Seasonal cycles of biogeochemical fluxes in the Scotia Sea, Southern Ocean: a stable isotope approach

The biological carbon pump is responsible for much of the decadal variability in the ocean carbon dioxide (CO₂) sink, driving the transfer of carbon from the atmosphere to the deep ocean. A mechanistic understanding of the ecological drivers of particulate organic carbon (POC) flux is key both to the assessment of the magnitude of the ocean CO₂ sink and for accurate predictions as to how this will change with changing climate. This is particularly important in the Southern Ocean, a key region for the uptake of CO₂ and the supply of nutrients to the global thermocline. In this study we examine sediment-trap-derived particle fluxes and stable isotope signatures of carbon (C), nitrogen (N), and biogenic silica (BSi) at a study site in the biologically productive waters of the northern Scotia Sea in the Southern Ocean. Both deep (2000 m) and shallow (400 m) sediment traps exhibited two main peaks in POC, particulate N, and BSi flux: one in austral spring and one in summer, reflecting periods of high surface productivity. Particulate fluxes and isotopic compositions were similar in both deep and shallow sediment traps, highlighting that most remineralisation occurred in the upper 400 m of the water column. Differences in the seasonal cycles of isotopic compositions of C, N, and Si provide insights into the degree of coupling of these key nutrients. We measured increasing isotopic enrichment of POC and BSi in spring, consistent with fractionation during biological uptake. Since we observed isotopically light particulate material in the traps in summer, we suggest physically mediated replenishment of lighter isotopes of key nutrients from depth, enabling the full expression of the isotopic fractionation associated with biological uptake. The change in the nutrient and remineralisation regimes, indicated by the different isotopic compositions of the spring and summer productive periods, suggests a change in the source region of material reaching the traps and associated shifts in phytoplankton community structure. This, combined with the occurrence of advective inputs at certain times of the year, highlights the need to make synchronous measurements of physical processes to improve our ability to track changes in the source regions of sinking particulate material. We also highlight the need to conduct particle-specific (e.g. faecal pellets, phytoplankton detritus, zooplankton moults) isotopic analysis to improve the use of this tool in assessing particle composition of the sinking material and to develop our understanding of the drivers of biogeochemical fluxes.

1 Introduction

The transfer of carbon from the atmosphere to the deep ocean via the biological carbon pump (Volk and Hoffert, 1985) is important for the sequestration of carbon and, combined with ocean circulation, is a main driver of decadal variability in the ocean carbon dioxide (CO₂) sink (DeVries, 2022). A mechanistic understanding of the processes controlling the magnitude and efficiency of the biological carbon pump is therefore key to the assessment and prediction of the ocean's role as a CO₂ sink and requires a robust characterisation of the composition of the sinking particles transferring particulate organic carbon (POC) to the deep ocean. The composition of particles affects the sinking rate, lability, and thus degree of remineralisation as they sink through the water column (e.g. Ploug et al., 2008; Giering et al., 2020).

Sediment traps enable visual assessment of sinking particles and have been deployed in numerous locations throughout the world's oceans to both quantify biogeochemical fluxes and characterise the nature of sinking material (e.g. Torres Valdés et al., 2014). Sediment traps can be susceptible to collection biases depending on the depth of deployment, trap design, hydrodynamic conditions, and properties of sinking particles (Buesseler et al., 2007). Moored sediment traps can underestimate the actual flux at depths shallower than ∼1500 m by collecting only a portion of the sinking material, though biases vary greatly between sites (Buesseler et al., 2007). Numerous studies have recorded the dominance of particular organisms or types of detrital material in trap material, highlighting the importance of ecosystem community structure for the magnitude and efficiency of the biological carbon pump. For example, faecal pellets, diatoms, diatom resting spores, and Acantharea have been observed as significant contributors to particle fluxes (González et al., 2009; Belcher et al., 2018, 2017; Manno et al., 2015; Gleiber et al., 2012; Rembauville et al., 2015; Roca-Marti et al., 2017). Such visual assessment of trap material is typically very time-consuming. Additionally, fragile material, such as salp faecal pellets (Iversen et al., 2017; Pauli et al., 2021), may break up in the sample manipulation processes, making it hard to account for visually. Biogeochemical methods such as the use of stable isotopes may offer additional insight into the drivers of POC fluxes (e.g. Henley et al., 2012).

Marine phytoplankton take up aqueous CO₂ (CO${}_{{\mathrm{2}}_{\text{(aq)}}})$ during photosynthesis, converting it to organic carbon. During this process, the lighter isotope (¹²C) is preferentially assimilated, which enriches the residual aqueous pool in the heavier isotope (¹³C). The stable isotopic composition of the POC (δ¹³C_POC) of the marine phytoplankton is therefore lower than that of the carbon source. Over large scales, the δ¹³C of marine phytoplankton has been found to be inversely correlated with [CO${}_{{\mathrm{2}}_{\text{(aq)}}}$] in surface waters (Rau et al., 1991). However, numerous other factors have been identified as impacting the δ¹³C_POC of surface waters and marine plankton. Phytoplankton growth rates, cell geometry, and non-diffusive uptake of carbon via carbon concentration mechanisms have all been highlighted as impacting the δ¹³C_POC of marine plankton and thus surface waters (Popp et al., 1999, 1998; Bidigare et al., 1999; Trull and Armand, 2001; Tuerena et al., 2019). This decoupling of the relationship between δ¹³C_POC and [CO${}_{{\mathrm{2}}_{\text{(aq)}}}$] presents challenges for palaeoceanographic studies but also the possibility of using the δ¹³C_POC of marine samples to infer information about community composition.

During photosynthetic uptake, the balance between supply and demand of carbon impacts δ¹³C_POC, regulated by the transport of inorganic carbon into the internal cell and fixation to organic carbon (Popp et al., 1999; Trull and Armand, 2001). A greater isotopic fractionation occurs in smaller phytoplankton cells, enabled by the higher cell-surface-area-to-volume (SA:V) ratios and increased amount of [CO${}_{{\mathrm{2}}_{\text{(aq)}}}$] diffusing across the cell membrane relative to the total carbon within the cell (Popp et al., 1998; Tuerena et al., 2019; Hansman and Sessions, 2016). Thus, a community dominated by large, fast-growing diatoms is expected to contribute to enriched δ¹³C_POC values compared to a community dominated by picoplankton. A study by Henley et al. (2012) in the coastal western Antarctic Peninsula attributed a large (∼10 ‰) negative isotopic shift in δ¹³C_POC to a near-complete biomass dominance of the marine diatom Proboscia inermis, highlighting the possible impact of shifts in species composition on stable isotopes. It may therefore be possible to use stable isotopes to gain information about the community composition of phytoplankton driving, for example, large spring pulses in POC flux. Additionally, siliceous phytoplankton, such as diatoms, require dissolved silica (silicic acid, DSi) to build their cell walls or frustules (amorphous SiO₂⋅nH₂O, referred to here as biogenic silica, BSi). During the uptake of DSi, diatoms fractionate the stable isotopes of silicon (²⁸Si, ²⁹Si, ³⁰Si), preferentially taking up the lighter isotopes during cell wall (frustule) formation (De La Rocha et al., 1997). This means that BSi fluxes and ratios of light ²⁸Si to heavy ³⁰Si (expressed as δ³⁰Si) in sinking particulate organic matter (POM) can be informative about DSi utilisation by siliceous phytoplankton. The fractionation of Si isotopes during diatom DSi utilisation is approximately −1.1 ‰, although estimates of this value vary in laboratory and field studies between −0.5 ‰ and −2.5 ‰ (Hendry and Brzezinski, 2014). Whilst some studies have shown that isotopic fractionation is independent of temperature, DSi concentration, and diatom species (e.g. De La Rocha et al., 1997), one in vitro laboratory culture experiment revealed a potential species effect, with polar species exhibiting more extreme fractionation (−2.09 ‰ for Chaetoceros sp. and 0.54 ‰ for Fragilariopsis kerguelensis; Sutton et al., 2013). The impact of water column dissolution on frustule δ³⁰Si is poorly constrained, with experimental evidence for either a small fractionation of −0.55 ‰ (Demarest et al., 2009) or a negligible impact (Wetzel et al., 2014; Egan et al., 2012; Grasse et al., 2021).

Additionally, the stable isotopes of marine nitrogen reveal information about the uptake of inorganic nitrogen sources by phytoplankton (Wada and Hattori, 1978) as well as trophic and food web processes (Michener and Lajtha, 2008). Nitrogen has two isotopes, ¹⁴N and ¹⁵N, and the ratio between these heavy and light isotopes is expressed as δ¹⁵N. Different sources of nitrogen can alter the stable isotopic composition of marine phytoplankton because ammonium characteristically has a lower value of δ¹⁵N than nitrate supplied from depth. As well as this, isotopic fractionation occurs during transfer through the food web, with a trophic enrichment of typically 2 ‰–4 ‰ between successive trophic levels (Montoya, 2007; Minagawa and Wada, 1984). Excretion and egestion processes can also impact δ¹⁵N; isotopic discrimination during the excretion of ammonium by zooplankton and fish results in ammonium that is ¹⁵N-depleted relative to the substrate catabolised (Montoya, 2007). Thus, there are several interacting processes impacting the degree of fractionation and subsequent isotopic ratios in particulate nitrogen (PN), and knowledge of δ¹⁵N ratios may provide insight into biogeochemical processes and the composition of the sinking flux.

In this study we examine the seasonal cycle of the magnitude and composition of vertical biogeochemical fluxes of particulate material collected by two sediment traps deployed for almost 1 year on a deep ocean mooring located in the northern Scotia Sea in the Atlantic sector of the Southern Ocean. The Scotia Sea, particularly the region downstream of South Georgia, is a hot spot for biological productivity, supported by higher iron availability (Korb et al., 2008; Matano et al., 2020). Diatoms dominate the phytoplankton assemblage, particularly in the summer months, with smaller contributions of dinoflagellates (Korb et al., 2012). The large, consistent phytoplankton blooms occurring in this region support high fluxes of POC to the deep ocean, with two peaks in POC flux occurring during the seasonal cycle: the first peak in austral spring and the second in late summer or early autumn (Manno et al., 2015). Faecal pellets (up to 91 % in late spring and early summer; Manno et al., 2015), krill exuviae (up to 47 % in summer; Manno et al., 2020), and diatoms, particularly resting spores (annual contribution of 42 %; Rembauville et al., 2016) have been shown to make large contributions to the POC fluxes in our study region. Here we use δ¹³C_POC, δ¹⁵N_PN, and δ³⁰Si_BSi alongside calculated fluxes of POC, PN, and BSi as tools to reveal information about sinking particulate organic matter and the processes influencing its production and subsequent flux to depth. A greater in-depth understanding of the composition and thus the drivers of POC flux in this important region is key to improving estimates of the current and future strength of the biological carbon pump and the ocean's role as a CO₂ sink.

2 Methods

2.1 Study area

This study was conducted in the open-ocean environment of the northern Scotia Sea in the Southern Ocean at a long-term observatory station, P3 (Fig. 1), where an oceanographic mooring is located. The mooring is part of the Scotia Sea Open Ocean Observatory (SCOOBIES: https://www.bas.ac.uk/project/scoobies/, last access: 10 August 2022), a programme designed to investigate the biological and biogeochemical influence of the large and persistent phytoplankton bloom to the northwest of South Georgia.

Figure 1 Location of P3 mooring site to the northwest of South Georgia. White lines indicate frontal positions of the Antarctic Polar Front (APF) (Moore et al., 1999), Southern Antarctic Circumpolar Current Front (SACCF) (Thorpe et al., 2002), and the southern boundary of the Antarctic Circumpolar Current (SB-ACC) (Orsi et al., 1995). Mean chlorophyll concentration (mg m⁻³) is shown for December 2018 from 8 d satellite chlorophyll data from the Ocean Colour CCI (version 5.0) (Sathyendranath et al., 2021, 2019).

2.2 Sediment trap deployment

Two sediment traps were deployed on the mooring array to collect sinking particles for the analysis of carbon, nitrogen, and biogenic silica fluxes and the analysis of δ¹³C_POC, δ¹⁵N_PN, and δ³⁰Si_BSi. The mooring was deployed from 25 January 2018, during research cruise JR17002 aboard the RRS James Clark Ross, to 1 January 2019, recovered during research cruise DY098 aboard the RRS Discovery. The mooring was located at 52.8036^∘ S, 40.1593^∘ W, to the northwest of the island of South Georgia in the Scotia Sea at a water depth of 3748 m. Sediment traps (McLane PARFLUX, 0.5 m² surface collecting area; McLane Research Laboratories Inc, Falmouth, MA, USA) were deployed at 400 and 2000 m (referred to hereafter as shallow and deep respectively) and were each equipped with 21 sample bottles. A baffle at the top of the trap prevented large organisms from entering and each sample bottle contained a formosaline solution (filtered seawater containing 2 % $v/v$ formalin, mixed with sodium tetraborate (BORAX; 0.025 % $w/v$), and 0.5% $w/v$ sodium chloride) to prevent mixing with the overlying water column and stop biological degradation. Previous studies have reported the effects of formalin on δ¹³C_POC and δ¹⁵N_PN to be small (± 1 ‰ and ±1.5 ‰ respectively; Mincks et al., 2008, and references therein). This equates to 13 % and 16 % of the maximum range measured in our study, which is small compared to the isotopic shifts we observed. Yet we stress that all δ¹³C_POC and δ¹⁵N_PN values given here are associated with this uncertainty. The sediment trap sample carrousel was programmed to rotate every 7–31 d depending on the season: shorter periods to coincide with austral summer and longer periods during austral winter (Table S1 in the Supplement). Seaguard™ current meters were deployed ∼50 m above the shallow sediment trap and 50 m below the deep sediment trap, set at a measurement interval of 2 h.

2.3 Trap sample processing

Each sample bottle from the sediment trap was processed on return to the laboratory. The supernatant was carefully removed using a syringe, and swimmers (zooplankton that are believed to have entered the trap actively whilst alive) were removed. Swimmers were removed by hand under a dissecting microscope and were not included in flux calculations. The material from each sediment trap sample bottle was split into a number of smaller aliquots for subsequent analysis using a McLane rotary splitter.

2.4 Organic carbon and nitrogen

For each sediment trap bottle from both deep and shallow traps, two or three splits were taken and each was analysed for POC and PN mass and δ¹³C_POC and δ¹⁵N_PN. Once split, the material was filtered onto pre-combusted (450 ^∘C, 16 h) 25 mm glass fibre filters (GF/F; nominal pore size 0.7 µm) and rinsed with Milli-Q water. Samples were air-dried and fumed for 24 h with 37 % HCl in a desiccator before finally being oven-dried at 50 ^∘C for 24 h. Filters and filter blanks were placed in sterile tin capsules, and POC and PN were measured on a CE Instruments NA2500 elemental analyser, calibrated using an acetanilide calibration standard with a known % C and % N of 71.09 % and 10.36 % respectively. Standards were interspersed regularly between samples to measure and correct for drift. Analytical precision was better than 1.0 % for POC and 1.1 % for PN. The POC flux (F, mg C m⁻² d⁻¹) for each sample was calculated using the following equation:

$$\begin{array}{}\text{(1)}& F=m/\left(A\times d\right).\end{array}$$

Here m is the mass of POC in the sample bottle (mg), d is the number of days that the sample bottle was open (7–31 d), and A is the surface area of the sediment trap opening (0.5 m²). The same calculation was carried out for PN.

δ¹³C_POC and δ¹⁵N_PN were analysed on a Thermo Finnigan Delta-V Advantage isotope ratio mass spectrometer that was in line with the elemental analyser. All δ¹³C_POC and δ¹⁵N_PN data are presented in the delta per mille (‰) notation relative to the appropriate international standard, according to Eq. (2).

$$\begin{array}{}\text{(2)}& \mathit{\delta }X\left(\mathrm{‰}\right)={\mathrm{10}}^{\mathrm{3}}\left(R_{\text{sample}}/R_{\text{standard}}-\mathrm{1}\right)\end{array}$$

R denotes the ¹³C $/$ ¹²C ratio for carbon or the ¹⁵N $/$ ¹⁴N ratio for nitrogen. R_sample refers to the relevant ratio in the sample. R_standard refers to the ratios in the international standards Vienna Pee Dee Belemnite (V-PDB) for δ¹³C and atmospheric nitrogen (AIR) for δ¹⁵N, both of which are calibrated against the PACS-2 marine sediment reference material. Multiple repeats of analytical standards gives a reproducibility of 0.2 ‰ for C and N, which is significantly smaller than the uncertainty associated with organic molecules in the formalin preservative (±1 ‰ and ±1.5 ‰ for C and N respectively; Mincks et al., 2008, and references therein).

2.5 Biogenic silica

Two splits were taken from each sample bottle from both deep and shallow sediment traps for the analysis of biogenic silica and silicon isotopes. Split material was filtered onto 25 mm, 0.4 µm polycarbonate filters and rinsed with Milli-Q water before drying at 50 ^∘C for 24 h. Material on the filters was solubilised via an alkaline extraction method (Hatton et al., 2019) carried out at the Bristol Isotope Group (BIG) laboratory. Sample material was digested in Teflon tubes with 0.2M NaOH at 100 ^∘C for 40 min. This was followed by neutralisation with 6M HCl. Biogenic silica (SiO₂, termed BSi) concentrations were measured chlorometrically by molybdate blue spectrophotometry (heteropoly blue method) (Strickland and Parsons, 1972) using a Hach DR3900 spectrophotometer set at a wavelength of 815 nm. Supernatants were stored for 7–11 months before column chemistry for isotope analysis. Fluxes of biogenic silica were calculated as for POC using Eq. (1).

For Si isotope analysis, supernatants and reference materials were purified by passing through cation exchange columns (Bio-Rad AG50W-X12, 200–400 mesh resin) pre-cleaned with HCl following Georg et al. (2006). Samples were acidified to a pH of 1–2 to ensure that all the silicon remained in solution. Samples were loaded onto columns and eluted with Milli-Q water to produce a 2.5 ppm solution, and concentrations were checked to confirm quantitative yields. Si isotopic composition was analysed within 24 h of column chemistry. Stable Si isotopic compositions are presented in standard delta notation (δ³⁰Si), as for δ¹³C_POC and δ¹⁵N_PN according to Eq. (2), where R is ³⁰Si $/$ ²⁸Si. These compositions are checked against δ²⁹Si (where R is ²⁹Si $/$ ²⁸Si) for mass dependence. The samples were measured at the BIG laboratory on a Finnigan Neptune Plus high-resolution MC-ICP-MS (Thermo Fisher Scientific). The Si solutions were spiked with magnesium spike (Inorganic Ventures MSMG 10 ppm), hydrochloric acid (1M HCl in-house distilled), and sulfuric acid (0.1 M H₂SO₄, ROMIL-UpA™ Ultra Purity sulfuric acid) and transferred from the autosampler via a PFA Savillex C-Flow nebuliser (35 µL min⁻¹) connected to an Apex IR desolvating nebuliser (Ward et al., 2022) and measured on the low-mass side to resolve any isobaric interferences (e.g. ¹⁴N¹⁶O⁺). All standards and samples were blank-corrected offline. The intensity of ²⁸Si in the 0.1 M HCl blank was <1 % of the sample intensity in all sample runs. Furthermore, we also measured Mg isotopes (²⁴Mg, ²⁵Mg, and ²⁶Mg) as an internal isotopic reference to correct for any mass-dependent fractionation (Cardinal et al., 2003). Measurements that resulted in large corrections (>0.3 ‰ on δ³⁰Si) underwent repeat analysis. Instrumental mass bias was further accounted for using a standard-sample bracketing method using a 2 ppm reference standard (NBS or RM8546) solution. Two splits were analysed for each sediment trap bottle as well as standards and sample blanks. Solutions obtained from each split were measured in replicate (n=2–3) alongside the continuous measurement of reference materials diatomite and LMG-08 to ensure reproducibility and to monitor data quality. Measurements of diatomite and LMG-08 yielded δ³⁰Si of +1.23 ‰ (SD ± 0.03, n=18) and −3.40 ‰ (SD ± 0.05, n=5) respectively, which agreed with published values (Reynolds et al., 2007; Hendry and Robinson, 2012; Grasse et al., 2017). Typical reproducibility between the sediment trap sample splits (coming from the same sediment trap bottle) was 0.034 ‰ (1 ± SD). A lithogenic correction (e.g. Closset et al., 2015) was not carried out on these samples given the high percentage of biogenic silica present in the samples (mean percentage BSi as SiO₂ of 17 %). BSi extraction methods show lower variability for marine sediments with BSi >15 %–20 % and do not show evidence of significant leaching of lithogenic material through time (Conley, 1998). However, even an extreme scenario of variable lithogenic contamination of 1 %–5 % of isotopically light marine clays (with δ³⁰Si of −2.3 ‰; Opfergelt and Delmelle, 2012) would only result in a potential systematic offset of 0.12 ‰, which, although this is larger than the uncertainty on an individual data point, is 1 order of magnitude smaller than the observed seasonal signal.

2.6 Chlorophyll and phytoplankton community composition

Surface chlorophyll concentrations were obtained from satellite-derived 8 d Ocean Colour CCI (version 5.0) (Sathyendranath et al., 2021, 2019). We present the monthly mean of these 8 d data for December at our study site (Fig. 1) as well as the 8 d chlorophyll concentration data from September 2017 to December 2018 (Fig. 2) averaged over a $\mathrm{1}\times \mathrm{1}{}^{\circ }$ bounding box around our study site (53–52^∘ S, 41–40 ^∘ W).

Light microscopy was used to assess phytoplankton and microzooplankton community composition of a small selection of samples from the two main productive periods. A biological method of sample preparation and analysis was chosen, comparable with Rembauville et al. (2015), to determine the quantity of empty and full cells. Following subsampling using the rotary splitter, samples for morphological taxonomic analysis were diluted to a standardised 25 mL. Samples were gently inverted using the Paul Schatz principle (figures of eight) for 1 min to homogenise them, and 2 mL was withdrawn using a modified pipette with widened opening. Several common diatoms in Antarctic waters are long and slim; in particular, Thalassiothrix antarctica has been recorded with an apical axis of up to 5 mm. To ensure such specimens remain intact and are not excluded from the pipetting process, a wide bore opening was used. The 2 mL subsamples were used to fill a 1 mL Sedgwick–Rafter counting chamber. Chambers were viewed using a compound light microscope (Nikon Eclipse 80i) with differential interference contrast at x200 magnification. For the larger, easily identifiable cells, the whole chamber was observed; for smaller cells a proportion of the chamber was examined depending upon cell abundance (at least 500 cells were counted). Only complete cells were enumerated to avoid over counting of fragmented specimens. Cells were determined as “full” or alive at time of collection if they possessed chloroplasts/plastids, pigment, a nucleus, or, in the case of Pronoctiluca, a distinct accumulation body; cells lacking these internal features were deemed as “empty” or dead at time of collection. Specimens were identified according to Hasle and Syvertsen (1997), Medlin and Priddle (1990), Priddle and Fryxell (1985), and Scott and Marchan (2005).

Cell biovolume and surface area estimates were calculated using geometrics and the appropriate shape-related equations for phytoplankton genera proposed by Hillebrand et al. (1999). Metrics used in the calculations were based on the average size of 10 randomly selected specimens belonging to a species or other taxonomic group within the samples.

3 Results

3.1 Environmental conditions

Mean current velocities were 0.11 (±0.06) and 0.06 (±0.03) m s⁻¹ for shallow and deep current meters respectively (Supplement Fig. S1). Maximum current speeds recorded reached 0.43 and 0.18 m s⁻¹ for shallow and deep meters respectively. The periods with currents substantially elevated above the mean were June for both traps and additionally in late August and September for the shallow trap, both for periods of ∼5–10 d. Both are periods of low fluxes during austral winter and are not the main subject of the study here, though it is likely that particle collection was biased at these times (Buesseler et al., 2007).

Satellite-derived estimates of surface chlorophyll show high concentrations during austral summer (January to March) peaking at 2.3 mg m⁻³, as well as during spring (November–December), peaking at 2.1 mg m⁻³ (Figs. 2, S2). Data coverage is limited in the winter due to cloud cover, but concentrations appear to be <0.4 mg m⁻³. We define here two productive periods (when chlorophyll concentrations were >0.4 mg m⁻³), which we refer to throughout the paper: productive period 1 – January to the start of April 2018; productive period 2 – September to the end of December 2018. We note that our sediment trap data begin on the 25 January, so we do not capture the start of period 1.

Figure 2 Seasonal cycle of satellite-derived surface chlorophyll concentration (green line, 8 d data from the Ocean Colour CCI (version 5.0 (Sathyendranath et al., 2021, 2019)). Day length at 53^∘ S is shown by the orange line. The two productive periods are highlighted by the shaded green regions, and the grey shaded bar shows the duration of the sediment trap sampling period.

3.2 POC, PN, and BSi fluxes

There is a clear seasonal cycle in POC, PN, and BSi fluxes, all tracking each other well (Fig. 3). Since two to three splits were analysed from each sediment trap bottle, we refer here to the mean flux for each sediment trap bottle based on the available splits for that bottle. POC fluxes were low during austral autumn and winter, with fluxes <10 and <7 mg C m⁻² d⁻¹ for shallow and deep traps respectively during the period March to October 2018 (Fig. 3a). Higher fluxes were measured in summer 2018 (productive period 1), reaching 25.3 mg C m⁻² d⁻¹ in late January 2018 in the shallow trap and 13.1 mg C m⁻² d⁻¹ in late February in the deep trap. The maximum POC fluxes measured occurred in early December 2018 (productive period 2), reaching 45.7 and 43.4 mg C m⁻² d⁻¹ in shallow and deep traps respectively. PN fluxes follow the same trends as POC fluxes, peaking at 4.2 and 2.4 mg N m⁻² d⁻¹ during period 1 and 10.8 and 8.2 mg N m⁻² d⁻¹ during period 2 in shallow and deep traps respectively (Fig. 3b). The mean POC:PN ratio (mol : mol) throughout the study period was 6.40 (±0.73) and 6.02 (±0.90) in shallow and deep traps respectively, with higher ratios in the productive periods compared to the winter months. Mean POC:PN ratios were 6.83 (±0.48) and 6.63 (±0.71) during period 1 and period 2 in the shallow trap and 6.40 (±0.63) and 5.51 (±0.87) in the deep trap. Over the winter months POC:PN was 5.83 (±0.54) and 6.26 (±0.87) in shallow and deep traps respectively.

BSi fluxes (Fig. 3c) track those of POC well. The lowest fluxes (<20 mg SiO₂ m⁻² d⁻¹) occurred in the autumn and winter (March–October), with the exception of a small peak of up to 39.7 mg SiO₂ m⁻² d⁻¹ in May 2018. During summer 2018 (productive period 1), BSi fluxes were high, reaching 129.1 mg SiO₂ m⁻² d⁻¹ in early February in the shallow trap and 84.3 mg SiO₂ m⁻² d⁻¹ in late February in the deep trap. By far the highest fluxes were observed in spring 2018 (productive period 2), peaking in early December at 562.4 and 285.4 mg SiO₂ m⁻² d⁻¹ in shallow and deep traps respectively. The mean BSi : POC ratio (mol : mol) throughout the study period was 29.82 (±17.80) and 25.86 (±11.72) in shallow and deep traps respectively. Higher BSi : POC ratios were observed in the shallow trap in period 1 (38.45 ± 10.96), and both shallow and deep traps in period 2 (36.94 ± 16.32 and 35.70 ± 12.10 respectively). BSi : POC ratios were lower in the deep trap during period 1 (23.64 ± 6.82). The correspondence in timing of the elevated fluxes of POC, PN, and BSi fluxes in the shallow and deep traps in spring (period 2) highlights that sinking rates must be sufficient (>114 m d⁻¹) for particles to travel the 1600 m between the two traps in the 14 d period that those sediment trap cups were open. In period 1, there was a time lag of 14 to 35 d between the timing of the maximum POC, PN, and BSi fluxes in the deep and shallow sediment traps. This suggests sinking rates of 46–114 m d⁻¹. However, we stress that this assumes vertical sinking, which as we discuss in Sect. 4 is not always the case.

Figure 3 (a) Particulate organic carbon (POC), (b) particulate nitrogen (PN), and (c) biogenic silica (SiO₂, BSi) fluxes (mg m⁻² d⁻¹) at deep (red shading) and shallow (blue shading) sediment traps. Shading indicates the maximum and minimum flux from two splits, with the solid line indicating the mean value. Coloured points show isotope ratios for (a) δ¹³C_POC, (b) δ¹⁵N_PN, and (c) δ³⁰Si_BSi with red open triangles and blue filled circles indicating deep and shallow sediment traps respectively. The legend shown in the top left hand corner of panel a applies to all panels. The maximum error in sediment trap δ¹³C_POC (±1 ‰) and δ¹⁵N_PN (±1.5 ‰) values is shown by scaled error bars in the bottom right corner and is associated with formaldehyde preservation (Mincks et al., 2008) since this vastly exceeds analytical error. For δ³⁰Si_BSi, the scaled error bar represents 2 ± SD (0.07 ‰) for the analytical sample replicates. For each sample, isotope ratios are given at the midpoint of the period that the sample cup was open.

3.3 δ¹³C_POC, δ¹⁵N_PN, and δ³⁰Si_BSi isotopes

δ¹³C_POC values of deep and shallow sediment trap samples track each other well and show the same order of enrichment and depletion (Fig. 3a). When describing the results for an individual sediment trap bottle, we give the mean of replicate splits from that sediment trap bottle unless otherwise stated. Initially, from January to March 2018, we see isotopically light δ¹³C_POC values between −27.40 ‰ and −28.56 ‰, before increasing to −24.38 ‰ and −25.07 ‰ in June in shallow and deep traps respectively. Over winter, δ¹³C_POC became more depleted (shallow −25.76 ‰ in October; deep −27.07 ‰ in August) with a slight divergence (2.17 ‰) in the tracking of deep and shallow δ¹³C_POC in August 2018. Coinciding with increasing chlorophyll concentrations, δ¹³C_POC became more enriched during the period September to December 2018 (−25.72 ‰ to −21.13 ‰ and −26.04 ‰ to −21.98 ‰ for shallow and deep traps respectively).

Comparison of flux-weighted δ¹³C_POC values confirms the carbon isotopic similarity of deep and shallow traps, particularly during period 2 (Table 1). These results also highlight the shift in both δ¹³C_POC and δ³⁰Si_BSi between period 1 and period 2.

Table 1 Sediment trap seasonal (January 2018–December 2018), period 1 (January 2018–start of April 2018), period 2 (September 2018–end of December 2018), and winter (April–end of August) flux-weighted mean δ¹³C_POC (‰), δ¹⁵N_PN (‰), and δ³⁰Si_BSi (‰) for shallow (400 m) and deep (2000 m) traps. Given that the analytical conditions were the same for all samples measured, we use the pooled variance over the applicable time period as a measure of uncertainty in these mean isotopic ratios. Degrees of freedom (dof) are based on cups with replicate isotopic measurements and are given in parentheses.

^∗ There were no replicates from the deep sediment trap sample bottles for Si isotopes during this period.

δ¹⁵N_PN values are less consistent between deep and shallow sediment trap samples, and there is more heterogeneity between sample splits. For the shallow trap we see values ranging between +0.13 ‰ and +2.96 ‰ (mean +1.42 ‰; SD 0.79 ‰) from January to June 2018, and, for the deep trap, values ranged between −1.95 ‰ and +3.04 ‰ (mean +0.60 ‰; SD 1.60 ‰) during this period. Values increase between June and August, reaching +5.42 ‰ and +2.10 ‰ in shallow and deep traps respectively. From August to December (shallow) and August to November (deep), we see a trend of decreasing δ¹⁵N_PN to +1.49 ‰ and −2.77 ‰ in shallow and deep traps respectively, with the decrease being of a similar magnitude (3.93 ‰ and 4.87 ‰ respectively) for both traps. Shallow δ¹⁵N_PN is consistently higher than deep δ¹⁵N_PN by 4.52 ‰ on average during this period (August to November). In the deep trap we see a final increase in δ¹⁵N_PN coinciding with the increase in PN flux from November to December 2018, reaching a mean of +0.71 ‰. The same increase in δ¹⁵N_PN is not apparent in the shallow trap.

Si isotope compositions in deep and shallow samples were quite similar, exhibiting the same seasonal patterns. Both deep and shallow traps showed an increase in δ³⁰Si_BSi from January to July 2018 (+0.71 ‰ to +1.24 ‰ in the deep trap and +0.24 ‰ to +1.26 ‰ in the shallow trap) with the steepest increase occurring from March to May (Fig. 3c). Sample splits generally showed good agreement with one exception during March 2018 when sample splits from the deep sediment trap were +0.52 ‰ and +1.41 ‰, highlighting the heterogeneous nature of the sediment trap material. Isotopic values in the deep trap were then quite steady over winter compared to the rest of the record, with an increase of 0.38 ‰ in the shallow trap between May and August. At the end of August, δ³⁰Si_BSi began to decrease steeply, reaching +0.68 ‰ and +0.86 ‰ in shallow and deep traps respectively in November 2018. Following this, δ³⁰Si_BSi increased rapidly to +1.72 (deep) and +1.89 ‰ (shallow) coinciding with the large increase in BSi fluxes at this time.

3.4 Phytoplankton community structure

Eight samples (four deep and four shallow, Table 2) were analysed by light microscopy for phytoplankton composition to cover the high-productivity periods 1 and 2. Diatoms, silicoflagellates, and dinoflagellates were observed, with a dominance of diatoms (>85 % by both abundance and biovolume). Microzooplankton were also recorded, in particular Radiolaria and tintinnids, though these were not dominant by biovolume or abundance. Only intact cells were identified and counted. In terms of abundance, during period 1, the diatoms Fragilariopsis spp. dominated both deep (58 %–66 %) and shallow (∼70 %) trap samples (Fig. 4a, c), whereas during period 2 the phytoplankton community structure was more mixed, with contributions from the diatoms Thalassionema nitzschioides, Chaetoceros small (<20 µm) centrics, and Fragilariopsis spp. Large centric diatoms (>20 µm) represented 15 %–20 % of the community by abundance in the deep trap during productive period 1 but <2.5 % in productive period 2. Interestingly we do not see these large centrics in the shallow trap during productive period 1, implying that sinking velocities were <76 m d⁻¹ for these large phytoplankton cells based on the duration that the first sediment trap bottle was open and the depth between the two traps.

In terms of biovolume, Fragilariopsis spp. were still a dominant component of the shallow trap sample in period 1 (∼33 %) but were <9 % of the community in the deep trap during period 1, with the large cells of the diatom Coscinodiscus dominating with 39 %–67 % (Fig. 4b, d). Diatoms, Corethron pennatum (shallow: 10 %–13 %; deep: 15 %), Rhizosolenia (shallow: 9 %–21 %), and large centric diatoms (>20 µm) (shallow: 10 %–17 %; deep: 16 %–20 %) as well as the silicoflagellate Dictyocha (shallow: 9 %–10 %; deep: 8 %) were also relatively high in terms of biovolume during period 1. During period 2, the community was quite mixed in terms of biovolume in the shallow trap (Fig. 4b). The deep trap had similar contributions from Fragilariopsis spp. (22 %–28 %), Dictyocha (14 %–15 %), Coscinodiscus (10 %), and small (<20 µm, 9 %–14 %) and large (>20 µm, 9 %–19 %) centric diatoms during period 2. Since there has been little known work on the δ³⁰Si of Dictyocha or indeed other silicoflagellates, we are not able to constrain the impact of this organism on our measured values. However, since the contribution by abundance was >5 % and diatoms were dominant (>85 %), their isotopic signature would need to be vastly different from that of diatoms to have an appreciable impact on our results.

Figure 4 Phytoplankton assemblage of (a, b) shallow and (c, d) deep sediment trap samples, according to abundance (a, c) and biovolume (b, d). Plots (a) and (c) show phytoplankton contributing >5 % by abundance, and plots (b) and (d) show phytoplankton contributing >5 % by biovolume. Other refers to all other counted taxa combined. Four samples were identified taxonomically for each trap. Green shading highlights productive periods 1 and 2, as per Fig. 2. Note that only intact cells were counted.

4 Discussion

In this study we measure the seasonal cycle of POC, PN, and BSi fluxes as well as the δ¹³C_POC, δ¹⁵N_PN, and δ³⁰Si_BSi values of sinking particles collected in shallow (400 m) and deep (2000 m) sediment traps in the Scotia Sea, Southern Ocean. Both the magnitude of fluxes and isotopic compositions were generally similar in the shallow and deep sediment traps, suggesting that most remineralisation occurred in the upper 400 m. This highlights that material reaching 400 m likely facilitates the transfer of carbon much deeper in the ocean, sequestering carbon for longer time periods (Kwon et al., 2009).

4.1 Seasonal flux cycles

The seasonal cycles of POC agree well with previously published work at the same location (Manno et al., 2015), with peaks in austral spring and late summer, though the peak POC fluxes recorded here (45.7 and 43.4 mg C m⁻² d⁻¹ in shallow and deep traps respectively) are higher than those observed in previous years (22.9 mg C m⁻² d⁻¹; Manno et al., 2015). A smaller additional peak in POC flux (<10 mg C m⁻² d⁻¹) occurred in April and May, in agreement with some previous years (Manno et al., 2015). PN fluxes followed the same seasonal trend as POC for both deep and shallow traps suggesting a similar source. The similar magnitude of POC:PN ratios in period 1 in the two traps supports consistency in the degree of degradation at these depths. The lower POC:PN ratios measured in the deep trap between August and October compared to the shallow trap are consistent with a divergence in δ¹⁵N_PN ratios and could indicate that material arriving at the two traps is not necessarily sourced from the same region and time period in surface waters. Given the slower sinking speeds at this low-productivity time of year, it is possible that material reaching the deep trap is sourced from upstream of where material reaching the shallow trap is sourced in the regional circulation system. Different source regions are likely characterised by different phytoplankton assemblages with different nutrient stoichiometry, and the time taken for source material to reach each of the traps may well lead to differences in the degradation state of organic matter, which could also lead to variations in POC:PN.

Our measured fluxes of BSi are higher than previously observed at this site at 2000 m (Rembauville et al., 2016). Maximum fluxes of 46.0 mg SiO₂ m⁻² d⁻¹ were recorded by Rembauville et al. (2016) in January 2012, which though of similar magnitude to our summer peak of 84.3 mg SiO₂ m⁻² d⁻¹ is 1 order of magnitude lower than the spring peak of 285.4 mg SiO₂ m⁻² d⁻¹ in December 2018. However, the Rembauville et al. (2016) record ends in November and therefore would not have captured the main peak in particle flux following the phytoplankton spring bloom in December (apparent in satellite surface chlorophyll; Fig. 2 in Rembauville et al., 2016). Additionally, we do not capture the first 3 weeks of January in our data. Interannual variability in export flux can be high due to the complexity of processes controlling the magnitude of export flux, such as community structure, nutrient limitation, and zooplankton activity. Closset et al. (2015) measured very high fluxes (>700 mg SiO₂ m⁻² d⁻¹) of BSi south of the Sub-Antarctic Front in the Australian sector of the Southern Ocean at 2000 m, and similarly high fluxes have been observed in other sectors (Fischer et al., 2002; Honjo et al., 2000). A study by Trull et al. (2001) measured fluxes of BSi in the range of 30–160 mg SiO₂ m⁻² d⁻¹ during the productive season in the same region as Closset et al. (2015), again highlighting the high interannual variability.

We define two main productive periods: productive period 1 from January to the start of April 2018 and productive period 2 from September to the end of December 2018 when chlorophyll concentrations were >0.4 mg m⁻³. Satellite data suggest the magnitude of chlorophyll concentration was similar during both productive periods but increasing in magnitude throughout period 2 and decreasing in period 1, consistent with the timing of sampling. The particle fluxes associated with productive period 2 were much higher than those during productive period 1, a difference that is particularly pronounced for BSi fluxes. The bloom during period 2 was more geographically widespread (Fig. S2), and thus it is possible that if more of the material reaching the trap was sourced from productive waters, this could have supported the higher fluxes observed at this time. The observed higher BSi fluxes in productive period 2 could also relate to the presence of more heavily silicified diatom species at this time, including the occurrence of resting spores (Chaetoceros spp.; Fig. 4 and Rembauville et al., 2016), increased aggregation (and thus sinking) potential, higher sinking rates, and/or reduced grazing pressure. The fact that we observed resting spores at the end of productive period 2 suggests that nutrients may have started to become limiting for at least some of the phytoplankton community (e.g. silicic acid and/or iron; Rembauville et al., 2016). POC and BSi fluxes track each other closely, and ratios suggest substantial export of biogenic silica (Fig. 5). This, combined with our visual observations of a dominance of phytoplankton material in the trap during the spring peak that was dominated by diatoms (Fig. 4), suggests an important role for diatoms in transferring organic carbon to the deep ocean at this time. This could be achieved if cells are large, through large mineral (silica) ballasted cells sinking at high velocities (Baumann et al., 2022) or through the bioprotection of internal organic matter from grazing and oxidation by the diatom silica frustules (Passow and De La Rocha, 2006; Armstrong et al., 2001; Smetacek et al., 2004).

4.2 Seasonal variations in isotope ratios

Despite the strong relationship between particulate fluxes of POC and BSi, the relationship between the δ¹³C_POC and δ³⁰Si_BSi isotope signatures is less pronounced (linear regression: R²=0.427, p<0.001; Fig. 5). This may relate to greater variation in the fractionation factor for δ¹³C compared to δ³⁰Si (Brandenburg et al., 2022) as well as differences in remineralisation of organic carbon and silicon in the frustule. Additionally, whereas most of the δ³⁰Si signal is from diatoms, the δ¹³C signal in the sediment trap material is also impacted by the presence of other organic material, e.g. zooplankton faecal pellets. We do not find significant relationships between δ¹⁵N_PN and δ¹³C_POC (p=0.63) or δ³⁰Si_BSi (p=0.60). We discuss results for each of the three main periods: productive period 1 (first export event), the winter flux hiatus, and productive period 2 (second export event).

Figure 5 Relationship between BSi and POC for data from both deep (triangles) and shallow (squares) sediment traps. (a) Regression between BSi and POC fluxes and (b) between δ¹³C_POC and δ³⁰Si_BSi. Regression lines are shown by dashed lines with coefficients and associated standard errors also shown. Error bars on isotope values represent the maximum error on sediment trap δ¹³C_POC (±1 ‰) associated with formaldehyde preservation (Mincks et al., 2008), and for δ³⁰Si_BSi, the scaled error bar represents 2 × SD (0.07 ‰) for the analytical sample replicates.

4.2.1 Productive period 1

During productive period 1, δ¹³C_POC is low, averaging −28.59 ‰ and −27.24 ‰ in shallow and deep traps respectively, close to that expected for Southern Ocean phytoplankton a employing typical C3 metabolism (i.e. diffusive CO₂ transfer into the internal cell pool and Rubisco carboxylation) (Raven, 1997). This is consistent with the dominance of diatoms (Fragilariopsis spp.) in the trap material, as Bacillariophyceae are known to employ a C3 metabolism (Table IV in Raven, 1997). Preferential uptake of ²⁸Si by diatoms (De La Rocha et al., 1997) during the late spring bloom of productive period 1 also explains the low δ³⁰Si_BSi values. BSi : POC ratios were elevated at the start of productive period 1, which may suggest that phytoplankton were heavily silicified. The contribution of non-siliceous phytoplankton was low during the periods analysed for phytoplankton composition (<2 %, with the exception of the shallow trap in the late January sample where the contribution was 6.7 %), though we cannot rule out higher contributions of non-siliceous phytoplankton during other periods which could account for the lower BSi : POC ratios at these times. After initial low values, we see a progressive increase in both δ¹³C_POC and δ³⁰Si_BSi, reflecting the progressive utilisation of both ¹²C and ²⁸Si as nutrient pools are consumed during the bloom. As such, the diatom cells reaching the sediment trap in late spring and early summer were utilising increasingly isotopically enriched C and Si for growth, leading to progressive isotopic enrichment of the cells sinking into the sediment trap. This observation fits with elevated but decreasing surface chlorophyll concentrations from February to April 2018. Increasing δ¹³C_POC and δ³⁰Si_BSi into the late summer may also partially reflect preferential remineralisation of the more labile ¹²C and ²⁸Si in particles as they sink through the upper 400 m of the water column. The lack of variation in δ¹³C_POC and δ³⁰Si_BSi between 400 and 2000 m in our study suggests that remineralisation may be limited between these depth or that there is no further fractionation effect. Whilst laboratory-based silica dissolution experiments are equivocal (Demarest et al., 2009; Wetzel et al., 2014), our findings agree with field studies that also indicate a lack of Si isotopic fractionation during diatom silica dissolution (Closset et al., 2015; Egan et al., 2012).

During productive period 1 there was no clear trend in δ¹⁵N_PN, with values between −1.95 ‰ and +2.96 ‰. We speculate that this mixed signal resulted from a combination of surface phytoplankton using both ammonium and nitrate as the inorganic nitrogen source and variability in the sediment trap material composition. Enrichments of 2 ‰–4 ‰ occur between successive trophic levels, and egestion and excretion can have varying isotopic effects (see Sect. 4.3); thus the presence of faecal pellets, animal moults, and carcasses could alter the isotopic composition of the sediment trap material. Additionally, any supply of ammonium through remineralisation would be utilised quickly because ammonium is kinetically favourable to nitrate (Glibert et al., 2016), resulting in particles with a decreased δ¹⁵N_PN compared to those produced by nitrate assimilation.

4.2.2 Winter hiatus

Between May and August, both δ¹³C_POC and δ³⁰Si_BSi showed little change, with a slight progressive decrease for δ¹³C_POC and an increase in δ³⁰Si_BSi. It is possible that the slight progressive trend towards a lighter carbon isotopic composition of sinking particles from −24.94 ‰ to −25.98 ‰ is driven by a mixture of older, isotopically heavier particles that have undergone partial remineralisation and the input of material of different isotopic composition from the small secondary peak in POC we observed in late April and early May. An input of smaller, more slowly sinking cells reaching the trap in increasing numbers following the initial late spring peak in production could drive the lower δ¹³C_POC at this time. Additionally, the pulse of material could be driven by a successive peak in production of a different phytoplankton community with a different isotopic signature. Korb et al. (2012) found an increasing presence of dinoflagellates from spring to summer as well as seasonal changes in the size structure of the phytoplankton community to the northwest of South Georgia, supporting either hypothesis. We do not have the species composition data from this time period to evidence this directly, but we suggest that the reduction in δ¹³C_POC does not relate to a mixing event and a resupply of ¹²C, due to the fact that δ³⁰Si_BSi continued to increase slowly. Given the generally lighter silicon isotopic composition of seawater below the photic zone, we would expect a mixing event to also result in a decline in seawater δ³⁰Si and consequently δ³⁰Si_BSi. This would mean that our hypothesised shift in phytoplankton species composition in the traps (May–August) did not impact Si fractionation to the same extent as carbon isotopes. Whereas size, growth rates, cell geometry, and different carbon acquisition mechanisms have all been highlighted as impacting the δ¹³C_POC of marine plankton (Popp et al., 1999, 1998; Bidigare et al., 1999; Trull and Armand, 2001; Tuerena et al., 2019), species-dependent Si fractionation by polar and subpolar diatoms has only been observed in the laboratory, not in the field (Annett et al., 2017; Cassarino et al., 2017; Sutton et al., 2013). δ¹⁵N_PN in the shallow trap showed a slight progressive decrease from April to July, before increasing in August to 5.42 ‰. The progressive decrease is consistent with the propagation of the surface signal of phytoplankton growth and fractionation, with a longer time lag than during spring and summer due to slower sinking rates during the low-productivity period. Decreasing δ¹⁵N_PN reflects the increasing influence of ammonium uptake, either in the same locale or upstream in the regional circulation system, which leads to lower δ¹⁵N_PN than nitrate uptake in the slowly sinking flux. The large range in δ¹⁵N_PN in the deep trap in July makes it difficult to determine with certainty a trend in δ¹⁵N_PN in the deep trap between July and October. Dissimilar trends in δ¹⁵N_PN between the two traps over the winter period also support the argument that material reaching these two traps may have a different source region or time period in surface waters (Sect. 4.1).

4.2.3 Productive period 2

At the start of productive period 2 (September) we saw a significant decrease in δ³⁰Si_BSi (∼0.5 ‰) in both traps suggesting resupply of ²⁸Si-enriched silicic acid to the euphotic zone via mixing. Interestingly, we did not see the same consistent shift in carbon isotopes; we measured a ∼1 ‰ decrease in the shallow trap δ¹³C_POC and a ∼1 ‰ increase in the deep trap δ¹³C_POC. We speculate that this mixing could bring waters of increased silicic acid concentrations to the surface, promoting the full expression of the isotope fractionation effect from phytoplankton uptake and thus lower δ³⁰Si_BSi in sinking particles. To match our observations, these mixed waters would need to be similar in dissolved inorganic carbon concentrations and δ¹³C, which could relate to the depth of mixing and differences in the depth at which POC and BSi are remineralised (Friedrich and Rutgers van der Loeff, 2002; Weir et al., 2020). We note that current velocities recorded at this time were elevated (Fig. S1), particularly in the deep trap, suggesting a shift in the surrounding velocity fields, which may have resulted in biased sample collection at this time through either over- or under-collection (Buesseler et al., 2007). Whereas δ¹³C_POC progressively increased during productive period 2, from −25.88 ‰ in September to −21.56 ‰ at the end of December (mean of deep and shallow traps), δ³⁰Si_BSi continued to decrease until November before showing a sudden increase from +0.74 ‰ to +1.80 ‰ at the end of the sampling period. This may suggest that DSi or co-limiting nutrients were replete, and uptake could occur unhindered until November 2018 when very high rates of production and the associated high fluxes of BSi increased the demand for DSi and led to the enrichment of δ³⁰Si in overlying waters and subsequently sinking siliceous phytoplankton. For carbon, uptake was sufficient from September to progressively deplete source waters in ¹²C, driving an increase in δ¹³C in surface waters and newly formed phytoplankton cells. BSi : POC ratios increased from September to December, suggesting that material reaching the traps was increasingly silicified.

Interestingly, unlike C and Si isotopes, we saw a divergence in the nitrogen isotopic composition of deep and shallow traps between August and December. The sharp increase in mean δ¹⁵N_PN from +1.32 ‰ in July to +5.42 ‰ in August 2018 in the shallow trap that initiated the divergence strongly suggests an advective change in source material. As noted above, this was a period of increased horizontal velocities and may have facilitated material reaching the two traps from different sources of differing initial composition and degradation states. The substantially lower δ¹⁵N_PN in the deep trap from August to November compared to that of the shallow trap is surprising. It would be expected, that, as particles sink and are progressively decomposed this would remove dissolved nitrogen depleted in ¹⁵N, thus increasing δ¹⁵N_PN in the particles. Indeed many studies have observed this trend of increasing δ¹⁵N with depth in suspended particles (Altabet et al., 1991, and references therein). However, like Altabet et al. (1991), we observe lower δ¹⁵N_PN in sinking particles in the deep sediment trap. This has also been observed previously in Antarctic waters (Wada et al., 1987). Though the reason for this is not well understood (Sigman and Fripiat, 2019), it appears to be a consistent phenomenon. Particles in our deep trap must therefore be gaining light nitrogen or losing heavy nitrogen and could reflect a different source composition. In agreement with Altabet et al. (1991), we suggest that lateral transport of low δ¹⁵N_PN from a region of increased ammonium-based production could explain this, highlighting a difference in the source of sinking particles to the two traps. Altabet et al. (1991) also suggests that, since protein nitrogen is 3 ‰ higher than bulk nitrogen, the selective decomposition of protein could explain the decrease in δ¹⁵N with depth, though why this would not be the case also for suspended PN is unclear. We observe the greatest divergence in shallow and deep N isotope compositions during periods of low PN flux (Fig. 3), consistent with the observations of Altabet et al. (1991), enabling a low flux of laterally supplied material to have an amplified impact on the isotope signal. In support of this, in December when particle fluxes increase sharply with the spring bloom, δ¹⁵N_PN in the deep trap increases more in line with that of the shallow trap, highlighting a switch from source material being dominated by lateral supply when vertical supply is negligible to the dominance of vertical supply from surface production following the phytoplankton bloom.

4.3 Drivers of shifting isotopic ratios

The mean flux-weighted isotopic composition measured during productive periods 1 (January to the start of April 2018) and 2 (September to the end of December 2018) suggests that the processes driving the flux of material at these times differ (Fig. 3, Table 1). The divergence in the δ¹⁵N_PN of deep and shallow trap material during period 2 limits our ability to compare the temporal shifts in mean isotopic ratios for nitrogen isotopes, so we focus here on δ¹³C_POC and δ³⁰Si_BSi. Since our record does not extend beyond December 2018 and we do not capture the first 3 weeks of January 2018 when fluxes were likely high, we do not record the initial value at this time; however, we would expect δ¹³C_POC to be even more negative at this time. We cannot determine if δ¹³C_POC and δ³⁰Si_BSi would return to values akin to that in period 1 in the following late-spring–summer season (January 2019). We saw a shift in δ¹³C_POC from a mean of −28.31 ‰ in January 2018 at the time of our first measurements to −25.88 ‰ in September at the start of period 2. This coincided with a change in community structure, with abundance dominated by Fragilariopsis spp. in period 1 to a more mixed community in period 2. Of the abundant phytoplankton species (>5 %, Fig. 4a, c), we find statistically significant linear relationships between δ¹³C_POC and percent abundance for Fragilariopsis spp. (empty: R²=0.926, p<0.001), Thalassionema nitzschioides (live: R²=0.774, p=0.004; empty: R²=0.844, p=0.001), and Chaetoceros spp. (resting spore) (R²=0.732, p=0.007). We stress this is based on only eight samples. Nevertheless, these robust samples show that there was a shift in phytoplankton community structure. Though Fragilariopsis spp. were mainly empty cells, colonisation by bacteria (Grossart et al., 2003; Kiørboe et al., 2003) may facilitate carbon transfer within and on these cells, and certainly the live cells of T. nitzschioides and resting spores of Chaetoceros spp. would act as agents of carbon transfer (Agusti et al., 2015; Salter et al., 2012; Rembauville et al., 2016).

We examine whether this shift in phytoplankton community composition is associated with a change in SA:V (Table 2) since greater fractionation of carbon in smaller phytoplankton cells with higher SA:V is well observed in the literature (e.g. Popp et al., 1998; Tuerena et al., 2019). There was a statistically significant (paired t test, p=0.008) difference in the community SA:V between productive periods, increasing from 0.35 µm² µm⁻³ in period 1 to 0.51 µm² µm⁻³ in period 2. However, this would result in increased isotopic fractionation during period 2, which is the opposite to what we observed. We note here, that as only intact cells were counted, the measured SA:V ratios may not fully account for the isotopic composition of the trap material due to the presence of fragmented material. It is possible that there was a change in the mechanism of carbon uptake with the more mixed phytoplankton community in period 2 using HCO${}_{\mathrm{3}}^{-}$ instead of CO₂ or employing carbon concentrating mechanisms (CCMs), both of which would result in higher δ¹³C_POC than the diffusive uptake of CO₂ and Rubisco carboxylation (Raven, 1997; Cassar et al., 2004). Studies show that there is much diversity amongst diatoms in the use of CCMs, and many are able to take up both CO₂ and HCO${}_{\mathrm{3}}^{-}$ (Trimborn et al., 2009; Roberts et al., 2007; Shen et al., 2017; Young et al., 2016). We suggest that species-driven differences in carbon uptake mechanisms account in part for the differing δ¹³C_POC that we observed during the two main productive periods.

Table 2 Phytoplankton cell community surface area to volume (SA:V) ratios measured in deep and shallow sediment traps for samples enumerated in both productive periods 1 and 2.

We also observed a shift in the mean flux-weighted δ³⁰Si_BSi ratios (Table 1) between period 1 and period 2. With the exception of one culture study (Sutton et al., 2013), systematic species-driven shifts in δ³⁰Si_BSi fractionation have not been observed (e.g. De La Rocha et al., 1997), suggesting that there may be an additional driver of the changing isotopic ratios. Since, prior to our first measurements there had been a long-lasting phytoplankton bloom (Fig. S2), we would expect production to have utilised much of the light ²⁸Si, resulting in particles with enriched δ³⁰Si_BSi reaching the trap in January 2018. However, we observe isotopically light mean values of +0.48 ‰ at the start of sampling at the end of January, suggesting that there must have been a resupply of ²⁸Si. Physical mixing, bringing deep and benthic waters rich in nutrients, including iron, to the surface waters around South Georgia, is known to support the large blooms occurring downstream of South Georgia (Matano et al., 2020; Nielsdóttir et al., 2012) and could supply both ¹²C-enriched dissolved inorganic carbon and ²⁸Si-enriched silicic acid. Additional nutrients could also be supplied to our study region by glacial discharge associated with isotopically light silicon isotopic signatures (Matano et al., 2020; Hatton et al., 2019) or benthic fluxes from shelf sediments, likely also releasing isotopically light DSi (Ng et al., 2020; Cassarino et al., 2020; Closset et al., 2022). Therefore, we suggest that low values (increased fractionation) of δ¹³C_POC and δ³⁰Si_BSi during period 1 relate to increased nutrient availability enabling the full expression of the isotopic fractionation and thus enabling isotopically light particulate material to reach the sediment trap.

The ocean circulation in our study region is complex and variable on fine spatial and temporal scales, affecting horizontal and vertical velocities (e.g. Boehme et al., 2008). It is clear from the currents measured at the depths of our two traps (Fig. S1), that both the direction and magnitude of the flow can vary within and between seasons and are not necessarily consistent between the two depths. There are thus potentially different source regions for material in the two traps at certain times of the year as suggested for example by δ¹⁵N_PN ratios in winter. We lack the full depth resolution of vertical and horizontal velocity fields and information on sinking rates to confirm this, but previous studies have highlighted variability in the locations of the Southern Antarctic Circumpolar Current Front and the Polar Front, as well as eddies generated from these fronts, in our study region (Moore et al., 1999; Boehme et al., 2008; Whitehouse et al., 1996). We suggest that variability in ocean current velocities could explain different isotopic ratios in periods 1 and 2, through the supply of material to the traps from different source regions with differing nutrient and remineralisation regimes. Different source waters would impact nutrient availability including iron supply, uptake, and recycling (Hawco et al., 2021; Ellwood et al., 2020), which in turn influences species composition, nutrient utilisation, and uptake rates (e.g. Meyerink et al., 2019). This highlights the importance of making synchronous and full-depth resolution measurements of physical processes, such as current strength and direction, to be able to distinguish between spatial and temporal drivers of shifts in species composition, particle flux, and isotopic composition.

Since trophic transfer is known to impact both carbon and nitrogen isotope compositions of organic matter, the presence of moults and faecal pellets in trap samples is also important to consider. An incubation study focussed on Euphausia superba found that the δ¹⁵N of the E. superba faecal pellets was always lower than that of the copepods they ingested, though still higher than that of POM (Schmidt et al., 2003). Additionally, Tamelander et al. (2006) measured faecal pellets produced by copepods with depleted ¹⁵N compared to the algal food source. Though a few studies on temperate and subtropical copepods showed that the faecal material had similar or slightly higher δ¹⁵N than the food source (Altabet and Small, 1990; Checkley and Entzeroth, 1985), there is no consistent fractionation effect of egestion for either δ¹⁵N or δ¹³C, which may relate to compositional differences (protein, carbohydrate, lipid) and their isotopic values (Tamelander et al., 2006). We are therefore not able to determine the impact of faecal pellets or moults on the isotopic composition of our samples. As phytoplankton material dominated at the times of peak flux, we suggest that the importance of faecal pellets and moults may be greater during periods of lower flux; however, we cannot rule out their contribution during the bloom periods. We suggest that it would be highly informative to conduct particle-specific isotope analysis of common particle types in sediment traps, such as faecal pellets, phytoplankton detritus, and zooplankton moults, to improve our ability to determine the impact of particle flux composition on bulk isotope compositions.

5 Conclusions

The seasonal cycles in primary productivity and nutrient uptake in surface waters at our study site in the Scotia Sea are reflected in the fluxes and isotopic ratios of sinking particulate material. We find that most remineralisation occurs in the upper 400 m of the water column, and below this the magnitude of the flux of sinking material is relatively consistent, supported by consistency in POC:PON ratios. We find that particulate fluxes of C and BSi are tightly coupled, which highlights the importance of siliceous material in the transfer of POC to depth. We suggest that a change in phytoplankton community structure can at least partly explain the shifts in carbon isotopic composition between the two productive periods measured here. Though complex, seasonal patterns in isotopic composition of particulate material reaching the sediment traps do reflect the degree and type of nutrient utilisation in the source surface waters. Our data also suggest an importance of laterally supplied material to the sediment traps and supports seasonal differences in source regions. Our results highlight the need for more detailed mechanistic understanding of the drivers of POC flux and biogeochemical cycling to improve estimates of the current and future strength of the biological carbon pump and the ocean's role as a CO₂ sink.