3.4.2. Mg/Ca Determination

On average, 15 *G. bulloides* tests from each sample were used for Mg/Ca determination. The specimens were cleaned using the standard protocol [39], omitting the reductive treatment, and were analysed for Mg/Ca ratios by inductively coupled plasma optical emission spectrometry (ICP-OES) using an Agilent 5100 instrument at the Godwin Laboratory for Paleoclimate Research, Department of Earth Sciences, University of Cambridge. Samples were dissolved in 0.1 M HNO<sup>3</sup> and centrifuged to remove any undissolved material. After an initial run to determine Ca concentration, samples were diluted to constant [Ca] (100 ppm). Mg/Ca ratios were determined by the intensity ratio method of [40] using calibration standards prepared according to [41]. Cleaning efficiency and diagenetic effects were monitored by measuring Fe/Ca, Mn/Ca, Al/Ca, Si/Ca, and Ba/Ca ratios.

The instrumental precision for the Mg/Ca ratios is ±0.51%, determined by replicate analyses of a standard solution containing Mg/Ca ratios of 1.3 mmol mol−<sup>1</sup> and a Ca concentration of 100 ppm. The accuracy of Mg/Ca ratios has been established by interlaboratory calibration [42]. Sample heterogeneity for Mg/Ca ratios has been shown to be much greater than instrumental precision [39,43]. We estimate the reproducibility of planktonic foraminiferal Mg/Ca ratios as ~8% from replicate analyses of *G. bulloides* picked from an Atlantic core-top sample.

#### **4. Results**

The record of *G. bulloides* shell mass attained from GeoB 8502-2 for the last 200 ky shows no distinct mode of variability (Figure 3a). In general, shell weights are stable, fluctuating only on a small scale (±1.1 µg, 1σ) around an average of 13.4 µg. They thus do not exhibit any glacial–interglacial cyclicity or follow the atmospheric *p*CO<sup>2</sup> fluctuations (Figure 3d). Lower shell mass weights are found within MIS 6, while values almost increase progressively after MIS 5.2. Superimposed on this pattern is a broad maximum in shell weight centred on the MIS 5/6 boundary of T-II with a duration of approximately 2300 yr, during which shell masses increase by 30% above average.

**Figure 3.** Climatic records of the last 200 ky (before present; BP): (**a**) *G. bulloides* average shell mass record from GeoB 8502-2. The grey line denotes the record's mean shell mass and the 1σ confidence interval is indicated by the shading; (**b**) specimen's preservation assessment using the Bulloides dissolution index (BDX'); (**c**) the fragmentation index of the coarse fraction; and (**d**) the record of atmospheric CO<sup>2</sup> from Vostok [44] is shown for comparison. The 1σ confidence interval for the shell mass and BDX' plots shown as transparent polygons. Numbers refer to Marine Isotopic Stages (MIS) or substages and grey shaded areas interglacial periods. Data are available in Supplementary Tables.

The multiproxy carbonate preservation assessment of the studied core is shown in Figures 2c and 3b. The record of fragmentation strikingly parallels that of the atmospheric *p*CO2, indicating the influence of carbonate chemistry changes on the study samples. Nonetheless, this influence does not prove

sufficient to severely alter foraminifera shell mass. In general, foraminiferal carbonate is found to be well preserved (Figures 2c and 3b), because any severe dissolution takes place after BDX' values of 3 [30]. The BDX' and the F.I. assessments (Figures 2c and 3b respectively) are in agreement, indicating that the preservation becomes better during the course of the last glacial period. However, regardless of the increase in foraminifera fragmentation after the Last Glacial Maximum, the *G. bulloides* specimens' ultrastructure was not found to be corroded. A similar disagreement between the dissolution proxies exists for MIS 6, where, although the surface ultrastructure of *G. bulloides* specimens shows signs of corrosion, overall foraminiferal breakage is found to be limited. The deteriorated shell preservation state during MIS 6, as suggested by the BDX', might be adequate to explain to some extent the slightly decreased *G. bulloides* shell weight values of this period, but on the other hand, the gradual decrease of the BDX' values towards the present may explain a gradual increase in mass.

The results of the geochemical analyses of the samples surrounding T-II are summarized in Figure 4 together with a focused record of *G. bulloides* shell weights, atmospheric *p*CO2, and relative sea level. δ <sup>18</sup>O ranges from −0.09 to 0.71‰ and, with the exception of the first sample in MIS 6, in general, they are found to be more depleted within the glacial period. The most depleted value is found near MIS 6.2 (133.9 ka), after which values start to increase and become heavier during the Eemian. Mg/Ca ratios range from 2.33 to 4.15 mmol/mol, but are in general consistent around 2.8 mmol/mol during this period. However, they are slightly more depleted prior to the termination and exhibit a spike with a ~46% increase in their values during the MIS 6.1 interstadial. This increase should indicate an environmental signal as all the contamination indicators (Fe, Al, and so on) were low for this sample.

The results of the µCT analysis of the samples covering the 22 ky interval surrounding the 30% increase in shell weight event during T-II are summarized in Table 1. Shells are on average ~16% thicker during the time that they are heavier (132.2 ka), with a mean thickness of 5.7 µm. They are also ~32% more voluminous/larger within the same sieve fraction during the same time, but with the highest size variability. Apart from being larger and thicker, they are also found to be more than threefold contaminated by clay infillings (19% of their interior void), thus the observed increase in weight is a result of both enhanced biomineralization and contamination. Shell densities or volume normalized weights do not change significantly (up to 7% from the densest during the time of the highest masses to the lightest during the interglacial), but values show high variability. In terms of test density or porosity, the heaviest and densest shells are only intermediate porous.

**Table 1.** Average biometric data of foraminifera from weighing and µCT (computed tomography) analysis. Test thickness is the ratio of calcite volume to calcite surface, and shell and test density are the ratio of the average shell weight to test and shell volume, respectively (see Section 3.3 for details). Sediment infilling is the specimen's internal volume percentage occupied by sediment impurities. The µCT analysis results for individual specimens are published in an accompanying paper [31].


Sample 865 at 139.1 ka should record the Penultimate Glacial Maximum (PGM) in the present archive, which, in the northern hemisphere, is placed at ~140 ka BP [45], because after that, the sea level and atmospheric CO<sup>2</sup> rise continuously (Figure 4). The low shell weights recorded during this period indicate reduced calcification, which is further supported by the µCT scans, which reveal small, thin, and highly porous shells of low density. The scanned shells closest to the peak of the Eemian interglacial period were found to be of average mass, intermediate size, thickness, and density, but were also the least porous. The interior of the Eemian and PGM specimens was found to be filled with sediment by only 5% and 4%, respectively, thus with minimum distortions to the weight measurements.

σ *δ* σ **Figure 4.** Climatic and geochemical records of the period surrounding the increased *G. bulloides* shell mass events during T-II: (**a**) average *G. bulloides* shell mass from GeoB 8502-2. The grey line denotes the mean shell mass of this period and the shading is the 1σ confidence interval. Enlarged circles denote the samples that have been X-ray microscopically analysed. The thickness of their envelope line is proportional to the average shell wall thickness shown in Table 1; (**b**) δ <sup>18</sup>Oshell of the foraminiferal calcite; (**c**) Mg/Ca ratios of the foraminiferal calcite; (**d**) the relative sea level record [46] is shown for comparison and (**e**) the record of atmospheric CO<sup>2</sup> from Vostok [44] is also shown for comparison. The 1σ confidence interval for the shell mass plot is shown as transparent polygons. Numbers refer to Marine Isotopic Stages and substages. Geochemical and shell mass data are available in the Supplementary Tables.

The geochemical analyses allow the reconstruction of the physical oceanographic parameters that characterize the period surrounding T-II at the study area and are shown in Figure 5a–c. Temperatures are relatively stable at around 17 ◦C and they are only slightly lower, in comparison with during MIS 5.5, for most of the late MIS 6 glacial. During the MIS 6.1 interstadial peak, the temperature rises to 20.3 ◦C, while after the Eemian peak, it drops to 14.8 ◦C. The salinity reconstructions show

strong changes in the salt content during this time, with a drop in salinity from ~34 to 29 psu for 3400 years between 139.1 and 135.7 ka, and then a ~24% increase to ~38 psu after the PGM before decreasing back to 34 at the peak interglacial. The same applies with the reconstructed densities of the water column that reveal lighter waters during the glacial and denser waters after MIS 6.1. Following the previous study [7], shell weights were plotted against the reconstructed seawater density values and a significant (R<sup>2</sup> = 0.43, *p* < 0.05, *n* = 12) relationship between these parameters was verified (Figure 5d), which becomes even more significant (R<sup>2</sup> = 0.74, *p* < 0.01, *n* = 12) if the (upper right) two heaviest (most contaminated by clay infillings) samples are omitted. It is evident that these two samples of excessive shell weights deviate from the resulting regression line. If this regression line dictates shell mass according to the shell's geochemical signal, then we can infer that these two samples are contaminated with 2–3 µg of sediment. μ

δ μ **Figure 5.** Geochemically reconstructed physical oceanographic parameters for the period surrounding T-II and their relation to average *G. bulloides* shell weights: (**a**) temperature estimates based on Mg/Ca ratios; (**b**) salinity reconstructions based on coupled Mg/Ca and δ <sup>18</sup>O measurements together with Barium (Ba) X-ray fluorescence (XRF) counts; (**c**) ambient seawater density reconstructions based on the preceding temperature and salinity estimates; (**d**) shell mass and seawater density regression plot. The two upper right points are the increased shell mass samples that deviate from the regression line by 2 and 3 µg, respectively. At the lower right, the contamination corrected regression equation is shown. Enlarged circles denote the samples that have been X-ray microscopically analysed. The thickness of their envelope line is proportional to the average shell wall thickness shown in Table 1. Data are available in the Supplementary Tables and XRF data from [47].

#### **5. Discussion**

The sieved based weight analysis of *G. bulloides* shells from the tropical Atlantic core GeoB8502-2 revealed a typical value of 13.4 µg and only small variability in the average mass throughout the last two climatic cycles, which does not follow the atmospheric *p*CO<sup>2</sup> record. This stability in planktonic foraminifera shell masses despite changes in atmospheric *p*CO<sup>2</sup> has also been reported in similar latitudes for the Pliocene [48], and may be the result of the environmental stability of the tropical regions. The tropics are generally an environment in which the physicochemical factors are not undergoing major changes through time [12] and this hydrological stability can explain the observed stability in the foraminifera shell weight record that we report here. The incoming solar energy varies considerably from tropical to polar latitudes. At the equatorial region, the average insolation is the highest with

only minor peak energy changes, while at middle and high latitudes, the regular periodicities in the Earth's orbit and tilt influence the amount and distribution of incoming energy and temporal changes are greater [49].

Test weights and the range of test weights in North Atlantic Pliocene and Pleistocene *G. bulloides* are similar despite the higher *p*CO<sup>2</sup> during the Pliocene [11]. However, the absolute *G. bulloides* shell mass values for specimens 300–350 µm in size reported here for a tropical Atlantic site are low compared with the North Atlantic records [48,50]. This is consistent with the influence of seawater densities on shell weights [7], as the lowest sea surface density values are found in the tropical waters, where salinity is lowest due to excess rainfall associated with the intertropical convergence zone, and increases toward the poles [51]. At higher latitudes, *G. bulloides* shell weight changes by around one-third over glacial–interglacial cycles [3,50,52], and a similar variability in test weight is seen through the Pliocene glacial–interglacial cycles at a similar location [48]. At these higher latitudes that are more sensitive to (latitudinal) changes in solar radiation through time, foraminifera shell weights follow the pronounced hydrological changes between the climatic cycles, with the most sensitive latitudes for explaining the glacial/interglacial ice-dynamics around 65◦ [53], leaving the tropical shell mass records almost invariable. The consistency in foraminifera shell mass may reflect hydrological stability in the study area, and thus stability of the ITCZ locations during the last 200 ky.

Both dissolution proxies (F.I. and BDX') agree that the biogenic carbonate preservation is generally good to very good. There is a clear match in the trends of both proxies for deterioration in the preservation of foraminifera specimens from MIS 2 to MIS 5.5. Increased specimen dissolution is also evident in the tomographs of the mid T-II sample [54]. This deteriorating trend, however, is not sufficient to cause obvious reductions in the weights of *G. bulloides* shells during this time interval. The two proxies disagree within MIS 6, where the degree of ultrastructural corrosion does not parallel the degree of specimens' fragmentation. This discrepancy points either to the application limits of the proxies or is the result of different test architecture. The F.I. is an indirect measure of shell dissolution and may be the result of ecologic factors like shell initial thickness or different environmental ones (e.g., degree of bioturbation), but in this case, it closely follows the atmospheric *p*CO2, implying an influence of carbonate chemistry. On the other hand, the BDX' assesses specimen dissolution directly, but it is more subjective than F.I. as a proxy. However, if specimens are smaller during the late MIS 6, as suggested by the CT analysis (Sample 865; Table 1), then they would have been less prone to fragmentation [55] even at a higher degree of corrosion, and this could explain the offset in the indices.

Carbonate dissolution in the equatorial Atlantic was traditionally believed to be greater during interglacial periods, mainly because of lower carbonate contents in the glacial sequences that fluctuated in response to Quaternary climatic oscillations [56–59]. However, Broecker, et al. [60] showed, in a core from the north Atlantic, that net carbonate input actually increased in the last glacial, but a proportionately higher detrital terrigenous influx diluted the carbonate, in accordance with later studies [61–63] showing that large glacial–interglacial differences in carbonate concentration are caused mainly by increased dilution by non-carbonate material during low sea levels. On the basis of the multiproxy dissolution assessment and the consistency of foraminifera shell weight, the present analysis shows that biogenic carbonate dissolution has overall been minimal in the eastern tropical north Atlantic, with a slight increase towards the peak interglacial. Slightly increased dissolution of fossil shells is to be expected during interglacials as a result of the biomineralization of dissolution susceptible higher-Mg calcite. Together with findings from the western equatorial Atlantic [63], the present study indicates that the average carbonate productivity in the tropical Atlantic has been constant during the last two climatic cycles. Furthermore, the strong carbonate preservation in the study site suggests that the Cape Verde plateau has remained under the influence of the non-corrosive NADW during the past 200 ky, the carbonate chemistry of which closely followed the atmospheric *p*CO2.

The homogeneity of the *G. bulloides* shell mass record is interrupted by a spike in the mass values during T-II, where weights increase by 30% above average. In order to explain this abnormality, a series of geochemical and tomographic analyses were performed that point to sediment contamination as the cause of the increase of the measured weights. The first elevated masses are recorded at 132.2 ka during MIS 6.1, where high Mg/Ca temperatures are also recorded, and persisted for approximately 2400 years. With the exception of MIS 6.1, when 20.3 ◦C was reconstructed, temperatures in the area are relatively stable around 16–17 ◦C and only slightly reduced during glacial times. For comparison, in the upwelling region of the area, the full range of glacial to interglacial temperature changes amounted to more than 6 ◦C [64], and such variations in SST most probably result from changes in the intensity of coastal upwelling, and in turn the consequence of fluctuations in trade-wind strength [65]. Peak warm stages, accordingly, correspond to a cessation of upwelling and trade winds. Indeed, temperatures of 25 ◦C are characteristic of the subtropical Atlantic outside the upwelling region [66]. T-II is defined by the fairly gradual δ <sup>18</sup>Oshell decrease between substages 6.2 and 5.5, as has been manifested for the study area before [67]. Coupled δ <sup>18</sup>Oshell – Mg/Ca derived salinities and densities reveal that, because temperatures are relatively stable during this time interval, it is salinity that dictates the sea surface density (SSD) and major hydrological changes in the area take place after MIS 6.1, when salinities return to normal marine conditions.

During the late MIS 6, the geochemical data suggest salinities as low as 29 for ~3500 years. The presence of a less-saline water surface layer will most often lead to reduced production, as higher density differences will slow down the upwelling system. T-II contains two major meltwater pulses [68] (MWP-2A and MWP-2B) that, on one hand, can explain low salinities, which are centered on 139 ± 1 and 133 ± 1 ka [69], and coincided within uncertainties with two North Atlantic cooling episodes [70]. MWP-2A indicates an early phase of ice-sheet retreat. MWP-2B is more convincingly resolved and marks a steep ~70 m sea-level rise (~70% of the glacial–interglacial change) at rates of 28 ± 8 m ky−<sup>1</sup> [68,69]. Although small scale perturbations in SSSs and SSDs such as during MIS 5.5 may have been explained by changes in the position of CVFZ and, accordingly, shifts in the influence of NACW and SACW, salinities in the order of 29 indicate the influence of continental discharge, as similar values are found at sites closest to the river mouth in western Africa [71–73]. Fluvial delivery is also suggested by the increase in Ba2<sup>+</sup> core X-ray fluorescence (XRF) data (Figure 5b), which are used to infer freshwater inputs [72,74]. Seawater Ba concentrations at oceanic sites influenced by riverine runoff have a notably high inverse correlation to salinity, because dissolved Ba is high in riverine water [75]. Saharan aquifers were found to be recharged during glacials [76] and such humid periods trigger the reactivation of the Tamanrasset river system in Western Sahara [15], part of which is the Cap Timiris canyon.

Shell weights closely follow SSD changes (*R* <sup>2</sup> = 0.43, *p* < 0.05, *n* = 12) during the time interval analyzed (Figure 5d). The two samples that exhibit abnormally elevated shell masses within the record are also shown to be outliers in Figure 5d plot and, according to the CT analysis performed in one of them, the specimens were four times more contaminated by clay impurities than the rest of the scanned samples. If these two contaminated samples are omitted, the correlation between weighed shell mass and geochemically reconstructed seawater densities becomes more significant (*R* <sup>2</sup> = 0.74, *p* < 0.01, *n* = 10). There is a weaker correlation with salinity (*R* <sup>2</sup> = 0.41, *p* < 0.05, *n* = 10), while there is no significant correlation between shell weights and atmospheric *p*CO<sup>2</sup> (*R* <sup>2</sup> = 0.16 *p* < 0.10, *n* = 12) during this period. Using the shell mass–density regression line to correct the elevated weights, a contamination of 2–3 µg is found. If these corrections are applied then the slope of the regression line (0.38; Figure 5d) matches that of Zarkogiannis, et al. [7] (0.39), which appears to be characteristic for *G. bulloides*. On the basis of this relationship, the low shell weights recorded prior to MIS 6.1 (Figure 4a) can be explained by the freshening of Mauritanian waters. Furthermore, during this time interval, the µCT analysis showed that foraminifera shells have the lowest volume normalized weights and they are smaller, thinner, and of highest porosity (Table 1). The lowest porosities are found close to the peak interglacial, when all other shell characteristics are intermediate. The heaviest, thickest, and densest shells are indeed recorded during the increased shell mass event (132.2 ka), when specimens are also largest, but of intermediate porosity (test density). Increased shell wall thicknesses and shell and

test densities are characteristics of increased biomineralization efforts and play a partial role in the observed increased shell masses.

Shell mass is an easy to measure, central feature of fossil planktonic foraminifera that reflects their physiology and provides clues regarding the ocean carbon cycle. Furthermore, the whole-shell weights of planktonic foraminifera picked from a narrow size range can provide a measure of the extent of surface ocean density changes. The consistency of the different shell mass records from Pliocene core material [11] to Miocene land section samples [77] proves that foraminifera weighing is a robust method for assessing the extent of pelagic biomineralization. However, the present study has shown that the degree of shell cleanliness can greatly affect weight measurements, imposing distortions to paleoceanographic interpretations. Nonetheless, the cleaning procedure remains unstandardized, and different laboratories currently use a variety of treatment methods. It becomes apparent that, in studies of planktonic foraminifera shell weight variations, it is of particular importance to ensure the removal of the detritus trapped within shell chambers and µCT scanning should complement such studies. Except for the degree of contamination, CT provides a wealth of information on the biometry and the preservation of the tests.
