High‐Resolution Coccolithophore Morphological Changes in Response to Orbital Forcings During the Early Oligocene

The global climate of the early Oligocene was characterized by initiated Antarctic glaciation and meridional overturning circulation, which then led to profound eutrophication in the upper ocean. Generating a high‐resolution coccolith record helps to understand the responses of marine phytoplankton to the newly established environment. Using highly resolved (∼6 kyr time‐resolution) marine sediment samples from Deep Sea Drilling Project Site 522 in the South Atlantic Ocean, we conducted a comprehensive morphological study on coccoliths from the genera Reticulofenestra, Dictyococcites, and Coccolithus, which dominated the study interval between ∼33.1 and 32.8 Ma. Our results showed that the size variations of the three measured genera were significantly correlated (p < 0.01) with each other, indicating homogeneous responses to the environmental changes. Moreover, spectrum analysis on integrated morphologic data of all measured coccoliths showed distinct obliquity (∼40‐kyr) and precession (∼23‐kyr and ∼18‐kyr) cycles. We suggest that these variations were mainly driven by temperate, short‐term ecological fluctuations, which periodically altered the nutrient conditions in the common living habitats of the studied coccolithophores. We proposed two tentative explanations focusing on the obliquity signal. First, the cyclic variation could result from obliquity‐modulated changes in ice volume and variations in ocean circulation intensity, which influenced nutrient export from deep waters to the upper ocean. Alternatively, the changes in coccolith size may indicate the strength of seasonality that influenced upper ocean mixing on the west coast of South Africa.

Despite current understandings and applications, constraining the influencing factors on the cellular size of coccolithophores at the geological time scale is still difficult, mainly for three reasons. First, unlike the controlled experiments in culture studies, multiple environmental variables (i.e., temperature, nutrients, and CO 2 concentration) covary in the modern and paleo-oceans. Second, different factors may have differentiated or species-specific effects on the growth of coccolithophores (e.g., Faucher et al., 2020). Third, fossil coccolith records retrieved from ocean sediments often have a low time-resolution. The seasonal growth brings great uncertainty when applying the conclusions of culture studies to fossil studies. Considering these difficulties, some authors attribute ecological changes that alter multiple physicochemical factors together in the upper ocean to explain phytoplankton size variations. For example, some studies have linked coccolithophore size variations with changes in specialized vertical or seasonal niches (e.g., Beaufort & Heussner, 2001;Jin et al., 2019;Lübke et al., 2015;Perrin et al., 2016). On this basis, it has been shown that not only mean sizes but also the size ranges could be indicative of diversified niches in response to climate change (Beaufort et al., 2022;Jin et al., 2022).
Deep-sea circulation that upwells nutrient from deep water to the surface ocean has increased the productivity in the mid-low latitudes since around the late Eocene (Coxall & Wilson, 2011;Norris et al., 2013). Macroevolutionary studies have shown that coccolithophores evolved to adapt to eutrophic conditions during the Eocene-Oligocene climatic transition (EOT) (e.g., Aubry & Bord, 2009;Dunkley Jones et al., 2008;Wei & Wise, 1990). However, it seems that different genera had heterogeneous responses to the EOT. For example, the genus Reticulofenestra reached its highest species diversity during the middle Eocene to early Oligocene, whereas the species diversity of Coccolithus dwindled at the same time (Aubry & Bord, 2009;Bown et al., 2004). In addition to that of long-term evolution, a high-resolution record for the upper ocean dynamics and orbital forcings is, thus far, very limited. Liebrand et al. (2018) showed that the astronomical control of ocean and atmospheric circulation could have induced bloom-like growth of the coccolithophore costal genus Braarudosphaera during reduced amplitude of precession. Except for this, the influence of orbital cycles is rarely observed in Paleogene species. This deficiency is possibly because any ecological factor should conduct its influence on the entire community of coccolithophores rather than a single species. Therefore, a comprehensive study on multiple species at the submillion-year time scale is needed to decipher orbital forcings on the growing conditions of marine phytoplankton during the early Oligocene.
The aim of this study is to investigate the environmental influences on the size of coccolithophores during the early Oligocene. We sampled the time interval between 33.10 and 32.82 Ma from Deep Sea Drilling Project (DSDP) Site 522. Relatively low and constant species diversity allows us to exclude evolutionary bias in the morphological changes (Schmidt et al., 2006). The morphological data of three key genera were analyzed to verify whether different kinds of coccolithophores that coexisted during the study interval had homogeneous responses to environmental changes, and then to investigate the driving mechanism(s).

DSDP Site 522
DSDP Site 522 (26°6.843ʹS, 05°7.748ʹE) is located in the eastern part of the South Atlantic Gyre, in the middle of the Angola Basin, with a modern water depth of 4,441 m ( Figure 1) (Hsü et al., 1984). The paleo-depth of this site is approximately 3,000 m (Tauxe et al., 1983), which is above the temporal paleo-CCD in the South Atlantic at approximately 4,200 m (Dutkiewicz & Müller, 2021). The sediments consist mainly of calcareous nannofossil ooze with a small amount of clay and foraminifera shells (Hsü et al., 1984). Many studies in this region have reported little to no effect of dissolution on carbonate content (Hartl et al., 1995;Liebrand et al., 2018). Benthic foraminifera show no evidence of dissolution at this site (Clark & Wright, 1984), which is consistent with the good coccolith preservation (Wei & Wise, 1990). In our study, 45 samples were taken every 6-10 cm from cores 32-X to 33-X.
The modern mid-low altitude Atlantic is linked with the Southern Ocean by the Atlantic Meridional Overturning Circulation (AMOC). The AMOC is driven primarily by wind-induced upwelling along the northern edge of the Antarctica Circumpolar Current (ACC), which then transports northward and subducts to become Antarctic Intermediate Water (AAIW) (e.g., Jayne & Marotzke, 2001). The AAIW may have prevailed at depths of ∼700-1,000 m during the late Paleogene, which ended the North Atlantic Deep Water (NADW) at the midlatitudinal South Atlantic (Katz et al., 2011). The AAIW ventilation in the early Oligocene could have enhanced the nutrient supply, stimulating productivity in the upper ocean (Corliss et al., 1984;Egan et al., 2013;Sarmiento et al., 2004). Moreover, the study region is near the Benguela Current, which is the eastern part of the South Atlantic Gyre, determining the development of the upwelling system on the west coast of South Africa (Andrews & Hutchings, 1980). In the south, the Agulhas leakage (orange arrows in Figure 1) that was formed by the retroflection off the southern tip of the African Peninsula might have existed in the early Oligocene (Langton et al., 2016). Modeling studies have shown that it is also an important dynamic that affects variability in the AMOC and Benguela Current and thus the upwelling system (Biastoch et al., 2008(Biastoch et al., , 2009. The in situ ecology could be very sensitive to changes in deep-water circulations and ocean currents.

Age Model
We adopted the age model established by Zachos et al. (1996), which suggests continuous sedimentation during the early Oligocene. The DSDP site 522 magnetostratigraphy yields sedimentation rates of ∼0.9 cm/kyr in our study interval Chron C12-C13 (Core 33, Sections 1 and 2). The Chron boundary ages (Ma) used in this study are from Cande and Kent (1992), which are ∼0.3 Ma different compared than those presented in Cande and Kent (1995) or the geological time scale (2004), which were used in recent studies. Nevertheless, the sedimentation rates are approximately the same using either standard geological time scale (∼0.93 and ∼0.94 cm/kyr, respectively). Our study does not include any coccolithophore evolutionary events that need absolute age control. We thus used the original age model from Zachos et al. (1996) for a better comparison between our coccolithophore data and benthic foraminiferal stable oxygen and carbon isotope data. According to the age model, the sampling resolution across the study interval is ∼6 kyr.

Morphological Analysis of Calcareous Nannofossil
Smear slides were prepared for coccolith analysis following the standard technique of using a glass rod and a drop of distilled water to distribute a thin layer of sediment on a glass microscope slide (Bown & Young, 1998). The thickness of the sediments was modified until the edges of each coccolith could be clearly seen under the light microscope. This method has been validated for the purpose of estimating the mean size of coccoliths and species complexes (Henderiks & Törner, 2006). Slides were then analyzed using a Zeiss Axio Scope A1 polarized light microscope with a magnification of 1000X, at Tongji University, Shanghai, China.
We measured multiple species of the genera Reticulofenestra (Hay et al., 1966) (including Cyclicargolithus (Bukry, 1971)), Dictyococcites (Black, 1967), and Coccolithus (Schwarz, 1894). The recognition was based on each of their common structural and morphological characteristics at the generic level, which were clearly distinguishable under the light microscope. All three genera are placolith-shaped (so-called placolith-bearing coccolithophores) with two superposed circular or elliptical shields. The genus Coccolithus is distinguished from Reticulofenestra and Dictyococcites by the weak birefringence in the marginal structure, which results in a dim outer circle under the cross-polarized field of view. The difference in the central area, in addition, separates Reticulofenestra and Dictyococcites. The former usually has a central opening, while the latter usually has a solid plug. Gallagher (1989) suggested merging Dictyococcites into Reticulofenestra, which is currently a widely used convention (Bown, 2005; Nannotax3: www.mikrotax.org/Nannotax3/). Nevertheless, due to the lack of genetic data, taxonomic classification is still under debate. The generic name Dictyococcites has been used by some authors in recent studies (e.g., Cappelli et al., 2019;Raffi et al., 2016). The discussion of the nomenclatures is beyond the scope of our study. Since the size range of Dictyococcites is separated from those of other Reticulofenestra species from our material (see Section 3.2 and 4.1), we use the generic name Dictyococcites to clarify the measured specimens.
In each sample, approximately 10-15 random fields of view were photographed under the microscope in both cross-and plain-polarized light to ensure confident taxonomic identification for morphological measurement. Each field of view typically contained ∼15-20 Reticulofenestra, ∼5 Dictyococcites and ∼5 Coccolithus specimens. The measurements were then conducted manually on the images, using the software Fiji (version 1.52 P) for biometric analysis. This study includes two datasets. For the purpose of studying the size variations in each genus, data set A was obtained by measuring 100 Reticulofenestra, 30 Dictyococcites and 30 Coccolithus specimens in each sample. The number of measurements was determined based on the relative abundances and species diversities of the three genera. On the other hand, to investigate the morphological variation in the entire coccolithophore community, data set B was obtained by measuring 100 randomly chosen placoliths without assigning a generic name to each specimen. Therefore, 260 data points in total were obtained in each sample, and the two sets of data were then compared and analyzed. Moreover, to test the validity and reproducibility of data set B, we randomly selected 100 data points out of 160 from data set A to generate a subdata set C. Correlation analysis was then conducted to compare data sets C and B.

Statistical Analysis
The coccolith MDI from Beaufort et al. (2022) was introduced to describe morphological diversity. It represents the divergence between the two size groups. A confident estimate of MDI may require a large number of coccolith measurements. In the study by Jin et al. (2022), the interquartile range (IQR) has been proven to be valid to describe the stability and diversity of coccolith size when the measuring points are limited by manual measurement (150-250 specimens per sample). Analogously, the standard deviation (σ), which has a similar mathematical meaning to IQR, is used to describe the changes in coccolithophore size structure, especially in data set B.
Correlation analyses were performed on eight independent variables, which included the mean size and standard deviation of all coccoliths (data set B), mean sizes of Reticulofenestra, Coccolithus and Dictyococcites (data set A) and the relative abundances of each of the three genera. The data were z score normalized ( − mean∕ ) before correlation analysis using the software PAST (V4.06, Hammer et al., 2001). At last, a spectrum analysis was performed on the evenly (6-kyr) interpolated mean size data for all placoliths (data set B) using the methods of REDFIT-Rectangle on PAST (V4.06) and Periodogram on software Acycle (v2.3) (Li et al., 2019).

Calcareous Nannoplankton Assemblage
In our study interval, the genus Reticulofenestra is the major component of the nannoplankton assemblage ( Figure 2). Its relative abundance varies from 50% to 80%. The lowest abundance can be found at 32.92 Ma, concurring with the peak of large species of Reticulofenestra (>9 μm). The genus Dictyococcites contributes up to 20% of the total abundance; its highest abundance can be seen at 32.86 Ma and 33.06 Ma. The genus Coccolithus contributes approximately 5%-15%, which is close to the relative abundance of Sphenolithus. In addition, Clausicoccus commonly occurred in the study interval, contributing approximately 2%-10% of the total assemblage, and its lowest abundance appears at around 33.02 Ma. The genus Discoaster is continuously observed in the sediments but with a very low abundance of less than 4%. Finally, the genus Helicosphaera occurred mainly in the lower parts of the study interval with only approximately 1% relative abundance. The abundance and the changing magnitude of other genera, such as Triquetrorhabdulus, Zygrhablithus, and Catinaster, etc., are too small to be important for our assemblage analysis (Figure 2).

Morphology and Species Composition of Measured Coccolithophores
Morphological measurements of all Reticulofenestra, Coccolithus and Dictyococcites coccoliths from data set A are shown in Figure 3. The histogram of all coccoliths (Figure 3a) shows that the size range of Reticulofenestra extends from 4 to 9 μm, centered at ∼6 μm. The rarely occurring large species (>9 μm) cause a long tail in the size structure of Reticulofenestra. Coccolithus and Dictyococcites greatly overlap, ranging from ∼6 to 14 μm, while the latter is slightly larger in size. On the size and circularity (width/length ratio, W/L; Figure 3b) scatter plot, the three measured genera are all similar in circularity, ranging from 0.75 to 1.00 (elliptical-circular). Circular species (circularity >0.98) within the genera Reticulofenestra and Coccolithus could be distinguished, which have a very narrow size range with a mean size of approximately 6.2 μm and approximately 8 μm, respectively. For the largest group, it is difficult to determine whether the few circular specimens are taxonomically different from the others due to the small sampling size.
The length and width of all Reticulofenestra from data set A are shown in Figure 3c. First, two distinct size groups of Reticulofenestra are recognized, in which the larger group (>9 μm) mainly comprises R. umbilicus, R. hillae and large R. dictyoda. The abundance of this group is lower than 5% (usually lower than 2%), which could not influence the mean sizes. In the small-medium-sized groups, the circular (W/L > 0.98) species is mainly Cyclicargolithus floridanus. This species is originally described as a genus, while it is now widely accepted as a species group of Reticulofenestra. Cy. floridanus is abundant (approximately 50%-60%) in the study samples, with a size range restricted to approximately 5-8 μm. The remaining species of this genus mainly include R. dictyoda and R. lockeri. The small species R. minuta are rarely observed in our study material and do not significantly alter the average size of the genus.
The size of Dictyococcites ranges from ∼6 to 16 μm (Figure 3d). It is difficult to apply size or circularity criteria to subdifferentiate this group at the species level and the data only indicate that the species composition of this genus is quite simple (or only consists of morphologically similar species). Most species are likely Dictyococcites Figure 2. Coccolithophores assemblage composition and the benthic foraminifera carbon isotope (LEOSS smoothed) data (Zachos et al., 1996). bisectus. Unlike Dictyococcites, the genus Coccolithus includes several species. The dominant species are C. pelagicus, ranging from ∼6 to 14 μm in size. Those rarely occurred, and specimens larger than 14 μm specimens are identified as C. miopelagicus and C. eopelagicus in this study. In addition, the most circular species C. formosus (LO = 32.9 Ma) also appeared in this study, and its size range overlaps with that of C. pelagicus, but it has higher circularity (>0.98) (Figure 3e). The relative abundance of C. formosus is also lower than 5%. Since we did not compare circularity in this study, its appearance would not cause any impact on the morphological analysis.
Changes in average coccolith size (±1σ) and abundance of the three measured groups are shown in Figure 4. In general, in the ∼300-kyr study interval, the magnitude of the size variations is limited (estimated from data set B, Figure 4a). Size variations of a single genus were calculated from data set A (Figures 4b-4d). For Reticulofenestra, the mean size varies between ∼6 and 7 μm. Between the highest (33.01 Ma) and lowest (32.92 Ma) abundances of this genus, the mean size does not show distinctive variations (Figure 4b). For the genus Coccolithus , the average size varies from ∼10 to 13 μm, and the main part is constrained to ∼9-14 μm (mean ±1σ). The highest abundance occurred at 32.92 Ma and corresponds to a relatively higher average size (∼10.6 μm), but our data do not show a clear covarying pattern between Coccolithus abundances and their morphology (Figure 4c). For example, larger Coccolithus are seen from the lower part of the study interval (∼33.08-33.02 Ma), while its abundances are relatively low (4%-10%). Finally, for Dictyococcites (Figure 4d), the size range is constrained between 8 and 12 μm, which is larger than Reticulofenestra and slightly smaller than Coccolithus. The abundances of Dictyococcites vary from ∼4% to 18% in our study material. The lowest abundances occurred at approximately 32.9 Ma, which corresponds to the interval of increased sizes for the genus.

Correlation and Spectrum Analysis
From data set B, the integrated mean size and their size divergence (σ) show a clear covarying pattern (Figure 4a); a higher mean size always cooccurs with a higher size divergence. Between data sets A and B, the increased mean size and size divergence corresponds to increased average sizes in each genus, such as at ∼32.90 Ma, ∼33.02 Ma and ∼33.07 Ma. In contrast, the correlation between abundance and morphology is weak. The results of spectrum analysis on the mean size of all coccoliths (data set B) show three periodicities of ∼40, ∼23, and ∼18-kyr ( Figure 5), which correspond to the obliquity and precession cycles. Each of them exceeds the 95% confidence interval, although the precession cycle might be artificial considering that our sampling resolution is only ∼6 kyr.
The morphological correlation between different coccolithophore genera was analyzed ( Figure 6, Table 1, and Figure S1 in Supporting Information S1). First, the sizes of Reticulofenestra, Dictyococcites, and Coccolithus are significantly correlated (p < 0.01) ( Figure 6, Table 1). Second, the mean size of each generic group (data set A) is positively correlated with the standard deviation (σ-all/μm) of data set B, indicating that increased size in any of the groups will lead to a higher size divergence for the community (as also seen in Figure 4a). Moreover, the mean size of all coccoliths is also positively correlated with the size of Reticulofenestra and Coccolithus (p < 0.01), while its link to the size of Dictyococcites is weak. Finally, the relative abundances of Dictyococcites and Coccolithus are both negatively correlated with the major composition of Reticulofenestra ( Figure 6, Table 1). However, the size variation is not correlated with the assemblage composition of coccolithophores, as none of the relative abundance data are correlated with the morphological variables from either data set A or data set B ( Figure 6, Table 1). Complete results of correlation analysis between abundances of other unmeasured coccolithophores are not shown, since their abundances are very low. To test the validity of data set B, correlation analysis between data sets C and B is shown in Table  S1 of Supporting Information S1. The results indicate a strong correlation (p < 0.05) between the two data sets, meaning that data set B is reliable and reproducible.

Physiological and Ecological Influences on Coccolithophore Cell Size
Our data show that the cyclic size variation in the entire coccolithophore assemblage is relatively independent of the percentage abundances of different groups or genera (Section 3.2). Correlation analysis suggests that the size changes in different groups of coccolithophores might arise from similar responses to the early Oligocene climatic and paleoceanographic changes.  From the late Eocene to early Oligocene, alkenone-based proxy and boron isotope data showed that atmospheric pCO 2 decreased from over 1,000 to ∼700 ppmv in 2-3 million years (e.g., Pearson et al., 2009;Zhang et al., 2013). Some researchers attribute the dramatic size decrease (several microns) in Reticulofenestra in the early Oligocene (∼32 Ma) to this pCO 2 decline (Henderiks & Pagani, 2008). Bordiga et al. (2015) also suggested that the decrease in the number of large species was driven by the CO 2 decline shortly after the EOT. Both studies argue that cell size is a significant physiological feature as it determines the surface area/volume (SA/V) ratio of unicellular phytoplankton. Theoretically, smaller cells with a higher SA/V ratio should have a higher efficiency in the diffusive uptake of CO 2 (Raven, 1998). Conversely, larger cells with lower uptake efficiency of carbon resources could become an evolutionary disadvantage for phytoplankton transfer from a high to low CO 2 world (Bolton & Stoll, 2013;Bolton et al., 2016;Henderiks & Pagani, 2008). However, a direct comparison between coccolithophore size and CO 2 concentration at the orbital time scale is lacking, as high-resolution proxy reconstruction of atmospheric pCO 2 is currently unavailable. With that potential caveat in mind, we suggest that CO 2 would play a role in long-term trends (Hannisdal et al., 2012) as the total carbon residual time in the ocean (∼100 kyr; Dickens et al., 1995) is much longer than our study time resolution (∼6 kyr).
The mean size and size divergence in coccoliths show cyclic changes ( Figures 5 and 7), which we interpret to be influenced by nutrient availability. Increased sizes and size ranges of coccoliths could indicate increased nutrient concentrations in the upper ocean during the early Oligocene. Our interpretation is supported by the favorable habitat of large Gephyrocapsa, which is one of the descendant genera of the lineage Reticulofenestra and Dictyococcites (Perch-Nilsen, 1971). The large species of the extant genus Gephyrocapsa have been shown to be especially nutrient-dependent in the northern Indian Ocean (Kleijne et al., 1989) and Western Equatorial Pacific (Hagino et al., 2000). In this context, many studies use large Gephyrocapsa species as indicators of eutrophic conditions (e.g., Andruleit & Rogalla, 2002;Andruleit et al., 2003;Bollmann, 1997;Jin et al., 2016).
The size of Coccolithus shows cyclic changes similar to those of Reticulofenestra and Dictyococcites in the early Oligocene (Figures 4 and 6). Little is known about Coccolithus morphological responses to ecological changes, as this genus is now constrained to cold waters (Cachão & Moita, 2000). The phylogenetic distance between the Orders Isochrysidales (Reticulofenestra and Dictyococcites) and Coccolithales (Coccolithus) (Sáez et al., 2004) suggests that there are great differences in physiological features. A culture study indeed showed that Coccolithus is more sensitive to carbonate chemistry (Walker et al., 2018). Fossil records of the Paleocene Eocene Thermal Maximum (PETM) showed that cell division rates and calcite production in Coccolithus were lowered by strong Figure 6. Correlation between coccolithophores morphology and assemblage.
Dashed lines indicate nonsignificant correlation. Solid lines represent significant correlation (p < 0.01). Thickened lines represent |R| > 0.5. Redlines indicate negative correlation (details in Table 1 and Figure S1 in Supporting Information S1). ocean acidification while the genus Toweius (Isochrysidales) was only slightly affected (Gibbs et al., 2013;O'Dea et al., 2014). Nevertheless, our data present evidence of a great correlation in size variation among Coccolithus, Reticulofenestra and Dictyococcites. We suggest that the different responses of the genera to environmental changes may only occur under rapid and drastically changing climates. Under relatively temperate and short-term influences, Coccolithales and Isochrysidales have similar growth rates (Daniels et al., 2014). In a recent study, Faucher et al. (2020) showed that Emiliania, Gephyrocapsa and Coccolithus were larger under P-limited condition than under N-limited condition. Therefore, we infer that the correlated size variation revealed from our data ( Figure 6) was caused by their similar responses to nutrient availability.

Coccolithophore Size Linked to the Upper Ocean Ecology
In the modern ocean, placolith-bearing coccolithophores, together with other phytoplankton, mainly thrive in the chlorophyll maximum zone located right above the nutricline, where balanced nutrient concentration and light intensity provide the most favorable conditions (Perrin et al., 2016;Poulton et al., 2017). Nitrate and phosphate are two major essential nutrient components each with different effects on the growth of phytoplankton and coccolithophores (Geider & La Roche, 2002;Müller et al., 2008Müller et al., , 2012. Phosphorus is required for nucleic acid and phospholipid membrane synthesis, especially during fast cell division in the exponential growth phase. P-limited conditions prevent the transfer from the stationary phase to the fast division phase for coccolithophores, resulting in larger cell sizes. On the other hand, N-limited conditions lead to decreased cell sizes as the main function of nitrate is to form amino acids (Müller et al., 2008). It is possible that the upper ocean is N-limited for the living coccolithophores in oligotrophic gyres (Moutin et al., 2008;Perrin et al., 2016). In this scenario, increased nutrient supply in general can switch the upper ocean from N-limited to nonlimiting conditions for coccolithophores and extend the niche to shallower water where light intensity is higher. The fast expansion (in both size and abundance) of the small-to medium-sized species (often seen as r-selected species) under nutrient-rich conditions could have eventually caused the extinction of large species at the evolutionary scale (often seen as K-selected species) (Imai et al., 2015).
The size variation in coccolithophores indicates cyclic changes in nutrient supply to the upper ocean during the early Oligocene, which could be modulated by either the intensity of large-scale deep-water circulations (e.g., AMOC) or more directly, wind-driven upper water dynamics (e.g., seasonality). With more preformed nutrients in the underlying deep waters (e.g., AAIW), productivity in the surface water is greatly increased by upwelling (Sarmiento et al., 2004). On the other hand, increased wind intensity over the ocean surface enhances mixing, which increases nutrient transport from the deeper ocean (below pycnocline) to the upper ocean (above pycnocline) (Winder & Sommer, 2012). Therefore, the enhanced seasonal difference in wind and mixing would lead to larger year-averaged coccolith mean size and size divergence, as a diversified seasonal ecological niche would be constructed (Beaufort & Heussner, 2001;Beaufort et al., 2011;Hopkins et al., 2021;Okada & McIntyre, 1979;Renaud et al., 2002;Suchéras-Marx et al., 2010;Zarubin et al., 2017). Hence, the increased size and size ranges of the three genera may indicate a widened or optimized ecological niche for the placolith-bearing coccolithophores.
Considering the periodicities revealed from our data, both possibilities imply controls from orbital forcings.

Dynamics of the Upper Ocean Ecology
The ∼40-kyr obliquity cycle in coccolith average size (data set B) is consistent with the obliquity cycle that was previously observed in benthic δ 18 O data (Zachos et al., 1996) (Figure 7). Increased sizes correspond to lighter benthic δ 18 O (lower ice volume) and high obliquity. The ∼40-kyr circle is also very similar to the Reticulofenestra size cycle (∼37-kyr) in the Late Miocene that was revealed by Beaufort (1992) at the same site. Distinct long (∼400-kyr) and short (∼100-kyr) eccentricity cycles revealed from both carbonate content variation and benthic foraminifera isotope data are evidence of Earth's orbital control on the South Atlantic paleoceanography in the early Oligocene (Diester-Haass & Zahn, 1996;Liebrand et al., 2018;Liu et al., 2004;Zachos & Kump, 2005).
The angle of the Earth's tilt modulates the amplitude of seasonal insolation received by high latitudes. High obliquity corresponds to decreased continental ice-sheets because increased summer insolation at high latitudes will cause a higher melting rate of the ice volume (Ruddiman, 2001). In the meantime, enhanced westerlies and the ACC would occur (Pena et al., 2008). In the early Oligocene, deepened ocean gateways led to the formation of the Antarctic ice-sheet and the initiation of the ACC, whose intensity and northward export deeply influenced the upper ocean at low latitudes. A strengthened ACC and upwelling from rebounded Circumpolar deep water (CDW) led to a more vigorous meridional overturning circulation, bringing more nutrients to the mid-low latitude upper ocean (Egan et al., 2013;Sarmiento et al., 2004). Therefore, it is possible that the ∼40-kyr cycle in the mean size of coccolithophores reflects the periodical changes in ACC intensity. Moreover, the AMOC intensity would further affect the dynamics of the Benguela Current and Agulhas leakage, thus having a more direct effect on our study site (DSDP 522). The increased Benguela Current strengthens the upwelling intensity. Previous studies have shown that the Agulhas leakage is obliquity-regulated and could deeply influence the strength of the AMOC by modulating heat and salinity transfer from the Indian Ocean to the South Atlantic (Caley et al., 2011(Caley et al., , 2012Knorr & Lohmann, 2003;Peeters et al., 2004). Based on Late Quaternary studies, these two currents covaried at the glacial-interglacial scale; during interglacial periods, the strengthening and southward shift of the westerlies enhanced the Agulhas Leakage by shifting the Subtropical Front southward, which intensified the Benguela Current (Bard & Rickaby, 2009;Peeters et al., 2004). Conversely, increased Hadley cells and northward migrated westerlies during glacial periods may have blocked or restricted the Agulhas leakage, and then slowed the Benguela Current (Beal et al., 2011;Biastoch et al., 2009;Caley et al., 2011;Petrick et al., 2015). Such mechanisms explain the precession amplitude cycles reflected by the bloom-like growth of the early Oligocene coccolithophores Braarudosphaera (Liebrand et al., 2018). Taken together, our results suggest that the larger size of coccolithophores during high-obliquity is attributed to the increased AMOC and Benguela upwelling. In contrast, smaller size is indicative of weakened AMOC, Benguela upwelling, and nutrient export during low obliquity (Figure 8a).
On the other hand, wind-driven mixing that brings nutrients from the ocean interior to the upper ocean could influence the ecology of the upper ocean, resulting in the seasonal growth of coccolithophores (Beaufort & Heussner, 2001;Beaufort et al., 2022;Hopkins et al., 2021;Jin et al., 2019;Renaud et al., 2002). This is also applicable to the study site because the west coast of South Africa also shows great obliquity-modulated seasonality in the late Quaternary (Chase, 2021;Dickson et al., 2010). According to the simulation of surface air temperature and precipitation, annual wet-dry differences could be amplified by the increased latitudinal temperature gradient in the South Atlantic Ocean during the early Oligocene (Elsworth et al., 2017). In this context, a larger coccolithophore community and increased size divergence would occur if the seasonal niche was enhanced during high obliquity ( Figure 8b). Currently, we cannot exclude any of the possibilities proposed above. If the ACC and ice volume played the determinate role, evidence of the connection between the Southern Ocean and the mid-low latitudes of the South Atlantic upper ocean in the early Oligocene is needed; if seasonality was the direct mechanism, a monsoon proxy, such as pollen records or aeolian dust deposition, will help verify the explanation.

Conclusions
We present a high-resolution (∼6 kyr) and multispecies coccolithophore morphological record obtained from the South Atlantic Ocean. Significant morphologic correlations (p < 0.01) between three major genera, Reticulofenestra, Coccolithus, and Dictyococcites suggest homogeneous responses to short-term ecological changes. In addition, the spectrum analysis showed obliquity (∼40 kyr) and precession (∼23-kyr and ∼18-kyr) cycles in the mean size of all measured coccoliths, which strongly indicate that orbital forcings influenced marine ecosystems during the early Oligocene. Possible physiological and ecological controls on the size of coccolithophores have been discussed, based on which we suggest that the observed size variations resulted from the changes in nutrient availability in the common living habitats of the placolith-bearing coccolithophores. Two mechanisms are proposed focusing on the obliquity cycle. First, during high obliquity, reduced ice-sheets and enhanced ACC could have increased nutrient export from the Southern Ocean, resulting in a larger cell size (and size divergence) of coccolithophores. Alternatively, weakened ACC during periods of low obliquity would cause a decrease in nutrient export and thus a smaller size. Second, the morphological variation may indicate changes in seasonality in the Benguela Current. During periods of high obliquity, enhanced winter monsoon and mixing could have widened the depth habitat for the placolith-bearing coccolithophores, increasing their mean size and size divergence. In contrast, limited living habitats during periods of low obliquity would lead to a smaller mean size and size divergence.

Data Availability Statement
Coccolith morphology and assemblage data used in this study are available at: https://www.ncei.noaa.gov/access/ paleo-search/study/37661. Figure 8. Two proposed mechanisms of the obliquity control on the morphology of coccolithophores. In mechanism 1, intensified ocean circulation during high obliquity is shown in thickened lines (ACC, Antarctica Circumpolar Current; CDW, Circumpolar Deep Water; AAIW, Antarctica Intermediate Water; AL, Agulhas leakage; BU, Benguela Upwelling). The darker green arrow indicates the increased nutrients carried by BU. In mechanism 2, dark blue indicates increasing nutrients and decreasing light intensity through depth. The black arrows represent the increased mixing during high obliquity. The size of coccolithophores is shown in the sketches, but the abundance is not necessarily increased in the high obliquity scenario.