Secular and orbital-scale variability of equatorial Indian Ocean summer monsoon winds during the late Miocene

. In the modern northern Indian Ocean, biological productivity is intimately linked to near-surface oceano-graphic dynamics forced by the South Asian, or Indian, monsoon. In the late Pleistocene, this strong seasonal signal is transferred to the sedimentary record in the form of strong variance in the precession band (19–23 kyr), because precession dominates low-latitude insolation variations and drives seasonal contrast in oceanographic conditions. In addition, internal climate system feedbacks (e

drives seasonal contrast in oceanographic conditions. In addition, internal climate system feedbacks (e.g. ice-sheet albedo, carbon cycle, topography) play a key role in monsoon variability. Little is known about orbital-scale variability of the monsoon in the pre-Pleistocene, when atmospheric CO2 levels and global temperatures were higher. In addition, many questions remain open regarding the 25 timing of the initiation and intensification of the South Asian monsoon during the Miocene, an interval of significant global climate change that culminated in bipolar glaciation. Here, we present new high-resolution (< 1 kyr) records of export productivity and sediment accumulation from International Ocean Discovery Program Site U1443 in the southernmost Bay of Bengal spanning the late Miocene and earliest Pliocene (9 to 5 million years ago). Underpinned by a new orbitally-30 tuned benthic isotope stratigraphy, we use X-Ray Fluorescence-derived biogenic barium variations to discern productivity trends and rhythms. Our data show strong eccentricity-modulated precession-band productivity variations throughout the late Miocene, interpreted to reflect insolation forcing of summer monsoon wind strength in the equatorial Indian Ocean. On long timescales, our data support the interpretation that South Asian monsoon winds were already 35 established by 9 Ma, with no apparent intensification over the late Miocene.
In contrast, a time interval in which records from different regions and proxies converge somewhat is the late Miocene. Magnetic records from the Chinese Loess Plateau are interpreted to show a 75 long-term intensification of the East Asian summer monsoon (EASM) from ~8.2-2.6 Ma . A late Miocene strengthening of Asian winter monsoons is inferred from South China Sea (Holbourn et al., 2018;Jia et al., 2003) and Andaman Sea (Lee et al., 2020) records. In the Arabian Sea, an intensification of upwelling and productivity at ~8 Ma is interpreted to reflect a strengthening of the South Asian summer monsoon (SASM) (Kroon et al., 1991;Singh and Gupta, 80 2014; Gupta et al., 2004), although other studies find evidence contrary to this Huang et al., 2007). Proposed monsoon intensifications during the late Miocene roughly coincide with strong global cooling , and a number of studies have implicated cooling and the ramp-up of Antarctic glaciation in monsoon strengthening Holbourn et al., 2018;Gupta et al., 2004). Until now, a lack of continuous, well-preserved marine sequences from 85 the South Asian monsoon region has stalled our understanding of its complex evolution during the Miocene.
The SASM is thought to have varied strongly on orbital timescales because monsoon strength responds, both directly and via internal feedback mechanisms, to insolation forcing. Model 90 simulations predict a stronger South Asian monsoon during summer insolation maxima at both precession minima and (to a lesser degree) obliquity maxima (Bosmans et al., 2018;Kutzbach, 1992, 1987;Kutzbach, 1981;Jalihal et al., 2019;Tabor et al., 2018). Precession is the dominant control on insolation and its seasonal distribution near the equator, and proxy-based Pleistocene SASM records show strong precession-band (19-23 kyr) variability (Kathayat et al., 95 and Prell, 2003;Caley et al., 2011;Singh et al., 2011;Shimmield and Mowbray, 1991;Rogalla and 105 Andruleit, 2005;Clemens et al., 1991;Ziegler et al., 2010). These records suggest that summer monsoon proxies significantly lag northern hemisphere summer insolation maxima in the precession and obliquity bands due to climate feedbacks. Orbital control on past SASM strength in the pre-Pleistocene, when boundary conditions were different, has so far only been investigated in the Andaman Sea over the latest Miocene-early Pliocene, where seawater oxygen isotope data 110 suggest high-amplitude precession and obliquity forcing of monsoon rainfall with significant phase lags .
In this paper, we investigate sediment accumulation and export productivity dynamics at millennial resolution in late Miocene sediments from southern Bay of Bengal (BOB) International Ocean 115 Discovery Program (IODP) Site U1443 (5°N, 90°E, Fig. 1). The late Miocene (11.6-5.3 Ma) is an interval of major global climate change, with long-term cooling between ~7.5 and 5.5 Ma  culminating in major high-latitude cooling events (Holbourn et al., 2018), and important carbon cycle shifts recorded in the marine and terrestrial realms potentially linked to atmospheric CO2 decline (Tauxe and Feakins, 2020;Steinthorsdottir et al., 2020). The region of 120 Site U1443 is strongly influenced by seasonally reversing monsoon winds today, and primary productivity is tightly coupled to the annual monsoon cycle ( Fig. 1a-b, Fig. 2). During the SASM, strong moisture-laden winds blow inland, driving surface circulation changes and increased mixed layer depth ( Fig. 1a-b) (Webster, 1987a, b;Schott and Mccreary Jr, 2001;Gadgil, 2003;Schott et al., 2009). Strong Ekman pumping mixes nutrients into the surface layer during the South Asian 125 summer monsoon and, to a lesser extent, the winter monsoon, stimulating biological productivity ( Fig. 2) (Lévy et al., 2007;Mccreary et al., 2009;Koné et al., 2009;Behrenfeld et al., 2005;Longhurst, 1995). This strong seasonal signal is transferred to the sedimentary record in the form of strong variance at orbital periods because insolation variations drive seasonal contrast. In this paper, we generate a new orbitally tuned age model based on benthic foraminiferal stable isotopes 130 spanning ~9 to 5 Ma. Core-scanning X-Ray Fluorescence (XRF) data are then used to reconstruct bulk, carbonate, and biogenic barium content and mass accumulation rates (MARs), shedding light on secular and orbital-scale export productivity and sedimentation changes over the late Miocene.

135
In the modern southern BOB in waters overlying Site U1443, both primary and export productivity are strongly controlled by seasonally reversing winds associated with the South Asian monsoon.
https://doi.org/10.5194/cp-2021-77 Preprint. Discussion started: 29 June 2021 c Author(s) 2021. CC BY 4.0 License. Figure 2 shows the annual cycle of wind stress, mixed layer depth (MLD), net primary productivity (NPP), and biogenic particle export based on recent oceanographic, satellite, and sediment trap data (see methods for details). In boreal summer (JJA) strong southwest winds mix the upper water 140 column, deepening the MLD to ~60m and entraining nutrients into the photic zone, leading to enhanced primary productivity and biogenic particle export (with a lag of ~3-4 weeks) ( Fig. 1a-b,   Fig. 2). During boreal winter, northeast winds deepen the MLD to a lesser extent, resulting in a second smaller peak in productivity during the winter monsoon (DJF) (Fig. 2). During the intermonsoon seasons, lowest wind stress is recorded leading to a shallow MLD, higher Sea Surface 145 Temperatures (SSTs), and a more stratified upper water column, resulting in increased oligotrophy and reduced biological productivity. The biannual productivity maxima observed in the surface ocean above Site U1443 is characteristic of monsoon-dominated tropical regions (Longhurst, 1995;Lévy et al., 2007). Sediment trap data suggest that primary productivity is the dominant control on organic carbon export at a location west of the northern end of the Ninetyeast Ridge (SBBT site; 150 5°N, 87°E, Fig. 1c). However, lithogenic mineral ballasting at this location is not negligible (average ~13% and 15% lithogenic particles in shallow and deep traps respectively) (Rixen et al., 2019;Unger et al., 2003) and could in part explain the bias towards the late summer peak seen in biogenic fluxes compared to NPP, as maximum concentrations of lithogenic particles at SBBT occur during the summer monsoon. While wind forcing is identified as the dominant factor 155 controlling biogenic particle fluxes at the SBBT site, advection of nutrient-and chlorophyll-rich waters originating from the eastern Arabian Sea via the Southwest Monsoon Current (SMC) may further contribute to the summer productivity peak in this region (Unger et al., 2003). During the summer monsoon, the relatively salty and nutrient-rich SMC flows eastwards south of Sri Lanka then turns northwards into the BOB (Fig. 1c) (Schott et al., 2009;Jensen, 2003). The SMC and 160 associated eddies have been shown to increase chlorophyll concentrations and average phytoplankton size along their paths as far east as 88-90°E, with the current's influence generally restricted to north of 6°N at this longitude (Jyothibabu et al., 2015;Vinayachandran et al., 2004;Webber et al., 2018). While river runoff and resultant salinity stratification during the summer monsoon supress primary productivity further north in the BOB (Prasanna Kumar et al., 2002), 165 seasonal surface salinity variations are very small (<0.2 psu) at 5°N (Zweng et al., 2013).
Accordingly, monsoon impacts on nutrients and productivity in our study area are limited to those driven by surface currents and wind mixing, and biogenic export fluxes during the SASM are similar (Particulate organic carbon, POC) or higher (CaCO3 and biogenic SiO2) than at sites further north in the BOB (Unger et al., 2003). Thus, modern data give us confidence that export 170 productivity at our site is likely a reflection of South Asian (primarily summer) monsoon wind 148-149 cm), following the revised shipboard splice (CCSF-D), spanning the interval ~5 Ma to 9.5 185 Ma based on initial bio-magneto-stratigraphy. Samples come from lithologic Units Ib and IIa, and sediments mainly consist of light grey to pale yellow nannofossil ooze with clay and foraminifers, and occasional volcanic ash . To minimise aliasing due to low estimated sedimentation rates (~0.5-2 cm/kyr), 1 cm thick half rounds (~18 cc) were sampled for micropaleontology. Cores were sampled at a depth resolution of 4 cm in the upper part of the study 190 interval (70.06-114.18 m) and 2 cm in the lower part (114.18-122.76 m) where sedimentation rate estimates were lower. U-channels for XRF scanning were sampled from archive halves of sediment cores at Kochi Core Centre (Japan) during the post-cruise sampling party for 39 sections included in the splice between U1443C 9H 2A (69.95 m CCSF) and U1443C 13H 5A (113.58 m CCSF).

Late Miocene benthic foraminiferal stable isotope data 215
Bulk sediment samples were oven-dried at 50°C, weighed, and washed over a 63 µm sieve in tap water at CEREGE. The >63 µm fraction was oven-dried at 50°C on a filter paper and weighed to determine percentage coarse fraction. The <63 µm fraction was centrifuged and dried at 50°C. sampling included overlap between cores increasing resolution. Six to twelve specimens of the epibenthic foraminiferal species Cibicidoides wuellerstorfi were picked from the >212µm fraction, with 6-8 well-preserved specimens selected for analysis. Tests were broken into fragments, cleaned in ethanol in an ultrasonic bath, and oven dried at 40°C. Stable carbon and oxygen isotopes were measured on a Thermo Scientific MAT 253 dual-inlet isotope ratio mass spectrometer (DI-IRMS) 225 coupled to Kiel IV carbonate preparation device at the Leibniz Laboratory, University of Kiel.
Based on long-term analysis of international and internal carbonate standards, precision (1σ) is better than ±0.08‰ for δ 18 O and 0.05‰ for δ 13 C. Results were calibrated using the National

Age model
Seven calcareous nannofossil bio-events dated between 5.04 Ma and 9.53 Ma were identified at 235 high-resolution in Site U1443 splice samples (Table S1) to increase the resolution of shipboard biostratigraphy (Robinson et al., 2016). To check for orbital periodicities prior to tuning, wavelet https://doi.org/10.5194/cp-2021-77 Preprint. Discussion started: 29 June 2021 c Author(s) 2021. CC BY 4.0 License.
analyses were performed on benthic isotope records in the depth domain and on the revised nannofossil-based age model (using a 4 th order polynomial fit, Fig. S1) in R using the biwavelet package (Gouhier et al., 2016;Grinsted et al., 2004) (Fig. S2a-d). All time-series were first 240 interpolated to constant depth or age resolution, such that the maximum resolution present was preserved (2 cm and 2 kyr for benthic isotope records in the depth and age domain, respectively).
Records were then filtered to remove signals with periods longer than one third of the length of the dataset using the "bandpass" function in the R package Astrochron (Meyers, 2014). Wavelets indicate that obliquity-driven cycles are present throughout (~0.53 m, Fig S2 a Fig. S1) and shipboard magnetostratigraphy  as preliminary age-depth tie-points (Fig. 3), an astronomical age model was constructed by tuning our monospecific benthic δ 18 O record to an eccentricity plus tilt (ET) target curve (Laskar et al., 2004) (Fig. 4). We did not include precession in our tuning target so as not to introduce assumptions 250 related to which hemisphere was controlling climate at our site, and because the temporal resolution of our benthic record does not permit accurate resolution of precession cycles in some intervals. We used a minimal tuning approach, tying ET maxima to benthic δ 18 O minima, with at most one tiepoint per ~100 kyr and often one every 200-300 kyr (Fig. 4, Fig. S1), so as not to artificially introduce frequency modulations (Zeeden et al., 2015). 255

XRF Scanning and Calibration
U-channels were scanned at 1 cm intervals at The Australian National University (ANU) on a third generation Avaatech XRF core scanner. All cores were scanned sequentially and standards measured daily were consistent across all runs. Core sections were covered with 4 micron-thick 260 HClO4) on a hot plate. Blank contribution was estimated to be negligible. The accuracy of measurements was evaluated using analysis of geostandards MAG-1(marine mud) and BE-N (basalt). The typical analytical uncertainty was better than 5%. XRF-derived element counts were converted into element concentrations by direct linear calibration. This allowed us to reduce uncertainties related to the variable matrix effect and physical properties such as moisture content 280 that typically change downcore. Additionally, we used calibrated XRF data to calculate biogenic barium concentrations and to estimate % CaCO3 and "carbonate-free basis" (cfb) elemental concentrations, permitting evaluation of the extent to which dilution by the dominant sediment constituent (here CaCO3) is driving trends and variability of more minor constituents in our records.
To represent the relative contributions of CaCO3 versus terrigenous sediment components, we use 285 the log count ratio of Ca/(∑( Al, K, Ti, Fe, Rb)), termed log(Ca/Terr).

Ba-based export productivity proxies
The accumulation of biogenic barium in sediments is a reliable proxy for export production in certain environments (Paytan and Griffith, 2007). Micron-sized barite (BaSO4) crystals are the main 290 carriers of particulate barium in the ocean, with a maximum in concentration occurring just below the euphotic zone (Bishop, 1988;Dehairs et al., 1980). Although the exact mechanisms governing the precipitation of barite in the water column are only now coming to light (Martínez-Ruiz et al., 2019), its formation is thought to be associated with decaying organic matter. Depth profiles of dissolved Ba suggest that passive adsorption of barite onto mainly biogenic particles as they sink 295 through the water column, combined with vertical mixing of dissolved Ba from the deep ocean and riverine input, can best explain Ba's nutrient-like water column distribution (Dehairs et al., 1980;Cao et al., 2016). Goldberg and Arrhenius (1958) first hypothesised that an increase in Ba accumulation rate in sediments underlying the equatorial Pacific divergence zone was directly linked to overlying high productivity, followed by similar observations in equatorial Indian Ocean 300 sediments (Schmitz, 1987). Subsequently, evidence for strong correlations between fluxes of Ba and organic carbon in Atlantic and Pacific sediment traps led to algorithms relating new productivity to particulate Ba flux (Dymond et al., 1992;Francois et al., 1995). A further study https://doi.org/10.5194/cp-2021-77 Preprint. Discussion started: 29 June 2021 c Author(s) 2021. CC BY 4.0 License.
focusing on the accumulation of barite (Babarite) extracted from core-top and late Pleistocene sediments refined its use as a proxy for export productivity . Although 305 significant Ba regeneration occurs in the uppermost few millimetres of sediment , barite dissolution is thought to cease after burial due to supersaturation in interstitial waters (Gingele and Dahmke, 1994;Dymond et al., 1992) and barite is not subject to burial diagenesis in oxic sediments (Paytan et al., 1993). Ocean sedimentary Ba has both a biogenic (Babio) and a terrigenous (Badetrital) component, so estimates of past export productivity using barium 310 must distinguish between these sources. This can either be done by chemical leaching of bulk sediment (assuming that all barite is Babio) e.g. , or by determination of total barium (Batotal) and subtraction of Badetrital using Al content and the terrigenous Ba/Al ratio, resulting in a record of Baxs (Batotal -Badetrital), see equation 1 (Dymond et al., 1992). (1) Direct comparisons of measurements of Babarite and Baxs suggest that non-barite phases of barium may be included in the calculation of Baxs; nevertheless Baxs is most representative of Babarite and therefore export productivity in oxic carbonate-rich sediments with low terrigenous, biogenic silica, 320 and organic carbon contents (Eagle et al., 2003;Averyt and Paytan, 2004;Gonneea and Paytan, 2006). The use of bulk Ba/Ti, Ba/Al, and Ba/Fe ratios is another approach to evaluate relative changes in export productivity (i.e., normalisation to an element presumed to be predominantly of terrigenous origin), but without precisely predefining the Ba/terrigenous ratio that could vary over time, and also removing the effect of dilution by a dominant sedimentary component such as 325
Here, we reconstruct changes in export productivity at Site U1443 over the late Miocene using XRF-derived Ba data and compare elemental count ratios of log(Ba/Fe), log(Ba/Ti), and log(Ba/Al), with [Ba]cfb and [Ba]xs calculated following equation (1), using a Ba/Alterrigenous value of 330 0.0075 g/g following Dymond et al. (1992). To verify consistency of trends, we also calculate [Ba]xs using [Ti] to represent Badetrital, applying a Ba/Titerrigenous ratio of 0.183 g/g (Mclennan, 2001)

Mass Accumulation rates 335
MARs of bulk sediment, CaCO3, [Ba]xs, and summed terrigenous elements (Al, K, Ti, Fe, and Rb) were calculated by multiplying concentrations by linear sedimentation rates (in cm/kyr) derived from our new age model and dry bulk densities (in g/cm 3 ). Dry bulk density values were estimated from high-resolution shipboard Gamma Ray Attenuation bulk density scanning data, and the linear relationship between all shipboard U1443 wet bulk density and dry bulk density measurements 340 (n=164, r 2 >0.99) . Units are g/cm 2 /kyr for bulk and CaCO3 MAR, µg/cm 2 /kyr for [Ba]xs MAR and mg/cm 2 /kyr for terrigenous MAR.

Time series analysis
Spectral analyses of benthic isotope and XRF data against tuned age were performed on filtered 345 records with a constant time step as described in Section 2.3 (0.5 kyr for XRF records and 2 kyr for isotope records). Multi-taper method (MTM) spectral analyses using a robust red-noise model were performed using Acycle (Li et al., 2019). Blackman-Tukey cross-spectral analyses were performed in Arand to assess phase and coherence (Howell et al., 2006). Wavelet analyses were performed in R using the biwavelet package (Gouhier et al., 2016;Grinsted et al., 2004). To illustrate precession-350 band variance and amplitude modulation, certain records (with identified significant precession variance) were filtered using a Tanner-Hilbert filter centred on 43 cycles/myr with bandwidth ±3 (designed to include both the ~24 kyr and ~22 kyr precession periods) in Acycle (Li et al., 2019).

XRF data and calibration
Scanning XRF results are shown in Figure Ma) were not converted to concentrations (Fig. 6). In brief, Al, Si, Ti, Fe, Rb, and K show similar trends, with a long-term small increase in concentrations 410 from 8.15 Ma to 5 Ma and spikes (particularly pronounced in Rb and K) corresponding in some cases to described ash layers . Ca and Ba show a minor long-term decrease over the study interval, while Sr and Mn increase from ~8.15 Ma to 6 Ma, then stabilise or decrease slightly. All elements show high-frequency variability throughout. For Ca and Sr, R 2 values were lower (0.39 and 0.42 respectively, Fig S6) due to the consistently high Ca and Sr contents and small 415 variability in the selected calibration samples. To estimate percent CaCO3, we therefore used a Ca/Fe ratio calibration rather than a direct linear calibration. We first used the linear relationship between Ca/Fe counts and Ca/Fe as determined by ICP-MS (Fig. S6, R 2 =0.93). Then %CaCO3 was calculated assuming that all Ca was contained in CaCO3 -a reasonable assumption at Site U1443 given the relatively low clay content in lithological subunits Ib and IIa . 420 Confidence in our method is provided by very good agreement with independent %CaCO3 estimates for the middle and early late Miocene interval of Site U1443 based on calibration of XRFderived counts of (Ca/∑(Ca, Al, Si, K, Ti, Mn, Fe, S)) to discrete CaCO3 measurements (Lübbers et al., 2019), including an overlapping interval based on an alternate splice from 112.8 and 113.6 m CCSF (Fig. 7b). 425

CaCO3 content, sediment accumulation patterns and Ba proxies
Late Miocene estimated % CaCO3 varies between ~60 and 90% with a small long-term decrease over the 9-5 Ma interval (Fig. 7b). This long-term trend is also visible in the log(Ca/Terr) record ( Fig. 7c), implying a small increase in the contribution of terrigenous material relative to CaCO3 in 430 Site U1443 sediments over time. Three % CaCO3 and log(Ca/Terr) minima between 5 and 6 Ma https://doi.org/10.5194/cp-2021-77 Preprint. Discussion started: 29 June 2021 c Author(s) 2021. CC BY 4.0 License. occur in identified ash layers. Log(Ba/Al), log(Ba/Fe) and log(Ba/Ti) show identical long-term and orbital-scale trends (Fig. S7), therefore we only discuss log(Ba/Fe) in the main text. Log(Ba/Fe) shows a long-term decrease between 9 and 5.3 Ma, and a smaller increase from 5.3 to 5 Ma (Fig.   7d).
[Ba]xs shows identical variability whether calculated using Al or Ti (Fig. 7e) Bulk sediment MARs vary between 0.5 and 2.1 g/cm 2 /kyr with a step increase from ~0.5 to 1.5 g/cm 2 /kyr occurring at 8.66 Ma (Fig. 7g), concurrent with a major sedimentation rate increase (Fig.  445 4d). CaCO3 MARs range from 0.4 to 2 g/cm 2 /kyr and co-vary with bulk sediment MARs, with the increasing difference between the two records reflecting a small long-term increase in non-CaCO3 components (Fig. 7g). This small increase is reflected in terrigenous element MARs, which vary between ~20-80 mg/cm 2 /kyr (excluding volcanic ash layers) (Fig. 7i). We note that absolute values of terrigenous MAR should be interpreted with caution, because this calculation does not include Si 450 as this element was not quantified in discrete samples. Nevertheless, a significant correlation between Al and Si counts (R 2 = 0.8, Fig. 6) suggests that Si is primarily of terrigenous origin, therefore trends in log(Ca/Terr) and terrigenous MAR in Fig. 7  Ma, 6.1-6.3 Ma, 7.5-7.7 Ma, and 7.8-8 Ma (Fig. 7f).

Late Miocene sedimentation patterns in the southern Bay of Bengal
We first examine the drivers of changes in sediment MAR identified in our record, and their 475 possible link to regional and global productivity trends. The three-fold increase in CaCO3 MAR at 8.66 Ma at Site U1443, originally described in Lübbers et al. (2019), could result from improved preservation and/or increased carbonate export by pelagic calcifiers (coccolithophores and/or foraminifera). Based on CaCO3 percentages, MARs, and benthic to planktic foraminiferal ratios, Lübbers et al. (2019) identified the mid to late Miocene "carbonate crash" in Site U1443 sediments 480 between ~12.2 Ma and 10 Ma, with a slow recovery from ~10 to 8.7 Ma, favouring an interpretation that the increase in CaCO3 MAR at 8.66 Ma reflects improved preservation. A record of planktic foraminiferal fragmentation between 9 and 8 Ma generated in the present study, interpreted to reflect a decrease in carbonate dissolution (Le and Shackleton, 1992), supports this interpretation (Fig. 7h). We see no change in log(Ba/Fe) concurrent with the CaCO3 MAR increase 485 at 8.66 Ma, which suggests that total export productivity at Site U1443 remained stable over this transition. However, our data suggest that an increase in coccolithophore production may have occurred. The contribution of foraminifera to total CaCO3 over our study interval is low (see >63 µm MAR in Fig. 7g), leading us to infer that higher CaCO3 MARs between 8.66 and 5 Ma are primarily driven by coccoliths. A 3-fold increase in sediment accumulation rate (SAR) at ~8. 6 Ma 490 with no change in CaCO3 content (%), implying a large increase in CaCO3 MARs, is also seen at shallower (2247 m) Deep Sea Drilling Project (DSDP) Site 216 on the NER near the equator (Fig.   1c) (Bukry, 1974;Mcneill et al., 2017;Pimm, 1974). This suggests that production played a role in driving regional carbonate MAR increases as well as improved preservation at deeper sites. coccolithophore productivity and calcification driven by weathering alkalinity inputs and regional nutrient changes (Si and Rosenthal, 2019).

500
Interestingly, the increase in bulk MARs at 8.66 Ma is driven by both CaCO3 and to a lesser extent non-CaCO3 components (clays), implicating another mechanism as well as improved carbonate preservation and increased coccolith export productivity affecting sedimentation at Site U1443. We suggest that an increase in coccolith CaCO3 flux to the seafloor could have led to increased scavenging by sinking biogenic aggregates of fine clays. Fine clays are present in the southern BOB 505 water column as a direct result of riverine flux (Rixen et al., 2019;Ramaswamy, 1993), and in nepheloid layers above the NER where high clay concentrations occur due to proximity to the sedimentary fan systems to the east (Nicobar Fan) and west (Bengal Fan) (Stow et al., 1990).
Recent studies of sedimentation patterns on the Bengal and Nicobar Fans, separated by the NER, interpret a large increase in SAR both on the NER and the Nicobar Fan at ~10-8 Ma to reflect 510 increased lithogenic sediment flux to the eastern Indian Ocean (Mcneill et al., 2017;Pickering et al., 2020b). Our data from the NER show that >75% of the 3-fold increase in SAR at 8.66 Ma is driven by biogenic CaCO3, thus we caution against using SAR at Site U1443/758 as representative of changes in sediment flux to the Bengal-Nicobar Fan system. Data from Site U1443, as well as from nearby DSDP Site 216 (Bukry, 1974;Mcneill et al., 2017;Pimm, 1974) (Fig. 1c), suggest that 515 increases in biogenic carbonate accumulation on the NER are decoupled, both temporally and mechanistically, from the increase in sediment delivery to the Nicobar Fan system. The gradual increase in terrigenous element and non-CaCO3 MARs over the 9-5 Ma interval seen at Site U1443 ( Fig. 7g,i) is part of a longer-term trend of increasing mineral flux in this region of the NER from the Miocene to the Pleistocene, beginning at ~12 Ma at ODP Site 758, that is thought to reflect 520 increased Himalayan erosion (Ali et al., 2021;Hovan and Rea, 1992).
Increases in the MAR of biogenic components (CaCO3, opal, organic carbon, phosphorus) between ~9 and 4 Ma have been measured in sediments from the Pacific, Indian and Atlantic Oceans (Farrell et al., 1995;Lyle and Baldauf, 2015;Van Andel et al., 1975;Grant and Dickens, 2002;Delaney and 525 Filippelli, 1994;Hermoyian and Owen, 2001;Dickens and Owen, 1999;Drury et al., 2020). This period of increased biogenic sedimentation, supported by independent paleoproductivity proxies (e.g., Diester Haass et al., 2005), is thought to reflect higher biological productivity and was dubbed the "biogenic bloom" by Farrell et al. (1995). A low-resolution CaCO3 MAR record from Site 758 shows higher values between 8 and 4 Ma, which in the absence of evidence for an increase in Pliocene (Dickens and Owen, 1999;Pierce et al., 1989;Si and Rosenthal, 2019), although improved age control for the Pliocene interval as well as independent paleoproductivity reconstructions are needed to verify this. Hypotheses to explain the biogenic bloom invoke a change in global nutrient cycling; i.e., a global increase in nutrient input, and/or redistribution of nutrients between basins 535 (Grant and Dickens, 2002), although the asynchronous timing of the biogenic bloom between regions, its variable expression, and its differentiation from the carbonate crash recovery complicate its interpretation. Diester Haass et al. (2006) hypothesised that changes in reconstructed productivity were correlated to the LMCIS at four Indo-Pacific sites, and tentatively proposed a link to a strengthened wind regime at this time. At Site U1443, we find no clear link between export 540 productivity or carbonate sedimentation and the LMCIS (Fig. 7). In the northern Indian Ocean, the influence of possible concurrent changes in monsoon strength on paleoproductivity and biogenic MARs must also be considered, and these are discussed in Section 4.3.

Significant variability at obliquity and precession periods has been identified in high-resolution late
Miocene-early Pliocene records of precipitation/runoff based on planktic foraminiferal δ 18 O and https://doi.org/10.5194/cp-2021-77 Preprint. Discussion started: 29 June 2021 c Author(s) 2021. CC BY 4.0 License. seawater δ 18 O in the nearby Andaman Sea . These authors suggest that, prior to 565 a distinct switch to obliquity-driven variability around 5.55 Ma, their records reflect strong precessional (insolation) control on South Asian monsoon rainfall from 6.2-5.55 Ma, with significant phase lags between proxies and precession. Wavelet analyses of log(Ba/Fe) and [Ba]xs show that precession-band (22-24 kyr) variability dominated throughout our 9-5 Ma study interval at Site U1443 (Fig. 9). Although phase relationships with insolation should be interpreted with 570 caution because of errors inherent to our late Miocene age model, export productivity appears to be coherent and in phase (within error) with the summer inter-tropical insolation gradient (SITIG, the insolation gradient between 23°N and 23°S on June 21 st ) (Fig. S3d). The SITIG has been proposed as a primary control on the strength of SASM winds, because a stronger SITIG increases the pressure gradient between the two limbs of the winter hemisphere Hadley cell, which drives 575 monsoon winds into the summer hemisphere (Bosmans et al., 2015). Our new Ba-based export productivity records corroborate the hypothesis that insolation played a dominant role in driving late Miocene South Asian summer monsoon wind variability in the equatorial sector of the SAM region, as predicted by general circulation models (Bosmans et al., 2018), with internal climate processes such as ice volume playing a more minor role than in the Late Pleistocene when large 580 glacial-interglacial cycles and related feedbacks drove variability in the Asian monsoon on 100-kyr timescales (Clemens et al., 2018;Clemens et al., 2021) and SASM wind proxies from the Arabian Sea and southern BOB record up to ~9 kyr phase lags relative to precession (Bolton et al., 2013;Caley et al., 2011;Clemens and Prell, 2003;Clemens et al., 1991).

585
In addition to the periods discussed above, a number of non-primary orbital periods termed heterodynes, which result from non-linear interactions between variables operating at Earth's primary orbital periods (Rial and Anaclerio, 2000;Thomas et al., 2016;Clemens et al., 2010), stand out in our late Miocene records (24,26,30,37,49, and 68 kyr periods; Fig. 8). For example, the 1/24 kyr heterodyne, prominent in all our records, could result from the interference of eccentricity 590 with precession, and the 1/30 kyr heterodyne seen in Ba records from an interaction between obliquity and precession. Several of these heterodynes have been previously identified in spectra of seawater δ 18 O that reflect Asian monsoon precipitation and runoff, both in the Andaman Sea (30 and 130 kyr during the Pleistocene, 27 and 30 kyr in the latest Miocene) (Gebregiorgis et al., 2018;Jöhnck et al., 2020) and in the East China Sea (29 and 69 kyr during the Pleistocene) (Clemens et 595 al., 2018), interpreted to suggest a strongly non-linear response of the monsoon to orbital forcing.
We favour the interpretation that the strong 24-kyr variability in our records reflects a primary  (Fig. 10). 600 Amplitude modulation of precession-scale variability in our productivity records broadly follows that of the SITIG (Fig. 10), suggesting a direct response of SASM winds to cross-equatorial insolation gradients during the late Miocene.
Cross-spectral analysis of our [Ba]xs productivity record with the Site U1448 seawater δ 18 O record 605 over the interval 4.95-6.19 Ma (where records overlap) shows >80% coherency and an in-phase relationship at the 30-kyr period, suggesting that during this time monsoon winds and precipitation/runoff in the BOB were to some degree coupled on orbital timescales (Fig. S3c), as is the case in the late Pleistocene (Clemens et al., 2021). Nevertheless, our [Ba]xs record over this time window contains stronger primary precession (22-24 kyr) and obliquity (41 kyr) signals than the 610 Site U1448 seawater δ 18 O record that cannot be explained by differences in resolution, highlighting that different climatic processes and feedbacks operating on orbital timescales must contribute to the two records (interpreted to reflect runoff/precipitation and wind, respectively) to different extents. In the late Pleistocene, strong obliquity-band and precession-band variance is found in Andaman Sea records of monsoon precipitation/runoff (Gebregiorgis et al., 2018), whereas records 615 of upper ocean stratification controlled by South Asian monsoon wind mixing at Site 758 (~100 m from Site U1443) show only precession-band variance (Bolton et al., 2013). The significant 41-kyr variability seen in late Miocene terrigenous elements at Site U1443 (Fig. 8a-c) could also suggest obliquity control on monsoon runoff into the BOB at this time, inferring a partial decoupling between monsoon winds (controlling open-ocean productivity) and runoff (controlling terrigenous 620 sedimentation and salinity/seawater δ 18 O) on orbital timescales, although additional records from regions closer to river sediment and freshwater sources are needed to corroborate this idea. It is important to note that whilst SAM expression above Site U1443 in the southern BOB is dominated by summer monsoon winds driving surface ocean currents and deeper mixing, oceanographic conditions in the northern and eastern BOB (e.g., Sites U1447 and U1448) are instead primarily 625 controlled by summer monsoon freshwater inputs . Runoff and direct precipitation during the SASM lead to strong salinity stratification in the northern parts of the BOB in the late summer and autumn that prevents upper ocean mixing (e.g., Sengupta et al., 2016). These regional differences in the manifestation of the monsoon must be considered when interpreting records from the heterogeneous BOB, and records from multiple locations and proxies 630 are needed to achieve a representative picture of the SAM subsystem. The 405 kyr eccentricity modulation of precession-scale export productivity variability broadly coincides with 405-kyr cycles in benthic δ 13 C at Site U1443, with higher export productivity occurring during benthic δ 13 C minima and eccentricity maxima on these timescales (Fig. 10). 635 Cross-spectral analysis indicates that log(Ba/Fe) and benthic δ 13 C are > 95% coherent at the ~405 kyr and 22-24 kyr periods, with an in-phase relationship in the precession band (-13°±25) and a near antiphase relationship on 405-kyr timescales (151°±27) (Fig. S3b). 405-kyr modulation of the ocean carbon cycle, primarily recorded in carbonate content and benthic δ 13 C records (Herbert, 1997;Drury et al., 2020;De Vleeschouwer et al., 2020;Westerhold et al., 2020;Pälike et al., 2012;640 Paillard, 2017;Holbourn et al., 2007) but also in productivity and monsoon-related dust records (Rickaby et al., 2007;Wang et al., 2010), has been observed throughout the Cenozoic and Mesozoic sedimentary record. In middle Miocene records, poor carbonate preservation noted during eccentricity maxima is interpreted as indicating transient shoaling of the carbonate compensation depth (Holbourn et al., 2007;Flower and Kennett, 1994). Here, we see a broad positive correlation 645 between log(Ba/Fe) and log(Ca/Terr) records on 405-kyr timescales (Figs. 9a,b), suggesting that carbonate content fluctuations at Site U1443 in the late Miocene were more strongly related to biogenic production than to dissolution on long eccentricity timescales. The coincidence of late Miocene eccentricity maxima with productivity maxima and benthic δ 13 C minima at Site U1443 is compatible with the hypothesis that during eccentricity maxima, a strengthened monsoon induced 650 405-kyr cycles in the marine carbon cycle via increased weathering and nutrient inputs, leading to enhanced marine biological productivity and deep-ocean organic carbon burial (Ma et al., 2011).

Late Miocene monsoon evolution
On long timescales, our Site U1443 biogenic Ba records show relatively stable (9 to 6.5 Ma) or 655 slightly decreasing (6.5-5.3 Ma) export productivity between 9 and 5 Ma ( Fig. 7d-f, Fig. S7). Based on sediment colour properties and XRF-derived Ba/Ti ratios in the preceding interval (~13.5-8.3 Ma), Lübbers et al. (2019) suggested that a shift towards a higher productivity regime at Site U1443 occurred at ~11.2 Ma, significantly earlier than the onset of the biogenic bloom at other sites and 2.5 Ma before the rise in CaCO3 MAR, and that this shift was potentially linked to an intensification 660 of the South Asian monsoon. An increase in export productivity at ~11 Ma is coherent with longterm changes in benthic foraminiferal assemblages at Site 758 (Gupta et al., 2004;Nomura, 1995) and at sites in the western tropical Indian Ocean (Smart et al., 2007). Reconstructions of Arabian Sea upwelling, export productivity, and deoxygenation (Bialik et al., 2020;Gupta et al., 2015;Huang et al., 2007;Zhuang et al., 2017), as well as the abrupt appearance of drift sediments in the 665 https://doi.org/10.5194/cp-2021-77 Preprint. Discussion started: 29 June 2021 c Author(s) 2021. CC BY 4.0 License.
Maldives Archipelago at ~13 Ma , point towards an intensification of seasonally-reversing South Asian monsoon winds between 13-11 Ma, consistent with Site U1443 export productivity records. Our new data suggest that similar levels of export productivity to those seen from 11.2 Ma to 9 Ma persisted until at least 5 Ma at Site U1443.

670
Compiled Asian monsoon proxy records spanning the 9 to 5 Ma interval show relatively stable long-term SASM strength (Fig. 11, see Fig. 1c for site map). In line with records from the Maldives and Arabian Sea Huang et al., 2007;Betzler et al., 2016;Zhuang et al., 2017), Site U1443 records do not corroborate the hypothesis that SASM winds intensified at ~7-8 Ma as suggested by some studies (Kroon et al., 1991;Singh and Gupta, 2014;Gupta et al., 2015;An et al., 675 2001). A recent modelling study suggests that the emergence of the Arabian Peninsula played a key role in the establishment of modern Somali Jet structure above the western Indian Ocean, initiating strong upwelling along the Oman margin during the Miocene (Sarr et al., in review). This provides a potential explanation for the temporal decoupling between western Arabian Sea records and SASM records from other regions, and suggests that Arabian Sea upwelling records should not be 680 interpreted as exclusive recorders of SASM strength on geological timescales. The long-term trend in our record shows broad agreement with a low-resolution clay mineralogy record from Site U1447 in the Andaman Sea showing gradual long-term decrease in smectite/(illite and chlorite) over the late Miocene (Fig. 11g), indicating strengthened physical weathering and/or weakened chemical weathering, attributed to the South Asian winter and summer monsoons respectively (Lee 685 et al., 2020). Also at Site U1447, records of potassium content (%K, Fig. 11x) are interpreted to show a shift in sediment provenance and/or an increase in physical weathering and erosion in the sediment source region between ~7 and ~6 Ma (Fig. 11h), potentially linked to an increase in monsoon rainfall intensity and global cooling (Kuhnt et al., 2020). Between 6.2 and 5 Ma, our equatorial Indian Ocean wind records show good long-term agreement with a seawater δ 18 O record 690 from the Andaman Sea , with a minimum in export productivity at ~5.3 Ma at Site U1443 coinciding with a maximum in seawater δ 18 O at Site U1448 (Fig. 11i, j). One interpretation of this could be a coupled reduction in both SASM wind intensity and runoff/precipitation over this interval, although Jöhnck et al. (2020)  In contrast to the relatively stable South Asian monsoon between 9 and 5 Ma, evidence for a step 700 strengthening of the East Asian winter monsoon during the late Miocene (~7 Ma) comes from the South China Sea (Holbourn et al., 2018) (Fig. 11a), whereas records from a site on the Chinese Loess Plateau suggest a more gradual intensification of the East Asian summer monsoon from 8.2 to 2.6 Ma (Fig. 11b). The step change in South China Sea surface water geochemistry is interpreted to reflect drying and cooling of the Asian interior and a related southward shift of the ITCZ leading 705 to an intensified dry winter monsoon over southeast Asia (Holbourn et al., 2018). The long-term increase in East Asian summer monsoon strength inferred from magnetic records in Chinese loess sequences is attributed to progressive Antarctic glaciation that drives an increased pressure gradient between the Australian High and Asian Low pressure cells, a mechanism supported by numerical simulations . However, we note that these single site records may not be 710 representative of the East Asian monsoon subsystem as a whole. Conversely, the apparent insensitivity of equatorial SASM wind intensity to global late Miocene sea surface cooling, which began at ~7.5 Ma and culminated in an SST minimum at 6 to 5.5 Ma (Fig. 11k)  , is consistent with climate modelling studies that show limited impact of different pCO2 scenarios on SASM wind patterns and strength (Kitoh et al., 1997;Sarr et al., in review). The South 715 Asian monsoon is widely considered to be a thermally direct circulation, driven by the thermodynamic contrast between the Indian subcontinent and the equatorial Indian Ocean that develops in summer, with changes in this gradient impacting the strength of onshore SASM monsoon flow (Lutsko et al., 2019;Acosta and Huber, 2020). The lack of a long-term trend in wind and surface circulation proxies over the 9 to 5 Ma interval (Fig. 11) suggests a relatively constant 720 land-sea temperature gradient despite global cooling. Thus, our data add to the body of evidence suggesting decoupling between the East Asian and South Asian monsoons on long timescales.

Summary & Conclusions
We present new equatorial Indian Ocean benthic δ 13 C and δ 18 O records and an age model spanning 725 the interval between 9 and 5 Ma (late Miocene-earliest Pliocene), and analyse sedimentation and productivity trends and cyclicity using XRF-derived records and MARs. Biogenic sediment MARs reveal a modest imprint of the late Miocene biogenic bloom at Site U1443 lasting until at least 5 Ma, primarily driven by fine-fraction (coccolith) CaCO3 accumulation, as noted at other sites (Si and Rosenthal, 2019). Nevertheless, the carbonate MAR record of Site U1443 is clearly influenced 730 by carbonate preservation as well as production over the late Miocene, so independent productivity proxies must be considered when defining the duration of the biogenic bloom. Our results indicate https://doi.org/10.5194/cp-2021-77 Preprint. Discussion started: 29 June 2021 c Author(s) 2021. CC BY 4.0 License. that step increases in SAR recorded on the Ninetyeast Ridge at ~8-9 Ma primarily reflect an increase in CaCO3 accumulation, and that this is likely independent from the increase in lithogenic sediment flux recorded in nearby Nicobar Fan sites, itself related to sediment re-routing within the 735 Nicobar-Bengal Fan system around the same time (Mcneill et al., 2017;Pickering et al., 2020a).
Our data show no long-term increase in export productivity between 9 and 5 Ma (and by inference no intensification of SASM winds), therefore these data support existing evidence for an early late Miocene (~13-10 Ma) establishment of strong seasonally reversing South Asian monsoon winds and Arabian Sea upwelling, with relatively stable or slightly weakening SASM winds over the 740 remainder of the late Miocene and earliest Pliocene between 9 and 5 Ma. Spectral and crossspectral analyses of XRF-based biogenic barium records reveal that export productivity in waters overlying Site U1443 was consistently paced by precession, with amplitude modulation of the precession signal on ~405 kyr timescales and no significant variability at glacial-interglacial (obliquity) timescales. Coeval late Miocene productivity maxima and benthic δ 13 C minima during 745 eccentricity maxima at Site U1443 provides support for the hypothesis that the monsoon may have paced changes in the carbon cycle on ~405 kyr timescales (Ma et al., 2011). Significant coherence and an in-phase relationship at the precession band between biogenic barium and the SITIG suggests direct forcing of South Asian monsoon winds by insolation gradients over the late Miocene, relatively unaffected by glacial boundary conditions and long-term global cooling trends 750 . In contrast, East Asian summer and winter monsoons appear to have intensified during the late Miocene in response to global cooling and Antarctic ice sheet growth and related feedbacks Holbourn et al., 2018), although more continuous records over the late Miocene are needed to understand regional trends due to the heterogeneous nature of Asian monsoon expression. 755

Data Availability
All data will be available on the www.

Competing Interests
The authors declare that they have no conflict of interest. 770 Yellow dots indicate the modern location of IODP Site U1443. Maps were created on the ERDDAP website using the datasets Wind Stress, Metop-A ASCAT, 0.25°, Global, Near Real Time, 2009-1245 and Ocean Climatology Ocean Mixed Layer Depth MLD T02 kriging (see methods for details). c: Regional bathymetric map showing modern locations of marine and terrestrial sites discussed in this study (white squares) and the location of Site U1443, a redrill of Site 758, in the modern ocean (yellow square) and its paleolatitude at 10 Ma (yellow star, ~2°N, data from paleolatitude.org). The red circle shows the location of the SBBT sediment trap. Green 1250 arrows show surface ocean circulation during the summer monsoon (July/August) and the eastward flow of waters from the Arabian Sea into the BOB via the Southwest Monsoon Current (SMC), after Schott et al. (2009).   Table S1), and shipboard magnetostratigraphy is also shown for the interval over which it could be reliably determined. Between ~90 and 128m CCSF, sediments in cores from all holes showed scattered directional signals during pass-through magnetic remanence 1270 measurement, which hindered any determination of polarity patterns in this interval across the whole site . Black and white zones = normal and reversed polarity, respectively; grey zones = magnetic polarity not clearly determined. All numbers are age assignations for boundaries/nannofossil events in Ma (Gradstein et al., 2012 rates, e: tuned benthic δ 13 C vs age. Tie-points between a and b are shown in orange (see Table 1). An age-depth plot showing ET tie-points and good agreement with biostratigraphic age control is shown in Fig S1. f and g: wavelet analyses of tuned isotope records; white shaded area shows cone of influence and contours show 95% significance level. Main orbital periods are shown on the right in kyr. h: Filtered tuned benthic isotope records compared to filtered ET (as in b). Top: 100-kyr 1285 filtered benthic δ 13 C (blue; Gaussian filter centred on 100 kyr with bandwidth ±25 kyr to include 95 and 125 kyr peaks) compared with filtered ET (grey, identical filter design). Bottom: 41-kyr filtered  (Holbourn et al., 2018) with age model revised in (Holbourn et al., In Press), equatorial Pacific IODP Sites U1338 (Drury et al., 2018;Drury et al., 1295 2016) and U1337 (Drury et al., 2017), and ODP Site 926 (Shackleton and Hall, 1997;Drury et al., 2017;Zeeden et al., 2013). All records are shown on their latest independent orbitally-tuned chronologies. We have excluded Caribbean ODP Site 999 from this figure because it is bathed in intermediate water masses due to basin geometry and sill depths (Bickert et al., 2004). Deep South Atlantic Site 704 (Müller et al., 1991) data are not plotted due to clear age model discrepancies 1300 when compared to orbitally-tuned records. All δ 18 O records are based on Cibicidoides wuellerstorfi or C. mundulus therefore no corrections are applied, following (Jöhnch et al., 2021). The Site 926 record includes δ 13 C corrections for some samples due to the multispecific nature of the record (Drury et al., 2017). No correction was applied to C. wuellerstorfi or C. mundulus δ 13 C values.