Particulate inorganic carbon pools by coccolithophores in low-oxygen–low-pH waters off the Southeast Pacific margin

A predicted consequence of ocean acidification is the decrease in coccolithophore-produced particulate inorganic carbon (PIC) pools. PIC is thought to enhance the sinking of particulate organic carbon (POC) to deeper waters, potentially influencing the depth of organic matter remineralization and subsurface O₂ levels. To explore these potential feedbacks, we examined the relationships between PIC, coccolithophores, carbonate chemistry, and dissolved O₂ in the Southeast Pacific open-ocean oxygen minimum zone – a region characterized by naturally low dissolved O₂, low pH, and high pCO₂ levels. Measurements of PIC and coccolithophore abundance from late spring 2015 and mid-summer 2018 revealed that coccolithophores, particularly Gephyrocapsa (Emiliania) huxleyi, were major contributors to PIC through the shedding of coccoliths. On average, about half of the PIC was attributed to reliably enumerated coccospheres and detached coccoliths, with significantly diminished pools below the euphotic zone. Temperature, O₂, and pH emerged as key factors associated with PIC variability. PIC pools and PIC : POC ratios in both surface and subsurface waters in this naturally low-pH–low-O₂ zone are lower than available data from most oceanic regions, with the exception of the Western Arctic. Our findings support the prediction that in upwelling regions with a shallow oxygen minimum zone, POC production is promoted by phytoplankton other than PIC-producing coccolithophores due to the injection of nutrient rich but low-pH water. This process decreases PIC : POC ratios, suggesting that the role of PIC in POC sedimentation might be decreased under such conditions. We emphasize that comparing PIC dynamics across diverse upwelling systems will be valuable for understanding how low-pH and low-O₂ conditions influence POC fluxes mediated by coccolithophores.

1 Introduction

The particulate inorganic carbon (PIC) pool is a key component of marine carbon cycles and atmospheric reservoirs (Ridgwell and Zeebe, 2005). It originates from various sources, including land-derived inputs to coastal margins (Cai, 2011) and biological processes such as phytoplankton calcification (Taylor and Brownlee, 2016). Coccolithophores, particularly the cosmopolitan species Gephyrocapsa (Emiliania) huxleyi, significantly contribute to PIC through coccolith production. Coccoliths are thought to enhance the sinking of organic matter (i.e., the “ballast effect”), facilitating the export of particulate organic carbon (POC) to deeper waters (e.g., Klaas and Archer, 2002), thereby influencing carbon sequestration and nutrient cycling in the global pelagic ocean (Monteiro et al., 2016; Balch, 2018). Changes in PIC dynamics, driven by shifts in coccolithophore communities across spatial and temporal scales, feed back into the ocean–atmosphere system (Balch et al., 2016; Claxton et al., 2022). For example, ocean acidification (OA) – characterized by decreasing pH and carbonate ion concentration – can impair coccolithophore calcification (Riebesell et al., 2000; Barcelos e Ramos et al., 2010; von Dassow et al., 2018; Kottmeier et al., 2022; von Dassow, 2022). This could lead to diminished CaCO₃ fluxes to the seafloor, increasing organic matter respiration and potentially decreasing both organic carbon flux and mid-water O₂ levels (Hofmann and Schellnhuber, 2009; Zhang et al., 2023). In this context, understanding the relationships between coccolithophore PIC, pH/pCO₂, and O₂ in natural oxygen minimum zones (OMZs) – where pH and O₂ levels can reach extremely low levels – is of particular interest.

While PIC measurements are more extensive in the Atlantic Ocean and the Atlantic sector of the Southern Ocean, the Pacific Ocean remains underrepresented in PIC and coccolithophore data. Nevertheless, research on coccolithophore distributions is increasing in the Indian Ocean, the subpolar Pacific, and the Southern Ocean (Balch et al., 2018; Oliver et al., 2023). A particularly notable basin-scale coccolithophore feature, the Great Calcite Belt, encircles the Southern Ocean and is primarily sustained by G. huxleyi growth (Balch et al., 2016; Balch and Mitchell, 2023). This feature, which decreases in the Pacific sector of the Southern Ocean, advances southward each year, peaking in austral summer (Hopkins et al., 2019). Temperature, competition with diatoms for nutrients, and Fe availability have been proposed as factors controlling coccolithophore distributions in the Southern Ocean (Oliver et al., 2023, and references therein). Early satellite data suggested that the Pacific hosts relatively lower surface PIC concentrations compared to the Atlantic (Brown and Yoder, 1994), but this is less clear in more recent analyses (Hopkins et al., 2019), and higher subsurface PIC concentrations and/or higher coccolithophore abundances have been reported in remote Southeast Pacific waters at depths beyond the detection limits of satellites (Beaufort et al., 2008; Oliver et al., 2023).

Vast OMZs are persistent features of the tropical and subtropical Eastern Pacific waters (Schmidtko et al., 2017), where low pH and high pCO₂ levels are common (Torres et al., 2002, 2011; Beaufort et al., 2011; Vargas et al., 2017, 2021). Although such conditions have been shown to inhibit coccolithophore growth and calcification in laboratory settings, field observations suggest more complex relationships (Beaufort et al., 2011; Müller et al., 2015; von Dassow et al., 2018). There is still limited data on how PIC, POC, pH, and O₂ interact within OMZ systems.

Here, we evaluated total PIC and the fraction derived from coccolithophores in OMZ waters, distinguishing pools in the well-lit waters above the oxycline from pools in the subsurface OMZ core below the oxycline. We first analyzed satellite imagery to assess how well the in situ sampling captured the spatial and temporal variability of surface PIC during two cruises. We also present the physicochemical structure of the upper water column from both cruises to evaluate how variations in the depth of the oxycline and carbonate system parameters were reflected in the discrete samples collected for coccolithophore and PIC analysis. Coccolithophore and detached coccolith diversity were analyzed to identify the major species contributing to PIC and to quantify the proportion of total PIC derived from coccolithophores. We further examined how low-pH and low-O₂ conditions may act as barriers to coccolithophore stocks and their associated PIC. Finally, total and coccolithophore-derived PIC were estimated separately for the upper layer, including the euphotic zone, where coccolithophores are presumed to grow, and for the oxygen-deficient core of the OMZ, where coccolithophore-derived PIC is assumed to originate from sinking particles. These values were then compared with available data from the same depths in other oceanic regions, obtained using comparable methodologies.

2 Materials and methods

2.1 Sampling

The sampling consisted of two cruises conducted in waters off the Southeast Pacific margin near northern Chile (Fig. 1a). In spring 2015 (LowpHOx 1), we sampled six stations along an inshore–offshore transect off Iquique (20° S; sts. T1–T6), along with six stations arranged latitudinally between 22 and 29.5° S (sts. L1–L6). During summer 2018 (LowpHOx 2), three of the inshore–offshore stations were replicated (sts. T1, T3, T5), along with seven stations sampled southward down to ∼34° S (sts. Lander and Hydro; Fig. 1a). In depth, the sampling crossed both the euphotic zone and OMZ core (Figs. 1b, 3c–e, 5a–c). The predominant oceanographic conditions during these two cruises, some reported previously by Vargas et al. (2021), provide the context for the new data presented here.

At each station, discrete seawater samples were collected, filtered, stored, and moved to the lab for the determination of PIC (Sect. 2.2) and coccolithophore standing stocks (Sect. 2.3). Due to seawater availability onboard in 2018, sampling for coccolithophores only reached 100 m, except for station T5, which reached a depth of 2000 m. This contrasts with the 2015 sampling reaching depths equal to or greater than 350 m (max. 1000 m in T3 and T4). As ancillary data, 1 m averaged temperature, salinity, O₂, and fluorescence continuous profiles, as well as discrete data on Chlorophyll a (Chl a), nutrients, POC, and carbonate system (Vargas et al., 2021) were extracted from Vargas et al. (2023a, b, c). The euphotic depth (Z_eu) was estimated by calculating the depth at which the photosynthetically active radiation (PAR) is reduced to 1 % of surface levels, using the attenuation coefficients (K_d490) provided by the Bio-Geo-Chemical products based on the Copernicus-GlobColour processor (Copernicus-GlobColour, 2023). The averaged K_d(490) values (3×3 pixels and 3 d average) were converted into K_d(PAR) values following the model of Morel et al. (2007), and the 4.6 optical depth was calculated as ln(100) $/$ K_d(PAR) (Morel, 1988). Where available, a good fit was found between the euphotic zone depth derived from satellite observations and PAR obtained from a sensor attached to the CTD (Fig. S1 in the Supplement). Lastly, the OMZ core was defined as the water layer where dissolved O₂ was below 20 µmol kg⁻¹, a commonly used threshold (e.g., Gilly et al., 2013), which we also use here to approximate the depth of the base of the upper oxycline.

2.2 Particulate inorganic carbon standing stocks

Monthly and weekly satellite-derived PIC products (November–December 2015 and January–February 2018) were obtained from the MODIS-Aqua mission (NASA Ocean Biology Processing Group, 2023; Balch and Mitchell, 2023). The data were then converted from mol CaCO₃ m⁻³ to µg C L⁻¹ by multiplying by 11 910.69 and plotted in R (R Core Team, 2024) using the RStudio IDE (Posit Software, PBC, 2024).

For in situ measurements of total PIC (PIC_Total), we used a modified version of the procedure introduced by Poulton et al. (2006). In summary, we filtered between 0.1–1.5 L of seawater (increasing with sampled depth) onto 25 mm polycarbonate filters with a 0.4 µm pore size. Prior to removal from vacuum, filters were immediately rinsed with a squirt of potassium tetraborate solution in 2015 or with seawater alkalinized with ammonium in 2018 (to maintain pH >8.0 while in storage) and stored in metal-free Falcon tubes at −20 °C until transit to the lab. Subsequently, Ca²⁺ was extracted with nitric acid and quantified using Inductively Coupled Plasma Atomic Mass Spectrophotometry facilities at the Bigelow Laboratory for Ocean Sciences. A correction was made to correct for potential excess of Ca²⁺ due to Na⁺ residues that might be left on the filters. This calculation indicated that residual seawater contributed on average 29.1±25.8 % of total Ca²⁺ in LowpHOx 1 samples and on average 35.2±21.5 % of total Ca²⁺ in LowpHOx 2 samples. The PIC concentrations were expressed in µg C L⁻¹. Additionally, we examined the association between calcification (PIC) and POC using PIC : POC ratios. These PIC, POC, and PIC : POC ratios were then categorized into two groups: above the oxycline and within the OMZ core.

Data from this study were compared with those reported for other regions (see Balch et al., 2018). PIC and POC data came from the SEABASS (Werdell et al., 2003) and BCO-DMO repositories (Balch, 2010). Depth intervals were chosen to detect broader ecological patterns robustly. We aimed to be roughly comparable with the categories above the oxycline (mostly corresponding to the Z_eu, where coccolithophores are growing) and below the oxycline, where coccolithophores are presumed present entirely due to sinking from the surface. However, OMZ systems are highly stratified, and eukaryotic phytoplankton growth is excluded from below the lower half of the oxycline (which also corresponds to a strong pycnocline) even when sufficient light penetrates for photosynthesis (Wong et al., 2023). In contrast, in non-OMZ systems, the lower limits of growing phytoplankton are less constrained, and coccolithophores are part of the “shade flora” (Balch, 2018). Therefore, data were binned over 0–100 m depths to represent the surface, and over 100–400 depths to represent the subsurface.

Figure 1 (a) Map of the Southeast Pacific margin showing the study site and stations sampled during late spring 2015 (circles) and mid-summer 2018 (crosses). (b) Sampling depth coverage for coccolithophores, highlighting areas crossing the OMZ core thresholds of 20 µmol kg⁻¹ (dotted and dashed red lines). Map produced by Ocean Data View (Schlitzer, 2024), with bathymetry based on the GEBCO chart (GEBCO, 2023).

2.3 Coccospheres and detached coccoliths standing stocks

For enumeration of coccospheres and detached coccoliths, between 0.1 to 1.0 L of seawater (increasing with depth) was filtered onto 25 mm 0.8 µm pore polycarbonate filters, dried in Petri dishes, and stored with desiccant until microscopy analyses. Coccosphere counts were conducted on filter slide preparations with oil immersion, using cross-polarized light microscopy (LM; Zeiss, Axioscope 5), analyzing 20 fields of view at 400× magnification covering 5.1 mm² of the filter area (corresponding to 1.9 to 16.3 mL of seawater). For counts of total detached coccoliths, 11 fields of view per filter were screened (224×165 µm per frame) at 630× magnification (oil immersion objective), covering 0.41 mm² of the filter area, corresponding to total volumes of 0.2–1.3 mL of seawater analyzed. On average, 171±307 coccospheres (range: 2–2099) and 708±664 coccoliths (range: 5–4067) were analyzed per sample. An issue arose where some filters from the inshore–offshore 2015 sampling (20° S; Stations T1–T6) exhibited excessive brightness under LM. The coccosphere and detached coccolith counts were made by scanning electron microscopy (SEM) analysis (Quanta FEG 250), as described in Díaz-Rosas et al. (2021). For the quantification of coccosphere abundances by SEM, between 28–48 images taken at 800–1500× magnification were examined per filter, covering from 5 to 6 mm² of the filter area and corresponding to a range of 2.1–18.4 mL of seawater analyzed. For total detached coccolith abundances, between 4–5 images were examined per filter, covering from 0.6 to 1.0 mm² of the filter area and corresponding to a range of 0.2–2.8 mL of seawater analyzed. On average, 166±280 coccospheres (range: 1–1141) and 861±879 detached coccoliths (range: 58–3815) were counted per sample. Layers of coccoliths detached from G. huxleyi (Fig. S2a–d) were added to the detached-coccolith counts. Collapsed coccospheres were included when they remained mostly intact, but when more disintegrated could not be accurately counted, especially as they were often less reflective than intact coccospheres and coccoliths (Fig. S2e–h). In a subset of samples, collapsed coccospheres were estimated to contribute <21 % (min. = 0 %, average = 7.1 %) of the total number of coccospheres. As expected, the standard error of the means (among the images obtained from the same sample) drops hyperbolically with the total number of coccospheres or coccoliths counted, whether with LM or SEM (Fig. S3), but remains higher in SEM due to the smaller size of SEM images. To check for differences between coccolithophore counts obtained through LM and SEM, five samples with varying abundances were analyzed using SEM as outlined above. The results showed highly linear relationships (R²>0.9), with slopes not significantly different from 1 and intercepts close to 0 (Fig. S4), supporting the combination of counts from both methods.

2.4 Diversity of coccospheres and detached coccoliths

Identification of coccolithophores and detached coccoliths by LM is limited, which may impact the estimation of coccolithophore-derived PIC relative to total PIC, as it relies on the estimated PIC quotas of coccospheres and coccoliths. To understand this effect, taxonomic classification by SEM was performed on samples from the LowpHOx 1 cruise (2015), focusing on samples from T1 to T6 as well as selected samples from L1 (at 5 and 25 m), L2 (at 5 and 50 m), and L3 (at 5 m). Between 6–11 and 4–5 images per filter, depending on magnification (ranging from 800× to 1500×), were examined for coccosphere and coccolith classification, respectively, following classifications of Young et al. (2003). Due to limitations in SEM time, it was not always possible to zoom to higher magnification to differentiate between G. parvula and G. ericsonii coccospheres, so they were merged into the G. parvula/ericsonii category. The only character distinguishing their coccoliths is a small bridge present in G. ericsonii and not G. parvula (representing a minor effect on PIC quotas). Also, the grouping of the two species is phylogenetically supported (Bendif et al., 2016, 2019). Moreover, as some of the small coccoliths detached from G. parvula/ericsonii might be overlooked (2.0 µm in length), the few coccoliths found in distal-shield view, and the coccoliths from G. huxleyi (3.6 µm in length) were merged into the Gephyrocapsa <4 µm category. Larger coccoliths (>5 µm in length) were classified into species or genus, being subtracted from total counts of coccoliths obtained in the same SEM image to obtain the number of coccoliths <4 µm. Most rare detached coccoliths were grouped in the miscellaneous category, that includes Syracosphaera spp., Acanthoica spp., Discosphaera tubifera, Umbellosphaera spp., and Umbilicosphaera spp. Moreover, the counts of detached coccoliths from Calcidiscus leptoporus might contain a few coccoliths of Oolithotus spp. as it was not always possible to differentiate them. In all surface samples taken in 2018, as well as in samples L4 to L6 from 2015, species dominating the coccosphere and detached coccolith pools were identified using LM (40× and 63× objectives, Zeiss Axioscope 5) following Frada et al. (2010).

2.5 Coccolithophore specific PIC quotas and estimation of coccolithophore-derived PIC

The PIC quotas of individual coccoliths and coccospheres were estimated following Young and Ziveri (2000) and used to calculate coccolithophore-derived PIC stocks (PIC_Cocco) from abundances of coccospheres and detached coccoliths. To obtain the coccolith mass (in pg CaCO₃), the cube of the mean distal-shield length of coccoliths (= biovolume in µm³) measured in the SEM images (see Table S1 in the Supplement) was multiplied by the respective taxonomic-specific shape factor K_s (Young and Ziveri, 2000) and by the density of calcite (2.7 g cm⁻³). The number of coccospheres was converted to coccoliths by using the compilation of coccoliths per coccosphere in Yang and Wei (2003). For G. huxleyi, the 17 coccoliths per coccosphere determined by Beaufort et al. (2011) was in agreement with those seen in our SEM images. We did not account for variation in coccoliths per coccosphere or in mass of coccoliths among G. huxleyi morphotypes. The PIC_Cocco standing stock (in µg C L⁻¹) included both coccospheres and detached coccoliths. Given the lack of taxonomic resolution for the 2018 samples, the G. huxleyi PIC quota (mass = 2.5 pg CaCO₃ per coccolith; see Table S1) was applied as the maximum threshold to all samples for consistency, a scenario justified as G. huxleyi was the most common coccolithophore in both the 2015 and 2018 samples (see below). Notably, PIC_Cocco quotas estimated using G. huxleyi as a uniform reference closely aligned with those derived from all-taxa coccolith mass factors across the 2015 T1–T6 transect (Fig. S5).

2.6 Statistical analyses

Rarefaction analyses using the iNEXT package (Hsieh et al., 2024) assessed sampling effort effects on diversity, based on the species richness and the exponential of Shannon entropy indexes.

Associations between selected pairs of variables across cruises were evaluated using simple linear regression. Key assumptions (linearity, homoscedasticity, normality, and independence) were assessed by visual inspection of scatterplots and residual plots. Trend lines were included only for relationships showing reasonable explanatory power (R²>0.5 or p value <0.05). All analyses were conducted in R (R Core Team, 2024) using the RStudio IDE (Posit Software, PBC, 2024).

To assess differences in PIC, POC, and PIC : POC ratios among the OMZ and other regions, we tested the assumptions of normality and homogeneity of variances. These assumptions were not met despite common data transformations. Therefore, we applied a robust ANOVA using 20 % trimmed means (Yuen, 1974; Kitchenham et al., 2017), implemented with the yuen function from the WRS2 package (Mair and Wilcox, 2020). Pairwise comparisons were conducted using a robust Tukey-type test (mcppb20), and results were summarized by letter groupings above each boxplot, where groups sharing a letter are not significantly different (p<0.05). A full table of all pairwise comparisons is provided in Tables S2–S8. Additionally, to test for differences between the two OMZ layers, we used a robust two-sample test with 10 % trimmed means (yuenbt), which provides bootstrapped p values and confidence intervals while reducing the influence of outliers. Letters indicating the results of this two-group comparison are shown above the corresponding boxplots.

3 Results

3.1 Oceanographic context

3.1.1 Spatial variability in PIC indicated by satellite

During both sampling periods (Fig. 2), satellite-derived surface PIC concentrations exhibited spatial variability, with peaks (>10 µg C L⁻¹) near 20° S at the coast. The high surface PIC extended further south in 2018, with prominent jets from the coast at approximately 21 and 23° S. In 2015, a wide band of moderately high PIC also occurred south of approximately 23° S and extended to 32° S. In situ sampling included most of the range of variability in surface PIC indicated by satellite.

Figure 2 Monthly (a) and weekly (b–d) satellite PIC during the LowpHOx 1 sampling (27–28 November/5–9 December 2015; open circles), along with monthly (e) and weekly (f–h) satellite PIC for the LowpHOx 2 sampling (30 January/3–9 February/12–13 February 2018; open circles).

3.1.2 Oceanographic context in situ

Surface (mixed) layer temperatures exceeded 21 °C to the north and offshore, while surface temperatures were below 17 °C near the coast, in the north and south of 21 and 24° S latitude in 2015, and south of 29° S in 2018 (Fig. 3a). The pycnocline (Fig. 3b) and oxycline (Fig. 3c) roughly paralleled the thermocline in both years. In 2015, the oxycline was always in the upper 50 m. In 2018, the depth of the oxycline increased to between 50 and 75 m south of 26° S. The estimated Z_eu was often near the base of the oxycline, but these sometimes crossed (Fig. 3c). The highest phytoplankton biomass (extracted chlorophyll and calibrated fluorescence values reaching or exceeding 3 mg m⁻³) always occurred shallower than Z_eu, though in some stations a defined primary chlorophyll maximum at the base of Z_eu was not present and fluorescence was highest near the surface, not near Z_eu (Fig. 3d–e).

Nitrate and phosphate levels were low above the pycnocline (<2 and <1 µM, respectively), and increased below (to approximately 20 and 2.5 µM respectively; Fig. 3f–g). Both pH and Ω_calcite declined sharply while pCO₂ increased with depth through the oxycline (Fig. 3h–j). Within the upper 100 m, pH varied negatively with pCO₂ and also positively with O₂, while nitrate varied positively with pCO₂ and negatively with pH and O₂, though the scatter was somewhat higher for the relationships pH-O₂ and nitrate-pCO₂ (Fig. S6). In general, O₂ was more restricted by depth compared to nitrate and pCO₂ (Fig. S6). Ω_calcite was always greater than 1 in the upper 100 m (Fig. S6). In fact, values of Ω_calcite below 1 were never observed – even down to 1000 m (Fig. 3k).

Figure 3 Spatial variation in temperature (a), neutral density (b), O₂ (c), fluorescence (d), Chl a (e), nitrate (f), phosphate (g), pH (h), Ω_calcite (i), and pCO₂ (j) within the upper 100 m, along with vertical distribution of Ω_calcite down to 1000 m depth (k) during 2015 and 2018. In panels (e)–(j) sample locations are indicated by solid black circles. Continuous profiles in panels (a)–(d) are 1 m binned. Dashed white lines in panels (c)–(e) indicate the estimated euphotic depth. The red line in panel (k) represents a power trend curve.

3.2 Coccolithophore species contributing to PIC

In 2015, G. huxleyi was the numerically dominant coccolithophore south of 20° S (Fig. S7). It co-dominated with G. parvula/ericsonii, the coccosphere standing stocks at 20° S, where there was no consistent variation in the relative abundance of these taxa (Fig. 4a). Despite the presence of G. parvula/ericsonii, its small coccoliths were underrepresented, reinforcing the overwhelming prevalence of G. huxleyi in the samples (Fig. S8a). The low diversity of both coccospheres and detached coccoliths is supported by the rapid saturation of rarefaction curves from samples at different stations, grouped by depth layers (Fig. S9). The relative abundance of larger taxa (principally G. oceanica in stations T1–T2, Helicosphaera spp. in T3, and C. leptoporus in T4) increased below the Z_eu, reaching 0 %–18 % and 0 %–15 %, respectively (Fig. 4), as total coccosphere and coccolith abundances declined sharply.

Only polarized LM data was available in 2018 (Fig. S10), and moderately sized Gephyrocapsa spp. dominated coccolithophore communities in all 2018 samples, as in 2015. Light microscopy cannot reliably distinguish G. huxleyi from G. parvula/ericsonii or G. muellerae (the bridge in G. ericsonii is too small to be reliably observed under LM and in both G. muellerae and G. ericsonii, the bridge may be absent in some coccoliths) but qualitative LM observations suggested that coccospheres and coccoliths smaller than G. huxleyi were very rare.

Above the Z_eu, the coccospheres and detached coccoliths of G. huxleyi and G. ericsonii/parvula dominated the PIC_Cocco pools (Fig. 4b–c). Coccosphere abundances reached maxima of 3.9×10⁵ and 5.2×10⁵ L⁻¹, and abundances of detached coccoliths reached maxima of 63×10⁵ and 49×10⁵ L⁻¹ in spring 2015 and summer 2018, respectively (Figs. S11–S12). Overall, coccosphere abundance tended to vary directly with detached coccolith abundance, although with a high scatter (Fig. S13a). Above the Z_eu, the average ratio of coccoliths to coccospheres was 40 in 2015 (range: 4–104; n=21) and 33 in 2018 (range: 6–62; n=31) (Fig. S13b). Only one sample had a ratio higher than 100. The ratio of detached coccoliths to coccospheres was not related to either coccosphere abundance or PIC_Cocco; notably, the three samples with the highest values for these variables all showed ratios below 20 (Fig. S13c–d).

In contrast, detached coccoliths of numerically rarer taxa, such as C. leptoporus, Helicosphaera spp., and G. oceanica, contributed significantly to PIC_Cocco below the Z_eu, sometimes exceeding the contributions of the smaller Gephyrocapsa species (Fig. 4c). However, because most samples were dominated by G. huxleyi, PIC_Cocco estimations using taxa-specific conversions generally showed little differences with PIC_Cocco estimations made with the assumption that all coccospheres and coccoliths had PIC quotas similar to G. huxleyi (Fig. S5a), so this simplification was made in estimating PIC_Cocco pools over the whole dataset.

Figure 4 Relative abundance of coccospheres (a) and estimated PIC masses from coccospheres (b) and detached coccoliths (c) off Iquique (20° S) in 2015. (b–c) Taxonomic breakdown of the relative contributions of coccospheres and detached coccoliths to PIC_Cocco quotas, expressed as percentages of the total PIC_Cocco pool. The PIC_Cocco pool estimate is based on abundances obtained via scanning electron microscopy.

3.3 Variability in PIC, POC, and PIC : POC ratios and the contribution of coccolithophores to PIC

In both years, PIC_Total and PIC_Cocco were the highest off 20° S, extending further from the coast in 2018 (Fig. 5a–b), similar to what was observed by satellite (Fig. 2) and coinciding with patches high in POC. Some of the patches or bands of moderately high surface PIC observed by satellite between 26 and 32° S (Fig. 2a–d) were also reflected in surface in situ PIC_Total and PIC_Cocco. A patch of high PIC in 2018 near 32° S, as indicated by satellite, was likewise reflected in surface in situ PIC_Total and PIC_Cocco. However, deeper euphotic peaks in coccoliths or PIC_Total, such as those observed in 2018 near 23–25° S (Fig. 5b), were not detected by satellite. Considering all stations and depths, PIC_Total ranged from 0.18 to 3.81 µg C L⁻¹ in 2015 and from 0.08 to 5.86 µg C L⁻¹ in 2018 (Fig. 5a). The estimated PIC pools produced by coccolithophores ranged from 0.02 to 3.60 µg C L⁻¹ in 2015 and from 0.01 to 3.69 µg C L⁻¹ in 2018 (Fig. 5b).

PIC_Cocco was well correlated with PIC_Total (Fig. 5d), with the 1:1 line representing an upper limit. Almost half of PIC_Total in the upper 100 m was directly accounted for by coccospheres or detached coccoliths (Fig. 5e). Below this depth, the proportion dropped substantially, with 70 % of PIC_Total below 100 m not accounted for by microscopic counts; where most PIC_Cocco was attributed to detached coccoliths (Figs. 5e, S8b).

Figure 5 Spatial variation in PIC_Total (a), PIC_Cocco (b), and POC (c) within the upper 100 m during 2015 and 2018, along with the relationship between PIC_Total and PIC_Cocco (d), and the bulk proportion of PIC_Total accounted for by PIC_Cocco (e) above and below 100 m depth. In panels (a)–(c) dashed white lines indicate the estimated euphotic depth and sample locations are indicated by solid black circles. In panel (d) the solid blue line represents the linear trend, with gray shading indicating 95 % confidence intervals. The dashed gray line represents the 1:1 relationship. Log-log axes highlight the performance of PIC_Cocco estimates over ∼2 orders of magnitude of PIC_Total concentrations. Deeper profiles of PIC_Total and PIC_Cocco are shown in Figs. S11–S12.

The PIC_Total and PIC_Cocco stocks were not statistically different (p<0.05) between the 2015 and 2018 cruises (result not shown), although mid-summer 2018 exhibited higher average and maximal values compared to late spring 2015 (Fig. 6a–b). On average, a unimodal thermal response peaking at ∼18 °C was observed for both PIC_Cocco and PIC_Total for depths above Z_eu (Figs. S16a, S17a). Low-to-moderate Chl a levels (<4 mg m⁻³) were associated with enhanced PIC (>2 µg C L⁻¹ PIC_Cocco, >3 µg C L⁻¹ PIC_Total), with PIC_Cocco returned to low levels at the highest Chl a concentrations (Figs. S16d, S17d). Overall, the average PIC_Total and PIC_Cocco pools below the Z_eu decreased by 50 % and 83 %, respectively (Figs. 6a–b, S11–S12).

Because Z_eu was sometimes below and sometimes within the oxycline, we separately analyzed the role of the O₂ and pH, consolidating data from both sampled years. In addition to the restriction of Z_eu, PIC was restricted by pH >7.7 and O₂ >60 µmol kg⁻¹ (Figs. 6c, S16c, g, S17c, g). One point within Z_eu showed a pH below 7.7 and another showed O₂ concentrations below 60 µmol kg⁻¹; both had low PIC. PIC_Cocco was higher at pH ∼7.9 (Fig. S16g). As a result, most PIC_Total was concentrated above the OMZ core (below the oxycline), at depths generally suitable for coccolithophore growth (Fig. 6d). Across all samples, POC ranged from 22.7 to 250.0 µg C L⁻¹ in 2015 and from 26.1 to 501.6 µg C L⁻¹ in 2018 (Fig. 5c). POC was closely correlated with Chl a (Fig. S14a). PIC_Total also correlated with POC (Fig. S14b), although POC concentrations were approximately two orders of magnitude higher. PIC : POC ratios ranged from 0.001 to 0.048 above 100 m depth (mean: 0.011±0.009) and from 0.002 to 0.054 below 100 m (mean: ±0.009) (Fig. S15). Within the OMZ core, both PIC and POC declined significantly, however, PIC : POC ratios were higher below the oxycline (Fig. 6d–f).

Figure 6 Violin plots showing the variation in PIC_Total (a) and PIC_Cocco (b), as well as PIC_Cocco in relation to O₂ and pH levels (c), above and below the euphotic depth in 2015 and 2018. Panels (d)–(f) display the values of PIC, POC, and PIC : POC ratios above and within the OMZ core during the same years. Only samples from the upper 100 m are included in panels (a)–(c). In panels (a)–(b), the number of samples per category is indicated in parentheses. In panel (c), horizontal and vertical white lines represent the projected pH for the year 2100 and the O₂ threshold for the Eastern Equatorial Pacific Ocean proposed by Stramma et al. (2008), respectively; the gray line indicates the least-squares model fit. In panels (d)–(f), all raw data are shown, with group comparisons based on robust two-sample bootstrapped tests (see Sect. 2.6).

3.4 Global comparison of SE Pacific PIC, POC, and PIC : POC to other ocean regions

PIC_Cocco originating from below Z_eu contributed minimally to PIC_Total above 100 m (Fig. 6c). This justified a simplified approach for global comparisons by binning all water columns into a 0–100 m interval, to correspond roughly to the Z_eu, and a 100–400 m interval to be comparable to the most oxygen-deficient core of the OMZ. The PIC values from the SE Pacific OMZ reported here overlapped with the lower end of values reported for other regions in both depth intervals; however, all other regions exhibited higher maximum values (Table 1). Within each depth interval, PIC values in the OMZ were significantly lower than in all other regions compared, except for the Western Arctic (Fig. 7a–b, e). In contrast, POC concentrations in the OMZ were comparable to or exceeded those in other regions across both depth intervals (Fig. 7c, f). In the 0–100 m interval, the PIC : POC ratio in the OMZ was significantly lower than in all regions except the Western Arctic (Fig. 7d). For the 100–400 m interval, the PIC : POC ratio was significantly lower than in all other regions, with mean values differing by nearly an order of magnitude compared to those in the Southern Ocean and Indian Ocean (Fig. 7g).

Table 1 Comparison of PIC standing stocks reported in surface layers (0–100 m depth) of different oceanic areas or shelf/coastal margins. Unless otherwise indicated, the PIC was measured chemically using mass spectrophotometry. AMT: Atlantic Meridional Transect; SP: spring; AU: autumn; WI: winter; SU: summer.

^a Outliers values were not included. ^b Extracted manually from Fig. 4c. ^c Extracted manually from Fig. 3b. ^d Estimated from underway “acid-labile backscattering” measurements calibrated with PIC measurements obtained with the same methodology outlined above. ^e Restricted to the uppermost surface waters. 1: Poulton et al. (2006); 2: Holligan et al. (2010); 3: Balch et al. (2018); 4: Daniels et al. (2012); 5: Painter et al. (2016); 6: Matson et al. (2019); 7: Oliver et al. (2023).

Figure 7 (a) Global map showing the annual O₂ concentrations at 100 m depth and the sampling locations for the well-mixed surface layer (0–100 m; panels b–d) and the stable subsurface layer (100–400 m; panels e–g). Values of PIC, POC, and PIC : POC ratios during late spring 2015 and mid-summer 2018 in the SE Pacific OMZ (this study), along with comparative data from other open-ocean and coastal margin regions (data from Balch et al., 2018), are presented. The Atlantic Ocean dataset includes samples from six cruises (AMT17-22). All raw data are shown, with group comparisons based on robust ANOVA and post hoc tests (see Sect. 2). The map was generated using Ocean Data View (Schlitzer, 2024), with annual O₂ climatology from the World Ocean Atlas 2018 (García et al., 2018).

4 Discussion

4.1 Coccolithophore species diversity and dominance off the Southeast Pacific margin

Unlike most other regions, where G. huxleyi is the dominant species (e.g., Winter et al., 2014), G. parvula/ericsonii were co-dominant with G. huxleyi in the coccolithophore communities at 20° S in 2015. The co-dominance of these three species was previously reported in winter samples from 2013 in the same zone (von Dassow et al., 2018; Díaz-Rosas et al., 2021). Likewise, “small Reticulofenestra complex” (presumably mostly G. parvula) and “small Gephyrocapsa complex” have been previously reported to be important in neighboring waters further offshore and to the north (Hagino and Okada, 2004), so these species are frequently important in the tropical Eastern South Pacific. To the south, co-dominance of G. huxleyi and G. muellerae has been reported previously (e.g., Díaz-Rosas et al., 2021). Thus, while coccolithophore richness remains low in the Eastern South Pacific margin, there is higher diversity within the genus Gephyrocapsa compared to other Eastern Boundary Currents (Ziveri et al., 1995; Henderiks et al., 2012; Guerreiro et al., 2013). Standing stocks of larger taxa contributed minimally to the total community above Z_eu but became more prominent below this depth. It should be noted that each detached coccolith of C. leptoporus is equivalent, in terms of PIC, to approximately 45 coccoliths of G. huxleyi. This suggests that the heavier coccoliths of these larger species may sink more efficiently (e.g., Menschel et al., 2016; Guerreiro et al., 2021).

4.2 Potential uncertainties in PIC measurements and the contributions of coccolithophores to PIC

Direct chemical measurements of PIC become increasingly difficult to measure as it decreases. In natural waters, where the PIC : POC ratios is usually much less than 1, the method of measuring total carbon before and after acidification becomes insensitive and problematic (Balch and Kilpatrick, 1996). Thus, it is preferable to measure acid-extractable Ca²⁺ from particulate matter. This direct chemical method of measuring PIC also has error from seawater retained on the filter by organic matter, which can trap Ca²⁺ even after gentle rinsing. This error still may increase proportionally as PIC : POC becomes lower, but it can be assessed and corrected for by measuring Na⁺ (Matson et al., 2019).

How much coccolithophores contribute to PIC remains an open question, as contributions from lithogenic sources (Daniels et al., 2012) as well as calcifying zooplankton (e.g., foraminifera, pteropods; Ziveri et al., 2023) complicate the relationship between PIC and coccolithophores. Off the northern Chilean coast, lithogenic PIC sources are limited. Fluvial inputs are negligible (Thiel et al., 2007) and the depth and topography of the Atacama Trench limits sediment resuspension (Xu et al., 2021). Therefore, we consider total PIC measured in this study to be likely mostly biogenic.

Estimates of PIC_Cocco by microscopy include sources of error which may underestimate PIC_Cocco relative to PIC_Total. One source of error relates to taxonomic and phenotypic variability in conversion factors, PIC quotas per coccolith, and estimates of the number of coccoliths per coccosphere, which might result in errors of up to 50 % in the estimation of PIC_Cocco using microscopic methods (Young and Ziveri, 2000). However, in this study, improved taxonomic resolution caused only minor changes in PIC_Cocco estimates due to the strong dominance of G. huxleyi and its close relatives in these waters. Difficulties in detection of smaller coccoliths, collapsed coccospheres, and fragmented coccoliths can also cause underestimation of PIC_Cocco (Barrett et al., 2014; Subhas et al., 2022). Despite these limitations, PIC_Cocco accounted for nearly half of direct PIC_Total measurements and PIC_Cocco values calculated from coccosphere and coccolith abundances were also linearly correlated with chemical measurements of PIC_Total, so we conclude that coccolithophores are major contributors to suspended PIC both above and below the oxycline in these OMZ waters.

4.3 Surface variation in coccolithophore stocks and PIC pools

Higher coccosphere and detached coccoliths were observed near 20° S in both 2015 and 2018, a pattern reflected in both in situ and satellite data. On average, higher values were recorded during mid-summer 2018 compared to late spring 2015, suggesting potential seasonality in PIC levels. This observation aligns qualitatively with satellite-PIC retrievals showing PIC concentrations in February 2018 extending further offshore and southward from the inshore 20° S sampled site. This is consistent with global satellite-PIC estimations showing higher inventories off the coasts of Chile from austral spring (October–November) peaking in summer (January–February), a seasonal pattern similar to that in the Benguela Current System (Hopkins et al., 2019).

The sites of higher PIC and coccosphere abundances observed in this study were associated with ratios of detached coccoliths to coccospheres that were relatively low (<16). This contrasts with the much higher detached-coccolith-to-coccosphere ratios (exceeding 250) reported during the termination phase of massive blooms or below the Z_eu (Balch et al., 1991; Holligan et al., 1993b; Lessard et al., 2005), suggesting that the coccospheres in the OMZ euphotic zone were actively growing. Higher coccolith PIC pools in the Z_eu were associated with SST ∼18 °C, suggesting a preference for the transition between nutrient-rich (cooler, upwelling) and nutrient-poor (warmer, offshore waters), similar to patterns observed in many other regions (Balch and Kilpatrick, 1996; Balch et al., 2000; Tyrrell and Merico, 2004; Zondervan, 2007; Matson et al., 2019). There is likely to be a roughly inverse relationship between coccolithophores and diatoms (Tyrrell and Merico, 2004; Menschel et al., 2016; Díaz-Rosas et al., 2021), consistent with classical paradigms of a succession from diatoms to coccolithophores and other flagellates as recently mixed waters stratify (Margalef, 1978).

4.4 Comparing euphotic zone coccolithophore stocks and PIC pools from OMZ and non-OMZ regions

Despite maximum coccolithophore abundances recorded in both years in this study being about twice the maximum values previously reported in the entire Chilean sector of the Southeast Pacific (Beaufort et al., 2008; Menschel et al., 2016; Díaz-Rosas et al., 2021) and the previously reported maximum for the entire Humboldt Current System (Hagino and Okada, 2006), coccolithophore stocks were still about an order of magnitude lower than the typical bloom abundances reported in other regions (e.g., Balch et al., 1991; Holligan et al., 1993a; Tyrrell and Merico, 2004; Poulton et al., 2013). Coccolithophore stocks in both the California Current and the European margin of the North Atlantic upwelling region are generally within the same range as the values reported here and previously for the Chilean OMZ (Silva et al., 2008; Ziveri et al., 1995; Guerreiro et al., 2013; Díaz-Rosas et al., 2021; Guerreiro et al., 2023). However, higher values have been reported in the Benguela Upwelling (Henderiks et al., 2012). In the California Current, the most similar Eastern Boundary Upwelling system to the Humboldt Current System, a G. huxleyi bloom was reported to reach abundances like those seen in Atlantic and Bering Sea blooms (Matson et al., 2019). Thus, while coccolithophore stocks in the Humboldt Current do appear to be lower than reported for other productive regions (Díaz-Rosas et al., 2021), we caution that it is still difficult to compare among upwelling regions.

This study presents the first in situ PIC_Total measurements from the OMZ waters off the Southeast Pacific margin, corresponding to late spring 2015 and mid-summer of 2018. PIC_Total values tended to be in the lower range compared to values documented so far for most of the rest of the ocean (Table 1, Fig. 7b). Comparable PIC data from other Eastern Boundary Upwelling regions is scarce. However, PIC_Total values as high as 67.3 µg C L⁻¹ were reported in a California Current G. huxleyi bloom (Matson et al., 2019; data not included in Fig. 7 because POC values not available). This picture becomes clearer when POC is considered as well. The POC values in the OMZ region were in a similar range to those of other highly productive regions (Fig. 7c). Also, the OMZ waters had significantly lower PIC on average compared to the majority of the Atlantic and Indian Oceans, despite tending to have higher average POC.

Although PIC correlated with POC, PIC was always about two orders of magnitude below POC in the Z_eu. In culture, PIC : POC ratios of G. huxleyi are around 1 (Paasche, 2001). These low PIC : POC ratios indicates that coccolithophores were only very minor contributors to total phytoplankton biomass in these waters. The Western Arctic Chukchi Sea is the only region where similarly low PIC and PIC : POC to the OMZ have been documented. This likely in part reflects that coccolithophores still have limited penetration into Arctic waters (Winter et al., 2014).

The findings of lower coccolithophore stocks, lower pools of PIC and lower PIC : POC in the SE Pacific OMZ compared to other regions are consistent with the possibility that the intrusions of low-pH–high-CO₂ waters into the Z_eu, as previously documented for this region (Torres et al., 2011; Vargas et al., 2021), might be limiting coccolithophore PIC in the region. Notably, among samples from within Z_eu it appeared that both low pH and low O₂ might limit coccolithophore PIC.

4.5 Subsurface PIC and PIC : POC ratios in the OMZ

As documented in other regions (Fernández et al., 1993; van der Wal et al., 1995; Ziveri et al., 2023), coccolithophore-derived PIC consisted predominantly of detached coccoliths, particularly in subsurface waters. The contributions to PIC of larger and rarer species with the heaviest coccoliths also increased below the Z_eu, as seen previously in other regions (e.g., Ziveri et al., 2007; Menschel et al., 2016; Guerreiro et al., 2021). The greater decrease in PIC_Cocco than PIC_Total below the Z_eu might be due to aggregation and fragmentation of coccoliths, suggesting the contribution of coccolithophores to subsurface PIC might be underestimated (Briggs et al., 2020).

Similar to the surface layer, PIC values in the subphotic zone (100–400 m) of the OMZ were lower than in the other regions compared (with the exception of the Western Arctic). The comparison of the subphotic PIC : POC ratios between the OMZ and other ocean regions was even more striking, as it was significantly lower than in all regions, including the Western Arctic. This pattern is consistent with a decreased availability of PIC as a ballast mineral in the OMZ.

An increase in the PIC : POC ratio below the Z_eu is often interpreted as evidence that PIC-rich aggregates sink faster, a phenomenon known as the ballast effect (Lee et al., 2009; Iversen and Ploug, 2010). This pattern was also observed in the SE Pacific OMZ (Fig. 6f). However, the increase in PIC : POC ratios from the upper to subsurface layer was smaller than in other regions with comparable data (Table 2), further suggesting a diminished role of PIC as ballast in OMZs.

Despite low pH, Ω_calcite remained above 1 throughout the OMZ euphotic and mesopelagic depths. Although we cannot discount the possibility that microenvironments with Ω_calcite below 1 exist where dissolution occurs, this does not align with the model in which low pH, associated with low O₂, leads to PIC dissolution, decreased ballasting by PIC, and consequently further O₂ depletion – at least within the depth interval between the upper and lower oxyclines. The lower oxycline lies at approximately 400 m and is very gradual, extending to well over 1000 m (Cornejo and Farías, 2012; Vargas et al., 2021). Data on Ω_calcite below 1000 m are limited (not shown), so we also cannot rule out that enhanced PIC dissolution may contribute to a positive feedback on O₂ depletion at greater depths.

Table 2 Comparison of the fractional increase in PIC : POC ratios between the surface (0–100 m) and subsurface (0–400 m) layers in the SE Pacific OMZ (this study) and other open-ocean and coastal regions (data from Balch et al., 2018). Only data points with both surface and subsurface PIC : POC ratios were included in the analysis.

4.6 Conclusions

Coccolithophores were major contributors to PIC in the OMZ region studied, but in these waters, PIC and PIC : POC ratios were lower in both the upper and subsurface waters of the OMZ, compared to other ocean regions. The observations are consistent with low pH potentially limiting coccolithophore growth and PIC accumulation in the euphotic zone, thereby decreasing the availability of PIC as a ballast mineral in OMZ waters. At 20° S, where PIC concentrations were highest, the oxycline is consistently the shallowest and sharpest, and subsurface waters are the most oxygen-deficient, reaching effective anoxia as indicated by elevated subsurface nitrite concentrations (Cornejo and Farías, 2012; Vargas et al., 2021). However, the waters above the oxycline and within the euphotic zone at this latitude did not exhibit the lower pH values observed farther south, where both coccolithophore abundance and PIC levels were also lower. Notably, the OMZ core – the most oxygen-deficient layer – never reached thermodynamically favorable conditions for calcite dissolution. These observations suggest that a local feedback loop between PIC dissolution and O₂ depletion below the euphotic zone is not significant within the OMZ core, though PIC dissolution might still play a role in deeper waters. Despite the decreased role of PIC suggested by our findings, we note that there is still no evidence that this results in slower POC sinking or shallower remineralization. In fact, some studies have reported efficient POC transfer from the euphotic zone to depth in OMZ systems (Cavan et al., 2017; Engel et al., 2017; Weber and Bianchi, 2020). Future studies that directly measure vertical fluxes of PIC, POC, and biogenic silica using sediment traps will be valuable for understanding how OMZs differ functionally from non-OMZ regions in terms of carbon export.