The Roles of Suspension-Feeding and Flux-Feeding Zooplankton as Gatekeepers of Particle Flux Into the Mesopelagic Ocean in the Northeast Pacific

Introduction

Zooplankton play diverse roles in the cycling of many elements in the ocean including iron, zinc, sulfur, and mercury (Fowler, 1977; Asher et al., 2016; Baines et al., 2016; Schmidt et al., 2016; Gorokhova et al., 2018). However, their greatest importance to global biogeochemistry is likely derived from their roles in the biological carbon pump (BCP; Buitenhuis et al., 2006; Turner, 2015; Steinberg and Landry, 2017). The BCP refers to the processes that transport organic carbon fixed by phytoplankton in the euphotic zone into the mesopelagic realm (Silver and Gowing, 1991; Ducklow et al., 2001; Siegel et al., 2016). The BCP leads to net transport of CO₂ from the surface ocean into the deep ocean where it can be sequestered for periods of decades to millennia (DeVries et al., 2012). Estimates of the present magnitude of the BCP range from 5 to 12 Pg C year^-1 (Henson et al., 2011; Laws et al., 2011; Siegel et al., 2014; DeVries and Weber, 2017), however, the responses of mesozooplankton and the BCP to future climate change remain unknown.

Research on the role of mesozooplankton in the BCP has focused primarily on the epipelagic zone where the relationship between zooplankton and the BCP can change based on the community composition of phytoplankton and zooplankton. Zooplankton can play an important role in combining smaller particles into large, rapidly sinking fecal pellets (Bruland and Silver, 1981; Komar et al., 1981; Turner, 2002; Wilson et al., 2008) and promote aggregation and sinking through discarded mucous feeding webs (Alldredge, 1976; Hansen et al., 1996; Robison et al., 2005). Zooplankton can also decrease the magnitude of the biological pump when their grazing activities exert top-down control on phytoplankton production (Glibert, 1998; Goericke, 2002) or fragment larger particles into smaller ones (Dilling and Alldredge, 2000), and also play an important role in nutrient regeneration in the euphotic zone (Frangoulis et al., 2005; Alcaraz et al., 2010; Saba et al., 2011). Active transport by diel vertically migrating zooplankton is also an important (potentially dominant) component of the biological pump in many marine ecosystems (Steinberg et al., 2000; Hannides et al., 2009; Bianchi et al., 2013; Stukel et al., 2018b; Archibald et al., 2019; Hernández-Leon et al., unpublished; Kelly et al., unpublished; Kiko et al., unpublished).

Although fewer studies have quantified zooplankton impacts in the mesopelagic, zooplankton play a substantial role in consuming, disaggregating, and transforming sinking particles (Steinberg and Landry, 2017). These organisms may play a substantial role in modulating marine snow flux in the mesopelagic (Lampitt et al., 1993) and have been hypothesized to play roles as “gatekeepers” that modulate carbon transfer from the euphotic zone to the mesopelagic (Jackson and Checkley, 2011). However, zooplankton are phylogenetically and functionally diverse with a wide array of feeding strategies (Kiørboe, 2011). It is thus important to consider the relative importance of zooplankton with different feeding modes. For instance, suspension-feeding salps and crustaceans may consume particles in relative proportion to their abundance in the water column, although they may show selectivity based on particle size or other characteristics (Fuchs and Franks, 2010). In contrast, a flux-feeding pteropod will intercept particles in proportion to the speed with which they sink through the water column (Jackson et al., 1993), while cruise-feeding zooplankton may search for and colonize large aggregates (Kiørboe and Thygesen, 2001). Yet other taxa may break aggregates apart due to their swimming and feeding behaviors or partially consume fecal pellets leading to breakage and decreased settling velocities (Goldthwait et al., 2004; Iversen and Poulsen, 2007). Understanding these interactions is important to understanding observed decadal scale changes in carbon flux attenuation (CFA) in the mesopelagic (Lomas et al., 2010).

In this study we investigate the relative importance of representatives of two groups of flux-feeding zooplankton (phaeodarians and euthecosome pteropods) with respect to sinking particle flux attenuation. Phaeodarians are a group of siliceous protists from the supergroup Rhizaria that typically thrive in the deep ocean (Nakamura and Suzuki, 2015). We focus in this manuscript on one group of large Phaeodaria (Aulosphaeridae) that have a typical diameter of ∼2-mm and are common in the shallow twilight zone in the CCE (Ohman et al., 2012; Biard et al., 2016, 2018). These mesopelagic phaeodarians are likely flux feeders that have relatively slow (for protists) growth rates and rely on the rain of sinking particles from above for their food (Gowing, 1986, 1989; Gowing and Bentham, 1994; Stukel et al., 2018a). Euthecosome pteropods are pelagic molluscs that produce large mucous feeding webs to trap food, including swimming zooplankton and sinking particles (Gilmer and Harbison, 1986; Lalli and Gilmer, 1989). We focus on the species Limacina helicina, which is common in cold waters worldwide from the Antarctic to the Arctic including the California Current (Hunt et al., 2010; Bednaršek et al., 2012). L. helicina produces a mucous feeding web with a typical diameter of 40–55 mm (Gilmer and Harbison, 1986). We draw our data from a decade’s worth of field campaigns with differing objectives and hence changing methodology. We thus do not intend this to be a definitive assessment of the role of flux feeders in the mesopelagic ecosystem. Rather we intend it as an initial quantitative investigation of the differing roles of suspension feeders and two specific taxa of flux feeders. We find that each of these flux feeders has the potential to mediate a substantial portion of the CFA in the shallow twilight zone, although their impact is greatly reduced in the deeper mesopelagic. In contrast, suspension-feeding zooplankton are abundant through the mesopelagic, but likely only play a substantial role in the attenuation of the flux of slowly sinking particles. There is, however, substantial uncertainty around our core conclusions, and we hope that this uncertainty will spur future targeted investigations.

Materials and Methods

Field Sampling

In situ measurements were made on six process cruises of the CCE LTER Program (P0704, April 2007; P0810, October 2008; P1106, June 2011; P1208, August 2012; P1408, August 2014; and P1604 April 2016, Figure 1). On these cruises, we used a quasi-Lagrangian sampling scheme to track water parcels for a period of 2–5 days while quantifying biotic and abiotic standing stocks and rates (Landry et al., 2009, 2012). After preliminary site surveys with a free-fall Moving Vessel Profiler (Ohman et al., 2012), quasi-Lagrangian experiments (hereafter referred to as “cycles”) were initiated with the deployment of a surface-tethered drifting sediment trap with a 3 × 1-m drogue centered at 15-m depth to track the mixed layer (Stukel et al., 2013). A second, identically drogued, experimental array was deployed and recovered daily while being used as a platform for in situ incubations (Landry et al., 2009). During each cycle, paired day–night Multiple Opening and Closing Net and Environmental Sensing System (MOCNESS) net tows were used to determine vertical patterns of mesozooplankton abundance (Powell and Ohman, 2015). On approximately 10 CTD-Niskin rosette casts per cycle, we used an Underwater Vision Profiler (UVP5) to determine vertical profiles of rhizarians (Ohman et al., 2012; Biard et al., 2018). We also measured the vertical flux (and flux attenuation) of sinking particles using the aforementioned drifting sediment traps and measurements of water column ²³⁸U–²³⁴Th deficiency (Stukel et al., 2015). We divided our Lagrangian experiments into oligotrophic cycles (<0.5 μg Chl a L^-1) or high biomass cycles (>0.5 μg Chl a L^-1). For details on cruise conditions, see Supplementary Materials.

Zooplankton Collection and Enumeration

Mesozooplankton depth-stratified abundances were determined from day–night MOCNESS tows (1 m² net, 202-μm mesh) on the P0704 and P0810 cruises (typically two pairs of tows per cycle). Each net collected organisms over a ∼50-m depth interval with consecutive nets tripped from a depth of 450 m to the surface. Samples were preserved in 1.8% formaldehyde and analyzed using a ZooScan digital scanner with ZooProcess software (Gorsky et al., 2010). Zooplankton were sorted into broad taxonomic groups (e.g., copepods, euphausiids, doliolids, etc.) using machine learning algorithms, the taxonomic assignments of 100% of the vignettes validated manually, then automatically sized (as Feret diameter). Organismal length was then used with allometric equations (see below) to quantify organism biomass and clearance rate. For additional details on ZooScan processing, see Stukel et al. (2013), Powell and Ohman (2015), and Ohman and Romagnan (2016). On two cycles from the P1208 cruise, these samples were also sorted to enumerate the abundance of pteropods, specifically L. helicina (Bednaršek and Ohman, 2015).

Because they are not well preserved in net tows, large rhizarians (>600-μm) were quantified using an UVP5 (Picheral et al., 2010; Biard et al., 2016). The UVP5 is an in situ imaging camera that was mounted downward facing on the bottom of the ship’s CTD-Niskin rosette and deployed an average of 10 times per Lagrangian cycle. It reliably images organisms that are >600-μm in diameter, although avoidance issues should be expected for strongly swimming taxa. The UVP5 images a volume of ∼1 L per image at ∼6 Hz. Data were analyzed as described in Biard et al. (2016, 2018). Briefly, images were automatically analyzed to separate out vignettes representing organisms or marine snow aggregates. ZooProcess software was utilized to generate morphometric information (e.g., diameter) and classify the organisms into broad taxonomic groups. Classifications were then 100% manually validated. Phaeodarian taxonomic groups included Aulosphaeridae, which was the dominant rhizarian present and is the only group of organisms enumerated by UVP5 that is utilized in this study.

Sediment Trap Deployments

VERTEX-style particle interceptor tube (PIT) sediment traps were deployed at the beginning and recovered at the end of each cycle (Knauer et al., 1979; Stukel et al., 2013). PITs consisted of a polycarbonate tube with 7-cm inner diameter and 8:1 aspect ratio with a baffle on top comprised of 13 smaller, beveled tubes. On the P0704 cruise, PITs were deployed at a depth of 100 m. On P0810, P1106, and P1208, PITs were deployed at a depth of 100 m and near the base of the euphotic zone (as estimated based on fluorescence profiles from MVP transects) if the base of the euphotic zone was shallower than 75 m. On the P1408 and P1604 cruises, PITs were deployed at the base of the euphotic zone, 100 m, and 150 m.

PITs were deployed with a dense formaldehyde-filtered seawater brine. Deployments lasted from 2.25 to 4.25 days. After recovery, the interface separating brine water from overlying ambient seawater was identified and the overlying seawater was immediately removed. Samples were then gravity filtered through a 200-μm mesh filter, the filter examined under a stereomicroscope, and swimming zooplankton removed. Three to five replicates per depth were filtered through pre-combusted GF/F filters and used for particulate organic carbon analyses (¼ to ½ of a tube). An additional three samples were filtered through pre-combusted quartz (QMA) filters and used for C:²³⁴Th ratio measurements (Stukel et al., 2019). On the P0704, P0810, and P1604 cruises, two replicates (one half tube each) per depth were saved in formaldehyde and analyzed under a stereomicroscope to quantify fecal pellet abundance (Stukel et al., 2013; Morrow et al., 2018). For detailed methods and information on additional analyses made from these traps, see Gutierrez-Rodriguez et al. (2018), Morrow et al. (2018), and Stukel et al. (2019).

²³⁴Th Analyses

Water column ²³⁴Th activity was measured (typically two profiles per cycle and 10–12 depths per profile, spanning the upper 200 m of the water column) using standard small volume techniques (Benitez-Nelson et al., 2001; Pike et al., 2005). Briefly, 4-L samples were spiked with tracer ²³⁰Th and thorium was co-precipitated with manganese oxide. Samples were beta counted on a RISO low-level background beta counter and re-counted >6 half-lives later. Samples were dissolved and spiked with ²²⁹Th. The ^229:230Th ratio was determined by inductively coupled plasma mass spectrometry to determine the yield of the initial thorium filtration. For additional details, see Stukel et al. (2019). ²³⁸U–²³⁴Th deficiency was quantified after determining ²³⁸U activity from relationships with salinity published in Owens et al. (2011). ²³⁸U–²³⁴Th deficiency was combined with sediment trap organic carbon flux measurements to estimate twilight zone CFA as outlined below.

Carbon Flux Attenuation

Within the CCE, which has high mesoscale variability and pronounced horizontal currents, we consider our drifting sediment traps to provide a more accurate estimate of carbon flux than ²³⁴Th (see Supplementary Materials). Our supposition that the sediment traps have no substantial over- or under-collection bias is supported by a total of 56 paired sediment trap and ²³⁸U–²³⁴Th deficiency measurements showing good agreement (see “Results” section). Consequently, we use sediment trap values of carbon flux at deployment depths (typically near the base of the euphotic zone and at 100 m) and utilize ²³⁸U–²³⁴Th measurements to generate smooth profiles of CFA above, between, and below sediment trap deployment depths. Specifically, carbon flux at a depth horizon (D) can be quantified from ²³⁸U–²³⁴Th deficiency using a steady-state without advection equation:

where CTh(D) is the C:²³⁴Th ratio of sinking particles at the depth horizon of interest, λ₂₃₄ is the ²³⁴Th decay constant, and Def(z) is the ²³⁸U–²³⁴Th deficiency at depth z, which is equal to the activity of ²³⁸U minus the activity of ²³⁴Th. CFA can thus be calculated from the first derivative of Eq. 1:

Stukel et al. (2019) found a strong relationship between the C:²³⁴Th ratio of sinking particles and the ratio of vertically integrated POC to vertically integrated total water column ²³⁴Th (^vC:²³⁴Th_tot). Using this equation allows us to determine CTh(D) as a smoothly varying function of D. For additional details see Supplementary Appendix S1.

Particle Sinking Speed

We estimate sinking speeds from microscopic analyses of fecal pellets collected in the sediment traps (Stukel et al., 2013; Morrow et al., 2018) and a relationship between fecal pellet size and sinking rate. A strong relationship between size and sinking speed is a consistent finding of studies that have quantified fecal pellet sinking rates (Small et al., 1979; Giesecke et al., 2010; Turner, 2015). Using fecal pellet sinking rate as a function of equivalent spherical diameter (ESD) data reviewed in Stukel et al. (2014), a power-law relationship suggests that sinking speed can be predicted from ESD:

where sinking speed is in units of m day^-1 and ESD is in units of mm. This relationship is derived from multiple studies spanning a range of taxonomic groups including appendicularians (Ploug et al., 2008), copepods and euphausiids (Smayda, 1971; Turner, 1977; Small et al., 1979; Yoon et al., 2001; Ploug et al., 2008), chaetognaths (Giesecke et al., 2010), and thaliaceans and pteropods (Bruland and Silver, 1981; Madin, 1982; Yoon et al., 2001). We thus expect it to be broadly representative of sinking rates of fecal pellets produced by the mixed zooplankton assemblages encountered in the southern CCE.

Flux-Feeding Calculations

The impact of flux-feeding zooplankters (such as a thecosome pteropod or phaeodarian) on particle flux can be quantified based on the effective cross-sectional area over which particles are collected (Jackson et al., 1993):

where F is the flux of sinking particles and N_x is the numerical concentration of suspension feeders (N_Aulo is the abundance of Aulosphaeridae and N_ptero is the abundance of pteropods; all a function of depth). σ_x is the cross-sectional area over which the organisms are intercepting particles: σ_x = π/4 × ESD_eff². Following Stukel et al. (2018a), ESD_eff is the effective diameter over which the organisms collect sinking particles and is dependent on both the diameter of the collection apparatus and the average cross-sectional area of sinking particles. For the CCE, the effective diameter of sinking fecal pellets was calculated as D_par = 403 μm. We then calculate:

where ESD_Aulo is the equivalent spherical diameter of Aulosphaeridae cells (measured separately by UVP5 for each cell) and ESD_web is the equivalent spherical diameter of a pteropod feeding web [assumed to be 45 mm based on summary in Gilmer and Harbison (1986)]. The factor of 1.25 in Eq. 6 represents the ratio of cell diameter (reported by UVP5) to diameter including radial spines.

Suspension-Feeding Calculations

The impact of suspension-feeding zooplankton (within which we include both true filter-feeders, e.g., appendicularians, and other organisms such as some herbivorous copepods that may use feeding currents that can be modeled as suspension-feeding) on particle flux attenuation can be calculated as:

where N_x is the number concentration of suspension feeders with a clearance rate of CR₀, and S is the sinking speed of the particles. The impact of suspension-feeding zooplankton thus depends on the range of sinking speeds of sinking particles. We calculated clearance rates using the allometric scaling relationship determined in Kiørboe (2011):

where CR₀ is the clearance rate (mL day^-1), BC is the individual carbon content (g), a = 7.31 (±0.27), b = 1.01 (±0.05), and the Q₁₀ used to account for temperature effects was assumed (following Hansen et al., 1997) to be 2.8.

We utilized zooplankton data from nighttime MOCNESS tows sorted into broad taxonomic and size groups using ZooScan. The choice to use only nighttime biomass likely leads to a conservative estimate of flux attenuation, because it excludes the activity of diel vertical migrants. This decision was made because these vertical migrants likely feed primarily in the surface layers and are not actively feeding at depth. The carbon biomass of individual organisms was determined using length:carbon relationships for different taxa as outlined in Table 1 of Stukel et al. (2013). We calculated clearance rates for four groups of potentially suspension-feeding mesozooplankton: copepods, euphausiids, appendicularians, and ostracods, while recognizing that some taxa include omnivores or predators. Nauplii and doliolids were also sorted in the samples, but their biomasses were low and hence are not considered further. Other taxa, including chaetognaths, were abundant, but are not suspension feeders.

Results

Carbon Flux and Carbon Flux Attenuation

We used two independent estimates of particle flux (sediment traps and ²³⁸U–²³⁴Th) to test whether or not our estimates of particle flux (and particle flux attenuation) suffered from methodological biases (Figure 2). Comparisons between ²³⁴Th flux collected in sediment traps and ²³⁴Th flux estimated from ²³⁸U:²³⁴Th disequilibrium [computed using a one-dimensional, steady-state model without upwelling, Savoye et al. (2006)] showed strong agreement, with occasional outliers. The median ratio of sediment trap-derived flux to deficiency-derived flux was 0.998 suggesting near perfect agreement on a typical deployment. However, the overall mean of the sediment trap dataset was 6% higher than that of the Th deficiency dataset, suggesting that when there was a substantial disagreement, the sediment trap results were likely to be higher. This is not surprising in a dynamic system, where bloom decay can lead to spikes in particle export on faster time-scales than the approximately monthly temporal integration time-scale of ²³⁸U–²³⁴Th deficiency approaches. The overall agreement between sediment trap and thorium-based approaches suggests that our sediment traps had neither an over- nor an under-collection bias. Furthermore, our results suggest that although ²³⁴Th-derived flux may not perfectly match with contemporaneous processes occurring in the surface layer, there is no reason to suspect that flux attenuation calculations based on ²³⁴Th measurements will have a systematic bias.

Across the 30 Lagrangian cycles, POC flux (measured by sediment trap) at 100 m depth ranged from 2.6 to 24.9 mmol C m^-2 day^-1. Export was higher in coastal areas than in the oligotrophic, offshore domain and there was a strong correlation between POC flux and both primary productivity and surface chl. When independent sediment trap export measurements were made at the base of the euphotic zone and at 100 m depth, the correlation (Spearman’s ρ) between flux at these two depths was 0.72 (p = 0.002). This across-depth correlation was substantially weaker when only data from frontal regions (P1106 and P1208 Cruises) were analyzed (ρ = 0.49, p = 0.36) than when non-front data were considered (ρ = 0.92, p = 4.5 × 10^-4). For all paired samples, the median ratio of sediment trap flux at 100 m to sediment trap flux near the base of the euphotic zone (shallow trap depths varied from 47 to 70 m) was 0.72, suggesting that approximately 28% of sinking POC could be expected to be remineralized before a depth of 100 m on these cycles.

To determine continuous profiles of carbon flux and CFA, we merged sediment trap data with ²³⁴Th data. We restricted this flux attenuation analysis to the depth range from the base of the euphotic zone to a depth of 150 m, because at deeper depths we cannot constrain the C:²³⁴Th ratio with certainty. The depth range of our flux attenuation calculations thus corresponds with the upper twilight zone, where zooplankton roles in flux attenuation have been hypothesized to be particularly important (Jackson and Checkley, 2011). CFA in the upper twilight zone was highly variable (Figure 3A). When comparing across all cycles, flux attenuation decreased with depth from a mean (across all cycles) flux attenuation of 913 μmol C m^-2 day^-1 decrease in flux over a 10 m depth range between 80 and 90 m (interquartile range was 57–874 μmol C m^-2 day^-1/10 m) to a mean of 495 μmol C m^-2 day^-1/10 m (interquartile = 328–692 μmol C m^-2 day^-1/10 m) between 140 and 150 m depth. Additional patterns can be seen when considering the oligotrophic and high biomass cycles separately. For the oligotrophic cycles, the base of the euphotic zone (1% light level) was typically in the range of 60–80 m. We thus considered flux attenuation starting at the deeper limit of this range. Flux attenuation increased with depth from a mean of 231 μmol C m^-2 day^-1/10 m (interquartile = -82 to 418.5) in the 80–90 m depth range to a mean of 503 μmol C m^-2 day^-1/10 m (interquartile = 233–799) in the 100–110 m depth range. Beneath this depth flux attenuation declined gradually (Figure 3B). For the high biomass cycles a similar pattern was seen. Flux attenuation over the 50–60 m depth range in the high biomass cycles averaged 595 μmol C m^-2 day^-1/10 m (interquartile = -338 to 1221) and increased to 1503 μmol C m^-2 day^-1/10 m (interquartile = 514–1.981) in the 80–90 m depth range.

In contrast to absolute flux attenuation, the percentage of flux remineralized over a 10-m depth range did not decrease beneath 100 m (Figures 3D–F). Instead, relative flux attenuation increased somewhat from near the base of the euphotic zone to 150 m. This pattern was relatively consistent when restricting analyses to the oligotrophic cycles (Figure 3E), but was less distinct in the high biomass cycles (Figure 3F). Nevertheless, across depth and regions, relative CFA was typically in the range of 6–10% of carbon flux remineralized over a 10-m depth range. The median relative flux attenuation in the shallow twilight zone varied from 6.0 to 9.0% of flux/10 m (mean ranged from 6.5 to 13%/10 m). This sets a reasonable expectation for the amount of flux attenuation that must be mediated by the combined mesopelagic microbial, mesozooplankton, and nekton communities.

Suspension-Feeding Zooplankton and Particle Flux Attenuation

Suspension-feeding crustaceans are the most abundant (by carbon biomass) ecological category of zooplankton in the CCE (Lavaniegos and Ohman, 2007). To quantify the potential role of these suspension feeders, we utilized data from nighttime MOCNESS tows in the CCE. Suspension-feeder biomass in the mesopelagic was typically dominated by copepods and euphausiids (Supplementary Figure S2). Between 100 and 450 m depth, copepod biomass was typically in the range of 1–3 mg C m^-3. Euphausiid biomass was almost always <1 mg C m^-3 at depths deeper than 200 m (and typically <0.4 mg C m^-3) but was greater in the shallow mesopelagic. Ostracods had substantially lower biomass, never exceeding 0.6 mg C m^-3 in the mesopelagic. Appendicularians, while occasionally abundant in the euphotic zone, were also relatively minor contributors to total suspension-feeder biomass in the mesopelagic.

Community clearance rates ranged from 2 to 81 L m^-3 day^-1 in the mesopelagic, which corresponds to mesozooplankton clearing between 0.2 and 8.1% of the water each day (Figure 4). Clearance rates were dominated by copepods and euphausiids, with copepods dominating beneath 150 m and both having substantial contributions at shallower depths.

The impact of suspension feeders on particle flux attenuation depends on the average sinking rates of particles. We chose P0704-1 as a cycle with typical community clearance rates and calculated flux attenuation for particles with a full range of sinking speeds (Figure 5). For particles with a sinking speed of 1 m day^-1, nearly all flux would be consumed by suspension feeders within 50 m of the depth of particle creation. By contrast, for particles sinking at a speed of 10 m day^-1, only ∼70% of flux would be expected to be consumed before particles reach a depth of 450 m. For particles sinking at a rate of >50 m day^-1, the impact of suspension feeders becomes negligible. Comparing data from all cycles, we see a similar pattern. If particles sink at 1 m day^-1, anywhere from ∼5 to 50% of sinking particles would be consumed over a 10 m depth range (Figure 6A). For particles sinking with a speed of 10 m day^-1, particle flux attenuation was typically <4%/10 m and for particles sinking at a speed of 100 m day^-1, particle flux attenuation was always <1%/10 m and typically <0.4%/10 m.

To investigate typical particle sinking speeds, we measured the abundance and size of recognizable fecal pellets collected in sediment traps. These pellets were almost always the dominant visually identifiable component of the sinking material, although in the oligotrophic regions the identifiable pellets typically comprised less than half of total carbon flux. We applied allometric–scaling relationships between sinking speed and fecal pellet size to estimate fecal pellet settling velocities. Across all samples, half of the fecal pellet carbon flux was mediated by pellets sinking at a speed slower than 141 m day^-1 (Figure 7). Only 10% of the flux was due to pellets sinking slower than 78 m day^-1 and only 1% was derived from pellets sinking slower than 46 m day^-1. Even for the cycle most dominated by small fecal pellets, half of the pellet flux was due to pellets sinking faster than 85 m day^-1. When considering these results in light of the calculated impact of suspension-feeding zooplankton on flux attenuation of particles with average sinking speeds in the range of 100 m day^-1, it becomes clear that suspension-feeding mesozooplankton are not likely to be playing a dominant role in CFA in the mesopelagic. This does not mean, however, that they play no role in flux attenuation. Rather, the abundance and activity of suspension feeders in the epipelagic may be the reason that so few slowly sinking particles were collected in sediment traps beneath the euphotic zone.

Flux-Feeding Zooplankton and Particle Flux Attenuation

To investigate the potential role of flux- feeding zooplankton in CFA, we quantified the abundances of two prominent taxa of flux feeders in the CCE: Aulosphaeridae (a phaeodarian) and L. helicina (a thecosome pteropod). These are certainly not the only flux-feeding zooplankton in the CCE, hence our calculations of the total contribution of flux feeders should be considered quite conservative.

Aulosphaeridae abundance consistently peaked in the shallow twilight zone between 50 and 150 m depth (Figures 8A–E). Abundances were substantially higher on the P0810 and P1106 cruises (peak cycle average abundances reaching >500 cells m^-3) than on the warm period cruises (P1408 and P1604, peak abundances <50 cells m^-3). Unlike most other plankton community and biogeochemical measurements, there was not a strong correlation between Aulosphaeridae abundance and primary productivity (cf. Biard and Ohman, 2019). Using Eqs 4 and 5, we quantified the potential impact of Aulosphaeridae on CFA (Figures 8F–H). Despite high variability, Aulosphaeridae often had an important role in flux attenuation. Near the depth of peak Aulosphaeridae abundance (∼100 m), the mean (across all cycles) flux attenuation attributable to Aulosphaeridae was 0.69%/10 m (median = 0.21%/10 m, interquartile range = 0.04–0.81%). The maximum Aulosphaeridae flux attenuation (Cycle P0810-1, 100–110 m depth range) was 4.2%/10 m. For comparison, the total flux attenuation mediated by all abiotic and biotic factors (as assessed using sediment traps and ²³⁴Th) was typically in the range of 6–10%/10 m. Thus, the Aulosphaeridae (a single family of phaeodarians) can play a substantial role in CFA, although at most times its contribution is <5% of total flux attenuation.

Limacina helicina was only quantified in MOCNESS samples from two cycles (P1208-3, which had a 31-m deep euphotic zone, and P1208-4, which had a 70-m euphotic zone). L. helicina abundances typically declined from peak abundances of ∼0.5–2 individuals m^-3 in the upper 100 m of the water column to <0.2 individuals m^-3 beneath 100 m depth (Figure 9A). We computed (using Eqs 4 and 6) the potential role of L. helicina in intercepting sinking particles and found that they could potentially intercept between 4 and 10% of sinking particles between the base of the euphotic zone and a depth of 400 m (Figure 9B). However, their impact was concentrated just below the base of the euphotic zone, where on Cycle P1208-3 they consumed an average of 1.2–1.6% of carbon flux/10 m and on P1208-4 they consumed an average of 0.45–0.81%/10 m (Figures 9C,D). As for the Aulosphaeridae, the evidence suggests that this species can intercept a substantial portion of the sinking particles, but their impact is concentrated on the region immediately beneath the euphotic zone.

Discussion

Zooplankton as Gatekeepers to the Mesopelagic

Zooplankton play diverse roles in the epi- and mesopelagic (Steinberg and Landry, 2017). Our results show that suspension-feeding mesozooplankton (especially copepods and euphausiids) are abundant in the twilight zone and can have substantial clearance rates on the magnitude of 50 L m^-3 day^-1 (Figure 4 and Supplementary Figure S2). However, these clearance rates are not sufficient to give them a meaningful impact on CFA of the particles sinking at ∼100 m day^-1, which dominate flux in the region (Figures 5, 6). Instead, their activity leads to near complete consumption of slowly sinking particles near the base of the euphotic zone. This is likely a general result, because studies focused on in situ measurement of particle sinking speeds often find typical velocities on the order of 100 m day^-1 beneath the euphotic zone (Alldredge and Gotschalk, 1988; Trull et al., 2008; Armstrong et al., 2009; McDonnell and Buesseler, 2010; Jackson et al., 2015). It may also partially explain the rapid decrease in marine snow abundance beneath the euphotic zone (Lampitt et al., 1993; Jackson and Checkley, 2011), particularly if these aggregates are assumed to have heterogeneous compositions that lead to a wide range of sinking speeds.

We also note that our calculations were based on nighttime net tows, thus largely excluding any potential impacts of diel vertically migrating suspension feeders. This decision was based on the assumption that diel vertical migrants feed primarily in the surface layers, not at depth. If we calculate clearance rates based on daytime net tows, we find increased mesopelagic clearance rates, particularly at depths between 150 and 250 m. However, this increased clearance rate (daytime values were typically 50% higher than nighttime values) still resulted in a low impact of suspension-feeders on rapidly sinking (100 m day^-1) particles (median across all cycles was <0.4%/10 at all depths), except for Cycle P0810-3. During this cycle, abundant vertically migrating euphausiids were present and, if they were feeding at depth, the total suspension-feeding community could have been responsible for attenuation of 4% of C flux/10 m during daytime hours (in the depth range 200–250 m). Notably, Euphausia pacifica (one of the dominant euphausiids in the CCE) has been shown to play an important role in disaggregation of marine snow in the euphotic zone (Dilling and Alldredge, 2000). However, it is unlikely that they swim as rapidly at their daytime resting depths as they do while actively feeding in the surface ocean. It is thus not currently possible to extrapolate their potential impacts on rapidly sinking particles.

Despite their comparatively weak capacity for intercepting rapidly sinking particles, suspension-feeding zooplankton in the mesopelagic have substantial carbon demands in the CCE (Kelly et al., unpublished) and many other ecosystems (Hernández-Leon and Ikeda et al., 2005; Steinberg et al., 2008b; Burd et al., 2010; Robinson et al., 2010; Schukat et al., 2013; Proud et al., 2017). Potential food sources include slowly sinking particles produced in the euphotic zone, free-living mesopelagic protists, particle-attached protists and microbes that may have been released from sinking particles, and carnivory on vertically migrating or mesopelagic resident zooplankton. Slowly sinking particles may also be generated from rapidly sinking particles through the swimming or feeding actions of mesozooplankton (Dilling and Alldredge, 2000; Goldthwait et al., 2004; Iversen and Poulsen, 2007). Indeed, such particle transformations, whether mediated by zooplankton, microbes, or abiotic processes, are likely to continually generate additional slowly sinking particles throughout the water column. However, these slowly sinking particles should not be expected to contribute substantially to particle flux in regions with abundant suspension-feeding zooplankton, because the suspension feeders will efficiently consume slowly sinking particles (e.g., Figure 6A).

In contrast, flux-feeding zooplankton can be substantial loss terms for rapidly sinking particles. The mean percentage flux attenuation for Aulosphaeridae in the shallow mesopelagic was 0.69%/10-m depth horizon corresponding to the depth of maximum Aulosphaeridae abundance (Figure 8). For comparison, in the same depth range total flux averaged 7.1%/10 m (Figure 3). This implies that a single family of giant Rhizaria (∼2-mm) may be responsible for nearly 10% of the flux attenuation in the layer immediately beneath the euphotic zone. Similarly, the pteropod L. helicina is abundant beneath the euphotic zone and may be responsible for ∼10–20% of total flux attenuation in the depth ranges where it is most common. These are certainly not the only flux feeders in the CCE. Many other pteropod and rhizarian taxa occur in the CCE (Kling and Boltovskoy, 1995; Bednaršek and Ohman, 2015; Biard and Ohman, 2019). Flux feeding has also been suggested for the copepods Neocalanus cristatus and Spinocalanus antarcticus (Dagg, 1993; Kosobokova et al., 2002) and the polychaetes Poeobius and Poecilochaetus (Hamner et al., 1975; Uttal and Buck, 1996; Christiansen et al., 2018). Flux feeding may also be a part-time feeding mode used by a diverse class of organisms during low food periods. For instance, in the absence of phytoplankton prey, the copepod Acartia tonsa will behave as a non-motile feeder that will occasionally feed on fecal pellets that pass within its detection radius (Poulsen and Kiorboe, 2005). It is unknown if such behavior is widespread amongst the abundant diel vertically migrating zooplankton that spend half of their life in the low-prey mesopelagic environment. Another similar strategy has been termed “active flux feeding” or “plume finding” by Stemmann et al. (2004b) and involves cruise-feeding copepods that detect the solute plume behind a sinking aggregate (Kiørboe and Thygesen, 2001), leading to capture rates that are dependent on sinking speed. We could not quantify the impact of this feeding strategy, because the taxa that utilize it have not been identified. However, if it is widespread amongst mesopelagic copepods [as assumed by Stemmann et al. (2004a)] it would lead to substantially greater attenuation of particle flux in the twilight zone.

Also of import, these flux feeders not only intercept and consume sinking particles, but also produce sinking fecal pellets. L. helicina fecal pellets have been shown to contribute 19% of the POC flux in a coastal bay near Antarctica (Manno et al., 2009), while “mini-pellets” produced by phaeodarians were abundant contributors to sinking flux in the Eastern Tropical Pacific and Northeast Atlantic (Gowing and Silver, 1985; Lampitt et al., 2009). Furthermore, due to their dense aragonite shells and siliceous tests, respectively, dead pteropods and phaeodarians likely contribute substantially to sinking flux and are often found in sediment trap material (Takahashi and Honjo, 1981; Bathmann et al., 1991; Fabry and Deuser, 1992; Michaels et al., 1995; Biard et al., 2018). Future research will need to focus on the roles of these organisms in transforming not just the quantity but also the quality of sinking particles in the mesopelagic.

Microbes and Zooplankton in the Mesopelagic

The abundance (and impact on flux attenuation) of suspension-feeding and flux-feeding zooplankton generally decreases with depth in the mesopelagic. In particular, the flux-feeding zooplankton that we believe play a disproportionately strong role in twilight zone flux attenuation both decreased sharply in abundance beneath a depth of ∼100 m. However, the percentage (but not absolute magnitude) of CFA increased slightly with depth between 80 and 150 m depth (Figure 3). While the substantial variability in flux attenuation profiles prevents us from making definitive conclusions about these patterns, this mismatch between flux attenuation patterns determined from bulk flux estimates (sediment traps and ²³⁴Th) and the abundance of flux feeders points to the importance of other sources of carbon remineralization. Stemmann et al. (2004a) suggested based on modeling results for the Mediterranean Sea that flux feeders may be particularly important in the shallow twilight zone, while microbial degradation becomes increasingly important at the lower particle concentrations experienced in deeper waters. Such a result seems plausible for multiple reasons. First, free-living zooplankton will have substantially greater difficulty obtaining sufficient energy for basal metabolic needs and respiration in the deeper ocean, while particle-attached microbes should have ample food at all depths. Second, the abundance of particle-attached microbes on a sinking particle may be expected to increase with age (and depth) of the particle, because it may be continually colonized by new microbes while it sinks, and these microbes are likely to multiply on the sinking particle.

Results from other studies in the CCE can highlight the potential impact of microbes. Simon et al. (1990) estimated that marine snow turnover times with respect to bacterial remineralization ranged from 20 to 100 d. This is similar to the turnover times expected to be mediated by suspension-feeding zooplankton, and suggests that bacteria have only a minor impact on the flux of rapidly sinking particles. However, Samo et al. (2012) measured bacterial carbon production rates (including particle-attached and free-living bacteria) at a depth of 100 m that ranged from ∼10 to 100 μmol C m^-3 day^-1. For comparison, we quantified that the total flux attenuation at this depth was 75 μmol C m^-3 day^-1 (median = 66 μmol C m^-3 day^-1; 95% confidence interval of up to 313 μmol C m^-3 day^-1). It is also unclear whether decomposition rates for marine snow aggregates are comparable to those of the fecal pellets that dominate sediment trap material in the CCE. Gowing and Silver (1983) suggested that fecal pellets collected near our study region had substantial contributions of interior bacteria that likely originated as enteric or ingested, but digestion-resistant bacteria. These communities would likely be substantially different, and with different biogeochemical impacts, than those found on aggregates.

Heterotrophic protists can also play dominant roles in the microbial communities consuming fecal pellets (Poulsen and Iversen, 2008). Gutierrez-Rodriguez et al. (2018) used 18S sequencing to investigate protistan sequences in sediment traps on our P1408 cruise. They found a consistent and substantial increase in the relative contribution of dinophytes in un-preserved trap samples relative to formaldehyde-preserved trap samples. Dinophytes typically increased from <10% of total protistan reads to >50%, despite deployment times that were only 3.25 days. Stramenopiles (mainly heterotrophic nanoflagellate taxa) also increased in the unpreserved samples. This suggests potentially rapid growth rates for these protists on sinking particles, with a commensurate role in feeding on either detritus contained in the sinking particles or other microbial taxa transported within the particles.

Multiple studies from other regions have investigated the differing roles of zooplankton and microbes in the mesopelagic (Simon et al., 2002; Robinson et al., 2010; Steinberg and Landry, 2017). Steinberg et al. (2008b) found that bacterial carbon demand exceeded zooplankton demand in the North Pacific subtropical gyre by a factor of 3, although the two were approximately equal in the subarctic gyre. The combined respiration of these two groups was found to be larger than that entering the mesopelagic through sinking particles, a result that has been consistently found in other regions including the subtropical and north Atlantic (Boyd et al., 1999; Reinthaler et al., 2006; Baltar et al., 2009). There are also likely to be extensive synergistic and antagonistic interactions between zooplankton and microbes in the mesopelagic. Zooplankton egestion and excretion at depth provide available organic matter for bacteria (Hannides et al., 2009; Saba et al., 2011; Kiko et al., 2016). Release of organic matter mediated by extracellular enzymes produced by bacteria may contribute to the formation of plumes behind sinking particles that aid zooplankton’s abilities to find such particles (Kiørboe and Thygesen, 2001; Arnosti, 2011). The microbial communities themselves may also serve as nutrient- and carbon-rich food sources for zooplankton (Wilson et al., 2010). It is even possible that fragmentation of particles by mesozooplankton serves to enhance microbial activity and trophic transfer to mesozooplankton (Mayor et al., 2014). Unraveling these potential interactions will require collaborative efforts between microbial and zooplankton ecologists and biogeochemists.

Epipelagic-Mesopelagic Coupling in the CCE

Our result that suspension-feeding zooplankton are abundant in the mesopelagic, but play only minor roles in consuming sinking particles, raises important questions about how they satisfy their metabolic demands. Based on results from our P0704 and P0810 cruises, Kelly et al. (unpublished) estimated that total mesozooplankton (mostly suspension-feeders) respiration in the mesopelagic ranged from 3.2 to 18 mg C m^-2 day^-1 across these cruises. These values substantially exceed the total carbon that we calculate suspension-feeders would be able to consume from rapidly sinking particles settling through the mesopelagic. However, the Kelly et al. (unpublished) results also suggest a resolution to this apparent imbalance. Diel vertically migrating taxa were abundant in the region and their active transport provides substantial energy subsidies to mesopelagic food webs. Mesopelagic resident organisms derive energy directly (through predation) and indirectly (through their release of dissolved organic carbon that supports microbial communities) from vertical migrants. Indeed, results showed that nearly half of the organic matter consumed by mesopelagic resident mesozooplankton was derived (directly or indirectly) from foodweb pathways that originated with active transport by vertical migrants rather than pathways originating from sinking particle flux (Kelly et al., unpublished).

This highlights an important reality of mesopelagic ecosystems. Although we most often think of these ecosystems as being supported by sinking particles (and to a lesser extent diel vertical migration), mesopelagic communities feature complex ecological relationships between particle-attached and free-living bacteria, protistan bacterivores, and zooplankton (Robinson et al., 2010). These zooplankton communities are diverse and have many feeding modes including suspension-feeding (likely on protists and suspended organic matter), flux-feeding, predation, and parasitism (Steinberg et al., 2008a; Wilson et al., 2010; Bode et al., 2015). The quality (in addition to quantity) of sinking particles also changes with depth as these particles are modified by microbial and zooplankton communities, new particles are created through defecation, mortality, molting, and discarded feeding webs, and particle size spectra are reshaped by aggregation and disaggregation (Wakeham et al., 1997; Burd and Jackson, 2009; Turner, 2015). Future studies are necessary to fully investigate the impacts of these different feeding traits on the BCP (Barton et al., 2013) and the impacts that future warming, ocean acidification, changes in optical attenuation, and potential shoaling of the oxycline may have on mesopelagic zooplankton communities (Richardson, 2008; Wishner et al., 2013; Cripps et al., 2014; Hauss et al., 2016; Ohman and Romagnan, 2016).

Conclusion

Our results offer compelling evidence that two flux-feeding zooplankton taxa (the phaeodarian Aulophaeridae and the pterpopod L. helicina) can exert substantial impact on sinking particles in the shallow twilight zone. Both taxa have the capacity to consume greater than 1% of sinking particles per 10-m vertical depth range at the depths at which they are most abundant, which equates to ∼10–20% of total CFA across those same depth ranges (typically the first 50–100 m beneath the euphotic zone). These are unlikely to be the only flux feeders in the mesopelagic. Indeed many other rhizarians and pteropods are found in the CCE and many other taxa are suspected of full or part-time flux-feeding behavior. Flux-feeders are likely particularly important in consuming rapidly sinking particles that would otherwise penetrate into the deep ocean. The abundance of suspension-feeding zooplankton in the ocean suggests that slowly sinking particles will have a relatively minor contribution to sinking carbon flux in the mesopelagic because they will be rapidly consumed. While our results suggest that feeding mode leads to very different biogeochemical importance for these functional groups of zooplankton, substantial work is needed to directly quantify the impact of these organisms in situ and to investigate the spatial and temporal distributions of each feeding trait amongst the diverse zooplankton communities found in the ocean (Barton et al., 2013). The qualitatively different impact of each feeding mode on remineralization length scales in the ocean further suggests that flux feeders should be included in modeling efforts undertaken to investigate marine carbon sequestration.

Data Availability

The datasets generated for this study are available on request to the corresponding author.

Author Contributions

MS and TK were responsible for the sediment trap and thorium measurements. MO was responsible for the MOCNESS zooplankton data. MO and TB were responsible for the UVP deployments. TB was responsible for the UVP rhizarian analyses. MS wrote the manuscript. All authors edited the final version of the manuscript.

Funding

This work was funded by the NSF Bio OCE grants to the CCE LTER Program: OCE-0417616, OCE-1026607, OCE-1637632, and OCE-1614359. A portion of this work was performed at the National High Magnetic Field Laboratory, which is supported by the National Science Foundation Cooperative Agreement No. DMR-1644779 and the State of Florida.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

We thank the captains and crews of the R. V. Melville, R. V. Thompson, and R. V. Sikuliaq as well as our many collaborators in the CCE LTER Program, without them this study would not have been possible. We also thank N. Bednaršek for the pteropod enumerations. Data used in this manuscript can be found on the CCE LTER DataZoo website: https://oceaninformatics.ucsd.edu/datazoo/catalogs/ccelter/datasets.

Supplementary Material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars.2019.00397/full#supplementary-material

References