In order to improve understanding of possible inter-species differences and microhabitat effects on the benthic Cd/Ca records, we analyzed Cd/Ca in three calcite (Cibicidoides pachyderma, Uvigerina peregrina and Globobulimina spp.) and one aragonite (Hoeglundina elegans) benthic foraminiferal species from core MD77-191. C. pachyderma is a shallow infaunal species U. peregrina and Globobulimina spp. are endobenthic species with intermediate and deep microhabitats, respectively (Fontanier et al., 2002). In core MD77-176, due to the limitation of calcitic species, we only measured Cd/Ca ratios in H. elegans shells. Moreover, Mn/Ca, Fe/Ca and Al/Ca ratios were also measured in all benthic foraminiferal samples to check the robustness of Cd/Ca results and the potential influence of contamination (i.e., oxides and sedimentary clay, Barker et al., 2003).

Each sample contained between 10 and 15 individuals picked from the 250–315 µm size fraction. Samples were gently crushed and cleaned to remove clays, organic matter and elemental oxides by using reductive and oxidative cleaning following previously published methods (Boyle and Keigwin, 1982; Barker et al., 2003). Each sample was dissolved in 0.075 N HNO3 and analyzed using a single collector sector field high-resolution inductively coupled plasma mass spectrometer (HR-ICP-MS) Thermo Element XR hosted at the GEOPS Laboratory (University Paris-Saclay, France).

The detailed instrumental settings and mother standard solutions are described in Ma et al. (2020). A blank consisting of the same 0.1 N HNO3 used to dilute the standards and samples was also analyzed. We removed the blank intensity values from all the raw intensities (including standards), and raw data were linearly drift-corrected by interspersing a drift standard every four samples. Standard curves were used to calculate elemental concentrations, coefficients of determination (r2) always being >0.9999 for all elemental ratios. The mean reproducibility and accuracy are 3.6 % and 7.5 %, respectively.

Benthic foraminiferal assemblages from core MD77-176 have already been published in Ma et al. (2019). For core MD77-191, a total of 72 samples were collected for benthic foraminiferal assemblage determinations. In each sample, benthic foraminifera (>150 µm) were extracted, counted and identified to species level following the taxonomical descriptions of various authors (e.g., Loeblich and Tappan, 1988; Jones, 1994; Holbourn et al., 2013). For core MD77-191, there is no material left in this old, small-diameter core, and so we used samples obtained earlier for stable isotope studies. Since the bulk weights of these samples were not recorded prior to sieving, we could not perform the calculation of the absolute abundance of foraminifera or accumulation rates. Thus, we only converted the individual counts to percentages with respect to the total benthic foraminifera present in each sample. In order to describe major faunal variations, we performed principal component analysis (PCA) on the variance–covariance matrix using the PAST software (Version 3.0, Hammer et al., 2001). Species present with a percentage >1 % in at least 1 sample were used for statistical analysis and diversity calculation.

To check the influence of oxide contaminants on the elemental ratios, Mn/Ca was systematically measured. The Mn/Ca of H. elegans from cores MD77-191 and MD77-176 ranges between 6.5–10 and 1–30 µmolmol-1, respectively. Such ranges are much lower than the 100 µmolmol-1 limit proposed by Boyle (1983). The Mn/Ca values obtained on the three calcite benthic foraminifera species from core MD77-191 – C. pachyderma (5–18 µmolmol-1), U. peregrina (3–23 µmolmol-1) and Globobulimina spp. (4–69 µmolmol-1) – are also all below 100 µmolmol-1 (Boyle, 1983). The Fe/Ca ratios are also lower than 1 mmol mol-1 in all samples from cores MD77-191 and MD77-176, in agreement with the limit proposed by Barker et al. (2003). In addition, Barker et al. (2003) concluded that no significant pollution by clay minerals would be expected when Al/Ca is <0.5 mmol mol-1. In all our samples, Al/Ca is below 0.5 mmol mol-1, indicating that the sample cleaning procedure was efficient.

All of the above results indicate that our samples were not affected by contamination.

Benthic foraminiferal species richness ranges between 16 and 36, and the total abundance fluctuates between 82 and 642 specimens (Table S2). Hyaline species are the dominant constituents (>80 %) and mainly consist of Bulimina aculeata, H. elegans, C. pachyderma, Uvigerina spp., Gyroidina broeckhiana, Globocassidulina subglobosa, Sphaeroidina bulloides, Gyroidinoides spp., Lenticulina spp., Melonis barleeanum and Globobulimina spp. (including Praeglobobulimina spp.) (in decreasing order of relative average abundance). Agglutinated taxa reach on average about 1.6 % and consist of Textularia sp., Martinottiella communis and Eggerella bradyi. The average percentage of porcelaneous species, characterized by Pyrgo elongata, Pyrgo murrhina, Pyrgo depressa, Pyrgoella irregularis, Quinqueloculina spp., Sigmoilopsis schlumbergeri and Spiroloculina spp., is about 5.1 %.

Furthermore, we merged species that share an ecological similarity, such as Globobulimina affinis, Globobulimina pacifica and Praeglobobulimina spp. into Globobulimina spp. A total of 74 samples and 55 groups and/or species were adopted to perform principal component analysis (PCA) in order to identify major faunal trends. The PCA analysis suggests that the benthic foraminifera could be grouped into three assemblages, with PC1 (positive and negative loadings) and PC2 (positive loadings) representing 42 % and 19 % of the total variance, respectively (Table 1). Moreover, compared with the total variance of PC1 and PC2, PC3 is the largest one and only explains 8 % of the total variance for the rest of the principal component scores (PCs). The species composition consists of H. elegans, Globobulimina spp. (positive loadings), Uvigerina peregrina and C. pachyderma (negative loadings) (Table 1). It seems that the main composition of assemblages (PC3) is quite similar to PC1 and does not show more information about the bottom conditions. Therefore, we only focus on PC1 and PC2 in the paper for the interpretation and do not present other PCs in the discussion.

Assemblage 1 can be defined as the combination of Bulimina aculeata and C. pachyderma, together with Pullenia bulloides and Ehrenbergina trigona (Figs. 3 and S2) and display high positive PC1 loadings. This assemblage dominated the foraminiferal record during the late Holocene (between 6 and 1.4 cal ka).

By contrast, assemblage 2, dominated by H. elegans and Bulimina manginata, exhibits high negative PC1 loadings, and corresponds to the record during the early Holocene (Figs. 3 and S2). Other quantitatively important contributors are C. wuellerstorfi and Globocassidulina subglobosa (Fig. S2).

Then, assemblage 3, dominated by Sphaeroidina bulloides and Gyroidinoides orbicularis, corresponds to the positive loadings of PC2, which is more important during the last deglaciation (Figs. 3 and S2). The associated species of assemblage 3 are Bulimina mexicana and Gyroidinoides soldanii (Fig. S2).

However, as the main composition of PC2 negative loadings is dominated by the same benthic species in assemblages 1 and 2, it is difficult to glean any additional information from this analysis. Thus, to clarify the discussion, we prefer to use three assemblages in the following rather than the two PCs.

In the modern ocean, benthic foraminifera Cd/Ca shows a positive correlation with Cdw and dissolved nutrients (phosphate and nitrate) (Boyle et al., 1976; Hester and Boyle, 1982). As aragonitic benthic foraminifera H. elegans faithfully records the bottom-water Cd concentrations (Cdw), Cd/Ca ratios can be converted to seawater Cdw with the appropriate relationship (Eq. 1), where the partition coefficient Dp≈1 for all water depths (Boyle et al., 1995; Bryan and Marchitto, 2010). 1Dp=Cd/CaforamCd/Cawater In contrast, the partition coefficient for calcite species changes with water depth. For water depths between 1150 and 3000 m, Dp was calculated based on the equation of Boyle (1992; Eq. 2). The seawater Ca concentration is assumed to be at a constant mean value of 0.01 mol kg-1 (Boyle, 1992). 2Dp=1.3+depth-1150×1.6/1850

The intermediate Cdw results based on the H. elegans Cd/Ca values of core MD77-191 range from 0.5 to 3.1 nmol kg-1 since 17 cal kyr BP (Fig. 4a), with a core top value of 0.80 nmol kg-1 in agreement with the estimated intermediate water depth modern Cdw (∼0.83 nmol kg-1) in the northern Indian Ocean (Boyle et al., 1995). The intermediate Cdw was also calculated from calcite benthic species C. pachyderma, U. peregrina and Globobulimina spp. from core MD77-191, with values ranging between 0.53–1.48, 0.52–1.04 and 0.26–0.65 µmolmol-1, respectively (Fig. 4a). The Cdw values of C. pachyderma and U. peregrina are within the same range. However, the H. elegans Cdw values are higher than those from the two calcite species, especially during the late Holocene. Moreover, the core top data of C. pachyderma and U. peregrina are also lower (∼0.7 and 0.69 nmol kg-1, respectively) than the modern estimated Cdw data (∼0.83 nmol kg-1) in the northern Indian Ocean (Boyle et al., 1995) (Fig. 4a). These depleted Cdw values may be related to the benthic foraminiferal microhabitat effect; indeed, U. peregrina is known to be strictly a shallow infaunal species, as well as C. pachyderma (Fontanier et al., 2002), differing from strictly epifaunal taxa, such as Cibicidoides wuellerstorfi (Mackensen et al., 1993).

Moreover, the deep infaunal Globobulimina spp. Cdw displays relatively much lower values and does not exhibit strong variations compared to the other species investigated in this study, displaying a general increasing trend from the last deglaciation to the Holocene. As Globobulimina spp. correspond to deep benthic infaunal species, this result may indicate a stable nutrient content of pore water as compared to other benthic taxa associated with bottom water (Fig. 4a). Thus, when tracking past changes in the bottom-water Cdw concentrations, the use of a strictly epifaunal species living at the water–sediment interface such as H. elegans appears to be more robust than using endofaunal species that live in contact with pore water.

Relative variations in the Cdw obtained from C. pachyderma and U. peregrina are in good agreement with the records obtained on H. elegans. Variations in H. elegans Cdw during the last deglaciation indicate a decrease of about ∼0.6 nmol kg-1 in the HS1 and YD periods, with a slight increase (0.9 nmol kg-1) during the warm B-A. Cdw results from core MD77-191 indicate a shift from the last deglaciation (∼0.7 nmol kg-1) to the late Holocene (∼1.59 nmol kg-1). During the Holocene, the Cdw records display relatively low values of around 0.9 nmol kg-1 in the 10–6 cal ka time interval and show a major shift at around 6.4 cal kyr BP with values rising up to 3.1 nmol kg-1.

For core MD77-176, the intermediate-water Cdw calculated from the H. elegans Cd/Ca records ranges between 0.6 and 1.7 nmol kg-1 over the past 18 cal kyr BP (Fig. 4b). Compared with intermediate Cdw from MD77-191, the Cdw record of core MD77-176 does not display any clear trend from the last deglaciation to the Holocene. However, a slight increase is observed since 6 cal kyr BP, in agreement with the MD77-191 intermediate Cdw records. In addition, even though the MD77-176 record has a lower time resolution, it displays a shorter maximum (1.3 nmol kg-1) during the 13.4–11 cal ka time interval.

To summarize, among the three calcite benthic taxa and the aragonitic benthic species H. elegans, the Cd/Ca records of H. elegans appear to be the most suitable for tracking past Cdw changes at intermediate water depth through time. Thus, in the following discussion, we will only focus on the intermediate Cdw calculated from the H. elegans Cd/Ca from both studied cores.

Comparing the geochemical records to the benthic assemblages, we can observe similar patterns. For core MD77-191 from the southeastern Arabian Sea, three benthic assemblages were identified since the last deglaciation. S. bulloides and Gyroidinoides orbicularis are major components of assemblage 3 (during the last deglaciation), together with B. mexicana and Gyroidinoides soldanii (Figs. 3 and S2). S. bulloides and B. mexicana are found in intermediate to high organic carbon flux rate regions (e.g., Schmiedl et al., 2000; Eberwein and Mackensen, 2006, 2008), while G. orbicularis and G. soldanii are associated with well-oxygenated and oligotrophic environments (Peterson, 1984; Burmistrova and Belyaeva, 2006; De and Gupta, 2010). Thus, assemblage 3 reflects mesotrophic environments and/or well-ventilated conditions during the last deglaciation. Although millennial-scale changes in the benthic foraminiferal assemblages during the last deglaciation could not be observed, assemblage 3 seems at least partly consistent with previous studies in the northern Indian Ocean based on multiple geochemical proxies (e.g., benthic δ13C, intermediate water [CO32-] and εNd records); these studies have revealed the presence of better-ventilated waters, which might correspond to AAIW, during the HS1 and YD (e.g., Yu et al., 2018; Ma et al., 2019, 2020).

Benthic foraminiferal assemblage 2 predominates during the early Holocene and is characterized by H. elegans and B. manginata as major contributors (Figs. 3 and S2). The other important contributors are C. wuellerstorfi and G. subglobosa. B. manginata is found in high organic carbon flux rate conditions (De Rijk et al., 2000; Eberwein and Mackensen, 2006, 2008). However, previous studies on H. elegans, C. wuellerstorfi and G. subglobosa indicate that these species correspond to high levels of dissolved oxygen and oligotrophic settings (e.g., Altenbach et al., 1999; Fontanier et al., 2002; Murgese and De Deckker, 2005, 2007; De and Gupta, 2010). Periods dominated by these taxa probably indicate high oxygen levels and an oligotrophic environment. This is consistent with previous studies in the area, based on benthic foraminiferal δ13C and Δ14C age difference (e.g., Naqvi et al., 1994; Bryan et al., 2010) (Fig. S3). Indeed, the glacial to Holocene benthic δ13C shifts (0.35 ‰–0.4 ‰, vs. Pee Dee Belemnite, PDB) at intermediate–deep water depth in the northern Indian Ocean are interpreted as reflecting an increased contribution of better-ventilated deep water NADW in IDW, during the Holocene (e.g., Naqvi et al., 1994; Ma et al., 2019) (Fig. S3). Furthermore, the increased benthic–planktonic (B-P) age offsets and depleted εNd records obtained from the same core site could also reflect the enhanced influence of NADW in IDW during the Holocene, which is characterized by well-ventilated conditions and depleted nutrient concentrations (modern Cdw, ∼0.2 nmol kg-1) (Poggemann et al., 2017; Yu et al., 2018; Ma et al., 2019). The impact of this change in the IDW composition can be recorded at intermediate water depth since the deep water masses are transformed to an upward flow during their pathway, thus being a potential contribution to intermediate-depth water masses (Naqvi et al., 1994; You, 2000). Although the intermediate benthic δ13C record from core MD77-191 is missing for the Last Glacial Maximum (LGM), the average value for the Holocene (∼0.31 ‰, vs. PDB) is consistent with previous studies carried out in the northern Indian Ocean; combined with the opposite trend between δ18Oivc records and intermediate-water temperature from MD77-191 (Ma et al., 2020), all these records suggest well-ventilated conditions (Fig. S3). To summarize, the predominance of benthic foraminifera assemblage 2 in the early Holocene seems to reflect better-ventilated water masses, related to an enhanced contribution of NADW in IDW at the core site, as already observed in previous studies (Poggemann et al., 2017; Yu et al., 2018; Ma et al., 2019, 2020).

By contrast, B. aculeata and C. pachyderma are major components of assemblage 1 (during the late Holocene), together with P. bulloides and E. trigona (Figs. 3 and S2). Living B. aculeata have a widespread distribution, with a preference for water depths ranging from 1500 to 2500 m, and are typically associated with high organic carbon fluxes (Mackensen et al., 1995; Almogi-Labin et al., 2000; Caulle et al., 2015). P. bulloides is a shallow infaunal species, which prefers mesotrophic environments and shows adaptability with respect to oxygen concentration in the Arabian Sea (Gupta and Thomas, 1999; Caulle et al., 2015). E. trigona is commonly recorded in low-oxygen habitats (Caulle et al., 2015). We thus interpret assemblage 1 as indicating relatively low-oxygen and meso- to eutrophic bottom-water conditions during the late Holocene (6–1.4 cal ka). However, the lower oxygen concentrations reflected by benthic fauna 1 seem to be the opposite of what would be expected under an enhanced influence of better-ventilated NADW in IDW during the Holocene in the northern Indian Ocean. Thus, another process has to be explored to combine our observations. To do that, we can use the relative abundance of Globigerina bulloides, a proxy for upwelling activity, that increased in the late Holocene in core MD77-191, suggesting an increased productivity in the southeastern Arabian Sea (Bassinot et al., 2011) (Fig. 5). This record is synchronous with the benthic foraminiferal assemblage 1 (during the late Holocene). Thus, increased surface productivity during the late Holocene could have introduced more organic matter into the intermediate water, leading to depleted oxygen conditions. By contrast, benthic assemblages 2 and 3 (during the last deglaciation and early Holocene; 17–6 cal ka) are associated with low G. bulloides abundances, suggesting lower productivity in the southeastern Arabian Sea during this period (Bassinot et al., 2011) and thus indicating that intermediate water masses were characterized by higher bottom-water oxygen conditions and a lower flux of organic matter. Therefore, all of these elements suggest that changes in primary productivity seem to be an important factor impacting the distribution of benthic assemblages at the core MD77-191 site, especially during the Holocene.

In order to examine the relationships between intermediate Cdw and these different processes (surface productivity and/or water mass ventilation) in the eastern Arabian Sea, we can compare the MD77-191 Cdw values with the relative abundance of G. bulloides and benthic foraminiferal assemblage analyses from the same core MD77-191, together with the records for Corg and the G. bulloides percentage obtained from core SK237 GC04 (1245 m, southeastern Arabian Sea, Naik et al., 2017) (Fig. 5). Indeed, the total organic carbon (Corg) could also be used as a qualitative indicator of past productivity and/or bottom-water ventilation changes (Naidu et al., 1992; Canfield, 1994; Calvert et al., 1995; Naik et al., 2017). Despite a lower resolution for MD77-191 H. elegans Cdw records, when compared to the Corg and the G. bulloides percentage from core SK237 GC04, all of them seem to exhibit similar trends at a long timescale even though some small-scale discrepancies can be observed at a millennial timescales (Fig. 5). From the last deglaciation to the late Holocene, the Cdw record displays a significant shift from ∼0.7 nmol kg-1 to about twice the values of ∼1.59 nmol kg-1. The intermediate Cdw values are thus extremely high during the late Holocene and synchronous with the higher values of Corg and G. bulloides percentage records. These observed similar trends suggest that the increased surface productivity at the core site during the late Holocene is associated with higher intermediate Cdw values. Moreover, previous studies have suggested that increased Cdw values (>1 nmol kg-1) could correspond to elevated surface productivity (Bostock et al., 2010; Olsen et al., 2016). However, at a millennial timescale, we also observed several decreases in intermediate Cdw values (∼0.81 nmol kg-1) during the late Holocene, reaching nearly similar values during the last deglaciation (Fig. 5). Thus, the variations in the Cdw values cannot be fully associated with variations in the surface productivity.

As mentioned before, during the Holocene, an increased influence of NADW in IDW was observed in the northern Indian Ocean (Yu et al., 2018; Ma et al., 2019, 2020). NADW is characterized by a depleted nutrient content (modern Cdw, ∼0.2 nmol kg-1; Poggemann et al., 2017), and its contribution to IDW may affect the intermediate Cdw by deep water masses upwelling when flowing northward. However, during the late Holocene, benthic foraminiferal assemblage 1 is associated with lower oxygen concentrations, which seem to be inconsistent with an enhanced influence of better-ventilated NADW in IDW in the northern Indian Ocean. Therefore, this apparent discrepancy seems to indicate that deep–intermediate water mass variations are not an important control during the Holocene in this area, although we could not fully exclude the influence of NADW in IDW at a millennial timescale. Moreover, there is no clear evidence for such a millennial-scale variability in the IDW and/or NADW circulation in the studied area. Thus, we suggest the intermediate Cdw at the core MD77-191 site may be mainly influenced by surface productivity, especially during the Holocene.

In the Bay of Bengal, the benthic assemblages of core MD77-176 suggest that the intermediate water masses were characterized by oligotrophic to mesotrophic conditions and/or well-ventilated environments during the Holocene (Ma et al., 2019), associated with much lower surface productivity (Fig. S4). This observation is in agreement with low primary productivity during the Holocene reconstructed by the relative abundance of coccolith species Florisphaera profunda from the same core MD77-176 in the northeastern BoB (Zhou et al., 2020). In the modern ocean, Prasanna Kumar et al. (2001) indicate that primary productivity in the BoB is much lower than in the Arabian Sea, the lower surface productivity resulting from the large freshwater input from rivers and direct rainfall resulting from enhanced Indian summer monsoon precipitation (e.g., Vinayachandran et al., 2002; Madhupratap et al., 2003; Gauns et al., 2005). Moreover, when we compare the average Cdw value of core MD77-176 from the BoB (∼0.9 nmol kg-1) with results from core MD77-191 in the Arabian Sea (∼1.2 nmol kg-1), lower values, especially during the late Holocene, are in agreement with the benthic assemblages.

To sum up, variations in the benthic assemblages seem to be associated with changes in the deep water mass ventilation and/or organic matter flux, linked to surface productivity. The benthic foraminiferal fauna are consistent with the Cdw record of core MD77-191 particularly during the late Holocene (6–1.4 cal ka). Thus, our results seem to show that the Cdw record is mainly controlled by changes occurring at the surface, especially during the Holocene. However, at millennial timescales, such during the HS1 and YD, the percentages of planktonic species G. bulloides from cores MD77-191 and SK237 GC04 all indicate modest paleo-productivity, the opposite of what is suggested by the results of core MD77-191 Cdw and the Corg record obtained from core SK237 GC04. This interval is also marked by enriched G. ruber δ18O values, indicating a weaker monsoon and reduced freshwater inputs (Naik et al., 2017). This apparent discrepancy may be related to changes in the intermediate water mass sources and/or ventilation during the last deglaciation.

So, in the next sections, we discuss (i) processes controlling surface productivity and (ii) changes in the intermediate-water circulation, both of them being potential drivers of the observed variations.

During the Holocene, the intermediate-water Cdw records obtained from cores MD77-191 and MD77-176 seem to display depleted values in the early Holocene, followed by an abrupt increasing trend at the middle Holocene and then reaching higher values on average (despite a short-timescale variability) during the late Holocene.

Of the two cores, core MD77-176, located in the northeastern BoB, shows the lowest intermediate Cdw (down to ∼0.83 nmol kg-1) during the 10–6 cal ka time interval. Observations described above suggest that this low in Cdw resulted from low primary productivity and thus reduced fluxes of organic matter to the intermediate depths. We attribute this evolution to monsoon variation. Indeed, the early Holocene Climate Optimum (10–6 cal ka) is characterized by enhanced monsoon precipitation (Marzin et al., 2013; Contreras-Rosales et al., 2014) (Fig. 6d–f) that resulted in increased freshwater discharge from the Ganges–Brahmaputra river system and from the Irrawaddy River. However, the distribution of chlorophyll in surface water of the western BoB suggests a low annual productivity, indicating that the BoB is not significantly influenced by the riverine nutrient input (Zhou et al., 2020). Thus, it is likely that this increase in freshwater drove pronounced ocean stratification in the northeast BoB, which could impede the nutrient transfer from the intermediate and/or deep layer to the euphotic upper seawater column and then induce low productivity. A similar low in Cdw values is observed in the reconstructed intermediate-water Cdw record from core MD77-191 during the early Holocene, with values descending to ∼0.92 nmol kg-1 in the 10–6 cal ka time interval. These low values of intermediate Cdw are coeval with low surface productivity as recorded by the G. bulloides percentage and low values in Corg content from SK237 GC04 in the Arabian Sea (Fig. 5). These variations are also recorded in changes in benthic assemblages, with the occurrence of assemblage 2 associated with high oxygen levels and an oligotrophic environment (Fig. 3). Off the southern tip of India, we cannot reject the possibility that increased monsoon precipitation and enhanced freshwater runoffs in the BoB during the early Holocene, inducing a stronger stratification, could explain part of the decrease in surface primary productivity. Yet, at this site, another explanation prevails which is related to the decrease in summer monsoon wind intensity that drives local Ekman pumping. As shown by Bassinot et al. (2011), the productivity variations at the southern tip of India are inversely related to the evolution of upwelling activity along the Oman Margin, to the west of the Arabian Sea. Based on a data–model comparison, Bassinot et al. (2011) showed that this anti-correlation can be attributed to the northward shift in the intertropical convergence zone (ITCZ) when boreal summer insolation reached a maximum in the early Holocene (Fig. 6a); this ITCZ location results in enhanced summer monsoon wind intensity and an increase in the associated Ekman pumping in the west of the Arabian Sea and along the Oman margin, while it weakens at the southern tip of India. This process may thus induce a decrease in surface productivity in the southeastern Arabian Sea.

In addition, Naik et al. (2017) pointed out the co-existence of low productivity during the early Holocene in the BoB and to the south of India, in agreement with our data that clearly show the impact of such a reduction in surface primary productivity on the intermediate-water Cdw. These authors suggested a direct relationship between intense monsoon rainfall and reduced surface productivity. However, the northeastern BoB received a much larger amount of river input than the southern tip of India during the early Holocene (Marzin et al., 2013). Thus, it seems reasonable to propose that the northeastern BoB is more affected by the salinity-related stratification effect, while the southern tip of India is more affected by the decrease in wind intensity (Bassinot et al., 2011) with enhanced stratification being potentially made stronger by an additional freshwater effect, although weaker than in the BoB. Ultimately, both climatic features (summer wind intensity and precipitation) are directly under the control of monsoon evolution resulting from the orbital forcing of low-latitude boreal summer insolation.

By contrast, higher intermediate Cdw values from core MD77-191 associated with higher G. bulloides relative abundances and Corg from core SK237 GC04 during the 5.2–2.4 cal ka time interval could indicate enhanced productivity during the mid to late Holocene (Naik et al., 2017) (Fig. 5). To a lesser extent, this is also observed in the records from the northern BoB for the same time period. These changes are consistent with the weakened summer monsoon intensity, with less rainfall during the late Holocene, as observed in the BoB using core MD77-176 seawater δ18O and core SO188-342KL δDAlk-ic records (Marzin et al., 2013; Contreras-Rosales et al., 2014; Fig. 6e–f). In addition, this is also strongly supported by the δ13Cwax records from the Lonar Lake on the Indian continent (Sarkar et al., 2015; Fig. 6d) and a progressive increase in monsoon summer winds to the south of India (Bassinot et al., 2011). These observations could also strongly support the hypothesis that the major control on surface productivity is linked to monsoon evolution in the BoB and at the southern tip of the Arabian Sea during the Holocene (Bassinot et al., 2011; Naik et al., 2017; Zhou et al., 2020).

During the last deglaciation, short events have been recorded at the site of core MD77-191 during the 16–15.2 (HS1) and 12.6–11 cal ka (YD) time intervals (Fig. 5). The low Cdw values in the MD77-191 record are coeval with reductions in Corg in core SK237 GC04 during the HS1 and YD periods (Fig. 5). According to previous studies, extremely high Cdw values (>1 nmol kg-1) were reported to have been associated with enhanced surface productivity (Bostock et al., 2010; Olsen et al., 2016). However, the range of values of intermediate Cdw (0.58–0.85 nmol kg-1, HS1; 0.5–0.8 nmol kg-1, YD) from core MD77-191 during the last deglaciation is much lower compared with the Holocene Cdw values (>1 nmol kg-1) and thus may be associated with other processes such as a better ventilation, changes in the water mass source and/or depleted surface productivity (Fig. 6). Significant decreases in G. bulloides relative abundance of SK237 GC04 (Naik et al., 2017) and MD77-191 records were observed from the HS1 to B-A (Bassinot et al., 2011), and thereafter slight increases occurred in the YD (Fig. 5). These high values at both core sites during the HS1 and YD may indicate an enhanced surface productivity during these intervals (Fig. 5). This should have led to increased intermediate Cdw and organic matter preservation under low oxygen concentration conditions during the HS1 and YD. However, despite a low resolution for the MD77-191 Cdw record during the last deglaciation, we do not observe high values of intermediate Cdw during the HS1 and YD (∼0.7 nmol kg-1) compared with the late Holocene (∼1.59 nmol kg-1), especially at 16.5–16 cal ka. Although we cannot fully discard the influence of surface productivity on the intermediate Cdw in these time intervals, this apparent discrepancy seems to provide another piece of evidence for the influence of changes in water masses and/or ventilation during the HS1 and YD, in line with previous studies and proxies in the northern Indian Ocean (Bryan et al., 2010; Yu et al., 2018; Ma et al., 2019, 2020).

Moreover, an increase in benthic δ13C values is observed during the HS1 and YD in the northern Indian Ocean (e.g., Duplessy et al., 1984; Curry et al., 1988; Naqvi et al., 1994; Jung et al., 2009; Ma et al., 2019) (Fig. S3). The increase in the different benthic δ13C records during the HS1 and YD in the western Arabian Sea, Pacific Ocean and BoB is interpreted as reflecting the northward expansion of AAIW (Pahnke and Zahn, 2005; Jung et al., 2009; Ma et al., 2019) (Fig. S3). The decreased B-P age obtained from marine sediment cores from the Arabian Sea and the Bay of Bengal during the same intervals could confirm enhanced vertical mixing in the Southern Ocean (Bryan et al., 2010; Ma et al., 2019). The transition in the εNd and Δ14C records during the deglaciation also indicates a strong northward penetration of AAIW within the North Atlantic and Bay of Bengal (e.g., Cao et al., 2007; Pahnke et al., 2008; Pena et al., 2013; Yu et al., 2018). In addition, during the HS1 and YD, a decrease in the [CO32-] record from core MD77-191 also suggests the release of CO2 from the deep ocean in the deglacial period through the expansion of AAIW (Ma et al., 2020). These time intervals are associated with better ventilation in the Southern Ocean (e.g., Anderson et al., 2009; Skinner et al., 2010), which led to enhanced vertical ventilation resulting in increased production of intermediate water masses (AAIW) (Anderson et al., 2009).

As mentioned before, previous studies have suggested an enhanced northward flow of southern sourced intermediate water mass AAIW, observed also in the Atlantic, Pacific and Indian oceans during the last deglaciation (e.g., Pahnke et al., 2008; Bryan et al., 2010; Poggemann et al., 2017; Yu et al., 2018; Ma et al., 2019, 2020), indicating that the source of intermediate water masses may be partly the same in these oceans. Thus, as the benthic δ13C values collected from the north Indian Ocean could better constrain the influence of AAIW in the two studied cores (Naqvi et al., 1994; Jung et al., 2009; Ma et al., 2019, 2020) (Fig. S3), we can also compare the range values of AAIW Cdw from both studied cores with data from the Atlantic and Pacific oceans at intermediate water depth during the HS1 and YD (Cdw, 0.3–0.9 nmol kg-1; Umling et al., 2018; Valley et al., 2017). Thereafter, we could get the ranges of Cdw–δ13C values of AAIW during these intervals, based on the benthic δ13C records in the Indian Ocean (Naqvi et al., 1994; Jung et al., 2009; Ma et al., 2019, 2020) as well as benthic Cdw values from the Pacific and Atlantic oceans (Valley et al., 2017; Umling et al., 2018) at intermediate water depths (Fig. 7). Unfortunately, the resolution of both intermediate Cdw and benthic δ13C from core MD77-176 (northeastern BoB) are very low for the HS1 and YD events, making it difficult to extract reliable information. Thus, we have decided to focus on the results from core MD77-191 (0.5–0.85 nmol kg-1) during these two time intervals; these results are in good agreement with the collected dataset (Fig. 7). Thus, the benthic Cdw results provide new evidence for tracking the northern flow of AAIW in the northern Indian Ocean, which increased during the HS1 and the YD.

Taken together, Cdw, the B-P age offset, benthic δ13C, εNd and Δ14C records reported from the northern Indian Ocean all suggest strong upwelling and enhanced northern flow of AAIW from the Southern Ocean during the HS1 and the YD. Thus, the variations in these records can provide strong evidence for the hypothesis that Southern Ocean upwelling played a vital role in the increase in atmospheric CO2 in the deglacial period (Anderson et al., 2009; Skinner et al., 2010, 2014). However, Kohfeld et al. (2005) suggested that although physical processes (such as ventilation) are involved in the glacial–interglacial atmospheric CO2 change, the biological pump may also contribute nearly half of the observed changes in CO2 during the glacial–interglacial transitions. As shown above, the HS1 event is characterized by reduced surface productivity, as revealed by the lower percentage values of G. bulloides in core MD77-191 (Bassinot et al., 2011) and by several studies of cores located in the eastern and western Arabian Sea within the Oxygen Minimum Zone (e.g., Schulz et al., 1998; Altabet et al., 2002; Ivanochko et al., 2005; Singh et al., 2006, 2011; Naik et al., 2017). This reduced productivity at a millennial timescale suggests that the entire biological factory was related to the reduced monsoon intensity during the North Atlantic Heinrich events (e.g., Singh et al., 2011; Naik et al., 2017). Thus, a weaker biological production could also have contributed to the two-step increase in atmospheric CO2 during the last deglaciation, at least for the HS1 period.