We produced smear slides from 71 freeze-dried samples taken from Hole 722B (Supplement S2) following the quantitative drop technique of Bordiga et al. (2015). On each slide, at least 47 field views were counted until at least 300 specimens were recorded or until over 190 field views were reached for samples containing very low abundances. During counting, nannofossils were identified to the species level whenever possible. The occurrence of diatom frustules (including pennate and centric forms), as well as other biogenic silica fragments (including silicoflagellates and radiolarian fragments), were quantitatively recorded without further taxonomic identification (Supplement S2). All recorded nannofossil taxa, as well as the siliceous skeletal fragments (including diatoms, silicoflagellates and radiolarian fragments), were then converted into absolute abundances per gram of sediment, according to Bordiga et al. (2015), with portions of the dataset already published (Bialik et al., 2020a). In addition to the above-described quantification, the high amount of biogenic silica recorded in some sections often dilutes absolute nannofossil abundances. We calculated nannofossil and diatom fluxes to alleviate the potential dilution of nannofossil abundance due to high biogenic silica fluxes for the studied interval (see Sect. 3.5).

For foraminifera analysis, 28 samples were freeze-dried, weighed, and wet-sieved using mesh sizes of 250, 125, and 63 µm. After sieving, sample residues were oven dried at 40 ∘C. For quantitative foraminiferal analyses, the size fractions > 250 and 250–125 µm were examined under a stereomicroscope (Zeiss V8). In each sample, at least 200 specimens were picked and identified. In eight samples, less than 200 specimens were found in the available material. When necessary, samples were split into smaller aliquots (splits). The total number of foraminifera in the sediment (N g-1) was calculated from the number of the counted specimen and the number of splits. Relative abundances (%) were calculated for each species (see Supplement S3 for details).

All applied statistical and ordination methods were performed using PAST4 (v. 4.11 released 13 September 2022; Hammer et al., 2001). The applied methods include correlation matrices between nannofossil taxa and XRF-based environmental proxy data for dust flux and Mn depletion and calcite-equivalent CaCO3 content (Supplement S4), as well as the abundance of diatoms (Supplement S1). Percentage data were then arcsine-transformed before cluster analyses and ordination methods. The arcsine transformation was applied to generate a statistically viable dataset suitable for the applied clustering and ordination methods (Sokal and Rohlf, 1995; Hammer and Harper, 2006; Auer et al., 2014, 2019; Bialik et al., 2021) and utilizes the universal paired group method with arithmetic mean (UPGMA) with Bray–Curtis distance. Cluster stability was further evaluated by using UPGMA clustering with Euclidian distance and Ward's method.

The contributing taxa of each cluster were subsequently evaluated based on similarity percentage (SIMPER) analysis (Bray–Curtis similarity). The correspondence of nannofossil variability within each sample with environmental parameters was investigated using the non-metric multidimensional scaling (nMDS), where geochemical proxy data (see Sect. 2; Fig. 3) were used as environmental variables and visualized as vectors within the two-dimensional coordinate space of the nMDS. Additionally, several diversity indices (see Supplement S1), including the Shannon H' diversity, were automatically calculated for the calcareous nannofossil assemblage (Hammer and Harper, 2006).

In addition to the paleobiological data generated for this study, we further use a suite of previously published geochemical proxy data (Bialik et al., 2020a), which we utilize as additional lines of evidence to anchor the observed assemblage variation within a multiproxy framework. In brief, we apply CaCO3 and TOC combined with fluxes of siliceous fragments (see Sect. 3.5 for details) as productivity proxies. Benchtop X-ray fluorescence-derived elemental ratios further supplement this interpretation, where we apply Mn / Al ratios to quantify Mn redirection (see Bialik et al., 2020a) based on the model of Dickens and Owen (1994). The available XRF data were also used to generate a dust flux proxy based on the elemental ratio of (K + Al) / (Fe + Ti + Zr), as defined by Kuhnt et al. (2015). This dust flux proxy allows determining the accumulation of Fe-, Ti-, and Zr-bearing heavy mineral phases compared to elements predominantly present in clay minerals (Al + K). We interpret this proxy as a qualitative proxy for wind-derived dust flux and thus varying wind strength at Site 722. Dust flux and wind speed are intrinsically linked to Africa's progressive aridification due to the uplift of the Arabian Peninsula (Zhang et al., 2014; Sarr et al., 2022). The published δ15N is also discussed in the context of the new assemblage data. Tripathi et al. (2017) interpret δ15N values > 6 ‰ as an indicator for significant water column denitrification in ocean basins with oxygenated bottom waters. Later, Bialik et al. (2020a) also used this proxy interpretation for the Middle to Late Miocene interval at Site 722, which will be followed herein.

To quantify flux rates, we applied moisture and density (MAD)-derived bulk density data generated during Leg 117 (Shipboard Scientific Party, 1989) to calculate mass accumulation rates (MAR). To calculate bulk MARs, we applied linear interpolated dry bulk density for each sample point using the calculation: BMAR=DBD×LSR10, where BMAR is the bulk mass accumulation rate (in g cm-2 kyr-1), DBD is the dry bulk density (in g cm-3) based on shipboard MAD data, LSR is the linear sedimentation rate (in metres per million years) calculated based on the age model of Bialik et al. (2020a), generated bulk MARs were subsequently used to calculate mass fluxes of TOC, and CaCO3 is given in g cm-2 kyr-1. Fossil fluxes are given as nannofossil accumulation rates (NAR), and diatom accumulation rates (DAR) as numbers of liths and/or frustules per cm-2 kyr-1, which are calculated by multiplying the BMAR with the number of individuals per gram of sediment.

Out of 28 samples, 1 sample (722B-34X-3W 30-32, ca. 10.2 Ma) was barren in planktonic foraminifera. In the remaining 27 samples, 27 taxa of planktonic foraminifera were identified (Supplement S3). The planktonic foraminifera preservation was overall good but decreases downhole. The foraminifera tests were found to be moderately pyritized. Of these taxa, five (Globigerinoides ruber, Globorotalia menardii, Neogloboquadrina acostaensis, Paragloborotalia mayeri) have their stratigraphic first or last occurrence within the studied interval. All recorded taxa were grouped according to their environmental preferences following established environmental assignments of either mixed-layer taxa, open-ocean thermocline taxa, open-ocean sub-thermocline taxa, upwelling taxa, or unknown (Table 2).

Through the studied interval, thermocline species and mixed-layer taxa are the most abundant (abundance reaches more than 50 %). Both mixed-layer and upwelling taxa increase in prevalence through the studied interval, while thermocline species decrease. A sharp drop in thermocline taxa occurs between 11 and 10 Ma, corresponding to the disappearance of Paragloborotalia mayeri, the dominant taxa until that time. Mixed-layer taxa remain at a near-stable level from 11 Ma onwards. Upwelling taxa are not represented in two samples between 11 and 10.8 Ma, after which this group exhibits a steady increase until the end of the studied interval. Sub-thermocline taxa are present between 9.0 and 9.5 Ma and account for only a small fraction (less than 3 % at most) of the assemblage.

Based on the above results, we interpret the analysed samples in the context of their taphogroups. Taphogroups represent the total preserved fossil assemblage deposited at a given time in the past. Therefore, samples assigned to contain the same taphogroup can be assumed to reflect similar local surface water conditions at Site 722.

Taphogroup 1a. TG1a is dominated by small reticulofenestrids. We therefore interpreted this TG as indicative of high nutrient levels, facilitating the proliferation of small bloom-forming placoliths (primarily Reticulofenestra minuta; see Table 1, Supplement S5). Small reticulofenestrids are commonly associated with high terrigenous nutrients in near-shore environments (see references in Table 1). However, as Site 722 was located in the open ocean throughout the study interval (Bialik et al., 2020a), changes in terrigenous nutrient sources are unlikely, and a different mechanism must be invoked for this dominance of small reticulofenestrids. Studies based on coccolithophore cultures indicate that the proliferation of small placoliths may result from nitrogen limitation in a highly productive open marine environment. For example, Paasche (1998) showed that modern-day coccolithophores tend to increase the formation of small placoliths during N limitation. Hence, we assume that the proliferation of small reticulofenestrids in the open-ocean results from increasing nitrogen limitation compared to other macro- or micronutrients. Such N-limited environments often persist in settings with high productivity due to rapid N-loss during denitrification (Paerl, 2018), which would fit with the above interpretation of small reticulofenestrid proliferation at Site 722, offshore Oman.

Taphogroup 1b. The presence of common C. floridanus in combination with abundant small- and medium-sized reticulofenestrids within this assemblage indicates elevated nutrient levels compared to a fully oligotrophic assemblage (see Table 1 and Supplement S5). The very high but not dominant abundance of small reticulofenestrids may also point to N-limited nutrient sources (see TG 1a). This will be analogous to the fringes of the modern-day Arabian Sea upwelling cell, where nitrogen may be the primary limiting nutrient (Anju et al., 2020), hinting at the presence of upwelling during TG1b, which was more confined to the coast of the Arabian Peninsula and did not fully reach Site 722. Compared to other TGs, the overall high diversity suggests intermittent (likely seasonal) oligotrophic conditions, which may also point towards phosphate co-limitation when upwelling is limited. We thus interpret TG 1b as reflective of open marine conditions with only somewhat elevated nutrient levels compared to an open-ocean gyre. Primary nutrient supply, however, is still controlled by nutrients derived through the remineralization of locally produced particulate organic matter (Cullen, 1991), likely supplied to the surface water through seasonal mixing during weaker summer monsoons.

Taphogroup 2. Within TG 2, common C. floridanus occurs together with medium and large reticulofenestrids, commonly associated with warmer water temperature, a deep nutricline, and potentially elevated nutrient conditions (Table 1 and Supplement S5). Therefore, we interpret this TG to reflect open marine conditions without directly indicating upwelling-derived nutrients. Nutrients were likely mainly derived through particulate organic matter remineralization, with low external nutrient influx through upwelling or terrigenous nutrients.

Taphogroup 3. Previous studies (Auer et al., 2014; Lohmann and Carlson, 1981) generally associated large reticulofenestrids with high nutrient conditions. Imai (2015) states that dominant large reticulofenestrids and common discoasterids indicate low nutrient conditions and a deep nutricline compared to a high abundance of small reticulofenestrids (Table 1 and Supplement S5). However, this interpretation is questioned by the association of TG 3 with high TOC, high dust flux, and high silica accumulation rates, indicating strong upwelling conditions (Fig. 4b). Although, similar co-occurrences of diatoms and discoasterids were previously recorded in the eastern equatorial Pacific and the Mediterranean (Backman et al., 2013).

While difficult to ascertain, the association of TG 3 with high dust flux and thus additional iron fertilization may represent exceptionally high primary productivity (Guieu et al., 2019). Furthermore, modern analogues based on large Geophyrocapsa taxa, descendants of the genus Reticulofenestra (Samtleben, 1980; Perch-Nielsen, 1985; Nannotax 3, 2023), are more abundant in high nutrient upwelling zones (Bollmann, 1997). However, seasonality between summer monsoon and weak or absent winter monsoon could also address this discrepancy in the interpretation of TG 3 with available environmental data. Diatom and coccolithophore accumulation occur in such a setting in different nutrient regimes. Modern-day culture studies of coccolithophores (Paasche, 1998) also show that the calcification of coccolithophores increases during nitrogen excess and phosphate limitation.

Therefore, we interpret TG 3 as indicative of likely the strongest summer monsoon-controlled upwelling for our Middle to Late Miocene study interval. Conversely, a relatively weak winter monsoon resulted in a deep nutricline for the rest of the year.

Taphogroup 4a. Taphogroup 4a is not dominated by a specific reticulofenestrid size range while also containing a diverse assemblage in general (see Table 1). We therefore interpret this TG to show weaker upwelling conditions compared to TG3 or TG 1a during transient climatic conditions. Furthermore, weaker productivity is implied by a stronger association of TG 4a with higher Mn / Al values (Fig. 4b).

Taphogroup 4b. The high dominance of large reticulofenestrids of TG 4b would suggest elevated, upwelling-derived nutrient levels in a temperate upwelling zone (see TG3 above). Furthermore, the level of calcification rates, as found in experimental studies by Paasche (1998), may also indicate P-limitation. High nutrient conditions are corroborated by the general association of TG 4b with high diatom fluxes, TOC, and dust flux in the nMDS (Fig. 4b).

Taphogroup 4c. Taphogroup 4c is defined by both medium and large reticulofenestrids (Table 1, Supplement S5). Therefore, we interpret this TG as indicative of weaker but sustained upwelling conditions. In addition, it shows some association with upwelling indicators such as dust flux and no association with the Mn / Al ratio in the sediment (Fig. 4b), indicating that it is only associated with an overall active upwelling zone and active Mn redirection and therefore OMZ conditions at Site 722.

Individual taphogroups represent specific eco-spaces, but to understand the relation and transitions between these eco-spaces in their temporal context, their variability must be considered in relation to other proxies within a multi-proxy approach. Integrating the analyses of nannofossil taphogroups (Table 1), planktonic foraminifera data (Fig. 5), and the abundance of diatom fluxes and geochemical data (Bialik et al., 2020a), we delineate temporal intervals in Site 722. These reflect stratigraphic intervals of specific environmental conditions in the WAS.

Interval 1 (base of study interval–13.4 Ma). This interval is characterized by variable taphogroups belonging to TG 1a, TG 2, TG 4a, and TG 4b. The variable taphogroups reflect a diverse and variable nannofloral assemblage in this interval. Overall, the nannofloral assemblages are characterized by a high abundance of Cyclicargolithus floridanus (Fig. 5). However, Cyclicargolithus floridanus abundances decline through the interval to its stratigraphic Top (T) occurrence at Site 722. In addition, we record abundant small reticulofenestrids and peaks of discoasterids (TG 4a, 4b). The average number of taxa in interval 1 is 14.9±2.1 (N= 13), with an average Shannon H′ diversity of 1.6±0.4. The planktonic foraminifera assemblage is dominated by thermocline-dwelling taxa (predominantly P. mayeri). Siliceous fragments are absent. We interpret this interval as a relatively low nutrient environment based on the above multi-group assemblage composition. In particular, the presence of TG 1a and 2 points to only moderately elevated nutrient concentrations in the surface waters at Site 722 during the MMCT. The common occurrence of Sphenolithus spp. and discoasterids suggests seasonally intermittent stratification. These results are consistent with the relatively warm SSTs recorded during this interval (Zhuang et al., 2017), further supporting a generally muted upwelling regime in the WAS during interval 1. These assumptions are corroborated by a more limited OMZ extent in the Indian Ocean compared to the later Miocene. At Site 722 this is shown declining Mn content. On the Maldives, high Mn concentrations and the absence of notable drift deposits, and thus lower wind intensity, also corroborate a generally weaker OMZ during this time (Bialik et al., 2020b; Betzler et al., 2016).

Interval 2a (13.4–12.0 Ma). Interval 2a is solely comprised by TG 4c. This taphogroup is characterized by a diverse assemblage with abundant R. pseudoumbilicus and common medium-sized reticulofenestrids and discoasterids. The average number of taxa is 16.6±2.2 (N= 7), with an average Shannon H′ index of 1.8±0.3. Siliceous fragments are absent. Planktonic foraminiferal assemblages are dominated by thermocline species with increased abundances of mixed-layer species compared to interval 1. Within interval 2a, a first slight increase in upwelling indicative taxa (primarily G. bulloides) is observed (Fig. 5). We interpret this interval as indicative of a first shallowing of the thermocline due to the initial strengthening of the wind-driven upwelling regime at Site 722. This intensification is likely related to an intensification of the monsoon system following the end of the MMCT (Betzler et al., 2018). The intensification of the monsoon system is also consistent with the establishment of an increased OMZ extent and drift deposits in the Maldives (Betzler et al., 2016).

Interval 2b (12.0–11.0 Ma). Interval 2b is comprised primarily of assemblages belonging to TG 4c, with one sample belonging to cluster 1b. The interval similar to interval 2a is characterized by assemblages (TG 4c) with abundant medium-sized reticulofenestrids that occur together with an increase in large reticulofenestrids (Figs. 2, 5). Furthermore, we detect a small but noteworthy increase in Umbilicosphaera jafari and a decline in Discoasteraceae. Moreover, the abundance of small reticulofenestrids is lower than in interval 2a. These differences within the assemblage are also the reason why interval 2 was separated into the two sub-intervals. The average number of taxa in interval 2b is 15.6±2.6 (N= 16), with an average Shannon H′ index of 1.5±0.3. The base of interval 2b also contains the first occurrence of diatoms within the section. Planktonic foraminifera mixed layer taxa decrease while upwelling taxa increase further in this interval.

We interpret this interval to mark a progressive intensification in the upwelling of high-nutrient subsurface waters. We base this on (1) the increase in siliceous fragments (diatoms and other siliceous biota, (2) higher abundances of upwelling indicative planktonic foraminiferal taxa, and (3) generally nutrient-adapted nannofossil taxa (i.e. medium- and large-sized reticulofenestrids; Beltran et al., 2014; Auer et al., 2015; Imai et al., 2015), which show progressive abundance increases. Intensified upwelling is consistent with increasing δ15N values and continuous cooling at Site 722 (Zhuang et al., 2017; Bialik et al., 2020a). Increased upwelling-derived nutrient access in the northern Indian Ocean is further supported by increased productivity and nitrogen utilization in the Maldives (Betzler et al., 2016; Ling et al., 2021). The upwelling intensification after 12 Ma is consistent with an overall increase in global atmospheric circulation and oceanic current strength, including the Indian Ocean South Equatorial Current (Fig. 6; House et al., 1991; Gourlan et al., 2008; Groeneveld et al., 2017; Betzler and Eberli, 2019).

Interval 3a (11.0–9.6 Ma). Interval 3a is characterized by a dominance of large reticulofenestrids (R. pseudoumbilicus) (TG 3) with intermittently common discoasterids and small reticulofenestrids (TG 4b). Notably, medium-sized reticulofenestrids show very low abundances compared to the previous intervals (Fig. 5). The abundance of Umbilicosphaera jafari is highly variable but overall common. At the same time, sphenoliths are rare in the lower part of the interval before increasing (up to ∼ 40 % of the assemblage) in the upper part. Within this interval, we also note the occurrence of variable abundances of small reticulofenestrids between ∼ 10.5 to 9.9 Ma. The average number of taxa is 14.3±5.1 (N= 22), with an average Shannon H′ index of 1.1±0.4. The high environmental variability within this interval is illustrated by alternations between assemblages belonging to TG 3, 4b, and 4c. Diatom fluxes increase significantly (Fig. 5). Diatoms generally dominate the phytoplankton assemblage, even outcompeting calcareous nannoplankton in terms of total abundance. High diatom abundances are especially prevalent within samples assigned to TG 3. Mixed-layer taxa dominate planktonic foraminifera assemblages and increase in this interval, together with upwelling taxa. Notably, thermocline species decline to less than half of their previous abundance. One sample (722B-34X-3W 30-32) is barren of planktonic foraminifers. The lack of foraminifera is likely due to the limited sample amounts washed for this study, in conjunction with the high accumulation rates of phytoplankton (diatoms and calcareous nannofossils) in this stratigraphic interval.

Based on the high abundance of diatoms and a generally high nutrient-adapted nannofossil assemblage, we interpret interval 3a as a peak in upwelling intensity at Site 722. This interpretation is consistent with previously published δ15N data from Site 722, sites U1466 and U1468, and other geochemical datasets in the Maldives (Bialik et al., 2020a; Ling et al., 2021). In addition, high productivity and OMZ expansion are further recorded by heightened TOC, uranium accumulation, and low Mn deposition within the northwestern Indian Ocean (Dickens and Owen, 1994, 1999; Betzler et al., 2016; Bialik et al., 2020a). This corresponds to an increase in Antarctic Bottom Water (AABW) formation due to the expansion of North Atlantic Deep Water (NADW), indicative of an intensified global thermohaline circulation (Woodruff and Savin, 1989). Increasing numbers of discoasterids in the upper part of interval 3a and decreasing diatoms numbers also point towards declining upwelling and thus seasonal nutrient depletion when no summer monsoon-derived upwelling occurs. This pattern of clear seasonality imparted on the plankton flux further amplifies within the next interval.

Interval 3b (9.6–8.8 Ma). Interval 3b continues to exhibit a dominance of large reticulofenestrids (R. pseudoumbilicus) (TG 3), although discoasterids noticeably decline and are replaced by higher abundances of sphenoliths (primarily Sphenolithus moriformis), with abundances of ∼ 40 % of the total assemblage. Small- and medium-sized reticulofenestrids are rare in this interval (Figs. 2, 5). The average number of taxa is 15±2.3 (N= 10), with an average Shannon H′ index of 1.4±0.3.

We thus interpret interval 3b to indicate decreasing upwelling intensity based on the increase in nannofossil taxa adapted to warmer and more stratified water masses, such as Discoaster spp. and Sphenolithus spp. (Lohmann and Carlson, 1981; Castradori, 1998; Negri and Villa, 2000; Blanc-Valleron et al., 2002; Gibbs et al., 2004a; Aubry, 2007; Villa et al., 2008; Schueth and Bralower, 2015). The waning upwelling of the northern Indian Ocean is corroborated by the proliferation of warm-water diatom taxa in the Maldives (Site 714; Boersma and Mikkelsen, 1990). Decreasing δ15N values support waning upwelling-derived productivity after 10 Ma at both Site 722 and in the Maldives and decreasing TOC fluxes at Site 722 (Gupta et al., 2015; Bialik et al., 2020a; Ling et al., 2021). It is, however, important to note that these changes are not reflected in the planktonic foraminifera community, which shows a continuously high presence of upwelling taxa (e.g. G. bulloides). One possibility would be that the upwelling cell became more seasonal, with nannoplankton-dominated photoautotrophic communities proliferating seasons with lower upwelling. However, primarily heterotrophic, non-symbiont-bearing taxa such as G. bulloides were still sustained by high primary productivity during monsoon season, as is the case in the present-day upwelling cell along the Oman margin (Schiebel et al., 2004; Rixen et al., 2019b).

We assume that this waning in upwelling is related to a decrease in the hemispheric temperature gradients, leading to a weaker summer monsoon wind system in the Indian Ocean. This reduction in temperature gradients is consistent with a decreasing trend in minimum deep-water temperatures, based on global benthic foraminifera compilations and deep-water records from the Ninety East Ridge (Site U1443; Fig. 1) (Lübbers et al., 2019; Westerhold et al., 2020). Furthermore, pollen data (Pound et al., 2012) suggest that progressive cooling of the Northern Hemisphere (NH) over the Middle to Late Miocene intensified. Northern Hemisphere cooling consequently reduced the asymmetry of hemispheric temperature gradients, thereby reducing summer monsoon wind intensity by muted northward migration of the intertropical convergence zone (ITCZ) in NH summer (Gadgil, 2018; Yao et al., 2023).

Interval 4 (8.8 Ma–top of study interval). Interval 4 – consisting of only three samples – is defined by the bloom of small reticulofenestrids (R. minuta) in the nannofossil assemblage. We also note an elevated abundance of Umbilicosphaera jafari and a marked decline in Sphenolithus spp. relative to interval 3b. This interval consists entirely of assemblages belonging to TG 1b (Figs. 2, 5). The average number of taxa is 17.3±0.5 (N= 3), with an average Shannon H′ index of 0.5±0.0. Despite the high number of nannofossil taxa in this interval, the low diversity directly results from the dominance of small reticulofenestrids. Siliceous fragments (primarily diatoms) persist but are much rarer than in interval 3. This reduction in diatom fluxes is part of an ongoing decrease in biogenic silica accumulation at Site 722, which culminates in a shift from phytoplankton- to zooplankton-dominated silica accumulation by ∼ 8 Ma (Nigrini, 1991; Prell et al., 1992). Planktonic foraminifera assemblages remain consistent with the upper part of interval 3, showing relatively high abundances of upwelling and mixed-layer taxa. We interpret this interval as a new nutrient regime, which likely led to a significant turnover in coccolithophore species around the same time (Young, 1990; Imai et al., 2015). However, the low sample number in this interval limits further interpretation.

Based on the intervals defined by the nannofossil taphogroups, a progression of plankton communities becomes apparent within the Middle to Late Miocene at Site 722. Their variation highlights the strong interactions between monsoon wind strength, nutrient availability, and primary productivity. Therefore, we link our new assemblage data with previous regional publications, highlighting a progressive upwelling increase leading to thermocline shoaling (Fig. 3; Zhuang et al., 2017; Bialik et al., 2020a). This thermocline shoaling in turn results in declining sea surface temperatures and increased surface water productivity through the upwelling of nutrient-rich thermocline waters along the Oman margin during this time.

Declining high Mn / Al ratios and diverse nannofossil assemblages point towards a relatively low nutrient regime between 15.0 and 13.5 Ma. Patterns of Mn decline have been observed since at least 15 Ma in the Maldives, which is in line with observations at Site 722 (Betzler et al., 2016, 2020a, b). This period thus represents a progressive increase in upwelling intensity during the MMCT due to globally declining SSTs and sea levels following the end of the MCO (Zhuang et al., 2017; Miller et al., 2020). Both nannoplankton and planktonic foraminifera reflect primarily open marine, low-nutrient conditions (Sexton and Norris, 2011; Lessa et al., 2020).

By 13.5 Ma, these progressive changes culminate in a first sustained community shift in nannofossil and planktonic foraminifera records (Figs. 2, 5). We consider these shifts to be a coupled response of Site 722 phytoplankton communities to increased surface water nutrient levels that subsequently allowed a population increase of heterotrophs such as foraminifera. These changes are consistent with establishing a more pronounced upwelling regime, which also resulted in the expansion of the OMZ further into the Indian Ocean, reaching the Maldives by ∼ 13 Ma. Furthermore, available TOC data still show low accumulation rates at Site 722 at this time, indicating that organic matter was still recycled mainly within the expanding OMZ (Bialik et al., 2020a).

By ∼ 12 Ma, another phytoplankton community shift (see interval 2b) leads to a size increase in the reticulofenestrids, lower nannoplankton diversity, and a higher abundance of thermocline-dwelling planktonic foraminifer taxa (Fig. 5). Together with increasing TOC fluxes (Fig. 3), all of these shifts point towards increasing productivity. These changes, however, happen without any significant changes in overall temperature within the upwelling zone (Zhuang et al., 2017). A northward shift in the Southern Hemisphere westerlies is recorded by 12 Ma (Groeneveld et al., 2017). We hypothesize that this shift and a potential increase in wind strength may have also increased nutrient concentrations in intermediate water masses within the sub-Antarctic frontal system. This interpretation would be in line with the effect increasing sea ice cover would have had on intermediate water nutrient concentrations based on modelling data and evidence from Southern Hemisphere records (Sarmiento et al., 2004; Sarmiento, 2006; Laufkötter and Gruber, 2018; Groeneveld et al., 2017). Such enhanced nutrient transport within the thermocline would reconcile increased productivity without increasing the total volume of upwelling – and consequently reducing SSTs – along the Oman margin. The first occurrence of diatoms within this interval may also point towards a shift in nutrient availability and increased phosphorus and silicon availability within the upwelling cell and likely globally (Keller and Barron, 1983). Decreasing P and Si limitation would thus provide more favourable conditions for highly efficient photosynthesizers, such as diatoms within the water column (Schiebel et al., 2004; Brembu et al., 2017; Sarmiento, 2006). Within the plankton community, we also note the first intermittent occurrences of elevated G. bulloides abundances, indicative of high-productivity upwelling conditions (Kroon et al., 1991; Gupta et al., 2015).

By 11 Ma, global climatic shifts and further decreasing sea levels (Miller et al., 2020; Westerhold et al., 2020) led to another step in the water masses upwelling in the WAS (Fig. 6). As a result of these water mass changes, diatoms dominate our phytoplankton record by 11 Ma, outpacing nannoplankton for the first time, while we note a first sustained occurrence (> 25 %) of G. bulloides. Therefore, we interpret this shift as the inception of sustained primary productivity within the upper water column of an upwelling cell supplied with enough Si, as well as P and N, to sustain a large diatom population (Brzezinski, 1985; Sarmiento, 2006; Closset et al., 2021).

However, the abundance of discoasterids and sphenoliths within our nannofossil record (Fig. 5) still needs to be reconciled with this interpretation. Both taxa are considered to be indicative of low nutrient conditions and increased stratification (Gibbs et al., 2004a; Schueth and Bralower, 2015; Karatsolis and Henderiks, 2023). This information is thus contrary to our recorded high abundances of mixed-layer-dwelling foraminifera and high-nutrient-adapted diatoms dominating the phytoplankton record. A possible way of integrating these opposite requirements is to evoke a highly seasonal upwelling cell with strong upwelling in one season and calm and stratified surface waters providing a deep thermocline and nutricline in the other.

This seasonal variability is most evident after 9.6 Ma when Sphenolithus abundances also increase together with overall nannofossil diversity (Fig. 5, interval 3b). These changes in the nannofossil community are also associated with decreasing diatom abundances and TOC fluxes, while upwelling indicative planktonic foraminifera taxa remain common. It thus seems that an initial spike in upwelling and therefore diatom accumulation waned again, pointing towards a significant reorganization of the upwelling cell after 9.6 Ma.

Within the topmost three samples of the record, belonging to interval 4, we note an increase in small reticulofenestrids corresponding to the base absence of Reticulofenestra pseudoumbilicus around 8.8 Ma, according to accepted nannofossil biostratigraphy (Young, 1990; Backman et al., 2012; Imai et al., 2015). We note that this significant size change and an increase in small placoliths are very pronounced within our WAS records from Site 722, in agreement with Young (1990). While we cannot contribute to the discussion of whether this assemblage shift constitutes an evolutionary-driven adaptation of taxa within the genus Reticulofenestra or purely an ecophenotypically driven size adaption (Young, 1990; Imai et al., 2015), we still note that a clear link to changing nutrient levels within the upwelling cell is becoming apparent. Imai et al. (2015) further hypothesized that the size shift is related to nutrient increases within the Indo-Pacific. Based on our records of high-nutrient conditions and likely at least intermittent seasonal eutrophication persisting from at least 11 Ma, we cannot completely follow their hypotheses that increasing nutrient levels within the surface ocean were the sole driver for this size shift. Therefore, we propose that changing nutrient limitation within the mixed layer may have played an important, as-of-yet unconsidered, role in defining the predominant assemblage structure within the WAS upwelling system during the Middle and Late Miocene (Fig. 7).

The modern productivity patterns and oxygen depletion in the northwestern Indian Ocean differ significantly from those observed in the studied period. For example, the increase in Mn content in the Maldives in the Pliocene (Betzler et al., 2016) suggests a significant reduction in Mn redirection, which continued until today. This is indeed visible in present-day oceanographic records, where elevated Mn concentrations are only found near the margins of the Arabian Sea (ThiDieuVu and Sohrin, 2013). Meanwhile, denitrification in the Eastern Arabian Sea appears to have only become significant during the Pliocene (Tripathi et al., 2017). These changes in productivity patterns thus may indicate that the WAS was potentially more productive during the Late Miocene than today and potentially even supported an expanded OMZ (Dickens and Owen, 1999, 1994).

Despite that, we note that even in the most productive parts of the Arabian Sea conditions are rarely eutrophic (Fig. 1a). As such, ascribing permanent eutrophic or even mesotrophic conditions to any of these assemblages is unlikely to be reasonable. On the other hand, nannofossil assemblages such as TG 3 with combined diatom occurrences possibly indicate the prevalence of mesotrophic and eutrophic conditions. Diatoms are generally less adapted to low nutrient levels, requiring higher P and N levels than coccolithophores to form blooms (Hutchins and Bruland, 1998; Litchman et al., 2006). If enough nutrients (including Si) are available, they tend to outcompete coccolithophores quickly and begin to dominate the mineralizing phytoplankton community (Schiebel et al., 2004; Brzezinski, 1985; Closset et al., 2021). Based on modern analogues, it seems likely that shifts in the nutrient content of upwelling waters may have played an important role in controlling the observed patterns in the plankton community along the WAS during the Middle to Late Miocene. In particular, after 13 Ma, a sustained and stable SAM regime seems to have existed during the Northern Hemisphere summer (Betzler et al., 2016). To disentangle these patterns, we focus on understanding observed patterns of the two dominant phytoplankton groups present within our record, with the context of their ecological preferences and primary nutrient requirements within our study interval.

The co-occurrence of diatoms, discoasterids, and sphenoliths in the upper part of the studied interval (Fig. 5) suggests that while nutrient levels were high, upwelling was likely highly seasonal. For the WAS, high seasonality may result from strengthening summer monsoon winds with no changes in winter monsoon winds (Schiebel et al., 2004; Rixen et al., 2019b; Sarr et al., 2022). Increasing summer but stable or absent winter monsoon conditions are likely the result of increased cooling in the Southern Hemisphere (Bialik et al., 2020a; Gadgil, 2018; Sarr et al., 2022). This asymmetric cooling strengthened the summer monsoon compared to the winter monsoon system, which only intensified ∼ 7 Ma (Gupta and Thomas, 2003; Holbourn et al., 2018; Rixen et al., 2019b).

The variability in wind and upwelling intensity and their interaction with nutrient availability thus likely also affected the community structure and size variability of primary producers on longer geological timescales. The community structure of primary producers then exerted control on first-level consumers, such as planktonic foraminifera.

Upwelling-derived TOC accumulation, primary productivity assemblages, and upwelling-indicative foraminifera show distinctive patterns, which are, however, not in complete agreement with wind proxies and the suggested expansion of the OMZ around 13 Ma (Betzler et al., 2016). These discrepancies resulted in a long-standing debate about the validity and usefulness of upwelling proxies as monsoonal indicators (Betzler et al., 2016; Clift and Webb, 2018; Bialik et al., 2020a; Yang et al., 2020; Sarr et al., 2022). We propose that this disagreement is primarily due to inadequate treatment of nutrient limitation and nutrient supply in conjunction with wind speed when evaluating primary productivity in the WAS (Figs. 5, 7).

Modern-day upwelling zones in the low to middle latitudes are generally well supplied in macro-nutrients, resulting in iron-limited environments or other micro- and nano-nutrient limitations (Moore et al., 2013). However, currently the fringing areas of upwelling zones are commonly N-limited through increased denitrification in underlying OMZs (Moore et al., 2013; Bristow et al., 2017; Anju et al., 2020; Buchanan et al., 2021; Ustick et al., 2021; Buttay et al., 2022). Within the WAS upwelling zone, major nutrients N, P, and to some degree minor nutrients such as Si are replenished through local recycling and intermixing with deep and intermediate water masses originating from Antarctica (Fig. 7; Sarmiento et al., 2004; Meisel et al., 2011; Sarmiento, 2006; Laufkötter and Gruber, 2018). Iron, a key micronutrient, is primarily supplied through dust and riverine influxes from surrounding continental sources (Kunkelova et al., 2022; Moore et al., 2013; Guieu et al., 2019).

Accepting that the wind regime had reached peak intensity by 13 Ma following a gradual increase from the end of the MCO (Betzler et al., 2016, 2018), the significant increase in diatom abundance and TOC accumulation after 12 Ma is not contemporary. Therefore, the availability of nutrients and the nutrient composition also played a key role in defining the variability between coccolithophore and diatom abundances within the WAS upwelling cell. Moreover, the shift in the reticulofenestrid morphotypes (Fig. 5) may also be linked to the state of nutrient limitation. Paasche (1998) also has shown that modern-day coccolithophores tend to increase the formation of small placoliths during N-limitation.

Therefore, the shift towards higher primary productivity after 12 Ma, including the first record of diatoms at Site 722, may indicate a change in nutrient composition along the WAS without necessitating a change in monsoon wind strength. Notably, during this time, the northward expansion of the Southern Hemisphere westerlies shifted the position of the polar and sub-Antarctic frontal system (Fig. 6). In particular, the Late Miocene sea ice expansion after 11 Ma strongly affected the Antarctic frontal system and, in turn, the nutrient enrichment of intermediate waters formed in this region (Groeneveld et al., 2017; Bijl et al., 2018; Laufkötter and Gruber, 2018). Here we propose that changes in the mode of intermediate-water formation significantly increased the nutrient availability in intermediate waters in the Antarctic frontal system, resulting in modern-like downwelling dynamics around Antarctica (Fig. 7). Furthermore, many modelling studies support the assumption that climatic changes affecting the Antarctic frontal system can strongly influence global productivity patterns (Sarmiento et al., 2004; Laufkötter and Gruber, 2018; Moore et al., 2018; Taucher et al., 2022). We therefore propose that the Middle to Late Miocene productivity changes in the WAS offer compelling evidence for this hypothesis.

Thus far, the discussion has focused on local aspects of the record in Site 722 in the WAS and northwestern Indian Ocean. However, the interconnected nature of the oceanic circulation and nutrient rejuvenation system means that critical mechanisms may be overlooked without a global perspective. For example, modelling evidence for nutrient transport and nutrient enrichment in low-latitude upwelling cells allows for the construction of a timeline of changes along the WAS and their interaction with plankton communities. Moreover, a complete oceanic perspective allows for contextualization into the broader evolution of the ocean-atmosphere system.

Initial plankton community structures agree with a generally low nutrient regime in a somewhat muted wind regime, based on a large amount of deep-thermocline-dwelling taxa in the foraminifera community, likely following phytoplankton primary productivity in the deeper photic zone (Lessa et al., 2020). In addition, the mixed layer is dominated by a diverse nannofossil assemblage (H′-diversity of around 1.5 within intervals 1 and 2). During the MMCT, wind shear strengthened by 13 Ma, resulting in a significant global shift in ocean–atmospheric circulation exemplified in the global reorganization of carbonate platform geometries and thermocline deepening and ventilation at Site 722, as shown by the increase in mixed-layer-dwelling planktonic foraminifera (Betzler et al., 2016, 2018; Betzler and Eberli, 2019; Lessa et al., 2020).

Modelling studies for the WAS link the initial intensification of upwelling and wind shear to a combination of increased latitudinal temperature gradients and the emergence of the Arabian Peninsula during the Middle Miocene (Zhang et al., 2014; Sarr et al., 2022; Yang et al., 2020). Notably, while OMZ expansion and Mn redirection are evident since at least ∼ 14 Ma at Site 722 (Bialik et al., 2020a), available productivity records support at most intermittently mesotrophic and likely P- and N-limited conditions before ∼ 12 Ma (Fig. 5). We thus propose that the upwelling cell in the WAS was wholly influenced by strong post-MMCT winds by 13 Ma. Productivity was still limited by the upwelling of comparably low-nutrient intermediate waters of local origin (Fig. 7). Likely originating in the marginal seas of the northwestern Indian Ocean, these water masses may have been remnants of the Tethyan Intermediate Water (TIW). While the Tethyan Seaway had terminated between 14 and 15 Ma (Bialik et al., 2019), TIW or a similar high-salinity mass (Woodruff and Savin, 1989; Smart et al., 2007) was still affecting the Northern Indian Ocean until at least 12 Ma. This remnant TIW can be considered a more potent form of the modern Red Sea and Persian Gulf Intermediate Waters (RSPGW; Fig. 7). These warm and salty intermediate waters may have played a much more substantial role in the WAS during the early stages of the uplift of the Arabian Peninsula (Woodruff and Savin, 1989; Tomczak and Godfrey, 2003; Chowdary et al., 2005; Smart et al., 2007; Acharya and Panigrahi, 2016). The influence of remnant TIW would also align with the high abundance of thermocline-dwelling taxa until 12 Ma, which we infer to be representative of a shallow and/or a poorly ventilated thermocline (Sexton and Norris, 2011; Lessa et al., 2020).

It thus seems likely that late Middle Miocene WAS upwelling may have been relatively nutrient poor. We speculate that these water masses may have suppressed primary productivity, muting the influence of the increasing Findlater Jets and the emerging Arabian Peninsula (e.g. Sarr et al., 2022) compared to today. Invoking significant TIW upwelling until at least 12 Ma would further reconcile the discrepancy between the occurrence of drift deposits in the Maldives, and thus strong monsoon winds and the first clear evidence for strong upwelling in the WAS, with the abundance increase in upwelling indicative planktonic foraminifera (e.g. G. bulloides; Fig. 5) and the first occurrence of diatoms at Site 722 (Fig. 5; Kroon et al., 1991; Huang et al., 2007b; Gupta et al., 2015; Bialik et al., 2020a). This change in nutrient availability is also reflected by a contemporary increase in medium-sized reticulofenestrids (R. antarctica and R. haqii), which are generally assumed to reflect higher nutrient availability due to upwelling (Fig. 5; Auer et al., 2019, and references therein).

Productivity in the WAS thereby only began to increase as remnant TIW got progressively supplanted by other more nutrient-rich water masses. At present, the waters upwelling in the Arabian Sea are primarily regarded to be ICW, which therefore also includes IIW, SAMW, and AAIW (You, 1997, 1998; Böning and Bard, 2009; Munz et al., 2017; Chinni and Singh, 2022). Today, AAIW and SAMW, forming in the northern branch of the Antarctic Divergence, control up to 75 % of low-latitude productivity (Sarmiento et al., 2004). We hypothesize that the increasing formation of AAIW and SAMW following the northward shift of the westerlies around 12 Ma (Fig. 6) may have modulated low-latitude productivity (Groeneveld et al., 2017; Laufkötter and Gruber, 2018; Moore et al., 2018; Taucher et al., 2022). This time also aligns well with the proposed inception of the northward shift of Southern Hemisphere climate belts and the invigoration of the South Equatorial Current (LeHouedec et al., 2012; Reuter et al., 2019). Following that, it can also be assumed that by 12 Ma the northward expansion of the Southern Hemisphere westerlies resulted in a near-modern Antarctic Divergence (Groeneveld et al., 2017; Laufkötter and Gruber, 2018).

This global change in circulation patterns was fully established by 11 Ma, with cool nutrient-rich SAMW and AAIW waters reaching Site 722, evidenced by a further SST drop (Zhuang et al., 2017). This resulted in the highest productivity in the WAS upwelling cell during the Miocene (Figs. 5–7). The Late Miocene high-productivity interval in the WAS is thus the result of intense summer-monsoon-dominated SAMW and AAIW upwelling, driven by the Findlater Jets and forced by steep latitudinal temperature gradients and favourable tectonic conditions on the Arabian Peninsula (Pound et al., 2012; Zhang et al., 2014; Sarr et al., 2022). Summer months were thus characterized by eutrophic P-, N-, and potentially Si-enriched waters, allowing the proliferation of diatoms and other siliceous organisms. Winter months, in contrast, favoured the accumulation of deep-dwelling discoasterids that utilized the nutrient-rich waters below a relatively deeper winter thermocline. Higher abundances of mixed-layer dwelling taxa also reflect the increased mixed-layer depth (Fig. 5). Expanding SAMW- and AAIW-fuelled high productivity that consequently also resulted in the highest recorded TOC fluxes between 11–10 Ma and a substantial OMZ expansion deep into the equatorial Indian Ocean (Dickens and Owen, 1994; Bialik et al., 2020a). Increasing OMZs also resulted in a global increase in denitrification, which is well-recorded in foraminifer-bound δ15N records, showing a trend from more oxygenated intermediate waters during the MCO to lower oxygenated waters in the Late Miocene in the Indo-Pacific (Auderset et al., 2022).

By 10 Ma, OMZs had reached a critical threshold, leading to another substantial change in nutrient conditions within the WAS upwelling. Through increased denitrification in the OMZ underlying the upwelling cell, nitrate and ammonia were lost through bacterial conversion to N2 (Sigman and Fripiat, 2019). Strong denitrification subsequently led to increasingly N-limited water masses upwelling within the WAS. Although concrete evidence is only presented for the WAS, these patterns could also have occurred globally, considering the clear evidence for decreasing ocean oxygenation during the Late Miocene (Auderset et al., 2022). The Late Miocene N-limitation in the WAS upwelling cell is chiefly expressed by a decline in diatom abundances after 10 Ma, in conjunction with overall community shifts in the nannofossil assemblage. Total upwelling intensity also remained consistently high, as indicated by the available SST record of Zhuang et al. (2017). Primary productivity thus remained relatively high, which is characterized by the continued presence and even dominance of large reticulofenestrids, diatoms, and the continuously high TOC concentration within the sediment (often > 1 wt %; Fig. 3). We thus assume that the drop in diatom abundance and intermittent decline in δ15N values at Site 722 (Figs. 3, 5) were not caused by decreasing upwelling intensity but rather a decline in P and Si availability and thus declining export of diatom-derived organic matter. The increase in sphenoliths within our Site 722 record (Fig. 5) could indicate increased environmental stress within the nannofossil assemblage (Wade and Bown, 2006). Sphenoliths are likely not a good indicator of long-term stratification changes (Karatsolis and Henderiks, 2023) in highly seasonal upwelling regimes like the WAS, as high TOC and thus sustained but lower diatom fluxes indicate continued upwelling after 10 Ma at Site 722. Sustained seasonal upwelling and high organic matter export (Fig. 3) are further inferred by decreasing organic carbon δ13C throughout this interval (Bialik et al., 2020a, and references therein).

By 8.8 Ma, the adaption of smaller reticulofenestrids may result in an evolutionary adaption to the continued N-limited nutrient availability in the WAS. We base this interpretation on the nutrient adaption of coccolithophorids based on modern culture experiments (Paasche, 1998). Although somewhat anecdotal, these offer the currently best explanation to reconcile the recorded history of Site 722 upwelling changes with the stark shifts in reticulofenestrids size ranges. It should be noted that these shifts have been recorded throughout the middle and low latitudes of the Indo-Pacific (Young, 1990; Imai et al., 2015). However, the full impact of this hypothesis needs to be tested further.

The data compilation of Young (1990) further shows that the recorded Late Miocene size shift was primarily limited to the low and middle latitudes, with larger reticulofenestrids persisting within in the higher latitudes. We propose that the transition in Reticulofenestra morphology from large to small morphotypes thus primarily represents a significant shift in nutrient limitation rather than total nutrient availability within the middle to low latitudes. We further argue that this turnover reflects N-limitation within the low and middle latitudes due to sustained and intense denitrification after 12 Ma (Auderset et al., 2022). Further studies, particularly on ultrastructural morphotaxonomy of reticulofenestrids, will be needed to fully disentangle the implications of the proposed N-limited nannofloral turnover.

The highly opportunistic small Reticulofenestra morphotype was subsequently also able to sustain phytoplankton blooms in the WAS, as evidenced by the significant increase in nannofossils within the sediment (Fig. 5). Furthermore, the high mass of small coccolith cells potentially also contributed to the re-establishment of strong denitrification, as evidenced by a rise in δ15N values after 8.8 Ma (Fig. 3), as their additional biomass contributed to OMZ re-expansion. Detailed records of Late Miocene OMZ strength throughout the Indian Ocean will be necessary to fully quantify the impact on local upwelling. Local tectonics also began to modify the region configuration at this time (Rodriguez et al., 2014), leading to bottom current intensification (Rodriguez et al., 2016), which may have also modulated subsequent OMZ dynamics (Dickens and Owen, 1999).