Sea Surface Temperatures Across the Coral Sea Over the Last Glacial‐Interglacial Cycle

Sea surface temperature (SST) across the Coral Sea is tightly coupled to the regional and larger‐scale ocean circulation and climate. Continuous records that reflect past changes in Coral Sea SST in high resolution are missing, however. Here, we present Mg/Ca‐ and alkenone‐based SST reconstructions from the northwestern Coral Sea that cover the past 130 kyr. Our SST estimates vary in line with southern hemisphere high latitude climate variability, linked to atmospheric CO2. Combining the newly generated with published records, we find that the SST changes across the Coral Sea show a large spatial heterogeneity during the last glacial‐interglacial cycle. Particularly prior to 60 ka, precession exerts a discernible influence on the meridional SST gradients across the Coral Sea, as well as between the Coral Sea and southern Western Pacific Warm Pool. We posit that these changes are linked to the regional trade winds and the South Pacific subtropical gyre circulation, and/or to changes in the El Niño Southern Oscillation.


Introduction
Sea surface temperature (SST) within the Coral Sea interacts with the larger-scale western Pacific Ocean and climate system in multiple ways: From the Coral Sea, waters are transported to the equatorial Pacific as well as to the high southern latitudes and the Indian Ocean via western boundary currents (WBCs), namely the Gulf of Papua Current (GPC) and the East Australian Current (EAC) (Figure 1), influencing both the Western Pacific Warm Pool (WPWP) (e.g., Hu et al., 2015) and the Australian and New Zealand climate (e.g., Shi et al., 2008;Sprintall et al., 1995).Meridional temperature gradients affect the strength of the WBCs, which in turn control the export of heat and energy.Furthermore, SSTs play an important role in regulating the regional hydroclimate (e.g., Shi et al., 2008).
Despite the importance of SSTs in the Coral Sea, the few existing records are mostly of low temporal resolution (e.g., Lawrence & Herbert, 2005;Russon et al., 2010) and neither continuous (e.g., Felis et al., 2014;Tachikawa et al., 2009), nor consistent, limiting our understanding of past changes in the temperature and circulation of the Coral Sea, and of their driving mechanisms.For instance, coral-based SST records from the GBR indicate a glacial-interglacial amplitude of 6°C (Felis et al., 2014), which is much higher than 1.5°C based on alkenones in sediments from the GBR (Lawrence & Herbert, 2005) and 2-3°C based on Mg/Ca records from the southeastern Coral Sea (Russon et al., 2010;Tachikawa et al., 2009).Furthermore, the Mg/Ca-based SST reconstructions from the southeastern Coral Sea depict peculiar patterns of change, such as an unusually early deglacial warming that precedes deglacial warming of the WPWP (e.g., Felis et al., 2014;Tachikawa et al., 2009).
Records of millennial-scale variability during the last deglaciation are also equivocal.An SST cooling associated with the Northern Hemisphere Younger Dryas (YD) period (12.9-11.7 ka) is evident in corals from Vanuatu (Corrège et al., 2004) and Tahiti (Asami et al., 2009), but not in corals from the GBR (Brenner et al., 2020;Felis et al., 2014).At the GBR, Felis et al. (2014) and Brenner et al. (2020) rather report a warming during the YD.However, the coral-based SST records are not continuous and, therefore can only provide snapshots of the deglacial warming pattern.Likewise, most of the other available records are of low temporal resolution and do not allow investigation of millennial-scale temperature variations.
Meridional SST gradients calculated between sites located along the northeastern Australian coast were used as an indicator for changes in the EAC strength and related changes in the regional ocean circulation and climate (e.g., Anderson et al., 1989;Barrows et al., 2000;Felis et al., 2014;Zhai et al., 2022).For instance, a study on fossil corals from the GBR suggests a steeper-than-today meridional SST gradient in the Coral Sea during the Last Glacial Maximum (LGM) and the last deglaciation (Felis et al., 2014).Based on this finding, the authors inferred a weaker South Pacific subtropical gyre circulation and EAC and, consequently, a northward expansion of cooler subtropical waters.These authors further suggested that the northward expansions of cooler subtropical waters might have restricted the southern boundary of the WPWP and southward migration of the Intertropical Convergence Zone (ITCZ) to a more northerly position.In addition, the heat transfer from low to high latitudes would have been reduced, resulting in a less effective cooling of the tropics during the LGM.However, the coralbased reconstructions are not continuous, and the interpretations did not take reconstructions from the southern WPWP boundary into account.In contrast, the meridional SST gradient calculated from marine sediment records of the southeastern Coral Sea and the WPWP during the LGM and last deglaciation was weaker than today (Felis et al., 2014 and references therein).Reconstructions from the Tasman Sea mostly suggest a cooling and  (Regoli et al., 2015), MD05-2925 (Lo et al., 2017), MD97-2125 (Tachikawa et al., 2009) and MD06-3018 (Russon et al., 2010) offshore New Caledonia and ODP site 820 (Lawrence & Herbert, 2005) offshore Queensland.Gray dots mark the positions of fossil coral sites from Great Barrier Reef (Felis et al., 2014) and Vanuatu (Corrège et al., 2004).VAN = Vanuatu, NOG = Noggin Pass, HYD = Hydrographer's Passage, HER = Heron Island.This figure was created with the Ocean Data View software (Schlitzer, 2021).

10.1029/2023PA004757
weakening of the EAC during glacial periods as compared to interglacial periods (e.g., Bostock et al., 2006;De Deckker et al., 2018;Sikes et al., 2009).Higher than present SSTs during marine isotope stage (MIS) 5e at multiple sites from the western Tasman Sea also imply a warming, a strengthening and a southward extension of the EAC during MIS 5e (Cortese et al., 2013).This is in contrast to southeastern Coral Sea records that indicate a cooler SST during MIS 5e.However, southeastern Coral Sea sites are too far from Australia to allow any meaningful conclusions on EAC changes.Taken together, our understanding of the past SST variability across the Coral Sea, its forcing mechanisms, and implications for the regional ocean and climate system is still limited.
Here, we present new planktic foraminiferal Mg/Ca-and alkenone-(U K′ 37 ) based records of SST from combined gravity cores GeoB22229-1 and GeoB22230-1 (Mohtadi et al., 2017), which were recovered offshore northeastern Australia (145°5'E, 15°38'S) in the northwestern Coral Sea.The combined records of these two cores provide the first continuous Coral Sea time series covering the last 130 kyr at relatively high (up to millennialscale) resolution.To address the conflicting Mg/Ca and U K′ 37 SST reconstructions, we perform paired Mg/Ca and U K′ 37 measurements, which allows for a direct comparison of the resulting SST records at the same core site.In contrast to most of the previously published Coral Sea records (Barrows & Juggins, 2005;Russon et al., 2010;Tachikawa et al., 2009), our cores were collected close to the Australian continent, where the South Equatorial Current (SEC) splits into the northward and southward directed WBCs.We study the SST evolution across the Coral Sea and explore potential forcing mechanisms and implications for the regional ocean and climate system on millennial to glacial-interglacial timescales.To this end, we also compare our SST records with records from the southern WPWP, particularly the Gulf of Papua (MD05-2930; Regoli et al., 2015) and the Solomon Sea (MD05-2925; Lo et al., 2017), as well as with records from the southeastern Coral Sea (MD97-2125; Tachikawa et al., 2009).

Modern Setting
The Coral Sea stretches from the East Australian coast to New Caledonia.At present, mean annual SSTs within the northernmost Coral Sea are typically around 27°C, successively decreasing toward the South (Figure 1).The main circulation system influencing the Coral Sea is the South Pacific subtropical gyre.Specifically, surface waters are transported into the Coral Sea by the westward-flowing South Equatorial Current (SEC).At the Australian coast, at around 14.5°S, the SEC bifurcates into the northward-flowing GPC (including North Queensland Current and Hiri Current as suggested by Burrage et al. (2012)) and the southward-flowing EAC (Kessler & Cravatte, 2013b;Kessler & Gourdeau, 2007;Ridgway et al., 2018).From the Coral Sea, the GPC flows into the Gulf of Papua, where it forms a semi-enclosed clockwise gyre.As it enters the Solomon Sea, it feeds the New Guinea Coastal Current (NGCC) and New Guinea Coastal Undercurrent (NGCUC).The total northward transport is estimated to be about ∼21-29 Sv (Gasparin et al., 2012;Hristova et al., 2014;Sokolov & Rintoul, 2000).The EAC flows southward into the Tasman Sea and transports ∼27-36 Sv of warm tropical waters poleward (Kessler & Gourdeau, 2007).At around 30°S, a part of the current detaches from the continent forming a series of eddies that flow eastward (e.g., Ridgway & Dunn, 2003).
The regional surface circulation varies on a seasonal time scale.Changes of the trade winds go along with meridional shifts of the latitude of SEC bifurcation and changes in the strength of the western boundary currents (e.g., Hu et al., 2015).During austral summer, when northwest winds prevail, the SEC bifurcation latitude is located at its northernmost position close to the equator.The GPC is weaker, while the EAC is stronger (Chen & Wu, 2015;Kessler & Gourdeau, 2007;Ridgway & Godfrey, 1997).During austral winter, when southeast winds prevail, the SEC bifurcation latitude shifts southward and the GPC strengthens, while the EAC weakens.Subsequently, the spatiotemporal variability of SST (Figure 2; Reynolds et al., 2002) is also dominated by the seasonal cycle.For instance, close to site GeoB22229-1, seasonal SST variations range roughly between 3.5 and 6.0°C.Differences in the seasonal cycle of SST lead to smaller SST gradients between the Gulf of Papua/Solomon Sea and the Coral Sea sites during austral summer (Figures 1b-1c and 2c-2d) and a larger SST gradient between the northwestern and southeastern Coral Sea sites during austral summer (Figures 1b-1c and 2e).
On interannual timescales, the regional surface circulation is influenced by the El Niño Southern Oscillation (ENSO).Changes in ENSO go along with shifts in the position of the SEC bifurcation latitude, with a more northern SEC bifurcation during El Niño conditions.The SEC and the northward flowing WBCs increase during El Niño, contributing significantly to interannual temperature changes in the equatorial Pacific Ocean (Davis et al., 2012;Kessler & Cravatte, 2013a;Melet et al., 2013;Sprintall et al., 2021).However, several studies show that only a weak ENSO signal is evident in the EAC transport, and only a small proportion of the interannual EAC variability is correlated with the El Niño Southern Oscillation (Hu et al., 2015;Ridgway & Dunn, 2007;Sprintall et al., 2021).SSTs at the Gulf of Papua, Solomon Sea and Coral Sea sites compared in this study do not show substantial variations in response to the ENSO (Figure 2b).While the SST gradients between the Gulf of Papua and the Coral Sea, and the SST gradient between the northwestern and southeastern Coral Sea sites do not show distinctive interannual variations (Figures 2d and 2e), the SST gradient between the Solomon and Coral Sea is generally smaller (larger) during El Niño (La Niña) (Figure 2c).This is in agreement with the higher sensitivity of the Solomon Sea to ENSO as compared to the Gulf of Papua and the Coral Sea sites apparent from ocean observations.On interannual to decadal timescales, the Pacific Decadal Oscillation (PDO) affects the regional climate similarly to ENSO, with positive phases having similar effects on the climate as El Niño.Therefore, the SST gradients between the Gulf of Papua/Solomon Sea and the northwestern Coral Sea are also generally consistent with the Pacific Decadal Oscillation (PDO) modulating ENSO (Figures 2c-2f).
At present, the EAC is additionally influenced by the Southern Annular Mode (SAM), through its impact on the position and strength of the southern hemisphere westerly wind belt, with an increased EAC strength during positive the Southern Annular Mode phases (e.g., Cai et al., 2005).However, SST (gradients) at the study sites do not show substantial variations in response to the Southern Annular Mode (Figures 2c-2e, and 2g).

Material
Gravity cores GeoB22229-1 and GeoB22230-1 were recovered offshore eastern Australia near Ribbon Reef (145°55'E, 15°28'S and 145°52'E, 15°26'S) from 1443 m and 968 m water depth, respectively.These shallow coring depths imply a good carbonate preservation, which is confirmed by the abundance of aragonitic pteropod tests in samples of both cores.The sites are located close to the SEC bifurcation at the Australian coast.Both cores were sampled on board at 4 cm steps; and later at 2 cm steps for the last deglaciation section of GeoB22229-1.The samples were freeze dried, washed over 63-and 150-μm sieves and dried at 40-60°C.Our SST estimates are based on Mg/Ca measurements on tests of the planktic foraminifera Globigerinoides ruber (white) (previously referred to as G. ruber (white) sensu strictu) as well as on the U K′ 37 unsaturation index.

Age Models
The age models of cores GeoB22229-1 and GeoB22230-1 are based on 10 AMS 14 C ages of planktic foraminifera and the alignment of stable oxygen isotope (δ 18 O) records of the planktic foraminifera G. ruber to the intermediate Pacific benthic LS16 stack of Lisiecki and Stern (2016).For radiocarbon dating, we used samples of mixed G. ruber, Globigerinoides elongatus, Trilobatus sacculifer and Orbulina universa tests.Comparisons of planktic and benthic δ 18 O records from the Solomon and southeastern Coral Seas (Lo et al., 2017(Lo et al., , 2022;;Tachikawa et al., 2009) suggest that temporal offsets between these records are negligible and we thus use the LS16 stack as alignment target for our record, even though our record is based on planktic foraminifera tests.
Radiocarbon ages were measured at the Analytical Center for Environmental Science, Atmosphere and Ocean Research Institute, University of Tokyo, Japan.The calibration of 14 C ages to calendar ages is based on the Marine20 calibration curve (Heaton et al., 2020) with a local reservoir age correction of ΔR = 153 years (Heron Island; Druffel & Griffin, 1999), using the CALIB14 calibration software (Stuiver & Reimer, 1993).The radiocarbon dates of GeoB22229-1 indicate an age reversal between 214 and 274 cm (14.1 and 12.3 ka; Table 1).
We excluded the 214 cm-sample from the age-depth model (see Figures S1 and S2 in Supporting Information S1 for details).Tie points were mostly set at MIS boundaries (Table 1).Between dating points, the chronology is based on linear interpolation.All ages are given as calibrated calendar ages before present (BP; present: CE1950).
Resulting age models indicate that GeoB22229-1 covers the last ∼119 kyr, while GeoB22230-1 extends back to 132 ka and fully covers MIS 5 (Figure 3).Combined records of these two cores can thus be used to reconstruct the climate evolution of eastern Australia over the entire last glacial-interglacial cycle.To validate the age models, we used G. ruber δ 13 C records and Ti/Ca elemental ratios (Figures S1 in Supporting Information S1).We found specimens of G. ruber (pink) up to a core depth of 718 cm in GeoB22229-1 and 446 cm in GeoB22230-1.
According to our age model, this corresponds to ages of ∼116.5 and 116.3 ka, which is in accordance with results from the WPWP records of Thompson et al. (1979).The last appearance of G. ruber (pink) in our two cores thus provides further confidence in our age models and the splicing of the GeoB222229-1 and GeoB22230-1 records.Core GeoB22229-1 does not show any clear signals of discontinuity, however from MIS 4 to the LGM, sediment accumulation was at least reduced, which makes thorough age modeling very difficult and we cannot exclude potential hiatuses during this interval.We therefore focus our analysis and interpretation on long-term, that is, orbital scale and glacial-interglacial variations, as well as on MIS 5 and the period from the LGM to the late Holocene, as the age model is more robust for these intervals, and, for the latter period, supported by AMS 14 Cages.

Stable Oxygen Isotope and Mg/Ca Analyses
To develop the age model beyond the limits of radiocarbon dating we used the stable oxygen isotope record of G. ruber (white) s.s.(250-355 μm).The samples were measured by the isotope laboratory at MARUM, University of Bremen on a ThermoFisher MAT 253 mass spectrometer, connected with automatic lines for carbonate preparation of type Kiel IV.All isotope values were calibrated against the international Vienna Pee Dee Belemnite (VPDB) standard by using an in-house carbonate standard, which has been calibrated to the National Bureau of Standards (NBS) 19 standard.During the measurement period, the analytical standard deviation was ±0.05‰.
For SST reconstructions, we measured the Mg/Ca ratio of tests of G. ruber (white) s.s.(250-355 μm; 30 specimens, where available).For sample preparation, we followed the procedure of Barker et al. (2003).Mg/Ca measurements were performed at MARUM, University of Bremen, on an Agilent Technologies 700 Series Inductively Coupled Plasma Optical Emission Spectrophotometer (ICP-OES), which is connected to a Cetax ASX 520 autosampler.The analytical standard deviation was better than 0.01 mmol/mol.To control instrumental drift and precision, an in-house standard solution with a Mg/Ca nominal value of 2.93 mmol/mol was measured after every fifth sample.The international ECRM 752-1 standard with a reported Mg/Ca of 3.76 mmol/mol (Greaves et al., 2008) was measured every fiftieth sample.The standard deviations were better than 0.02 mmol/ mol (0.2%) for the in-house standard solution and 0.007 mmol/mol (0.05%) for ECRM.We corrected the Mg/Ca ratios of all samples for the reported ECRM value.Replicate measurements of roughly 10% of the samples indicate a reproducibility of ±0.15 mmol/mol for GeoB22229-1 and ±0.18 mmol/mol for GeoB22230-1.To control cleaning efficiency, Al/Ca, Fe/Ca and Mn/Ca were measured alongside with Mg/Ca.Al/Ca was under the ICP-OES detection limit.Fe/Ca and Mn/Ca were mostly below 100 μmol/mol, typical values for clean, uncontaminated samples (Barker et al., 2003) (Figure S3 in Supporting Information S1).Besides, none of these ratios shows a correlation to Mg/Ca, implying that Mg/Ca is not affected by contamination.Generally, Mg/Ca is additionally influenced by salinity and carbonate chemistry (e.g., Arbuszewski et al., 2010;Gray & Evans, 2019;Hönisch et al., 2013;Kısakürek et al., 2008), but studies disagree on the magnitude of this influence.Recent multivariate calibrations, which account for secondary influences of salinity and pH on foraminiferal Mg/Ca, indicate that univariate calibrations, which assume a Mg/Ca-temperature dependency of 9% per °C, may overestimate the thermal component (Gray & Evans, 2019;Gray et al., 2018;Saenger & Evans, 2019;Tierney et al., 2019).However, the comparative application of uni-and multivariate Mg/Ca-temperature calibrations to a suite of downcore records provided confidence that, despite the confounding nonthermal effects, estimated temperatures are within the calibrations' uncertainty (Rosenthal et al., 2022).Furthermore, past changes in salinity and pH at our core sites are unconstrained.We thus do not apply any correction factors to account for changes in salinity or carbonate chemistry.
To convert Mg/Ca to temperature, we used the species-specific Mg/Ca-temperature calibration of Anand et al. (2003) with an assumed exponential constant of 0.09.Regional Mg/Ca-temperature calibrations are not available, but the application of this equation to the core top Mg/Ca results in an SST value of 26.8°C, which is close to the 1981-2022 mean OISST of 26.9°C (Reynolds et al., 2002).Uncertainties of SST estimates were assessed by propagation of the uncertainties introduced by Mg/Ca measurements (see above) and the Mg/Ca- temperature calibration (Anand et al., 2003) as described by Mohtadi et al. (2014).Resulting uncertainties average about 1.0°C.For comparison, we also applied the multivariate equations of Gray et al. (2018) and Gray and Evans (2019).Consistent with Rosenthal et al. (2022), we find that estimates of absolute SST differ from those calculated with the univariate calibration equation of Anand et al. (2003), overestimating modern SST at the core site (Figure S4 in Supporting Information S1).In addition, application of the multivariate equations results in slightly larger amplitudes of past SST changes, but overall, the downcore SST variability is in agreement with the variability that results from the application of the univariate calibration equation.

Alkenone Analysis
We additionally used the alkenone unsaturation index (Brassell, 1986) to reconstruct SST.Sample preparation and measurements were done at the German Federal Institute for Geosciences and Natural Resources in Hannover.All samples were extracted with a Dionex Accelerated Solvent Extractor 200 in three cycles using dichloromethane/methanol (95:5; v:v) as eluent.Extracts were dried under a stream of nitrogen, saponified with 0.5 ml 1-propanolic KOH (5%) for 24hr at 20°C followed by a solid phase clean-up using silica gel columns to remove the KOH.These purified extracts were analyzed by gas chromatography with an HP-6890 instrument equipped with an HP PTV Inlet on a DB-1 capillary column (30 m × 0.25 mm i.d.; film thickness 0.25 μm) coupled to a flame ionisation detector.Afterward, the samples were injected in dichloromethane using a cool injection program with solvent venting.Hydrogen was the carrier gas at a flow of 0.9 ml/min.A temperature program of 2 min isothermal at 56°C, 56-150°C at 24°C/min, 150-320°C at 4.7°C/min and 10 min isothermal was used, giving a good separation of all major compounds.Alkenones were identified by retention times.Quantification was performed relative to external calibration with n-C36 alkane (Doose-Rolinski et al., 2001).Precision and accuracy of the alkenone analyses were controlled by replicate measurements.The analytical standard deviation was better than 0.006 units.Calculated U K′ 37 was converted to temperature using the temperature calibration of Prahl et al. (1988), which yields a core top SST of 27.2°C, closest to modern SST values, although it slightly exceeds present-day temperatures, that is, other commonly applied U K′ 37 -temperature calibrations (e.g., Conte et al., 2006;Müller et al., 1998;Pelejero & Grimalt, 1997;Tierney & Tingley, 2018) result in higher SST, overestimating modern values.

Mg/Ca-SST Reconstructions
Mg/Ca of GeoB22229-1 varies between 3.6 and 5.5 mmol/mol, which converts to SSTs between 23.3 and 27.9°C (Figure 3).Maximum Mg/Ca values of GeoB22230-1 are about 5.9 mmol/mol, which converts to 28.7°C.The core top SST of GeoB22229-1 of 26.8°C is in agreement with the modern mean annual SST of 26.7-26.9°C(Locarnini et al., 2018;Reynolds et al., 2002) at this site.The GeoB22229-1/GeoB22230-1 record shows clear glacial-interglacial variations with elevated temperatures during MIS 5e and the Holocene.From MIS 5d to MIS 5a, SSTs are relatively cool with relative minima during MIS 5d and b, where SSTs are even slightly lower than during MIS 4.During MIS3 (from 40 ka to 23 ka), SSTs vary around 24.7°C, and decrease slightly during the LGM (23-19 ka), with estimates ranging from 24.1 to 24.5°C.Minimum temperatures of 24.0°C are reached at ∼18.4 ka.Afterward, SST sharply increases by about 3.7°C, up to 27.7°C at 11.6 ka, but the warming trend is interrupted by a temperature drop of about 1°C between 14.8 and 12.2 ka.During the Holocene, SST varies between 25.8 and 27.9°C (average 26.6°C; excluding one outlier that indicates an SST of 24.2°C) with a slight cooling trend toward the middle Holocene and a warming afterward.The record indicates an overall LGM cooling of 2.3°C relative to the Holocene.

U K′ 37 -SST Reconstructions
The alkenone unsaturation ratio, U K′ 37 , varies between 0.919 and 0.986 for GeoB22229-1, and between 0.945 and 0.997 for GeoB22230-1, which converts to SSTs between 25.9 and 28.2°C (Figure 3).The U K′ 37 record depicts elevated temperatures during MIS 5 and the Holocene.From MIS 5 to the LGM, U K′ 37 -SSTs indicate a steady cooling, which is interrupted by a slight temperature drop during MIS 4.During the LGM at ∼20.3 ka, the record indicates minimum SSTs of 25.9°C.Afterward, SST increases by about 1.9°C, up to 27.8°C at 11.3 ka, but the warming trend is interrupted by a temperature drop of about 0.5°C between 14.2 and 12.2 ka.During the Holocene, the record depicts a slight cooling trend.The record indicates an overall LGM cooling of ∼1.5°C relative to the Holocene.
Both the Mg/Ca-and U K′ 37 -inferred SST records of GeoB22229-1 and GeoB22230-1 show an offset between ∼114 and 110 ka (Figure 3).Such a mismatch could result from inconsistencies of the age models.However, any shift of the records along the age axes would result in increasing mismatches of the isotope and/or XRF records of the cores (cf. Figure 3 and Figure S1 in Supporting Information S1).

Mg/Ca-Versus Alkenone-Based SST Estimates
The patterns of the Mg/Ca-and alkenone-based SST records of GeoB22229-1/GeoB22230-1 are in general agreement and share some common features (Figures 3 and 4).For instance, both records indicate warmer SSTs during MIS 5e compared to the Holocene.In addition, they both show a similar timing of deglacial temperature rise, and an SST drop during the last deglaciation.These features are discussed in more detail in Sections 5.2 and 5.3.The consistency of the Mg/Ca-and alkenone-based SST records at this point show that these common features do not depend on the choice of proxy.However, absolute SST estimates and their variability differ between the records.Specifically, SSTs calculated from U K′ 37 exceed Mg/Ca-based SSTs, independent of the choice of U K′ 37 -temperature calibration (Figure S5 in Supporting Information S1).As pointed out in Section 3.4, core top SST also slightly exceeds present-day temperatures.It thus appears likely that the alkenone-based SST record overestimates reconstructed temperatures.The reason for this mismatch is unclear.It is noteworthy that U K′ 37 is in the same range of values as nearby core ODP site 820 (Lawrence & Herbert, 2005), providing confidence in the reliability of our data.Potentially, the existing calibrations are not suitable to accurately capture Coral Sea temperatures, and/or U K′ 37 does not reflect mean annual, but austral summer SSTs.However, seasonal differences between Mg/Ca and U K′ 37 records were not observed in the southeastern Coral Sea records of MD97-2125 (Tachikawa et al., 2009).
In addition, our alkenone-based record exhibits smaller amplitudes of SST variability compared with the Mg/Cabased record.Notably, the record is close to the limit of this proxy's sensitivity around 28°C and thus potentially underestimates amplitudes of past SST changes and temperatures at the high end.This holds true even when applying the Bayesian calibration of Tierney and Tingley (2018), which predicts larger amplitude SST changes than the calibration of Prahl et al. (1988) (Figure S5 in Supporting Information S1).Mg/Ca and U K′ 37 are prone to different uncertainties, due to disparate life cycles of the proxy producing organisms, and different secondary effects, which might explain the generally smoothed variability of the U K′ 37 record compared with the Mg/Ca record.Based on our data alone, we cannot fully explain the differing amplitudes.However, the differing amplitudes of Mg/Ca and U K′ 37 records from paired measurements of the same core shows that previously reported differences in the amplitudes of Coral Sea Mg/Ca and U K′ 37 records, for instance, during glacial-interglacial transitions, do not reflect spatial variations of SST variability, but more likely, proxy biases or seasonal differences.

SST Evolution Across the Coral Sea
In order to investigate the spatiotemporal evolution of SST across the Coral Sea during the last glacialinterglacial cycle, and to put it into a larger paleoenvironmental context, we compare our SST record with records from the equatorial Indo-Pacific as well as with records from the southeastern Coral Sea and adjacent areas.The comparison is mainly based on Mg/Ca-records.This is not to favor Mg/Ca-derived reconstructions, but because a number of Mg/Ca-based reconstructions are available from the Coral Sea and southern WPWP and can be used for comparison.Comparison of the interglacial periods in our new SST records indicates higher SSTs during MIS 5e compared to the Holocene (Figures 3 and 4), supporting the hypothesis of an overall warmer GBR (Dechnik et al., 2017) during the last interglacial period.Our reconstruction of higher SSTs during MIS 5e is in line with reconstructions from both the equatorial (e.g., Hollstein et al., 2020;Lo et al., 2017;Regoli et al., 2015;Tachikawa et al., 2014) and the southwestern (e.g., Cortese et al., 2013;Pahnke et al., 2003) Pacific Ocean, including alkenone unsaturation records (Pelejero et al., 1999) southwest Pacific Ocean have previously been related to a warming and southward extension of the EAC during MIS 5e that went along with an increased southern hemisphere subtropical gyre circulation (e.g., Cortese et al., 2013).Warmer temperatures across the GBR are supported by a southward extension of reefs along the GBR during the last interglacial (Dechnik et al., 2017;Pickett, 1981;Pickett et al., 1989).In contrast, records from the southeastern Coral Sea, namely MD06-3018 (Russon et al., 2010) and MD97-2125 (Tachikawa  (Brenner et al., 2020;Felis et al., 2014).Note the different scaling of the y-axis.(e) Southern hemisphere proxy SST stack (Shakun et al., 2012), (f) δD from the EDC ice core (Bazin et al., 2013b) and Antarctic composite ice core atmospheric CO 2 (Bereiter et al., 2015b).Triangles indicate dating points of GeoB22229-1/GeoB22230-1. Records MD05-2930, MD05-2925 and MD97-2125 are shown on modified age scales (see Section 5.3 and Figure S6 in Supporting Information S1).Likewise, the calibrations of the southeastern Coral Sea records were adapted according to the ones we used for our records.Colored shadings represent 1σ-uncertainties.Gray bars mark the Antarctic Cold Reversal (ACR), Last Glacial Maximum (LGM) and MIS 4. Note the axis break at 30 ka. et al., 2009) indicate lower SSTs during MIS 5e than during the Holocene.The reason is unclear, but regional effects (see Section 5.3) might play a role.
For the LGM, our Mg/Ca-based SST reconstructions of GeoB22229-1 indicate a cooling of 2.3°C relative to the Holocene.This is consistent with estimates from Mg/Ca-based records from the Gulf of Papua (Regoli et al., 2015) and Solomon Sea (Lo et al., 2017) and southeastern Coral Sea (Russon et al., 2010;Tachikawa et al., 2009).Hence, regional reconstructions reflect a largely consistent amplitude of glacial cooling across the Coral Sea and adjacent areas.The few available alkenone-based SST records indicate a weaker glacial cooling of 1.5°C (e.g., this study; Lawrence & Herbert, 2005), but potentially underestimate SST changes (see Section 5.1).As pointed out in the introduction section, coral-based SST estimates indicate a larger LGM cooling of 4-6°C (Figure 4; Brenner et al., 2020;Felis et al., 2014).However, these estimates are uncorrected for changes in seawater Sr/Ca and consequently provide upper limits for the estimated temperature change.
The onset and pattern of the following deglacial temperature rise differ between the available SST records from across the Coral Sea.The Mg/Ca-based SST record of GeoB22229-1 depicts a canonical Southern Hemisphere pattern of deglacial warming, as represented by the Southern Hemisphere SST stack (Figure 4; Shakun et al., 2012).The deglacial temperature rise started around 18.6 ka and lasted until about 11.6 ka, the beginning of the Holocene.This is in line with many other reconstructions from the tropical Indo-Pacific realm (e.g., Dang et al., 2020;Lea et al., 2000;Schröder et al., 2018;Tachikawa et al., 2014) including the Gulf of Papua and Solomon Sea records, which indicate a similar onset of deglacial warming but a later Holocene SST peak than our record (Figure 4; Lo et al., 2017;Regoli et al., 2015).In contrast, the SST records from the southeastern Coral Sea depict a substantially different timing of deglacial warming (Russon et al., 2010;Tachikawa et al., 2009).Particularly the SST record of MD97-2125 shows an earlier deglacial SST rise, starting around 20.4 ka, with maximum SST between 15 and 14 ka (Figure 4), preceding the deglacial change in surface water δ 18 O from the same site (Tachikawa et al., 2009).Such a phase shift between SST and δ 18 O cannot be observed in our record (Figure S8 in Supporting Information S1).In addition, the patterns of temperature rise during the deglaciation are equivocal.This will be discussed in more details in Section 5.3.Taken together, the few available Mg/Ca-based SST records depict a spatially highly variable SST evolution across the Coral Sea over the last glacial-interglacial cycle.
The differing temporal evolution of SSTs and the resulting time-varying changes in the spatial distribution of SSTs are coupled to changes in the regional atmospheric and oceanic circulations.To better understand the longterm variability of the regional climate and ocean circulation, we calculate SST gradients between our northwestern Coral Sea records and previously published records from the Gulf of Papua (MD05-2930; Regoli et al., 2015)/the Solomon Sea (MD05-2925; Lo et al., 2017) and the southeastern Coral Sea (MD97-2125; Tachikawa et al., 2009).To ensure comparability of the records, we updated the age models of these cores according to our age-depth models (supporting information, Figures S6 and S7 in Supporting Information S1).In addition, we applied the species-specific calibration of Anand et al. (2003) also to the southeastern Coral Sea records.The Gulf of Papua and Solomon Sea records were converted to temperature by applying the multispecies calibration of Anand et al. (2003) in the original publications.We kept this conversion, as it provides reasonable results, in accordance with a study from modern WPWP core top sediments (Hollstein et al., 2017).To calculate SST gradients, we resampled all records at intervals of 1,000 years.
Prior to 60 ka, SST gradients calculated between the individual sites (ΔSST) are highly variable (Figure 5).In principle, age model uncertainties could lead to offsets between the SST peaks of the individual records.However, considering that the MIS 5 peaks of the δ 18 O records are in line (Figure 5), aligning the maxima in SST would necessarily lead to offsets between the δ 18 O records of the individual cores.This also relates to phase shifts between δ 18 O and SST during terminations, which are observed in the southeastern Coral Sea, but are less apparent in the northeastern Coral Sea, Gulf of Papua and Solomon Sea records (Figure S8 in Supporting Information S1).
The ΔSST records show cyclic variations that, by visual comparison, are coherent with varying climatic precession.The SST gradients between the Gulf of Papua/Solomon Sea and our site are generally larger, when precession is low (Figure 5).This can be attributed to the comparatively strong, precession-controlled SST variability at the Gulf of Papua/Solomon Sea sites with higher SST during precession minima (Lo et al., 2022;Regoli et al., 2015).Conversely, the SST gradient between our site and the southeastern Coral Sea tends to be smaller, when precession is low.The opposed variability of the SST gradients resembles the pattern of seasonal ΔSST fluctuations observed in the modern ocean (see Section 2), although the amplitude of past ΔSST changes is larger than the seasonal variability observed today, and the records are supposed to reflect mean annual SSTs.By analogy to the seasonal variability observed today, the study area is characterized by southeasterly trade winds during low precession, a southward displaced SEC bifurcation latitude and, consequently, a stronger GPC and a weaker EAC, which could also be associated with a northward movement of the South Pacific Subtropical Front.During high precession, trade winds are northwesterly and the SEC bifurcation latitude is displaced northward, the GPC is weaker and the EAC stronger, while the South Pacific Subtropical Front moves south.
The precession control on the spatiotemporal SST variability could also be conveyed through changes in ENSO, although the influence of ENSO on the regional SST and circulation pattern is comparatively weak today.A number of studies have suggested precession-controlled changes in the state of ENSO, with either more ENSO variability/activity or more El Niño-like conditions (e.g., Beaufort et al., 2001), or less ENSO variability/activity or more La Niñalike conditions (e.g., Hollstein et al., 2020;Jian et al., 2022;Tudhope et al., 2001) under low precession.The latter would be characterized by a warming of Gulf of Papua/Solomon Sea surface waters, a more southern SEC bifurcation and a weaker GPC, and be associated with a larger SST gradient between our site and the Solomon Sea.The precession-relation is less obvious in the ΔSST record between our site and the southeastern Coral Sea, which is consistent with the low impact of ENSO on the present SST at and SST gradients between these sites.SST gradients between the western and eastern equatorial Pacific are more sensitive to changes in the state and activity in ENSO with larger SST gradients during La Niña.In agreement with our inferences, reconstructions of past changes of these zonal SST gradients also suggest a weaker ENSO activity or more La Niña-like conditions under precession minima (Dang et al., 2020;Zhang et al., 2021).We observe that gradients in SST and the oxygen isotope composition of seawater (δ 18 O SW ) vary along with each other (Figure S9 in Supporting Information S1).That means, larger SST gradients are accompanied by larger δ 18 O SW gradients, which would be supportive for our interpretation of the SST gradient records.However, we do not deepen the discussion of the δ 18 O records, as we used the G. ruber δ 18 O records to generate the age model of GeoB22229-1 and GeoB22230-1, which impedes the interpretation of the δ 18 O (δ 18 O SW ) records.
During the past ∼60 kyr, the SST gradient calculated between the Gulf of Papua and our site remains relatively constant (Figure 5), while the gradient between our site and the southeastern Coral Sea sites shows a substantial decrease during the LGM and last deglaciation, between ca.20 and 10 ka.This suggests a reduced influence of precession-driven insolation changes on the ΔSST records, likely due to the smaller amplitude of precession cycles and the overriding effect of changing glacial-interglacial boundary conditions from the LGM to the Holocene.From the relatively constant SST gradient between the Gulf of Papua and our site we infer that the SEC bifurcation latitude remained relatively stable during the past 60 kyr.This interpretation also implies that previously suggested changes in the EAC strength during the LGM and last deglaciation (Felis et al., 2014) were neither caused by, nor did they substantially affect the SEC bifurcation latitude.Consequently, our results at least challenge the hypothesis of Felis et al. (2014) of a substantial northward expansion of cooler subtropical waters, a subsequent restriction of the southern WPWP boundary and southward migration of the ITCZ to a more northerly position during the LGM and deglaciation.
The change in the SST gradient across the Coral Sea during the last deglaciation reflects the regionally diverse evolution of SSTs and is mainly connected to the anomalously early SST rise recorded at site MD97-2125.As  and c) SST records of GeoB22229-1/GeoB22230-1 (blue and orange; this study), MD05-2930 (pink;Regoli et al., 2015), MD05-2930 (red;Lo et al., 2017), MD97-2125 (black;Tachikawa et al., 2009).All records were resampled at 1000-year intervals.(d) ΔSST calculated between MD05-2930 and GeoB22229-1/ GeoB22230-1, (e) MD05-2925 and GeoB22229-1/GeoB22230-1, and between (f) GeoB22229-1/GeoB22230-1 and MD97-2125.Yellow curves show variations in the Earth's orbital precession (Laskar, 1990).Gray bars mark the Last Glacial Maximum (LGM), MIS 4 and MIS 6.Note that the yaxes in (d) and (e) are inverted.pointed out, it is not accompanied by substantial changes in the meridional SST gradient toward the Gulf of Papua.Thus, the decrease in the SST gradient across the Coral Sea during the last deglaciation is mechanistically different from the precession-modulated spatial SST variability, and related changes in the surface ocean circulation are difficult to assess.A decreasing SST gradient between the northwestern and southeastern Coral Sea sites during the LGM and last deglaciation does not necessarily reflect changes in the strength of the EAC, as site MD97-2125 is located far offshore between New Caledonia and the Chesterfield Islands (Figure 1).Here, the surface ocean circulation is complex and the core site is not predominantly influenced by the EAC today (e.g., Cravatte et al., 2015;Ridgway & Dunn, 2003).Unfortunately, records from the southern Coral Sea along the main route of the EAC are missing.Continuous, well-dated records that track the past SST variability along the northeast Australian coast are needed to better infer past variations in the EAC strength.

SST Evolution and Potential Relation to Southern Ocean Climate
A striking feature of our Coral Sea SST record is its strong similarity to climatic records from the southern high latitudes such as thermal changes inferred from the δD record of the EDC ice core (Bazin et al., 2013b) and global atmospheric CO 2 inferred from Antarctic ice cores (Bereiter et al., 2015b) (Figure 4).This coherence does not only apply to the glacial-interglacial SST variability, but includes to shorter-term variations, such as millennialscale changes during the last deglaciation, and the Holocene SST evolution.More specifically, the Coral Sea SST record is marked by a warm MIS 5e, and cooler SSTs during the second half of MIS 5 with values close to glacial ones, as it is known from southern high latitude climate records and atmospheric CO 2 concentrations (e.g., Toggweiler & Lea, 2010).Similarly, the onset and pattern of the SST rise during the last deglaciation, as well as the evolution during the Holocene, with an early Holocene SST peak, which is followed by a cooling until about 8 ka and a slight warming trend afterward, are synchronous with Antarctic temperature records and atmospheric CO 2 concentrations.Prior to the LGM, the millennial-scale SST variability is not resolved, thus a potential linkage at millennial timescales prior to the LGM cannot be examined here.In agreement with previous publications from the southern WPWP (e.g., Lo et al., 2017) our finding implies that Coral Sea temperatures are directly linked to changes in global atmospheric CO 2 (e.g., Lo et al., 2017) and/or to southern high latitude climate by oceanic pathways (e.g., Lo et al., 2017;Pena et al., 2008).However, this linkage is less well represented in other Coral Sea records.For instance, Tachikawa et al. (2009) showed that warming of the southeastern Coral Sea waters preceded the global atmospheric CO 2 rise.This discrepancy between records from the northwestern and southeastern Coral Sea again reflects the spatial heterogeneity of past SST changes across the Coral Sea.Particularly the regional SST evolution during the last deglaciation and its' relation to high latitude climate has been addressed in numerous studies (e.g., Kiefer & Kienast, 2005;Lo et al., 2014).The SST records of GeoB22229-1 depict a temperature drop between 14.8 and 12.2 ka, coeval with the Antarctic cold reversal (ACR; 14.0-12.4ka), representing the Southern Hemisphere see-saw response to changes of the Atlantic Meridional Overturning Circulation (AMOC) during the Bølling-Allerød.The interpretation of millennial-scale events requires a robust chronology and synchronization of age scales.As pointed out in Section 3.2, the radiocarbon dates of GeoB22229-1 indicate an age reversal between 214 and 274 cm (14.1 and 12.3 ka).Sea-level reconstructions from Noggin Pass (Figure 1) (Yokoyama et al., 2018), which is located close to our core site indicate a sharp increase in sea level between ∼13.8 and 13.0 ka (Figure S1 in Supporting Information S1).As a result, we cannot rule out that the observed SST drop is an artifact that is created by sediment remobilization and -deposition.
Hitherto available records of the last deglaciation are equivocal.A potentially ACR-related SST plateau or drop has been observed in several records from offshore southern Australia (e.g., Calvo et al., 2007;Sikes et al., 2009) and New Zealand (Pahnke et al., 2003), from the eastern Pacific Ocean off southern Chile (Kaiser et al., 2005), as well as in some records from the equatorial western Pacific Ocean (Lo et al., 2014;Moffa-Sanchez et al., 2019) and the Indonesian Seas (e.g., Levi et al., 2007;Schröder et al., 2016).An SST plateau during the ACR, and a warming trend during the YD is also depicted by the Gulf of Papua and Solomon Sea records (Lo et al., 2017;Regoli et al., 2015).However, previously published Coral Sea SST reconstructions indicate a steady SST increase during the last deglaciation.This could partly be related to the rather low resolution of the available proxy records, but is supported by other records from the (wider) Indo-Pacific area, that indicate a continuous deglacial SST warming, without interruptions during the ACR (e.g., Kiefer & Kienast, 2005 and references therein).Our finding is also in contrast to studies by Andres et al. (2003) and Morigi et al. (2003), who postulate a Southern Ocean (Great Australian Bight and New Zealand-Ross Sea) SST cooling during the Northern Hemisphere Younger Dryas (YD) period.However, they base their interpretation on planktic foraminiferal δ 18 O records, not on Paleoceanography and Paleoclimatology 10.1029/2023PA004757 independent SST proxies.An SST cooling associated with the YD is also evident in corals from Vanuatu (Corrège et al., 2004) and Tahiti (Asami et al., 2009), but not in corals from the GBR (Brenner et al., 2020;Felis et al., 2014).At the GBR, Felis et al. (2014) and Brenner et al. (2020) rather report a warming during the YD.However, the coral-based SST records are not continuous and therefore, can only provide very limited information about the deglacial warming pattern.
The differing response (or the lack thereof) of the individual records to changes in atmospheric CO 2 and southern high latitude climate variability might be due to the overprinting by local and regional influences.In addition, if changes in southern high latitude climate are conveyed by oceanic pathways, for instance via the South Pacific subtropical gyre circulation, this could particularly affect sites that are directly influenced by the gyre circulation and WBCs.For instance, our Coral Sea site is directly influenced by the South Pacific subtropical gyre, whereas the southeastern Coral Sea sites are located within the gyre and not directly influenced by the WBCs.

Summary and Conclusions
We provide the first continuous SST records from the Coral Sea that cover the last glacial-interglacial cycle in up to millennial-scale resolution.The patterns of the Mg/Ca-and alkenone-based SST records of GeoB22229-1/ GeoB22230-1 are in general agreement and indicate elevated SST during MIS 5e and a cooling of 1.5 (U K′ 37 ) to 2.3 (Mg/Ca) °C during the LGM as compared to the Holocene.Deglacial warming was interrupted by an SST drop of up to 1°C during the Antarctic Cold Reversal.Comparing our record to previously published records, we find that the hitherto available Mg/Ca-based SST reconstructions depict a spatially heterogenous SST evolution across the Coral Sea over the past ∼130 kyr.Our record of estimated Coral Sea SSTs shows strong similarities to climatic records from Antarctica on glacial-interglacial to millennial timescales, which is attributed to changing atmospheric CO 2 and/or climate in southern high latitudes.Particularly prior to ∼60 ka, precession-controlled insolation changes modulate the spatiotemporal evolution of SST.Precession exerts a noticeable influence on the meridional SST gradients across the Coral Sea, and between the Coral Sea and the Gulf of Papua/Solomon Sea, which we interpret to reflect changes in the regional trade winds and ocean circulation and/or changes in the state or activity of ENSO.To allow for a more direct assessment of past variations in EAC strength and their impacts, and provide a link to published records from the Tasman Sea, continuous, high-resolution records from the southern Coral Sea, particularly along the main route of the EAC are needed and could be a focus of future Coral Sea studies.
, although the timing and detailed structure of SST during MIS 5 differs between the individual records.Warmer SSTs in the

Table 1
AMS 14 C Ages, Calibrated Calendar Ages, Reported as Median Probability Ages, and Tie Points of Cores GeoB22229-1 and GeoB22230-1 HOLLSTEIN ET AL.