A total of 21 samples from the spliced record between Holes A (10 samples) and B (11 samples) of Walvis Ridge ODP Site 1264 (28°31.955^′ S, 2°50.730^′ E; water depth of 2507 m; Shipboard Scientific Party Leg 208, Zachos et al., 2004; Fig. 1) were selected from depth intervals that coincide with, or fall close to (maximum offset ± 3 cm), the astronomical tuning ties presented by Drury et al. (2021). Sample ages correspond to the age model of Drury et al. (2021), covering the time interval from 5–3 Ma with an average time step of ∼ 100 kyr (Table S1 in the Supplement). This sample resolution does not resolve fluctuations in nannofossil assemblages on orbital timescales (< 100 kyr). Nevertheless, it is adequate for capturing sustained changes in community structure that marked the termination of the biogenic bloom. Microscopy slides were prepared with the “drop technique” (Bordiga et al., 2015). The initial screening revealed a high density of coccoliths, so in order to increase the dispersal of particles within each field of view, the sediment suspensions were further diluted to a bulk-weight equivalent concentration of 0.06 mgmL-1 (Data Table 1 in 10.5281/zenodo.19854421, Karatsolis et al., 2026). The suspension was ultrasonicated, and the slides were prepared with a drop volume of 1.9 mL on each cover slip. Micropaleontological analysis was conducted at × 1000 magnification and included the examination of at least 300 specimens from at least 10 fields of view (FOVs). The relative abundance (%) for the most common nannofossil species was calculated. Species of the main taxonomic groups (genera), namely Reticulofenestra and Gephyrocapsa, were classified into small (< 3 µm), medium (3–5 µm), and large (> 5 µm) size categories, a standard cut-off that allows for comparison with previous studies (including results presented in Karatsolis et al., 2020). Error envelopes representing 95 % confidence intervals (CIs) were calculated for the nannofossil relative abundances (%) using PAST 4 software (Hammer et al., 2001; Suchéras-Marx et al., 2019). Absolute abundances for the abovementioned common taxa were estimated using Eq. (1) (e.g., Koch and Young, 2007): 1Abs. Abundance (AA)=N×Af×n×W, where A is the coverslip area (mm2), N the number of nannofossils counted in all investigated fields of view, F the surface area of each field of view (FOV; mm2), n the number of FOVs, and W the dry bulk sediment weight on the coverslip (g).

The error estimate for the absolute abundances represents a reproducibility of ± 15 % (Bordiga et al., 2015), estimated for the total nannofossil abundance.

Nannofossil accumulation rates (NARs; in  Ncm-2kyr-1) were calculated using Eq. (2): 2NAR=AA×LSR×DBD, where AA is absolute abundance (Ng-1); LSR the linear sedimentation rate at each sample depth (Drury et al., 2021; since the samples were taken at core depths corresponding to astronomical ties, the sedimentation rate was calculated for each tie and the next one below); and DBD (gcm-3) the dry bulk density, which was estimated by linearly interpolating between two adjacent measurements (above and below the sample age) of calibrated GRA dry bulk density (Drury et al., 2021).

The relative carbonate mass contribution of each nannofossil taxon was calculated following Suchéras-Marx and Henderiks (2014), based on nannofossil size and shape (volume) considerations (Table S1). For taxonomic groups that were only identified at the genus level, the average of several mass estimates for late Miocene and Pliocene species was applied (Suchéras-Marx and Henderiks, 2014; Young and Ziveri, 2000; Table S2). Most unidentified nannofossils were placolith-like nannofossils within the size range of small- and medium-sized Reticulofenestra. For Calcidiscus spp., the mean mass estimate for medium-sized (5–8 µm) species was used because specimens > 8 µm were not common in our samples. The nannofossil carbonate mass (NCmass) contribution within the bulk sample (gg-1 bulk) was calculated following Eq. (3): 3NCmass=AA×MM×10-12, where NCmass was calculated for each nannofossil taxon, with AA being its absolute abundance (Ng-1) and MM its mean mass (in pg). The total nannofossil CaCO3 mass contribution (gg-1 bulk) was then calculated by summing all taxon-specific NCmass estimates. The nannofossil-derived CaCO3 mass accumulation rates (CMARnannos, gcm-2kyr-1) were calculated by multiplying NAR by NCmass (for individual taxa and summed total). For the NW Australian shelf Site U1463, NCmass was calculated in the same way, multiplying previously published nannofossil absolute abundances (Karatsolis and Henderiks, 2023) by their respective coccolith masses (Suchéras-Marx and Henderiks, 2014; Young and Ziveri, 2000; Tables S1 and S2 in the Supplement).

Calculations of foraminifera-derived CaCO3 mass accumulation rates (CMARforams, gcm-2kyr-1) were based on two assumptions. The first was that in a high-carbonate content nannofossil ooze (> 94 % CaCO3), all the coarse fraction belongs to the carbonate fraction of the sediment. The second assumption was that this coarse fraction mostly corresponds to foraminifera shells. Based on these assumptions, the coarse-fraction (> 63 µmwt%) measurements from Keating-Bitonti and Peters (2019) were resampled at the sample depths used in this study and multiplied by the LSR and DBD (see NAR description above).

Biostratigraphic marker species, such as Reticulofenestra pseudoumbilicus (> 7 µm), Pseudoemiliania lacunosa, and Sphenolithus abies, were documented in order to evaluate any significant age discrepancies between observed biohorizons (first or last occurrences) and the astronomically tuned ages in the spliced record (Drury et al., 2021). Special focus was given to Amaurolithus primus, a species with a last occurrence (top) that has been astronomically calibrated in the equatorial Pacific (ODP Leg 138) between the Nunivak (4.48 Ma) and Sidufjall (4.8 Ma) geomagnetic subchrons to 4.5 Ma (Raffi et al., 2020, and references therein), falling very close to the timing of the end of the biogenic bloom, as suggested by Karatsolis et al. (2022). This biohorizon has a low degree of reliability (“D”: indistinct and poorly defined in terms of abundance change and which is demonstrated as diachronous; Raffi et al., 2006, 2020). Yet, it was the only calcareous nannofossil biohorizon available for constructing the initial age model between 4–5 Ma at ODP Site 1264, identified between section 3H-7-19 and 4H-1-40 at Hole A and between 4H-7-11 and 5H-CC at Hole B (shipboard biostratigraphy; Shipboard Scientific Party Leg 208, Zachos et al., 2004, and Drury et al., 2021). The mid-point depth between these sections has an age of ∼ 4.1 Ma (4115 ka) in the astronomically tuned age model (Drury et al., 2021), which supports a diachroneity of this bioevent between the equatorial Pacific and South Atlantic. We investigated > 200 FOVs to verify the presence or absence of Amaurolithus spp. in all samples (Data Table 3 in Karatsolis et al., 2026).

In order to identify different phases in overall CaCO3 fluxes, we calculated median CMAR values for three distinct time intervals (Fig. 2). These correspond to the interval before the biogenic bloom (10–7.8 Ma), the interval of the biogenic bloom as defined by Drury et al. (2021) (∼ 7.8–3.3 Ma), and the interval between the suggested termination of the biogenic bloom in this study (following visual inspection of stepwise changes in the CMAR record) and that by Drury et al. (2021) (4.1–3.3 Ma). This provides a first-order statistical evaluation of the timing that would most likely resemble the end of the biogenic bloom at ODP Site 1264, before comparison with the nannofossil assemblages.

Defining the beginning and end of distinct paleoceanographic and paleoclimatic time intervals, such as the biogenic bloom, can be challenging. Disagreements between studies are common, mainly because of differences in age model accuracies or sampling resolution and definitions based on visual inspection of the available data. The calculation of median CMAR values at Site 1264 allows for a statistical estimation of the main breakpoint in carbonate burial rates (Fig. 2) and suggests that they significantly dropped after 4.1 Ma (median of ∼ 2 gcm-2kyr-1; Phase II), resembling the accumulation rates before the onset of the biogenic bloom (∼ 1.8 gcm-2kyr-1) rather than during (∼ 2.8 gcm-2kyr-1; LMBB; see Drury et al., 2021). This is in good agreement with the interval of the highest CaCO3 content (96 %–97.5 %) that lasted until ∼ 4 Ma (Drury et al., 2021).

This observation suggests that the CMAR reduction at Site 1264 occurred in two main steps, at 4.1 Ma (end of Phase I) and at 3.3 Ma (end of Phase II), where the latter was denoted as the termination of the biogenic bloom (Drury et al., 2021). These steps may relate to compositional changes in the main carbonate producers. The end of Phase I falls close to the depth in core, where the top of Amaurolithus primus was identified by shipboard biostratigraphy (Zachos et al., 2004; Drury et al., 2021). This bioevent is considered to have low reliability (Raffi et al., 2020; see above), but were it taken at face value, it would suggest that a first step of CMAR reduction at Site 1264 is in good agreement with the proposed termination of the biogenic bloom suggested for low latitudes, at ∼ 4.5 Ma. However, we were not able to identify any A. primus with clear morphological characteristics, and some nannoliths that demonstrated resemblance were instead classified as broken A. delicatus (Eric De Kaenel, personal communication, 2022; top at 3.933 Ma) or A. brevigracilis (Eric De Kaenel, personal communication, 2022). The latter has a top dated at 4.64 Ma in the Atlantic Ocean (Blair et al., 2017a).

When dated on the tuned age model of Drury et al. (2021), the top occurrence of Amaurolithus brevigracilis (this study) has a midpoint-depth age of 4.099 ± 0.060 Ma, which is ∼ 500 kyr younger than the calibrated bioevent age (Data Table 4 in Karatsolis et al., 2026). Other biohorizons identified in this study (Fig. 2), such as the top of Sphenolithus abies (midpoint-depth age 3.622 ± 0.093 Ma), the top of Reticulofenestra pseudoumbilicus (midpoint-depth age 3.793 ± 0.083 Ma), and the first consistent presence (Base common) of Pseudoemiliania spp. (midpoint-depth age of 4.099 ± 0.060 Ma), actually show good correspondence, with published bioevent ages of 3.61 + 0.09 Ma, 3.82 Ma (Raffi et al., 2020), and 3.81 ± 0.11 Ma, respectively (Nannotax3 database; Data Table 4 in Karatsolis et al., 2026). Combined, our biostratigraphic observations do not justify any revisions to the astronomically tuned age model. Instead, the apparent ∼ 500 kyr discrepancy between the shipboard assignment of the top of A. primus (Zachos et al., 2004; Drury et al., 2021) or top of A. brevigracilis (this study) and the astronomically tuned ages for the same stratigraphic level (Drury et al., 2021) could be attributed to a delayed extinction of this taxon in higher latitudes, taxonomic misidentifications, reworking, or significant changes in sedimentation rates, which might have influenced the astronomical tuning of the records. Since an a priori assumption of the synchronicity of the end of the biogenic bloom is not the premise nor aim of this study and most biostratigraphic evidence shows good alignment, we retain the astronomical tuning ages for Site 1264 in order to investigate the nannofossil assemblage patterns across the main steps in CMAR reduction.

We report on the relative and absolute abundances of calcareous nannofossils (Fig. S1), including the most common taxa belonging to the Noelaerhabdaceae family (Fig. 3a–d), namely small Reticulofenestra and Gephyrocapsa species (< 3 µm), as well as medium-sized (3–5 µm) and large (> 5 µm) Reticulofenestra. The relative abundance of small Reticulofenestra increased gradually from ∼ 5 to ∼ 4.2 Ma, reaching very high values of ∼ 90 % at ∼ 4.4 Ma. After that, a progressive long-term decrease began, and relative abundances gradually dropped to ∼ 20 % towards the late Pliocene (Fig. 3a). Although no rapid and sustained decrease in relative abundance was observed across the Pliocene, two step-wise decreases of 20 % and 35 % occurred between ∼ 4.1–3.9 and ∼ 3.5–3.2 Ma, respectively (Fig. 3a). On the other hand, the relative abundance of small Gephyrocapsa – although consistently present for the interval between ∼ 4.2–3.3 Ma – never exceeded ∼ 8 % (Fig. 3a). The presence of small Reticulofenestra and the good preservation of coccoliths reduce the likelihood that the low abundances of small Gephyrocapsa at ODP Site 1264 are the result of dissolution. Among other prominent groups, medium-sized Reticulofenestra (3–5 µm) became progressively less abundant between 5–4.4 Ma, decreasing from ∼ 35 % to ∼ 2 %. Following that, they increased in relative abundance again after ∼ 3.8 Ma, reaching ∼ 60 % at ∼ 3.2 Ma (Fig. 3b). This rise in abundance could be related to the reported first occurrence of R. minutula (3–5 µm) between 3.70–3.92 Ma (Nannotax 3; Young, 1998). Large Reticulofenestra species (> 5 µm) constituted 3 %–10 % of the assemblage from ∼ 5–4.5 Ma and virtually disappeared after this time (Fig. 3a). Among other taxa, Calcidiscus spp. were consistently present in the assemblage, with relative abundances between ∼ 2 %–10 % across the study interval (Fig. S1). Together, the abovementioned taxa comprised > 87 % of the assemblage in all samples. Absolute abundances (Ng-1) show similar patterns for the respective taxa (Fig. 3b). Superimposed on their overall dominance (> 50 %), distinct peaks in small Reticulofenestra, with concentrations of > 30 billion coccoliths per gram bulk sediment, are centered around 4.4 Ma and at 3.5 Ma (Figs. 3c and S1). The most pronounced and sustained reduction in absolute abundances of this taxon occurred after 3.5 Ma. Finally, NARs were dominated by small Reticulofenestra and were the highest between ∼ 4.7–4.1 Ma (> 50–80 × 10⁹ nannofossilscm-2kyr-1) and ∼ 3.5 Ma (75 × 10⁹ nannofossilscm-2kyr-1), with two distinct reductions at ∼ 4.8 and between ∼ 4.1–3.8 Ma. Consistently low NARs (< 30 × 10⁹ nannofossilscm-2kyr-1) were established after 3.3 Ma (Fig. S1).

The NCmass records for ODP Site 1264 and IODP Site U1463 provide additional information regarding the CaCO3 contribution of different nannofossil taxa relative to their abundance. The stacked plots of mass contributions (Fig. 4a–d) show that Noelaerhabdaceae species and Calcidiscus spp. were the main contributors of carbonate across the late Miocene and Pliocene at both sites. At Site 1264, the carbonate mass contribution from large placoliths such as Calcidiscus spp. is much higher compared to their relative abundance and remained relatively stable (between ∼ 26 %–50 % of total NCmass) across the study interval. In contrast, the relative contributions from Reticulofenestra species changed notably between ∼ 4.1–4 Ma. Before this time, all Reticulofenestra species contributed uniformly, with values ranging between ∼ 10 %–20 % of the total NCmass (Fig. 4c). Small Reticulofenestra dominated between 4.6–4.1 Ma, with values reaching up to ∼ 32 % of the total NCmass (Fig. 4c). After 4 Ma the relative carbonate mass contribution was dominated by medium Reticulofenestra (∼ 10 %–45 %), with small Reticulofenestra at lower contributions of ∼ 10 %–22 % until 3.4 Ma. At the same time, the mass contribution from large Reticulofenestra species plummeted, following their reduction in absolute abundance. After 3.4 Ma, small Reticulofenestra no longer contributed significantly to the total NCmass, with medium Reticulofenestra remaining as the main representative of the genus. All other species had rather stable carbonate mass contributions of 10 %–20 % across the study interval, with half of these contributions most of the time coming from Helicosphaera spp.

A different pattern is observed at the NW Australian shelf Site U1463. The main change here occurred at 4.5 Ma. Before this time, Calcidiscus spp. contributed minimally (rarely exceeding 15 %), and the carbonate mass was mainly derived from small- and medium-sized Reticulofenestra species. After 4.5 Ma, Calcidiscus showed an increased contribution to the total NCmass, with values ranging between ∼ 20 %–40 %. Between 4.4 and 4.2 Ma, medium Reticulofenestra was the main carbonate contributor of the Noelaerhabdaceae (up to ∼ 62 %), with small Reticulofenestra playing a much smaller role (∼ 5 %–12 %). After 4.2 Ma, small Gephyrocapsa started contributing markedly to the total NCmass and became the most effective carbonate mass producer (sometimes reaching > 45 %) within the Noelaerhabdaceae, at least until 3.5 Ma.