This study utilises material recovered at ODP Site 1264 on the Angola Basin side of Walvis Ridge in the Southeast Atlantic (Fig. 1; 28∘31.955′ S, 2∘50.730′ E, 2505 m water depth; Shipboard Scientific Party Leg 208, 2004a), which was drilled during ODP Leg 208 to provide a Cenozoic deep-sea record of the South Atlantic (Shipboard Scientific Party Leg 208, 2004a, b). At Site 1264, a continuous ∼ 316 m shipboard stratigraphic splice was developed back to the early Oligocene using magnetic susceptibility and 600/450 nm colour reflectance data (Shipboard Scientific Party Leg 208, 2004a). Oligocene–early Miocene XRF ln(Ca / Fe) data were used to verify the shipboard splice and produce a revised composite depth (rmcd) scale (Liebrand et al., 2016, 2018). Liebrand et al. (2016) filled four short Oligocene–early Miocene core gaps at Site 1264 with data from Site 1265 (Fig. 1; 28∘50.101′ S, 2∘38.354′ E, 3059 m water depth; Shipboard Scientific Party Leg 208, 2004b) to provide a continuous sedimentary sequence. Site 1264 and the relevant intervals of Site 1265 are characterised by high biogenic carbonate, with shipboard analysis indicating an average CaCO3 content of 92 %–96 % (Shipboard Scientific Party Leg 208, 2004a, b). At Site 1264, shipboard linear sedimentation rates (LSR) derived from bio-magnetostratigraphy were exceptionally low for the early–middle Miocene (19–12 Ma; LSR = 3.9–5.4 m Myr-1, average of 4.7 m Myr-1). Higher shipboard LSR occurred during the Oligocene–early Miocene (30–19 Ma; LSR = 5.3–9.3 m Myr-1, average of 7.1 m Myr-1) and early Plio-Pleistocene (3–0 Ma; LSR = 4.5–7.4 m Myr-1, average of 6.0 m Myr-1). The highest shipboard LSR occurred in the late Miocene–early Pliocene (12–3 Ma), where LSR average 15.9 m Myr-1 (7.7–30.5 m Myr-1). The shipboard LSR for the Oligocene–early Miocene (30–17 Ma) and Plio-Pleistocene (5.3–0.0 Ma) were confirmed by previous studies on Sites 1264 and 1265, which also support the shipboard notion that these sites are excellent recorders of orbital-scale climate dynamics (Liebrand et al., 2011, 2016, 2017, 2018; Bell et al., 2014, 2015).

To assist with splice verification, astronomical tuning and data interpretation, we compiled composite core images for ODP Site 1264 from both cropped line scan images and core table-top photos (JANUS – http://www-odp.tamu.edu/database/, last access: May 2020 for core table-top photos and November 2014 for line scan images) using Code for Ocean Drilling Data (CODD 2.1 – http://www.codd-home.net, last access: May 2021​​​​​​​; Wilkens et al., 2017). After cropping the original line scan images, core section images were compiled into a single image for each core and scaled to depth using the “Includes_Core_Image_Assembly” functions. The line scan images obtained during ODP Leg 208 “Walvis Ridge” are redder in colour than they appear in the core table-top photos. This is likely an artefact of the line-scanning calibration, which had only recently been introduced at the time (Fig. 1). The individual core images were also compiled from lighting-corrected table-top core box photos using the “Includes_Core_Table_Photos” functions. These table-top core box images more realistically represent the original colour of the cores. However, because the core-box-derived images are exceptionally white, we preferentially use the line-scan-derived images, as the sedimentary cyclicity at Site 1264 is better visible in these images, and thus these are more beneficial to evaluating and revising the stratigraphic splice together with XRF and physical property data (see Fig. 2; Sect. 2.3 and 3.1). The individual core box and line scan core images were then combined into a single composite image along the revised Site 1264 splice using the “SpliceImages” function. The individual core box and line scan core images from Site 1265 (Westerhold et al., 2017) were spliced into the four Oligocene–early Miocene core gaps in the Site 1264 splice image using the Site 1264 to Site 1265 ties from Liebrand et al. (2016) updated to accommodate any splice revisions (see also Sect. 3.1 and Fig. S5 and Table S5 in the Supplement). This resulted in a continuous spliced image of the sedimentary succession at Site 1264–1265 spanning the early Oligocene to present day.

XRF core scanner data were collected at ODP Site 1264 between 0 and 205 rmcd (revised metres composite depth, this study) to connect with previously published XRF core-scanning data spanning 205–315.96 rmcd (Liebrand et al., 2016). The new XRF data were generated in four measurement campaigns in 2011 (195–205 rmcd), 2013 (29.21–153.28 rmcd), 2017 (141.49–195.12 rmcd) and 2018 (0–33.35 rmcd). Ca, Fe, K, Mn, Si and Ti were measured directly at the core surface of Site 1264 archive halves using a 10 kV run at 1–2 cm resolution over a 1.2 cm2 area with a slit size of 10 mm down-core and 12 mm cross-core with XRF Core Scanner II/III (AVAATECH serial no. 2/12) at the MARUM – University of Bremen. In 2018, Ba and Sr were additionally measured with the same slit conditions using a 50 kV run across intervals that proved more problematic to accurately correlate between holes. The following settings were used: in 2011, 10 kV/0.15 mA/10 s count time/Cl-Rh filter on the MARUM XRF III (see also Liebrand et al., 2016); in 2013, 10 kV/0.2 mA/15 s count time/Cl-Rh filter on the MARUM XRF III; in 2017, 10 kV/0.15 mA/15 s count time/no filter on the MARUM XRF II; and in 2018, 50 kV/0.5 mA/7 s count time/Cu filter and 10 kV/0.035 mA/7 s count time/no filter on the MARUM XRF III. The split core surface was covered with a 4 µm thick SPEXCerti Prep Ultralene1 foil to avoid contamination of the XRF measurement unit and desiccation of the sediment. Selected intervals were rerun during successive campaigns to account for differences in measurement intensity between datasets, including with the published Liebrand et al. (2016) data. All data were inspected directly following collection, and outliers were removed if they were clearly associated with cracks and/or uneven sediment surface. A further four measurement points, often at section ends, were removed because intensities were unrealistically low. The XRF datasets collected in 2011, 2013 and 2017 were calibrated to the 2018 dataset using an individual linear regression for each element (see Supplement and Figs. S1 and S2). The 2011 XRF dataset includes the data originally published in Liebrand et al. (2016). All XRF core-scanning intensity data and derived information are reported in Table S1.

Liebrand et al. (2016) showed that the XRF ln(Ca / Fe) ratio shows a strong positive correlation with shipboard coulometric CaCO3 data from Site 1264 (Shipboard Scientific Party Leg 208, 2004a). We use a similar approach to generate a continuous record of CaCO3 content using the combined Site 1264 XRF ln(Ca / Fe) dataset calibrated to the shipboard CaCO3 data (%CaCO3=80.238±1.069+(2.526±0.188×ln⁡(Ca/Fe)); r2 = 0.622; Fig. S6). The Oligocene–early Miocene calibration used by Liebrand et al. (2016) is within the 2σ uncertainty of the new %CaCO3 calibration, which equates to ±2.2 % in the calibrated %CaCO3 dataset. The uncertainty in the calibration likely originates from the scatter of the shipboard coulometry-derived %CaCO3 data that were used in the calibration. This uncertainty only pertains to the absolute %CaCO3 values. The trends and cyclicity observed in the calibrated CaCO3 data are independent of this uncertainty, as these patterns are present in the raw ln(Ca / Fe) time series.

The new and recalibrated %CaCO3 data from Site 1264 were combined with the %CaCO3 data from Site 1265 (Liebrand et al., 2016) to form a ∼ 315 m/∼ 30 Myr continuous record of CaCO3 content at Walvis Ridge. Bulk and CaCO3 mass accumulation rates (MARs; g cm-2 kyr-1) were calculated using the following formulas: 1MARBulk=ρdry×LSR,2MARCaCO3=ρdry×LSR×%CaCO3100,3MARdetrital=MARBulk-MARCaCO3. The LSR (here in cm kyr-1) were calculated using the new astrochronology (Sect. 4). Dry bulk density ρdry (g cm-3) was estimated using the shipboard gamma ray attenuation (GRA) bulk density data calibrated to the shipboard discrete dry density data (Shipboard Scientific Party Leg 208, 2004a) (Fig. S7). The uncertainty in the MARs is difficult to quantify. The largest uncertainties affecting bulk, CaCO3 and detrital MARs arise from uncertainties in the ρdry, which was calculated using shipboard GRA and discrete dry density data, and the LSR, both of which are difficult to estimate. CaCO3 MARs additionally have ±2.2 % 2σ calibration uncertainty. However, as %CaCO3 is so high at Site 1264, the %CaCO3 calibration uncertainty will have a smaller affect compared with the changes in LSR. Because detrital MARs are low and calculated using the difference between bulk and CaCO3 MARs, changes in detrital MARs should be treated cautiously.

The line scan core photos and XRF data, especially the ln(Ca / Fe) and Ba data, show that there are several misalignments in sedimentary features when using the shipboard composite depth scale, leading to duplicated and/or missing intervals in the shipboard splice (Fig. 2). These misalignments are especially pronounced in intervals where the shipboard physical property data was low amplitude due to very high CaCO3 content, e.g. during the late Miocene–early Pliocene interval. Predominantly using the XRF ln(Ca / Fe) ratio, the shipboard splice was verified between 0 and 205 rmcd and revised where needed (Figs. 2 and S3; Tables S2 and S3). Where the inter-hole correlation based on the ln(Ca / Fe) ratio was ambiguous, the Ba data were used (see Fig. S3). Revisions were made between 27 and 149 rmcd and generally resulted in changes of less than 0.6 m relative to the shipboard composite depth (Tables S2 and S3), with the exception where Core 1264A-11H was shifted by -1.26 m relative to Core 1264B-11H to improve the correlation (Fig. 2). The core images, the new ln(Ca / Fe) data, and shipboard 600/450 nm colour reflectance and MS data (Shipboard Scientific Party Leg 208, 2004a) were also used to map off-splice intervals of Holes 1264A and 1264B onto the splice between 0 and 196.13 rmcd (Table S4).

Liebrand et al. (2016, 2018) previously revised the Oligocene–early Miocene interval of the composite depth scale, stratigraphic splice and mapping pairs for Site 1264, as well as the Site-to-Site correlation between Sites 1264 and 1265. The revisions in the upper sedimentary succession (0–205 rmcd) result in a cumulative shift of -1.96 m to the Liebrand et al. (2016, 2018) Site 1264 composite depth scale, stratigraphic splice, mapping pairs and Site 1264 depths in the Site 1264 to Site 1265 correlation between 196.13 and 315.96 rmcd (Tables S2–S5). Furthermore, the line scan composite core photos showed that the mapping of Core 1264B-29H to the splice, erroneously corrected in Liebrand et al. (2018) from original mapping in Liebrand et al. (2016), should be adjusted (see Fig. S4 and Table S4). The composite depth scale and splice revisions resulted in two small gaps in the Plio-Pleistocene Site 1264 isotope record (Bell et al., 2014): ∼ 15 cm (∼ 17 kyr) between 27.25–27.40 rmcd and ∼ 25 cm (∼ 11 kyr) between 54.18–54.43 rmcd, which were filled with new isotope data (Westerhold et al., 2020).

The XRF-derived CaCO3 content at Site 1264 is generally high throughout, ranging between minimum values of 92 % and maximum values of 97 % (Fig. 3). The range of observed %CaCO3 variability is close to the 2.2 % uncertainty associated with the calibration. However, we are confident that both the long-term trends and short-term variability discussed below represent true changes in carbonate content, as these patterns originate in the original ln(Ca / Fe) ratio. The calibration uncertainty is most relevant to the absolute carbonate content. The recalibrated CaCO3 content span 93 %–96 % between ∼ 205 and ∼ 316 rmcd (early Oligocene–early Miocene), which agrees within error with the original calibrated CaCO3 content reported and discussed in detail by Liebrand et al. (2016). The lowest CaCO3 content (92 %–93 %) occurs between 205 and 190 rmcd (middle Miocene). The CaCO3 content shows especially clear 0.5 m cycles in this interval. CaCO3 content increases slightly to 94 % between 190 and 180 rmcd (middle Miocene) and then remains around 94 %–95 % until ∼ 118 rmcd (early late Miocene). The CaCO3 content initially displays short 0.6 m cycles, but after ∼ 185 rmcd (middle Miocene), 0.2–0.3 m cycles are superimposed upon ∼ 1–2 m cycles. CaCO3 content then undergoes a two-step rapid increase to 96 % between ∼ 118 and 110 rmcd and again to 97 % between 105 and 100 rmcd (both latest Miocene). CaCO3 content remains around 97 % until 90 rmcd, after which the CaCO3 content decreases slightly to around 96 % until 40 rmcd (early Pliocene). Between ∼ 118 and ∼ 40 rmcd (latest Miocene–early Pliocene), ∼ 1.0 m cycles and occasionally ∼ 0.5 m cycles are prevalent, although the amplitude of the short-term cycles is reduced compared to the deeper interval. CaCO3 content slowly drops to 95 % by 15 rmcd (Pleistocene) and decrease further to 93 %–94 % in the upper 15 m of the record. Short-term cycles are less well expressed in this upper interval. The Si and K intensities are comparable throughout the record, although Si is generally slightly higher than K (Fig. 3). Both elements, together with Fe and Ti intensities, display the same short-term variability and long-term trends (Figs. 3 and S2), indicating that these elements reflect changes in aluminosilicates. As the trends of Si and K are inverse to those seen in the CaCO3 content, this supports that Site 1264 is predominantly composed of carbonate and clay, with minimal influence of biogenic silica. The amplitude of changes in Si and K becomes much smaller relative to CaCO3 content changes between ∼ 115 and 0 rmcd compared to between ∼ 315 and 115 rmcd.

Because CaCO3 content accounts for > 90 % of the sediment mass, the bulk (0.3–4.7 g cm-2 kyr-1) and CaCO3 (0.3–4.5 g cm-2 kyr-1) MARs are remarkably similar, with trends controlled almost completely by variability in LSR (Fig. 3). LSR also strongly affect detrital MARs; however, these remain low throughout at Site 1264 (0.01–0.2 g cm-2 kyr-1). The Oligocene-early Miocene CaCO3 MARs generally oscillate between 1 and 2 g cm-2 kyr-1 from ∼ 315 to 205 rmcd. MARs are very low (∼ 0.3–0.7 g cm-2 kyr-1) between 205 and 190 rmcd, before slowly increasing to 1.0–2.5 g cm-2 kyr-1 between 190 and ∼ 118 rmcd. The highest CaCO3 MARs (2.5–4.5 g cm-2 kyr-1) occur between ∼ 118 and 35 rmcd, with values decreasing back to 1–2 g cm-2 kyr-1 after ∼ 35 rmcd. The highest frequency variability in the bulk and CaCO3 MARs results from changes in dry bulk density (Shipboard Scientific Party Leg 208, 2004a); however, this variability is smaller than that variability reflecting the changing resolution of the astrochronology (see Sect. 4.2).

Here, we describe the imprint of cyclic patterns on the Site 1264 CaCO3 content (Fig. 3). New cyclostratigraphy covers the upper ∼ 205 m of the sedimentary succession at Site 1264, which corresponds to strata of middle Miocene to late Pleistocene age. Cyclostratigraphy of the lowermost ∼ 111 m from Site 1264 (between ∼ 205 and ∼ 316 rmcd), which corresponds to the early Oligocene to early Miocene time interval, was previously described in great detail (Liebrand et al., 2016). A cycle interpretation and age model were also previously presented for the upper ∼ 57 m from Site 1264 (Bell et al., 2014); however, due to several splice revisions between 27 and 55 rmcd (see Sect. 3.1) we briefly re-evaluate the cycle imprint on this part of the record (see Sect. 4.1.3 and 4.1.4).

The upper ∼ 205 m is split into four intervals that are characterised by distinct cyclic patterns and/or average sedimentation rates (Sect. 4.1.1, 4.1.2, 4.1.3 and 4.1.4). We apply an 11th order polynomial age model computed on selected (i.e. high-quality) biostratigraphic and magnetostratigraphic depth-age points to obtain a first-order approximation of the durations of the cycles that we identify in the depth domain (see Table S6, Figs. S8 and S9). After applying the polynomial age model, the record was tuned to generate an astrochronology (Sect. 4.2; Figs. 4 and S10).

Two published astrochronologies exist for Site 1264–1265: (1) an Oligocene–early Miocene one (30 to 17 Ma), based on tuning CaCO3 content to eccentricity (Liebrand et al., 2016), and (2) a Plio-Pleistocene one (5.3 to 0 Ma) based on a correlation of the Site 1264 benthic foraminiferal δ18O record to the LR04 Plio-Pleistocene benthic δ18O stack (Lisiecki and Raymo, 2005; Bell et al., 2014). Because of the splice revisions between 27 and 149 rmcd at Site 1264, we re-evaluated the Bell et al. (2014) chronology in the early Pliocene prior to 3.5 Ma/∼ 27 rmcd. The Oligocene to early Miocene astrochronology remains unchanged, but we updated the depth–age tie points to accommodate the cumulative -1.96 m shift in the revised composite depth scale of the overlying sedimentary sequence. The cyclostratigraphic analyses in the depth-domain indicate that the combined Site 1264–1265 CaCO3 content record is suitable for developing an astrochronology for the interval 17–3.5 Ma (see Sect. 4.1) using the flexible best-practice guidelines outlined in Sinnesael et al. (2019). Because of the variable imprint of eccentricity (E), obliquity (T) and precession (P) recorded at Site 1264, it was not possible to implement a uniform tuning strategy for the entire record. In all, we employed three distinct strategies to achieve a 30 Myr astrochronology for Site 1264 (Table S7 and Fig. S10): I.

30–8.0 Ma: CaCO3 content tuned to eccentricity; obliquity is also used in 2.4 Myr minima (when stable in the solution). a.

30–9.7 Ma: CaCO3 content tuned to E (visually aided by δ18O, where available).

Previous work at ODP Site 1264 shows that the recovered sediments record the orbital climate variability of the Southeast Atlantic for the last 30 Myr (Shipboard Scientific Party Leg 208, 2004a; Bell et al., 2014, 2015; Liebrand et al., 2016). Oligocene to early Miocene carbonate and benthic δ18O records from Site 1264 show that the early Coolhouse was dominated by large ∼ 110 kyr eccentricity-driven variability in the Antarctic ice sheet (for full discussion, see Liebrand et al., 2016, 2017). By the Pliocene, Atlantic benthic δ13C gradients indicate that North Atlantic Deep Water (NADW) heavily influenced Southeast Atlantic Site 1264 (for full discussion, see Bell et al., 2014, 2015). The new complete and continuous depth (∼ 316 m; Sect. 3.1; Fig. 3) and age (∼ 30 Myr; Sect. 4; Fig. 4) model presented here constitutes a reference framework for future palaeoclimatic and palaeoceanographic studies at Site 1264. Furthermore, the new data (Fig. 5) enable investigation of how long-term climate trends and orbital-scale climate variability impacted this region, especially between 17 and 5 Ma, for which no high-resolution Southeast Atlantic records previously existed.

At Site 1264, CaCO3 content is very high, with all long- and short-term variability occurring between 92 % and 97.5 % CaCO3 (Fig. 5). CaCO3 content varied between 94 % and 96 % during the Oligocene-early Miocene (30–18.5 Ma), with MARs of ∼ 1–2.5 g cm-2 kyr-1 and are discussed in greater detail in Liebrand et al. (2016). The lowest CaCO3 content (92 %–93.5 %) and MARs (∼ 0.3–0.7 g cm-2 kyr-1) occur between ∼ 18.5 and 14.4 Ma, which broadly coincides with the Miocene Climatic Optimum (MCO; 17–14.7 Ma; Shevenell et al., 2004; Holbourn et al., 2005) (see Sect. 5.2 for discussion). Broadly concurrent with cooling in the lead up to the middle Miocene Climate Transition (mMCT; ∼ 13.9 Ma), CaCO3 content increases and remains between 94 % and 96 % during the early late Miocene (14.4–8.0 Ma), coincident with MARs of ∼ 1–2.5 g cm-2 kyr-1 (see Sect. 5.2 for discussion). The highest CaCO3 content (96 %–97.5 %) and MARs (2.5–4.5 g cm-2 kyr-1) occur between 8 and 4 Ma (Fig. 5), potentially indicating high carbonate productivity coincident with the known age of the global late Miocene–early Pliocene Biogenic Bloom (LMBB; acronym from Lyle et al., 2019) (see Sect. 5.3 for further discussion).

Carbonate deposition is strongly affected by the balance between biogenic carbonate productivity (mostly in the surface water) and carbonate dissolution in the water column and at the sea floor. Sedimentary processes, such as dilution with terrigenous material and/or the removal of fine-grained material through winnowing, can affect both the amount and composition of the carbonate preserved. The relative importance of biogenic productivity versus dissolution is discussed in detail in Liebrand et al. (2016) for the Oligocene to early Miocene, in Sect. 5.2 for the early–middle Miocene and in Sect. 5.3 for the late Miocene–early Pliocene. Over the last 30 Myr, detrital MARs were low, indicating that dilution with terrigenous material was not a major contributing factor in controlling carbonate deposition at Site 1264. Winnowing may have removed fine fraction material, including coccolith carbonate, thereby reducing carbonate deposition at Site 1264. By comparing MARs between nearby sites recovered during DSDP Leg 74, Shackleton et al. (1984) suggested that winnowing may have affected parts of the Walvis Ridge. They suggested that winnowing was especially pronounced at DSDP Site 526 (1054 m water depth) since the late Oligocene. Site 1264 is situated on a very gentle slope above the lysocline and carbonate compensation depth (palaeowater depths: 2–2.5 km). Winnowing likely had less of an effect on Site 1264 compared to Site 526, as Site 1264 is not positioned on the shallowest parts of the Walvis Ridge bathymetry. Nonetheless, Shackleton et al. (1984) also found some indication of winnowing at DSDP Site 525 (2467 m water depth) since the late Pliocene. Independent constraints on winnowing are not available for the entire 30 Myr interval; however, detailed fine fraction weights are available between 30 and 17 Ma (Liebrand et al., 2016; their Fig. 2). If these data are interpreted as a proxy for winnowing, this would suggest that winnowing is modest during the middle Oligocene, increasing during late Oligocene warming and relatively high across the Oligocene–Miocene Transition (OMT; Fig. 5). During the early Miocene (post OMT, pre-middle Miocene) winnowing is comparable to late Oligocene values (Fig. 5). There is evidence for winnowing to have increased towards the condensed middle Miocene part of the Site 1264 record, as there is an increase in both high-resolution and low-resolution percent > 63 µm coarse fraction (%CF) (Liebrand et al., 2016; Keating-Bitonti and Peters, 2019) (Fig. 5). However, between 18.5 and 8 Ma, the Site 1264 %CF varies within a 5 % range, suggesting the amount of winnowing remained stable (Fig. 5; Keating-Bitonti and Peters, 2019). After ∼ 3 Ma, %CF gradually increases from 20 % to 40 % (Fig. 5), which is the largest increase seen in the entire record and could indicate that Site 1264 is affected by winnowing at this time. The presence of winnowing is also supported by the fact that deeper Walvis Ridge Site 1266 and Site 1267 both have higher sedimentation rates than Site 1264 in the last 3 Ma, whereas the opposite would be expected if deep-sea dissolution alone was considered (productivity should affect all sites similarly).

The influence of the long 405 kyr eccentricity on CaCO3 deposition at Site 1264 is complicated. The imprint of the 405 kyr cycle on CaCO3 content was not constant during the Oligocene interval at Site 1264 (Liebrand et al., 2016). This contrasts with the clearer imprint of a 405 kyr cycle on CaCO3 deposition during the Miocene between ∼ 21 and 5 Ma. For periodicities shorter than 405 kyr, we can recognise three distinctly different orbital imprints on the variability in CaCO3 content (Figs. 5 and 6).

The ∼ 110 kyr eccentricity is the dominant driver between 30 and ∼ 13 Ma (Fig. 6a).

The three pacings observed in Southeast Atlantic CaCO3 deposition broadly coincide with major developments in climate, the cryosphere and/or the carbon cycle over the last 30 Myr. At Site 1264, the strong expression of ∼ 110 kyr eccentricity-driven %CaCO3 variability between 17 and 13 Ma parallels the dominant pacing of %CaCO3 during the Oligocene–early Miocene (30–17 Ma; Figs. 5 and 6a; for further details, see Liebrand et al., 2016). The prevalence of ∼ 110 kyr eccentricity pacing at Site 1264 is in line with the wider understanding that Oligocene to middle Miocene unipolar Coolhouse climate was predominantly paced by short-term eccentricity during widespread global warmth (Pälike et al., 2006; Tian et al., 2013; Holbourn et al., 2014, 2015; Beddow et al., 2016, 2018; Voigt et al., 2016; Kochhann et al., 2016; Liebrand et al., 2016, 2017; De Vleeschouwer et al., 2017, 2020; Westerhold et al., 2020). The strong ∼ 110 kyr cyclicity observed in marine archives is attributed to eccentricity-driven changes in ice volume and/or deep-sea temperature, likely associated with changes in atmospheric CO2 (Pälike et al., 2006; Holbourn et al., 2015; Liebrand et al., 2017; Greenop et al., 2019).

The orbital imprint on CaCO3 content shifts between 14 and 13 Ma, when eccentricity-modulated precession cycles progressively become more clearly superimposed on the larger ∼ 110 kyr cycles. These precession cycles remain the main driver of carbonate deposition until ∼ 8 Ma, although obliquity cycles are visible during the 2.4 Myr eccentricity minima from ∼ 12.6 to 12.2 Ma and ∼ 9.7 to 9.3 Ma, when the imprint of precession and ∼ 110 kyr eccentricity is muted (Figs. 5 and 6b). Strong obliquity was also observed in benthic δ18O data from the South China Sea during the ∼ 9.7–9.3 Ma node (Holbourn et al., 2013). The strong obliquity intervals that are observed across multiple marine archives support the idea that obliquity exerts greater control on the climate system as a whole when the orbital configuration is characterised by long-term eccentricity minima coincident with long-term obliquity maxima (Holbourn et al., 2013, 2018; Drury et al., 2017; Levy et al., 2019). The shift to stronger precession pacing occurs after global cooling and the reglaciation of Antarctica across the mMCT (∼ 13.9 Ma; Holbourn et al., 2005). Some precession-driven %CaCO3 cycles were observed and superimposed on larger ∼ 110 kyr eccentricity cycles between 23.5 and 19.5 Ma at Site 1264 (Fig. 5; Liebrand et al., 2016). However, the relative amplitude of eccentricity and precession is different in the middle to late Miocene compared to the Oligocene–early Miocene. In the Oligocene–early Miocene, the amplitude of the ∼ 110 kyr eccentricity cycles in CaCO3 content were greater than the precession-driven CaCO3 content cycles. In contrast, we observe a decrease in strength of the ∼ 110 kyr eccentricity cycles concurrent with the strong precession pacing of the CaCO3 content between 14 and 8 Ma (Fig. 5). The influence of early–middle Miocene climate evolution on Southeast Atlantic carbonate deposition is discussed further in Sect. 5.2.

Although some power remains in the ∼ 110 kyr eccentricity bandwidth, the orbital imprint seen in CaCO3 content changes around 7.7 Ma to a strong obliquity–precession interference pattern, which remains visible until ∼ 3.3 Ma (Figs. 5 and 6c). The onset of prevalent obliquity–precession pacing of %CaCO3 observed at Site 1264 after ∼ 7.7 Ma has been observed globally in benthic δ18O records and is associated with increased influence of high-latitude processes, such as enhanced glacial activity and high-latitude cooling (Drury et al., 2016, 2017, 2018b; Holbourn et al., 2018; see also Sect. 5.3). Although benthic δ18O data are not available at Site 1264 between 8.0 and 6.0 Ma, the obliquity–precession interference pattern is visible in the benthic δ18O record between 6.0 and 3.3 Ma (Fig. 4). Relative to the Oligocene–early late Miocene, the amplitude of the variability in CaCO3 content is reduced during the latest Miocene and early Pliocene. Concurrent with the waning influence of ∼ 110 kyr eccentricity at Site 1264, the highest CaCO3 content (96 %–97.5 %) of the entire record occurs between 8 and 4 Ma. The influence of the complex late Miocene climate system on carbonate deposition is discussed in Sect. 5.3.

After 3.3 Ma, the short-term orbital imprint is more difficult to characterise. The wavelet analysis shows that ∼ 110 kyr eccentricity influence increases in the Plio-Pleistocene compared to the latest Miocene (Fig. 5). The influence of some obliquity and precession forcing on %CaCO3 remains until ∼ 0.9 Ma, when the ∼ 110 kyr eccentricity pacing characteristic of the middle Pleistocene appears after the mid-Pleistocene Transition (MPT; Bell et al., 2014). Compared to Site 1264, the transition from ∼ 40 to ∼ 110 kyr pacing is recorded more clearly at nearby Site 1267, where it is visible in the benthic δ18O data, composite core photos and physical property data (physical property data from Shipboard Scientific Party Leg 208, 2004c; benthic δ18O data from Bell et al., 2014; composite core photos from Westerhold et al., 2017). This difference in expression of the MPT may partly relate to water depth differences between the sites, as the deeper Site 1267 (4356 m water depth) may record a stronger deep-water signal compared to Site 1264 (2507 m water depth). Alternatively, winnowing may have obscured some of the cyclicity at Site 1264, considering the indication that both Sites 1264 and 525 (both ∼ 2.4–2.5 km water depth) were affected by winnowing in the late Pliocene–early Pleistocene. Nonetheless, although the onset of the Pleistocene ∼ 110 kyr cycles is not exceptionally clear at Site 1264, it is apparent that these cycles only appear after 0.9 Ma at both Sites 1264 and 1267 (Fig. 5), which is considerably later than in the eastern equatorial Pacific, where ∼ 110 kyr cycles first appear in carbonate records at 1.6 Ma (Lyle et al., 2019).

Benthic foraminiferal δ18O records are only available for the Oligocene–early Miocene (30–17 Ma; Liebrand et al., 2011, 2016) and the Plio-Pleistocene (5.3–0.0 Ma; Bell et al., 2014). It is therefore not yet possible to track the evolution of the relationship between the climate–cryosphere system (encompassed by benthic δ18O) and South Atlantic carbonate deposition over the last 30 Myr. However, in contrast to the in-phase %CaCO3–benthic δ18O relationship on ∼ 110 kyr eccentricity periodicities between 30 and 17 Ma (Liebrand et al., 2016), the new Site 1264 %CaCO3 data have an inverse relationship with benthic δ18O on obliquity periodicities for the last 6.0 Myr (Figs. 4 and S11). This points to a considerably different relationship between the cryosphere and controls on carbonate deposition at Site 1264 in the late Miocene–Pleistocene compared to the Oligocene–early Miocene. The in-phase Oligocene to early Miocene %CaCO3-benthic δ18O relationship on ∼ 110 kyr periodicities observed at Site 1264 has been observed elsewhere for the Oligocene through to the middle Miocene, including across the mMCT (Holbourn et al., 2014, 2015; Kochhann et al., 2016; Liebrand et al., 2016; Beddow et al., 2018; Tian et al., 2018). It is possible that the %CaCO3–benthic δ18O relationship changed from in-phase on 110 kyr eccentricity periodicities to anti-phase on obliquity periodicities, concurrent with the ∼ 7.7 Ma shift in CaCO3 deposition from a predominantly eccentricity–precession-paced system to one that is more controlled by obliquity–precession. Such an interpretation would further support the notion that the Earth's system underwent a major shift in its response to orbital forcing in the late Miocene–early Pliocene, with Northern Hemisphere high-latitude processes steadily growing in importance in the latest Miocene (Kirtland Turner, 2014; Drury et al., 2017, 2018b; De Vleeschouwer et al., 2020).

The early–middle Miocene marks a warm interval where Antarctic ice volume underwent major change and climatic trends deviated from the overall Cenozoic cooling pattern (Miller et al., 1991; Shevenell et al., 2004; Holbourn et al., 2005, 2014, 2015; Tian et al., 2013, 2014; Super et al., 2018). The MCO (defined in the benthic foraminiferal δ18O as between 17 and 14.7 Ma; Holbourn et al., 2015) was characterised by pervasive global warmth and more humid conditions, together with lower meridional temperature gradients and greatly reduced continental ice sheets in Antarctica compared to the present day (Lear et al., 2000, 2015; Billups and Schrag, 2002; Shevenell et al., 2004; John et al., 2011; Pound et al., 2012; Gasson et al., 2016; Levy et al., 2016). Distal marine records that track variations in land ice volume and deep-sea temperatures are marked by a strong ∼ 110 kyr eccentricity pacing, coupled with large 400 kyr driven carbon cycle perturbations (Monterey Excursion) (Shevenell et al., 2008; Holbourn et al., 2014, 2015; Tian et al., 2014; Kochhann et al., 2016; Ohneiser and Wilson, 2018). The MCO warmth, ice volume decrease and carbon cycle perturbations have been hypothesised to be driven by increased atmospheric CO2 levels associated with volcanic degassing from the Columbia River Flood Basalts, with the earliest eruptions occurring after ∼ 17.2 Ma (Foster et al., 2012; Barry et al., 2013; Greenop et al., 2014; Kasbohm and Schoene, 2018; Moore et al., 2018, 2020; Super et al., 2018; Cahoon et al., 2020; Sosdian et al., 2020). The warm MCO conditions were reversed ∼ 13.9 Ma during the mMCT when major continental ice sheets reappeared on Antarctica associated with a large decrease in atmospheric CO2 and global temperatures (Shevenell et al., 2004; Holbourn et al., 2005; Foster et al., 2012; Pound et al., 2012; Badger et al., 2013; Lear et al., 2015; Gasson et al., 2016; Levy et al., 2016; Super et al., 2018, 2020).

Benthic foraminiferal δ18O records are not yet available at Site 1264 for this interval, so it is not possible to recognise the MCO using this dataset. Nonetheless, the lowest %CaCO3 content (92 %–93.5 %) and MARs (∼ 0.3–0.7 g cm-2 kyr-1) occur between ∼ 18.5 and 14.4 Ma and broadly coincide with the MCO (Figs. 5c and 7c). Low detrital MARs (bulk–CaCO3 MARs), Si and K intensity indicates that biogenic silica and detrital input remains relatively constant and minimal across this interval. The low detrital MARs at Site 1264 (average 0.09 g cm-2 kyr-1) are comparable to the non-carbonate MARs of nearby sites drilled during Leg 74, particularly DSDP Site 525 (Shackleton et al., 1984). Dilution was therefore not the main driving factor of the early–middle Miocene low %CaCO3 content at Site 1264. Winnowing could have removed the < 63 µm fraction at Site 1264 (Fig. 5); however, such winnowing also tends to remove both small CaCO3 and detrital particles, ultimately raising the overall CaCO3 content but lowering the CaCO3 MAR (Marcantonio et al., 2014). A 10 % increase in the percent > 63 µm coarse fraction (%CF) after ∼ 18.5 Ma (Fig. 5; Liebrand et al., 2016) indicates that some winnowing occurred. However, between 18.5 and 8 Ma, the Site 1264 %CF varies within a 5 % range but never increases to the high %CF values seen in the Plio-Pleistocene (Fig. 5; Keating-Bitonti and Peters, 2019). This indicates that increased dissolution and/or decreased productivity likely also drove the early–middle Miocene low CaCO3 content at Site 1264. An increase of B / Ca concentration at Sites 1264 and 1266 after 15.5 Ma (Kender et al., 2014) indicates that dissolution influenced the early–middle Miocene low CaCO3 content at Site 1264.

The recovery of %CaCO3 content ∼ 14.5 Ma agrees well with the end of the MCO ∼ 14.7 Ma (Holbourn et al., 2015). However, at Site 1264, the decreasing CaCO3 content starts ∼ 18.5 Ma, which is ∼ 1.5 Myr before the decrease in benthic δ18O normally associated with the onset of the MCO (Fig. 5). During the early–middle Miocene, low %CaCO3 and CaCO3 MARs were observed at multiple sites in the eastern equatorial Pacific Ocean (EEP; DSDP Site 574; IODP Sites U1335 to U1338), initially decreasing after 18–17.5 Ma, before recovering to early Miocene values by 15–14.5 Ma (Piela et al., 2012; Kochhann et al., 2016). Multiproxy evidence at these EEP sites indicates that the low %CaCO3 and CaCO3 MARs values were associated with increased deep-sea dissolution rather than decreased productivity, with the peak dissolution occurring at the onset of the MCO (Piela et al., 2012; Kochhann et al., 2016). This dissolution horizon has been traced regionally across the equatorial Pacific as the “Lavender” seismic unconformity, with the dissolution potentially linked to the intensification of proto-NADW formation leading to increased corrosive Antarctic Bottom Water (AABW) reaching the Pacific (Mayer et al., 1985). This hypothesis could not be tested at the time due to the absence of any comparable Atlantic carbonate records. However, the new evidence of low %CaCO3 and CaCO3 MARs at Site 1264 in the Southeast Atlantic indicates that dissolution occurred in the Atlantic and the Pacific during the early to middle Miocene. Increased dissolution across ocean basins indicates a global forcing, supporting suggestions that the dissolution seen in the Pacific was associated with elevated atmospheric pCO2, increased carbon storage in the deep ocean and shoaling of the carbonate compensation depth during the early–middle Miocene global warmth (Pälike et al., 2012; Piela et al., 2012; Kochhann et al., 2016).

Assuming that the low CaCO3 content and MARs at Site 1264 is dissolution driven (e.g. see also Kender et al., 2014), rather than reflecting a decrease in carbonate rain, there is evidence that carbonate dissolution preceded the MCO by ∼ 1.5 Myr in the Southeast Atlantic (Site 1264) and ∼ 1.0–0.5 Myr in the equatorial Pacific (Piela et al., 2012; Kochhann et al., 2016). Few early–middle Miocene atmospheric CO2 or sea surface temperature (SST) records extend back to 18.5 Ma, but long-term trends in early–middle Miocene TEX86-derived SSTs from the North Atlantic Ocean indicate that SSTs may have been at MCO levels since ∼ 20 Ma (Super et al., 2018; Fig. 7a). It is not yet clear whether elevated SSTs prior to the MCO are a global phenomenon. However, the likely dissolution-induced lows in Southeast Atlantic and equatorial Pacific CaCO3 deposition up to ∼ 1.5 Myr before the MCO indicate that the MCO itself was preconditioned by elevated temperatures and atmospheric pCO2.

Shortly after the MCO, the overall Cenozoic cooling trend resumes across the mMCT with the reappearance of large ice sheets on Antarctica around 13.9 Ma. At Site 1264, the %CaCO3 values increase after ∼ 14.5 Ma, which could reflect decreased deep-sea dissolution and/or increased surface ocean productivity. Between 14 and 13 Ma, the orbital imprint on CaCO3 at Site 1264 progressively shifts from ∼ 110 kyr eccentricity-dominated pacing to precession-dominated pacing superimposed on the ∼ 110 kyr eccentricity cycles (Figs. 6a, b and 7c). In comparison to the Oligocene–early Miocene, the ∼ 110 kyr eccentricity cycles are more muted, and the precession cycles are of higher amplitude and more clearly expressed during the middle to late Miocene (Figs. 5 and 6b). The change in orbital imprint at Site 1264 after the mMCT may indicate that productivity in this region became more sensitive to precession forcing following changes to ocean circulation and/or the hydrological cycle driven by the reglaciation of Antarctica, global cooling and increased meridional temperature gradients. The shift towards strong precession pacing occurs as carbonate content recovers after the MCO. Site 1264 likely experienced increased carbonate deposition during the MCO as indicated by low LSR and MARs between 18.5 and 14.4 Ma. The increased preservation of precession cycles at Site 1264 after 14 Ma (i.e. after the mMCT) could also reflect a shift in deep-water circulation patterns bringing less corrosive deep waters to Site 1264, which is supported by the increase in B / Ca at Sites 1264 and 1266 (Kender et al., 2014). This deep-water change would have enabled better preservation of precession-driven productivity cycles after the mMCT compared to the early–middle Miocene.

The latest Miocene (∼ 8–5.3 Ma) was a complicated and dynamic interval when climate and ecosystems recognisable to the present-day first appeared (Herbert et al., 2016). There is abundant evidence for a global and long-lasting increase in primary productivity in the global surface ocean during the late Miocene to early Pliocene (Farrell et al., 1995; Dickens and Owen, 1999; Diester-Haass et al., 2002, 2005). This late Miocene–early Pliocene Biogenic Bloom (LMBB; as defined by Lyle et al., 2019) has been recognised between ∼ 8 and 4 Ma in upwelling and oligotrophic areas of all major oceanic basins (Kroon et al., 1991; Dickens and Owen, 1999; Hermoyian and Owen, 2001; Diester-Haass et al., 2002, 2004, 2005; Grant and Dickens, 2002; Liao and Lyle, 2014; Lyle and Baldauf, 2015; Lyle et al., 2019). At Site 1264–1265, the highest CaCO3 content (96 %–97.5 %) occurs between ∼ 8 and 4 Ma, and the highest bulk and CaCO3 MARs (∼ 2–4.5 g cm-2 kyr-1) are found between ∼ 7.8 and 3.3 Ma (Figs. 5 and 8c), which falls within the broad timing associated with the LMBB. As %CaCO3 is so high at Site 1264, CaCO3 MARs account for most of the variability in the bulk MARs. Similarly, Lyle et al. (2019) showed that the LMBB is expressed between ∼ 8 and 4.4 Ma in the bulk and CaCO3 MAR of six sites in the eastern equatorial Pacific (EEP), with CaCO3 MARs accounting for most of the bulk MAR variability (Fig. 8d). Despite the influence of palaeogeographical heterogeneity on the absolute EEP MARs, it becomes apparent after normalisation that common productivity patterns are visible across the EEP (Lyle et al., 2019; Fig. 8b and d). The Site 1264 CaCO3 MARs are generally higher than the EEP sites (Fig. 8c and d), except for ODP Sites 849 and 850, which are the two highest sedimentation EEP sites (Lyle et al., 2019).

The exact cause of the LMBB is poorly understood; however, key hypotheses suggest the increased primary productivity was caused by (1) increased nutrient input into the surface ocean through increased weathering/dust input and/or (2) changes to the global distribution of nutrients through changes in atmospheric and oceanic circulation patterns. The widespread documentation of the LMBB shows that the expression and timing is regionally variable (Liao and Lyle, 2014; Lyle et al., 2019; Sutherland et al., 2019). However, most of these records are low resolution and insufficient for accurately constraining regional differences, so we cannot yet distinguish between global changes to the nutrient budget and changes to the regional distribution of nutrients in the ocean (e.g. changes in ocean circulation and/or upwelling). The availability of orbital-scale %CaCO3 and MARs from Site 1264 and the EEP provide the opportunity to compare the LMBB at high resolution for the first time. Based on increased CaCO3 MARs, the timing of the LMBB at Site 1264 and the EEP generally agrees well (Fig. 8b), which corroborates the global nature of the LMBB. After increasing from 8 Ma onwards, MARs peak between 7.2 and 6.6 Ma at both Walvis Ridge and the equatorial Pacific (Fig. 8b), which supports a global LMBB optimum occurring between ∼ 7.0 and 6.4 Ma (Lyle and Baldauf, 2015; Lyle et al., 2019). Site 1264 also has the highest absolute %CaCO3 values between 7.2 and 6.6 Ma, indicating high productivity of carbonate producers at this time (Figs. 5 and 8a). However, the LMBB extends to 3.3 Ma at Site 1264, in contrast to the western EEP, where the LMBB ends ∼ 4.4 Ma (Fig. 8b). In the far eastern equatorial Pacific near South America, high CaCO3 MARs continue to ∼ 3 Ma (Fig. 8 in Lyle et al., 2019), which is further evidence for regional variability of the termination of the LMBB production interval. The recognition of global patterns and temporal heterogeneity in the expression of the LMBB between the Pacific and the Southeast Atlantic could reflect different regional responses to a single climatic forcing and/or multiple driving forces.

Constraining which primary producers drove the LMBB at different regions will be useful for disentangling regional and global patterns. During the latest Miocene–early Pliocene (∼ 8–3 Ma), the new Site 1264 %CaCO3 data displays an inverse relationship with the low-resolution record of the percent > 63 µm coarse fraction (%CF) (Keating-Bitonti and Peters, 2019; adapted to this study's new composite depth and age model; Figs. 5c and 8a). The %CF specifically shows the opposite trend to %CaCO3 across the LMBB: decreasing %CF from 8 Ma, with the lowest %CF values occurring ∼ 7 Ma, in line with the maximum values in %CaCO3. Through the middle to late Miocene and early Pliocene, the LSR at Site 1264 are either similar or higher at Site 1264 (2505 m) relative to deeper Site 1266 (3806 m). The available %CF and %CaCO3 from Site 1264 also do not display a strong relationship prior to 8 Ma. This all suggests that any winnowing at Site 1264 was minimal and stable for the middle to late Miocene and early Pliocene. The inverse %CF-%CaCO3 relationship therefore could indicate that the LMBB was predominantly driven by a change in the calcareous phytoplankton (coccolithophores) versus foraminifera ratio at Site 1264. Based on Si intensity, there is no evidence that biogenic silica producers play a major role in the LMBB at Site 1264 (Fig. 5). This contrasts with the EEP, which is upwelling dominated and where a combination of calcareous (coccolithophores) and siliceous (diatoms) phytoplankton drove the LMBB (Lyle and Baldauf, 2015; Lyle et al., 2019).

Although we cannot yet accurately distinguish global increases in nutrient delivery to the ocean versus the regional redistribution of nutrients causing localised increased primary productivity, we can consider links between this prolonged productivity event and the dynamic changes observed during the late Miocene. Terrestrial and sea surface temperatures decreased rapidly during the late Miocene cooling between ∼ 7.0 and 5.4 Ma (Pound et al., 2011, 2012; Herbert et al., 2016). There is evidence for dynamic ice sheet activity, although there is no major late Miocene increase in benthic δ18O records suggesting that there was no long-term expansion in continental ice sheet extent or substantial deep-sea cooling (Hodell et al., 2001; Drury et al., 2016, 2017, 2018b; Holbourn et al., 2018; Tian et al., 2018). The carbon cycle underwent major change in the atmospheric, terrestrial and marine realms, with evidence for an atmospheric pCO2 decrease around ∼ 8–7 Ma (Bolton and Stoll, 2013; Herbert et al., 2016; Mejía et al., 2017; Tanner et al., 2020), the rise of terrestrial C4 plants on land (Cerling et al., 1997; Behrensmeyer et al., 2007; Uno et al., 2016; Tauxe and Feakins, 2020) and the globally synchronous marine late Miocene carbon isotope shift (LMCIS; ∼ 7.5–6.9 Ma) linked to global changes in oceanic circulation (Haq et al., 1980; Hodell and Venz-Curtis, 2006; Reghellin et al., 2015, 2020; Drury et al., 2017, 2018a).

At Site 1264, the onset of elevated CaCO3 MARs (∼ 7.8 Ma) roughly coincides with the shift from eccentricity–precession pacing to pervading obliquity–precession pacing of %CaCO3, which infers an increased influence of high-latitude processes in the Southeast Atlantic (Fig. 5). The onset of strong obliquity pacing is also observed ∼ 7.7 Ma as asymmetric (i.e. sawtooth-shaped) benthic δ18O cycles, which have a characteristic “interglacial–glacial” anti-phase relationship with benthic δ13C on obliquity timescales (Drury et al., 2017). The appearance of strong obliquity forcing in multiple systems shortly after 8 Ma implies increased influence of high-latitude climate processes, such as increased glacial activity and high-latitude cooling. There is widespread evidence that the late Miocene cooling was especially pronounced in the high latitudes and reached near-modern gradients around 5.4 Ma (Pound et al., 2012; Herbert et al., 2016). The growing importance of the high-latitudes in the latest Miocene is further supported by evidence that deep-sea stable δ13C and δ18O switched from in-phase to anti-phase on eccentricity timescales (Kirtland Turner, 2014; De Vleeschouwer et al., 2020) as a result of continental carbon reservoirs shrinking during cold periods due to increased extent of low-carbon Arctic biomes, such as ice sheets, polar deserts and tundra (De Vleeschouwer et al., 2020). There is also ice-proximal evidence for enhanced glacial activity in both the Northern Hemisphere and Southern Hemisphere, potentially indicating early transient bipolar cryosphere activity in the latest Miocene (O'Connell et al., 1996; Fronval and Jansen, 1996; Wolf-Welling et al., 1996; Kong et al., 2010; Williams et al., 2010). Increased glacial weathering after 8 Ma may have contributed to the onset of the LMBB through increased the nutrient influx into the ocean. An increased nutrient flux may also be driven by enhanced chemical weathering through Himalayan uplift and the intensification of the Indian and Asian monsoon systems in the latest Miocene (Kroon et al., 1991; Filippelli, 1997; Zhisheng et al., 2001; Holbourn et al., 2018; Yang et al., 2019). Finally, the LMBB may be partly driven by increased nutrient input into the ocean as a result of widespread continental aridification coupled with trade wind intensification due to greater meridional gradients during the late Miocene cooling (7–5.4 Ma) (Hovan, 1995; Filippelli, 1997; Diester-Haass et al., 2006; Tipple and Pagani, 2007; Lyle et al., 2008; Pound et al., 2012; Herbert et al., 2016).

Regional variability in the LMBB may in turn be driven by regional differences in the extent of late Miocene cooling (Pound et al., 2012; Herbert et al., 2016), as well as regional diachrony in aridification (Molnar, 2005; Schuster et al., 2006; Lyle et al., 2008; Dupont et al., 2013). The LMCIS (∼ 7.5–6.9 Ma) has been linked to the onset of near-modern thermohaline circulation with NADW percolating further into the South Atlantic (Hodell and Venz-Curtis, 2006; Drury et al., 2017; Keating-Bitonti and Peters, 2019). A major shift in oceanic circulation would likewise affect the redistribution of nutrients around the globe, thereby potentially contributing to regional differences in nutrient supply. All these aspects go some way to explain why the LMBB began after ∼ 8 Ma; however, it is unclear why the LMBB continued into the Pliocene, and especially why it continued until 3.3 Ma at Site 1264. Further work disentangling global versus regional productivity patterns is needed in future to explore causal links in greater detail.