4. Discussion

In this discussion, we first put our results into the overall cruise context and then focus on the causes that drive the differences between the suspended and sinking particles as measured with the MSC. Specifically, we discuss why – despite all particles being small – some particles sank while others remained suspended. Furthermore, we discuss the potential mechanisms that can explain the presence of small, slow sinking particles in the mesopelagic. We finally address two aspects of the methodology: (1) comparison of sinking particle fluxes derived from different methods, and (2) interpretation and implications of the observed outliers in our data set including the potential role of ‘patchiness’ on ocean particle distributions.
4.1 Comparison with other particulate biochemical measurements performed during EXPORTS: consistency in responses by size
A variety of methods and sampling equipment were used to measure the biogeochemical composition of particulate matter during the first EXPORTS field campaign (Siegel et al., 2021). Values of POC and PON measured in the MSC time zero fraction were slightly higher than those obtained from the 12-L Niskin bottles sampled from the CTD/rosette system from the same ship (Graff et al., in review). The differences between the two measurements may be explained by a combination of the following: (1) under sampling of sinking particles by Niskin bottles compared to the MSC, (2) slight overestimation of MSC POC values due to a possibly higher POC blank than the one used to correct the measured POC masses (see section 2.4), and (3) spatiotemporal differences in sampling (Graff et al., in review).
Our PIC concentrations were low, but consistent with low productivity regimes in the global ocean (Mitchell et al., 2017) and complimentary measurements made at Station P during EXPORTS (Roca-Martí et al., 2021), suggesting a limited presence of calcifying microorganisms. In situ pump observations showed low PIC concentrations and small contributions to particle stocks during EXPORTS (Roca-Martí et al., 2021). Our PIC estimates were higher than the in situ pump observations, which may be attributed to the use of larger filter pore size with in situ pumps (QMA; ~ 1.0 μm nominal pore size) compared to the smaller pore size of glass fiber filters used with Niskin bottles and MSC analyses (GF/F; ~ 0.3 µm nominal pore size after combustion, Nayar and Chou, 2003) (Graff et al., in review). Indeed, during EXPORTS, most of the PIC was associated with the smallest size particle fraction measured by in situ pumps (i.e., in 1–5 µm fraction rather than in the 5–51 µm or > 51 µm fractions; Roca-Martí et al., 2021).
Our measurements of bSi standing stocks were in good agreement with measurements performed on seawater collected with Niskin bottles (Brzezinski et al., 2022) and measurements obtained with in situ pumps (Roca-Martí et al., 2021), which used similar pore sizes (0.6, 0.6 and 0.8 µm, respectively). Our molar bSi-to-POC ratios associated with suspended and sinking particles were in the same range as the 1–51 μm particle size fractions (bSi:POC = 0.03–0.16, Roca-Martí et al., 2021), except for one outlier (0.35 at 500 m; see Section 4.5). Molar bSi-to-POC ratios measured from larger sized particles with in situ pumps and drifting sediment traps were higher (Roca-Martí et al., 2021; Estapa et al., 2021). This comparison suggests that the MSC sampled on average small particles (< 51 μm), under-sampled large particles, and that using GF/F filters allowed us to better represent particles < 1 μm in respect to in situ pumps. Overall, the concentrations and flux values (see also section 4.4) obtained by the MSCs agreed with other measurements obtained during the cruise, and we are therefore confident that the MSC observations are representative of the conditions present in situ during our visit.

4.2 Partitioning of TEP and ballast minerals in suspended and sinking particles

When comparing sinking to suspended particles, we expected to find either larger particles in the sinking fraction compared with the suspended fraction, and/or a higher ratio of ballast materials-to-POC. However, neither expectation was met (Figure. 1, 3 and 4 ). Both fractions were characterized by small particles, which - in a simplistic view - would suggest that sinking particles contained a higher fraction of ballast material than suspended particles. Contrary to this idea, however, suspended particles were characterized by a higher ratio of ballast material in the mesopelagic. This begs the question why the small non-ballasted particles sank (and, why the ballasted particles did not sink).
The partitioning of particles into suspended and sinking pools must have been controlled by another factor. So far, we looked at the composition in context of the traditional constituents: POC, PIC and silicate. However, the molecular makeup of POC can have a large effect on particle density. For example, diatoms have been shown to be positively buoyant (i.e., float) (Villareal, 1988; Moore et al., 1996; Woods et al., 2008). However, the phytoplankton community consistent predominantly of very small non-diatom cells (McNair et al., 2021). Alternatively, TEP are positively buoyant and thus could reduce the sinking velocity of aggregates (Engel and Schartau, 1999; Azetsu-Scott and Passow, 2004). Overall, we observed low concentrations of water column TEP between 5 and 500 m (17.3–1.2 µg GXeq L-1). This result is in line with low productivity systems such as the Southern Ocean (from undetectable values to 48 µg GXeq L-1; Ortega-Retuerta et al., 2009), the tropical western Pacific Ocean (5.3–40 µg GXeq L-1; Yamada et al., 2017), the Eastern Mediterranean Sea (5.7–25.1 µg GXeq L-1; Ortega-Retuerta et al., 2019) and the Indian and the subtropical South West Pacific (< 20 µg GXeq L-1, Engel et al., 2020). Despite the observed low TEP concentrations, TEP-C accounted for approx. 15% of the suspended POC pool. According to theoretical calculations, a 1-mm aggregate composed solely of TEP and diatoms can sink only if less than 5% of its carbon content consists of TEP (Mari et al., 2017). Here we show that the relative contribution of TEP-C was on average three times that threshold (15%) in suspended particles while in sinking particles on average only 3% of the POC was TEP. While the model used diatom as primary particle pool, in our case most of the suspended particle pool consisted of very small cells (Synechococcus spp, picoeukaryotes and nanoeukaryotes; McNair et al., 2021), and possibly fragments of old cells, minerals, and detritus. Nonetheless, our data support the model implication (Mari et al., 2017) and provide evidence that the chemical composition of particulate organic carbon, and especially the contribution of exudates, is critical to understand particle sinking velocity.
Hence, the higher TEP-to-POC ratio in suspended particles compared to sinking particles may have led to the difference in the partitioning of the two particle fractions as collected by the MSC by reducing the sinking velocity of small, ballasted particles that - in the absence of TEP - would have been part of the collected sinking fraction. Whereas the suspended POC was associated with ballasting material and consisted of a large fraction of TEP, the sinking POC consisted of relatively small fractions of ballasting material and TEP. This result suggests that the ratio of ballasting material-to-POC cannot be used as a sole predictor for sinking velocity, but that organic carbon composition needs to be considered.
The importance of TEP for potentially reducing sinking velocity and thus flux was first suggested by Mari et al. (2017), and here we present the first in situ measurements suggesting that high TEP concentrations reduce particle sinking velocities. Yet, our observations are limited to one field campaign only, and hence we cannot generalize. We observed this phenomenon in a low productivity system where particle concentrations were low and TEP concentrations were high. Our dataset was collected during late summer, when the system was characterized by low concentrations of chlorophyll a , nutrients, particles, and ballast minerals (Siegel et al., 2021; Roca-Martí et al., 2021; Brzezinski et al., 2022). The TEP-to-particle ratio may have been higher compared to other seasons, due to stressful conditions for phytoplankton growth and high light levels (Ortega-Retuerta et al. 2009), implying that the observed mechanisms could be seasonal (Mari et al., 2017). Regardless, our data suggests that future biological pump studies should characterize POC composition, including TEP, to help gain a mechanistic understanding of how these materials influence particle sinking velocities and, in turn, the role of particle dynamics in ocean biogeochemical cycles across different productivity systems and seasons.

4.3 Potential mechanisms explaining small particles at depth

Our observations suggest that suspended particles were increasingly reworked with depth, e.g., a larger fraction of POC was remineralized with depth, leading to the relative increase in the ratio between ballasting material and POC with depth. Sinking POC did not decrease with depth, suggesting that either the attenuation of the sinking particle population was negligible, or an input of sinking material balanced the expected losses. If we assume that sinking particles collected at 500 m were produced at 30 m (within the mixed layer) and sank at 18 m d-1, those particles would require 26 days to reach 500 m. We can estimate the loss rate assuming a temperature-dependent remineralization rate of 0.04 d-1 measured experimentally using a temperature of 7 °C (average temperature between 30 and 500 m from CTD) and a Q10 of 3.5 (Iversen and Ploug, 2013; Giering et al., 2017). If particles were to sink undisturbed, 30% of the POC would have been remineralized by bacteria during that time window (26 days). Yet, such decrease was not observed.
One theoretical explanation for a lack of flux attenuation of sinking particles would be rapid physical transport processes (Omand et al., 2015; Giering et al., 2016; Siegel et al., 2023). Particle subduction or convective mixing can, however, be excluded as during our visit density stratification beneath the mixed layer was strong and consistent and lateral density gradients were weak (Siegel et al., 2021).
Alternatively, an input of sinking particles at depth may have balanced their loss. Large sinking phytodetrital aggregates were rare, implying that disaggregation of sinking aggregates was not the prevailing source of small sinking particles. However, if suspended POC existed as loosely packed aggregates within a buoyant TEP matrix (e.g., low-density aggregates; Mari et al., 2017), their disaggregation would allow dense components to sink, albeit at a relatively slow velocity. Such disaggregation could be caused by zooplankton (Passow and Alldredge, 1999) or by microbially mediated degradation processes (Passow, 2002). During microbially mediated degradation, lighter C-components may be consumed preferentially, leaving denser and more recalcitrant material in the sinking fraction (Hamanaka et al., 2002). In fact, extracellular particulate carbohydrates released by phytoplankton (i.e., TEP) are remineralized by bacteria at a faster rate than non-TEP carbon (i.e., 0.53 and 0.21 d-1, respectively; Harvey et al., 1995; Mari et al., 2017).
Particle flux was, in fact, dominated by fecal pellets during EXPORTS (~80%; Durkin et al., 2021; Steinberg et al., 2023). Specifically, mini-pellets (ESD < 100 μm) contributed up to 46% of the total carbon flux in the upper mesopelagic (Durkin et al., 2021), with mesozooplankton and salp pellets contributing the majority of the sinking flux (Durkin et al., 2021; McNair et al., 2023). Fragmentation of fecal pellets, e.g., due to sloppy feeding or swimming motions (Steinberg and Landry, 2017) could have produced small and dense particles, which may have constituted the bulk of the small sinking particle pool we collected. Such pellet fragments would likely contain a high percentage of organic carbon and a low percentage of TEP, similar to our sinking particle pool. Though we did not observe large fast sinking pellets in the MSCs, this absence could have been caused by an overall low abundance of pellets or because the MSCs were deployed during the daytime when a large fraction of the zooplankton community likely resided at depth (e.g., Steinberg et al., 2023).
Microzooplankton grazing of suspended POC, and the resulting production of mini-pellets (Stemmann et al., 2004), may have also contributed to the production of small, dense sinking particles. Consumption of small particles by zooplankton and incorporation of their carbon and biogenic silica into sinking particles (Dagg et al., 2003) is especially important for carbon flux if it occurs within relatively short-path food webs (Richardson, 2019). This mechanism could also have been important during our study: Compound-specific isotope analysis (CSIA-AA) performed during our field campaign (Connor Shea, 2023, pers. commun.) found that the mesopelagic zooplankton food web at Station P was mainly (72–96 %) based on small particles (< 6 µm).
Although we cannot say with certainty why sinking flux of POC did not decrease with depth, likely a series of biologically mediated aggregation or fragmentation processes led to a replenishment of the small sinking particle pool with depth. The described mechanisms are potentially significant contributors in dictating the efficiency of the biological carbon pump, especially in stable low-productivity systems, characterized by small particles in low concentrations.

4.4 Comparison of fluxes derived from different methods during EXPORTS

Our background POC fluxes compared to the cruise mean POC fluxes obtained using both neutrally buoyant and surface-tethered sediment traps (Estapa et al., 2021) and water column Thorium-234 assessment (Buesseler et al., 2020) were similar in the upper 100 m, but larger below 100 m (Figure 6 ). These differences between the estimates for depths > 100 m existed even when the lower-bound estimate (assuming a sinking velocity of 18 m d-1; Section 3.4) was used. In contrast, bSi flux estimates obtained from the MSC were in good agreement with those derived from the traps, but lower than the Thorium-234-based values (Roca-Martí et al., 2021) (Figure 6 ).
These three approaches for determining sinking particle fluxes differ in the design, spatiotemporal resolutions, and size of collected particles; thus, may not be directly comparable and comparisons should be made cautiously. The approaches mainly diverge in: (1) Temporal coverage: the MSC provided virtually instantaneous estimates, whereas drifting sediments trap (Gardner, 1977) and water column Thorium-234 carbon flux determinations (Buesseler et al., 1992) integrate over periods of 3–6 days (time to collect sinking particles) and 24 days (the half-life of234Th), respectively. (2) Spatial coverage: the MSC deployments have the smallest spatial resolution among these methods with 39 discrete deployments as each MSC deployment assesses the flux from a single ~ 100 L sample. Drifting sediment traps sampled sinking particles from an area of ≤ 10 km2(Siegel et al., 2008). Due to the long integration times and abundance of samples (nearly 1000), 234Th particle fluxes average over even larger spatial scales. (3) Minimum size of collected particles: differences in pore sizes of filters can lead to differences in measured biogeochemical concentrations. Here we measured POC using filters with a pore size of ~ 0.3 μm, while POC from sediment traps (Estapa et al., 2021) and water column Thorium-234 samples (Buesseler et al., 2020) were determined using filters with a pore size of ~ 1 μm. The filter pore sizes used to estimate bSi were 0.6 μm, ~ 1 μm and 0.8 μm for MSC, sediment traps and Thorium-234, respectively. Additionally, drifting sediment traps often underestimate small particles due to hydrodynamics (Buesseler et al., 2007). (4) Maximum size of collected particles: the MSC likely under samples rare particles, such as large particles, as it collects a limited amount of seawater. During EXPORTS, sediment trap samples were highly impacted by zooplankton “swimmers” and fluxes were corrected to account for the removed biomass (Estapa et al., 2021).
The potential sources of discrepancies between fluxes measured from sediment traps and water column Thorium-234 were addressed in Estapa et al. (2021). Briefly, Thorium-234 fluxes measured in sediment traps were roughly three-fold smaller than water column Thorium-234 fluxes. The different temporal coverages did not play a major role due to the stability in production and exports before the EXPORTS cruise (McNair et al., 2023). The mismatch was largely attributed to under-sampling of both small particles (< 32 μm) due hydrodynamic biases and rare, large particles (> 1 mm) and zooplankton active migrant flux by the sediment traps (Estapa et al., 2021). The size of particles collected in the polyacrylamide gel traps, which ranged from 75 to 1461 μm (median of 282 μm), supported this hypothesis (Estapa et al., 2021). Estimates from the MSC account for the contribution of those small particles close to ~ 0.3 μm. Thus, the discrepancy between MSC and sediment traps may be partially explained by an under-characterization of very small (< 1 µm) particles by the traps. This statement is also valid when comparing Thorium-234-based POC fluxes if we consider the pore size of the filters used and that Thorium-based fluxes were estimated assuming the POC/Th ratio of mid-sized particles (5–51 μm) (Buesseler et al., 2020).
However, the MSCs did not capture large, rare particles such as aggregates, salp fecal pellets nor the active flux due to migration. We would therefore have expected to see the opposite trend: higher flux measured using Thorium-234 and traps, if large, rare particles played an important role in overall flux at the site. This comparison suggests that POC flux from the MSCs is complementary to those measured from Thorium-234 and traps. Most importantly, it suggests that small sinking particles were a relevant part of the sinking POC flux in the upper mesopelagic zone, likely exceeding the contribution of rare, large sinking particles. The fact that the MSCs and sediment traps produced similar estimates of bSi fluxes reinforced our finding that the population of sinking particles collected by the MSCs included small sinking POC missed by the other methods, whereas bSi, which was mostly associated with slightly larger particles (Roca-Martí et al., 2021; Estapa et al., 2021; Brzezinski et al., 2022), was equally sampled by all three methods.

4.5 Outliers – Patchiness or artifacts?

The dataset obtained from the MSC holds the potential of highlighting the ecosystem’s patchiness through the collection of spatially heterogeneous features. Patchiness in the ocean exists both vertically and horizontally spanning from the microscale to the mesoscale (Cassie, 1962; Siegel, 1998; Robinson et al., 2021) and is in general difficult to map with most oceanographic approaches (McNair et al., 2023). Several of our biogeochemical measurements resulted in values considerably higher than average concentrations for their collection depth (Table S5 ). These “outliers” may be attributed to patchiness or could be due to methodological artifacts. Methodological artifacts could be caused by misfiring of the MSCs at a shallower depth than intended or by contamination of the filter. The latter is improbable as replicate filters and comparisons between time zero and suspended plus sinking particle fractions would have indicated “contamination”. Additionally, filters were always visually inspected for visible particles, such as macrozooplankton, which, if present, were removed. These outliers were specific to the MSC, no outliers were detected by the 12-Liter Niskin bottles of the CTD on board R/V Roger Revelle .
Three of the five outliers in POC were observed on August 16, at 55 m, 95 m, 195 m, one on August 21 at 195 m, and one on September 1 at 350 m. The outlier concentrations of sinking POC measured at 55 and 95 m, which were an order of magnitude higher than any other sinking POC values, are likely reflective of in situ patchiness: the respective time zero and suspended POC concentrations were also within the upper range of the values measured at those depths. The POC measured in the time zero fraction at 55 m equals the sum of POC measured in the suspended and sinking particle fraction, confirming that the MSC samples were uncontaminated. Associated sinking PON values were also marked as outliers and the molar C-to-N ratios were reasonable (10.4 and 12.5 at 55 and 95 m, respectively), albeit somewhat higher than those of the time zero and suspended fractions (7.3 and 6.4 and 8.3 and 9.7 at 55 and 95 m, respectively). Lastly, while the associated sinking TEP were high (with the observation at 95 m being an outlier), suspended TEP concentrations agreed with TEP measured from water collected with Niskin bottles at those depths (Figure S3 ). These observations and the fact that the MSC deployment depths were chosen to specifically target “interesting” areas as seen by the CTD or UVP profiles, make us believe that the 100-L sample we collected may have included patches of unusually high biochemical concentrations.
The three POC outlier concentrations of suspended POC at 195, 195 and 350 m were associated with outlier values of PON, and could reflect POC and PON concentrations measured at 50–65 m. However, the associated suspended TEP concentrations (2.8, and 3.1 µg GXan equiv L-1) suggest samples stemmed from below 50–65 m (Figure S3 ). Although uncertain, these too could reflect patchiness.
One value of biogenic silica (176 nmol Si L-1, 500 m) was more than a factor of 3 higher than the bSi measured on other days at that depth either with the Niskin bottles or the MSC (57 and 45 nmol Si L-1, respectively). However, associated POC concentrations were consistent with the deployment depth and the molar bSi-to-C ratio of this observation (0.35) matched the one measured for particles > 5 μm (bSi:POC = 0.16–0.94 ) collected with in situ pumps (Roca-Martí et al., 2021). Thus, we assume that this sample included a higher-than-average presence of diatoms or rhizaria. Hot spots of bSi accumulation have previously been found in low productivity systems (Crombet et al., 2010) and this depth was targeted because the UVP profile directly before the MSC deployment indicated a Rhizaria maximum. Similarly, the elevated PIC concentration associated with sinking particles (21.5 μg PIC L-1) could potentially be attributed to the presence of foraminifera test fragments or coccoliths.
In summary, the “outliers” measured with the MSC are likely, at least in part, reflective of patchiness in concentrations of small particles in this low productivity system. Patchiness in the distribution of salp swarms was also a prevalent feature during this study (Steinberg et al., 2023). The sporadic nature of patchiness is challenging to quantify and its importance with respect to standing stocks and fluxes of an entire ecosystem hard to assess, especially in a low productivity system.