Seawater samples were collected during two UK-GEOTRACES cruises in the South Atlantic Ocean (GA10; Fig. 1). The first cruise (D357) took place during austral spring 2010 (18 October to 22 November 2010), sampling the south-eastern Atlantic on board the RRS Discovery. During D357, two transects were completed between Cape Town and the zero meridian that represent early austral spring (D357-1) and late austral spring (D357-2), respectively. The second cruise (JC068) took place during austral summer 2011/2012 (24 December 2011 to 27 January 2012), along the same transect of the first cruise and continuing along 40∘ S between Cape Town and Montevideo, Uruguay, on board the RRS James Cook. For JC068, we present here only the repeat transect data between Cape Town and 13∘ W that represent the south-eastern Atlantic aspect of this transect. The stations occupied during the three transects were not identical but rather represent a coverage of the Southern Ocean and sub-tropical waters present. Where stations were reoccupied during one or more transects, they have the same station number.

All sampling bottles were cleaned according to the procedures detailed in the GEOTRACES sample handling protocols (Cutter et al., 2010). Seawater and particulate samples below 15 m depth were collected using a titanium-frame CTD with 24 trace metal clean, 10 L, Teflon-coated OTE (ocean test equipment) Niskin bottles deployed on a plasma rope. Sub-samples for dissolved trace metal analysis were filtered through 0.8/0.2 µm cartridge filters (AcroPak™ 500, Pall) into 125 mL low-density polyethylene bottles inside a class 1000 clean-air container. Each sub-sample was acidified to pH 1.7 (0.024 M) by addition of 12 M hydrochloric acid (HCl, UpA, Romil) under a class 100 laminar-flow hood. Vertical sampling for dissolved trace metals was augmented by surface samples collected at each station using a towed “fish” positioned at approximately 3–5 m depth. Fish samples were filtered in-line and acidified as described for samples collected from the titanium sampling system. Particulate samples were collected onto acid-clean 25 mm, 0.45 µm, polyethersulfone membrane disc filters (Supor^®, Pall) and stored frozen (-20 ∘C) until shore-based analysis.

Dissolved Co was determined in the ISO-accredited clean room facility (ISO 9001) at the University of Plymouth (UK) using flow injection with chemiluminescence detection, modified from the method of Cannizzaro et al. (2000) as described by Shelley et al. (2010). Briefly, dCo was determined in UV-irradiated samples using the reaction between pyrogallol (1,2,3-trihydrobenzene) and hydrogen peroxide formed in the presence of Co. Standards (20–120 pM Co) were prepared in 0.2 µm filtered low-dCo seawater (16.5 ± 5.2 pM, n=15) by serial dilution of a 1000 ppm Co ICP-MS standard (Romil, UK). The accuracy of the analytical method was validated by quantification of dCo in SAFe (S and D2) and GEOTRACES (GD) reference seawater (Table 1). There was no detectable analytical dCo blank, and the limit of detection (3σ of the lowest concentration standard) was 1.98 ± 0.87 pM (n=15).

Dissolved Zn was determined using flow injection coupled with fluorescence detection, modified from the method of Nowicki et al. (1994) and described previously for this GEOTRACES section by Wyatt et al. (2014). The accuracy of the analytical method was validated by quantification of dZn in SAFe (S and D2) reference seawater (Table 1). The blank for dZn determination was 0.14 ± 0.13 nM, and the limit of detection (3σ of the lowest concentration standard) was 0.01 ± 0.01 nM (n=15).

Measurement uncertainties were estimated after the Nordtest approach (Worsfold et al., 2019), where a combined uncertainty (uc) is computed from day-to-day within-lab reproducibility and uncertainties associated with the determination of reference materials (Table 1). This approach creates higher uncertainties than those previously published for dZn and dCo analyses but provides a more realistic estimation of analytical uncertainty. During this study, the uc for dZn and dCo analysis was 22 % and 19 %, respectively, similar to the 13 %–25 % reported by Rapp et al. (2017) for the determination of trace metals, including dZn and dCo, by online pre-concentration and high-resolution sector field ICP-MS detection. The elevated uc within our data results from the greater uncertainty surrounding the very low dZn and dCo concentrations in the SAFe reference sample (Table 1), whereas the dZn and dCo uc using only the Safe D2 are <5 %. Hereafter, when presenting low dZn and dCo concentrations for comparison with phytoplankton biological requirements (Sect. 3.5), we apply a fixed uc of 20 % to our data.

Total particulate trace metals (i.e. pZn, pCo, pTi) were determined using inductively coupled plasma mass spectrometry (Thermo Fisher XSeries-2) following a sequential acid digestion modified from Ohnemus et al. (2014). Potential interferences (e.g. 40Ar16O on 56Fe) were minimized through the use of a collision and reaction cell utilizing 7 % H in He and evaluation of efficiency and accuracy assessed using certified reference material (CRM). Full details of the method and CRM results can be found in Milne et al. (2017).

The dissolved macronutrients phosphate (PO43-), silicic acid (Si(OH)4 but referred to as Si hereafter) and nitrate (determined as nitrate + nitrite, NO3-) were determined in all samples for which trace metals were determined, in addition to samples collected from a stainless-steel rosette. Macronutrients were determined using an AA III segmented-flow AutoAnalyzer (Bran and Luebbe) following colorimetric procedures (Woodward and Rees, 2001). Salinity, temperature and depth were measured using a CTD system (Seabird 911+), whilst dissolved O2 was determined using a Seabird SBE 43 O2 sensor. Salinity was calibrated on board using discrete samples taken from the OTE bottles and an Autosal 8400B salinometer (Guildline), whilst dissolved O2 was calibrated using a photometric automated Winkler titration system (Carritt and Carpenter, 1966). Mixed-layer depths (MLDs) were calculated using the threshold method of de Boyer Montégut et al. (2004), where MLD is identified from a linear interpolation between near-surface density and the depth at which density changes by a threshold value (0.125 kg m-3).

Measurements of phytoplankton pigment biomass and community structure were made on discrete samples collected using a 24-position stainless-steel CTD rosette equipped with 20 L OTE Niskin bottles. For chlorophyll a analysis, samples were filtered (0.7 µm Whatman GF/F), and then the filters were extracted overnight in 90 % acetone (Holm-Hansen et al., 1965). The chlorophyll a extract was measured on a pre-calibrated (spinach chlorophyll a standard, Sigma) Turner Designs Trilogy fluorometer. High-performance liquid chromatography (HPLC) samples (0.5–2 L) for accessory pigment analyses were filtered (0.7 µm Whatman GF/F), flash-frozen in liquid nitrogen and stored at -80 ∘C prior to analysis using a Thermo HPLC system. The matrix factorization program CHEMTAX was used to estimate the contribution of taxonomic groups to total chlorophyll a (Mackey et al., 1996). Concentrations of nanophytoplankton, Synechococcus, Prochlorococcus and photosynthetic picoeukaryotes were analysed by analytical flow cytometry (AFC) using a FACSort flow cytometer (Becton Dickenson, Oxford, UK) according to the methods described in Davey et al. (2008) and Zubkov et al. (2003).

The prominent water masses along the D357 and JC068 transects (Fig. 2) were identified by their characteristic thermohaline and macronutrient properties (Sarmiento et al., 2004; Ansorge et al., 2005; Browning et al., 2014). Wyatt et al. (2014) provide a more detailed description of the JC068 hydrography along the entire GA10 section. Whilst we aim to compare the near-shore versus offshore distributions of micro- and macronutrients, note that sub-Antarctic mode water was not sampled for trace metals during the D357-2 late-spring transect, and therefore only the early-spring and the summer values are discussed for SASW hereafter.

Potential sources of trace metals to surface waters of the south-eastern Atlantic include atmospheric inputs from South Africa and Patagonia (Chance et al., 2015; Menzel Barraqueta et al., 2019) as well as interactions with shelf and slope waters of the Agulhas Bank (Bown et al., 2011; Boye et al., 2012; Paul et al., 2015). During the D357 spring transects, elevated mixed-layer dZn and dCo concentrations (up to 4.57 nM and 50 pM, respectively; Sect. 3.2) were observed at stations closest to the Agulhas Bank shelf and slope (Stations 0.5, 1, 1.5 and 2). Here, we compare these metal elevations with respect to the aforementioned sources. Firstly, we encountered only brief, light rain during the study and thus minimal wet deposition of atmospheric aerosol. By combining the median flux of atmospheric dry deposition for soluble Zn and Co for the south-eastern Atlantic (Zn: 6.0 nmol m-2 d-1; Co: 0.05 nmol m-2 d-1; Chance et al., 2015) with the mean mixed-layer depth (34 m) for STSW during D357, dust dissolution is estimated to add approximately 5.5 and 0.05 nmol m-3 dZn and dCo, respectively, over a 1-month period. These inputs are low compared with the mixed-layer metal inventories, representing <1 % of dZn and dCo concentration in STSW during the D357 transects (Table 2), and would not be sufficient to generate distinct mixed-layer maxima. It is likely, therefore, that the dZn and dCo elevations originated from the advection of metal-enriched waters from the western Agulhas Bank, a region identified as a distinct source of both dissolved and particulate trace metals to the south-eastern Atlantic (Chever et al., 2010; Bown et al., 2011; Boye et al., 2012; Paul et al., 2015) and/or from the leakage of Indian Ocean water into the south-eastern Atlantic via the AC.

The detachment of Agulhas rings and filaments from the AC during its retroflection back towards the Indian Ocean constitutes a source of Pb to the surface south-eastern Atlantic along the D357 transects (Paul et al., 2015). Whilst we observed elevated mixed-layer dZn and dCo at ∼ 15∘ E during both D357 transects, the absence of metal enrichment across the depth of the AC salinity maxima (Figs. 2 and 3) suggests that the signal must be entrained from elsewhere. Furthermore, dZn concentrations from the AC along the east coast of South Africa do not exceed 0.5 nM in the upper 200 m (Gosnell et al., 2012). It is therefore likely that the dZn and dCo enrichment was derived from the Agulhas Bank. The AC has been shown to meander over and interact with the Agulhas Bank, forming eddies and filaments on the shoreward edge of the AC proper that tend to move northwards along the western shelf edge and into the south-eastern Atlantic (Lutjeharms and Cooper, 1996; Lutjeharms, 2007), potentially delivering shelf-derived sedimentary material. We found no evidence of a fluvial signature in our data, and no significant fluvial source for trace elements to our study region has been reported in the literature. We focus here on the more likely scenario of sedimentary inputs as the driver of mixed-layer dZn and dCo elevations at the shelf and slope stations during D357. Despite no available particulate trace metal data for the D357-1 early-spring transect for direct comparison with the highest dZn and dCo elevations, we observed elevated mixed-layer particulate Zn (pZn; 0.08–1.40 nM) and Co (pCo; 8–49 pM) at stations closest to South Africa during the D357-2 late-spring transect (Stations 0.5, 1 and 1.5; Fig. S1), coincident with elevated dZn (0.05–1.82 nM) and dCo (1–43 pM). Furthermore, for the upper 500 m at Stations 0.5 and 1, we found strong positive correlations between particulate aluminium and titanium (pAl / pTi; slope=41.7 mol mol-1, Pearson's r=0.99, n=15), as well as particulate Fe and titanium (pFe / pTi; slope=10.2 mol mol-1, Pearson's r=0.99, n=15), indicative of a strong lithogenic source. Whilst there are presently no South African sedimentary data against which we can compare our water column values, our pAl / pTi and pFe / pTi slope ratios are in excess of upper crustal mole ratios (34.1 and 7.3 mol mol-1, respectively; McLennan, 2001). These 500 m ratios are also steeper than the aggregate slopes for the full depth of the Atlantic Ocean away from hydrothermal sources (32.1 and 7.4 mol mol-1, Pearson's r>0.97, n=593; Schlitzer et al., 2018). Given the refractory nature of lithogenic pTi across diverse oceanic environments (Ohnemus and Lam, 2015), this may suggest the resuspension and dissolution of Agulhas Bank sediments enriched in dAl and dFe, followed by westward offshore transport, a common feature of the Bank's physical circulation during spring and summer (Largier et al., 1992). Such processes may in turn provide an additional source of dZn and dCo to STSW of the south-eastern Atlantic. For example, Little et al. (2016) proposed that oxygen-deficient, organic-rich, continental-margin sediments may constitute a significant global sink within the marine Zn cycle. These sediments could additionally provide a local source of dZn following remineralization. Recent model outputs have likewise highlighted oxygen-deficient boundary sediments as a dominant external source of Co to the oceans (Tagliabue et al., 2018). Given that oxygen-depleted (<45 µM) bottom waters are prevalent across the western Agulhas Bank (Chapman and Shannon, 1987; Chapman, 1988), considered to arise from high organic-matter input to sediments and its bacterial decomposition, a sedimentary source of dZn and dCo appears likely.

In addition to seasonal variations in the lateral advection of continentally derived trace metals, the lower dZn and dCo concentrations in STSW during summer compared with spring likely reflect differences in biological utilization. Here, we examine the micro- and macronutrient concentration inventories to assess the trace metal stoichiometry of the south-eastern Atlantic over seasonal timescales. The data were grouped into STSW and SASW regimes, with STSW defined by θ≥15 ∘C. This isotherm was located at a mean depth of 144 ± 96 m across the study compared with a mean mixed-layer depth of 39 ± 10 m, and thus the inventories for SASW were determined over this depth for comparison with STSW (Table 2). Early- and late-spring STSW samples in the depth range 20–55 m that clearly exhibited continentally derived elevated dZn and dCo were removed from the analysis in order to compare stoichiometry with respect to biological processes. For SASW, micronutrient sampling did not occur during late spring and therefore only early-spring and summer values are compared.

Distinct temporal trends in the stoichiometric relationship with PO43- were evident for both dZn and dCo (Fig. 4). Within STSW, the dZn / PO43- inventory ratio ranged from 699–1876 µmol mol-1 (Table 2), with the highest value observed during early spring and the lowest during summer. Combined with summer dZn concentrations 4 times lower than early spring, this suggests strong biological uptake of dZn alongside PO43- between seasons. In contrast, lower dZn / PO43- ratios of 372 and 188 µmol mol-1 were observed in SASW during early spring and summer, respectively. Here, the absolute change in dZn concentration between spring and summer was lower than for STSW but was greater for PO43-, likely reflecting the increased availability of PO43- in these Southern-Ocean-derived waters (Table 2) and an open-ocean phytoplankton community that has lower trace metal requirements than its counterparts north of the STF. Such dZn / PO43- ratios sit at the lower end of cellular Zn / P reported for the diatom-type and haptophyte-type phytoplankton typical of this region (∼ 100–1100 µmol mol-1; Twining and Baines, 2013, and references therein), highlighting the importance of micronutrient processes with respect to Zn availability.

In contrast to dZn, the spatiotemporal variation observed for STSW dCo / PO43 was small, with ratios ranging from 82–129 µmol mol-1 (Table 2), likely reflecting external inputs to the oceans and biological Co requirements that are typically 4 times less than for Zn (Ho et al., 2003; Roshan et al., 2016; Hawco et al., 2018). The STSW dCo / PO43- ratio decreased between early- and late-spring transects, potentially in part due to the westward expansion of STSW during late spring (Fig. 2) and subsequent mixing with SASW depleted in dCo relative to PO43- (Fig. 3). This dilution is likely also true of dZn and Si, yet their STSW concentration inventories may be sufficiently high so as to mask this effect. Unfortunately, an insufficient quantity of late-spring SASW data are available with which to affirm this postulation. The highest dCo / PO43- ratio was observed during summer due to the preferential biological removal of PO43- relative to dCo.

In SASW, dCo / PO43- was consistently low, with ratios of 23 and 26 µmol mol-1 for early spring and summer, respectively. Much higher inventory ratios of ∼ 580 µmol mol-1 can be calculated over similar depths for open-ocean North Atlantic waters (GA03 Stations 11–20; Schlitzer et al., 2018), likely reflecting an elevated atmospheric Co input and/or an extremely low surface PO43- inventory (Wu et al., 2000; Martiny et al., 2019).

Our results provide evidence for the greater availability and preferential removal of dZn relative to dCo in the upper water column the south-eastern Atlantic based on STSW dZn / dCo inventory ratios of 19, 17 and 5 mol mol-1 for the three transects and SASW ratios of 13 and 7 mol mol-1 for early spring and summer, respectively (Table 2). With relatively consistent inter-seasonal dCo inventories for STSW and SASW, indicating a more balanced ecophysiological regime with regard to dCo organization, the change in dZn / dCo inventory stoichiometries principally reflects changes in dZn concentration. We postulate that the inter-seasonal variations in dZn and dCo availability and stoichiometry of the south-eastern Atlantic reflect changes in the relative nutritional requirement of resident phytoplankton and/or biochemical substitution of Zn and Co to meet nutritional demand.

Here we discuss the principle phenomena that together likely explain our observations of seasonally decreasing dZn / dCo inventory stoichiometries in STSW and SASW of the south-eastern Atlantic, i.e. the preferential removal of dZn relative to dCo, leading to low dZn availability, and differences in phytoplankton assemblages with different cellular-metal requirements.

Satellite images show elevated surface chlorophyll concentrations across the south-eastern Atlantic STF compared with waters farther north and south, with peak concentrations observed during summer in January 2012 (Fig. 1). Profiles of total chlorophyll a concentration (Fig. S2) also show maximum summer values in the upper water column of STSW (1.02 mg m-3) and SASW (0.49 mg m-3) compared with spring values (<0.61 and <0.36 mg m-3, respectively). This is consistent with the hypothesis that increasing irradiance, coupled with shallower mixed-layer depths (de Boyer Montégut et al., 2004), results in enhanced growth conditions across the STF between September and February (Browning et al., 2014). Diagnostic pigment analyses (Fig. 5a) indicated that eukaryotic nanophytoplankton, specifically Phaeocystis-type haptophytes, dominated the early-spring STSW chlorophyll a content (73 %) but with a reduced contribution during summer (20 %). Maximum growth rates for cultured Phaeocystis antarctica have been achieved under elevated Zn concentrations (Saito and Goepfert, 2008), and thus, the dominance of this haptophyte would likely contribute to the removal of dZn between spring and summer. Furthermore, an increased summer diatom contribution (13 % chlorophyll a compared with near-zero during spring transects) would have further reduced the dZn inventory, with diatoms having at least 4 times higher cellular-Zn / P ratios than co-occurring cell types (Twining and Baines, 2013).

The fact that both Phaeocystis and diatomaceous nanophytoplankton maintain a contribution to the summer STSW chlorophyll a complement, when dZn availability is low, is intriguing. Both P. antarctica and the large coastal diatoms Thalassiosira pseudonana and Thalassiosira weissflogii have been shown to be growth-limited in culture by free-Zn2+ concentrations ≤10 pM (Sunda and Huntsman, 1992; Saito and Goepfert, 2008). A simple estimate of summer STSW free-Zn2+ availability, based on North Atlantic organic-complexation data (>96 %; Ellwood and Van den Berg, 2000), indicated that free Zn2+ averaged 6.3 ± 5.3 uc pM, suggesting the potential for growth limitation of these phytoplankton. In addition, when comparing the south-eastern Atlantic dZn stoichiometry with the cellular requirements of phytoplankton grown under growth-rate-limiting conditions (Fig. 6), we found summer STSW dZn : PO43- to be in deficit of the requirements of coastal T. pseudonana but not those of the smaller, open-ocean diatom Thalassiosira oceanica. The variation in cellular Zn / P between small and large phytoplankton is related to the higher surface-area-to-volume ratio of smaller cells and the limitation of diffusive uptake rates at low Zn2+ concentrations (Sunda and Huntsman, 1995). This would suggest that the lower dZn availability in summer STSW should influence phytoplankton species composition by selecting for smaller organisms with lower cellular-Zn requirements, confirmed by a ratio of picophytoplankton to nanophytoplankton at least 4 times higher during summer compared with spring values. The comparison further implies that the presence of Phaeocystis and diatoms in summer STSW may be linked with their metabolic Zn–Co–Cd substitution capability, potentially allowing them to overcome some portion of their Zn deficiency. Largely connected to carbonic anhydrase enzymes, several species of eukaryotic phytoplankton are capable of biochemical substitution of Zn, Co or Cd to maintain optimal growth rates under low-trace-metal conditions (Price and Morel, 1990; Sunda and Huntsman, 1995; Lee and Morel, 1995; Lane and Morel, 2000; Xu et al., 2007; Saito and Goepfert, 2008; Kellogg et al., 2020). For example, metabolic substitution of Co in place of Zn has been observed to support the growth of P. antarctica, T. pseudonana and T. weissflogii in media with Zn2+<3 pM (Sunda and Huntsman, 1995; Saito and Goepfert, 2008; Kellogg et al., 2020). Thus, the lower mixed-layer dCo inventory of summer STSW relative to early spring may be in part related to enhanced dCo uptake through biochemical substitution alongside the growth of phytoplankton with distinct Co requirements.

In contrast to Phaeocystis, Emiliania huxleyi-type haptophytes were near-absent in spring STSW (<5 % chlorophyll a; Fig. 5a) and increased in contribution during summer (18 %). E. huxleyi appear to have a biochemical preference for Co over Zn (Xu et al., 2007), which could potentially be a contributing factor to the increased fraction of this haptophyte in summer STSW. Based on Co organic-complexation data for south-eastern Atlantic STSW (>99 %; Bown et al., 2012), however, even the maximum dCo concentration of 56 pM (estimated free Co2+=0.56 ± 0.11 uc pM) observed for STSW during this entire study would limit the growth of cultured E. huxleyi in the absence of Zn or Cd (Sunda and Huntsman, 1995; Xu et al., 2007). This is supported by inter-seasonal dCo / PO43- stoichiometries in deficit of the cellular requirements of cultured E. huxleyi (Fig. 6). Despite this, Xu et al. (2007) showed that E. huxleyi can maintain significant growth at only 0.3 pM Co2+ in the presence of Zn, with the limitation by and substitution of these metals reported to occur over a range of free-ion concentrations (0.2–5 pM) that are relevant to summer conditions of the south-eastern Atlantic. This assessment implies an additional need for Zn in phytoplankton nutrition due to low dCo availability throughout the south-eastern Atlantic, which may accelerate the decrease in dZn / dCo inventory ratio between seasons.

The elevated summer STSW chlorophyll a concentrations were accompanied by increased cell concentrations of the Synechococcus and Prochlorococcus (up to 100 and 400 cells µL-1, respectively) relative to early-spring abundance (Fig. 5b). This pattern suggests an inter-seasonal community shift towards smaller picocyanobacterial cells that is coincident with decreased dZn availability. Synechococcus and Prochlorococcus are thought to have little or no Zn requirement and relatively low Co requirements (growth-limited by ≤0.2 pM Co2+; Sunda and Huntsman, 1995; Saito et al., 2002). This, alongside their small cell size, hence greater capacity for acquiring fixed nitrogen under conditions where this nutrient is depleted, may allow these prokaryotes to flourish following depletion and export of Zn associated with Phaeocystis and diatom blooms. This supposition is supported by a persistently high abundance of Synechococcus and Prochlorococcus (>1000 cells µL-1) relative to eukaryotic nanophytoplankton in the dZn-depleted surface waters of the Costa Rica Dome (Saito et al., 2005; Ahlgren et al., 2014; Chappell et al., 2016). Here, surface dCo concentrations were maintained above those of surrounding waters by the biological production of Co-binding ligands (Saito et al., 2005). The increased abundance of these prokaryotic autotrophs in summer STSW of the south-eastern Atlantic may have also contributed to the inter-seasonal decrease in dCo inventory.

In contrast to STSW, cell counts of eukaryotic phytoplankton and prokaryotic cyanobacteria in SASW varied little between early spring and summer (Fig. 5b), indicative of a more balanced ecophysiological regime. The fractional contribution of Phaeocystis (Fig. 5a), the dominant contributor to the SASW chlorophyll a complement, was similar between transects at 54 % and 44 %, respectively, whilst the contribution of E. huxleyi increased from 19 %–33 % between spring and summer, respectively. Whilst it is proposed that the low Fe supply rate to these waters provides a dominant control on phytoplankton biomass and composition (Browning et al., 2014), low dZn and dCo availability may also be important drivers of such change. The summer SASW dZn inventory (0.08 ± 0.07 uc nM) and stoichiometry with PO43- (Fig. 6) indicate growth-limiting conditions for Phaeocystis and E. huxleyi in the absence of cambialistic metabolism (Sunda and Huntsman., 1995; Saito and Goepfert, 2008; Xu et al., 2007). The presence of these phytoplankton therefore indicates that Zn biochemical substitution occurs in oceanic waters of the south-eastern Atlantic. A lower Co half-saturation growth constant for cultured P. antarctica (Km=∼0.2 pM Co2+) compared with E. huxleyi (Km=∼3.6 pM Co2+) further suggests that Phaeocystis species may more effectively occupy low-dZn and low-dCo environments (Saito and Goepfert, 2008), such as SASW of the South Atlantic.

Conversely, the absence of a significant diatom contribution to summer SASW chlorophyll a (Fig. 5a) relative to early spring is surprising as the summer dZn / PO43- inventory ratio is in excess of the cellular-Zn / P requirements of typical oceanic diatoms such as T. oceanica (Fig. 6). Furthermore, whilst the dCo / PO43- ratio of summer SASW is in deficit of the cellular Co / P below which growth limitation of T. oceanica may occur, this species has been shown to grow effectively at Co2+ < 0.1 pM in culture in the presence of Zn (Sunda and Huntsman, 1995). The low diatom fractional contribution to summer SASW may be instead related to low Fe availability (Browning et al., 2014) and stress-induced Si exhaustion. In support of this, we calculate summer SASW mixed-layer Si concentrations (0.9 ± 0.3 µM) to be 50 % of early-spring values (1.8 ± 0.2 µM), with a dissolved NO3- / Si stoichiometry of 3.8 mol mol-1, close to the 4 mol mol-1 shown to limit diatom growth in culture (Gilpin et al., 2004) and in contrast to the 2.9 mol mol-1 calculated for early spring.