Decadal resolution record of Oman upwelling indicates solar forcing of the Indian summer monsoon ( 9 – 6 ka )

The Indian summer monsoon (ISM) is an important conveyor in the ocean–atmosphere coupled system on a trans-regional scale. Here we present a study of a sediment core from the northern Oman margin, revealing early to midHolocene ISM conditions on a near-20-year resolution. We assess multiple independent proxies indicative of sea surface temperatures (SSTs) during the upwelling season together with bottom-water conditions. We use geochemical parameters, transfer functions of planktic foraminiferal assemblages and Mg /Ca palaeothermometry, and find evidence corroborating previous studies showing that upwelling intensity varies significantly in coherence with solar sunspot cycles. The dominant∼ 80–90-year Gleissberg cycle apparently also affected bottom-water oxygen conditions. Although the interval from 8.4 to 5.8 ka BP is relatively short, the gradually decreasing trend in summer monsoon conditions was interrupted by short events of intensified ISM conditions. Results from both independent SST proxies are linked to phases of weaker oxygen minimum zone (OMZ) conditions and enhanced carbonate preservation. This indicates that atmospheric forcing was intimately linked to bottom-water properties and state of the OMZ on decadal timescales.


Introduction
The Indian summer monsoon (ISM) is the dominant driver for intraseasonal changes of wind directions and precipitation patterns in one of the world's most densely populated regions in southern Asia.For the agricultural development and economic prosperity of the region, it is therefore vital to assess the variability in the monsoon system on societally relevant decadal to centennial timescales.This might help to better understand how potential driving forces might be controlling ISM variability and how it might develop under future climate scenarios.To precisely determine possible changes under climatic conditions in response to anthropogenic impact, it is vital to understand how ISM variability is modulated on societally relevant decadal to centennial timescales.In comparison to modern times, the early to mid-Holocene is marked by different orbital configurations, affecting Northern Hemisphere insolation (A.L. Berger, 1978) and monsoon circulations (Kutzbach and Street-Perrott, 1985).This period was marked by a steep decline in solar insolation changes, whilst climatic conditions were largely unaffected by human-induced greenhouse gas emissions.Furthermore, climatic conditions during the early to mid-Holocene have been attributed to cultural and societal turnover in Africa Published by Copernicus Publications on behalf of the European Geosciences Union.

Arabian Sea
Oxygen studied by Gupta et al. (2005).Black arrows show the schematic pathways of major currents (solid lines; SC is Somali Current and EAC is East Arabian Current) and near-surface winds (stippled lines) during the summer monsoon.White contour lines refer to the summer (July to September) sea surface temperature (SST) in 10 m water depth.Colour shading indicates oxygen concentration (mL L −1 ) in 650 m water depth, indicating a strong oxygen deficiency in the modern depth of the sampled sediment core.SST (Locarnini et al., 2010) and oxygen concentration (Garcia et al., 2010) are from World Ocean Atlas 2009.Panel (b) shows the depth profile of oceanographic data (salinity, temperature and oxygen) measured during cruise M74/1b at the station of core SL163.Grey shading refers to the location of the modern oxygen minimum zone (OMZ).A subsurface salinity maximum in 200-400 m water depth indicates the location of Persian Gulf Water (PGW).The map was created using the software ODV ver.4.7.4 (Schlitzer, 2015).
( Kuper and Kröpelin, 2006) and Mesopotamia (Kennett and Kennett, 2007).Monsoonal winds develop when summer (July to September) heating over the continent forms a low-pressure zone over continental Asia, which then interacts with the highpressure zone over the southern Indian Ocean and drives strong, moisture-laden southwesterly winds.Along the eastern coast off Somalia and Oman, the alongshore winds induce coastal upwelling of cold and nutrient-rich deeper water layers, which cool surface water temperatures and fuel biological productivity within the euphotic zone (Findlater, 1969;Wyrtki, 1973).Below the photic zone, remineralisation of sinking organic matter consumes oxygen, which leads, in combination with the lateral supply of low-oxygenated intermediate water masses (You and Tomczak, 1993), to the formation of a pronounced mid-depth oxygen minimum zone (OMZ).Upwelling and OMZ intensity, together with the biological uptake and release of carbon dioxide (CO 2 ), nitrous oxide (N 2 O) and other greenhouse gases from and into the atmosphere (Farías et al., 2009;Paulmier et al., 2011;Ward et al., 2009), play a critical role in the global climate system.
Quantitative reconstructions of upwelling intensity have the advantage of being directly related to wind strength and thus ISM intensity (Murtugudde et al., 2007).Numerous studies attempted to quantitatively reconstruct sea surface temperatures (SSTs) in the northwestern Arabian Sea to study changes in upwelling and thus ISM intensity (Emeis et al., 1995;Clemens et al., 1991;Naidu and Malmgren, 1996;Dahl and Oppo, 2006;Huguet et al., 2006;Anand et al., 2008;Godad et al., 2011).Accordingly, ISM variations are modulated by Northern Hemisphere summer insolation changes controlled by orbital parameters of the earth's precessional cycle.However, previous reconstructions of Arabian Sea SST did not resolve variations on societally relevant decadal to centennial timescales.Here, we present a high-resolution study of early to mid-Holocene ISM variability based on a multi-proxy study comprising of SST fluctuations and OMZ intensity.Reconstructions are based on a comparison of Mg / Ca measurements of the upwellingrelated species Globigerina bulloides and on transfer functions based on planktic foraminiferal (PF) census counts.To this end we first evaluate species response to potential ecological driving parameters and convert fossil assemblage counts to summer SST using a regionally validated transfer function (Munz et al., 2015).We test the ecological importance and statistical significance of the reconstructions and evaluate summer SST variability in comparison to other proxy records.We further assess the influence of surface parameters, upwelling SST and primary productivity on variations in OMZ intensity and deep-water carbonate preservation.
Sea (Fig. 1), namely the Somali Current (SC) and as its northward extension the East Arabian Current (EAC), promoting an overall anti-cyclonic circulation pattern (Shetye et al., 1994;Tomczak and Godfrey, 1994).The alongshore surface currents induce an offshore-directed deflection of surface waters and an upwelling of deeper, cold and nutrient-rich subsurface waters through the process of Ekman pumping.During boreal winter (December-March), weaker, dry and cold northeasterly winds (IWM, Indian winter monsoon) lead to increased surface water cooling and evaporation, especially in the northeastern part of the basin, where the IWM enables deep convective mixing and a second productivity peak during winter (Madhupratap et al., 1996;Schulz et al., 2002;Munz et al., 2015Munz et al., , 2016)).Enhanced nutrient availability during summer upwelling nourishes primary productivity within the photic zone near the Somali and Omani coasts.This promotes OMZ conditions at intermediate water levels (∼ 100-1000 m; Fig. 1).
The water masses in the Arabian Sea are derived from several major sources.High evaporation leads to highly saline surface water, which is therefore named Arabian Sea High-Salinity Water (ASHSW) (Shetye et al., 1994;Prasanna Kumar and Prasad, 1999).Thermocline waters are mainly sourced from an intermediate water mass (500-1500 m water depth) that originates in the central Indian Ocean (Indian Ocean Central Water, IOCW) as a mixture of Antarctic Intermediate Water (AAIW) and Indonesian Intermediate Water (IIW) (Emery and Meincke, 1986;You, 1998).Although intermediate waters have a relatively low pre-formed oxygen saturation when they arrive in the Arabian Sea (0.07-0.52 mL L −1 ), they exert a strong influence on the OMZ of the Arabian Sea.Additionally, intermediate waters are sourced from the inflow of highly saline waters forming through excess evaporation over precipitation in the Persian Gulf and Red Sea.This leads to the formation of high salinity subsurface waters (35.5-36.8 psu) and subduction under less-saline waters after entering the Arabian Sea.The Persian Gulf Water (PGW) shows a pronounced salinity maximum at 200-400 m water depth, whereas the Red Sea Water (RSW) has an equilibrium depth of 500 m at its maximum northward extent at ∼ 18 • N (Shetye et al., 1994).Although the PGW is a relatively young water mass and originates from a closer source region compared to the other intermediate water masses, the high-salinity tongue of the modern PGW increases the oxygen concentration at mid-depth by less than 0.5 mL L −1 (Fig. 1).

Samples and chronology
During RV Meteor cruise M74 on Leg 1b in September 2007 (Bohrmann et al., 2010), piston core SL163 recovered the uppermost 955 cm of the sedimentary sequence from 650 m water depth within the central part of the present-day OMZ (Fig. 1) at the northern Oman margin (21 • 55.97 N, 059 • 48.15 E).Multicorer MC681 recovered the undisturbed uppermost 53 cm of near-surface sediments at the same position.The sedimentary sequence from the intervals 1-56 cm of core SL163 and 1-53 cm of MC681 is considerably different from the topmost 1 cm and the deeper intervals of SL163.From smear slide analysis and scanning electron microscopy this deposition can be described as faintly bioturbated, olivebrown organic-rich diatomaceous nannofossil silty clay.The uppermost 1 cm from both cores as well as the deposition below 56 cm of SL163 is described as olive foraminiferal nannofossil ooze.We did not find any evidence for interruptions of the stratigraphical record in the form of slumps, turbidites, erosional surfaces or phosphorites below the unconformity at 56 cm core depth.In this study, we focus only on samples from the foraminiferal nannofossil ooze between 56 and 400 cm of SL163 and the core top (0-1 cm) of MC681 and therefore assume that the record we present here is deposited homogeneously and undisturbed.
Core SL163 was sampled in continuous 1-2.5 cm intervals.The age model of core SL163 and the core top sample of MC681 is based on 14 accelerator mass spectrometry 14 C datings (Table 1), measured at Beta Analytic, Miami, USA, and the Leibniz Laboratory in Kiel, Germany.Datings of core SL163 were mostly based on a monospecific analysis of handpicked individuals of Neogloboquadrina dutertrei (Table 1).This species is commonly known to thrive at thermocline depths (e.g.Fairbanks et al., 1982), but in the western Arabian Sea it shows a relatively shallow habitat within the upper 35 m of the water column (Peeters and Brummer, 2002).For two samples (52.5 and 58.5 cm) with insufficient foraminiferal carbonate content, the bulk organic fraction was dated.To avoid a potential bias introduced by dating different organic compounds that might have been produced during different seasons or within different water masses, the core top sample of MC681 was sampled twice using mixed planktic foraminifera and bulk organic fraction.Both analyses yielded comparable results (Table 1), indicating a similar synthesis and general comparability of both dating methods.
Radiocarbon dates were calibrated to calendar years using the Marine13 calibration curve (Reimer et al., 2013) within the program clam ver.2.2 (Blaauw, 2010) for the statistical software environment R ver.3.2.1 (R Core Team, 2015).We assume a regional reservoir correction of R = 231 ± 31 years, which is the weighted mean of the four closest (maximum distance ∼ 700 km) R determination points (Southon et al., 2002;von Rad et al., 1999) from the 14CHRONO Marine Reservoir Database (http://calib.org/marine/).Added to the global marine reservoir age of 400 years, this correction factor is consistent with the recent age of our undisturbed core top sample of MC681 (Table 1).An age-depth model was then built for SL163 using a smooth spline regression with a mixed-effect model (10 000 iterations) within clam.

Planktic foraminiferal faunal analyses and transfer function approach
Samples were freeze-dried, washed with tap water over a 63 µm screen, oven-dried at 40 • C, and dry-sieved into the size fractions 150-250, 250-315 and > 315 µm.PF were identified according to the taxonomic framework of Hemleben et al. (1989), Bé (1967), and Bé and Hutson (1977).Following the procedure of Pflaumann et al. (1996), PF counts were conducted on a sub-split of each size fraction that contained a minimum of 100 PF individuals, yielding a summed fraction (> 150 µm) to contain > 300 identified PF individuals.Relative species abundances were calculated on the summed fraction > 150 µm after multiplying each size fraction with the respective split factor.To evaluate the secondary influence of calcite dissolution, we calculated fragmentation indices following Le and Shackleton (1992).
Quantitative reconstructions from PF faunal abundances were based on a modern calibration dataset of n = 603 surface samples spanning the Indian Ocean from 30 • S to 25 • N and 33 to 119 • E recently compiled by Munz et al. (2015).To enhance the signal-to-noise ratio, rare species with < 0.5 % average abundance were removed from the calibration dataset prior to further analyses (Kucera et al., 2005), which resulted in the exclusion of 16 rare species out of 33 taxa.The dataset was then further subsampled to contain only samples most similar to the fossil assemblages of core SL163.This removes noisy effects introduced by modern samples covering ecological conditions outside the range of the fossil samples (Kucera et al., 2005).We followed the approach presented in Munz et al. (2015) and restricted the calibration dataset to samples that had the same range along the first and second axes of a joint principal component analysis together with the fossil dataset (using squareroot-transformed species abundances to dampen the effect of a few dominant taxa).The resulting subsampled dataset used for further analyses contained n = 96 modern surface samples (Fig. 2).
Environmental gradient analysis was then carried out on the 96 subsampled calibration samples using the package vegan ver.2.3-1 (Oksanen et al., 2015) for R.An initial detrended correspondence analysis (DCA) revealed a short gradient length (0.75 for the longest gradient on DCA axis 1), suggesting an ordination method based on linear species response (Birks, 1998).Redundancy analysis (RDA) was therefore chosen for the construction of an ordination model to examine species response to several potential controlling environmental gradients.Because of the extreme seasonality of surface water productivity owing to the monsoonal circulation, four different approaches were used to assess modern productivity values.Net primary productivity estimates from the vertically generalized production model (VGPM), the Eppley-VGPM and the carbon-based productivity model (CbPM) were accessed through the Ocean Productivity Home Page (http://www.science.oregonstate.edu/ocean.productivity/).Satellite-derived chlorophyll α measurements of SeaWiFS and Aqua MODIS sensors averaged over the period 1998-2010 were obtained from NASA Ocean Biology (http://oceancolor.gsfc.nasa.gov/cms/).Sea surface salinity (SSS) and SST were extracted from the 10 m depth level of the World Ocean Atlas 1998 (Conkright et al., 1998).Transfer function analysis and statistical significance tests of the reconstructions were performed with the R packages rioja ver.0.9-5 (Juggins, 2015) and palaeoSig ver. 1. 1-3 (Telford, 2015).We used two common techniques for paleoenvironmental reconstructions, the Imbrie-Kipp method (IK) (Imbrie and Kipp, 1971) and weighted-averaging partial-least-squares (WA-PLS) regression (ter Braak and Juggins, 1993), to reconstruct SST from the fossil samples.The number of factors to extract for IK was based on the Kaiser-Guttman criterion, the parallel analysis after Horn (1965) and the analysis of optimal coordinates (reviewed by Courtney and Gordon, 2013).WA-PLS model complexity was evaluated from the model of lowest root mean square error of prediction (RMSEP) using bootstrapping cross validation (n = 999 cycles).

Mg / Ca analyses and palaeothermometry of G. bulloides
For Mg / Ca analyses, approximately 30 individuals of G. bulloides were picked from the size fraction 250-315 µm, gently crushed between two glass plates under the microscope to open the chambers and cleaned following the protocol of Barker et al. (2003).Trace elemental measurements were performed with an inductively coupled plasma optical emission spectrometer (ICP-OES) using an Agilent Technologies 700 series with autosampler ASX-520 Cetac and micro-nebulizer at the MARUM Center for Marine Envi-ronmental Sciences, University of Bremen, Germany.Precision of ICP-OES measurements was determined using an in-house standard with a Mg / Ca ratio of 2.93 mmol mol −1 after every five samples.The average relative standard deviation from n = 55 measurements was 0.155 % (1σ = 0.005).As a standard reference material, we analysed the international limestone standard ECRM752-1 with an Mg / Ca ratio of 3.75 mmol mol −1 (Greaves et al., 2008) prior to every batch of 50 samples.The average of Mg / Ca values over all ECRM752-1 measurements was 3.714 (n = 8; 1σ = 0.041).Three replicate measurements of every sample were used to estimate analytical precision and yielded an average relative standard deviation of 0.068 % (n = 234 samples, 1σ = 0.004).To estimate SST from Mg / Ca we used the species-specific calibration equation T = 1/0.102×LN (Mg / Ca/0.528) published by Elderfield and Ganssen (2000) for G. bulloides.This Mg / Ca-temperature relationship was previously used for calibrating Mg / Ca measurements of G. bulloides in the Arabian Sea (Anand et al., 2008;Ganssen et al., 2011).Reconstruction errors were estimated using the error propagation approach of Mohtadi et al. (2014).
Samples were screened for potential contamination by iron-manganese coatings and clay minerals not successfully removed by the cleaning technique (Barker et al., 2003).Out of 142 samples, 10 showed Fe / Ca or Mn / Ca values of more than 0.1 mmol mol −1 and were excluded prior to further analyses.No correlations exist between Mg / Ca and element / Al ratios of Al / Ca (r = 0.06) and Fe / Ca (r = 0.16) and only weak correlations for Mn / Ca (r = 0.36).Palaeotemperature estimates based on trace elemental concentrations in foraminiferal calcite can potentially suffer from post-depositional dissolution and preferential removal of more solution-susceptible Mg-rich calcite (e.g. Brown and Elderfield, 1996;Dekens et al., 2002).Pteropod concentration was always > 0, except for 11 samples where they were likely diluted by a high concentration of foraminiferal shells.The persistent presence of aragonite indicates that selective dissolution could not affect Mg / Ca ratios, as planktic foraminifera build their shells from calcite, the more dissolution resistant polymorph of calcium carbonate compared to aragonite.

Bulk geochemical analyses
Bulk geochemical analyses of the sediment were performed on average every 5 cm with X-ray fluorescence (XRF).Concentrations of manganese (Mn) and vanadium (V) were quantitatively analysed as an indicator for bottomwater redox conditions and state of the OMZ.After fusion of the samples with lithium metaborate at 1200  Brussels).The precision for major elements was generally better than ±0.5 % and better than 5 % for trace elements.To decipher a possible secondary alteration of the geochemical composition, the values were expressed as enrichment relative to the detrital background, or average shale, as expressed by Wedepohl (1971) and Wedepohl et al. (1991).Accordingly, it is calculated as the element / Al ratio of the sample divided by the element / Al ratio of average shale.Hence, element / Al ratio and enrichment factor are co-varying, with a factor > 1 indicating enrichment and < 1 indicating depletion relative to the detrital background.Biogenic opal was determined photometrically after wet alkaline extraction of biogenic silica (BSi) using a modification of the DeMaster method (DeMaster, 1981).About 30 g of dry sediment per sample was digested in 40 mL of 1 % sodium carbonate solution (Na 2 CO 3 ) in a shaking bath at 85 • C.After treatment with 0.021 M HCl, the neutralised supernatant was analysed after 3, 4, and 5 h and the amount of BSi was estimated from the linear intercept through the time course aliquots.Slope correction was used to prevent an overestimation of BSi by dissolution of clay minerals at low BSi concentrations (Conley, 1998).Biogenic opal was then determined by multiplying the BSi concentrations with a factor of 2.4.Duplicate measurements revealed a mean standard deviation of 0.13 %.

Spectral analyses
Spectral analyses of the proxy records from planktic foraminiferal transfer functions, Mg / Ca SST, and OMZ intensity with the multi-taper method (MTM; Mann and Lees, 1996) were computed with the SpectraWorks software kSpectra© ver.3.4.5 and a red noise null hypothesis (Ghil et al., 2002) using the default setting of p = 2 and K = 3 tapers.A Gaussian band-pass filter was used to reveal the signature of the dominant cycles in the data.After resampling the time series to the average sampling rate, filtering was carried out using the software program AnalySeries ver.2.0.8 (Paillard et al., 1996).A cross wavelet transform (XWT; Grindsted et al., 2004) of both temperature proxy time series was calculated with the bi-wavelet package ver.0.17.10 for R. MTM and XWT analyses were conducted on trendremoved time series interpolated to regular average sample spacings using piecewise cubic polynomial interpolation (function "pchip" of the signal package ver.0.7-6 for R).To estimate a linear relationship of the low-frequency signals between the differently spaced time series, a new common time axis was produced where signals were consecutively averaged into 60-year bins with a 20-year overlap.

Age control
The 1-2.5 cm sample spacing yielded an average temporal sampling distance of ∼ 19 years over the entire interval.The sample spacing fluctuates between < 10 years in the upper 50 cm of the core and close to 50 years in the lowermost 15 cm of the core.The sharp lithofacies change above the studied interval at 56 cm core depth is marked by a sedimentation hiatus of ∼ 3600 years (Table 1).Based on the accumulation rates above and below the unconformity, this corresponds to a thickness of the missing sedimentary sequence of ∼ 1.5 m.The diatomaceous ooze above the unconformity at 56 cm core depth of SL163 showed an excursive increase in the sedimentary water content of < 17 %.One possible reason for the sedimentation hiatus might thus be gravitational instability due to the high water content of the organicrich diatomaceous silty clay deposited at the steeply inclined northern Oman margin.However, further discussion of this issue is beyond the scope of this study and will be presented elsewhere.
We observed two samples where the age-depth relationship was reversed within the lower half of the core (Fig. 3).However, the maximum age deviation is lower than the 2σ probability of both dating points, enabling the fit of a smooth spline model with continuous deposition rates and continuously increasing ages.As we did not find any indications of interruptions or other inconsistencies of the sedimenta-Clim.Past, 13, 491-509, 2017 www.clim-past.net/13/491/2017/tion within the studied interval, we are confident that our age model represents a continuous deposition.Instead, we assume the offset to be either the result of changing water mass ages due to changing upwelling intensities, as it was shown for the Peruvian upwelling (Fontugne et al., 2004), or due to mixing by bioturbation.

PF faunal analyses of core SL163 and palaeothermometry
A total of 29 PF morphospecies were identified in core SL163, whereof six species showed a total average abundance of > 5 % (Fig. 4).The overall PF fauna is dominated by Globigerina bulloides (39.2 %), followed by Globigerinita glutinata (11.5 %), Globigerinoides ruber (11.2 %), Globigerina falconensis (9.2 %), Globigerinoides sacculifer (8.3 %) and Globigerinella siphonifera (7.4 %).Core top studies (Bé and Hutson, 1977;Hutson and Prell, 1980;Prell and Curry, 1981), plankton tow casts (Peeters and Brummer, 2002) and sediment trap studies (Curry et al., 1992;Conan et al., 2002;Conan and Brummer, 2000;Mohan et al., 2006) in the Arabian Sea indicate that G. bulloides is the dominant species during upwelling season.Relative abundances of G. bulloides were used in a number of studies to express upwelling and ISM intensity (Naidu and Malmgren, 1996;Anderson et al., 2002;Gupta et al., 2005).The high numbers of G. bulloides throughout the studied interval of core SL163 suggest highly elevated primary productivity during the early to mid-Holocene summer upwelling at this station.Modern surface water properties and plankton productivity at the northern Oman margin are dominated by the seasonal upwelling during boreal summer.Because PF assemblages are dominated by species produced during the upwelling season (Curry et al., 1992), we investigated the environmental control on the PF fauna during summer (July to September; J-A-S).RDA results indicate, that summer SST correlates best with the first RDA axis and is the strongest determinant in the explanation of PF assemblages among the investigated parameters (Table 2).Transfer functions were therefore calibrated to summer SST.
The three methods used for determining the factor numbers to retain for the IK transfer function approach suggested a number of n = 4 factors.For WA-PLS, a two-component model showed the best model performance (r 2 = 0.65) and lowest cross-validated error estimates (RMSEP = 0.95 • C).Performance estimates for both methods are given in Table 3. Statistical significance of the reconstructions was tested using a novel method of random forest reconstructions (Telford and Birks, 2011).The analysis shows that summer SST can be reconstructed from fossil PF assemblages of core SL163 with a high statistical significance (Table 3; Fig. 5).Reconstructed summer SST from both techniques (IK and WA-PLS) show a high linearity (r = 0.82; p < 2.2 × 10 −16 ), indicating a low model-specific bias (Kucera et al., 2005).The resulting consensus of reconstructed summer SSTs from the Measured Mg / Ca values range between 3.9 and 7.3 mmol mol −1 (Fig. 4), corresponding to water temperatures ranging from 19.6 to 25.7 • C. The Mg / Ca value of 6.66 mmol mol −1 of the modern core top sample at this station (MC681) yields SST estimates (T = 24.85• C) close to modern summer SST of the upper 50 m of the water column (WOA1998 = 24.81• C), indicating that the calibration equation of Elderfield and Ganssen (2000) is applicable to the Mg / Ca temperature dependence of G. bulloides at the northern Oman margin.Error propagation (Mohtadi et al., 2014) yields an average propagating 1σ error of 0.89 • C. We are aware of a potential bias introduced by a genotype-specific fractionation of Mg / Ca ratios, which was recently shown to occur between two species of G. bulloides that prefer warm water (type I) and cool water (type II) in the Arabian Sea www.clim-past.net/13/491/2017/Clim.Past, 13, 491-509, 2017  (Sadekov et al., 2016).However, the high numbers of G. bulloides observed throughout the studied interval of core SL163 suggest intense upwelling conditions and we therefore assume that we mainly sampled specimens of the coolwater genotype, type IIf, which is restricted to upwelling stations of the northern Oman margin (Sadekov et al., 2016).

Proxies of OMZ conditions and carbonate preservation
Trace-element distributions of manganese (Mn) and vanadium (V) were used as an indicator for bottom-water redox conditions and state of the OMZ, following Tribovillard et al. (2006).V is enriched throughout the studied interval and enrichment relative to average shale ranges from 1.0 to 1.7.With average values of 0.92, Mn is mostly depleted except for four short phases centred at 6.0, 6.3, 7.2 and 8.4 ka BP (Fig. 6g).The concentration of pteropod fragments ranges from 0 (pteropod barren) to 5.2 × 10 3 numbers per gram dry weight and PF fragmentation ranges between 0.1 and 12.2 % (Le and Shackleton, 1992).

Interpretation of the two independent planktic foraminiferal SST proxies
A direct comparison of the Mg / Ca SST of G. bulloides and assemblage-based (consensus of IK and WA-PLS) summer SST reveals no linear relationship (r = 0.02, p = 0.91), although the binned time series show a weak positive linearity (r = 0.24, p < 0.06).This indicates that the lowfrequency signals of both time series are linearly related.Furthermore, the time series of both proxy records reveal a common trend of increasing temperatures over the record (Fig. 6a-b) and are coherent on a wide range of frequencies.Significant (> 95 % confidence) coherence was found on ∼ 1300-, ∼ 110-and ∼ 60-year periods, as well as on ∼ 150-, ∼ 90-and ∼ 40-year periods (> 99 % confidence).
Both SST records are thus in good general agreement, but Mg / Ca SSTs are on average 2.3 • C colder.This relative offset of the SST reconstructions might be attributed to timing differences during the recording of the respective proxy signal.G. bulloides is interpreted to carry the upwelling signal during peak production of this species (Peeters et al., 2002;Anand et al., 2008).Highest flux rates of G. bulloides are found from May to October, but with a clear maximum from late July to September (Curry et al., 1992;Conan and Brummer, 2000;Peeters et al., 2002).Furthermore, several stud-  2015).The null distribution is produced by generating 999 random environmental variables and reconstructing these variables from PF assemblages of core SL163 using (a) the Imbrie-Kipp method (IK) and (b) the weighted-averaging partial-least-squares (WA-PLS) method.Grey shaded histograms indicate the amount of variance of the fossil data explained by the 999 random environmental variables, whereas the black vertical lines indicate the variance explained by the reconstruction of summer productivity (VGPM) and summer SST using the different methods.The red dashed line represents the 95th percentile of the null distribution; the black dashed line represents the variance explained by the first RDA axis.Thus, the proportion of variance explained by the IK method for the reconstruction of summer SST from the fossil dataset is higher than 95 % (p = 0.015) of reconstructions of 999 random environmental variables.The summer SST reconstruction from the WA-PLS method is higher than 99 % of the random reconstructions (p = 0.007).The VGPM-based reconstruction of summer surface primary productivity is not significant (IK: p = 0.76; WA-PLS: p = 0.16).
ies in the Arabian Sea (Peeters et al., 2002;Schiebel et al., 2004;Friedrich et al., 2012) and in the Java upwelling region (Mohtadi et al., 2011) indicate that G. bulloides thrives in coastal upwelling areas mostly within the mixed-layer and upper thermocline waters in the uppermost 50-60 m of the water column.Assemblage-based SST reconstructions are calibrated to the 10 m depth level of summer temperatures.We therefore interpret Mg / Ca values of G. bulloides to represent calcification temperatures of mixed-layer and upper thermocline waters during peak production of the late summer upwelling and assemblage-based SST reconstructions to represent a shallower SST average from July to September.

Evidence for early to mid-Holocene monsoon variability
The range of reconstructed water temperatures of 6.16 • C and a mean absolute deviation of 0.72 • C from Mg / Ca measurements of G. bulloides (upwelling SST) suggest that upwelling temperatures at the northern Oman margin fluctuated strongly and rapidly during the relatively short interval of the early to mid-Holocene covered by SL163.The amplitude of temperature fluctuations recorded by G. bulloides is approximately twice as high compared to oxygen isotope temperatures of the same species at the Somali margin during the early Holocene (Jung et al., 2002).Records from the Arabian Sea revealed that strongest ISM conditions occurred during the early Holocene around 8-11 ka BP (Staubwasser et al., 2002;Gupta et al., 2003;Thamban et al., 2007), followed by a gradual weakening around 7 ka BP, which is concomitant with the time period of relatively warm assemblage-based SST (Fig. 7).However, the overall trend in our record is not consistent and instead indicates gradual weakening of ISM conditions established as early as 8 ka BP.Consistently low SSTs suggest that upwelling and ISM intensity was strongest during the intervals ∼ 7.5-8.1 ka, followed by events of strong ISM at ∼ 7.0-7.3and 6.1 ka BP (Fig. 6a).Weaker ISM conditions are reflected by warmer-than-average SST values at 6, 6.5-6.9, 7.4 and 8.2 ka BP.Other palaeoceanographic records from the Arabian Sea (Fig. 7c-d) and stalagmite records from the nearby Hoti and Qunf caves (Fig. 6i-j) indicate low monsoonal precipitation at 6.3, 7.4 and 8.3 ka BP (Fleitmann et al., 2007;Neff et al., 2001), suggesting that summer monsoonal winds (reflected by low summer and upwelling SST) and amount of precipitation are in-phase during the early to mid-Holocene.This is an interesting finding, as Fleitmann et al. (2004) show an apparent decoupling of ISM precipitation amount and wind intensity (represented by percentage of G. bulloides) since the year ∼ 1900 AD, indicating that higher upwelling intensities do not necessarily involve increasing precipitation.Moreover, the most noticeable events of weak ISM conditions are in concordance with advection of North Atlantic drift ice from Bond et al. (2001) (Fig. 7e), centred at ∼ 5.9 (Bond event 4), ∼ 7.4 (Bond event 5) and ∼ 8. indicate phases of increased surface water productivity.Aragonite preservation is indicated by the concentration of pteropod fragments (f) and bottom-water oxygen conditions, and state of the oxygen minimum zone (OMZ) is shown with relative enrichment or depletion of manganese (g) and vanadium (h).Note the inverse relationship of both elements, as explained in the text.Stable oxygen isotopes of (i) Qunf Cave (Fleitmann et al., 2007) and (j) Hoti Cave stalagmites (Neff et al., 2001) expressing precipitation in Oman.Vertical solid green bars indicate periods of low biogenic opal content, concomitant with decreased OMZ intensity and increased aragonite preservation.Error bars for panel (a) were calculated by propagating the errors introduced by the Mg / Ca measurements and the Mg / Ca-temperature calibration (Gaussian error propagation; see Mohtadi et al. (2014)).Grey shading gives the uncertainty estimates for the transfer functions in panel (b), based on the sample-specific root mean square error of prediction (RMSEP) using bootstrapping cross validation.(e) Hematite-stained grains (HSG, %) indicate ice rafting events and cold spells in the North Atlantic (Bond et al., 2001).Bond events 4 (∼ 5.9 ka BP) and 5 (∼ 7.4 ka BP) are indicated.temporaneous occurrence of North Atlantic cold spells and weak Asian monsoon was previously found in speleothem (Dykoski et al., 2005;Fleitmann et al., 2007;Liu et al., 2013), lake (Menzel et al., 2014) and other marine records (Gupta et al., 2003).Thus, our findings are further evidence corroborating the presence of an atmospheric teleconnection linking North Atlantic climate to summer monsoonal conditions in tropical Asia during the early to mid-Holocene.However, in order to significantly assess millennial-scale cycles in the upwelling SST off the Oman margin, further studies are needed, as the time span of ∼ 2.5 ka covered by our record is likely too short.(d, e) with the record of sunspot numbers (Solanki et al., 2004) indicates that both records significantly co-vary on a wide range of frequencies.All time series were trend-removed prior to the analyses.
MTM analysis revealed statistically significant periodicities of assemblage-based summer SST at ∼ 1300 and 75-95 years per cycle (Fig. 8a).Mg / Ca-based upwelling SSTs are modulated on frequencies at 110-130, 80-90 and ∼ 40 years (Fig. 8b).The longer ∼ 110-130-year cycle was previously found in a number of records from the Asian monsoon realm (Berger and von Rad, 2002;Dykoski et al., 2005;Gupta et al., 2005) and is close to the 132-year subharmonic of the Hale cycle (Attolini et al., 1990), previously identified to be modulating the Oman upwelling system during the Holocene (Gupta et al., 2005).The ∼ 80-90-year cycle has been observed in several studies of ISM variability (Neff et al., 2001;Fleitmann et al., 2003;Dykoski et al., 2005;Gupta et al., 2005) and was interpreted to be most likely influenced by the 88-year solar Gleissberg cycle.Turner et al. (2016) found, however, that periodicities within the range of the Gleissberg cycle are also common in random-walk simulations and could be statistical artefacts from the sampling resolution and the age model applied.In fact, the overall error of the reservoir correction of ±31 years is close to the observed periodicities, which might give an indication of why we found the Gleissberg cycle but not the longer-period (∼ 200-210 years) de Vries solar cycle.However, Duan et al. (2014) showed with a speleothem record from Dongge Cave (China), that coherency of eastern Asian monsoon precipitation and the solar Gleissberg cycle is persistent over the last 4 kyr, whereas coherency with the de Vries cycle was only observed over a short 1 kyr long period.Other Holocene Asian monsoon records also showed periodicities similar to the Gleissberg cycle, while lacking the longer de Vries cycle (e.g.Wang et al., 1999;Sagawa et al., 2014).It could thus be hypothesised that the de Vries cycle is not detectable in our record, due to a transient link of the solar-monsoon relationship, potentially suppressed by changes in the oceanatmosphere system in the El Niño-Southern Oscillation domain (Berkelhammer et al., 2010).We therefore tested a possible solar component on the decadal-scale forcing of our SST records by evaluating the coherence of both time series with the record of reconstructed sunspot numbers (Solanki et al., 2004).The coherence pattern reveals that both SST records and sunspot numbers are coherent on a wide range of periodicities (630, 190-230, 160, 110-130, 80-90, ∼ 70, ∼ 50 and ∼ 40 years per cycle; Fig. 8d and e).The Mg / Cabased SST record shows the strongest coherency with cycles of the sunspot record in both the ∼ 88-year Gleissberg and the ∼ 190-230-year de Vries periodicities.This observation further strengthens the hypothesis that ISM variability is not only controlled by orbital-scale insolation forcing, indicated by the long-term trend of warming temperatures and decreasing ISM intensity, but also by solar forcing.
To illustrate the signature of the observed cycles in the data, band-pass-filter outputs of both SST records are shown in Filtered components were extracted using a Gaussian band-pass filter with a central frequency of 0.0118 cycles year −1 (∼ 85-year cycle) and 0.0083 cycles year −1 (∼ 120-year cycle), respectively, and a bandwidth of 0.001 cycles year −1 using the program AnalySeries 2.0.8 (Paillard et al., 1996).
(TSI) are offset by a lag of ∼ 400 years.A direct forcing on this bandwidth seems unlikely, as modelling studies indicate a response of surface temperatures to solar forcing within less than 50 years (Seidenglanz et al., 2012).However, within the limits of our age model, comparison of the amplitude variations in the filter outputs from both SST time series and the TSI record of Steinhilber et al. (2012), filtered on the ∼ 85-year Gleissberg bandwidth, apparently indicates synchrony of our SST records and TSI with maximum amplitude modulations of all three records centred at ∼ 7.4 ka BP.Our new high-resolution record of summer and upwelling SST therefore provides further strong evidence that solar forcing persistently modulated ISM variability during the early to mid-Holocene after the summer insolation maximum, previously found from speleothem (Neff et al., 2001;Burns et al., 2002;Fleitmann et al., 2003;Dykoski et al., 2005;Wang et al., 2005;Duan et al., 2014) and marine records (von Rad et al., 1999;Wang et al., 1999;Gupta et al., 2005).

Carbonate preservation and bottom-water redox conditions
In order to study the interplay of ISM intensity and OMZ fluctuations, we compared our new multi-proxy reconstruction of summer SST to OMZ reconstructions of Böll (2014).This study previously evaluated ISM intensity and OMZ conditions during the deposition of SL163 in a spatial context using alkenone-derived annual mean SST, stable nitro-gen isotopes and published OMZ reconstructions, concluding that changes in OMZ intensity were linked to variability in intermediate water ventilation and monsoon strength.
Our new multi-proxy records that primarily respond to summer upwelling conditions enable the study of the interplay of ISM intensity and OMZ fluctuations during the early to mid-Holocene in more detail.Under oxic bottom-water conditions, Mn is precipitated as Mn oxyhydroxides.Under suboxic or hypoxic bottom-water conditions, MnO can be reduced to soluble Mn 2+ and escape into the water column (Calvert and Pedersen, 1993;Calvert et al., 1996;Schnetger et al., 2000;Böning et al., 2004).Mn is mostly depleted in core SL163, indicating Mn loss through the OMZ, except for three short intervals of Mn enrichment centred at ∼ 6.0, 6.3 and 7.2 ka BP, as well as from 8.3 ka BP to the end of the record (Fig. 6g).However, precipitation and fixation of Mn 2+ in the sediment as MnCO 3 can occur when bottom waters become completely anoxic (Tribovillard et al., 2006;McKay et al., 2015).This finding corroborates the study of Böll (2014), who found increased denitrification, indicating more intense OMZ conditions during the intervals 5.9, 6.1, 6.4 and 7.2 ka BP (based on the same age model).V is precipitated under anoxic conditions (Emerson and Huested, 1991) and constant enrichment values between 1.0 and 1.7 relative to average shale also indicate permanent anoxic bottomwater conditions during the deposition of core SL163.Carbonate preservation was assessed by PF fragmentation indices and the abundance of pteropod fragments in the sieve Clim.Past, 13, 491-509, 2017 www.clim-past.net/13/491/2017/fraction > 150 µm.Pteropod shells primarily consist of aragonite, the less stable polymorph of calcium carbonate as compared to calcite (Morse et al., 1980).In the northern Arabian Sea, aragonite has a modern compensation depth (ACD) of ∼ 500 m (W.H. Berger, 1978;Böning and Bard, 2009) and a lysocline at ∼ 1 km water depth (Böning and Bard, 2009), thus approximately 350 m below the station of core SL163.However, in high-productivity environments supra-lysoclinal dissolution can occur well above this depth, induced by respiration of organic matter and metabolic release of CO 2 (e.g.Milliman et al., 1999).Thus, enhanced pteropod preservation indicates either ACD deepening (Reichart et al., 2002) or less supra-lysoclinal dissolution due to decreased production of organic matter.Phases of increased concentrations of pteropod fragments are linked to the short intervals of slightly enriched Mn values (Fig. 6f-g).This indicates short phases of less corrosive bottom waters due to potentially weaker OMZ conditions.Furthermore, the variability in Mn / Al values (expressed as Mn enrichment) is modulated on dominant ∼ 170-and ∼ 310-year cycles (Fig. 8c), which are approximately one-half (170) and one-fourth (340) the ∼ 80-90-year frequency of the Gleissberg cycle.This suggests that bottomwater oxygenation state is modulated on the same frequencies as SST variability and upwelling intensity.
The intervals of increased pteropod concentration and lower OMZ intensity at ∼ 6, 6.3, 7.2 and 8.3-8.5 ka BP are linked to intervals of low biogenic opal and thus occurred when surface water productivity was diminished.The former three of these intervals also correspond to low abundances of the upwelling indicator species G. bulloides (Fig. 6d), but during the latter interval at 8.3-8.5 ka BP G. bulloides abundances are high (> 45 %).In addition, Mg / Ca measurements of G. bulloides at 8.3 ka BP are > 7 mmol mol −1 and inferred SSTs are > 25.5 • C. A similar feature of a very short warm excursion of Mg / Ca temperatures from G. bulloides with concomitantly high abundances of this species around 8 ka BP was observed by Anand et al. (2008) in their record from the Somali upwelling system.It may be speculated here whether a potential intrusion of the warm-water genotype of G. bulloides (Sadekov et al., 2016) during the short-lived period around 8.3 ka BP had occurred, leading to erroneously warm Mg / Ca temperature estimates during this period.
Regardless of the apparent difference in absolute temperatures, wavelet analysis and cross-wavelet transform (XWT) reveals three distinct intervals, two in the ∼ 40-90-year band (6-6.2, 7.2-7.6ka BP) and one in the ∼ 110-130-year band (8.1-8.4 ka BP), where both records show significant common power (Fig. 10).The intervals correspond well with the phases of enhanced pteropod preservation and lower OMZ intensity.The cross-wavelet phase angle of XWT further indicates that both SST records are phase-shifted during these intervals.The arrows in Fig. 10 within the significant areas point mostly upwards, which indicates either that assemblage-based SSTs lead Mg / Ca SSTs by π/2, i.e. 90 • , or that assemblage-based SSTs lag Mg / Ca SSTs by 3π/2, i.e. 270 • .Although these phases are relatively short-lived and both scenarios seem reasonable from the visual comparison of the two SST time series, a systematic lead of faunal SST reconstructions compared to alkenone-derived SST was previously found by Cayre and Bard (1999) in a study from the eastern Arabian Sea.However, their study observed a delay of several thousands of years, which was potentially produced by strong productivity changes (Bard, 2001).It should be noted that the phase relationships at the longer wavelength could be erroneous, as the behaviour is not consistent and partly truncated by edge effects below the cone of influence.
Several studies in the Arabian Sea indicate an alternating influence of different water masses during the deglacial and early Holocene (Zahn and Pedersen, 1991;Jung et al., 2001;Schmiedl and Leuschner, 2005;Gupta et al., 2008;Jung et al., 2009).The site of core SL163 is within the modern range of intermediate water masses, which form a mixture of RSW, PGW and IOCW down to ∼ 1500 m water depth (Emery and Meincke, 1986;Shetye et al., 1994).During the last glacial stage, the outflow from the two mediterranean basins connected to the Arabian Sea (Red Sea and Persian Gulf) was highly suppressed due to lowering of the global sea level potentially close to or even below the respective sill depths (Rohling and Zachariasse, 1996).Gupta et al. (2008) discussed the finding that ventilation changes at the northern Oman margin during the early Holocene are distant to RSW and are more likely attributed to an alternating contribution of North Atlantic Deep Water to the deeper water masses.In comparison to their study site, our core site is further north, close to the modern Ras Al Hadd frontal zone, where PGW enters the Arabian Sea.At the station of SL163, modern PGW reaches down to 400 m (Fig. 1).During phases of more restricted water exchange and increased evaporation, subsurface salinities in the Persian Gulf might have been higher, which could have enabled PGW to reach the core site at deeper depth.However, the modern oxygen input of PGW is very low and OMZ conditions at this depth are still fully hypoxic.Instead, the concomitant occurrence of high common cross-wavelet power of the SST records (Fig. 10) with less corrosive bottom waters indicates that deep-water conditions are linked to SST variations and atmospheric forcing was intimately linked to the deep-water properties.

Conclusions
Among the environmental variables investigated, summer SST primarily controls the underlying variability in PF assemblages from the northwestern Arabian Sea.Statistically significant reconstructions of summer SST from PF assemblages and Mg / Ca SST from G. bulloides are colder than present summer temperatures, indicating vigorous monsoonal winds and upwelling intensity during the early to mid-Holocene.ISM conditions were generally strongest around 8 ka BP and gradually decreased towards 5.8 ka BP.The trend of weakening monsoon conditions was interrupted by decadal-scale episodes of intensified ISM around ∼ 7 and 6.2 ka BP, which is concomitant with pluvial episodes indicated by the Oman speleothem records.Summer SST fluctuations furthermore reveal that upwelling intensity was coupled to decadal-to-centennial-scale sunspot cycles.Low monsoonal conditions, indicated by warm upwelling SST and summer SST and low surface water productivity, are associated with enhanced pteropod preservation and weaker OMZ conditions.Enrichment of manganese as a proxy for bottom-water oxygenation varies on multiples of the dominant ∼ 80-90-year Gleissberg cycle, suggesting that upwelling intensity and OMZ conditions are modulated by solar activity.We therefore conclude that, instead of intermediate water mass changes, atmospheric forcing is the main driver of OMZ conditions.Data availability.The data presented in this paper are available electronically at the PANGAEA Data Publisher for Earth & Environmental Science (doi:10.1594/PANGAEA.874696,Munz et al., 2017).
Competing interests.The authors declare that they have no conflict of interest.

Figure 1 .
Figure 1.(a) Map of the Arabian Sea, showing the location of gravity core SL163 and multicorer MC681 (white cross) at the northern Oman margin (21 • 55.97 N, 059 • 48.15 E) in 650 m water depth, as well as Integrated Ocean Drilling Program (IODP) core 723A (black cross) studied byGupta et al. (2005).Black arrows show the schematic pathways of major currents (solid lines; SC is Somali Current and EAC is East Arabian Current) and near-surface winds (stippled lines) during the summer monsoon.White contour lines refer to the summer (July to September) sea surface temperature (SST) in 10 m water depth.Colour shading indicates oxygen concentration (mL L −1 ) in 650 m water depth, indicating a strong oxygen deficiency in the modern depth of the sampled sediment core.SST(Locarnini et al., 2010) and oxygen concentration(Garcia et al., 2010) are from World Ocean Atlas 2009.Panel (b) shows the depth profile of oceanographic data (salinity, temperature and oxygen) measured during cruise M74/1b at the station of core SL163.Grey shading refers to the location of the modern oxygen minimum zone (OMZ).A subsurface salinity maximum in 200-400 m water depth indicates the location of Persian Gulf Water (PGW).The map was created using the software ODV ver.4.7.4(Schlitzer, 2015).

Figure 2 .
Figure 2. Scatterplot of the joint principal component analysis (PCA) of n = 603 modern calibration samples (blue triangles) and fossil downcore samples of SL163 (red crosses).The variance of the first and second PCA axes is indicated.N = 96 modern samples within the red polygon are most similar to the fossil samples along the first and second PCA axes, and were used for the calibration of transfer functions.

Figure 3 .
Figure3.Age-depth relationship of the early to mid-Holocene section of core SL163 (55-400 cm).Blue areas indicate the conventional 14 C-calibrated ages; the black line indicates the interpolation between the dated samples using a smooth spline fit and the 95 % confidence level (grey shading).

Figure 4 .
Figure 4. Time series of the six most common planktic foraminiferal species (> 5 % average abundance) and Mg / Ca measurements of G. bulloides from the studied interval in core SL163.Dashed lines are the real data; thick solid lines give the 3 pt running average.

Figure 5 .
Figure5.Results of the significance test of assemblage-based palaeoenvironmental reconstructions using the R package palaeoSig(Telford, 2015).The null distribution is produced by generating 999 random environmental variables and reconstructing these variables from PF assemblages of core SL163 using (a) the Imbrie-Kipp method (IK) and (b) the weighted-averaging partial-least-squares (WA-PLS) method.Grey shaded histograms indicate the amount of variance of the fossil data explained by the 999 random environmental variables, whereas the black vertical lines indicate the variance explained by the reconstruction of summer productivity (VGPM) and summer SST using the different methods.The red dashed line represents the 95th percentile of the null distribution; the black dashed line represents the variance explained by the first RDA axis.Thus, the proportion of variance explained by the IK method for the reconstruction of summer SST from the fossil dataset is higher than 95 % (p = 0.015) of reconstructions of 999 random environmental variables.The summer SST reconstruction from the WA-PLS method is higher than 99 % of the random reconstructions (p = 0.007).The VGPM-based reconstruction of summer surface primary productivity is not significant (IK: p = 0.76; WA-PLS: p = 0.16).

Figure 6 .
Figure 6.Time series of reconstructed SST from (a) the Mg / Ca ratio of G. bulloides (blue) indicating upwelling SST and (b) the consensus from the assemblage-based transfer function methods, IK and WA-PLS (red), indicating average summer SST (July-September).Thin lines and symbols represent real data; thick dashed lines represent the 3 pt running averages.Relative abundance of G. bulloides from IODP site 723A from Gupta et al. (2005) (c) and SL163 (d), together with biogenic opal content (e),indicate phases of increased surface water productivity.Aragonite preservation is indicated by the concentration of pteropod fragments (f) and bottom-water oxygen conditions, and state of the oxygen minimum zone (OMZ) is shown with relative enrichment or depletion of manganese (g) and vanadium (h).Note the inverse relationship of both elements, as explained in the text.Stable oxygen isotopes of (i) Qunf Cave(Fleitmann et al., 2007) and (j) Hoti Cave stalagmites(Neff et al., 2001) expressing precipitation in Oman.Vertical solid green bars indicate periods of low biogenic opal content, concomitant with decreased OMZ intensity and increased aragonite preservation.Error bars for panel (a) were calculated by propagating the errors introduced by the Mg / Ca measurements and the Mg / Ca-temperature calibration (Gaussian error propagation; seeMohtadi et al. (2014)).Grey shading gives the uncertainty estimates for the transfer functions in panel (b), based on the sample-specific root mean square error of prediction (RMSEP) using bootstrapping cross validation.

Figure 8 .
Figure8.Results of the MTM spectral analysis show the frequency distribution and identified significant periodicities of the (a) assemblagebased SST record, (b) the SST record using Mg / Ca ratios of G. bulloides and (c) the Mn / Al ratio as a proxy for OMZ conditions.Black lines indicate 90 and 95 % significance levels against red noise.Cross-spectral coherence of both SST records (d, e) with the record of sunspot numbers(Solanki et al., 2004) indicates that both records significantly co-vary on a wide range of frequencies.All time series were trend-removed prior to the analyses.

Figure 9 .
Figure 9.Comparison of filtered assemblage-based SST (red line, upper panel) and G. bulloides Mg / Ca SST (blue lines, lower panel) with filtered reconstructed total solar irradiance (TSI) from Steinhilber et al. (2012) (grey lines); 88 years per cycle are illustrated in the upper panel and 132 years per cycle in the lower panel.Filtered components were extracted using a Gaussian band-pass filter with a central frequency of 0.0118 cycles year −1 (∼ 85-year cycle) and 0.0083 cycles year −1 (∼ 120-year cycle), respectively, and a bandwidth of 0.001 cycles year −1 using the program AnalySeries 2.0.8(Paillard et al., 1996).

Figure 10 .
Figure 10.Continuous wavelet transform of (a) assemblage-based summer SST and (b) Mg / Ca-based upwelling SST.The 10 % significance level against red noise is shown as black contour lines.(c) Cross-wavelet transform (XWT) of the trend-removed Mg / Cabased and assemblage-based SST time series.The 5 % significance level is shown as thick contours.Relative phasing of both records is indicated by arrows.Arrows within the significant intervals point mostly upwards, which indicates a consistent phase shift of both records by 90 or 270 • , respectively, of assemblage-based SST relative to Mg / Ca-based SST.The cone of influence, where edge artefacts might be introduced, is indicated by hatching and white shading, respectively.

Table 1 .
(Reimer et al., 2013)on dating with the organic compound used, dating results and calibrated ages using the Marine13 calibration curve(Reimer et al., 2013), and a reservoir correction of R = 231 ± 31years.Analyses were conducted at Beta Analytic in Miami, FL, USA (Beta), and the Leibniz Laboratory in Kiel, Germany (KIA).

Table 2 .
Results of the redundancy analysis of the adjusted calibration dataset.Highest correlation with RDA1 axis and highest explanation of the PF assemblage variance (bold numbers) is shown by temperature (SST).

Table 3 .
Cross-validated root mean square error of prediction (RMSEP) as absolute values and relative to the range of the target variable, as well as the coefficient of determination.The results of the significance analysis after Telford and Birks (2011) from both transfer function methods are given; p indicates the statistical significance against the null distribution of 999 random reconstructions.