Revisiting the Geographical Extent of Exceptional Warmth in the Early Paleogene Southern Ocean

To assess zonal temperature and biogeographical patterns in the Southern Ocean during the Paleogene, we present new multi‐proxy air‐ and sea‐surface temperature data for the latest Paleocene (∼57–56 Ma) and the Paleocene‐Eocene Thermal Maximum (PETM; ∼56 Ma) from the northern margin of the Australo‐Antarctic Gulf (AAG). The various proxy data sets document the well‐known late Paleocene warming and, superimposed, two transient late Paleocene pre‐cursor warming events, hundreds of kyr prior to the PETM. Remarkably, temperature reconstructions for the AAG and southwest Pacific during the latest Paleocene, PETM and Early Eocene Climatic Optimum (∼53–49 Ma) show similar trends as well as similar absolute temperatures east and west of the closed Tasmanian Gateway. Our data imply that the exceptional warmth as recorded by previous studies for the southwest Pacific extended westward into the AAG. This contrasts with modeling‐derived circulation and temperature patterns. We suggest that simulations of ocean circulation underestimate heat transport in the southwest Pacific due to insufficient resolution, not allowing for mesoscale eddy‐related heat transport. We argue for a systematic approach to tackle model and proxy biases that may occur in marginal marine settings and non‐analog high‐latitude climates to assess the temperature reconstructions.

(Dunkley Jones et al., 2013;Frieling et al., 2017;Hollis et al., 2019;Lunt et al., 2021;Zhu et al., 2019). This suggests that such models provide accurate reflections of global climate states and meridional gradient changes under high radiative forcing. However, even in simulations where the majority of the reconstructed sea-surface temperature (SST) patterns and deep-ocean temperatures are consistent with model output throughout the Eocene, absolute temperature reconstructions from several regions, notably the Arctic and the southwest (SW) Pacific Ocean, and in particular the area around the Tasmanian Gateway (TG) and Zealandia, are still much (>10°C) warmer than those in the simulations (Cramwinckel et al., 2018;Evans et al., 2018;Frieling et al., 2017;Lunt et al., 2021). In contrast to the reconstructions showing exceptional regional warmth, the Southern Ocean, including the south Pacific Ocean, was likely the dominant locus of cold deep-water formation during much of the Paleogene (Huck et al., 2017;Pak & Miller, 1992), highlighting the importance of resolving the paleoceanography and mechanistic understanding of the enigmatic warmth in and around the SW Pacific (Bijl et al., 2009;Douglas et al., 2014;Hollis et al., 2009).
Regional paleoceanography, especially the gradual opening of the TG, would have affected regional temperature trends through re-routing of warm versus cold ocean currents (Bijl, Bendle, et al., 2013;Bijl, Sluijs, et al., 2013;Cande & Stock, 2004;Sijp et al., 2014Sijp et al., , 2016. Recent efforts to constrain the consequence of gradual gateway opening suggest that the regional climatic impact of both the Drake Passage and TG is limited, unless both allow relatively deep throughflow simultaneously (Sauermilch et al., 2021), a situation that does not seem to have occurred until ca. 26 Ma (van de Lagemaat et al., 2021). Therefore, even if a shallow (<300 m water depth) connection existed during the Paleocene and early Eocene (Bijl, Bendle, et al., 2013;Sauermilch et al., 2021), it should have had a negligible impact on paleoceanography and heat transport. On the other hand, in very high-resolution model simulations (<1°), warm mesoscale eddies reach further south than any current in low-resolution runs. Therefore, substantial differences in modeled surface-water temperatures in parts of the SW Pacific arise between the low-resolution and eddy-permitting model runs. Such oceanographic features can presumably reduce the temperature difference between the SW Pacific and the AAG directly surrounding the TG (Nooteboom et al., 2022).
Unfortunately, however, no well-dated temperature proxy or biogeographical data from the northern margin of the AAG, presumably the warmest place in this region, are available for some key intervals of the Paleogene, notably the latest Paleocene (ca. 57-56 Ma) and the PETM. These periods, along with the EECO, deserve particular attention as they are targeted by community data-model comparison efforts such as DeepMIP (Hollis et al., 2019;Lunt et al., 2017Lunt et al., , 2021. However, the absence of data hampers the comparison of temperatures on both sides of the TG, the reconstruction of regional oceanography, and the establishment of a regional temperature response pattern. Consequently, in-depth assessment of model performance is limited, which is crucial in light of the apparent proxy-model mismatch. To fill this data gap, we present new multi-proxy SST and MAAT estimates for two expanded late Paleocene-PETM sedimentary successions from the AAG, deposited in shallow-marine settings at ∼60°S paleolatitude (Otway Basin (OB), Victoria, Australia, Figure 1) Huurdeman et al., 2021). To reconstruct temperature, we applied a suite of lipid biomarker proxies (Table 1) and palynological tools. We supplement the MAAT reconstructions of the latest Paleocene based on nearest-living relative (NLR) analyses of sporomorph assemblages and branched glycerol dialkyl glycerol tetraethers (brGDGTs) (Huurdeman et al., 2021). In addition, we analyze novel proxies proposed to reflect paleotemperature based on branched glycerol monoalkyl glycerol 10.1029/2022PA004529 3 of 24 tetraethers (brGMGTs) (Baxter et al., 2019;Naafs, McCormick, et al., 2018) and the relative abundance of isoprenoid glycerol dialkyl glycerol tetraethers (isoGDGTs) with five cyclopentane moieties (%GDGT-5) (Naafs, Rohrssen, et al., 2018). Furthermore, we applied the TetraEther index of tetraethers consisting of 86 carbon atoms (TEX 86 ) paleothermometer to estimate sea (sub-)surface temperature (Kim et al., 2010;Schouten et al., 2002) and support this record by assessing the relative abundances of crenarchaeol-isomer to total crenarchaeol (f(cren')) that depends on temperature in culture studies (Bale et al., 2019;O'Brien et al., 2017;Sinninghe Damsté et al., 2012). Dinocysts produced by thermophilic dinoflagellates and mangrove palm pollen were used to acquire minimum temperature thresholds. Temperature estimates for the AAG are paired with previously published data from the SW Pacific to assess temperature differences across the TG.

Materials and Setting
Samples from the Latrobe-1 core (38° 41′ 35″S, 143° 09′ 00″E) and the Point Margaret outcrop (∼3 km east; 38° 43′ 28.8″S, 143° 10′ 35″E) near Port Campbell, Victoria, Australia were analyzed. A description of sample collection, lithology and stratigraphy of the section was given by . Point Margaret, a coastal cliff outcrop section, is composed of shallow-marine deposits marked by a gradual up-section transition from (pro-)deltaic muddy sandstones to sandy silt and mudstones, with 0.5%-3% organic matter (Figure 2, Figure S9 in Supporting Information S1). Combined bio-and carbon isotope stratigraphy suggest that the upper ∼35 m of the Point Margaret section cover at most ∼500 kyr of the latest Paleocene (56.5-56 Ma), and the base of the Paleocene section is deemed younger than ca. 57 Ma . The top of the outcrop is covered by modern soils truncating the "body" of the CIE of the PETM, as suggested by carbon isotope stratigraphy.
The Latrobe-1 core, although in some intervals marked by sparse recovery, shows a similar lithology dominated by sandy-siltstones as the Point Margaret outcrop. The Latrobe-1 core extends into the EECO based on dinocyst and sporomorph biostratigraphy . Sediments were deposited in a subsiding trough system, allowing for rapid and almost continuous sediment accumulation during continental rifting (Sauermilch et al., 2019). In contrast to offshore deposits of the same age, the onshore successions analyzed here were not positioned in the depocenter of the OB and were never deeply buried by later Paleogene or Neogene material (Arditto, 1995). Indeed, organic microfossils suffered minimal oxic degradation or even compression and show no signs of thermal alteration. Since the Latrobe-1 core recovery is sub-optimal in sandy strata and in critical intervals, and since the material has degraded during core storage , we refrain from constructing a composite or spliced section with the Point Margaret succession.
The regional oceanography of the SW Pacific in the early Paleogene is thought to have been characterized by the Antarctica-derived Tasman Current (TC) in the southernmost sector ( Figure 1a, Bijl et al., 2011;Huber et al., 2004;Sauermilch et al., 2019, and references therein). Further north, the influence of the lower-latitude Proto-East Australia Current becomes pronounced. The low latitude Proto-Leeuwin Current entered the AAG in the west and extended progressively further east as the AAG widened during the Cenozoic.
We here focused on the latest Paleocene and PETM  in both sections.

Methods-Palynology
Detailed dinocyst assemblage data were generated for 94 samples from Point Margaret and 20 samples from the Latrobe-1 core. A minimum of 200 and 300 specimens were determined to the species level for dinocysts and  Figure S1 in Supporting Information S1. Other abbreviations: PLC, Proto-Leeuwin Current; TC,Tasman Current; PEAC, Proto East Australia Current. Biogeography and simplified model-based currents; red, purple and blue arrows indicate low-latitude, transitional and Antarctic-derived surface currents and eddies, respectively (Bijl, Bendle, et al., 2013;Nooteboom et al., 2022;Sauermilch et al., 2019). (b) Regional modeled sea-surface temperature (SST). Model run represents a high pCO 2 (6x pre-industrial pCO 2 ) of CESM1.2 (Lunt et al., 2021;Zhu et al., 2019). Note higher SST in the Australo-Antarctic Gulf (box 1) compared to the southwest (SW) Pacific (box 2) at the same latitude.  Hopmans et al. (2004) and Weijers et al. (2006) 0.3/location dependent 5 of 24 terrestrial palynomorphs, respectively, in the palynological samples used for biostratigraphy by . Dinocyst taxonomy follows G. L. Williams et al. (2017) except for the subfamily Wetzelielloideae, where we follow Bijl et al. (2016). We refer to Huurdeman et al. (2021) for details on the pollen and spore taxonomy.

Methods-Organic Geochemistry
A total of 114 samples was analyzed for branched and isoprenoid GDGTs and brGMGTs; 94 for Point Margaret and 20 for the Latrobe-1 core. 54 of the samples from Point Margaret had already been extracted and analyzed for brGDGTs by Huurdeman et al. (2021), and resulting chromatograms were used for identification and quantification of isoGDGTs and brGMGTs here. The remaining samples were prepared following the same procedure as used by Huurdeman et al. (2021). In short, 10-15 g of freeze-dried sediment was extracted using an accelerated solvent extractor using 9:1 (volume:volume) dichloromethane:methanol and subsequently separated into apolar and polar fractions over activated aluminum oxide. The polar fractions, containing the GDGTs and GMGTs were dissolved in 99:1 hexane:isopropanol, filtered over a 0.45 μm polytetrafluoroethylene filter, and measured on an Agilent 1260 ultra-high performance liquid chromatograph-mass spectrometer (UHPLC-MS) with settings according to Hopmans et al. (2016). GDGTs and GMGTs were identified by detecting the [M+H]+ ions in selected ion monitoring mode at m/z 1302,1300,1298,1296,1292,1050,1048,1046,1036,1034,1032,1022,1020,1018, and using m/z 744 for the internal standard (a synthetic C 46 ). A minimum peak area cut-off (3,000 units) was applied for individual components, which here typically amounts to absolute concentrations below 1 ng g −1 dry sediment or ∼10 μg g −1 total organic carbon (TOC).

MBT' 5me and CBT' 5me
The methylation and cyclization of brGDGTs, quantified in the Methylated of Branched Tetraethers (MBT' 5Me ) and Cyclisation of Branched Tetraethers (CBT' 5Me ), respectively, was originally linked to soil temperature and pH (Weijers et al., 2007). Subsequent improvements to the liquid chromatograph-mass spectrometer (LC-MS) methodology resulted in identification and clearer separation of brGDGT isomers with a methylation on the fifth or the sixth position (indicated with an '), which allowed disentangling these two environmental controls on brGDGT distributions .
Here, we apply the MBT' 5Me based on the methylation of 5-methyl brGDGTs to extend the existing late Paleocene and PETM MAAT reconstruction (Huurdeman et al., 2021) to the base of the Point Margaret section. To verify soils as primary source of the brGDGTs in these sections, brGDGT distributions are compared to modern global soil and peat data sets ( Figure S11 in Supporting Information S1) (

brGMGTI and Other brGMGT-Based Proxies
Branched glycerol monoalkyl glycerol tetraethers (brGMGTs) are produced by unknown bacteria and are characterized by an additional covalent C-C bond linking their two alkyl chains (Morii et al., 1998;Schouten et al., 2008). The presence of this bond is thought to improve the stability of the membrane under extreme environmental conditions, such as higher temperature. We identify brGMGT compounds with m/z of 1048, 1034, and 1020 (Table 1). In addition to peats (Naafs, McCormick, et al., 2018) and lake sediments (Baxter et al., 2019), these compounds are also produced and recorded in the marine realm (Liu et al., 2012), particularly in oxygen-minimum zones (Xie et al., 2014). Following the outline in Sluijs et al. (2020), we use the nomenclature of Baxter et al. (2019) to identify the compounds that are used to calculate the brGMGTI and MBT based on H-shaped compounds (HMBT acyclic ) indices (Baxter et al., 2019;Naafs, McCormick et al., 2018). The %brGMGTs relative to total brGMGTs and brGDGTs, the brGMGTI, and HMBT acyclic have all been proposed as temperature proxies (Table 1). Even though several recent papers have reported the occurrence of these compounds in Paleogene marine sediments Cramwinckel et al., 2022;Sluijs et al., 2020), their deep-time application remains largely untested. We here explore the temperature-proxy potential of brGMGTI, HMBT acyclic and %brGMGTs in our shallow-marine setting ( Figure 2, Figures S4-S6 in Supporting Information S1).

TEX 86
The degree of cyclization of isoGDGTs produced by marine Thaumarchaeota is captured in the TetraEther index of tetraethers consisting of 86 carbon atoms (TEX 86 ), and can be used to reconstruct SST . Prior to calculating TEX 86 -based temperatures (Kim et al., 2010;Schouten et al., 2002), we test whether the isoGDGT distributions generally align with those observed for the global core-top database used to build the calibration. In this, deviation from the core-top isoGDGT distributions is indicated by a high value of delta ring index (ΔRI). Such differences in isoGDGT distributions might be related to high soil-derived lipid input, which can be assessed using the branched and isoprenoid tetraether (BIT) index (BIT), or to isoGDGT contributions from methanogens and methanotrophs, which can be identified through the methane index (MI) (Hopmans et al., 2004;Weijers et al., 2011;Zhang et al., 2011Zhang et al., , 2016. Together, these indices aid in identifying samples in which the isoGDGT composition is altered by non-thermal effects and thus may not yield reliable TEX 86 -based SST results. For samples that passed the ΔRI cut-off (<0.3), the ratio of GDGT-2 over GDGT-3 (hereafter [2/3]) is employed to assess possible contributions of isoGDGTs derived from deep-dwelling Thaumarchaeota (Table 1, see also Section 4).

f(cren')
f(cren') might be useful to gain insight into SST trends from samples with substantial non-thermal isoGDGT overprints, as identified by ΔRI (Table 1). This is because crenarchaeol is exclusively produced by Thaumarchaeota (De La Torre et al., 2008;Sinninghe Damsté et al., 2002) and therefore substantially less sensitive to overprints by methanogenic or methanotrophic archaea than GDGT-1, -2, and -3 that make up most of the TEX 86 . Weijers et al. (2006) indeed noted that the abundance of crenarchaeol in soils is relatively low, particularly in more acidic soils.
The temperature dependency of f(cren') may be the result of membrane adaptation within Thaumarcheota populations, as the stereoisomer results in slightly different membrane packing (Bale et al., 2019;Schouten et al., 2013;Sinninghe Damsté et al., 2018). In the natural environment, the temperature dependency may also originate from shifts in the Thaumarcheotal populations, as has been observed in the water column of lake Chala (Baxter et al., 2021). In the latter case, the positive temperature correlation may come from more dominant (sub-)surface-dwelling Thaumarchaeota group I.1b that synthesizes more cren' relative to Thaumarchaeota from group I.1a, for example, due to changing oxygenation, stratification or nutrient distribution within the water column (Baxter et al., 2021). However, strains of Thaumarchaeota group I.1b have so far not been detected in the marine environment.

%iGDGT-5
In addition to the standard marine isoGDGTs that are part of the TEX 86 , we also calculate the abundance of isoprenoid GDGT-5 as a percentage relative to isoGDGTs 1-3 (Naafs, Rohrssen, et al., 2018). The occurrence of GDGT-5 (>1%) is currently restricted to regions with MAAT of at least 19°C and the compound is found almost exclusively in acidic peats. While %GDGT-5 has been applied to Paleogene lignites (Naafs, Rohrssen, et al., 2018), the proxy has not yet been applied in (shallow) marine settings. While we detect and calculate %GDGT-5 here, we note that %iGDGT-5 in marine settings will be influenced by additional non-peat sources of GDGTs 1-3, such as marine Thaumarchaeota, methanogens and methanotrophs. The calculated %GDGT-5 thus represents a minimum estimate of the %GDGT-5 of the initial peat material.

Data Compilation for the Southern Ocean
We focus on late Paleocene, PETM and EECO marine and terrestrial temperature proxy records for the Southern Ocean, ranging from New Zealand in the southwest Pacific Ocean in the east to the AAG in the west, between paleolatitudes of 50°S and 65°S Seton et al., 2012) (Figure 1, Table 1). Both quantitative (GDGT-based proxies, foraminiferal Mg/Ca, NLR-based estimates) and semi-quantitative/qualitative (temperature-indicative dinocyst and mangrove palm pollen) temperature information is incorporated. Absolute temperatures are compared within proxies, within the selected time slices. Time-slices "latest Paleocene" (57-56 Ma) and "PETM" (∼55.9 Ma) were identified based on carbon-isotope data (δ 13 C) if available (i.e., for Ocean Drilling Program (ODP) Site 1172 and Deep Sea Drilling Project Site 277) and/or on pollen-based biostratigraphy. We here use records from the upper Lygistepollenites balmei zone and the Spinizonocolpites prominatus subzone to represent the latest Paleocene and PETM, respectively. Consequently, the included Paleocene data have a maximum age of ∼57 Ma , in compliance with previous compilations and DeepMIP (Dunkley Jones et al., 2013;Frieling et al., 2017;Hollis et al., 2019). Only data comprising the "body", that is, the stable period of anomalously low δ 13 C within the PETM CIE, were used in order to focus on the period of sustained peak warmth. Consequently, the rapid onset and more gradual recovery of the CIE, that is, with δ 13 C values, and potentially temperature, intermediate between background and peak PETM values, were not included. A broad interval covering the EECO, ca. 53-49 Ma, as defined by Westerhold et al. (2018) previously identified by bio-, magneto-and isotope stratigraphy is used for the EECO data compilation (Hollis et al., 2019). As it is often challenging to assign absolute ages for terrestrial deposits, we also include localities that were determined to be "Ypresian" and "Early Eocene" in age (Hollis et al., 2019).

BrGDGTs
Distributions of brGDGTs in sediments from Point Margaret are generally indistinguishable from those in modern soils and peats, with negligible riverine or marine sedimentary contributions throughout the entire succession ( Figure S11 in Supporting Information S1), in line with the findings of Huurdeman et al. (2021) for the top of the succession. We here extended the MBT' 5Me -MAAT estimates for the entire late Paleocene part of the Point Margaret section (Figure 2d). There is a long-term warming trend (∼4°C) from ∼18 to 22°C from the base of the section to the onset of the PETM warming around 50.8 m. We find two minor pre-cursor warming events (PW#1 and PW#2) at ∼33 and 46 m superimposed on the long-term trend, which are also supported by f(cren') and dinocysts. The NLR record only shows a response during the first pre-cursor warming (Figures 2b  and 2c). Potentially similar fluctuations in MBT' 5Me further down the section (e.g., ∼27 m, Figure 2d) cannot be confirmed as warming events, as these fluctuations are not mimicked by the trends in other proxies. While the amplitude of these warming events does not exceed the calibration error of the temperature proxies (Figure 2), the analytical precision for the lipid-based parameters (f(cren'), MBT' 5Me ) is at least an order of magnitude higher and sufficient to detect more subtle temperature signals. While we cannot provide absolute estimates of the magnitude of warming during the events, the correspondence of several proxies and analytical precision of the methods strongly suggests that these periods represent anomalous warming.

Isoprenoid GDGT-5
In the Point Margaret sediments we find high %GDGT-5 (0%-10% Figure 2e). The occurrence and abundance of this compound in our shallow marine setting is likely explained by substantial input of peat-derived material, consistent with the presence of Sphagnum spores (Huurdeman et al., 2021). Although there is substantial scatter and a few (n = 13) samples without GDGT-5 throughout the section, GDGT-5 is mostly present (n = 81) and comprises up to ∼10% of the isoGDGT assemblage (GDGTs 1-3) (Figure 2, Figure S9 in Supporting Information S1). Such %GDGT-5 values are close to the maximum percentage observed for modern tropical peats (Naafs, Rohrssen, et al., 2018). In the upper part of the section, %GDGT-5 reaches a maximum of 3% on a relatively stable latest Paleocene (43-50 m height) average ∼1.5%. In this interval, %GDGT-5 appears to broadly follow the rise and fall in MBT' 5Me . However, a correlation between these proxies is absent as the trends in %GDGT-5 and MBT' 5Me are stratigraphically offset.
However, as GDGTs 1-3, used to calculate %GDGT-5, are also derived from marine Thaumarcheota and possibly other non-terrestrial sources ( Figures S4 and S8 in Supporting Information S1), the calculated %GDGT-5 for all our samples must be regarded as a minimum estimate of the original peat-derived source. Considering that the calculated terrestrial fraction of GDGTs 1-3 is ∼30%-50% ( Figure S8 in Supporting Information S1) in some late Paleocene intervals, the relative abundance of GDGT-5 in the source material could well have been similar to, or higher than some modern tropical peats (Naafs, McCormick, et al., 2018;Naafs, Rohrssen, et al., 2018). A single-point within the body of the CIE reaches ∼6 %GDGT-5 ( Figure 2e). However, it is difficult to gauge the value of this observation: starting from the onset of the CIE, %GDGT-5 and Sphagnum spores are no longer consistently present. The erratic occurrence and abundance of GDGT-5 may have occurred through sea level rise so that Point Margaret becomes relatively more distal. GDGT-5-containing terrestrial material is deposited nearer to shore, becomes extremely diluted by marine GDGTs 1-3 or the peat lands are lost as a consequence of warming. The greater abundance of marine dinocysts and lower BIT index values ( Figure S9 in Supporting Information S1) and the occurrence of a fern-spike (Huurdeman et al., 2021) in this interval suggests that changes in both the terrestrial and marine realm may have affected the %GDGT-5 record across the PETM in this shallow-marine setting.

BrGMGTs
The Point Margaret section yields a suite of brGMGTs that have recently been documented in mineral soils and river sediment (Kirkels et al., 2022), peats (Naafs, McCormick, et al., 2018;Tang et al., 2021), Paleogene lignites (Naafs, McCormick, et al., 2018), African lakes (Baxter et al., 2019) and both modern marine surface (Liu et al., 2012) and Paleogene marine sediments Cramwinckel et al., 2022;Sluijs et al., 2020). Naafs, McCormick, et al. (2018) show that the relative abundance of brGMGTs over regular brGDGTs is positively correlated with MAAT in peats. In the Point Margaret section, the amount of brGMGTs relative to brGDGTs is positively correlated to MBT' 5Me , but is marked by higher values (>10% of total brGDGTs) than all modern peat samples at the same MAAT ( Figure S5 in Supporting Information S1). This may indicate that we significantly underestimate absolute temperatures or that additional sources, for example, marine (sedimentary) organisms, add brGMGTs in our setting (Baxter et al., 2021;Kirkels et al., 2022). Baxter et al. (2019) calibrated the brGMGT distributions, formulated as brGMGTI, in tropical lake sediments to temperature. Although the application of the brGMGTI proxy outside tropical lakes is unvalidated, we note that the brGMGT-derived MAAT response at our site follows the trends of other temperature proxy records, although the amplitude is somewhat greater than that based on MBT' 5Me ( Figure 2d). We also find high scatter in the brGMGTI record in some intervals, particularly around the onset of the PETM CIE, 49-51 m, which may suggest the mixing of different brGMGT sources, likely peat and in situ marine.
Despite the relatively limited range of temperature covered in our data, the HMBT acyclic (Naafs, McCormick, et al., 2018) shows a strong correlation with for example, MBT' 5Me and f(cren') ( Figure S7, Text S1 in Supporting Information S1). However, accurately assigning variability to either air or SST, or another, indirect control is difficult without proper source identification (Kirkels et al., 2022). Collectively, it is noteworthy that both HMBT acyclic and brGMGTI follow MBT' 5Me trends and that brGMGTI produces similar MAAT despite the lack of an environment-specific calibration.

Terrestrial Palynomorphs and NLRs
The expanded late Paleocene record ( Aside from the NLR approach to estimate MAAT, we use the well-known climatic envelopes of fossil pollen taxa such as S. prominatus (Nypa), a mangrove palm that only occurs in regions with MAAT > 22°C at present (Reichgelt et al., 2018). This species has a first consistent appearance at 50.57 m at Point Margaret (Figures 2f), 23 cm below the CIE onset (Huurdeman et al., 2021), and occurs within the PETM CIE body (299.67 m below surface) and the EECO in the Latrobe-1 core. A single occurrence of Nypa is registered during the first pre-cursor warming (∼33 m). The presence of Nypa implies that coastal MAAT was, at times, at least 22°C in the northern AAG just prior to and during the PETM and during the EECO.

TEX 86
We generated isoGDGT data for the Point Margaret outcrop and Latrobe-1 borehole. The isoGDGT distributions in most samples from the Point Margaret section and the Latrobe-1 bore have ΔRI values >0.3 signaling non-pelagic contributions to the isoGDGT pool (Zhang et al., 2016). In most samples, soil-derived isoGDGT input, as indicated by the BIT index and contributions from methanogenic and methanotrophic isoGDGT producers, as derived from the MI exceeded generally proposed cut-offs (0.3 for BIT (Hopmans et al., 2004;Weijers et al., 2007) and 0.3 for MI (Zhang et al., 2011)) ( Table 1). GDGT concentrations in the Latrobe-1 samples are <1 ng g −1 sediment or <10 mg g −1 TOC for most compounds, generally sufficient to identify, but not properly quantify, isoGDGTs, and insufficient to identify penta-and hexamethylated brGDGTs and brGMGTs. This implies that brGDGT distributions could not be used to calculate MBT' 5Me and MAAT for these samples. We also note that the recorded GDGT concentrations are remarkably low compared to the sediments from the nearby (∼3 km) Point Margaret, which may be the result of oxidation during long-term (40-50 year) dry storage. A similar effect was noted for dinocysts .  (Kim et al., 2010) are ∼27°C for the late Paleocene and ∼32°C during the PETM, implying a 5-6°C warming during the PETM.

f(cren')
In our samples that are marked by very high terrestrial input f(cren') may provide supporting information on relative SST changes. The temperature-dependency of f(cren') is based on the link between temperature and f(cren') in cultures as well as the modern core-top calibration data set (Bale et al., 2019;Kim et al., 2010;Schouten et al., 2002;Tierney & Tingley, 2015). In the core-top data, SST explains a substantial part of the variability in f(cren') (linear R 2 = 0.6) in waters with an SST above 10°C, although this is slightly less than the traditional TEX 86 (linear R 2 = 0.75) in the same data set Tierney & Tingley, 2015) ( Figure S1 in Supporting Information S1). Moreover, at nearby ODP Site 1172 Sluijs et al., 2011), virtually all variation in TEX 86 is captured (R 2 = 0.98; Figure S3 in Supporting Information S1) by f(cren') across the PETM.
The concentration of crenarchaeol decreases with increasing soil acidity and crenarchaeol rarely exceeds a relative abundance of 10% of isoGDGTs in acidic soils (pH < 6). The same applies to peats and Paleogene lignites, where crenarchaeol rarely exceeds 5% Naafs, McCormick, et al., 2018;Naafs, Rohrssen, et al., 2018). Based on the cyclization of brGDGTs (quantified in the CBT' index; De Jonge, ; De Jonge, Stadnitskaia, et al. (2014)), we find that soils with low pH (values <5 in Paleocene, ∼5.5 during the PETM) dominate the brGDGT distribution at our site (Huurdeman et al., 2021). Collectively, the brGDGTs appear predominantly derived from acidic soils and peats. When we model the contribution of terrestrial isoGDGTs from brGDGT using the approach of Sluijs et al. (2020), we find that the majority of crenarchaeol in our samples is most likely derived from marine Thaumarchaeota and not soils, despite high BIT ( Figure S8 in Supporting Information S1).
In the Point Margaret section, we therefore employ f(cren') to assess the temperature trends in the marine realm (Figure 2b). Similar to the data from ODP 1172, in Point Margaret, f(cren') show a correlation with the scarce TEX 86 data (R 2 = 0.96, p = 0.002, n = 5, Figure S3 in Supporting Information S1) and broadly reproduces the subtle long-term rise in MBT' 5Me -and NLR-derived temperatures in the late Paleocene (0-50 m), as well as the two late Paleocene transient precursor warming episodes (∼33 and 46 m) (Figures 2b and 2f). The precursor warming events are pronounced in f(cren'), whereas the response in brGDGTs and NLR appears more subdued.
Moreover, f(cren') rises just before (50.57 m) the onset of the CIE (50.8 m), whereas the rise in MBT' 5Me slightly lags the CIE (∼51 m). SST rise directly prior to the CIE, as recorded here in f(cren'), has also been recognized elsewhere (Frieling et al., 2019;Secord et al., 2010;Sluijs et al., 2007;Thomas et al., 2002). The presumed temperature signal obtained from f(cren') is supported by the coeval, consistent appearance of Nypa pollen (Huurdeman et al., 2021, Figure 2f). The delayed response (up to a few kyr) in MBT' 5Me compared to vegetation-derived MAAT was attributed by Huurdeman et al. (2021) to (a) differences in transport time and/or (b) reworking of Paleocene or even older soil materials (John et al., 2012) and clay-bound organic matter (Schneider-Mor & Bowen, 2013), including brGDGTs. The different transport time may result in an apparent delay in warming in peat and soil-derived components (brGDGTs, and (peat-derived) brGMGTs). Warming based on above ground vegetation (palynomorphs), especially coastal elements (mangrove palms) and marine compounds (f(cren'), dinocysts), would be more synchronous with the warming. These processes may have also played a role in suppressing the temperature change inferred from MBT' 5Me relative to other proxies during the pre-cursor warming events.

Dinocysts
The dinocyst assemblages at Point Margaret are dominated by low-salinity tolerant taxa throughout most of the late Paleocene section (<50 m) ( Figure S9 in Supporting Information S1) (Frieling & Sluijs, 2018). Intervals with higher abundances of inner-neritic or coastal taxa are also recorded, likely signaling periods of diminished freshwater influx. Thermophilic dinocyst taxa, here mainly Apectodinium, become regular constituents of the assemblage around ∼26 m and open marine taxa (e.g., Spiniferites spp.) are progressively more abundant upsection and suggest more distal or open marine conditions.
The late Paleocene and PETM SST trends as reconstructed through f(cren') are supported by progressively higher percentages of thermophilic dinocysts toward the top of the Paleocene section (Figure 2c), even though these taxa are outnumbered by low-salinity tolerant taxa during pre-CIE warming and onset of the PETM CIE (ca. 50-50.9 m, Figure S9 in Supporting Information S1). In addition to rough trends, the appearance and relative abundance of selected extinct thermophilic dinocysts, notably Apectodinium spp. and F. reichartii provide constraints on minimum SST (Frieling & Sluijs, 2018). The first abundance events of Apectodinium (>10%) are found during the precursor warming events in the latest Paleocene (at ∼33 and ∼46 m, Figure 2c) at Point Margaret. These events are not registered in the Latrobe-1 core, which may be due to low sampling resolution and/or poor recovery in the respective core intervals. A third abundance increase is recorded at Point Margaret during the CIE, which is mirrored by a similar event in the Latrobe-1 core ( Figure S10 in Supporting Information S1). F. reichartii is never abundant (maximum: 5% at Point Margaret), and occurs consistently only during peak CIE. A single late Early Eocene (EECO) abundance event of Apectodinium is found in the Latrobe-1 core. Following observations of Frieling and Sluijs (2018) we arrive at most likely minimum SST estimates ∼20-25°C for the latest Paleocene (based on occasional Apectodinium abundance; Latrobe-1 core and Point Margaret), 25-30°C for the PETM (based on F. reichartii) for the Point Margaret section and 20-25°C for the EECO in the Latrobe-1 core. At Point Margaret, the relative abundance of these thermophilic taxa follows the long-term late Paleocene SST rise, as well as short-term variations ( Figure 2) observed in other temperature proxies (f(cren'), MBT' 5Me , brGMGTI) in detail except for a short interval around the CIE onset. The broad patterns in thermophilic dinocysts and fcren' seem to be confirmed by the limited Latrobe-1 core data ( Figure S10 in Supporting Information S1), with the note of caution that both the dinocyst and lipid biomarker record may be compromised by degradation during core storage.

Integrated Regional SST for the Australo-Antarctic Gulf and SW Pacific
The early Paleogene climate of the SW Pacific has been intensely studied with a range of proxies (Hollis et al., 2019). The majority of SST data is based on TEX 86 , and second planktonic foraminiferal Mg/Ca ratios. Briefly, the SW Pacific TEX 86 H and Mg/Ca records show SSTs of ∼26-30°C in the late Paleocene (Table 2), rising to 31-33°C during the PETM. For the EECO, results are somewhat more variable and carbonate-based proxies show somewhat lower temperatures on average (∼26°C) compared to TEX 86 H (31-32°C) (Hollis et al., 2019).
Fewer data were available for the AAG and prior to this study, none for the late Paleocene and PETM. Although some caution is warranted due to high BIT in our samples, the new data suggest that TEX 86 -based SSTs in the late Paleocene (27°C), PETM (∼32°C) and the published data from Site U1356A (32°C) were indistinguishable from those in the SW Pacific in the same intervals. This is supported by semi-quantitative lines of evidence, particularly the occurrence and abundance of thermophilic dinocysts; the abundance of Apectodinium and occurrence of F. reichartii during the PETM are mirrored east and west of the TG (Figures 3 and 4). Similarly, high SSTs during the EECO of Site U1356 are accompanied by high relative abundances of Apectodinium.  Bijl, Bendle, et al. (2013), Bijl, Sluijs, et al. (2013) Note. Data sources: (Bijl, Bendle, et al., 2013;Bijl et al., 2009Bijl et al., , 2021Carpenter et al., 2012;Contreras et al., 2014;Crouch et al., 2014;Greenwood et al., 2003;Hines et al., 2017;Hollis et al., 2009Hollis et al., , 2012Hollis et al., , 2015Huurdeman et al., 2021;Inglis et al., 2015;Pancost et al., 2013;Pross et al., 2012;Reichgelt et al., 2022;Sluijs et al., 2011).

Integrated Regional Mean Air Temperatures for the Australo-Antarctic Gulf and SW Pacific
Available late Paleocene MAAT reconstructions for the SW Pacific region are mostly derived from vegetation-based (e.g., NLR, leaf-margin analyses) approaches. MAATs have been reconstructed for several localities, but, due to the nature of the proxies, MAAT is available for a relatively small number of samples per location compared to SSTs. Localities include Konkon-1 and Poonboon-1 in the Bass Basin (e.g., Contreras et al., 2014), and Cambalong Creek, on the southeast Australian coast (Greenwood et al., 2003) (Figure 1), Site 1172 and Mid-Waipara, New Zealand, which together arrive at an average of ∼16°C, with MAAT rising to ∼20°C during the PETM. The EECO MAAT estimates are based on MBT' 5Me from Site 1172 (Bijl, Bendle, et al., 2013;Bijl et al., 2021) and Mid-Waipara , and yield MAAT of ∼21-22°C.
Within the AAG realm, late Paleocene MAAT reconstructions are now available for the Point Margaret outcrop (Huurdeman et al., 2021; this study) and Latrobe-1 (this study). The abundance of GDGT-5 in the uppermost Paleocene of the Point Margaret outcrop indicates MAAT > 19°C, the NLR-based estimates (Latrobe-1 & Point Margaret) are 16-19°C, and the Point Margaret MBT' 5Me estimates are 21-22°C. Both NLR and MBT' 5Me indicate a MAAT rise during the PETM to 20-22°C and 23°C, respectively. BrGMGTI derived MAAT estimates suggests slightly higher temperatures during the PETM (∼24°C) and a temperature increase comparable to NLR-based estimates (∼4°C), although these results should be treated with some caution, as brGMGTI estimates were only calculated for Point Margaret and this novel proxy remains unvalidated in Paleogene marine settings. MAAT estimates for the EECO are derived from only few localities (Table 2). This includes Lowana Road, also known as Regatta Point, in the Sorrell Basin, western Tasmania, Site U1356A on the Antarctic Margin and two recent ensemble (NLR and leaf-morphology) estimates from Dinmore and Deans Marsh, Australia   (Table 2). The localities show somewhat divergent plant-based MAAT estimates; NLR shows MAAT ∼18°C at Site U1356A, while a higher MAAT is reconstructed for Lowana Road (∼24°C). The ensemble MAAT estimates from Dinmore and Deans Marsh fall between these estimates. MBT' 5Me estimates from Site U1356A align with the average of all vegetation-based estimates (∼20-21°C). Similar to the SST estimates, MAAT estimates from the SW Pacific and AAG within the same proxy are indistinguishable for the same intervals.  Bijl et al., 2021;Sluijs et al., 2011). (c) Global deep ocean carbon and oxygen isotope stack based on benthic foraminifera (Westerhold et al., 2020). Events were correlated based on succession of dinocyst events and the PETM carbon isotope excursion (CIE) (∼56 Ma) and absolute dates from Point Margaret (first occurrence of G. australiformis; ) and magnetochron C25n-C24r reversal at Site 1172 . Note that sea-surface temperature at Point Margaret is represented by only five TEX 86 estimates, four within the body of the CIE and one in the late Paleocene. Relative abundances for thermophilic dinocysts and Apectodinium are given in % of total dinocysts. The rarer Florentinia is plotted amplified 10-fold.
Terrestrial micro-and macrofossil evidence yields a very similar picture: mangrove palm pollen (Nypa) are found throughout the entire studied area during the PETM (Figure 4) and also appear during the EECO (e.g., Latrobe-1 core and Lowana Road (Carpenter et al., 2012)).

Late Paleocene Warming Events
Multiple, independent, temperature proxies reflect two late Paleocene transient warming events superimposed on subtle or step-wise long-term warming (Figure 2). While the amplitude of warming does not exceed the calibration error for some proxies (NLR, MBT' 5Me ), the correspondence between the various proxies gives confidence these intervals represent pre-cursor warming events. Importantly, these events do not seem to be local. The second precursor warming (PW-2) at ∼46 m in the Point Margaret section has an equivalent at Site 1172 ( Figure 3). This may also hold true for the event at ∼33 m, although at Site 1172 only very subtle increases in Apectodinium abundance (∼614.5 m below sea floor) are registered . Available bio-and magnetostratigraphic age-depth tie points indicate average late Paleocene accumulation rates were 0.6 cm kyr −1 (Sluijs et al., 2011) at Site 1172 and ∼7 cm kyr −1  Point Margaret. Based on these values, PW-2 may precede the PETM by some 100 kyrs. By extrapolation this would imply that PW-1 at 33 m is another ∼200 kyr older, but we note that in marginal settings sedimentation rates can strongly vary on short timescales. Notably, the second precursor warming interval coincides with an increase in open marine and thermophilic dinocyst percentages at the expense of low-salinity tolerant taxa. This suggests more distal conditions or reduced fresh water and siliciclastic flux (Figure 2, Figure S9 in Supporting Information S1). It thus seems likely the second precursor warming interval (PW#2, Figure 2) at Point Margaret represents a period of sediment starvation resulting in a condensed section. At the equivalent level, a distinctly more open marine dinocyst signal is also seen at Site 1172 and no sediment starvation is apparent there. This implies the warming signal appears independent of fluctuations in sedimentation.
Although their exact timing remains unclear, both pre-cursor warming events fall within in an interval with minor deep ocean carbon isotope fluctuations (Cramer et al., 2003;Westerhold et al., 2018Westerhold et al., , 2020 (Figure 3c). Given current constraints on their age, PW#1 and #2 appear to precede previously recognized precursor carbon isotope events that occur closer (<100 kyr) to the PETM CIE (e.g., Babila et al., 2022;Bowen et al., 2015). Although the raw data suggests that a subtle decrease in δ 13 C org co-occurs with the precursor warmings, this could also result from coeval small changes in organic matter sourcing. We find a greater proportion of marine dinocysts as well as slightly lower BIT index values during these events ( Figure S9 in Supporting Information S1), which is particularly relevant as a higher proportion of marine organic matter could skew δ 13 C org to more depleted values (Sluijs & Dickens, 2012). Therefore, evidence for any CIE occurring at the same level as either of the transient precursor warming events is weak. However, as small δ 13 C fluctuations are notoriously difficult to detect in organic carbon we cannot exclude a relation between deep-ocean events and the pre-cursor warming events.
Even if the relation of these subtle transient warming events to the variability recorded in the deep ocean is difficult to constrain, the existence of such events is noteworthy as they exceed the (regional) variability observed in most of the Paleocene (Figure 2). Although these events can only be revealed in high-resolution data generated for background climates, such data is currently scarce. Yet, resolving such signals from background noise could prove essential to understand (Paleogene) climate and carbon cycle behavior (Armstrong McKay & Lenton, 2018;Bowen et al., 2015;Sluijs et al., 2007).

Potential for brGMGT Proxies
Collectively, we find that the strong correlations with other reconstructed environmental parameters including MAAT and SST support a temperature-related response in brGMGTs. However, their common presence in lakes, peats and marine sediments implies that it is challenging to accurately assign observed variability to either air or SST, or other parameters indirectly related to temperature (Kirkels et al., 2022). Despite this, we note that both HMBT and brGMGTI follow MBT' 5Me -based MAAT trends and brGMGTI produces similar absolute MAAT estimates despite the lack of an environment-specific calibration.
Intriguingly, the HMBT not only corresponds in trend with MBT' 5Me , but the ratios between the compounds (HMBT acyclic ; H1020c/H1020c + H1034a + H1048) are also virtually identical to MBT acyclic (defined as brGDGT-Ia/brGDGT-Ia + IIa + IIIa, Table 1), supporting the notion that these compounds have a shared origin and/or mechanistic purpose in microbial membranes. In addition, the increase in brGMGT abundance relative to that of regular brGDGTs (%brGMGT) across the onset of the CIE may imply that the formation of H-shaped compounds represents an additional temperature adaptation (Morii et al., 1998;Naafs, McCormick, et al., 2018) and/or that production of brGMGTs increased relative to that of brGDGTs in specific (i.e., marine) source areas (Kirkels et al., 2022). Despite these unknowns, the clear correlation to reconstructed environmental parameters and the ubiquitous presence of brGMGTs in these (shallow) marine settings such as sampled at Point Margaret highlight the potential for new paleoenvironment proxies based on brGMGTs once their origin and function are better resolved.

No Temperature Differences Between the Australo-Antarctic Gulf and the Southwest Pacific?
Within proxies, we find that absolute temperatures are within calibration error east and west of the TG. Moreover, marine and terrestrial temperature trends in the AAG and the SW Pacific seem similar for all analyzed intervals (i.e., late Paleocene, PETM and EECO; Figure 4). While paleolatitude is similar, the AAG and SW Pacific are thought to be affected by very different surface ocean currents (Figure 1). It is difficult to reconcile with the proposed large-scale ocean circulation patterns, that is, the warm low-latitude Proto-Leeuwin Current in the AAG and cooler higher-latitude Tasman Current in the SW Pacific, across the analyzed interval (Figure 1). It also contrasts with modeled differences in SST and MAAT between the areas east and west of the TG.
It is remarkable that not only the trends (Figure 3), but also the reconstructed absolute TEX 86 -based temperatures are similar across analyzed sections in the marine realm (Figure 4). While a seasonality bias in SST proxies could affect latitudinal gradients through dominance of warm-season productivity at higher latitudes (e.g., Antoine et al., 1996), it is unlikely that such effects would eliminate zonal differences (Figure 4). Modern examples, such as the SST difference between the eastern and western North Atlantic that can exceed 5°C (e.g., Gouretski & Koltermann, 2004), support the notion that substantial zonal differences, such as those expected across the TG, should be detectable in proxy data.
The reconstructed temperatures and trends for the AAG relative to those in the SW Pacific increase the geographical extent of the discrepancy between modeled and proxy-derived temperatures in the high southern latitudes (Hollis et al., 2012;Lunt et al., 2021). At the same time, the findings for the AAG imply that the model-data discrepancy is not limited to the SW Pacific, but extends into the AAG (Lunt et al., 2021). Moreover, this zonal pattern did not notably change during intervals of both transient (PETM) and multi-million-year global warming (Late Paleocene-EECO). The temperature patterns exist within the marine and terrestrial realms and are evident in fundamentally different proxies for both realms. This reinforces the existence of anomalously high SSTs in the AAG and particular the SW Pacific and it appears unlikely that the discrepancy can be resolved by an improved mechanistic understanding of a single SST proxy. While the proxy-derived MAAT for both regions is within calibration error, the absolute reconstructed MAATs are often ca. 10°C below SST (see also e.g., Bijl et al., 2021) and in relatively close agreement with modeled MAAT at high pCO 2 (e.g., Lunt et al., 2021;Reichgelt et al., 2022).
It remains uncertain how accurate the reconstructed absolute mean annual temperatures from the individual proxies are. For example, culture experiments emulating the non-analog high-latitude conditions, such as the seasonal contrasts in light conditions in combination with high-temperature, are yet lacking. Constraining proxy behavior under climate conditions such as those that prevailed in the high southern latitudes during the early Paleogene might prove crucial to assess the value of currently available and forthcoming data. In the following section, we explore and revisit new and previously proposed options that may merit further attention in order to improve our understanding of deep-time high-latitude climate.

Spatial Biases in the Proxy and Modeled Temperature Signals
In general, the inherent heterogeneity of hinterlands and, by extension, sourcing and transport of terrestrial components, particularly pollen and spores, gives rise to several challenges and may complicate a robust comparison between localities (e.g., Inglis et al., 2019). Challenges include changes in the catchment area, including vegetation source, river flow path, coastal proximity, altitude, and spatial integration. While this may affect some interpretations that rely on whole assemblages or presence/absence data (NLR), we suggest that this is likely a relatively minor issue for lowland or coastal taxa and indeed much of the study area. We find that this assumption is warranted by the apparently synchronous appearance of Nypa across the TG, the relatively short time span of the studied time interval, and the fact that all records come from passive margins, implying that major tectonic changes in the catchment area are unlikely. However, comparing to localities further offshore or regions with strong (paleo)relief will invariably include some of these factors.
As the above factors mostly affect terrestrial proxy data, it is unlikely that invoking one single effect (e.g., seasonal biases, sourcing) resolves much of the model-data discrepancy. However, until recently, one effect on marine temperature proxies may have been largely overlooked. There is a dominance of records from near-shore, shallow and coastal environments in the compilation, an inherent (preservation) bias of many deep-time temperature reconstructions. Modern marginal marine settings generally experience greater influence of nearby landmasses and, partly as a consequence, more pronounced seasonal SST variations (∼10°C) compared to open marine or oceanic (typically <5°C) (Hirahara et al., 2014;Judd et al., 2020), and it is reasonable to assume that this was similar in the Paleogene. A greater mean annual temperature range potentially exacerbates any seasonal bias that may exist in proxy data for example, by further amplifying warm-season dominated proxy signals.
Lastly, the low-resolution (1° and greater) models the (paleo)climate modeling community relies on tend to strongly over-or underestimate temperature in specific regions due to lack of fine-scale oceanographic features such as meso-scale eddies. The effects of this are most pronounced in regions associated with eastern and western boundary currents (Judd et al., 2020). Comparing the mostly nearshore paleoclimate reconstructions to low horizontal resolution model simulations may be complicated by such effects (Judd et al., 2020;Nooteboom et al., 2022), especially for regions with complex (paleo)geography. As these factors are challenging to constrain and the impact is likely to be site-specific it is difficult to gauge whether and how this may influence our ability to constrain and compare regional temperature patterns.

Influence of Paleogeography
On a global scale and over latitudinally averaged zones, climate models can now reproduce Eocene proxy data (Cramwinckel et al., 2018;Evans et al., 2018), but an accurate representation of the global, local and regional paleogeography becomes important for finer-scale model-data comparisons (Frieling et al., 2017;Lunt et al., 2016;Nooteboom et al., 2020Nooteboom et al., , 2022. The paleogeography of the region around the TG includes many continental blocks of uncertain paleobathymetry (S. E. Williams et al., 2019), which means that even if fully coupled simulations were to be run in higher spatial resolution, substantial uncertainties in paleobathymetry/paleogeography may still impact temperature distribution. However, extreme end-member early Paleogene (prior to ca. 50 Ma) TG geographies with either deep throughflow or high topography have predictable climatic and oceanographic consequences Sauermilch et al., 2021;Sijp et al., 2011Sijp et al., , 2016 that remain unsupported by the combination of tectonic, biogeographic and temperature proxy data (Baatsen et al., 2018). This implies that such drastic changes in paleogeographic boundary conditions are not primary candidates to resolve the regional discrepancy between data and models.
Although the paleobathymetry of the SW Pacific itself has received less attention than Southern Ocean gateways (Bijl, Bendle, et al., 2013;Lagabrielle et al., 2009;van de Lagemaat et al., 2021), recent work has suggested that sectors of the now submerged continental plates of Zealandia may have been shallow or even emerged above sea level during the Paleogene (Sutherland et al., 2019). The exact influence of bathymetric features on the surface and deep ocean flow and heat distribution in this region is yet unknown, but likely important for the exact configuration and shape of the South Pacific polar gyre and thereby the direction of the Proto-East-Australian Current, as has been argued for other regions of deep-water formation (Coxall et al., 2018;Vahlenkamp et al., 2018). Apart from regional or local details in paleogeography, the use of either a hotspot or paleo-magnetic reference frame for absolute paleolatitude reconstructions may have a large impact on modeled oceanography at the sites used in this study. The type of framework does not notably affect the positions of the sites relative to each other, but the paleomagnetic framework shifts localities around the TG ca. 5 (±5)º latitude north Seton et al., 2012;van Hinsbergen et al., 2015), relative to the spin axis of the Earth. While this may seem trivial, much of the region is within a latitudinal band that is highly sensitive to such changes (Baatsen et al., 2020). Specifically, placing the same regional geography at lower latitudes implies that there is a higher probability of wind-driven surface currents entering the AAG and the SW Pacific through the Proto Leeuwin Current and Proto East-Australia Current, respectively, an effect that is independent of model resolution (Baatsen et al., 2018;Nooteboom et al., 2022). Ultimately, the minor shifts in paleolatitude may therefore have major impact on the origin and temperature of water masses bathing sites east of the TG.

Low-Latitude Current Invasion Into the SW Pacific and Australo-Antarctic Gulf
Intriguingly, recent high-resolution (0.1°) ocean model simulations show an invasive Proto East-Australia current in the middle Eocene, penetrating as far south as ∼55°S (Nooteboom et al., 2022), bringing it within reach of some SW Pacific sites (e.g., Site 277, New Zealand) unlike previous simulations (e.g., Hollis et al., 2012;Huber et al., 2004). A shallow connection between the AAG and the SW Pacific may have existed in the early Paleogene and would be in line with a superficial similarity of the dinocyst assemblages from Site 1172 and Point Margaret and Latrobe-1. However, dinocyst bioprovinces are generally not well-defined in the Paleocene and earliest Eocene, with the majority of taxa likely having a cosmopolitan distribution (e.g., Frieling & Sluijs, 2018), implying similarity on either side did not necessitate an open TG and associated warm or cold through flow.
While the observed biogeographic separation in the Middle and Late Eocene Cramwinckel et al., 2020;Huber et al., 2004) may be interpreted as the expression of a temperature or oceanographic difference, most modern and extinct dinocysts, including thermophilic taxa such as Apectodinium have a wide temperature tolerance (Frieling & Sluijs, 2018;Prebble et al., 2013;Zonneveld et al., 2013). Therefore, it is much more likely that a combination of local environmental parameters, including, for example, nutrient availability, coastal proximity and salinity , ultimately determined the assemblage characteristics and therefore regional biogeography Zonneveld et al., 2013). In this sense, previous interpretations of corresponding modeled high or low SST and biogeography may have overstated the influence of SST on dinocyst biogeography.
With the currently available evidence from emergent high-resolution (0.1°) ocean model runs (Nooteboom et al., 2022) we consider "warm"-current invasion into the SW Pacific and AAG as the leading mechanism for forcing similar temperatures east and west of the TG. This however does not yet explain the extremely high temperatures in the high-latitude AAG or SW Pacific. Particularly the extremely high temperatures obtained from the various SST proxies remain difficult to approach in climate models that, for other regions and for regional MAAT proxies, produce satisfactory results.

Conclusions
The southwest Pacific Ocean (∼50-60°S paleolatitude) was anomalously warm through much of the early Paleogene, and proxy-derived SSTs exceed modeled SST by ∼10°C. Our data extend the area with extremely high proxy temperatures westward into the AAG, with broad implications for reconstruction of meridional temperature gradients and polar amplification that would be based on zonally averaged temperature or temperature patterns and general ocean circulation.
The new multi-proxy temperature records from the AAG reveal a subtle long-term or step-wise Late Paleocene warming on land and in the ocean, and, superimposed, two Late Paleocene transient "precursor" warming events, some ∼300-400 and ∼100 kyr prior to the PETM. The origin, geographical extent and magnitude of these transient events remain uncertain, but the existence of such relatively pronounced (regional) variability is remarkable.
The new data also emphasizes the persistence of high, but similar absolute temperatures and temperature evolution on both sides of a likely closed TG through the warmest periods of the Paleogene (late Paleocene, PETM and EECO). A strong influence of low-latitude ocean currents on both sides of the TG is not expected based on marine microfossil distributions or low-resolution models, yet should not be discarded as a mechanism that contributed to excessive regional warmth and particularly similar temperatures east and west of the TG.
A scenario with (seasonal) low-latitude influences on both sides of the TG may become a preferred scenario when high-resolution, eddy-resolving, modeling can be shown to accurately represent surface water conditions in the Paleocene-Eocene Southern Ocean. Moreover, the difference between low and high-resolution climate model runs may shed some light on SST over-or underestimates east and west of the TG. In addition, a more accurate representation of seasonality in the coastal-marginal marine settings may aid in resolving the influence of proxy biases.
However, even if part of the model-data discrepancy can be resolved by higher-resolution climate modeling, including an accurate representation of paleogeography, it is likely other challenges, such as the offset between SST and MAAT estimates, still limit our understanding of these distinctly non-analog climates as they prevailed in the southern mid to high-latitudes. Some of these directly complicate comparison of proxy data to climate models, such as the influence of paleogeographic and paleobathymetric boundary conditions; factors that are both difficult to reconstruct and to accurately represent in models.