Introduction

The last deglacial warming has been typically attributed to the retreat of icebergs over the northern high latitudes due to the increasing summer insolation1. However, it was also demonstrated that sea surface temperature (SST) changes in the southern hemisphere and Indo-Pacific warm pool (IPWP) preceded northern hemisphere and global ice volume variations2,3,4,5, and that the southern hemisphere itself could induce global warming directly through its response to orbital forcing2,6. Furthermore, the ocean-atmospheric teleconnections between the tropical Pacific and Southern Ocean have a profound impact on global climate6,7. However, the role of the IPWP, the world’s greatest heat engine8, and its thermal linkages with the southern high latitudes during the last deglaciation are still ambiguous.

The El Niño-Southern Oscillation (ENSO) affects the heat distribution in the equatorial Pacific and is a major force behind the teleconnections influencing global climate8. This inherently unstable oceanic–atmospheric oscillation between positive (El Niño) and negative (La Niña) phases exhibits a periodicity of 2 to 8 years9. As a result, changes in the zonal equatorial temperature gradient across the Pacific have been used as a proxy for ENSO-like variations10,11. However, the ENSO evolution since the last glacial epoch is controversial owing to divergent SST changes in the equatorial Pacific under heterogeneous hydrological conditions. The last glacial maximum (LGM) SST cooling in the IPWP has a range of 2–5 °C based on geochemical proxies10,12,13, while cooling in the eastern equatorial Pacific (EEP) cold tongue varies between 0.3 and 3.5 °C14,15. Some previous ENSO investigations included cores that were distant from the central western Pacific warm pool (WPWP)16,17, or in areas close to shore that were impacted by coastal upwelling, fresh water runoff and coastal currents13,18,19. Other ENSO investigations were based on the meridional SST gradient between the EEP cold tongue and the northeastern Pacific warm pool, which is under debate as an indication of ENSO or seasonality assessment20,21. These discrepancies ultimately result in opposing assertions about the prevalence of La Niña-like10,13,17,19,20 or El Niño-like conditions5,11,12,14,16 in the last glacial period.

In this study, we integrated 340 published SST data sets (planktonic foraminiferal Mg/Ca and Uk’37 alkenone index) spanning the last 30 kyr from the tropical and subtropical Pacific Ocean and eastern Indian Ocean (Fig. 1) to gain a comprehensive spatial and temporal overview of thermal variability: (1) We quantified the variations in the boundaries of the WPWP, the warmest region within the IPWP, by analyzing the temporal SST distribution in the IPWP. (2) We monitored ENSO-like transitions since the last glacial period by comparing SST anomaly (SSTA) stacks between the warm pool core area and the EEP cold tongue (Fig. 1). (3) We spatially elucidated the thermal linkage of the tropical Pacific to the Southern Ocean and their respective role during deglacial warming, by combining a new ocean heat content (OHC) record (based on Mg/Ca ratio of five planktonic foraminiferal species from core KX22-4 in the WPWP, Supplementary Material) and seven published subsurface temperature (subT) records in the Pacific (Fig. 1).

Fig. 1: Locations of published sea surface temperature (dots) and subsurface temperature (yellow stars) records from the Indo-Pacific region referred to in this study.
figure 1

Red star marks the ocean heat content record reconstructed from core KX22-4 in this study. The base map shows modern annual mean sea surface temperature distribution drawn with Ocean Data View software83 based on the World Ocean Atlas 2018 dataset84. The isothermals of 28 and 24 °C delineate the modern extents of the Indo-Pacific warm pool (IPWP) and eastern Pacific cold tongue. The dashed rectangles mark the East China Sea (ECS), South China Sea (SCS), Eastern Indian Ocean (EIO), west and east part of the western Pacific warm pool (WWP and EWP), central and eastern equatorial Pacific (CEP and EEP), southwest, south-central and southeast Pacific (SWP, SCP, and SEP), respectively. The white rectangles mark the areas within the Indo-Pacific warm pool region, all areas are classified based on cluster analysis (Supplementary Fig. 1). The insert map shows the areas considered for El Niño-Southern Oscillation-like investigations (blue boxes).

Full size image

Results and discussion

ENSO-like variations of the IPWP thermal state since 30 ka

The ENSO-like variations depicted by the ΔSST between the WPWP and EEP since the last glacial period are shown in Fig. 2d. The glacial zonal ΔSST was considerably smaller (0.67 ± 0.24 °C) than the modern annual mean ΔSST (Fig. 2d), suggesting a more El Niño-like state during the last glacial. This is also supported by transient simulation (TRACE) results, which display 0.82 ± 0.27 °C less glacial cooling in the EEP than in the WPWP, in contrast to the early Holocene (5–10 ka) (Supplementary Fig. 2)22. This asymmetric cooling may be linked to the inhibited upwelling of subsurface waters in the cold tongue region, ascribed to the deeper thermocline in the EEP and reduced Walker circulation12,23. With a 2.41 ± 0.63 °C cooling during the glacial (Supplementary Table 1), the WPWP migrated slightly toward the equator24 (Fig. 3a, c) and its latitudinal extent was reduced by about 4° compared to the Holocene (Fig. 3d and Supplementary Figs. 3, 4), meanwhile its west boundary shrank markedly eastward to ~150°E (Fig. 3d and Supplementary Figs. 3, 4). A prevailing weaker and eastward tropical Pacific Walker circulation25 in the glacial consequently contributed to a reduced northward heat transfer through the weakened Kuroshio Current26 and the cooling of Indonesian seas by the weakened Indonesian Throughflow19.

Fig. 2: Thermal variations of the Indo-Pacific warm pool and El Niño-Southern Oscillation-like evolution since 30 ka.
figure 2

a Variations of planetary radiative imbalance (PRI)61. b 0–300 m ocean heat content (OHC) record from core KX22-4 in this study (Supplementary Fig. 17). c El Niño events recorded in lake Laguna Pallcacocha27 in the eastern equatorial Pacific. d Annual zonal sea surface temperature difference between the central western Pacific warm pool (WPWP) and eastern equatorial Pacific (EEP) (red) and its standard deviation (red shade), the dotted horizon line represents the modern annual mean sea surface temperature difference based on the World Ocean Atlas 2018 sea surface temperature data, and black solid line denotes orbital precession85. e Annual mean sea surface temperature anomaly (SSTA) stacks in the central western Pacific warm pool (red, Mg/Ca) and eastern equatorial Pacific (blue, Mg/Ca and Uk’37, Supplementary Fig. 5) computed using a sliding rectangular window of 1 kyr width by STATNARY 1.2, as well as the late winter insolation at 30°S (black). f Oceanic ΔpCO2 record from Ocean Drilling Program Site 1238 in the eastern equatorial Pacific32. g–i Zonal and meridional heat transfer (sea surface temperature gradient) between the eastern part of the warm pool (EWP) and East China Sea (ECS), the eastern part of the warm pool and eastern Indian Ocean (EIO), the equatorial Pacific (Eq) and South Pacific (Supplementary Fig. 9). Yellow and blue bars indicate the early Holocene (EH) and last glacial (LG), respectively. The vertical dash lines denote the precession maximum and minimum at ~22 and ~12 ka, respectively.

Full size image
Fig. 3: Variations in the spatial extent of the western Pacific warm pool since 30 ka.
figure 3

a Comparisons of meridional and b zonal variations in the extent of the western Pacific warm pool (range of top 1 °C) in the last glacial (blue) and early Holocene (red), based on sea surface temperature distributions. Trend lines are orthogonal polynomial fittings (OPF) based on Mg/Ca (solid) and mixed temperature proxies (Mg/Ca and Uk’37, dashed), with 90% confidence (shadows). Gray lines are discrepancies between the last glacial and early Holocene. Mg/Ca results are shown by circles and plus signs, while Uk’37 results are represented by diamonds and crosses. The spatial extent of the western Pacific warm pool marked by dotted boxes is based on sea surface temperature data from Mg/Ca, which are more uniformly distributed compared to sea surface temperature based on Uk’37 (Supplementary Fig. 10). Mg/Ca based sea surface temperature distribution did not differ much from annual sea surface temperature distribution (Uk’37 included) in the western Pacific warm pool region with no discernible seasonal variability (−0.07 ± 0.76 °C), divergence was more marked at higher latitude owing to the data weighted towards sea surface temperature based on Uk’37 (Supplementary Fig. 7). Please refer to the supplementary material for a discussion of the seasonality of proxies. c Meridional displacements of the north and south boundaries and location of maximum sea surface temperature within the western Pacific warm pool. d Variations in the latitudinal extent and west boundary of the western Pacific warm pool.

Full size image

In contrast, the meridional extent of the WPWP in the early Holocene approximated from 19°N to 14°S (Fig. 3a, c), consistent with the present 28 °C isotherm (Fig. 1), and its west boundary in the early Holocene was at ~120°E (Fig. 3b, d). As evidenced by the precipitation records from lacustrine sediments27,28 and modeling studies29,30, a gradually increased frequency of El Niño-like events occurred over the late Holocene (Fig. 2c, d). The attenuated Walker circulation caused a decline in tropical heat export from the IPWP via the Indonesian Throughflow to the Leeuwin Current offshore western Australia from the mid to late Holocene31. Besides, this period was characterized by a reduction in thermocline nutrients and CO2 leakage in the EEP32,33. Because El Niño-like conditions during the Holocene were not as pronounced as during the last glacial, this relatively mild El Niño-like tendency would have also contributed to CO2 release into the atmosphere.

During the deglaciation, more intense warming in the WPWP than in the EEP resulted in a larger zonal ΔSST contrast and a more La Niña-like state (Fig. 2d, e and Supplementary Fig. 5). Results of the WPWP spatial development from TRACE reveal that the WPWP first reached 28 °C at ~14 ka (Supplementary Fig. 6) during the Bølling-Allerød Warm interval22, and then remained above 28 °C since ~12 ka (Supplementary Fig. 6) following the Younger Dryas cold event, in phase with the precession minimum (Fig. 2d). Meanwhile, the extent of the WPWP began to expand substantially at ~12 ka when La Niña-like conditions peaked, the WPWP OHC also reach highest values (Fig. 2b, d and Fig. 3d). This thermal expansion was in line with a record from the East China Sea which showed that warm water diatom species increased since 12 ka34.

The more intense deglacial warming in the WPWP may be attributed to the presence of a barrier layer, a feature well-recognized in modern observations and models35,36. The barrier layer forms beneath the fresh surface water and at the base of the mixed layer in the WPWP due to the vigorous atmospheric deep convection there35. It can effectively restrict the entrainment of the cold thermocline water to the mixed layer, favoring heat built up at the sea surface35,37,38,39. Thus, the barrier layer along the equatorial Pacific is intimately linked to ENSO activity36,39,40,41. During the deglaciation, the strong atmospheric convection in the WPWP10,42 benefited the development of the barrier layer and contributed to the buildup of surface heat. To explore this further, we examined the Mg/Ca-SST and residual seawater δ18O (δ18Osw-iv) data of Globigerinoides ruber and Trilobatus sacculifer from core KX22-4 in the east part of the warm pool (Supplementary Fig. 7). The SST contrast between these two species somewhat decreased in the deglaciation (remained about 0.5 °C), which means that T. sacculifer inhabits the mixed layer as G. ruber. While the δ18Osw-iv contrast between them dramatically rose due to increased δ18Osw-iv of T. sacculifer, which implies that T. sacculifer resides in waters that are much saltier than G. ruber, probably within the barrier layer that contributed to the prevalence of La Niña-like conditions37,38,43.

Deglacial coupling between the IPWP and Southern Ocean

Both atmospheric and oceanic processes are involved in the teleconnections between the tropical Pacific and Southern Ocean44,45. Data reanalysis and modeling experiments show that the Southern Ocean heat uptake alters the meridional circulation and is accompanied by anomalous easterlies, resulting in a zonally asymmetric La Niña-like pattern of SST change through atmospheric process42,46,47. In this study, we focus on the oceanic tunnel between the IPWP and the Southern Ocean. The Southern Hemisphere supplies up to 70% to the water masses in the equatorial undercurrent48 and serves as an external thermal source to discharge/recharge the tropical ocean49. According to observations and simulations, the shallow meridional overturning circulation/subtropical overturning cells connect the extratropical and tropical oceans. These processes include the subduction of subtropical water, equatorward geostrophic flow in the thermocline to the equatorial Pacific either through its interior or western boundary pathway, and upwelling at the equator48,50,51,52,53 (Fig. 4g). Subduction of South Pacific waters includes Subtropical Mode Water, Subantarctic Mode Water and Antarctic Intermediate Water51,54. Subantarctic Mode Water and Antarctic Intermediate Water formed in the Southern Ocean are transported northward into the adjacent subtropical gyres55.

Fig. 4: Hypothesized spatial heat transfer process in the Pacific during the last deglaciation.
figure 4

a–f Variations of sea surface temperature (SST), subsurface temperature (subT) and their vertical gradient since 30 ka along the path of the Kuroshio Current (KC) in the East China Sea (ECS)17,86, along the path of the Indonesian Throughflow (mirror the variation in the eastern Indian Ocean: EIO)63, in the eastern equatorial Pacific (EEP)64, in the east part of western Pacific warm pool (EWP), in the southwest Pacific (SWP)87,88,89 and south-central Pacific (SCP), respectively, as shown on the map. g The bold sea surface temperature, subsurface temperature and their temperature difference (ΔT) lines are polynomial fitting results for the raw data (gray), the P value of the polynomial fitting for ΔT is 0.0003 for East China Sea and <0.0001 for all the other regions. The inverted triangles indicate 14C-dated points. The black downward arrows mark the start of subsurface temperature warming. g TRACE modeled temperature anomaly in the deglaciation relative to glacial period (15 minus 20 ka). The red dashed arrows in the upper panel indicate surface heat transfer from the western Pacific warm pool. The sun marked warming in the south Pacific subduction area. Two vertical sections marked by dotted lines are shown in the lower panel, a hypothesized oceanic tunnel of heat transfer (red arrow) between the Southern Ocean and equatorial Pacific is sketched out.

Full size image

Around 75% of the heat absorbed by the ocean over the historical period has come from the Southern Ocean, north of the Antarctic Circumpolar Current56,57, and heat is preferentially stored where surface waters are subducted to the north58. Modern heat gained by the Southern Ocean south of 30°S57 is carried through the subpolar-subtropical gyre toward the equator and plays an important role in ventilating the equatorial thermocline, on which ENSO variability heavily depends49,50,59,60. The overturning circulation causes the eastern edge of the warm pool to move westward during La Niña events. This alters the thermocline structure and surface stratification, which in turn causes the thick barrier layer to move back and forth along the equator and is thought to have an impact on ENSO evolution53. When more heat entered the ocean with increased insolation61 as shown by the raised planetary radiative imbalance in the late last glacial epoch (Fig. 2a), a weaker Atlantic Meridional Overturning Circulation in a cold climate period would cause greater heat to accumulate in the Southern Ocean by impeding heat transfer into the boreal high latitudes62. As the excessive heat in the southern hemisphere was transported equatorward into the interior of the ocean by subduction58, subT of the eastern part of the WPWP rose earlier than the SST at ~22 ka (Fig. 4d). This is consistent with a previous study, which showed that subT warming began about 4000 years before the deglacial pCO2 and SST rise in the IPWP63. An earlier increase in WPWP subT and a decrease in the vertical temperature gradient also occurred in the south Pacific (Fig. 4e, f), in contrast to the delayed subT warming and the increase in the vertical temperature gradient in the East China Sea and eastern Indian Ocean regions (Fig. 4a, b). This earlier increase probably supports a thermocline linkage between the WPWP and Southern Ocean during the deglaciation. Furthermore, the temperatures of the EEP cold tongue, which is fed by the equatorial undercurrent from the western Pacific, also rose earlier64,65 (Fig. 4c). Evidences from Δ14C also indicated a Southern Ocean source for the waters in the EEP and even eastern North Pacific (23.5°N) via the equatorial undercurrent66,67,68 during the deglaciation. The results of TRACE showed that the upper water column of the WPWP absorbed heat from the south and lower waters, and released mainly to the north (Supplementary Fig. 8) in the deglacial warming, and the SST gradient between the equatorial and south Pacific reduced (Fig. 2i).

A continuous subduction of the extratropical warm anomaly to the WPWP thermocline is hypothesized to affect ENSO-like transitions since 360 ka, paced by orbital precession69. The results of earlier modeling simulations have shown that orbital precession exerts a major impact on ENSO-like processes through the strength of ocean-atmosphere feedbacks70,71. Precessional forcing on the subtropical South Pacific affects the stratification of the tropical Pacific, altering the strength of ocean upwelling and the thermocline tilt to the surface wind anomalies70. When the precessional parameter decreased in the deglaciation (Fig. 2d), the overturning circulation that mainly formed in winter by intense vertical mixing72 was warmed due to the rising insolation in austral winter (Fig. 2e), and this warmer signal was further transferred to the western Pacific69. Preceding thermocline anomalies in the western and central deep tropical Pacific contribute to the recharge phase of ENSO63,73, and modern observations also indicate that the change in the WPWP OHC is tightly related to ENSO activity74. During the deglaciation, rise in WPWP subT, OHC, as well as ENSO-like variations were quite correlated with the decrease in the precession parameter (Fig. 2b, d and Fig. 4d).

We argue that the rapid warming in the subsurface layer and reduced vertical temperature gradient in the WPWP also favored the intense deglacial SST increase and a La Niña-like mode. The intense SST rise in the WPWP increased the meridional and zonal temperature gradients (Fig. 2g, h) and reinvigorated heat transport to the high latitudes5,63. Ekman transport converged surface warm waters to the subtropics75, lowering the ΔSST between the WPWP and the East China Sea (Fig. 2f). The recharge of the equatorial warm pool supplied the Indonesian Throughflow and the Equatorial Undercurrent, thereby feeding the downstream surface waters of the cold tongue in the eastern Pacific76. The enhanced Walker circulation, in turn, facilitated the buildup of warm water westward by intensifying the equatorial easterly winds77. Under the La Niña-like state with more heat gathered in the WPWP, the Pacific thermocline tilted and the shoaling of thermocline in the EEP, currently the largest source of CO278,79, facilitated the outgassing of CO2 (Fig. 2e)11,32,33,80, further accelerating the global warming.

In this study, we compiled 340 SST and 7 subT published records from the Pacific and eastern Indian Ocean, as well as a new OHC record from the WPWP, to elucidate the thermal evolution of the IPWP region and its coupling the Southern Ocean during the last deglaciation. The zonal SST gradient in the tropical Pacific Ocean was reduced and El Niño-like conditions dominated in the last glacial, as the WPWP contracted towards the equator and the east. During the deglaciation, La Niña-like conditions prevailed as a result of the intense warming in the WPWP, which may have been related to the development of the barrier layer within the mixed layer, favoring heat built up at the sea surface. Both the subT and OHC in the warm pool rose earlier than the SST, and in phase with South Pacific subTs and orbital precession, implying that heat exchange between the tropical upper water column and the extratropical Southern Ocean enhanced WPWP warming paced by the precession. The WPWP began to substantially expand when La Niña-like conditions peaked at ~12 ka, and the enhanced zonal and meridional thermal gradients further facilitated global warming. Our results underscore the pivotal role of the thermal coupling between the WPWP and the Southern Ocean during the last deglacial warming and its relevance for future global warming.

Methods

Compilation of regional SST and SSTA stacks

We integrated 340 published SST records spanning the last 30,000 years from 302 stations in the Pacific and eastern Indian Oceans to reconstruct temporal and spatial distribution and ENSO-like variability since the last glacial. The data set is comprised of 188 planktonic foraminiferal calcite Mg/Ca records based on G. ruber and of 152 Uk’37 records (Fig. 1). Moreover, 7 published subT records based on Mg/Ca measurements of planktonic foraminiferal thermocline species and one new OHC record (Supplementary material) from core KX22-4 (Fig. 1) were incorporated to spatially investigate vertical thermal distribution and to explore the key process involved in global warming during deglaciation.

We directly applied the original SST data considering that the authors have supplied the optimal local SST estimations (Supplementary Data). All of the age models are either radiocarbon-dated or calibrated to the LR04 reference curve81. In this study, we specifically focused on the contrast between two key time intervals, i.e., the early Holocene (5–10 ka) and last glacial period (17–29 ka). We considered the last glacial period rather than the LGM (19–23 ka) because the timing of the lowest glacial SST was not consistent in the published records, and the averaged SST for the LGM and last glacial did not differ significantly (0.10 °C, Supplementary Fig. 10). Additionally, in view of the fact that different approaches such as cleaning methods between laboratories, estimations of the dissolution effect would bring offsets to Mg/Ca-based SST reconstructions, we estimated SSTA relative to the mean value of the core top 1 kyr for each downcore SST data.

Cluster analysis was applied to classify the 169 sites in the western Pacific and eastern Indian Ocean areas into four regions (Fig. 1 and Supplementary Fig. 1): the South China Sea, eastern Indian Ocean, west and east part of the WPWP, which is beneficial to explore the spatial thermal transfer between regions. Cluster analysis was performed on SST data (Mg/Ca and Uk’37) covering both the early Holocene and last glacial periods using software Past 4.08 with Ward’s method82. Longitude and latitude were taken as references for cluster analysis. Zonal and meridional heat transfer was evaluated by calculating SST differences between regions.

We used STATNARY 1.2 (https://www.marum.de/Prof.-Dr.-michael-schulz/Michael-Schulz-Software.html) to calculate SST/SSTA stacks for all sites within each region with a sliding rectangular window of 1 kyr width (Fig. 2e, and Supplementary Figs. 5c, 12 and 13a, c), which could estimate the time-dependent mean of an unevenly spaced time series directly without the requirement of interpolation. We calculated the SST difference between regions after the data was linearly interpolated to 1 kyr time interval (Fig. 2d, g–i, Supplementary Figs. 5a, b, 7a, 9, and 11b).

Investigation of WPWP evolution and ENSO mode

We used temporal variations in SST distribution based on 160 SST records from a meridional region (35°N–35°S, 110°E–130°E) and 95 SST records from a zonal region (10°N–10°S, 90°E–180°E) to estimate meridional/zonal shifts of the WPWP during the last 30 kyr. To investigate variations of the WPWP boundaries, data from all sites were linearly interpolated to 1 kyr time interval, orthogonal polynomial fitting (OPF) was applied to data of every 1 kyr SST distribution. We defined the range of the WPWP by the warmest 1 °C based on Mg/Ca-SST trend line through time (Fig. 3), because a uniform criterion is invalid for both glacial and interglacial periods. Then, the positions of the WPWP boundaries and SST maximum could be recognized.

We used SST/SSTA data from sites within the central WPWP (117°E–161°E, 5°S–4°N, modern 29 °C isotherm) to explore the ENSO evolution. The SST variability within the 29 °C isotherm is consistent with that within the 28 °C isotherm but with an offset of 0.45 °C (Supplementary Fig. 11). Then we compared these data with published SST/SSTA data from the cold tongue of EEP (Fig. 1 and Supplementary Table 1), which are commonly used to characterize ENSO variations12,21, as the large SST anomalies in the cold tongue of EEP are correlated strongly with the Niño-3 index21.