ODP Site 747 (54∘48.68′ S, 76∘47.64′ E; 1695 m water depth) lies on the Kerguelen Plateau in the Southern Ocean (Fig. 1; Schlich et al., 1989). At present, the site is situated south of the polar front and is bathed by Circumpolar Deep Water (CDW) with a temperature of ∼ 1–2 ∘C (Belkin and Gordon, 1996; Billups and Schrag, 2002). The middle Miocene geographic position of Site 747 relative to Antarctica was similar to today (e.g. Abrajevitch et al., 2014) with a paleolatitude between 51 and 56∘ S at 16–12 Ma (van Hinsbergen et al., 2015; Torsvik et al., 2012). Middle Miocene benthic foraminiferal species found at Site 747 are indicative of a lower bathyal to abyssal depth at that time (Schlich et al., 1989). The clumped isotope record generated in this study covers the depth interval from 62.64 mb.s.f. (metres below sea floor, Sample 747A-7H-5, 14–16 cm) to 85.36 mb.s.f. (Sample 747A-9H-8, 75–77 cm) in Hole 747A. In total, 191 samples (15–20 cm3, mostly calcareous nannofossil ooze with foraminifera) were taken continuously with a mean temporal resolution of around 20 kyr (Table S1 in the Supplement). We slightly rescaled the originally assigned shipboard sample depths to account for core expansion (Table 1 of Schlich et al., 1989), similar to previous studies focusing on the middle Miocene section of Hole 747A (e.g. Abrajevitch et al., 2014; Majewski and Bohaty, 2010).

We revised the Hole 747A age model by integrating six magnetostratigraphic tie points (Abrajevitch et al., 2014; Majewski and Bohaty, 2010) on the GTS2012 timescale (Gradstein et al., 2012), three benthic foraminiferal δ13C-based tie points associated with the “Monterey” carbon isotope excursion (using the nomenclature of Holbourn et al., 2007), and one peak warm event visible in benthic foraminiferal δ13C and δ18O (Kochhann et al., 2016) (Fig. S1 and Table S2 in the Supplement). For the δ13C-based tie points, we used the high-resolution isotope stratigraphies of IODP Sites U1335, U1337 and U1338 in the eastern equatorial Pacific Ocean (Holbourn et al., 2014; Kochhann et al., 2016; Tian et al., 2018) as a reference (Fig. 2c). In addition, we included a hiatus at the core break between Cores 7H and 8H, identified by previous studies (e.g. Majewski and Bohaty, 2010). δ18O and δ13C time series of Sites 747, 761, 806, U1335, U1337 and U1338 are shown in Fig. 2 with isotope-based age tie points for Site 747 as black crosses. The age models for ODP Sites 761 and 1171 (not shown) are from Leutert et al. (2020). For ODP Site 806, we utilized a previously published orbitally tuned age model from ∼ 14.1 to ∼ 13.3 Ma. For the older and younger parts of the Site 806 record (∼ 16.6–14.1 and ∼ 13.3–11.6 Ma), we updated biostratigraphic events from Shipboard Scientific Party (1991) and Chaisson and Leckie (1993) to the GTS2012 timescale (Gradstein et al., 2012) and applied polynomial curve fits (Fig. S2 and Table S3).

Each sediment sample was freeze-dried, washed over a 63 µm sieve, oven-dried at 50 ∘C and then dry-sieved into different size fractions. We mainly picked tests of Cibicidoides mundulus from the 250–355 µm size fraction for our measurements. For samples with low abundances of benthic foraminifera in this size fraction, the > 355 µm size fraction was also included. The interval from ∼ 16.0 to ∼ 15.3 Ma was additionally complemented with measurements on Cibicidoides wuellerstorfi. No inter-species offsets in benthic foraminiferal Δ47 have been found in previous studies (e.g. Modestou et al., 2020; Piasecki et al., 2019). To assess inter-species δ18O and δ13C offsets, however, both Cibicidoides species were measured separately in 36 sediment samples (Table S1). Middle Miocene benthic foraminifera (and more specifically Cibicidoides) from Site 747 were previously described as well preserved (e.g. Abrajevitch et al., 2014; Billups and Schrag, 2002), and our examination confirms this impression (Figs. S3 and S4). We note that some of the analysed specimens of C. mundulus and C. wuellerstorfi closely resemble the sensu lato morphotype of the respective species (shown in Fig. 2 of Gottschalk et al., 2016).

Prior to isotope analysis, we cracked open the picked specimens and ultrasonicated the test fragments in deionized water (3 × 30 s) and methanol (1 × 10–30 s) to remove adhering sediment. Test fragments were rinsed with deionized water once between each ultrasonication step and at least 3 times at the end of the cleaning. The cleaned test fragments were subsequently oven-dried at 50 ∘C.

Low abundances of carbonate ions containing both 13C and 18O isotopes require stringent analytical procedures and comparably large sample sizes to obtain clumped isotope temperatures that are precise enough for Cenozoic ocean temperature reconstructions. We achieve the necessary precision by averaging over ∼ 30–40 clumped isotope values measured on small (∼ 100 µg) carbonate samples (Fernandez et al., 2017; Hu et al., 2014; Meckler et al., 2014; Schmid and Bernasconi, 2010). Results from adjacent samples are pooled to achieve this number of measurements (e.g. Grauel et al., 2013; Rodríguez-Sanz et al., 2017), due to the generally low abundance of mono-specific benthic foraminifera (allowing for only one to five individual measurements per sample, Fig. S5b). Producing a low-resolution clumped isotope temperature record with this approach yields higher-resolution δ18O and δ13C time series in parallel (Tables S1 and S4).

Clumped isotope measurements were performed using two Thermo Scientific MAT 253 Plus mass spectrometers at the University of Bergen, Norway, and one Thermo Scientific MAT 253 mass spectrometer at ETH Zurich, Switzerland. All mass spectrometers were coupled to Thermo Fisher Scientific Kiel IV carbonate preparation devices. CO2 gas was extracted from carbonate samples with phosphoric acid at a reaction temperature of 70 ∘C. A Porapak trap included in each Kiel IV carbonate preparation system was kept at -20 ∘C to remove organic contaminants from the sample gas (Schmid et al., 2012). Between each run, the Porapak trap was heated at 120 ∘C for at least 1 h for cleaning. Every measurement run included a similar number of samples and carbonate standards. Four carbonate standards (ETH-1, ETH-2, ETH-3 and ETH-4) with different isotopic compositions and ordering states were used for monitoring and correction of the results (Table S5). External reproducibilities (1 standard deviation) in corrected Δ47 values of ETH-1, ETH-2, ETH-3 and ETH-4 were typically between 0.030 ‰ and 0.040 ‰ (Table S6). External reproducibilities (1 standard deviation) for δ18O and δ13C values of the same standards (given relative to VPDB) were 0.03 ‰–0.10 ‰ and 0.02 ‰–0.06 ‰, respectively. More details on isotope analysis and data processing can be found in Appendix A.

We converted the sample Δ47 values (averages over ∼ 30–40 separate measurements each) into temperature (T, in ∘C) using a calibration based on various recent datasets from core-top-derived foraminifera, corrected with the same carbonate standards as used in our study (Eq. 2 of Meinicke et al., 2020): 1T=0.0431×106Δ47-0.1876-273.15. This combined calibration has been recommended for foraminifer samples (Meinicke et al., 2020). We note that the individual datasets in this compilation (Meinicke et al., 2020; Peral et al., 2018; Piasecki et al., 2019) are all in good agreement with a travertine-based calibration (Kele et al., 2015; recalculated by Bernasconi et al., 2018) spanning a wider temperature range (6–95 ∘C). For consistency, previously published Δ47-based ocean temperatures from ODP Sites 761 (Modestou et al., 2020) and 1171 (Leutert et al., 2020) originally based on the travertine calibration were recalculated with the calibration equation of Meinicke et al. (2020) (Tables S7 and S8 in the Supplement).

The Δ47 signal from individual analyses (Fig. S5a) is by nature much noisier in comparison to δ18O and δ13C (Fig. 2), necessitating an averaging of Δ47 over many adjacent samples before interpreting the data in terms of calcification temperature. We have averaged our Δ47 data using two approaches (Fig. 3), each with different advantages: (1) we averaged results from around 30–40 individual measurements from neighbouring samples, avoiding averaging across hiatuses and intervals with no measurements. These BWT averages are shown as filled circles, with horizontal lines indicating the averaging intervals (circles are plotted at average ages of the respective groups of measurements) and vertical lines indicating 68 % (solid) and 95 % (dashed) confidence intervals. The number of measurements used for the calculation of each mean temperature value is listed at the top of Fig. 3. (2) In addition, 400 kyr moving averages based on 30 or more measurements are shown as solid lines, whereas those based on fewer measurements are dotted. The latter approach does not require a decision on each averaging interval and may thus be better suited for inter-site comparison. We note that small-scale features in the moving average curves (around 1 ∘C or less) are likely caused by the scatter in the underlying individual Δ47 measurements and should not be interpreted as real climate signals. Furthermore, signal changes during rapid transitions can be “smoothed out” to some extent. A comparison of our smoothed clumped isotope temperature curves to different LOESS non-parametric regressions of the data is shown in Fig. S6. We propagated analytical and calibration uncertainties in Δ47-based temperatures (as described in Appendix A of this study and the supporting information of Huntington et al., 2009) and report combined uncertainties as 68 % and 95 % confidence intervals.

Δ47-based BWTs were used in combination with benthic foraminiferal δ18O (δ18Oforam) to calculate δ18Obw (reported relative to VSMOW) with Eq. (9) of Marchitto et al. (2014): 2δ18Obw=δ18Oforam+0.27+0.245×BWT-0.0011×BWT2-3.58. For these calculations, benthic foraminiferal δ18O values of the taxon Cibicidoides were averaged over the same intervals as have been used for Δ47 averaging. For Site 747, we used the δ18O values from this study (measured on C. mundulus and C. wuellerstorfi), whereas the foraminiferal δ18O values for Site 761 were compiled from existing studies (Holbourn et al., 2004; Lear et al., 2010; Modestou et al., 2020). Alternative oxygen isotope temperature equations were also tested (Fig. S7).

The isotope records of Site 747 (Fig. 2) display features typical of middle Miocene sequences, including the stepped increase in benthic δ18O across the MMCT and the pronounced δ13C maxima associated with the Monterey carbon isotope excursion (e.g. Holbourn et al., 2007, 2014; Kochhann et al., 2016; Vincent and Berger, 1985). From ∼ 16.0 to ∼ 15.3 Ma, we analysed stable isotope compositions of both C. mundulus and C. wuellerstorfi, allowing for a direct assessment of species-specific effects on the isotopic compositions of these two different epifaunal species (Fig. 2a and c). δ18O values measured on C. mundulus and C. wuellerstorfi appear indistinguishable, whereas a consistent offset of up to ∼ 0.5 ‰ exists between the δ13C values of these species at Site 747. Similar δ13C offsets between C. mundulus and C. wuellerstorfi have been previously observed for the sub-Antarctic Atlantic during the Quaternary (Gottschalk et al., 2016). Our δ13C values from the middle Miocene underscore the need to carefully examine inter-species offsets in δ13C before combining different species to produce a single δ13C curve.

Independent of the averaging approach, Δ47-based BWTs at Site 747 are highest (8.9 ± 1.3 ∘C, uncertainties 95 % confidence level) from around 16.0 to 14.4 Ma during the Miocene climatic optimum (MCO; Fig. 3b). Thereafter, during the early phase of the MMCT, BWTs decrease by 4.2 ± 2.3 ∘C (difference between mean BWT value from ∼ 16.0 to ∼ 14.4 Ma and mean BWT value from ∼ 14.4 to ∼ 13.6 Ma). The cooling appears to partly coincide with an overall increase in benthic foraminiferal δ18O from around 15 to 14 Ma, reflecting bottom-water cooling and/or global ice sheet growth. However, the Δ47-based cooling is much more pronounced than the δ18O data would suggest. Even if the δ18O signal was influenced by BWT only, then the gradual ∼ 0.5 ‰ increase in benthic δ18O would correspond to a cooling of roughly 2 ∘C (e.g. Marchitto et al., 2014). During the subsequent distinct stepped increase in benthic δ18O around 13.9–13.7 Ma, the Δ47-based BWT record on the other hand does not provide any evidence for a significant cooling. To the contrary, the Site 747 Δ47 record reveals a transient warming starting at or just after the stepped benthic δ18O increase. The magnitude of this transient warming in the later phase of the MMCT is 3.2 ± 3.1 ∘C (difference between mean BWT estimates for ∼ 14.0–13.6 and ∼ 13.6–13.2 Ma). The warming appears to some extent also imprinted in the benthic δ18O signal at Site 747, as visible in the slight δ18O decrease between 13.7 and 13.6 Ma. A hiatus prevents us from drawing any inferences about bottom-water conditions from ∼ 13.2 to ∼ 12.6 Ma. In the youngest interval covered by our study (∼ 12.6–12.2 Ma), bottom-water conditions are comparably cold again (5.8 ± 2.1 ∘C).

Comparison of our Δ47-based BWTs from Site 747 with Δ47-based BWTs from Site 761 off northwest Australia in the Indian Ocean (Modestou et al., 2020) reveals good agreement, where temperatures are based on at least 30 Δ47 measurements (solid lines of the moving averages), with the Site 747 BWTs being slightly lower. Temperature averages from < 30 measurements (dotted lines) are less certain and thus not the focus of our interpretation here (see Sect. ). Note that we processed the Δ47 measurement values from Site 761 (Modestou et al., 2020) in the same way as our results from Site 747 (e.g. temperature calibration, smoothing) to optimize comparability of BWTs from these two middle Miocene reference sites. Since modern BWTs at Sites 747 and 761 are similar (∼ 1–3 ∘C; see Fig. 1), we expect middle Miocene temperature differences between Sites 747 and 761 to also be small, although the middle Miocene water depths of these sites may have been somewhat different from today. Our study confirms the similarity of BWTs at these sites for large parts of the studied interval, suggesting a close to modern meridional temperature gradient around 2000 m water depth in a scenario of substantially (by up to ∼ 9 ∘C) warmer bottom waters. Unfortunately, the period of most pronounced BWT change at Site 747 during the MMCT is characterized by very low data density at Site 761, due to low benthic foraminiferal abundances resulting in few measurements, and possibly a hiatus (core break between Cores 5H and 6H from Site 761 around 14.1 Ma). This leaves open the question of whether the substantial early MMCT cooling around 14.5–14.0 Ma and the subsequent warming were restricted to particular regions in the Southern Ocean or whether they were more widespread features.

Interestingly, a Mg/Ca record of the infaunal benthic foraminifer O. umbonatus from ODP Site 806 in the equatorial Pacific (present water depth 2521 m; Lear et al., 2015) indicates BWT trends that are similar to those reconstructed from Δ47 at Site 747 during the MMCT (Fig. 3). Even though the Site 806 Mg/Ca record is of limited temporal resolution (∼ 100–200 kyr), this low-latitude record provides evidence that the early cooling and the subsequent warming reconstructed at Site 747 could have indeed been of larger-scale or even global significance.

Other available Mg/Ca-based BWT records covering the MMCT do not show the same features. Similar to Site 806, Mg/Ca ratios were also measured on the infaunal species O. umbonatus at Site 761 (Lear et al., 2010). This approach yields BWTs that are within uncertainty of those from Δ47 measured at the same site (Fig. S8; regardless of whether or not the Mg/Ca-based BWTs have been corrected for changes in saturation state; Modestou et al., 2020) and show no indication of the substantial BWT changes derived from Δ47 at Site 747 and Mg/Ca at Site 806 (Lear et al., 2015). However, we note that Mg/Ca-based BWT estimates from Site 761 have been deemed less reliable than those from Site 806, due to unusual and variable pore water chemistry at Site 761. Mg/Ca records from Southern Ocean Sites 747 (Kerguelen Plateau; Billups and Schrag, 2002) and 1171 (South Tasman Rise; Shevenell et al., 2008) measured on the epifaunal species C. mundulus also do not show the large temperature swings (Fig. 3). The observed discrepancies between the sites could suggest a regional and/or depth-related differentiation in water mass properties, related to transient ocean circulation changes during the MMCT (see further discussion below). However, especially in the light of the discrepancies between BWTs estimated from Mg/Ca (Billups and Schrag, 2002) and Δ47 at Site 747, another possible explanation is additional non-thermal controls on Mg/Ca and/or Δ47, which may be related to seawater chemistry during test precipitation and/or post-depositional alteration, such as dissolution.

To the best of our current knowledge, seawater chemistry does not appear to significantly influence Δ47 signatures in foraminifera over the range of natural variation (e.g. Tripati et al., 2015; Watkins and Hunt, 2015). On the other hand, it has been shown that Mg/Ca signatures can be affected by changes in seawater Mg/Ca (Evans and Müller, 2012) and carbonate ion saturation (Elderfield et al., 2006; Yu and Elderfield, 2008). On the timescales considered here, the latter is more likely to be important. The relatively few Mg/Ca-based BWTs from Site 747 can be directly compared to our BWTs based on Δ47 from the same site (Fig. 4a). Δ47- and Mg/Ca-based BWTs appear to diverge most pronouncedly in times of increased dissolution (high percentage of benthic foraminiferal tests and fragments), indicating fluctuations in bottom-water carbonate ion saturation (Diester-Haass et al., 2013; Fig. 4b and c). Mg/Ca-based temperatures from Site 747 were measured on foraminiferal tests of the epifaunal species C. mundulus; compared to infaunal foraminifera, this species lives in more direct contact with bottom water and may thus be more prone to saturation-state-related effects (Elderfield et al., 2006; Lear et al., 2015). The observation of diverging Mg/Ca- and Δ47-based BWTs in times of increased dissolution supports the interpretation of a possible saturation state effect on the Mg/Ca signatures of C. mundulus (see Fig. S9 for sensitivity calculation).

In addition to saturation state effects, variable dissolution itself (Fig. 4b and c) could have influenced foraminiferal Mg/Ca and/or Δ47 signatures. For planktic foraminifera, dissolution controlled by bottom-water saturation has the potential to significantly lower initial Mg/Ca signatures and thus also the estimated ocean temperatures in certain burial settings (e.g. Regenberg et al., 2014). Dissolution may also impact the Mg/Ca signatures of benthic foraminiferal tests, although the tests of benthic foraminifera appear generally denser and more resistant to dissolution than those of planktic foraminifera (e.g. Berger, 1973; Pearson et al., 2001). The effects of dissolution on benthic foraminiferal Mg/Ca have thus received little attention. Similarly, dissolution effects on benthic foraminiferal Δ47 signatures have not yet been specifically assessed. While there is currently no evidence for a significant dissolution effect on foraminiferal Δ47 (e.g. Breitenbach et al., 2018; Leutert et al., 2019) or variable dissolution of benthic foraminiferal calcite at Site 747 during the interval of this study (Fig. S4), a potential effect of dissolution cannot be fully ruled out. We thus note that this aspect warrants further study but interpret Δ47-based temperatures as unaffected by dissolution in the absence of indications otherwise. The good agreement of our Δ47-based BWT estimates from Site 747 with the infaunal Mg/Ca BWT record from Site 806 (Lear et al., 2015) lends support to this interpretation.

The observation of a pronounced early MMCT bottom-water cooling and subsequent warming during the later MMCT at Site 747 is surprising and suggests previously unrecognized changes in deep-water properties surrounding one of the major climate transitions in the Cenozoic era. Upper-ocean temperature records from the Southern Ocean are sparse, but existing data (Kuhnert et al., 2009; Leutert et al., 2020; Shevenell et al., 2004) do not show similarity to the temperature pattern we reconstruct for the deep ocean. Instead, the multiproxy temperature dataset from ODP Site 1171 on the South Tasman Rise indicates that the cooling in the upper waters of the Southern Ocean was synchronous with the benthic δ18O increase reflecting a substantial expansion of the Antarctic ice sheet (Fig. 5; Leutert et al., 2020). This observation suggests that the Southern Ocean BWT signal reconstructed from Site 747 benthic foraminiferal Δ47 reflects changes in deep-water properties rather than a high-latitude surface ocean response. Occurring in an interval of overall decreasing pCO2 (Foster et al., 2012; Sosdian et al., 2018; Super et al., 2018), the early MMCT deep-ocean cooling might reflect ice-sheet-related changes in deep-ocean circulation. Recent model results suggest that the spatial extent of the Antarctic ice sheet in particular may have played an important role for BWT during the MMCT because albedo changes affect the hydrological cycle and the regions of deep-water formation around Antarctica (Bradshaw et al., 2021). Alternatively, circulation changes at that time could have been related to tectonic processes accompanying the opening of the Drake Passage and the Scotia Sea (e.g. Dalziel et al., 2013; Lagabrielle et al., 2009; Pérez et al., 2021) and/or the closing of the eastern Tethys gateway (e.g. Hamon et al., 2013; Steinthorsdottir et al., 2020; Woodruff and Savin, 1989). However, large uncertainties in the timing of these ocean gateway changes, which may have affected Southern Ocean bottom waters and Antarctic ice volume to different extents, hamper an unambiguous correlation. Overall throughout the middle to late Miocene, climate modelling indicates that intermediate to deep waters in the Southern Hemisphere may have been warmer than modern values due to differences in ocean currents related to the open Central American Seaway (e.g. Burls et al., 2021). Although this does not immediately help explain the sequence of events observed in our record during the MMCT, it may at least shed some light on the elevated temperatures during the MCO and the rebound to warmer temperatures after the observed cooling at the MMCT.

Given the lack of similar data from a range of locations and water depths, it is difficult to assess how widespread the observed deep-ocean cooling was and whether the cooling reflects variations in the properties of a single bottom-water mass or rather shifts in the boundaries between different water masses. Results from a climate modelling study indicate spatially heterogeneous temperature changes in large parts of the Southern Ocean during the MMCT, caused by a complex interplay between winds, ocean circulation and sea ice (Knorr and Lohmann, 2014). Nevertheless, the similarity between the early MMCT BWT decreases observed at Site 747 in the Southern Ocean and at Site 806 in the deep tropical Pacific (Lear et al., 2015; Fig. 3) suggests that the temperature signal was transferred from the Southern Ocean region covered by our Site 747 record into the Pacific Ocean basin. This interpretation may imply deep-water formation in the Southern Ocean and an ocean gateway configuration similar to today, with an active Antarctic Circumpolar Current and continuous export of deep-ocean water masses formed in the Southern Ocean to lower latitudes.

Compared to the reconstructed early bottom-water cooling, the warming during the later phase of the MMCT starting at or just after the stepped main increase in benthic δ18O (∼ 13.9–13.7 Ma) is even more enigmatic. It could signify a return to the circulation state before the early MMCT bottom-water cooling. Alternatively, substantial ice expansion could have led to increased stratification and shielding of deeper waters in the Southern Ocean, resulting in a warming of these water masses. Majewski and Bohaty (2010) measured δ18O on middle Miocene benthic (Cibicidoides spp.) and planktic foraminifera (e.g. Globigerina bulloides) at Site 747 across the MMCT. These authors documented a marked increase in the calculated δ18O differences between Cibicidoides spp. and G. bulloides (vertical δ18O gradient) during the main increase in δ18O and interpreted this signal as a surface freshening. A freshening in the upper waters of the open Southern Ocean may be related to an increase in meltwater input from a growing ice sheet in Antarctica and possibly the melting of northward-exported sea ice (e.g. Crampton et al., 2016; Sangiorgi et al., 2018; Sigman et al., 2004). An upper-ocean freshening across the MMCT was also reconstructed at Site 1171 (Leutert et al., 2020; Shevenell et al., 2004). At high southern latitudes, salinity has a large effect on stratification (e.g. Kuhnert et al., 2009). We hypothesize that a Southern Ocean freshening concurrent with Antarctic ice sheet expansion may have decreased convective vertical mixing resulting in a shielding of upper-ocean waters from comparably warm deeper waters. This stratification mechanism may have influenced Southern Ocean BWTs during the late MMCT, explaining the transient bottom-water warming and the different trends in upper-ocean temperature and BWT. An increase in stratification starting between 14 and 13.5 Ma is also supported by an increase in dissolution at that time (Figs. 4 and S9), which may be related to reduced ventilation and an increase in CO2 storage in the deep ocean. A similar mechanism may, in principle, have acted in the opposite direction during the earlier cooling.

Further clues can be obtained from the evolution of δ18Obw, which we can calculate from measured benthic foraminiferal δ18O in combination with Δ47-based BWTs (Fig. 5e). Due to the comparably large random errors in our Δ47-based BWT estimates, the propagated uncertainties in δ18Obw are also large. In addition, the foraminiferal δ18O values used in the calculations could also include an ocean pH component (Zeebe, 1999) and/or reflect foraminiferal species-specific effects on 18O fractionation that were different than those included in existing calibrations (e.g. Bemis et al., 1998; Marchitto et al., 2014). However, other systematic biases in our δ18Obw estimates may be smaller compared to alternative methods (e.g. paired benthic foraminiferal Mg/Ca and δ18O measurements) because Δ47 signatures seem to be insensitive to foraminiferal species-specific vital effects and environmental parameters other than temperature (e.g. Leutert et al., 2019; Peral et al., 2018; Piasecki et al., 2019; Tripati et al., 2015; Watkins and Hunt, 2015).

At Site 747, the MCO is characterized by variable δ18Obw values ranging from around -0.1 ‰ to 0.7 ‰ (∼ 16.0–14.4 Ma; Fig. 5e). For the cold BWT period during the MMCT, δ18Obw is overall lower than before with values from around -0.3 ‰ to 0.1 ‰ (∼ 14.4–13.6 Ma), followed by comparably high post-MMCT values of ∼ 0.7 ‰–1.0 ‰ (∼ 13.6–12.2 Ma). All reconstructed δ18Obw values are consistently higher than expected for minimal ice (i.e. -0.89 ‰ according to Cramer et al., 2011). Overall, our δ18Obw values from Site 747 correspond well to those reconstructed with a similar approach at Site 761 (Modestou et al., 2020; Fig. 5) and those based on Mg/Ca BWTs at Site 806 (Lear et al., 2015). Thus, mounting evidence from various sites and proxies suggests that high δ18Obw values represent a robust feature of the middle Miocene. Taken at face value, these results suggest the presence of substantial ice sheets primarily in Antarctica and possibly also on Greenland (e.g. Thiede et al., 2011) in times of warm bottom waters (e.g. Lear et al., 2015; Modestou et al., 2020). Short-lived (orbital-scale) minima in global ice volume during peak MCO interglacials (e.g. Levy et al., 2016) may not be visible in the Δ47-based records from Sites 747 and 761 due to their temporal resolution and possible averaging over glacial and interglacial climate states. In addition to the extent of global ice volume, however, the δ18Obw may also reflect an Antarctic ice sheet oxygen isotopic composition that was different from today (Langebroek et al., 2010) and/or variations in deep-ocean salinity (e.g. Modestou et al., 2020). At Site 747, the latter seems especially likely during the cold BWT period in the MMCT where δ18Obw is low (∼ 14.4–13.6 Ma), possibly reflecting a cold and fresh water mass bathing the site.

The increase in δ18Obw after the cold BWT period likely includes both a salinity and ice volume component, given that it occurs close in time to the main stepped benthic δ18O increase starting between 13.8 and 14.0 Ma (Fig. 5d). This marked feature of the MMCT may reflect an increase in Antarctic ice volume during a prolonged period of low seasonal contrast over Antarctica (declining eccentricity, decreasing amplitude variations in obliquity, Fig. 5a), as pointed out by Holbourn et al. (2005). The inferred ice volume increase is supported by ice-rafted detritus records from two study sites off the coast of East Antarctica – Wilkes Land IODP Site U1356 and Prydz Bay ODP Site 1165 (Pierce et al., 2017) – as well as multiproxy evidence for an episode of maximum ice sheet advance (MISA-4) recorded in the ANDRILL (AND)-2A drill core from the western Ross Sea, Antarctica (Levy et al., 2016). Similarly, an earlier period of maximum ice sheet advance documented in the Ross Sea around 14.7–14.6 Ma (MISA-3) corresponds to a maximum in δ18Obw at Site 747 suggesting a larger global ice volume. Unfortunately, the section from 14.4 to 13.8 Ma is missing in the AND-2A core (Levy et al., 2016) preventing us from a final assessment of the extent to which the minimum in δ18Obw reflects a substantial global ice volume minimum or rather another factor such as low salinity (as discussed above). More proxy environmental data from Antarctica and its continental shelves as well as additional BWT and δ18Obw records from different sites and water depths in the Southern Ocean will allow for a better understanding of the intriguing features of the MMCT recorded at Site 747.