The core-top samples (n=28) employed in this study were retrieved from different depths in the basin of SCS (Fig. 1) during the R/V Sonne cruises (SO-95) (Table 1). Toothpick samples from each location were used to prepare smear slides, without any chemical or physical treatment, following standard micropaleontological procedures (Marsaglia et al., 2015). Unfortunately, the surface sediments were already depleted resulting in not having enough material to perform dissolution experiments using the same samples. For the dissolution experiment, we employed 240 mg of dry sediment obtained from a Late Pleistocene sample from the Western Equatorial Pacific (Ocean Drilling Program – ODP – 807A-2H-2W; 57–59 cm). The distribution of coccolithophore species belonging to the Noëlaerhabdaceae family in the sample ODP 807 is 41 % of Gephyrocapsa oceanica, 34 % of Gephyrocapsa caribbeanica, and 23 % of small Gephyrocapsa. These taxa are thicker, particularly G. caribbeanica, than the thinner Noëlaerhabdaceae species commonly found in the SCS (e.g., E. huxleyi; Roth and Berger, 1975; Roth and Coulbourn, 1982). The sediment sample was suspended in 120 mL Milli-Q water, and then the suspension was separated into six centrifuge tubes, each with a volume of 20 mL and containing the equivalent of 40 mg of sediment. Sodium hexametaphosphate (NaPO3)6 (Calgon^®) has traditionally been used in pretreatment of samples with calcareous microfossils, particularly foraminifera (Olson and Smart, 2004; Smart, 2008). However, it has been observed that application of this chemical agent dissolves these microfossils due to complexation of Ca with phosphates, an effect which varies with the exposure time (Feldmeijer et al., 2013). Therefore, we added 100 mg of Calgon^® into 100 mL Milli-Q water, resulting in a concentration of 1.6 mM, to conduct our dissolution experiment. Different volumes of Calgon^® solution (0, 0.4, 0.8, 2, 4, 6 mL) were added to each of the six subsamples. The Calgon^® is very corrosive to the fine carbonate particles, and the reaction between Calgon^® and carbonate could be simplified in two steps. First, the (NaPO3)6 hydrolysis releases the sodium trimetaphosphate (Na3P3O9). Then, the calcium in the solution is exchanged with sodium and precipitate as Ca(PO3)2, CaNa(PO3)3, and CaNa4(PO3)6, strongly reducing the free calcium concentration in the solution. The decrease in calcium concentration promotes carbonate dissolution. In theory, adding 1 mol (NaPO3)6 would result in the dissolution of 3 mol CaCO3 at maximum. So, there could be ∼ 80 % carbonate left even after adding 6 mL Calgon^® solution. The particles in all tubes were kept suspended gently by a rotating disaggregation wheel as described previously (Stoll and Ziveri, 2002) for 2 d to achieve a full reaction between carbonate and (NaPO3)6. Slides were prepared for coccolith morphological analyses using the drop technique as described by Bordiga et al. (2015) to trace the variations of coccolith amount during dissolution.

The morphological parameters of coccoliths in the dissolution experiment and surface sediment were analyzed using a polarized microscope (Zeiss Axio Scope HAL100), configured with circularly polarized light and a Zeiss Plan-APOCHROMAT 100×/1.4 oil objective, and a coupled camera (Zeiss Axiocam 506 Color). For every sample, at least 40 fields of view were photographed. After species identification and selection of coccolithophores images belonging to the Noëlaerhabdaceae family (Emiliania huxleyi >2 µm, and small Gephyrocapsa spp.), each sample had between 100 and 400 (average of 250 per sample) coccolithophore images. The relationship between the color of coccolith images and thickness was calibrated using a reference calcite wedge, the thickness of which had been carefully quantified (González-Lemos et al., 2018). After calibration, all images were analyzed in the Matlab-based software, C-Calcita (Fuertes et al., 2014), to obtain the coccolith morphological parameters, including length, volume, and mass. The length-shape factor of each coccolith, ks, was calculated using the formula by Young and Ziveri (2000) based on the coccolith mass and length obtained from C-Calcita: ks=mass2.7×length3. Beyond the traditional morphological parameters, we calculated the normalized ks variation, which is the ratio between the standard deviation of ks (σ) over the mean ks (σ/ks). The goal of this novel parameter is to provide a new dimension to trace the dissolution process in coccoliths, especially when the coccolith assemblage is diverse. For example, if the coccoliths dissolve at different speeds in the assemblage due to differential sensitivity to acidification, a small increase of σ/ks would be expected at the beginning of the dissolution because of the σ increase and ks decrease. Then, after all fragile coccoliths dissolve, leaving only thicker coccoliths in the assemblage, the σ/ks should show a decreasing trend which could be mainly caused by a decrease in σ.

Annual means of different physical, chemical, and biological variables in both 50 m depth and bottom water for the location of the surface samples (Table 1) were extracted from different databases, interpolated to the geographical location of the surface sediment samples. Here the 50 m depth was selected because it is the depth at which the highest concentration of Noëlaerhabdaceae coccolithophorid is observed in the SCS (Jin et al., 2016). Seawater temperature, salinity, nitrate, and phosphate concentrations (used as a proxy of phytoplankton) at 50 m were obtained from the World Ocean Atlas 2001 (Fig. 1b, c, d, e; Conkright et al., 2002). Sea surface chlorophyll a concentration data were based on MODIS data (2003–2016) extracted from Oregon State University Ocean Productivity (http://www.science.oregonstate.edu/ocean.productivity/, last access: 16 July 2021). Annual averaged concentrations of total alkalinity (TALK) and dissolved inorganic carbon (TCO2) were extracted from Goyet et al. (2000) (Fig. 1f, g). Then the carbonate ion concentration, pH (Fig. 1h), and pCO2 for the depth of 50 m and ΩCa for the seafloor depth were calculated by CO2SYS macro for Excel^® (Pierrot et al., 2006) (Fig. 1i) using salinity, temperature, pressure, total phosphate, total silicate, TALK, and TCO2 at the corresponding depth (50 m or depth of the surface sediment sample). The light intensity at 50 m water depth was calculated using a model of penetration of photosynthetic active radiation (PAR) from surface to depth (Buiteveld, 1995; Murtugudde et al., 2002), monthly climatologies of PAR from the MODIS ocean database (http://oceancolor.gsfc.nasa.gov/cgi/l3, last access: 16 July 2021), and the diffuse attenuation coefficient for downwelling irradiance at 490 nm (Kd490) and Eq. (1) in Lin et al. (2016).

Pearson correlation and redundancy analysis (RDA) were employed to explore the relationship between morphological features of the coccoliths in surface sediment samples and the environmental data. All statistical analyses were performed using the PAST 4.06 software (Hammer et al., 2001).

In the dissolution experiment, mean ks decreased with increasing volume of Calgon^® solution (Fig. 2a). The mean ks varied between 0.12 (0 mL Calgon^®) and 0.04 (6 mL Calgon^®) (Fig. 2a). The σ/ks represents variation in preservation among coccoliths within each sample. Higher differences in σ were observed in samples containing 2, 4, 0.8, 0.4, 0, and 6 mL, respectively (Fig. 2b). Increasing the amount of Calgon^® solution up to 2 mL showed a decrease in mean ks and an increase in σ. Samples with 4 and 6 mL Calgon^® solution showed a reduction in mean ks and σ among coccoliths (Fig. 2b). The lowest mean ks (0.04) and the maximum mean length (3.95 µm) were recorded under the higher Calgon^® solution (6 mL) amount (Fig. 2c). Increased amounts of Calgon^® solution also resulted in a gradual increase in coccolith length leading to a negative correlation between length and ks (r=-0.68; p>0.05), but this was not significant due to the small number of observations (Fig. 2c).

Overall, the mean ks, thickness, and volume in the core-top sampling stations (Fig. 3) presented higher values in shallower depths. The mean ks varied between 0.03 and 0.07, and the mean thickness was between 0.25 and 0.44 µm, with maximum values recorded at station 17931 located in northern SCS at 1005 m water depth. The mean length of coccolith varied between 3.23 and 3.78 µm, with the highest values recorded at 2195 m water depth (site 17949) in northern SCS but without a significant trend along depths. The mean volume of coccoliths ranged between 1.70 and 2.97 µm3, and the mean mass was between 4.61 and 8.03 pg, with maximum values for both recorded in the shallowest station (e.g., 17930), at 629 m water depth in northern SCS.

In general, the degree of dissolution varied according to the depth of the sediment samples. The calcite saturation, ΩCa, decreases with colder temperature, higher pressure, and higher CO2 concentration in deep ocean. The normalized ks variation presents different trajectories associated with light, strong, or no dissolution (Fig. 4a). The shallowest stations in the south (<15∘ N) and north SCS (>15∘ N) show a linear and increasing trend between ks and σ/ks. For the samples below 2000 m, there is no clear pattern of variation related to the mean ks standard deviation. However, samples below 3000 m are mainly located on the left upper part of the plot, in a similar position as the samples treated with 4 and 6 mL of Calgon^® in the normalized ks variation comparison of the dissolution experiment (Fig. 3b). The mean ks vs. mean length shows a negative correlation (r=-0.62; p<0.05), with the deepest samples showing larger size coccoliths and lower mean ks (Fig. 4b).

We analyzed the correlations between the biological and environmental datasets (Table 2). Although some of the surface variables were autocorrelated (e. g., TALK–salinity, pH–pCO2, and nitrate–phosphate), they were included in our analyses because some studies have identified a strong influence of these parameters on coccolith morphology during the life-cycle (e.g., Chen et al., 2007; Jin et al., 2016). Significant correlations can be found (p<0.05) between several morphological parameters of coccolith and bottom water carbonate chemistry (ΩCa), with a correlation coefficient r = 0.67 between mean ks and ΩCa, r=0.66 between mean volume and ΩCa, and r=0.66 between mass and ΩCa. The mean thickness of the coccolith shows a significant correlation with ΩCa at the sample depth (r=0.41), and with the concentrations of nutrients nitrate and phosphate at 50 m (r=0.44 and 0.4, respectively). Surprisingly, the mean length showed no significant correlation to any environmental variables except with PAR (r=0.35).

The results of RDA can provide another critical perspective on the control of environmental variables on coccolith morphology. The RDA1 and RDA2 explain together 58.3 % of the total variations in coccolith morphological data. The surface sediment samples, color-coded by different depth intervals, are distributed along the axis of RDA1, which is the most important and explains 54.6 % of the total variance (Fig. 5a). Among the environmental variables, ΩCa shows the highest correlation to RDA1 (r=-0.67; p<0.05). The results of both the correlation analyses and the RDA show that ΩCa in bottom water is the most important environmental variable driving the morphological dataset, which shows a high correlation with mean ks (r=0.69; p<0.05) and could explain up to 47 % (R2=0.47) of the variance observed in mean ks (Fig. 5b). The RDA2 explained 3.69 % of the variance and is mainly correlated to the salinity, temperature, pH, phosphate, TALK, and pCO2 (Fig. 5a). The null response of coccolith length to any environmental parameter is also observed in the RDA plot by its position near the center of the ordination space, significantly contrasting with other morphological parameters (Fig. 5a).

In this study, we evaluate fossil coccolith responses to dissolution under laboratory experiments and field settings. In the dissolution experiment, ks values are higher than the modern coccoliths in the SCS due to the higher abundance of the relatively thicker G. caribbeanica in the downcore sediment sample. Though the absolute values of ks cannot be directly compared between the dissolution experiment (Fig. 2b) and surface sediments (Fig. 4a), the trajectory of morphological variations during the dissolution experiment does provide important diagnostic information to explain phenomena observed in the surface sediment samples.

First, the phenomenon that coccolith length increased with the decrease of ks could be observed in both the dissolution experiment (Fig. 2c) and natural surface sediments (Fig. 4b). The laboratory experiment showed that under controlled conditions (known changes in water chemistry and uniform species composition), the coccolith morphology variations (mean length and mean ks) reflected different degrees of dissolution. We also observed a length-related dissolution pattern, where smaller coccoliths gradually dissolve with the increase in Calgon^® concentration, leading to a higher average length but a lower mean ks. The mean ks and mean length relationships in the surface samples (Fig. 4b) show a similar trend to the laboratory observations (Fig. 2c). Thus, the observed trend and the largest size and lowest ks in the surface sediment samples are explained by the dissolution of the smallest species due to the lower ΩCa at the deepest samples and increasing the abundance of the larger coccoliths.

Second, changes in the σ/ks ratio in the dissolution experiment reflect a slight and gradual increase in dissolution and then a decrease with the highest concentrations of Calgon^® (Fig. 2c). In the laboratory experiment, the subsample with no Calgon^® solution presented well-preserved coccoliths with high mean ks and a small standard deviation. As the amount of Calgon^® solution added to each subsample increases, small coccoliths start dissolving preferentially, decreasing the mean ks and increasing the standard deviation (Fig. 2b). With higher amounts of Calgon^® solution (4 and 6 mL), the small coccoliths are completely dissolved, resulting in an assemblage dominated by larger coccoliths (Fig. 2c). Under these highest dissolution stages, the larger coccoliths are also partially dissolved, then both mean ks and σ decrease (Fig. 2b). In this way, the σ reflects how differential dissolution size selection affects the composition of the assemblages. Hence, samples that are more (less) susceptible to dissolution result in more homogeneous (heterogeneous) assemblages regarding carbonate preservation.

However, the trajectory of the normalized ks variation in surface sediments seems more complex than in the dissolution experiment (Fig. 4a). First, there is a group of samples with a positive correlation between σ/ks and ks from shallow areas of the north and south SCS. The depth of samples from northern and southern SCS regions is similar, so we argue that this feature is not caused by dissolution but due to the assemblage composition differences in both parts of the SCS. The coccolithophores have multistage blooms in the north SCS, with a peak of G. oceanica in late winter, when coccolithophore fluxes are highest due to strong water column mixing and renewed nutrient inventory, and another of E. huxleyi in early spring (Chen et al., 2007; Jin et al., 2019). In contrast, E. huxleyi is the dominant species in the more oligotrophic south SCS (Fernando et al., 2007b) due to its higher competitiveness in situations of lower nutrient concentration (particularly nitrate) compared to G. oceanica (Eppley et al., 1969; Rhodes et al., 1995). So, even without any influence from dissolution, the assemblages in the north SCS should feature a higher species diversity and, thereby, a higher σ/ks compared with the coccolith in the south SCS. Hence, the variety of the coccolith assemblages in the surface sediment samples results in different trajectories in the normalized ks variation plotting. But the general trend of the normalized ks variation in surface sediment is still following the trends observed in the dissolution experiment: (1) ks decreases with dissolution, (2) σ/ks increases slightly when dissolution starts, and (3) then it decreases with greater dissolution.

Previous studies have evaluated changes in the calcification of Noëlaerhabdaceae coccoliths in glacial–interglacial cycles through analyses of the coccolith mass and attributed morphological variations mainly to water column nutrient availability and carbonate chemistry parameters, related to the coccolithophores physiological response (e.g., Beaufort et al., 2011). Su et al. (2020) found that the environmental dynamics of the surface photic zone controlled Noëlaerhabdaceae coccoliths' calcification in northern SCS (MD05-2904). Similarly, higher calcite production recorded by increased coccolith mass has been attributed to the increased [CO32-] in the surface water column in the South Indian Ocean and North Atlantic Ocean in modern sediments (Beaufort et al., 2011). Dissolution effects were thought to be less likely drivers of changes in the morphology of coccolith (Beaufort et al., 2011; Su et al., 2020), which is a reasonable assumption for the coccoliths depositing in shallow sediments above the lysocline. These interpretations are partially sustained by the findings of Beaufort et al. (2007), who found no significant coccolith dissolution during the settling in sediment traps deployed between 250 and 2500 m. The former study proposed that most of the dissolution occurs in the euphotic zone and possibly in the guts of grazers, therefore, discarding the impact of bottom water chemistry and/or post-burial processes on coccolithophore weight.

In our set of samples in the SCS, the RDA results show that mean thickness and length significantly correlate to nitrate and phosphate at 50 m (Table 2). This observation agrees with Jin et al. (2016), who found that biometric attributes of E. huxleyi correlated with nutrient concentrations in the plankton samples in the East China Sea (ECS). Nutrient variables are important for coccolithophore calcification (Raven and Crawfurd, 2012) and morphological parameters, at least in species of the Noëlaerhabdaceae family (Båtvik et al., 1997; Paasche, 1998). However, based on the extended evidence of our study, including carbonate chemistry at the depth of the sediment samples in the SCS, we observe evidence that several of the morphological parameters measured are not just influenced by primary biomineralization. Still, abiogenic post- or syn-depositional processes override this signal in the sediment samples in this region. The highest correlations between coccolith morphology, namely mean ks, volume, and thickness, with the bottom water calcite saturation, ΩCa, indicates that the calcium carbonate preservation conditions could strongly override some of the morphological parameters in fossil coccoliths (Table 2, Fig. 5a). We suggest that the mean ks of coccolith could be a potential proxy for the carbonate dissolution in the bottom water, especially in sites near or below the lyscocline.

Carbonate dissolution may also happen within the shallow sediment (Sulpis et al., 2021; Subhas et al., 2022). Based on our current dataset and using only the morphological variations, we cannot distinguish where the dissolution happens at the time of deposition in the sediment water interface, or post-burial in the first centimeters of the seafloor sediment. For the deep ocean deposits with lower sedimentary rates, such as the deepest parts of the SCS (Huang and Wang, 2006), the exposure time of particles to bottom water should be longer than that along the continental slope. Thus, we suggest that the major dissolution in the deep SCS happens on the sediment–water boundary instead of within pore water. Interestingly, the ks of coccolith in the surface sediment of the ECS is much lower, as low as 0.01 (Jin et al., 2019), than those in our study, which is higher than 0.04. However, the ks of coccoliths during the laboratory dissolution experiment performed by Jin et al. (2019; Fig. 9a in that study) shows the same range as our measurements. The ECS samples are from the continental shelf with high sedimentary rates and organic carbon content (Jin et al., 2019). In these settings, the coccoliths continuously dissolve after being buried within the first centimeters of the seafloor sediments in response to organic matter remineralization and CO2 release, resulting in a ∼ 30 %–50 % decrease in coccolith mass (Jin et al., 2019). Therefore, the sedimentary environment has to be individually evaluated to understand which process is controlling the dissolution of coccolithophores at the seafloor. More detailed work, such as in situ pore water chemistry measurements, would be necessary to fully reveal the fate of coccolith dissolution in different burial scenarios (Holcová and Scheiner, 2023).

Among all the morphological parameters, we find the mean ks of coccolith is a more robust dissolution proxy compared to the other measured morphological parameters. Firstly, we observe a higher correlation coefficient between mean ks of coccolith and ΩCa compared with other morphological parameters. Secondly, although volume, mass, and thickness are also highly correlated with ΩCa, these morphological parameters vary more with the feature of different coccolithophores, including variations in coccolith circularity and cell sizes (Young and Ziveri, 2000; Bolton et al., 2016). Thirdly, the thickness is a morphological pattern sensitive to the upper ocean's preservation and surface ocean nutrients conditions during biomineralization (Table 2). Another important feature of ks is its high sensitivity to dissolution. As shown in Fig. 4, the ks of coccoliths have already begun to decrease even though the water depths are only at ∼ 2000 m, which is below the modern lysocline but above the CDD in the SCS (Wang et al., 1995; Luo et al., 2018). Finally, the dissolution effects on morphological attributes of mean ks agree well with the laboratory dissolution experiment, in which each subsample's mean ks reflected different preservation stages (Fig. 4).

Despite a noticeable degree of uncertainty due to the mixing of life cycle and post-mortem signals in the sedimentary record, similar findings of calcite dissolution modifying coccolith's morphology in waters at or below saturation suggest that the conclusions drawn from the present study are not unique to the SCS. In the sub-Antarctic and Antarctic zone, dissolution signals affecting coccolithophores were manifested as a decrease in mass and distal shield length of E. huxleyi coccoliths preserved in surface sediments (Rigual-Hernández et al., 2020b; Vollmar et al., 2022). Based on this collective evidence, a key reasonable question could be can the morphological variation of coccoliths be employed to trace their evolution safely or instead be a good proxy for carbonate preservation?

On longer timescales, the morphological variations of coccoliths have been employed to trace coccolithophores evolutionary trends. Bolton et al. (2016) first measured the ks of Noëlaerhabdaceae in the last 15 million years. They found that the decrease of coccolith ks paralleled the reduction of atmospheric pCO2 since the late Miocene and interpreted this as a decrease in biomineralization. More recent works by Beaufort et al. (2022) and Jin et al. (2022b) focused on the coccolithophore evolution over the last 2 million years by measuring coccolith mass, highlighting the role of seasonality and local environments in the evolution and production of Noëlaerhabdaceae. Similarly, Guitián et al. (2020) studied size trends across different regions between Oligocene to the Early Miocene, concluding that cell size distribution was controlled by multiple competing factors, with a strong selective pressure from CO2 decline as a potential mechanism. This study examined dissolution by looking at, among others, the fragmentation and etching of coccoliths and found that temporal trends in lith size distributions were not significantly affected. This agrees with our observations since the mean length in SCS surface sediments does not correlate with any saturation state-related parameter. However, Guitián et al. (2022), using a new calibration in the C-Calcita software that enables estimations of coccolith thickness up to 3.1 µm, found that elliptical ks (kse) was inversely correlated with the relative abundance of dissolution-resistant nannoliths. This was interpreted as a dissolution control on the elliptical shape factors in coccolithophores between Oligocene and Miocene, as it was found in our surface sediment samples. Therefore, we propose that for studies focusing on coccolithophore evolutionary histories, it would be safer to select a shallow sediment core with low organic carbon content, high clay content, and always lying above the carbonate lysocline (Guitián et al., 2020).

One useful way to identify dissolution in these studies covering geological timescales could be plotting the σ/ks against ks. If an increase of σ/ks is detected in the sediment coccolith without any significant variations in coccolith assemblage or with an increase in dissolution-resistant species (Guitián et al., 2022), it should be interpreted as dissolution. Another way to determine separate evolutionary/ecological influences on ks variations could be to measure the ks of coccolith across a close spatial gradient which includes different depositional depths. Significant variations in the morphological attributes of the fossil coccolithophores would likely be caused by different saturation through time at different sites. Related to this last suggestion, coupling downcore morphological assessment in coccolithophores with other calcareous proxies measurements, such as size-normalized weight of planktic foraminiferal tests (Lohman, 1995; Broecker and Clark, 2001; Barker and Elderfield, 2002), which include recent advances in morphological analyses in large microfossils (Iwasaki et al., 2015, 2019), may provide an even more precise and safe quantitative estimates of past deep-carbonate chemistry parameters.