Decoupled Antarctic and Subarctic export productivity under intensified Southern Hemisphere westerlies during the last interglacial

Abstract

The Last Interglacial (128–113 ka B.P.) provides key insights into climate-carbon interactions under warm, high-CO₂ conditions. While Antarctic temperature-driven Southern Ocean ventilation has been proposed to modulate atmospheric CO₂, role of Southern Hemisphere westerlies (SHW) remains unclear. Here we synthesize 15 Antarctic and 5 subarctic Pacific export productivity records over two glacial-interglacial cycles. We find that Antarctic and subarctic productivity generally tracked Antarctic temperature and atmospheric CO₂, but decoupled during Marine Isotope Stage 5, notably the Last Interglacial: high Antarctic productivity and CO₂ persisted despite reduced subarctic productivity and over 6 °C Antarctica cooling. We suggest that poleward-intensified SHW strengthened upwelling of Circumpolar Deep Water, sustaining Antarctic productivity and atmospheric CO₂, while altered Southern Ocean deep-water supply to the subarctic Pacific likely suppressed upwelling, limiting CO₂ release. These findings identify SHW as key in polar ocean dynamics and carbon cycling, with implications for future climate under projected SHW intensification.

Introduction

Physical and biological processes at high-latitude oceans play a crucial role in the global carbon cycle, influencing the uptake, storage, and transport of carbon, thereby modulating the Earth’s climate. In these regions, the Southern Ocean and the subarctic Pacific Ocean are particularly important^1,2. Here, wind-driven upwelling or density-driven overturning bring nutrient- and CO₂-rich deep waters to the surface, fueling phytoplankton growth and simultaneously releasing oceanic CO₂ to the atmosphere^3,4,5. Because phytoplankton incorporate nutrients and CO₂ into biomass during photosynthesis, the balance between upwelled CO₂ and the biologically fixed CO₂ directly affects the net air-sea CO₂ exchange. Thus, either an enhancement in export productivity or a reduction in vertical exchange would help to minimize the CO₂ emission, thereby contributing to low atmospheric CO₂ during glacial periods^6,7,8.

Paleoceanographic records reveal a simultaneous decline in export productivity in both the Antarctic zone of the Southern Ocean and the Subarctic Pacific 2.7 million years ago, coinciding with intensified Northern Hemisphere glaciation and global cooling³. Over the past 800 ka, glacial-interglacial variations of similar magnitude were observed in both regions^6,7,9, tracking Antarctic air temperature and atmospheric CO₂ concentration^10,11. The observed correlation among polar export productivity, Antarctic air temperature, and atmospheric CO₂ has garnered widespread debate over whether it reflects a mechanistic linkage^6,9 or mere coincidence³. Nonetheless, one point remains clear: the critical role of temperature in regulating polar ocean dynamics, ultimately affecting export productivity and global atmospheric CO₂, is consistently emphasized^5,6.

Marine Isotope Stage (MIS) 5 (128–71 ka B.P.) provides a well-documented interglacial interval for investigating ocean-carbon-climate interactions under elevated temperatures and atmospheric CO₂ conditions^11,12. During the Last Interglacial, the warmest interval (MIS 5e, 128–113 ka B.P¹³) in MIS 5, the Earth’s climate exhibited warming similar to future projections, yet atmospheric CO₂ remained relatively stable at 270–280 ppm from 125 to 113 ka B.P., despite pronounced Antarctic temperature variations^10,11,14,15. This asynchrony challenges the conventional relationship of a direct temperature-CO₂ coupling. Increasing evidence points to the role of Southern Hemisphere westerlies^4,16,17, which has been proposed to enhance deep-ocean upwelling and facilitate the oceanic carbon release on millennial timescales⁴. However, it remains unresolved whether the Southern Hemisphere westerlies played an equally, or perhaps more, influential role in oceanic vertical exchange and CO₂ regulation during the Last Interglacial. This uncertainty particularly concerns the mechanisms influencing atmospheric CO₂ controls during MIS 5d-5a, a period that followed the climatic optimum of the Last Interglacial but preceded the next glacial inception.

In this study, we examine the large-scale ocean-atmosphere processes influencing export productivity in the Antarctic zone and subarctic Pacific, with a focus on the role of Southern Hemisphere westerlies and their links to atmospheric CO₂. To assess regional differences in export productivity and carbon cycling, we synthesize 15 export productivity records from 12 sediment cores in the Antarctic zone and 5 export productivity records from 10 sediment cores in the subarctic Pacific (see Fig. 1 and Methods; Supplementary Tables 1–2). Our results reveal that during MIS 5, Antarctic and subarctic Pacific export productivity decoupled, breaking the long-term covariation with atmospheric CO₂. This decoupled evolution was most pronounced during the Last Interglacial, when Antarctic export productivity remained high and tracked atmospheric CO₂, while subarctic Pacific export productivity dropped significantly, even as Antarctic air temperature declined by over 6 °C. This asynchronous pattern suggests that poleward-intensified Southern Hemisphere westerlies, rather than Antarctic temperature alone, enhanced Circumpolar Deep Water upwelling, supporting high Antarctic export productivity and contributing to CO₂ rise.

Fig. 1: Modern nitrate distribution and ocean circulation in the Southern Ocean and subarctic Pacific.

a The distribution of nitrate in the global surface ocean, with the nitrate data from the 2013 annual World Ocean Atlas⁹⁹. The map was created using MATLAB. The orange dots are sediment sites for the compiled Antarctic export productivity (For details, see Supplementary Table 1). The light green dots are sediment sites for the compiled subarctic Pacific export productivity (For details, see Supplementary Table 2), while the additional yellow dots are sediment sites discussed in this study. The Juan de Fuca Ridge (JdFR) is a regional group comprising Sites AT26-19-05PC, 09PC/06MC, 12PC-TC/10MC, 35PC, 38PC, and 39BB. We use a JdFR productivity stack because productivity trends across these six sites are closely similar, allowing them to be combined into a single stack that represents the regional signal while minimizing redundancy among individual cores³⁵(Supplementary Table 2). The black dotted line indicates the Polar Front (PF), the orange line indicates the section in Fig. 1b; b Schematic of modern ocean circulation in the Southern Ocean and open subarctic Pacific, with the oxygen concentrations data from the 2013 annual World Ocean Atlas along the orange section¹⁰⁰, visualized using Ocean Data View¹⁰¹. AAIW Antarctic Intermediate Water, CDW Circumpolar Deep Water, AABW Antarctic Bottom Water, NPIW North Pacific Intermediate Water, PDW Pacific Deep Water.

This study offers an alternative view to the traditional emphasis on Antarctic temperature as the dominant factor controlling atmospheric CO₂ during past warm periods and underscores the critical role of Southern Hemisphere westerlies in influencing oceanic carbon cycling. Given projections of poleward-intensified westerlies under global warming, further investigation into their impact on polar ocean circulation and carbon cycles is crucial for understanding future climate change.

Results and discussion

Disrupted relationship between Antarctic and subarctic Pacific export productivity during MIS 5

Biogenic (or excess) barium, commonly referred to as bioBa, precipitates in seawater, and its flux to the sea floor is tightly linked to overall export productivity¹⁸. Likewise, the content and fluxes of biogenic opal have long been used as proxies for past export productivity, particularly in the diatom-rich Antarctic Zone of the Southern Ocean^4,19. However, biogenic opal accumulation can be influenced by various factors such as diatom species composition and iron availability^20,21, in addition to particle aggregation, differential dissolution, and post-depositional preservation effects^22,23,24,25. To mitigate potential biases from a single proxy and strengthen our export productivity reconstructions, we developed independent stacks for biogenic opal (BSi content and flux) and biogenic barium (flux, concentration, and elemental ratios (Ba/Ti, Ba/Al, and Ba/Fe)). A combined stack integrating both proxies was constructed to provide a more comprehensive and robust qualitative framework. Each of the three stacks was independently reconstructed using two distinct methods, resulting in two versions for each stack type (see Methods). To enhance the robustness of our compilation of the Southern Ocean export productivity, we applied a common age model based on the AICC 2012 age scale (see Methods; Supplementary Fig. 1). Chronological uncertainty (0-3.84 ka) is minor relative to the temporal scope of our analysis (Supplementary Tables 3–10).

Despite potential influences from local climate variability and differential preservation of biogenic opal and biogenic barium, the two versions of each stack exhibit exceptionally strong agreement, with Spearman’s correlation of ρ = 0.94 for the biogenic opal stacks, ρ = 0.98 for the biogenic barium stacks, and ρ = 0.97 for the combined stacks (Supplementary Figs. 2, 3). All stacks exhibit synchronous glacial-interglacial patterns (Fig. 2a, b, and Supplementary Fig. 2), broadly consistent with previously published biogenic opal and barium records from the Antarctic Zone of the Southern Ocean^26,27,28,29(Supplementary Figs. 4–8), indicating that they capture the true underlying export productivity signals, not proxy-specific artifacts. Independent palaeoecological data on the relative abundance of Fragilariopsis kerguelensis within the permanent open-ocean zone south of the Polar Front in our study area^30,31, which can be used as an indicator of productivity changes in the pelagic Southern Ocean, closely track the stacks (Supplementary Fig. 9), further supporting the interpretation that the reconstructed stacks capture key signals of Antarctic export productivity.

Fig. 2: Export productivity (EP) stack in the Antarctic zone of the Southern Ocean and Subarctic Pacific, and the associated paleoclimate over the past 160 ka.

a Antarctic air temperature stack (ATS)¹⁰; b EP inferred from biogenic opal stack, biogenic barium stack, (biogenic opal+ biogenic barium) stack and its confidence intervals (CI) of 95% in the Antarctic zone of the Southern Ocean (this study); c Subarctic Pacific EP stack and its confidence intervals of 95% (this study), EP at Site 882 inferred by Ba/Al in the subarctic Pacific^6,9 and ATS¹⁰; d Antarctic (biogenic opal+ biogenic barium) stack and compilation of the atmospheric CO₂ records¹¹. The dashed lines mark the Termination I (TI) and Termination II (TII) events.

To facilitate comparison with Antarctic records, we compiled five independent export productivity reconstructions from the subarctic Pacific, using biogenic barium concentrations^6,32,33,34 and sedimentary ²³¹Pa/²³⁰Th as proxies³⁵ and applying two stacking methods consistent with those used in Antarctic productivity (Fig. 1; see Methods and Supplementary Table 2). Unlike in the Southern Ocean, we excluded opal-based records because post-burial diagenetic alteration of biogenic silica can systematically underestimate subarctic Pacific productivity, as evidenced by discrepancies between opal and excess silica fluxes during older intervals (e.g., MIS 5)³⁵. For some sites, detailed age-model data are not publicly available, limiting the possibility of further improving the chronology. Therefore, we retained published age models to maintain consistency and minimize additional uncertainty in the stack construction (Supplementary Table 2).

The two subarctic Pacific productivity stacks constructed using two independent methods show remarkably consistent patterns, with a high correlation of 0.97 (Supplementary Fig. 10). The temporal patterns of the stacks are consistent with published biogenic barium and ²³¹Pa/²³⁰Th records from the subarctic Pacific^6,32,33,34, with correlation coefficients exceeding 0.7 (Supplementary Figs. 11–13), supporting the reliability of the stacks as representative proxies and enabling interhemispheric comparisons across the glacial-interglacial cycles. Given potential chronological uncertainties in the stacks, which may affect the interpretation of millennial-scale variability, we further incorporated the high-resolution, astronomically tuned Ba/Al record from ODP Site 882 (Hole 882 A) as a complementary dataset into our analyses⁶.

Generally, the Antarctic and subarctic Pacific export productivity stacks exhibit consistent patterns, broadly covarying with published Antarctic air temperature over the two glacial-interglacial cycles^6,9,10(Fig. 2a–c). Remarkably similar millennial-scale increases in export productivity were observed in both regions during Antarctic warming events³⁶, such as Termination I and II (Fig. 2a–c). However, during MIS 5, the Antarctic and subarctic Pacific records exhibited a pronounced deviation from their long-term covariation (Figs. 2a–c and 3b). In particular, during the late Last Interglacial (125–113 ka B.P.), subarctic Pacific export productivity declined alongside an abrupt ~6 °C drop in Antarctic air temperature, whereas Antarctic export productivity continued to rise post-Termination II and remained elevated^6,9,10 (Fig. 2a–c). To investigate this anomaly, we applied linear regression across the full record. The regression fit is moderate when all data are included (R² = 0.58, n = 149), but it improves markedly once MIS 5 is excluded (R² = 0.90, n = 92; Fig. 3b). Notably, the largest deviation occurred during the late Last Interglacial, around 115–113 ka B.P, after which the deviation gradually diminished as Antarctic productivity converged toward the declining subarctic Pacific trend, with deviation persisting until MIS 5a, when the two regional productivity stacks finally reconverged (Fig. 3b).

Fig. 3: Relationships among Antarctic and subarctic Pacific productivity, Antarctic temperature, and atmospheric CO₂.

a, b are the cross plot of the Antarctic (biogenic opal+ biogenic barium) stack (this study) versus Antarctic air temperature stack (ATS)¹⁰ and the subarctic Pacific export productivity (EP) stack (this study), respectively; c is the cross plot of ATS¹⁰ versus subarctic Pacific EP stack; d–f are cross plots of atmospheric CO₂ concentrations¹¹ versus ATS¹⁰, Antarctic (biogenic opal+ biogenic barium) stack and subarctic Pacific EP stack, respectively. TI Termination I, MIS Marine Isotope Stage, TII Termination II.

Wind-driven upwelling, not temperature, controls Antarctic export productivity

Unlike the pelagic subarctic Pacific, which is unaffected by sea ice and experiences higher boreal summer sea surface temperature³⁷, the Southern Ocean is highly influenced by variable sea ice extent. Several sediment cores used in this study are situated within the austral winter sea ice limit coverage, including Sites PC493, MD07-3134, U1537, U1361A, and PS1506-1^19,27,29,38. Sea ice, which limits sunlight vital for photosynthesis in phytoplankton, could potentially restrict export productivity³⁸. However, despite the regrowth of sea ice after Termination II, Antarctic export productivity remained high during the late Last Interglacial and exhibited three distinct high peaks throughout MIS 5, suggesting that sea ice did not significantly affect regional export productivity³⁹ (Fig. 4b and d). In the Subantarctic Zone, iron fertilization has been shown to enhance export productivity during ice ages, with increased deposition of iron-rich dust^8,40,41. In contrast, iron fertilization does not appear to have been impactful in the Antarctic Zone³⁸, which exhibits lower export productivity during glacial periods⁴⁰.

Fig. 4: Comparisons of export productivity (EP) between the Antarctic and the subarctic Pacific Ocean, and the associated paleoclimate during the Last Interglacial (LIG).

a Obliquity and daily mean top-of atmosphere insolation at 65°N and 65°S summer solstice; b Antarctic air temperature stack (ATS)¹⁰ and the sea ice indicated by the EPICA Dome C (EDC) ssNa flux record³⁹; c Subarctic Pacific EP stack (this study), Ba/Al at Site 882⁹, an indicator of EP, and the subarctic upwelling index, calculated as the normalized opal mass accumulation rate at Site U1343 minus the normalized δ¹⁵N difference between Sites U1343 and 1012⁶⁹; d (opal+ barium) stack in the Antarctic Zone of the Southern Ocean (this study), δ¹⁵N_db at Site MD12-3394¹⁷ and mean grain size of the sortable silt plus the fine sand fractions ($\overline{{{{\rm{SSFS}}}}}$, 10–125 μm) from Site PS97/085-3⁵⁴, a proxy for Antarctic Circumpolar Current (ACC) strength; e (opal+ barium) stack in the Antarctic Zone of the Southern Ocean (this study) and the Subtropical Convergence zone (STC) position in the Cape basin based on planktonic foraminifer data at Site GeoB-3603-2⁵⁶; f (opal+ barium) stack in the Antarctic Zone of the Southern Ocean (this study) and atmospheric CO₂¹¹. The dashed lines indicate the boundaries between the early and late phases of the Last Interglacial. MIS 6 Marine Isotope Stage 6, TII Termination II.

In the modern Southern Ocean, export productivity in the Antarctic Zone of the Southern Ocean is dominated by diatoms such as Fragilariopsis kerguelensis, with silica-rich frustules that facilitate the formation and export of particulate organic carbon from the surface ocean^30,31. The zone of high diatom productivity is largely located south of the Polar Front^22,25, a region characterized by upwelling driven by Southern Hemisphere westerlies or buoyancy from heat fluxes^4,5. Thus, the nutrient supply via upwelling is expected to play a crucial role in limiting the phytoplankton growth^4,28, and the export productivity is widely regarded as a key indicator of upwelling^4,26. Previous studies suggested that Antarctic air temperature plays an important role by directly affecting Southern Ocean surface buoyancy forcing over glacial-interglacial cycles⁵, which in turn influences upwelling intensity and nutrient supply for productivity. Hence, one would expect export productivity and Antarctic air temperature to exhibit a close relationship across the entire glacial-interglacial cycles¹⁰. We then tested this hypothesis using linear regression across the entire record, which yields a moderate fit (R² = 0.76, n = 149). Excluding MIS 5, however, results in a much stronger fit (R² = 0.90, n = 92; Fig. 3a), revealing that MIS 5 is an interval during which the relationship deviates from the long-term mean relationship. More importantly, during the late Last Interglacial, Antarctic export productivity remained elevated despite an abrupt 6 °C drop in Antarctic air temperature (Fig. 4b and d), suggesting that Antarctic air temperature cannot fully account for the variability in Antarctic export productivity.

The Southern Hemisphere westerlies are crucial for promoting upwelling, and their poleward shift is known to intensify northward Ekman transport, facilitating the upwelling of nutrient-rich deep waters^4,16. This mechanism is thought to influence export productivity over millennial timescales^4,26, and likely operated in the same way during MIS 5. During Termination II, massive freshwater discharge and surface cooling in the North Atlantic substantially weakened the Atlantic Meridional Overturning Circulation (AMOC)⁴², reducing northward heat transport and, via the bipolar seesaw^43,44, promoting heat accumulation in the Southern Ocean and Antarctic warming⁴⁵. This resulting interhemispheric temperature contrast, with the Southern Hemisphere warmer than the Northern Hemisphere, appears to have triggered a southward shift of the Intertropical Convergence Zone (ITCZ)⁴⁶ and poleward intensification of the Southern Hemisphere westerlies⁴, thereby enhancing Southern Ocean upwelling and surface productivity, consistent with the co-variation of Antarctic air temperature and export productivity¹⁰ (Fig. 4b and d). As the climate entered the early Last Interglacial, Antarctic air temperature records indicate peak temperatures¹⁰, while evidence suggests that AMOC had begun to recover⁴². However, proxy data show that surface waters in the North Atlantic, Labrador, and Nordic Seas remained cold and fresh, with reduced ventilation⁴⁷, indicating that this recovery was incomplete and spatially heterogeneous, with the Southern Hemisphere remaining warm while the high-latitude Northern Hemisphere stayed relatively cold. Thus, the persistent hemispheric temperature contrast would have continued to sustain a bipolar seesaw-like state, reinforcing Southern Ocean upwelling and export productivity during the early LIG.

Orbital forcing likely further amplified changes in Southern Hemisphere westerlies, thereby sustaining Southern Ocean upwelling and high export productivity during the late last Interglacial. Previous studies indicate that summer insolation peaked at 65°N around 127 ka B.P. before declining, while it was at a minimum at 65°S and increased thereafter⁴⁸ (Fig. 4a), creating an interhemispheric insolation contrast that amplified the meridional temperature gradient. Simultaneously, declining obliquity during the late Last Interglacial would have steepened the meridional temperature gradient (Fig. 4a), further reinforcing the Southern Hemisphere westerlies¹⁷. This reorganization of the wind system closely resembles a prolonged positive phase of the Southern Annular Mode (SAM), the dominant mode of atmospheric variability in the Southern Hemisphere^49,50,51. In modern observations and model simulations, a positive SAM phase is associated with enhanced upwelling of deep waters around Antarctica and reduced upper-ocean stratification^52,53. By analogy, such a positive SAM-like circulation during the late Last Interglacial would have likely intensified the upwelling of nutrient-rich Circumpolar Deep Water (CDW), brought nutrients to the surface, thereby promoting productivity. This interpretation is supported by consistent evidence of intensified Antarctic Circumpolar Current (ACC)^54,55, increased nitrate supply^17,28, and higher export productivity²⁸ in the Antarctic Zone of the Southern Ocean (Fig. 4d). Additionally, the significant southward shift of the Subtropical Convergence and ACC front provides further evidence of a more vigorous and expansive wind-driven circulation system during this period^56,57 (Fig. 4e). These obliquity-driven changes, by steepening the meridional temperature gradient and reinforcing the Southern Hemisphere westerlies, likely reached their maximum impact around 113 ka, when obliquity was at its minimum and Antarctic export productivity most evidently diverged from temperature (Fig. 3a). Their influence appears to have persisted until MIS 5a, when Antarctic temperature and export productivity had largely returned to the linear trends observed across glacial-interglacial cycles outside MIS 5 (Fig. 3a).

Remote influence of Southern Ocean circulation on subarctic pacific export productivity

Although a strong coupling in export productivity between the Antarctic and subarctic Pacific over most of the past 160 ka was established by previous studies^6,7 and our results (Fig. 2b, c), a significant weakening of this relationship during MIS 5 is identified. Notably, Antarctic export productivity remained high, likely due to intensified upwelling driven by poleward strengthened Southern Hemisphere westerlies, while a strong regression fit between subarctic Pacific export productivity and Antarctic air temperature was still observed (R² = 0.77, n = 149; Fig. 3c), implying a mechanistic connection or coincidence^3,6,9.

For the subarctic Pacific, local processes such as enhanced upper-ocean stratification due to surface warming or freshening, reduced winter mixing, and changes in sea-ice extent may have influenced nutrient resupply and thus productivity³⁵. However, our export productivity stack integrates sites spanning a wide range of hydrographic settings across both the western and eastern sectors, as well as marginal seas such as the Bering Sea. Some sites, such as pelagic ODP Site 882, were largely free from sea ice even during glacial times⁶, ruling out sea-ice cover as a major limitation for the observed productivity changes. Moreover, the co-variation between subarctic Pacific productivity (including Ba/Al at ODP Site 882) and Antarctic temperature is observed not only across interglacial-glacial cycles but also on millennial timescales during the Northern Hemisphere cooling⁶ (Fig. 2c). Given that the subarctic Pacific is unlikely to have been the primary driver of the climatic shifts documented in Antarctic ice cores, these co-variations may reflect a substantial influence of the Southern Hemisphere climate on subarctic Pacific biological productivity.

In the modern subarctic Pacific, where deep-water formation is absent⁵⁸, ventilation relies largely on the lateral supply of nutrient- and CO₂-rich abyssal water. This abyssal Pacific water, believed to comprise Antarctic Bottom Water (AABW) and recirculated North Atlantic Deep Water⁵⁹ (NADW) that is contained within Lower CDW (LCDW) advected from the Southern Ocean. Together, these water masses form Pacific Deep Water in deeper layers of the subarctic Pacific⁶⁰. Observations along the GP16 section of the East Pacific Rise indicate that west of 120°W, AABW contributes over 40% of the water column, while LCDW fractions vary patchily from 0 to 50% between 2000 and 5000 m⁶¹. Consequently, variations in AABW or LCDW export from the Southern Ocean could penetrate into the subarctic Pacific, potentially modulating nutrient availability and export productivity.

Compared with the glacial maxima (e.g., MIS 6), during rapid climate excursions that resulted in centennial to millennial-scale Northern Hemisphere cooling^62,63, particularly across Terminations I and II, Antarctic warming linked to the bipolar seesaw was accompanied by an enhanced export of AABW^64,65. Transient simulations support this rationale, indicating an acceleration of enhanced AABW intrusion to the subarctic Pacific⁶⁶. In parallel, sortable-silt and magnetic-grain records document a southward shift of the ACC frontal system and weaker flow along its northern boundary^67,68, likely limiting northward LCDW advection. The increased inflow of AABW, combined with reduced contributions of LCDW, promoted the accumulation of nutrient-rich deep water in the subarctic Pacific, and favored higher surface ocean productivity, as reflected in deglacial peaks in diatom productivity and Bio-Ba burial^6,9 (Fig. 4c). Moreover, an independent upwelling index, calculated as the normalized opal mass accumulation rate at Site U1343 minus the normalized δ¹⁵N difference between Sites U1343 and 1012⁶⁹ also indicates enhanced deep-ocean ventilation during these intervals (Fig. 4c). This state likely persisted into the early Last Interglacial, prior to the full recovery of AMOC, under the continued dominance of AABW in the abyssal Pacific water mass properties.

By the late Last Interglacial, following AMOC recovery and enhanced ventilation of NADW and surface waters^42,47, a strengthened upper overturning cell, driven by NADW and wind-induced CDW upwelling, coincided with a decline in AABW export^64,65. A concurrent poleward migration of the ACC further restricted LCDW access to the abyssal Pacific^67,68. This reorganization of deep-water sources reduced nutrient return to the Pacific surface and is consistent with the observed decline in productivity and deep-ocean ventilation^6,9,69 (Fig. 4c). Overall, while Pacific Deep Water integrates contributions from multiple deep-water sources, subarctic productivity and ventilation change aligns most closely with variations in AABW volume transport over glacial, deglacial, and interglacial intervals^6,9,64,65,66. Based on linear fits between Antarctic temperature and productivity, these observations suggest that Antarctic temperature may have remotely influenced subarctic Pacific ventilation and productivity primarily through its impact on AABW export. Yet, the precise mechanism remains to be elucidated.

Implications for ocean-atmosphere CO₂ exchange in polar oceans

Currently, upwelling processes in both the Southern Ocean and subarctic Pacific significantly influence the exchange of CO₂ between the ocean and atmosphere^3,4, a process that plays a pivotal role in regulating atmospheric CO₂ concentrations. A substantial amount of evidence suggests the Southern Ocean circulation’s key role in this regulation^5,6,7, particularly over glacial-interglacial timescales, during which atmospheric CO₂ levels closely correspond to Antarctic temperature⁷⁰, distinct from temperature trends in mid- to high-northern latitudes⁷¹. Our results are consistent with this general framework but emphasize the role of wind-driven Southern Ocean upwelling, rather than Antarctic temperature itself, in modulating ocean-atmosphere CO₂ exchange during MIS 5.

Compared to the Penultimate Glacial Maximum (e.g., MIS 6), the increased meridional thermal gradient between the hemispheres during Termination II triggered an intensification of the Southern Hemisphere westerlies^4,43,44, promoting vigorous Southern Ocean upwelling and CO₂ outgassing. Concurrently, enhanced AABW export coincided with stronger deep-ocean ventilation, supporting elevated CO₂ outgassing in the subarctic Pacific^6,9,69. This positive climate-carbon feedback amplified global warming and appears to have extended into the early interglacial, as reflected in the synchronous variability of Antarctic temperature, Antarctic and subarctic Pacific export productivity, and atmospheric CO₂ (Figs. 4b–d and 4f, 5a, b). Proxy records, however, reveal that this tight coupling broke down during the late Last Interglacial (125–113 ka B.P.), showing atmospheric CO₂ remained stable at ~270–280 ppm^11,14,15 despite a > 6 °C Antarctic cooling¹⁰ (Fig. 2a, d). This striking offset challenges the conventional view, supported by numerous studies, that higher Antarctic temperatures during interglacial periods are typically associated with higher atmospheric CO₂ levels^10,72. Our statistical analysis further reveals that the decoupling between Antarctic temperature and CO₂ is not confined to peak Last Interglacial but extends across the entire MIS 5, as evidenced by a moderate fit across the full record (R² = 0.82, n = 149) that strengthens markedly when MIS 5 data are excluded (R² = 0.92, n = 92; Fig. 3d). We therefore suggest that the conventional view of a tight Antarctic temperature-CO₂ coupling oversimplifies the processes at play during MIS 5. Interestingly, our reconstructed export productivity, strongly linked to Southern Ocean upwelling^4,26, exhibits a strong positive fit with atmospheric CO₂ concentrations over the two full glacial-interglacial cycles (R² = 0.92; n = 149; Fig. 3e), confirming the persistent role of Southern Ocean upwelling in influencing ocean-atmosphere CO₂ exchange.

Fig. 5: Schematic illustration of polar export productivity and associated ocean circulation.

The illustration depicts three key intervals: a Marine Isotope Stage (MIS) 6; b Termination II (TII) and early Last Interglacial; c Late part of the Last Interglacial. AAIW Antarctic Intermediate Water, CDW Circumpolar Deep Water, AABW Antarctic Bottom Water, NPIW North Pacific Intermediate Water, PDW Pacific Deep Water. Phytoplankton productivity (green dots). Southern Hemisphere westerlies (a circle with a dot in the center).

Notably, during warm periods, high export productivity in the Antarctic zone of the Southern Ocean always coincided with elevated atmospheric CO₂^6,7, even during the late Last Interglacial, when Antarctic temperatures dropped by more than 6 °C (Fig. 2a, b, and d). This apparent paradox of elevated export productivity coinciding with increased atmospheric CO₂ can be explained by an imbalance between upwelling supply and biological nutrient utilization^7,17, as supported by low δ¹⁵N_db records indicating high surface nitrate supply during interglacial periods^17,28. Vigorous upwelling may have continuously transported carbon- and nutrient-rich deep waters to the surface, exceeding the capacity of nutrients and CO₂ assimilation through photosynthesis⁷³. As a result, nutrient consumption was incomplete, constraining net biological CO₂ fixation and allowing a portion of the upwelled carbon to escape to the atmosphere, a phenomenon analogous to that observed in contemporary high-nutrient, low-chlorophyll (HNLC) regions of the Southern Ocean. Thus, an HNLC-style imbalance between supply and utilization likely persisted during the Last Interglacial, maintaining elevated atmospheric CO₂ even under high export productivity. This interpretation is further supported by the co-evolution of export productivity with increased nitrate supply^17,28 in the Antarctic zone of the Southern Ocean and rising CO₂ (Fig. 4d and f).

In contrast to the persistent role of Southern Ocean upwelling in influencing ocean-atmosphere CO₂ exchange, the subarctic Pacific shows a weakened coupling from CO₂ during MIS 5, as indicated by a moderate fit across the entire record (R² = 0.66, n = 149), and a substantially stronger fit after the exclusion of MIS 5 (R² = 0.88, n = 92; Fig. 3f). This decoupling parallels to the asynchronies observed among Antarctic versus subarctic Pacific export productivity, Antarctic export productivity versus Antarctic temperature, and Antarctic temperature versus atmospheric CO₂, with maximum deviation around 115–113 ka (Fig. 3a, b, d and f).

In the Southern Ocean, a positive SAM-like westerlies pattern likely intensified Circumpolar Deep Water upwelling, thereby sustaining CO₂ outgassing even as Antarctic temperatures declined (Fig. 5c). This interpretation is further supported by the co-evolution of export productivity with poleward-intensified Southern Hemisphere westerlies⁷⁴, a strengthened ACC⁵⁴, and rising atmospheric CO₂¹¹ (Fig. 4d and f). In contrast, ventilation in the subarctic Pacific appears to have weakened, as indicated by a low deep-ocean ventilation⁶⁹ (Fig. 4c), reducing the transfer of deep-ocean carbon to the surface (Fig. 5c). This inference is supported by the Ca/Al ratio at ODP Site 882, showing a lower deep ocean saturation state in the Subarctic Pacific⁶. Strengthened Southern Hemisphere westerlies during the late Last Interglacial enhanced Southern Ocean upwelling and carbon outgassing, likely forming positive climate-carbon feedback that may have initially amplified warming or prolonged the warm period. However, the stabilization of atmospheric CO₂ at ~270–280 ppm, along with Antarctic cooling, suggests this feedback was eventually limited, possibly due to a new equilibrium established in winds and ocean circulation, to the finite nature of readily ventilated deep-ocean carbon reservoirs, or adjustments in Earth’s orbital parameters.

Given that the Last Interglacial is a period warmer than the preindustrial by 1.0-1.5 °C with comparable atmospheric CO₂ concentrations, the contrasting responses between the Southern Ocean and the subarctic Pacific provide insights into how polar oceans may potentially respond to future CO₂ changes in warming scenarios. Climate models project a reduced but stable AMOC⁷⁵, alongside a poleward-downward spread of Circumpolar Deep Water⁷⁶ and a > 40% slowdown of Southern Ocean abyssal circulation driven by reduced AABW formation till 2050^76,77. Concurrently, climate simulations under scenarios of elevated greenhouse gas concentrations suggest a shift toward a more positive phase of the SAM, characterized by stronger, more poleward Southern Hemisphere westerlies⁷⁸. As Southern Hemisphere westerlies continue to intensify^79,80, upwelling in the Southern Ocean is expected to strengthen^76,81, contributing to higher atmospheric CO₂, amplifying and prolonging global warmth. In contrast, a slowdown of AABW, together with a concurrent poleward migration of the ACC that further restricts LCDW access to the abyssal Pacific, would be likely linked to reduced deep-water ventilation in the subarctic Pacific and limit ocean-atmosphere CO₂ exchange. This future scenario, although projected to unfold over relatively short timescales, may be analogous to conditions during the Last Interglacial. Such potential parallels offer valuable insights into how changes in polar ocean-atmosphere interactions, driven by climate factors such as Southern Hemisphere westerlies and temperature variations, may evolve and subsequently affect atmospheric CO₂ under warming scenarios.

Methods

Core material in the Southern Ocean

In this study, we have compiled 15 export productivity proxy records from 12 marine sediment cores from the Antarctic Zone of the Southern Ocean. These sediment cores include MD11-3353, PS75/083-1, PS1768-8, MD07-3134, U1537, TN057-14PC4, ODP 1094, PS75-072-4, PS1506-1, PC493, U1361A, and MD12-3394 (Supplementary Table 1). Notably, some of these cores contain multiple reconstructions of export productivity employing diverse methodologies. The data span an extensive period of 160 ka here, covering two glacial-interglacial cycles. Among the 15 proxy records, MD07-3134 and TN057-14PC, though not covering the LIG interval targeted in this study (spanning 92.5-0 ka B.P. and 80.6-6.6 ka B.P., respectively), provide detailed Holocene and millennial-scale productivity data, allowing for the compilation of a more comprehensive and robust Antarctic productivity stack. The sediment cores are situated between 50°S and 72°S, retrieved from an average water depth of 2976 m, with depths ranging from 1568 to 3713 m. Sites MD11-3353, PS1768-8, PS75/83-1, PS75-072-4, PS75/83-1, TNO57-14PC, ODP 1094, and MD12-3394 are located in the Antarctic Zone of the Southern Ocean, north of the maximum winter sea ice line^{4,7,17,28,82,83,84}. Sites MD07-3134, U1537, U1361A, and PS1506-1 are located within the winter sea ice limit, north of the modern maximum summer sea-ice line⁸⁵, while Site PC493 is located close to the modern maximum summer sea ice limit²⁹.

Age model of Southern Ocean sediment cores

Among the 15 export productivity proxy records, proxy records from Sites PS75/83-1, U1537, U1361A, and MD12-3394 are reconstructed using the AICC 2012 age scale^{17,26,82,86,87}. In order to enhance the robustness of our compilation of the Southern Ocean export productivity, a common age model was applied. Thus, we have refined certain published age models, ensuring consistency across all proxy records using the AICC 2012 age scale. Specifically, the initially published chronology for Site MD11-3353 relied on the alignment of sea surface temperatures (SST) with the EPICA Dome C (EDC) δD record^72,84. We have subsequently revised this age model on the AICC 2012 age scale⁸⁸. Similarly, the chronology for Site PS1768-8 was refined through the correlation of SST to EDC δD record, also on the AICC 2012 age scale^72,88,89. For Site MD07-3134 age scale, we enhanced and adjusted additional magnetic susceptibility (MS) of tie points to achieve the correlation to EDC dust records and East Antarctic Dronning Maud Land (EDML) nssCa²⁺ flux records on the AICC 2012 age scale^39,40,88,90. We refined the chronology for Site TN057-14PC based on the alignment of the Eucampia Antarctica record with the EDC δD record on the AICC 2012 age scale^4,88. The originally published chronology for ODP Site 1094 relied on the graphical alignment of N. pachyderma (s.) δ¹⁸Ο record with the EDC δD record^7,72. We have updated this age model on the AICC 2012 age scale⁸⁸. The originally published chronology for Site PS75-072-4 was established through the AMS ¹⁴C and correlation of N. pachyderma (s.) δ¹⁸Ο record to the EDC δD record^28,72. We have also refined this age model on the AICC 2012 age scale⁸⁸. The originally published chronology for Site PS1506-1 was based on the correlation of N. pachyderma (s.) δ¹⁸Ο record to the EDC δD record^27,72, we refined this age model on the AICC 2012 age scale⁸⁸. The originally published age model for Site PC493 was established via tuning of both δ¹⁸O_p and δ¹⁸O_cib data to the LR04 global δ¹⁸O_cib stack^29,91. We adjusted this age model by the alignment of both δ¹⁸O_p and δ¹⁸O_cib data with the EDC δD record on the AICC 2012 age scale^72,88. Additionally, we have incorporated estimates of age uncertainty. The chronological uncertainty associated with these records ranges from 0 to 3.84 ka, which is small in comparison to the temporal scope of our focused analysis. More details can be seen in Supplementary Fig. 1 and Supplementary Tables 1-9.

Southern Ocean export productivity reconstructions

Biogenic barium serves as a proxy for the export of organic matter from the surface ocean to the seafloor, providing a direct measure of export productivity¹⁸. In the Antarctic Zone of the Southern Ocean, diatom-derived biogenic opal offers an additional record of past productivity changes^4,19. Thus, we have compiled data on both biogenic opal and biogenic barium proxy records, including the opal content, opal flux, barium flux and concentration, and various barium ratios such as Ba/Ti, Ba/Al, and Ba/Fe. We have developed a biogenic opal stack and a biogenic barium stack, each serving as a proxy for export productivity. Given the distinct nature of the biogenic opal and biogenic barium proxies—derived from different indicators such as opal content and opal flux for the opal stack, and barium concentration along with various barium ratios like Ba/Ti, Ba/Al, and Ba/Fe for the barium stack—these proxies are inherently independent of each other. Consequently, the combined use (biogenic opal+biogenic barium stack) provides a more comprehensive and robust qualitative framework to investigate past changes in export productivity in the Antarctic Zone of the Southern Ocean. Specifically, the biogenic opal stack incorporates opal content at Sites MD07-3134³⁸, U1537 (biogenic opal content is calculated by the correlation of Si/Al and biogenic opal content¹⁹), PS75/083-1⁸², PS1768-8⁸³, and MD11-3353⁸⁴, as well as the opal flux data at Sites TN057-14PC⁴, ODP 1094⁷, and PS75-072-4²⁸. Meanwhile, the biogenic barium stack incorporates barium concentrations from Site PS1506-1⁹², barium flux data from Site PS1768-8⁹³, Ba/Ti ratio from Site PC493²⁹, Ba/Al ratio from Site U1361A⁸⁶, and Ba/Fe ratios from Sites ODP 1094⁷, MD12-3394^17, and PS75-072-4²⁸. Collectively, these 15 proxy records have been integrated into a comprehensive combined (biogenic opal+biogenic barium) stack.

To develop biogenic opal, biogenic barium, and a combined (biogenic opal+biogenic barium) stack, we employed two different methods: the inverse standard deviation weighting method and principal component analysis (PCA). The integration of these approaches enhances our data reliability and reveals key patterns, leading to more robust and interpretable results.

The inverse standard deviation weighting approach is based on a well-established principle in quantitative data analysis, where variables are weighted inversely proportional to their variability (standard deviation). Such weighting approaches ensure more stable records contribute more strongly to the composite stack, thereby minimizing the influence of noisy records. To prepare the data for this analysis, we first resampled all 15 records at a 1 ka resolution, ensuring continuity in the data and filling the missing values. Following this refinement, each record was then normalized using the Max-Min normalization, defined as X_norm = (X_sample – X_min)/(X_max–X_min). This transformation rescales the data to the range [0, 1], allowing for direct comparison of the magnitude and timing of changes in export productivity across different records. It also reduces biases in weight calculations caused by differences in data scale and improves the stability and consistency of the resulting weight distribution. Recognizing that the degree of dispersion directly reflects data stability, where a higher standard deviation corresponds to a greater fluctuation or dispersion, we calculated the standard deviation for each normalized record to quantify variability. Weight coefficients were then assigned as the reciprocal of the standard deviation of each normalized record (w_i = 1/σ_i), so that more stable and reliable records contribute more to the final stack. The weights were then normalized to sum to 1, and weighted averages were calculated for the biogenic silica stack, the biogenic barium stack, and their combined stack (biogenic silica + biogenic barium). The result of biogenic opal, biogenic barium, and a combined (biogenic opal+biogenic barium) stack can be seen in Fig. 2, Supplementary Fig. 2 and Supplementary Data 1⁹⁴. The correlation of biogenic opal, biogenic barium, and a combined (biogenic opal+biogenic barium) stack with the associated export productivity records at different sites in the Antarctic Zone of the Southern Ocean can be seen in Supplementary Fig. 4.

PCA was employed to identify the main trends in the data, revealing the underlying patterns and relationships among the proxy records. This analysis was conducted using PAST v4.07 software⁹⁵. First, we standardized all 15 export productivity records. This normalization process is crucial as it mitigates the influence of variables with larger ranges, allowing for a more accurate comparison of patterns across the proxy records. The individual normalized records were then resampled at a 1 ka resolution. Since many records did not span the full period of 0-160 ka, we addressed the missing values by employing mean value imputation. Following these steps, we performed the PCA analysis to extract the principal components. The result of principal components on related export productivity records is shown in Supplementary Figs. 2, 5, 6⁹⁴.

The PCA of the biogenic opal records, biogenic barium records, and the combined biogenic opal and biogenic barium records reveal a dominant trend in export productivity represented by the first principal component (PC1). This component also displayed high values during interglacial and lower values during glacials (Supplementary Fig. 2). PC1 of the biogenic opal records explains 53.614% of the total variance, while PC2 accounts for only 16.202% (Supplementary Fig. 5a). For the biogenic barium records, PC1 explains 68.203% of the total variance, with PC2 contributing just 11.496% (Supplementary Fig. 5b). In the combined records, PC1 accounts for 57.64% of the total variance, while PC2 explains only 11.654% (Supplementary Fig. 5c). Notably, all principal components except PC1 fall below the broken stick curve (Supplementary Fig. 5), indicating that only PC1 appears to carry enough signal for interpretation. Additionally, strong correlations between PC1 and the various variables further support the significance of PC1 in capturing the underlying dynamics of the dataset (Supplementary Fig. 6).

The most significant observation is that PC1 of the biogenic opal records, biogenic barium records, and their combined counterpart showed a high degree of overlap with the respective stacks generated using the inverse standard deviation weighting method (Supplementary Fig. 2). Given that the data are not normally distributed, Spearman’s correlation was applied, showing ρ values exceeding 0.94 across all three stacks (Supplementary Fig. 3). Importantly, both PC1 and the respective stacks correlate strongly with previously published biogenic opal and biogenic barium records in the Antarctic Zone of the Southern Ocean^26,27,28,29 (Supplementary Figs. 4, 6–8). This consistency in findings enhances the reliability of the data, suggesting that both the PCA and inverse standard deviation weighting method capture the Antarctic export productivity patterns.

Subarctic Pacific export productivity stack

To facilitate comparison with the Southern Ocean records, we have additionally compiled 5 export productivity proxy records from 10 marine sediment cores located in the subarctic North Pacific using two independent methods consistent with those applied to Antarctic records, allowing direct comparison across datasets. These cores include ODP 882⁶, MR98-05-3PC³⁴, JPC17³³, ODP887³², and a group of six closely spaced cores from the Juan de Fuca Ridge (JdFR) area³⁵: AT26-19-05PC, AT26-19-09PC/06MC, AT26-19-12PC-TC/10MC, AT26-19-35PC, AT26-19-38PC, and AT26-19-39BB (Supplementary Table 2). Spatially, these cores are distributed between 44.96°N and 54.22°N, and from 164.98°E to 49.39°W, spanning a range of hydrographic settings across the subarctic Pacific, including both the western and eastern subarctic sectors and marginal seas such as the Bering Sea, with water depths ranging from 2209 to 5507 m. Given the geographic proximity of the JdFR cores, which exhibit similar productivity signals and trends³⁵, we adopted an existing regional productivity stack derived from these six sites, which provides a continuous record over the past 160 ka and serves as a representative index for this sector³⁵. This approach ensures better temporal continuity and reduces redundancy in overlapping signals. Combined with records from other subarctic locations including JPC17 (118.8–0.6 ka)³³, MR98-05-3PC (142–3.5 ka) ³⁴, ODP 882 (149.7–8.7 ka)⁶, and ODP 887 (160-0 ka)³², these data enable the construction of a subarctic North Pacific productivity stack covering at least two full glacial-interglacial cycles.

Similar to the Southern Ocean, biogenic barium is widely used as a paleoceanographic proxy for export production in the subarctic Pacific. However, opal records from the subarctic Pacific may underestimate past productivity due to diagenetic alteration of biogenic silica after burial. This is evidenced by the observed divergences between opal and excess silica fraction fluxes during older intervals (e.g., MIS 5), when excess silica suggests higher productivity than opal³⁵. In addition, although sedimentary ²³¹Pa/²³⁰Th in the modern Pacific Ocean is largely influenced by opal scavenging⁹⁶, it is relatively insensitive to preservation and diagenetic effects^97,98, making it a robust tracer for reconstructing past changes in productivity³⁵. Therefore, to reconstruct export productivity stack in the subarctic Pacific, we compiled records of both biogenic barium and ²³¹Pa/²³⁰Th applying methods consistent with those used for the Antarctic records.

All sediment cores included in the compilation are well dated using well-established age models, primarily based on radiocarbon dating, magnetostratigraphy, biostratigraphy, and astronomical tuning (Supplementary Table 2). For some sites, detailed age-model data are not publicly available, so we used the original published age models to maintain consistency and avoid introducing additional uncertainties in the stack construction. While some spatial variability in export productivity exists among subarctic North Pacific sites, the stack constructed from multiple cores using two independent methods displays consistent patterns (Supplementary Data 2⁹⁴). The PCA of the subarctic Pacific records shows that the first principal component (PC1) captures the dominant trend in export productivity (EP), explaining 61.3% of the total variance, whereas the second component (PC2) explains only 14.8% (Supplementary Fig. 10). This pattern closely aligns with the stack generated using the inverse standard deviation weighting method, as well as the previously published biogenic barium and ²³¹Pa/²³⁰Th records from the subarctic Pacific^{6,32,33,34,35} (Supplementary Fig. 10). To quantify this relationship, we calculated Pearson’s correlation coefficients, given that the subarctic data generally follow a normal distribution. The two stacks exhibit a strong correlation of 0.97 and maintain correlations above 0.7 with all previously published individual biogenic barium and ²³¹Pa/²³⁰Th records. This consistency supports the reliability of the stacks as representative proxies, thereby enabling robust interhemispheric comparisons across glacial-interglacial cycles.