10.1029/2010PA001950) Crossref, Google Scholar
26
Wade BS, Houben AJP, Quaijtaal W, Schouten S, Miller KG, Katz ME, Wright JD, Brinkhuis H
. 2012 Multiproxy record of abrupt sea-surface cooling across the Eocene–Oligocene transition in the Gulf of Mexico. Geology 40 , 159–162. (doi:10.1130/G32577.1) Crossref, ISI, Google Scholar
27
Bohaty SM, Zachos JC, Delaney ML
. 2012 Foraminiferal Mg/Ca evidence for Southern Ocean cooling across the Eocene–Oligocene transition. Earth Planet. Sci. Lett. 317–318 , 251–261. (doi:10.1016/j.epsl.2011.11.037) Crossref, ISI, Google Scholar
28
Liu Z, Pagani M, Zinniker D, DeConto R, Huber M, Brinkhuis H, Shah SR, Leckie RM, Pearson A
. 2009 Global cooling during the Eocene–Oligocene climate transition. Science 323 , 1187–1190. (doi:10.1126/science.1166368) Crossref, PubMed, ISI, Google Scholar
29
Goldner A, Huber M, Caballero R
. 2013 Does Antarctic glaciation cool the world? Clim. Past 9 , 173–189. (doi:10.5194/cp-9-173-2013) Crossref, ISI, Google Scholar
30
Knorr G, Lohmann G
. 2014 Climate warming during Antarctic ice sheet expansion at the Middle Miocene transition. Nat. Geosci. 7 , 376–381. (doi:10.1038/ngeo2119) Crossref, ISI, Google Scholar
31
Hill DJ, Haywood AM, Valdes PJ, Francis JE, Lunt DJ, Wade BS, Bowman VC
. 2013 Paleogeographic controls on the onset of the Antarctic circumpolar current. Geophys. Res. Lett. 40 , 5199–5204. (doi:10.1002/grl.50941) Crossref, ISI, Google Scholar
32
Pfuhl HA, McCave IN
. 2005 Evidence for late Oligocene establishment of the Antarctic circumpolar current. Earth Planet. Sci. Lett. 235 , 715–728. (doi:10.1016/j.epsl.2005.04.025) Crossref, ISI, Google Scholar
33
Lyle M, Gibbs S, Moore TC, Rea DK
. 2007 Late Oligocene initiation of the Antarctic circumpolar current: evidence from the south Pacific. Geology 35 , 691–694. (doi:10.1130/G23806A.1) Crossref, ISI, Google Scholar
34
Barker PF, Thomas E
. 2004 Origin, signature and palaeoclimatic influence of the Antarctic circumpolar current. Earth-Sci. Rev. 66 , 143–162. (doi:10.1016/j.earscirev.2003.10.003) Crossref, ISI, Google Scholar
35
Stickley CE
et al. 2004 Timing and nature of the deepening of the Tasmanian Gateway. Paleoceanography 19 , PA4027. (doi:10.1029/2004PA001022) Crossref, Google Scholar
36
Barker PF, Filippelli GM, Florindo F, Martin EE, Scher HD
. 2007 Onset and role of the Antarctic circumpolar current. Deep Sea Res. II 54 , 2388–2398. (doi:10.1016/j.dsr2.2007.07.028) Crossref, ISI, Google Scholar
37
Cox PM, Betts RA, Jones CD, Spall SA, Totterdell IJ
. 2001Modelling vegetation and the carbon cycle as interactive elements of the climate system. In Meteorology at the millennium (ed. R Pearce), pp. 259–279. San Diego, CA: Academic Press. Google Scholar
38
Gordon C, Cooper C, Senior CA, Banks H, Gregory JM, Johns TC, Mitchell JFB, Wood RA
. 2000 The simulation of SST, sea-ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Clim. Dyn. 16 , 147–168. (doi:10.1007/s003820050010) Crossref, ISI, Google Scholar
39
Jones C
. 2003 A fast ocean GCM without flux adjustments. J. Atmos. Ocean. Technol. 20 , 1857–1868. (doi:10.1175/1520-0426(2003)020<1857:AFOGWF>2.0.CO;2) Crossref, ISI, Google Scholar
40
Markwick PJ
. 2007The palaeogeographic and palaeoclimatic significance of climate proxies for data-model comparisons. In Deep-time perspectives on climate change: marrying the signal from computer models and biological proxies (eds M Williams, AM Haywood, FJ Gregory, DN Schmidt), pp. 251–312. London, UK: Geological Society for the Micropalaeontological Society. Google Scholar
41
Cox PM
. 2001Description of the TRIFFID dynamic global vegetation model. Technical note 24. Exeter, UK: Hadley Centre, Met Office. Google Scholar
42
Lunt DJ, Ross I, Hopley PJ, Valdes PJ
. 2007 Modelling Late Oligocene C4 grasses and climate. Palaeogeogr. Palaeoclimatol. Palaeoecol. 251 , 239–253. (doi:10.1016/j.palaeo.2007.04.004) Crossref, ISI, Google Scholar
43
Inglis GN
et al. 2015 Descent towards the icehouse: Eocene sea surface cooling inferred from GDGT distributions. Paleoceanography 30 , 1000–1020. (doi:10.1002/2014PA002723) Crossref, Google Scholar
44
Huber M, Nof D
. 2006 The ocean circulation in the southern hemisphere and its climatic impacts in the Eocene. Palaeogeogr. Palaeoclimatol. Palaeoecol. 231 , 9–28. (doi:10.1016/j.palaeo.2005.07.037) Crossref, ISI, Google Scholar
45
Parish TR, Waight KT
. 1987 The forcing of Antarctic katabatic winds. Mon. Weather Rev. 115 , 2214–2226. (doi:10.1175/1520-0493(1987)115<2214:TFOAKW>2.0.CO;2) Crossref, ISI, Google Scholar
46
Gasson E
et al. 2013 Uncertainties in the modelled CO₂ threshold for Antarctic glaciation. Clim. Past Discuss. 9 , 5701–5745. (doi:10.5194/cpd-9-5701-2013) Crossref, Google Scholar
47
Allison LC, Johnson HL, Marshall DP, Munday DR
. 2010 Where do winds drive the Antarctic circumpolar current? Geophys. Res. Lett. 37 , L12605. (doi:10.1029/2010GL043355) Crossref, ISI, Google Scholar
48
Abernathey R, Marshall J, Ferreira D
. 2011 The dependence of Southern Ocean meridional overturning on wind stress. J. Phys. Oceanogr. 41 , 2261–2278. (doi:10.1175/JPO-D-11-023.1) Crossref, ISI, Google Scholar
49
Munday DR, Johnson HL, Marshall DP
. 2015 The role of ocean gateways in the dynamics and sensitivity to wind stress of the early Antarctic circumpolar current. Paleoceanography 30 , 284–302. (doi:10.1002/2014PA002675) Crossref, Google Scholar
50
Assmann KM, Timmermann R
. 2005 Variability of deep water formation in the Ross Sea. Ocean Dyn. 55 , 68–87. (doi:10.1007/s10236-004-0106-7) Crossref, ISI, Google Scholar
51
Miller KG, Wright J, Katz M, Browning J, Cramer B, Wade BS, Mizintseva S
. 2008A view of Antarctic ice-sheet evolution from sea-level and deep-sea isotope changes during the Late Cretaceous–Cenozoic. In Antarctica: a keystone in a changing world, pp. 55–70. Washington, DC: National Academies Press. Google Scholar
52
Douglas PMJ, Affek HP, Ivany LC, Houben AJP, Sijp WP, Sluijs A, Schouten S, Pagani M
. 2014 Pronounced zonal heterogeneity in Eocene southern high-latitude sea surface temperatures. Proc. Natl Acad. Sci. USA 111 , 6582–6587. (doi:10.1073/pnas.1321441111) Crossref, PubMed, ISI, Google Scholar
53
Meredith MP, Woodworth PL, Hughes CW, Stepanov V
. 2004 Changes in the ocean transport through the Drake Passage during the 1980s and 1990s, forced by changes in southern annular mode. Geophys. Res. Lett. 31 , L21305. (doi:10.1029/2004GL021169) Crossref, ISI, Google Scholar
54
Gill AE, Bryan K
. 1971 Effects of geometry on the circulation of a three-dimensional southern-hemisphere ocean model. Deep Sea Res. 18 , 685–721. Google Scholar
55
England MH
. 1993 Representing the global-scale water masses in ocean general circulation models. J. Phys. Oceanogr. 23 , 1523–1552. (doi:10.1175/1520-0485(1993)023<1523:RTGSWM>2.0.CO;2) Crossref, ISI, Google Scholar

speicherbutenway.blogspot.com

Source: https://royalsocietypublishing.org/doi/10.1098/rsta.2014.0419

Antarctica Has Been Continuously Glaciated for the

1. Introduction

2. Experimental design

3. Results

4. Discussion

(a) Mechanisms

(b) Implications

5. Conclusion

Competing interests

Funding

Footnotes

References

0 Response to "Antarctica Has Been Continuously Glaciated for the"

Post a Comment

Iklan Atas Artikel

Iklan Tengah Artikel 1

Iklan Tengah Artikel 2

Iklan Bawah Artikel