Core MD12-3412 was collected during the MD191/MONOPOL expedition of the French R/V Marion Dufresne in 2012 at a water depth of 2368 m in the northeastern Bay of Bengal (18°18.62^′ N, 89°34.26^′ E; 32 m long) using a giant piston (Calypso) corer. Sediments of the Bengal Fan are deposited by turbidity current deposits via channel levee systems during active fan progradation and by hemipelagic sedimentation during periods of local fan inactivity, making it a mixed-sedimentation environment with both turbidite and hemipelagic deposits (Fauquembergue et al., 2019).

The lithology of core MD12-3412 was studied in detail by Fauquembergue et al. (2019) using grain size analyses, XRF elemental data, X-ray imaging, microscopy, and physical property data. These authors describe alternating intervals of fine-grained (clay, 4–15 µm grain size) hemipelagic sediments and 91 thin (1 cm scale) turbidite layers, identified by their sharp basal contacts, coarser grain size fining upwards, and elevated Si/Al and Zr/Rb ratios at their base (Fauquembergue et al., 2019). Analysis of turbidite frequency showed that periods of higher turbidite activity mainly occurred during the glacial periods MIS 6 and MIS 2–4, as illustrated by peaks in median grain size (Fauquembergue et al., 2019; Fig. 2b). Clay mineral assemblages and Sr–Nd isotopic compositions confirm a consistent sediment source from the Ganges–Brahmaputra system with a more minor contribution of sediments from the western part of the Indo–Burman Ranges, with higher contributions from the Indo–Burman Ranges during glacial stages (Joussain et al., 2016). Regardless of sediment source, smectite/(illite + chlorite) ratios suggest more detrital material from highland areas of river basins during glacial periods (Joussain et al., 2016).

A total of 232 samples were taken from the core (resulting in a mean time resolution of 1.2 kyr) and were prepared and analyzed for calcareous nannofossils at CEREGE, Aix-en-Provence. Coccolith samples were taken at 10 cm intervals in the core without discrimination between hemipelagic and turbidite layers. Microscope slides were prepared using a quantitative random settling technique modified from Beaufort et al. (2014), allowing absolute coccolith abundances to be calculated. Samples were weighed (∼ 5 mg), suspended in tap water, and briefly ultrasonicated to disaggregate. 1 mL of sample solution was then placed into decantation vessels, with pre-weighed (on a Mettler Toledo XP2U microbalance) glass coverslips (12×12 mm) positioned at the bottom. The samples were left undisturbed for 4 h, allowing all particles to settle onto the coverslips. Water was then gently removed using a pipette. Following this, samples were dried in the oven overnight at 50 °C. Dry coverslips were then weighed and mounted onto standard microscope slides using Norland Optical Adhesive No. 74.

Automated image acquisition was performed on a Leica DM6000 microscope, equipped with bidirectional circular polarization (Beaufort et al., 2021) and fitted with an automated XY stage holding two slides (16 samples), at 1000× magnification. In each sample, 150 fields of view (FoVs; area 125× 125 µm each) were imaged, with each image composed of a stack of seven images at 5 µm z intervals, using a Hamamatsu black and white digital camera (C11440). Nannofossil classes were identified using SYstème de Reconnaissance Automatique de COccolithes (SYRACO), an automated recognition system based on artificial neural networks (Beaufort and Dollfus, 2004; Dollfus and Beaufort, 1999). The software analyzes specimens across 33 morphological classes of coccolithophores. The underlying model architecture of SYRACO is continuously updated to enhance accuracy and processing speed. In this study, we employed a combination of ResNet50 and YOLOv8 architectures. SYRACO also enables detailed morphometric analysis of each identified coccolith, including measurements of length and mass (e.g., Beaufort et al., 2022).

To confirm our results based on the automated system, we also carried out traditional manual coccolith counts to determine the relative abundance of F. profunda on a subset of 10 samples spanning the study interval. For these 10 samples, two experienced micropaleontologists (Bolton and Beaufort) independently carried out counts. Random FoVs were counted at 1000× magnification on a Leica DMRBE microscope under circular-polarized light, until a total of at least 200 coccoliths or 20 FoVs were counted.

Coccolith absolute abundances (CAs; number of coccoliths per g of sediment) were calculated, both for total coccoliths and for the two main groups (F. profunda and Noelaerhabdaceae coccoliths), according to the equation 1CA=A×Nf×n×W, where A = the total area of the coverslip (mm²), N = the total number of coccoliths counted, f = the area of one field of view (mm²), n = the number of fields of view counted, and W = the weight of dry sediment on the coverslip (g).

Relative abundances of groups/species were also calculated, and 95 % confidence intervals were determined using multiple proportion confidence intervals, calculated with PAST software (version 4.03).

Total sediment mass accumulation rates (tMARs; g m⁻² yr⁻¹) were calculated as 2tMAR=SR×DBD, where SR = sedimentation rate (cm kyr⁻¹) and DBD = dry bulk density (g cm⁻³). DBD was calculated from wet bulk density (WBD) data measured on core MD12-2412 at 2 cm resolution using a multi-sensor core logger. For this calculation, we applied a linear regression between DBD and WBD, developed from discrete measurements at the nearby International Ocean Discovery Program (IODP) Site U1446, in the northwestern Bay of Bengal (Clemens et al., 2016). 3DBD=1.5402×WBD-1.5453 Coccolith accumulation rates (ARs; number of coccoliths m⁻² yr⁻¹), both for total coccoliths and for specific groups (F. profunda and Noelaerhabdaceae coccoliths), were calculated using the equation 4AR=CA×SR×DBD. Coccolith mass accumulation rates (cMARs; g of coccolith calcite m⁻² yr⁻¹) were calculated using the equation 5cMAR=DBD×SR×cM, where cM is coccolith mass (total mass of coccolith CaCO₃ per g of sediment).

We also calculated cMAR for F. profunda coccoliths and Noelaerhabdaceae coccoliths (NoMAR).

To obtain cM, we used the coccolith CaCO₃ mass measured by SYRACO, calculated as 6cM=total coccolithCaCO3mass on coverslipweight of sample used, where total CaCO₃ mass on coverslip = CaCO3mass by SYRACOarea of coverslip analyzed × area of total coverslip.

Spectral analyses (multi-taper method, robust AR(1) noise model) and phase and coherence analysis on coccolith AR and MAR records were performed using Acycle v2.8 (Li et al., 2019). The cAR and cMAR records were filtered to isolate significant variance in the precession band, using a Gaussian filter and a frequency range of 0.04–0.06 cycles kyr⁻¹ (17–25 kyr), also in Acycle v2.8.

Following Fauquembergue et al. (2019), we applied a depth correction to Calypso core MD12-3412 based on the correlation of magnetic susceptibility measurements with those in a CASQ (9 m long gravity core) from the same location, to correct for any potential sediment thickness bias in the upper part of the Calypso core. An age model for the upper 13.46 m corrected depth of core MD12-3412 was published by Fauquembergue et al. (2019) based on seven radiocarbon dates, the identified Toba ash layer (dated at ∼ 73.7 ± 0.3 thousand years ago (ka); Mark et al., 2017), and tuning of a high-resolution Globigerinoides ruber oxygen isotope (δ18O) stratigraphy to the LR04 benthic foraminiferal δ18O stack (Lisiecki and Raymo, 2005).

Here, we propose a revised age model for the interval 0 to 22.35 m corrected depth of core MD12-3412, based on the same radiocarbon and Toba age–depth tie points as Fauquembergue et al. (2019) but with a different tuning of the G. ruber δ18O record and the inclusion of additional G. ruber δ18O data between 13.46 and 22.35 m corrected depth (80 new samples). For all G. ruber analyses from core MD12-3412, 15 tests of G. ruber (sensu stricto) were picked from the 250–315 µm size fraction. New and published isotopic analyses on G. ruber tests were conducted at the LSCE using an ISOPRIME mass spectrometer. Samples were calibrated to PDB values with a laboratory standard referenced to NBS19. The internal reproducibility, estimated from replicate analyses of the standard, was ±0.06 ‰ for δ18O (1σ). We first performed a tuning using imposed radiocarbon and Toba age–depth tie points using the “get age estimate” MATLAB code, which automatically tunes isotope records to the updated global benthic δ18O stack (Ahn et al., 2017). We then performed some manual adjustments, based on visual assessment and inferences based on expected sedimentation rates given the frequency and thickness of turbidites identified by Fauquembergue et al. (2019) and on the presence of coccolith biostratigraphic markers (Fig. 2).

The main difference between our revised age model and that of Fauquembergue et al. (2019) occurs below 12 m corrected depth, where we suggest that the minimum in G. ruber δ18O of -3.54‰ at 16.30 m corresponds to MIS 7c rather than MIS 7e (Figs. 2 and S1 in the Supplement; Table S1 in the Supplement). This tuning results in significantly higher sedimentation rates for MIS 7 than in the original age model, coherent with the high frequency of turbidites identified during this interval (Fig. 2a), with low coccolith abundances suggesting dilution (see Sect. 3.2), and with the stratigraphic position of the lowest confirmed occurrence of coccoliths belonging to the species E. huxleyi at 15.60 m (the first occurrence of this species is dated at 265–291 ka; Raffi et al., 2006). The end of the acme of Gephyrocapsa caribbeanica (dated at ∼ 300 ka; Beaufort et al., 2022, and references therein) near the core base (25.30 m) further supported a younger age for these sediments than previously suggested. Sedimentation rates are on average 10 cm kyr⁻¹ and are in the range ∼ 2 cm kyr⁻¹ (early Holocene, MIS 5a–e) to ∼ 17 cm kyr⁻¹ (MIS 7d–e) (Fig. 2b). Our age model is supported by the median grain size record (Fig. 2a; Fauquembergue et al., 2019) that shows highest turbidite frequency during periods of highest estimated sedimentation rate. The original and revised age–depth models for core MD12-3412 are shown in Fig. S1.

The calcareous nannofossil assemblage mainly consists of F. profunda, Noelaerhabdaceae (E. huxleyi, G. caribbeanica, Gephyrocapsa oceanica, Gephyrocapsa ericsonii, Gephyrocapsa muellerae), Umbilicosphaera spp., Syracosphaera pulchra, and Helicosphaera spp. In addition, Rhabdosphaera, Discosphaera, Pontosphaera, Calcidiscus, Ceratolithus, Calciosolenia, and Umbellosphaera were more rarely observed. However, these minor groups were not included in our dataset because of low numbers and because of the presence of false positives in these groups related to a high abundance of detrital carbonate in some intervals (i.e., detrital particles falsely identified as coccoliths).

Visual assessment of samples from throughout the study interval under the microscope indicated that samples from 17.35 to 22.35 m corrected depth (237.29 to 269.88 ka) were severely diluted by terrigenous particles and contained few coccoliths (Fig. 3c). On average, 4.3 × 10⁹ coccoliths per g of sediment were quantified between 0 to 17.35 m corrected depth, and 1.4 × 10⁹ coccoliths g⁻¹ were quantified between 17.35 m and 22.35 m corrected depth (see Sect. 3.4 for details). This dilution effect was confirmed by manual counts of F. profunda relative abundance in 10 samples, including 5 from the 17.35 to 22.35 m interval (Fig. S2). Therefore, although data are included in all graphs, we do not consider relative abundance data reliable in this interval because of low total coccolith counts (the start of this interval is indicated with a dotted line in Figs. 4 and 5), and values from this interval are not included in average values stated below.

Calcareous nannofossil assemblages are well preserved throughout the record. This is attested to by the presence of delicate coccoliths with intact central area features, e.g., tiny (1–2 µm) Gephyrocapsa ericsonii coccoliths with bridges, Syracosphaera pulchra with intact central area grills, and Umbellosphaera tenuis coccoliths. In addition, several whole coccospheres were observed. The high clay content in this core likely favored coccolith preservation. In the diluted interval with fewer coccoliths below 17.35 m, preservation remained good even though coccoliths were much sparser in slides.

Florisphaera profunda constitutes on average 80 % of total coccoliths (minimum 64 %, maximum 93 %) (Fig. 4b), and Noelaerhabdaceae coccoliths constitute on average 18 % (range 6 % to 35 %) (Fig. 4c). The highest percentage of F. profunda is observed during MIS 3, with lowest abundances occurring in MIS 7. F. profunda percentages and trends are broadly confirmed with manual counts in a subset of samples (Fig. S2). The other main coccolith groups combined generally constitute < 2.5 % of the total assemblage in the interval MIS 1 to 7 (Fig. 4d–f). These minor coccolith groups (Helicosphaera, Umbilicosphaera, S. pulchra) show no clear trends or rhythms in abundance over the study interval.

Within the Noelaerhabdaceae, we grouped together small (< 2.5 µm) and medium to large (> 2.5 µm) coccoliths and calculated their relative abundance as a percentage of total Noelaerhabdaceae (Fig. 5b, c). The small group comprises the identified morphospecies E. huxleyi (coccolith length averaged over the entire study interval = 1.9 µm), G. caribbeanica (average coccolith length = 2.3 µm), and G. ericsonii (average coccolith length = 1.7 µm). The medium to large group comprises G. oceanica (average coccolith length = 3.6 µm), G. muellerae (average coccolith length = 2.6 µm), and grouped Reticulofenestra coccoliths (average coccolith length = 3.3 µm). The most abundant size class throughout was small Noelaerhabdaceae (Fig. 5b), sometimes constituting > 95 % and on average 86 % of all Noelaerhabdaceae coccoliths. Medium- to large-sized Noelaerhabdaceae ranged between 1 % and 61 %, with an average of 13.7 % (Fig. 5c). We observe glacial–interglacial variations in the relative abundance of the two Noelaerhabdaceae size groups from MIS 1 to 6, with small Noelaerhabdaceae most relatively abundant in MIS 2, 4, and 6 (glacials) (Fig. 5b). Medium to large Noelaerhabdaceae coccoliths show highest relative abundances during MIS 1, 3, and 5 (interglacials) (Fig. 5c). To illustrate the dominance of small Noelaerhabdaceae coccoliths during glacial periods, we plot the ratio of small coccoliths to medium and large coccoliths (Fig. 5d). The largest and heaviest coccoliths are observed in MIS 1, 3, and 5 interglacials. Average (whole-population) Noelaerhabdaceae coccolith mass varied from 1 to 5.8 pg (mean 2.2 pg, standard deviation 0.9 pg; Fig. 5g), while Noelaerhabdaceae coccolith length varied between 1.7 and 3 µm (mean 2.1 µm, standard deviation 0.2 µm; Fig. 5h). The higher relative contribution of small Noelaerhabdaceae coccoliths during glacials MIS 2, 4, and 6 (Fig. 5b–d) occurs alongside a higher total Noelaerhabdaceae AR and a higher NoMAR (Fig. 3e, f; see Sect. 3.4), indicating that Noelaerhabdaceae coccoliths are more numerically abundant when the smaller size group dominates during glacial periods.

Total coccolith absolute abundances (CAs) over the last 279 kyr averaged 3.5 × 10⁹ coccoliths g⁻¹ and varied between 0.2× 10⁹ and 14.5 × 10⁹ coccoliths g⁻¹ with highest values during the latest Holocene, relatively elevated values from MIS 1 to MIS 5, and lowest values during MIS 7 and 8 (Fig. 3c). Coccolith accumulation rates (cARs) vary between 0.03 × 10¹² coccoliths m⁻² yr⁻¹ in MIS 7 to 2 × 10¹² coccoliths m⁻² yr⁻¹ in MIS 1 (Fig. 3d). The average F. profunda AR was 298 × 10⁹ coccoliths m⁻² yr⁻¹, whereas the average Noelaerhabdaceae AR was 73 × 10⁹ coccoliths m⁻² yr⁻¹. The cAR is generally higher during glacials (MIS 2, 4, and 6; with the exception of a trough in the middle of MIS 2) and lower during interglacials (MIS 1, 3 and 5), although this trend breaks down in MIS 7–8, where CA is low and turbidite frequency and intensity are high (Fig. 3b–d).

Coccolith MAR (cMAR; Fig. 3e) is on average 0.65 g coccolith CaCO₃ m⁻² yr⁻¹ (range 0.09 to 2.8 g coccolith CaCO₃ m⁻² yr⁻¹). cMAR shows similar patterns to cAR, with generally higher values during glacial periods of the last 200 kyr. The relative contribution of F. profunda to total cMAR (Fig. 3e) is lower than its numerical contribution (cAR; Fig. 3d) because of the low mass of individual F. profunda coccoliths relative to Noelaerhabdaceae and other coccoliths. On average, F. profunda coccoliths contribute 65 % to total cMAR, whereas Noelaerhabdaceae coccoliths contribute 24 % (Fig. 3e). Noelaerhabdaceae cMAR (NoMAR) ranges from 0.02 to 0.9 g m⁻² yr⁻¹ (Fig. 3e). On average, cMAR from the main nannofossil groups (Noelaerhabdaceae, F. profunda, Umbilicosphaera, Helicosphaera, and S. pulchra) makes up 0.47 % of the total sedimentary MAR, tMAR (Fig. 3f), at Site MD12-3412. Although this low coccolith CaCO₃ contribution is not surprising in the Bengal Fan sedimentary environment, we note that this is likely to be an underestimate because some coccoliths (e.g., coccoliths in aggregates or coccolith fragments) are inevitably missed in the automated image analysis workflow.

Spectral analyses of coccolith accumulation rate (total cAR, F. profunda AR, and Noelaerhabdaceae AR) and cMAR (total coccoliths) records reveal > 95 % significant variance at the 100 kyr frequency and > 95 % significant variance in the precession band (19–23 kyr) (Fig. 6). Precession-band variability is absent from the total sediment MAR record (Fig. 3f) and must thus be related to coccolith production and/or export flux from the surface ocean. Coherence and phase analysis between the MD12-3412 G. ruber δ18O record and the cAR record shows that the two records (from the same core and on the same age model) are highly coherent and in phase (within error) in the precession band (Fig. S3).

Before we can interpret our calcareous nannofossil data in terms of coccolith export and primary productivity fluctuations, we must be certain that coccolith sedimentation patterns in core MD12-3412 primarily reflect in situ production and export rather than transport by turbidity flows. We infer that this is the case based on several lines of evidence. Firstly, turbidites as represented by median grain size peaks (Fig. 3b; Fauquembergue et al., 2019) do not consistently co-vary with coccolith abundances (Fig. 3c). Turbidites did, however, play a major role in sediment deposition at this site, especially during sea-level lowstands (glacials) when the channel was most active (Fauquembergue et al., 2019; Joussain et al., 2016). Highest turbidite frequency aligns with periods of highest estimated sedimentation rates, supporting our revised age model (Fig. 2). Our results show that cMAR constitutes only a very small portion of total sediment MAR (< 0.5 %) at this site, as a result of dilution by terrigenous material transported in the Bengal Fan system, although this is likely an underestimation of coccolith calcite contribution due to automated recognition techniques (i.e., some coccoliths are likely missed because they are not recognized, they occur in aggregates, or they are fragmented). Total % CaCO₃ values over the Holocene at nearby core MD12-3617 (16°30′ N, 87°47′ E) vary between 2 % and 7 % (Moreno et al., 2020), illustrating the low relative contribution of CaCO₃ to sediments in the region.

Coccolithophore AR (cAR) and cMAR trends in core MD12-3412 co-vary with sedimentation rate changes, with generally higher values during higher-sedimentation-rate glacial intervals (Fig. 3d, e). However, in the sedimentary setting of our study site, sedimentation rates cannot provide a first-order indication of biological export productivity, as is the case in open-ocean pelagic sedimentation realms where almost all sediment is made up of biogenic components (CaCO₃, opal). Our data show that the lowest coccolith abundance occurs during the period with the most intense turbidite activity (Fig. 3b, c), and we also find that cAR and cMAR records have unique spectral characteristics (Fig. 6). Thus, we think that few coccoliths are transported to the site in turbidity currents, and we infer that higher cAR and cMAR are indicators of increased coccolith carbonate export from the overlying water column, reflecting increased coccolithophore export productivity. In core MD12-3412, no difference in G. ruber δ18O trends was noted when foraminifera were picked exclusively from hemipelagic intervals versus when they were picked without discriminating the origin of the sequences (Fauquembergue et al., 2019), supporting the idea that transport of marine carbonate microfossils via turbidity currents was negligible.

The mass accumulation rate of Noelaerhabdaceae coccoliths, NoMAR, has been shown to be driven primarily by coccolith flux in the tropical Indo-Pacific, where this group generally makes up around half of the total coccolith mass (Beaufort et al., 2022). A dominant control of AR on NoMAR is also the case in core MD12-3412 (Fig. 6e–g). NoMAR values in core MD12-3412, along with the contribution of Noelaerhabdaceae to total coccolith MAR (Fig. 3e), are slightly lower than the range documented in an Indo-Pacific stack of seven tropical cores over the late Pleistocene (∼ 0.5 to 2 g m⁻² yr⁻¹) but show quite similar trends (Beaufort et al., 2022) (Fig. 7d). NoMAR values at our central northern BoB site are similar to those at Site U1448 in the Andaman Sea and at Site U1446 in the northwestern BoB and are higher than those at southern BoB Site U1443 (Fig. 7).

Spectral analyses of the cAR and cMAR records reveal significant (> 95 % CI) variance at two main orbital frequencies: ∼ 100 kyr (a period of Earth's orbital eccentricity and of the late Pleistocene glacial–interglacial cycles) and ∼ 19–23 kyr (Earth's orbital precession periods) (Fig. 6). On ∼ 100 kyr timescales, higher cAR and cMAR occur during glacial periods (Fig. 3e). More efficient export of coccolith CaCO₃ via ballasting by terrigenous particles during glacials is unlikely to have driven this trend, given that compiled BoB monsoon records suggest a weaker glacial monsoon rather than a stronger one (e.g., Haridas et al., 2022). Our results from the central northern BoB are coherent with total CaCO₃ records from the northwestern BoB, showing peak carbonate MAR or content during glacials, interpreted to reflect increased productivity (Da Silva et al., 2017; Panmei et al., 2018; Phillips et al., 2014). In these studies, ISM weakening during glacial periods is suggested to lead to increased productivity and enhanced CaCO₃ deposition via reduced salinity stratification and increased nutrient input from below. We also document a strong precession component in the cAR record (Fig. 6a) that is highly coherent and in phase with the precession component of the G. ruber planktic δ18O record from the same core (Fig. S3). This suggests that South Asian monsoon variability on precessional timescales (Cheng et al., 2022) impacts both coccolithophore productivity and upper-ocean seawater δ18O (recorded by G. ruber) at our northern BoB study site, presumably via precipitation and runoff.

Several mechanisms could be invoked to explain monsoon impacts on productivity at northern BoB Site MD12-3412. Firstly, stronger wind-driven mixing during ISM maxima could break up stratification, increasing bottom-up nutrient inputs and fueling productivity, as proposed for southern BoB ODP Site 758 (of which Site U1443 is a re-drill) (Bolton et al., 2013). However, although wind intensity is highest during the ISM in our study region, modern data indicate that, in the northern BoB (in contrast to the southern BoB), the mixed layer is much shallower during the summer monsoon season than in winter (∼ 40 m vs. 60 m; Fig. 1), making this hypothesis unlikely. Secondly, productivity at Site MD12-3412 might be suppressed during ISM maxima due to increased runoff and salinity stratification, as proposed for several sites in the northwestern and northeastern BoB (Bolton et al., 2024; Phillips et al., 2014; Thirumalai et al., 2025; Zhou et al., 2020). To assess this hypothesis, we compared productivity variations on precession timescales recorded in the cAR record with a summer monsoon multi-proxy stack from an Arabian Sea core (Caley et al., 2011) and with a stratification record based on planktic foraminiferal δ18O gradients from southern BoB ODP Site 758 (Bolton et al., 2013) (Fig. 8). We note that these three records are on independent age models, so some differences in phasing between them may occur related to age model uncertainty. Figure 8 shows that, in 75 % of cases, maxima in northern BoB productivity on precession timescales coincide with minima in ISM strength, as suggested by the Arabian Sea ISM stack and by southern BoB stratification, which is higher during ISM minima due to weaker winds (green bars in Fig. 8). In the remaining 25 % of cases, the relationship is reversed, with northern BoB productivity maxima occurring during times of maximum monsoon strength and minimum southern BoB stratification (gray bars in Fig. 8). Based on this, we infer than coccolithophore productivity at northern BoB Site MD12-3412 was generally higher during ISM minima, when runoff plus precipitation was reduced and salinity stratification likely weakened, allowing nutrients to mix into the upper water column in the absence of a thick barrier layer. In the modern northern BoB, the barrier layer, defined as a parcel of water sitting between the base of the mixed layer and the top of the thermocline due to salinity stratification, is present year-round with a thickness of up to 60 m and a spring minimum and late-winter maximum thickness (Mignot et al., 2007; Thadathil et al., 2007).

However, in our record, notable exceptions to the above pattern occur during which productivity (cAR) peaks are associated with a strong monsoon (Fig. 8); thus other factors or mechanisms must be at play. The frequency and intensity of cyclones and eddies that trigger phytoplankton blooms in the BoB (e.g., Kuttippurath et al., 2021), along with interannual climate phenomena such as the El Niño–Southern Oscillation and the Indian Ocean Dipole (Currie et al., 2013), also have the potential to affect productivity patterns in the relatively oligotrophic central northern BoB, yet these relatively short-term events are challenging to reconstruct in the fossil record. The complexity of productivity dynamics in the stratified northern BoB was recently highlighted by a study showing suppressed productivity during both maxima and minima in monsoon strength over the last 20 kyr (Thirumalai et al., 2025), although this study was not long enough to resolve precession-scale variability.

In addition to the higher cAR and cMAR values discussed above, evidence for higher glacial productivity might also come from the higher relative abundance of different size classes of Noelaerhabdaceae coccoliths in core MD12-3412 (Fig. 5b, c). The ratio of small (< 2.5 µm) to medium to large (> 2.5 µm) Noelaerhabdaceae coccoliths in core MD12-3412 shows a clear glacial–interglacial pattern, with a dominance of small coccoliths during glacial intervals of the last 200 kyr (Fig. 5d). In the modern ocean and in the fossil record, a dominance of smaller Noelaerhabdaceae species is generally associated with higher-nutrient environments and higher coccolith carbonate export (Beaufort et al., 2022; Flores, et al., 1995; Flores et al., 2014; Hagino and Okada, 2004; Okada and Wells, 1997; Rickaby et al., 2007; Wells and Okada, 1996). In core MD12-3412, a higher ratio of small to larger Noelaerhabdaceae is accompanied by a higher NoMAR and a higher total cMAR during glacials, which we interpret to show enhanced coccolith CaCO₃ export and higher productivity. This is consistent with studies showing that a weaker ISM prevailed during late Pleistocene glacials, resulting in relaxed stratification and more nutrient entrainment into the mixed layer (Banerjee et al., 2024; Bolton et al., 2013; Clemens et al., 2021; Haridas et al., 2022; Zhisheng et al., 2011). However, unlike cAR and cMAR records, the size ratio of Noelaerhabdaceae shows no significant spectral power in the precession band (not shown), suggesting that the relative abundance of small vs. large Noelaerhabdaceae morphotypes may also be responding to other forcing mechanisms, for example, temperature or evolutionary processes. Aside from size classes within the Noelaerhabdaceae, none of the other main (F. profunda) or more minor (Umbilicosphaera spp., Helicosphaera spp., S. pulchra) coccolith groups display either clear glacial–interglacial variability or precession-band variance (Fig. 4). This is perhaps not surprising for the minor groups, given the relatively small glacial–interglacial temperature changes and the likelihood that this part of the BoB remained relatively salinity-stratified throughout the studied time interval. However, the lack of orbital-scale variance in the relative abundance of F. profunda in a tropical, stratified region such as the northern BoB is surprising.

Calcareous nannofossil assemblages in core MD12-3412 from 0–200 ka are numerically dominated by F. profunda coccoliths in terms of relative abundance (∼ 60 %–90 %; Fig. 4b). In addition, F. profunda is the main contributor to cAR and cMAR (Fig. 3d, e). This is consistent with studies in the BoB that show an unusually high dominance (60 %–90 %) of F. profunda in water samples (Liu et al., 2020), sediment traps (Stoll et al., 2007), and sediment cores (Bolton et al., 2024; Robinson et al., 2016; Zhou et al., 2020), especially in the salinity-stratified northern parts of the BoB. Typically, F. profunda lives in the deep photic zone, characterized by relatively stable, nutrient-rich waters and low light (Ahagon et al., 1993; Molfino and McIntyre, 1990; Quinn et al., 2005). The dominance of F. profunda in the BoB highlights its affinity for the distinct hydrographic characteristics of this region (Liu et al., 2020), and the low light levels in its sub-euphotic zone habitat have led to suggestions that this species may be mixotrophic (Poulton et al., 2007). A strong correlation between F. profunda fluxes and organic carbon fluxes in BoB sediment traps suggests that DPZ productivity represents an important part of total productivity (Stoll et al., 2007). This is supported by our finding that total coccolith MAR is dominated by the F. profunda contribution in core MD12-3412, despite this species' low individual mass (Fig. 3).

The relative abundance of F. profunda is a well-established proxy for primary productivity in much of the tropical open ocean, with high abundances typically associated with overall low productivity (Beaufort et al., 1999; Bolton and Stoll, 2025; Hernández-Almeida et al., 2019; Saavedra-Pellitero et al., 2022). However, this index is based on the vertical stratification of coccolithophore communities and thus does not necessarily directly reflect primary productivity in the upper photic zone. The percentage of F. profunda has been applied as a paleoproductivity indicator at continental shelf sites in the northeastern BoB (core MD77-176; Zhou et al., 2020) and in the northwestern BoB (Site U1446; Bolton et al., 2024), with maximum F. profunda abundances occurring during the Holocene period of maximum monsoon runoff and stratification as indicated by independent proxies or models. In contrast, our results suggest this proxy may not be universally applicable in the BoB. At our central northern BoB site, F. profunda relative abundances are quite constant over the last 200 kyr (Fig. 4b), despite the fact that coccolith export by this DPZ coccolithophore species is a major part of total coccolithophore productivity and increases in concert with total cAR/cMAR, with significant variance at precessional timescales. At this location, F. profunda coccoliths are not more relatively abundant when Noelaerhabdaceae coccoliths show lower ARs, but, rather, their absolute abundance increases in concert with Noelaerhabdaceae and total coccolithophore productivity, generally during weaker ISM intervals. One mechanism to explain this synchronized coccolithophore productivity increase in the upper and lower photic zones could be the influence of increased turbidity caused by river runoff during periods of strong ISM. In this scenario, high concentrations of suspended particles in the upper water column could limit light penetration to the deep photic zone and reduce F. profunda productivity despite the availability of nutrients, whereas salinity stratification simultaneously limits nutrient input into the upper photic zone from below, impacting Noelaerhabdaceae productivity. A greater number of long-term coccolithophore assemblage records from different regions of the BoB are needed to shed light on the factors controlling deep-photic-zone productivity in this unique oceanographic region.