Two cruises were undertaken in the GCB during 2011 and 2012 (http://www.bco-dmo.org/project/473206). The Atlantic sector of the Southern Ocean (GCB1) was sampled from 11 January to 16 February 2011 onboard the R/V Melville between Punta Arenas, Chile and Cape Town, South Africa (Balch et al., 2016; Fig. 1). The Indian sector of the Southern Ocean (GCB2) was sampled from 18 February to 20 March 2012 onboard the R/V Revelle between Durban, South Africa and Fremantle, Australia (Fig. 1). Water samples were taken at 27 stations across a latitudinal gradient ranging from 38 to 60∘ S and a longitudinal gradient ranging from 60∘ W to 120∘ E during the GCB cruises, which enabled sampling of the major oceanographic features of this region.

Water samples were collected from the upper 30 m of the water column using a Niskin bottle rosette and CTD profiler for sea surface temperature, salinity, chlorophyll a (Chl a), nitrate plus nitrite (NOx), ammonia (NH4), phosphate (PO4), silicic acid (Si(OH4)), and carbonate chemistry. Nutrient analyses of NOx, PO4, Si(OH4), and NH4 were run on a Seal Analytical continuous-flow AutoAnalyzer 3, while salinity was determined using a single Guildline Autosal 8400B stock salinometer (S/N 69-180). Chlorophyll a was sampled in triplicate following Joint Global Ocean Flux Study (JGOFS; Knap et al., 1996) protocols. Mixed layer depths were calculated from processed CTD data by applying a criteria of a 0.02 kg m-3 density change from the 5 m value (Arrigo et al., 1998). Daily photosynthetically active radiation (PAR, mol PAR m-2 d-1) was estimated from 8-day composite Aqua MODIS data from the closest time and latitude–longitude point (averages were taken where necessary). Mixed layer average irradiance (E‾MLD) was calculated from daily PAR following Poulton et al. (2011).

Water samples were collected for total dissolved inorganic carbon (CT) and total alkalinity (AT) following standardized methods and analyzed using a Versatile Instrument for the Determination of Titration Alkalinity (VINDTA) with a precision and accuracy of ±1 µmol kg-1 (Bates et al., 1996, 2012). The remaining carbonate chemistry parameters were calculated from the CT and AT values using CO2SYS (Lewis and Wallace, 1998) and CO2calc (Robbins et al., 2010) with the carbonic acid dissociation constants of Mehrbach et al. (1973) refitted by Dickson and Millero (1987). This includes computation of the saturation state (Ω) for calcite (i.e., Ωcalcite).

Samples for biomineralizing phytoplankton community structure were taken from the upper 30 m of the water column. One-liter seawater samples were collected and prefiltered through a 200 µm mesh to remove any large zooplankton. Seawater samples were gently filtered through a 25 mm, 0.8 µm Whatman^® polycarbonate filter placed over a 200 µm backing mesh to ensure an even distribution of cells across the filter. Filters were rinsed with ∼ 5 mL of potassium tetraborate (0.02 M) buffer solution (pH = 8.5) to prevent salt crystal growth and PIC dissolution, air-dried, and stored in petri slides in the dark with a desiccant until further analysis.

To identify coccolithophores to the species level, each sample was imaged using the SEM methodology of Charalampopoulou et al. (2011). A central portion of each filter was cut out and gold coated, and 225 photographs were taken at a magnification of 5000 × (equivalent to ∼ 1 mm2; GCB1) or 3000 × (∼ 2.5 mm2; GCB2) using a Leo 1450VP SEM (Carl Zeiss, Germany). Detached coccoliths and whole coccolithophore cells (coccospheres) were identified following Young et al. (2003). Diatoms and other recognizable protists were identified following Hasle and Syvertsen (1997) and Scott and Marchant (2005). Where a confident species level identification was not possible, cells were assigned to the level of genera (e.g., Chaetoceros spp. or Pappamonas sp.). Each species identified was enumerated using the freeware ImageJ (v1.44o) for all 225 images or until 300 cells (or coccoliths) were counted. A minimum of 10 random images was picked for enumeration when species were in high abundance (> 1000 cells mL-1). The abundance of each species was calculated following Eq. (1): CellsmL-1=(C×F/A)/V, where C is the total number of cells (or coccoliths) counted, A is the area investigated (mm2), F is the total filter area (mm2), and V is the volume filtered (mL).

Multivariate statistics (PRIMER-E v.6.1.6; Clarke and Gorley, 2006) were used to examine spatial changes in coccolithophore and diatom abundance, species distribution, and the influence of environmental variability on biogeography (e.g., Charalampopoulou et al., 2011, 2016). Environmental data were initially assessed for skewness, most likely due to strong chemical gradients across fronts. Heavily left-skewed variables (NOx, silicic acid, and NH4) were log⁡(V+0.1) transformed to reduce skewness and stabilize variance. Other environmental data, including SST, salinity, E‾MLD, NOx, silicic acid, NH4, pH, pCO2, and Ωcalcite, were then normalized to a mean of zero and a standard deviation of 1, and Euclidean distance was then used to determine spatial changes in these parameters. A principal component analysis (PCA) was used to simplify environmental variability by combining the more closely correlated variables and the relative influence of the environmental variables within the data (Clarke, 1993; Clarke and Warwick, 2001; Clarke and Gorley, 2006).

Coccolithophore and diatom species diversity was assessed as the total number of species (S) and Pielou's evenness index (J′), which assesses how evenly the count data were distributed between the different species present (before further statistical analysis). Species with cell counts of less than 1 cell mL-1 and/or consistently representing less than 1 % of the total cell abundance were excluded from multivariate statistical analysis to reduce the influence of rare species. Analysis of coccolithophore and diatom community structure was carried out on standardized and square-root-transformed cell abundance (to reduce the influence of numerically abundant species) using a Bray–Curtis similarity matrix. Bray–Curtis similarity describes the percentage of similarity (or dissimilarity) between different communities according to their relative species composition. To identify which stations had a statistically similar biomineralizing phytoplankton community across the GCB, a SIMPROF routine (1000 permutations, 5 % significance level) was applied to the Bray–Curtis similarity matrix. SIMPROF identifies, based on pairwise tests of the calculated Bray–Curtis percentage similarity, whether the similarities between samples are smaller and/or larger than those expected by chance and groups those that are statistically distinct (Clarke et al., 2008). The phytoplankton species driving the differences between the groups were identified through a SIMPER routine and presented using nonmetric multidimensional scaling (nMDS; Clarke, 1993; Clarke and Warwick, 2001; Clarke and Gorley, 2006). SIMPER allows for the statistical identification of which species are primarily responsible for differences between groups of samples and breaks down the Bray–Curtis similarity into individual species contributions.

A BEST routine was applied to environmental and plankton data to determine the combination of environmental variables that “best” described the variability in coccolithophores and diatoms across the GCB. The BEST routine statistically searches for relationships between the biotic and abiotic patterns and to identify which environmental variable(s) explained most of the variation in species distribution. Spearman's rank correlations were used to further investigate the relationship between the key environmental variables identified in the BEST routine and selected coccolithophore and diatom species.

The GCB cruises crossed various biogeochemical gradients associated with the Antarctic Circumpolar Current (ACC) fronts and subcurrents, with most parameters following a recognizable latitudinal (or zonal) pattern. The position of the oceanic fronts referred to in the text relates to those defined in Fig. 1 (see also Balch et al., 2016). Sea surface temperature decreased southwards from 21 ∘C north of the STF to 1.1 ∘C close to 60∘ S (Table 1). The calcite saturation state (Ωcalcite) decreased from 5.2 north of the subtropical front to 2.6 close to 60∘ S (Table 1). Macronutrient concentrations generally increased southwards with a distinct divide across the SAF. NOx ranged from below detection limits (< 0.1 µM) to as high as 28 µM, with higher concentrations generally south of the Subantarctic Front (> 12 µM) and lower concentrations (< 7 µM) north of the Subantarctic Front (Table 1). PO4 followed a very similar pattern with concentrations generally greater than 1 µM south of the Subantarctic Front and < 0.6 µM to the north. Silicic acid concentrations were divided by the PF, being generally less than 2 µM to the north and up to 78.5 µM to the south (Table 1). E‾MLD was highest on the Patagonian Shelf (∼ 40 mol PAR m-2 d-1) and generally less than 10 mol PAR m-2 d-1 south of the Subantarctic Front (Table 1). There was no distinct latitudinal trend in pH or pCO2. Surface water pH was generally greater than 8.06, ranging from 8.03 on the Kerguelen Plateau to 8.13 in the Subtropical Front southwest of Australia (Table 1). Surface water pCO2 ranged from 299 to 444 µatm with both extremes in the vicinity of the Atlantic STF (Table 1). Chl a concentrations were variable across the oceanic gradients, highest on the Patagonian Shelf (2.78 mg m-3), and on average less than 1 mg m-3 in the South Atlantic compared with less than 0.5 mg m-3 in the southern Indian Ocean (Table 1).

The most frequently occurring and abundant size group within the coccolithophore and diatom counts were the nanoplankton (cells 2–20 µm). Large diatom species (cells > 20 µm) were found in higher numbers (up to 50 cells mL-1) south of the PF. Consideration of community biomass would potentially reduce the dominance of the nanoplankton relative to microplankton in the GCB. However, converting from cell size to biomass is not straightforward for diatoms, as highlighted by Leblanc et al. (2012), and to avoid such issues we consider species abundance only. Total cell abundances were less than 1000 cells mL-1 at most stations (Table 2), which are indicative of late summer, non-bloom conditions. In the South Atlantic, the highest abundance of coccolithophores was on the Patagonian Shelf (station GCB1-16; 1636 cells mL-1) and the highest abundance of diatoms was east of the South Sandwich Islands (station GCB1-77; 6893 cells mL-1; Table 2). In the southern Indian Ocean, coccolithophore abundance was highest near the Crozet Islands (station GCB2-27; 472 cells mL-1), and diatom abundance was highest at the most southerly station (station GCB2-73; 538 cells mL-1; Table 2). There were no stations in the southern Indian Ocean where coccolithophore and diatom abundances were greater than 1000 cells mL-1 (Fig. 2, Table 2). Additionally, the silicifying chrysophyte Tetraparma sp. was particularly abundant east of the South Sandwich Islands (station GCB1-77) at a cell density of 2000 cells mL-1, though they were present in low numbers (< 5 cells mL-1) at three more stations in the South Atlantic and absent throughout the rest of the GCB.

Coccolithophores dominated the biomineralizing community at 12 stations in terms of abundance north of the PF (Fig. 2, Table 2). On average coccolithophores contributed approximately 38 % to total (coccolithophore and diatom) abundance in the GCB. Coccolithophores were greater than 75 % of the total abundance at only one station to the north of South Georgia (station GCB1-59) and never accounted for 100 % of total cell numbers. Twenty-eight species of coccolithophores were identified as intact coccospheres across the GCB. Coccolithophore diversity decreased south towards 60∘ S, with the highest coccolithophore diversity (19 species) found in the vicinity of the STF in the eastern part of the southern Indian Ocean (station GCB2-106), while coccolithophore abundance was more evenly distributed between the different species in the lower latitudes (i.e., high J′; Table 2). Emiliania huxleyi was the most numerically abundant coccolithophore at all but four stations and was encountered in the mixed layer at all stations except one (station GCB2-73, the most southerly station in the Indian Ocean). Other coccolithophore species (e.g., Syracosphaera spp. and Umbellosphaera spp.) were present north of the PF throughout the GCB and were most abundant north of the STF. At stations south of the SAF (50∘ S) only one (E. huxleyi) or two species (E. huxleyi and Pappamonas sp.) were observed as intact coccospheres.

Diatoms dominated 15 stations in terms of biomineralizing plankton abundance across all environments sampled (Fig. 2, Table 2), being found in every sample analyzed and contributing 62 % (on average) to total (coccolithophores + diatoms) abundance. Diatoms made up 100 % of the total cell counts at the most southerly station in the southern Indian Ocean (station GCB2-73) and 99.7 % east of the South Sandwich Islands (station GCB1-77; Fig. 2). Seventy-six species of diatom were identified as intact cells across the entire GCB. The most frequently occurring species in the GCB were small (< 5 µm in length) Fragilariopsis spp. The highest abundance of diatoms in the South Atlantic Ocean (6893 cells mL-1) was dominated by F. nana east of the South Sandwich Islands (station GCB1-77). The highest diatom abundance in the southern Indian Ocean (538 cells mL-1) was dominated by F. pseudonana at the most southerly station (station GCB2-73) sampled. Another frequently dominant diatom was Pseudo-nitzschia spp., which was most abundant north of the PF (Table 2).

Diatom species richness increased south towards 60∘ S with the contribution of the different diatom species to the total biomineralizing plankton abundance fairly even (J′ > 0.5, Table 2), except at stations (stations GCB1-70, GCB1-77, GCB2-27, and GCB2-63) where Fragilariopsis spp. < 5 µm were dominant (> 70 % of the diatom population, J′ < 0.5). The highest diatom species richness (32 species) was found in the GCB south of the SAF (station GCB2-36) at a temperature of 8 ∘C, in HNLSiLC conditions (NOx 18.9 µM, silicic acid 1.7 µM, 0.21 mg Chl a m-3).

Three of the environmental variables were removed from the statistical analysis following a Spearman's rank (rs) correlation analysis (Table S1 in the Supplement). NOx and PO4 had a strong significant positive correlation (rs=0.961, p<0.0001), so NOx was deemed representative of the distribution of both nutrients. Sea surface temperature displayed significant negative correlations with both CT (rs=-0.981, p<0.0001) and AT (rs=-0.953, p<0.0001), so sea surface temperature was taken as being representative of these two variables of the carbonate chemistry system.

The variation in environmental variables across the GCB was examined using a principal component analysis (PCA), which simplifies environmental variability by combining closely correlated variables into principal components in order to account for the greatest variance in the data with the fewest components. The first principal component (PC1) accounted for 58 % of the variation in environmental variables, with an additional 17 % of environmental variation described by PC2 (Table 3). PC1 describes the main latitudinal gradients of environmental changes across the GCB (decreasing SST, increasing macronutrients). PC1 is a predominantly linear combination of SST, salinity, NOx, silicic acid, NH4, and Ωcalcite; there is a significant positive correlation of PC1 with SST and salinity and a significant negative correlation with all other variables (Table 3). PC2 represented the environmental variation in the GCB occurring independently of latitude and was driven predominantly by variation in pCO2, with weaker influences from E‾MLD and pH (Table 3). PC2 had significant positive correlations with pCO2 and E‾MLD and a negative correlation with pH.

The SIMPROF routine identified the stations in the GCB that had statistically similar coccolithophore and diatom community composition through a comparison of Bray–Curtis similarities. Six statistically significant groups (p<0.05) were defined across the GCB (Fig. 3). Three of these groups (A, B, C) were specific to the South Atlantic Ocean (Fig. 3). For example, groups A and B represented individual stations GCB1-46 and GCB1-117, respectively, in the subtropical region of the South Atlantic Ocean. The most southerly stations in the South Atlantic Ocean (stations GCB1-70 and GCB1-77) defined group C (Fig. 3). Groups D, E, and F included stations across the GCB in both ocean regions. Here, group D was defined by eight stations sampled predominantly north of the SAF, while group F was defined by 11 stations predominantly sampled south of the SAF (Fig. 3). These statistically defined similar community structures indicate that although the GCB covers a wide expanse of ocean, the community structure is consistently latitudinally defined across its longitudinal range.

A SIMPER routine statistically identified the species that define the difference between (and the similarity within) the statistically different community structures defined by the SIMPROF routine (Table 4). The abundance and distribution of four phytoplankton species (E. huxleyi, Pseudo-nitzschia spp., F. nana, and F. pseudonana; Fig. 4) were identified as having the most significant contribution to differences in community structure across the GCB (Table 4). Emiliania huxleyi and F. pseudonana were the most numerically dominant coccolithophore and diatom species, respectively, across the GCB (Table 2). Fragilariopsis pseudonana was the numerically dominant diatom (> 30 %) at seven stations in the southern Indian Ocean (Table 2). The diatom with the highest abundance, F. nana (6797 cells mL-1), was almost exclusively found in the South Atlantic Ocean (Table 2) and the more frequently occurring Pseudo-nitzschia spp. was present at all but one station.

The nonmetric multidimensional scaling (nMDS) plot of the Bray–Curtis similarities (Fig. 5) shows the station distribution with respect to the SIMPROF-defined groups (Fig. 5a), the four main species (Fig. 5b–e), and also holococcolithophores (Fig. 5f). The more closely clustered the stations, the more similar their biomineralizing species composition. Groups A and B were defined by the absence of E. huxleyi (Fig. 5b) and the presence of either holococcolithophores (group A; Fig. 5f) or the diatom Cylindrotheca sp. (group B). Group C was defined by the dominance of F. nana (Table 4; Fig. 5d) and low contributions from E. huxleyi and Pseudo-nitzschia spp. (Table 2; Fig. 5b, e), resulting in a significant difference from the other groups. Group D had high total species diversity overall (19–41 species; Table 2) and was defined by similar relative abundances of E. huxleyi and Pseudo-nitzschia spp., which were not found elsewhere (Fig. 5b, e). Group E, composed of stations north of the SAF (Figs. 3, 5a), included E. huxleyi, U. tenuis, and holococcolithophores (Table 4, Fig. 5b, f). The low abundance and diversity (3–125 cells mL-1, 7–11 species; Table 2) of diatoms within group E separated it from the other groups. The combination of E. huxleyi, F. pseudonana, and Pseudo-nitzschia spp. that defined group F (Table 4, Fig. 5b, c, e) represented stations on the Patagonian Shelf and south of the SAF (Figs. 3, 5a). The almost monospecific E. huxleyi coccolithophore community (Table 2) in group F highlights its strong dissimilarity from the other community structure groups identified (Fig. 5).

The influence of environmental variables on the biogeography of coccolithophores and diatoms in the GCB was assessed using the BEST routine. The strongest Spearman's rank correlation (rs=0.55, p<0.001) between all possible environmental variables and the biogeographical patterns observed came from a combination of five variables: (1) SST, (2–4) macronutrients (NOx, silicic acid, NH4), and (5) pCO2. This was followed by a correlation of rs=0.54 (p<0.001) that included these parameters and Ωcalcite. Salinity was included in the third-highest correlation, whereas E‾MLD and pH did not rank as significant factors in the BEST analysis.

Studies of Southern Ocean phytoplankton productivity have generally focused on the microphytoplankton (Barber and Hiscock, 2006) as these species contribute around 40 % to total oceanic primary production (Sarthou et al., 2005; Uitz et al., 2010). However, nanoplankton and picoplankton are becoming increasingly recognized as important contributors to total phytoplankton biomass, productivity, and export in the Southern Ocean (e.g., Boyd, 2002; Froneman et al., 2004; Uitz et al., 2010; Hinz et al., 2012), as the dominant size group in both post-bloom (Le Moigne et al., 2013) and non-bloom conditions (Barber and Hiscock, 2006).

In this study, coccolithophores were generally numerically dominant at stations sampled north of the PF, particularly around the Subantarctic Front, whereas diatoms were dominant at stations south of the PF (Fig. 2). There was also a significantly different species distribution (a priori ANOSIM; R=0.227, p<0.01) north and south of the Subantarctic Front, which has been previously identified as the divider between calcite- and opal-dominated export in the Southern Ocean (e.g., Honjo et al., 2000; Balch et al., 2016). Diatoms were more abundant (∼ 570 cells mL-1) than coccolithophores (∼ 160 cells mL-1) on average in the entire GCB. This is in contrast to Eynaud et al. (1999) for the South Atlantic Ocean at a similar time of year, who reported a peak in coccolithophore cell abundance in the vicinity of the PF (a feature that was not observed in this study). These differences are likely due to the variability of Southern Ocean plankton on short temporal scales (Mohan et al., 2008), including variability in the seasonal progression of the spring bloom (Bathmann et al., 1997).

The coccolithophore E. huxleyi and diatoms F. pseudonana, F. nana, and Pseudo-nitzschia spp. (Fig. 4) were all identified as being central to defining the statistical similarities within, and the differences between, the different biomineralizing phytoplankton groups (Table 4, Fig. 5). Three of these species (E. huxleyi, F. nana, and F. pseudonana) are part of the nanoplankton, whilst Pseudo-nitzschia spp. is at the lower end of the size range of the microplankton (Pseudo-nitzschia spp. is > 20 µm in length but < 5 µm in width) and contributes significantly to biomass in Southern Ocean HNLC regions (Boyd, 2002). Emiliania huxleyi and Fragilariopsis spp. smaller than 10 µm have been identified as two of the most abundant biomineralizing phytoplankton further south in the Scotia Sea (Hinz et al., 2012). Our results further highlight that nanoplankton have the potential to contribute a significant proportion to GCB community composition alongside the larger phytoplankton (including large diatoms) typical of HNLC regions.

Abundances of HNLC diatoms, such as F. kerguelensis (< 10 cells mL-1), T. nitzschioides (< 20 cells mL-1), and large Chaetoceros spp. (< 10 cells mL-1), were lower than those observed in other studies (e.g., Poulton et al., 2007; Armand et al., 2008; Korb et al., 2010, 2012). Furthermore, the absence of the diatom Eucampia antarctica (< 1 cell mL-1) in this study does not reflect the typical assemblage (sometimes > 600 cells mL-1) found in previous studies (e.g., Kopczynska et al., 1998; Eynaud et al., 1999; de Baar et al., 2005; Poulton et al., 2007; Salter et al., 2007; Korb et al., 2010). Low abundances of the large-celled diatoms in the silicic-acid-replete regions may partly relate to the small filter area analyzed using SEM; in this study the area imaged equates to a relatively small volume of water (2–6 mL depending on magnification) relative to the larger volumes (10–50 mL) often examined for light microscopy in other studies. Large, rare cells may not be enumerated from such small sample volumes; however, the numerically abundant nanoplankton groups were well represented in SEM images. Conversely, samples preserved in acidic Lugol's solution for light microscopy analysis are biased towards larger species since small diatoms (< 10 µm) are not clearly visible and coccolithophores are not well preserved (Hinz et al., 2012). In the future, a combination of both imaging techniques is recommended to fully express the phytoplankton community structure of the Southern Ocean.

The importance of coccolithophores in the GCB was examined via species composition and the abundance of intact cells, focusing on areas identified as having high PIC reflectance from underway sampling and satellite observations (Balch et al., 2014, 2016; Hopkins et al., 2015). Higher species diversity of coccolithophores occurred north of the STF (i.e., 6–19 species; Table 2). Coccolithophores are diverse in the stratified and low-nutrient waters associated with lower latitudes (Winter et al., 1994; Poulton et al., 2017). Only a few species are found in the colder waters south of the STF (Mohan et al., 2008), the most successful being E. huxleyi, which was observed at an abundance of 103 cells mL-1 at 1 ∘C in this study in the South Atlantic (station GCB1-70). The 2 ∘C isotherm has been previously assumed to represent the southern boundary of E. huxleyi (e.g., Verbeek, 1989; Mohan et al., 2008), and interannual variability could be influenced by the movement of the southern front of the Antarctic Circumpolar Current (Holligan et al., 2010). The Southern Ocean E. huxleyi morphotype (Cook et al., 2011; Poulton et al., 2011) may therefore have a wider temperature tolerance than its Northern Hemisphere equivalent (Hinz et al., 2012) and has been observed poleward of 60∘ S further east in the Southern Ocean (Cubillos et al., 2007) and across the Drake Passage (Charalampopoulou et al., 2016). There were three distinct E. huxleyi occurrences (the Patagonian Shelf, north of South Georgia, and north of the Crozet Islands) within the GCB where E. huxleyi contributed > 50 % of the total cell counts of biomineralizing phytoplankton. Emiliania huxleyi was most abundant (1636 cells mL-1) on the Patagonian Shelf and was the most frequently occurring coccolithophore across the entire GCB. The main E. huxleyi occurrences are further discussed below to examine why this species is so widely distributed in the GCB.

The environmental variables that best describe coccolithophore and diatom species distribution in this study were SST, macronutrients (NOx, silicic acid, NH4), and pCO2 (Spearman's rank correlation = 0.55, p<0.001), with the second-highest correlation (Spearman's rank correlation = 0.54, p<0.001) including the calcite saturation state (Ωcalcite). The inclusion of pCO2 and Ωcalcite as important factors indicates a potential influence of carbonate chemistry on coccolithophore and diatom distribution (and vice versa) in the GCB. However, Ωcalcite had a very strong positive correlation (r=0.964, p<0.0001) with SST (Table S1), and therefore separating the influences of the two variables was impossible in this study due to the tight coupling between carbonate chemistry and temperature (as also observed by Charalampopoulou et al., 2016).

Relating carbonate chemistry to phytoplankton distribution, growth, and physiology is an important step when considering the potential effects of climate change and ocean acidification on marine biogeochemistry. In this study, no significant correlation (Spearman's r=0.259, p=0.164, n=27) occurred between pH and Chl a. The inclusion of pCO2 and Ωcalcite as influential factors in the statistical results describing GCB species biogeography highlights the importance of understanding phytoplankton responses to carbonate chemistry as a whole rather than as individual carbonate chemistry parameters (Bach et al., 2015). Of the four major species driving the differences in biomineralizing plankton community composition and biogeography across the GCB, only F. pseudonana abundance was positively correlated with pCO2 (Pearson's product moment coefficient = 0.577, p<0.05, n=16).

The response of diatoms to increasing pCO2 is not straightforward (e.g., Boyd et al., 2015), with some studies implying that large diatoms may be more successful in future climate scenarios (e.g., Tortell et al., 2008; Flynn et al., 2012), although changes in nutrient and light availability (via stronger stratification) may prevent a permanent switch in phytoplankton community structure (Bopp et al., 2005). The carbonate chemistry system is complex, as biological activity also impacts the concentration of each of the components. Organic matter production reduces total dissolved inorganic carbon (CT), and hence pCO2 via photosynthesis, and increases alkalinity (AT) through nutrient uptake, while subsequent respiration and remineralization of organic matter has the opposite impact. The simultaneous actions of biological and physical processes result in seasonal and localized changes in the carbonate system, which are often difficult to decouple.

In our study, there was no significant correlation between E. huxleyi and Ωcalcite (Pearson's product moment = 0.093). However, the waters of the GCB remained oversaturated (Ωcalcite > 2) throughout, and the relationship between coccolithophores, calcification, and carbonate chemistry is now recognized as being complex and nonlinear (e.g., Beaufort et al., 2011; Smith et al., 2012; Poulton et al., 2014; Rivero-Calle et al., 2015; Bach et al., 2015; Charalampopoulou et al., 2016; Marañón et al., 2016). Hence, significant gaps remain in our understanding of the in situ coccolithophore response to increasing pCO2, reduced pH, or decreasing Ωcalcite. Notably, a significant positive correlation between Pseudo-nitzschia spp. and Ωcalcite also existed (Pearson's product moment correlation = 0.5924, p<0.01, n=19) across the GCB despite there being no presently known detrimental effect on diatoms with low saturation states. However, due to the tight coupling of temperature and Ωcalcite (and Pseudo-nitzschia spp. and temperature), the correlation is more likely to be temperature driven.