Carbon and Calcium Carbonate Export Driven by Appendicularian Faecal Pellets in the Humboldt Current System off Chile

The role of appendicularian faecal pellet (FPa) size fractions on coccolithophore-derived particulate organic carbon (POC) and calcium carbonate (CaCO3) export to the deep sea was assessed from sediment traps within a period of ten years (1995–2004) off Coquimbo (CQ, 30°S) and five years (2005–2009) off Concepción (CC, 36°S) in the Humboldt Current System (HCS) off Chile. The composition and size distribution of 1,135 FPa samples from sediment traps deployed at 2,300 and 1,000 m depths showed non-linear, inverse relationships between the FPa size-fractions and their volume-specific POC and CaCO3 contents, which were up to ten times higher for small (<100 µm in diameter) than large (>100 µm) FPa. On average, 13 and 2% of the total POC and CaCO3 fluxes, respectively, were contributed mainly by small FPa (90%), with maxima during the autumn and summer. Thus, a non-linear, exponential model of volume-specific POC and CaCO3 contents of FPa substantially improved vertical flux rate estimates. In the HCS, annual carbon flux based on a non-linear FPa carbon load was double the estimate assuming a linear-volume to carbon load for FPa (345 and 172 kton C y−1). We recommend a widespread consideration of this non-linear model in global carbon estimates.

efficiency of the total POC flux temporally and spatially depend on changes in zooplankton and phytoplankton community compositions, as well as on changes in the biological processes (i.e. zooplankton grazing rate, faecal pellet production rate, phytoplankton aggregation, microbial degradation) that affect the dynamics of particle flux from the euphotic zone in oceanic provinces 16 and upwelling areas off central Chile 12 .
In the HCS, El Niño (EN) events show oceanographic changes such as intrusion of oceanic, low-nutrient, warm, oxygenated waters into the coastal areas that trigger positive sea-surface temperature anomalies. During EN events, the plankton assemblages shift to smaller species during the warm phase; however, the total phytoplankton biomass, primary production, and vertical flux of carbon do not change 3,[17][18][19] . Conversely, during "La Niña" events, the shallowed nutricline leads to increased nutrient supply, phytoplankton biomass, and primary production 20 .
Several hundred sediment trap studies since the 1970s have reported the pivotal role of zooplankton faecal pellets (FP) in the carbon biogeochemical cycle 14,[21][22][23] . This plethora of information demonstrates the importance of zooplankton FP in the export of particulate organic carbon (POC) from the photic zone to the deep sea 24,25 . The contributions of FP produced by different functional groups, including euphausiids, appendicularians, copepods, salps, and microzooplankton, to vertical flux of POC also have been reported [24][25][26] , where appendicularians stand out because of their high FP production rates (Supplementary Table S1) and carbon fluxes (Supplementary  Table S2), but relatively low sinking rates (Supplementary Table S3). The high variability of those estimates seems to relate to many physical and biological interactions that affect export versus recycling of the carbon and nitrogen in FP. For example, appendicularians usually represent a minor fraction of total mesozooplankton biomass; however they can process large amounts of food in a short time and produce numerous FP that contribute to export production 27,28 , playing significant roles in the export of the POC 14 and calcium carbonate (CaCO 3 ) to the deep sea  in many disparate areas of the world's oceans.
Coccolithophores also are an important component of the phytoplankton in the HCS off Chile and the world ocean 6,30,31 . Their transport to deep layers of the ocean has been principally by means of FPa, where a linear relationship between POC or CaCO 3 contents with volume-derived FPa size has been assumed 32 . The reasoning is that the shells (coccoliths) increase the FPa sinking speed 33 , although there is little information on the appendicularian diets and the physical and chemical composition of FPa, which are usually assumed to be invariable and independent of the size of the animal and the extruded faecal pellet. We tested this assumption in order to determine the role of FPa of different sizes in the fluxes of POC and biogenic carbonate to the deep sea at two sites (CQ, 30°S; and CC, 36°S) ( Table 1), in the HCS off Chile (Fig. 1).

Results
fpa size-spectrum. The size-spectrum of FPa sizes (diameter and length) and volumes were analysed from 5,102 FPa isolated from 201 sediment trap samples from CQ and CC in the HCS, covering most of the study period. We found that 89% of the FPa were between 10 and 100 µm in diameter, highlighting the pivotal role of small FPa in fluxes of carbon and calcium carbonate in the oceanic region of the HCS off central Chile (Fig. 2). The remaining FPa (11%) were between 101 and 390 µm. carbon and calcium carbonate contents in fpa. The inverse relationship between the FPa size fractions and their volume-specific POC content (mg C mm −3 ) was best represented by a power function (f(x) = 0.0087x −0.556 ; R 2 = 0.85). This was obtained from measurements of 1,135 FPa isolated from sediment trap samples collected at stations CQ and CC during most of the study period. The POC in FPa was highly variable (0.02-0.77 mg C mm −3 ), with the highest volume-specific values (~0.2-0.77 mg C mm −3 ) in small FPa (Fig. 3a) (Fig. 3). We also expressed the average concentration of carbon per FPa (mg C FPa −1 ) for each size group (small, medium, and large) as functions of their average biovolumes. Different slopes in each of the linear equations among the three size groups denote a highly non-linear and inverse relationships between FPa carbon content and their biovolumes in FPa ( Fig. 3b-d).

Station Position
Coccolith abundance, CaCO 3 content, and coccolithophore composition within FPa. The number of coccolithophore plates in the FPa matrix were quantified by species. The FPa ranged between 0.08-0.35 mm in diameter and 0.0006-0.080 mm 3 in biovolume. The coccolith abundance and coccolithophore-derived calcium carbonate (CCa) concentration showed non-linear inverse exponential functions when related with its biovolume-specific (per mm 3 ) value at both sites. At CQ, the mean coccolith and CCa contents were 5.1 to 24.3 × 10 6 coccoliths mm −3 and 0.1 to 0.3 mg CaCO 3 mm −3 , respectively. The maximum biovolume-specific abundance of coccoliths and CCa were observed in the smallest FPa (2.1 × 10 7 coccoliths mm −3 ; 0.24 mg CaCO 3 mm −3 ) and the minimum in the largest (1.3 × 10 7 coccolith mm −3 ; 0.18 mg CaCO 3 mm −3 ) (Fig. 4a,b). In contrast, at the CC site, these parameters ranged from 7.8 × 10 6 coccoliths mm −3 and 0.091 mg CaCO 3 mm −3 to 2.9 × 10 6 mm −3 and 0.027 mg CaCO 3 mm −3 , respectively. The smallest FPa showed higher coccolith abundance and CCa concentration (6.2 × 10 6 coccoliths mm −3 ; 0.064 mg CaCO 3 mm −3 ); in contrast, the largest FPa had the least (5.5 × 10 6 coccoliths mm −3 ; 0.056 mg CaCO 3 mm −3 ) (Fig. 4a,b). On average, seventeen genera/species of coccolithophore were found in the FPa from both sites; however, there were important differences between the CQ and CC sites. The abundance of coccoliths and CCa concentrations were ~3 and 1.6 times higher in CQ than in CC site. The main coccolith type observed was Emiliania huxleyi, which represented >70% of the abundance of all recorded species and was ~3 times more abundant in CQ than the CC site.   The CCa concentrations in FPa were dominated by coccoliths belonging to the species Calcidiscus leptoporus, which contributed ~45% of total carbonate biomass, while E. huxleyi and Helicosphaera carteri together contributed ~40%. These three species represented >80% of the carbonate biomass in FPa in both sites ( Table 2).  In the CC site, FPa-mediated fluxes averaged 19.8% (3.1 × 100)/15.6) and 4% (0.86 × 100)/21.36) of the total POC flux to the 1,000 m (average 15.6 mg C m −2 d −1 ) and 2,300 m (average 21.4 mg C m −2 d −1 ), respectively. While that the general average was 7.7% of the total POC flux (19.7 mg C m −2 d −1 ) (Fig. 6a,b and Table 3).
The total PIC/POC ratios for the CQ site (1.2) was higher than for CC site (0.28). On the other hand, the PIC/POC ratio was slightly higher in the FPa from CC-site (0.037, from both depths), than for CQ-site (0.022), respectively (Table 3).
In the CQ site, maximum fluxes of FPa-mediated POC ranged between 2.3 and 4.4 mg C m −2 d −1 and occurred during summer-autumn 1995 (Fig. 5b) For the CQ site, the FPa-mediated POC flux during 1995 (1.78 mg C m −2 d −1 ) was statistically different (Mann-Whitney test, n = 16, p < 0.01) from all other years combined; similarly, 1996-1998 (0.  Table 2. Coccolithophore species found in the matrix of appendicularian faecal pellets (FPa) collected in sediment traps deployed at the CC and CQ sites off Chile and their contributions to total average carbonate and standard deviation (SD), The species-specific abundance of coccoliths per mm 3 (Fig. 6b,d). No significant seasonal differences in FPa-mediated POC or CaCO 3 were observed (Mann-Whitney test, N = 35, p > 0.05) during El Niño, La Niña, or "normal" periods (see Figs 5 and 6) in either of the two study areas (Mann-Whitney test, N = 11, p > 0.05).

Discussion
The Biological Carbon Pump is an important component of the global carbon cycle in which the factors influencing zooplankton faecal material export efficiency and its controlling factors, particularly for PFa export, remain poorly understood. Overall, the magnitude and efficiency of the total POC flux in time and space depend on changes in the zooplankton and phytoplankton community compositions, as well as on changes in biological processes (i.e. zooplankton grazing rate, FP production rate, phytoplankton aggregation, microbial degradation) that affect the dynamics of particle flux from the euphotic zone in oceanic provinces 16 and upwelling areas off central Chile 11 . The bulk of the faecal material usually is removed quickly from the upper water column due to microbial degradation and zooplankton activity 34,35 , although zooplankton faecal material still is a prominent fraction of total POC export in many areas of the world oceans 13,14 .
The composition of the faecal material collected in sediment traps deployed at the CQ and CC sites was represented mainly by three functional groups: appendicularians, euphausiids, and undetermined (where high degradation, amorphous or broken pieces precluded classification).
Published estimates of the POC content of FPa and its contribution to carbon export from the HCS have assumed an average carbon:volume ratio of 0.048 mg C mm −3 14,32 , but we demonstrated in this study that FPa size and both carbon and carbonate contents were related non-linearly (Figs 3a, 4b). For example, the mean carbon content per mm −3 in small FPa was five times higher than in large FPa (0.26 and 0.041 C mm −3 , respectively). This result is important for quantitative estimates of the biological pump intensity, assuming that the PFa contents would be released in the same proportions as found in the sediment traps in this study, where more than 90% of the FPa were small. Our exponential conversion increased the contribution of the FPa-mediated carbon flux by ~50% (1.04 mg C m −2 d −1 ) over that obtained using a linear conversion factor between carbon content and FPa-size 14 . The bar at the bottom shows "El Niño" (black), "La Niña" (gray), and "normal" (white) periods (www.cpc.noaa.gov/products/ analysis_monitoring/ensostuff/ensoyears.shtml). n.d. along the horizontal axes show periods of no data due to equipment failure.
The calcareous components found in the matrix of FPa from the CQ and CC sites were mainly coccolith plates (mean 10.3 × 10 6 plates mm −3 , Fig. 4A), from which the most abundant coccoliths were Calcidiscus leptoporus, Helicosphaera carteri and Emiliania huxleyi, which together averaged 86% of the CCa or 2.4 × 10 −13 mg CaCO 3 µm −3 in PFa isolated from sediment trap samples deployed at the CQ and CC sites ( Table 2).
While existence of coccoliths and coccospheres embedded in the matrix of FPa have been reported 28 , but no prior quantitative information on their contribution to carbonate flux in the ocean has been previously made. The coccoliths and coccospheres within the matrices of FPa are important in the export of carbonate and increase the sinking velocity of FPa to the deep ocean. Previous work has demonstrated that 1-mm particles that are 90% carbonate by mass (and 10% POM) sink ~fivefold faster than particles that are 60% carbonate by mass (and 40% POM) and ~1.75 times faster than particles that are 90% silicate by mass (and 10% POM) 33 . An individual coccolith has a sinking velocity of 0.1 m d −1 or slightly higher, depending on species types and calcification degree. For example, E. huxleyi, G. oceanica, and C. leptoporus coccolith settling velocities (0.3, 0.7, and 4.3 m d −1 , respectively) were higher than diatom sinking rates, which ranged from 0.07 m d −1 (Thalassiosira weissflogii) to 0.2 m d −1 (Thalassiosira oceanica) 36 . This information clearly suggests that FPa loaded with coccoliths and coccospheres can reach a higher sedimentation velocity than one loaded with diatom frustules, demonstrating the effectiveness of coccoliths and coccospheres in the transport of carbon and carbonate to the deep ocean.
The absence of a linear relationship between FPa-size and POC and CaCO 3 content may be attributable to differences in the size spectrum of the food ingested, with small FPa being more uniform and compact than large PFa. To test this hypothesis, we isolated large and small FPa from a sediment trap deployed at 2,300-m depth during different seasons between 1995 and 2004 and analysed the particle content qualitatively. Large FPa contained many relatively large pieces of diatom frustules (20-63 µm diameter), mostly from the genus Pseudo-nitzschia, but also from the abundant Thalassiosira spp. and Planktoniella sol. In addition, the remains of radiolarians (groups Spummellaria, Nasselaria, and Phaeodaria), tintinnid loricae (genera Codonellopsis, Dictiocysta, Protorhabdonella, Undella, and Dadayiela), and silicoflagellates, such as Dictiocha fibula, were frequently observed. In contrast, microscopic analysis of small FPa demonstrated a preponderance of coccoliths, with occasional remains of diatoms and tintinnid loricae embedded in the FPa matrix. Large oikopleurid species have food-concentrating filters www.nature.com/scientificreports www.nature.com/scientificreports/ with a coarser mesh than small ones 37 , suggesting a direct relationship between pore dimensions of appendicularian filters and body size. Our study suggested that large appendicularians utilized a broad food-size spectrum (diatoms, coccolithophores, tintinnids, radiolarians, etc.), while the diet of small appendicularians was more restricted, which partially explained the small mineral skeletons in small FPa.
The FPa biovolume to POC (or CaCO 3 ) content models that should be used to estimate the role of FPa in biomineral vertical fluxes follow size-dependent non-linear relationships, indicating that the FPa size fraction needs to be considered to improve the accuracy of these estimates. We strongly recommend a more taxa-specific analysis by microscopy, which is more time consuming but essential when dealing with size-dependent FPa-mediated POC and carbonate export processes. This view is supported by findings of changing composition of sinking particles across a region with unchanging carbon flux, suggesting that variability in the mechanisms of carbon flux 38 or the role of phytoplankton biominerals as ballast 39 may not be reflected by bulk measurements. In addition, coccolithophore calcium carbonate is usually considered to have a low carbon flux and high export efficiency 24 . Thus, CCa is generally not included in calculations of FPa carbon export, resulting in up to 4% underestimation of total carbon flux to deeper areas of the HCS.
High seasonal and inter-annual variability in POC flux was observed in the time series station off the CQ site (Fig. 5). The estimated average FPa-driven POC flux at the CQ site (0.86 mg C m −2 d −1 ) was slightly higher than that reported for the same area during 1993-1995 (~0.7 mg C m −2 d −1 ) 14 , while that recorded for the CC site was substantially higher (1.5 mg C m −2 d −1 ) than at the CQ site (Figs 5b, 6b).
Overall, the use of the FPa-size-based carbon content model of this study increased the estimated contribution of FPa to total POC export by a factor of 34% (CC site) and 46% (CQ site) over those obtained from a linear FPa-volume to carbon content model.
The highest average FPa flux occurred during the productive summer period, coinciding with intensification of the winds favouring upwelling and Ekman transport 14 , which may enhance local productivity and carbon export. We did not find significant differences (p > 0.05) in the PFa-mediated fluxes of POC or CaCO 3 among El Niño, La Niña, and "normal" periods (showed in Figs 5 and 6), suggesting that the diet of the appendicularians was not severely affected by these events.
Reports during the El Niño event in the HCS show that the seasonal upwelling ceases, the water column became warmer, and the thermocline and nutricline deepens significantly during the passage of coastal-trapped waves 40,41 . These events decrease phytoplankton productivity and alter the trophic chain, changing the biological composition and physical dynamics of the coastal and marine ecosystems 18,41 . Both phytoplankton and zooplankton shift toward small-sized species that have been claimed to prevail during El Niño 17,18,40 . Thus, it is highly probable that the composition and size spectrum of small food particles used by appendicularians, down to colloids (<0.2 µm) 42 , is little affected by El Niño events 37,43 , a situation that might be extrapolated to future global warming scenarios on the FPa-mediated carbon and CaCO 3 export fluxes.   www.nature.com/scientificreports www.nature.com/scientificreports/ For the CQ site, fluxes of calcium carbonate and POC into sediment traps below 2,300 m were correlated (y = 6.889x + 15.417, n = 127; R 2 = 0.51) (Fig. 7a), which is consistent with published data for various geographic regions 3,14,[44][45][46][47][48][49][50][51][52][53] , like the PIC/POC ratios ( Table 3). Some of these oceanic and coastal areas (such as the CQ site) exhibit relatively low productivity 1 that favours the export of CaCO 3 (mean 48.1 mg m −2 d −1 ) over POC (mean 4.7 mg m −2 d −1 ) (Fig. 5a,c) and possibly carbonate-driven POC export as the main ballast mechanism, which would explain the high PIC/POC ratio (>1) observed in this site. Thus, biominerals increase substantially sinking velocity because of their high densities and may serve to protect organic matter from degradation during transit to the ocean floor [53][54][55] . Conversely, POC and CaCO 3 at CC, were not significantly correlated (y = 1.090x + 27.56, n = 30; R 2 = 0.16) (Fig. 7b), and the PIC/POC ratio was lower (<1) than the CQ site, which may be attributable to a very productive upwelling centre that supports massive diatom blooms 1 and favours the export of both CaCO 3 (mean 46.7 mg m −2 d −1 ) and POC (mean 19.7 mg m −2 d −1 ) (Fig. 6a,c) with perhaps silicate-driven POC export as the main ballast mechanism.
In the HCS, an average appendicularian abundance is estimated at 21.  Table S4). We estimated a total area of the HCS off Chile of 910,000 km 2 , assuming a total latitudinal length of 2,600 km between 18.5° and 41.5°S and a width of 350 km 58 . Because most appendicularians occur within the upper 50 m of the water column 59,60 , an estimate of 1.08 × 10 9 appendicularians within this upper layer would produce 7.12 × 10 19 FPa y −1 along the whole HCS. Moreover, considering the size-spectrum of FPa found at both study sites (Fig. 2), we assumed that ~90% of the FPa produced are small FPa (<100 µm of diameter) and 10% large. As the small FPa had an average biovolume and carbon content of 0.26 mg C mm 3 and 0.000364 mg C, respectively, while equivalent values for a large FPa were 0.041 mg C mm 3 and 0.0015 mg C (Fig. 3a and Supplementary  Table 4). Finally, we estimated that 3.4 × 10 16 mg C y −1 or 0.034 Gt C would be released annually as FPa in the upper 50 m water column of the HCS (Supplementary Table S4).
Given that the FPa vertical flux averaged 1.04 mg C m −2 d −1 (Figs 5b, 6b), we can estimate that 345.5 × 10 6 kg carbon (or 0.000345 Gt carbon) are exported annually down to 2,300 m depth in the HCS, which represents 1% of the FPa carbon standing stock within the upper 50 m of the water column, suggesting extensive recycling of FPa between the photic zone and deeper layers of the HCS (Supplementary Table 4). We lack data on POC fluxes and degradation rates between 2,300 m of depth and the sea floor (circa 4,500 m depth, in CQ and CC), but information on POC recycling in many disparate regions of the world oceans indicate that an additional POC loss of ~30% might occur 61 . This suggests that ~0.7% of the standing stock of the FPa-mediated carbon export from the upper 50 m stratum could reach the ocean floor along the HCS off Chile. This is in line with reports suggesting that <3% of the generated upper production reaches bathypelagic depths (>1,000 m) in the deep sea 62 , mainly due to the highly complex and variable food web in the water column 63 that removes most of the sinking particles, such as the FPa.
Our non-linear FPa-size derived carbon to volume ratio estimated an average POC flux in the HCS off Chile to be 345 kton C y −1 , twice as much the estimated with the linear model (172 kton C y −1 ). This study only includes the HCS off Chile, a very productive ecosystem that might differs from the rest of the global oceans (in food supply quality and quantity). Nevertheless, a widespread characteristic of the model is its non-linear, FPa-size derived nature, that we recommends to be considered for future estimated where FPa are key vehicles of PIC and/or POC carbon export. Analysis of poc in fpa. Intact FPa (1,114 units, 528 from CQ and 586 from CC) of various sizes between 50 and 340 µm in diameter were isolated from sediment trap samples that covered most of the study period, and measured (major and minor axes of the FPa ellipsoid shape) by using a stereo microscope (Leica model MZ6). After a gentle wash with de-carbonated mineral water, using petri dish, to maintain the integrity of the FPa and photographed (Nikon Coolpix model 4500) using the stereo-microscope, then they were placed on 23 Whatman GF/F glass fibre filters in batches depending on their size (from 14 large to 142 small FPa). The images were processed in ImageJ software to estimate the volume of each faecal pellet and the total volume of each batch. Finally, subsamples of FPa on GF/F filters were acidified with 0.2 mL (~4 drops with a Pasteur pipette) of 2 N HCl to remove particulate inorganic carbon, washed with distilled water to remove excess acid, dried at 50 °C, and stored in hermetic plastic bottle with silica gel to await organic carbon (CHN) analysis, as described above. Due to the high variability in the diameters of the FPa used for elemental carbon analysis in the time series data from 1995 and 2004 (Fig. 2), average POCs of 0.26 mg mm −3 were used for small FPa (<100 µm diameter) and 0.041 mg mm −3 for large FPa (>100 µm diameter).
Total carbonate, coccolith abundance, and coccolithophore composition in FPa. 110 intact FPa (50-340 μm in diameter) were removed from sediment trap samples deployed at CQ (44 FPa pooled in 16 bunches) and CC (66 FPa pooled in 18 bunches) throughout the study period. This was made with the aid of a stereomicroscope (Leica model MZ6) at ×4 magnification and placed in a Petri dish with pre-filtered de-carbonated mineral water to prevent coccolith dissolution. Faecal pellets were washed five to eight times with filtered de-carbonated water and photographed. Images were processed with ImageJ software to estimate faecal pellet size and biovolume (Figs 2, 4) by applying prolate spheroid geometry 66 .
In order to analyse the composition of coccospheres and coccoliths within FPa, 1-10 pellets (number depending on size) were removed from the Petri dish, gently placed in 2.5-mL vials containing 2 mL de-carbonated mineral water and sonicated at 50 to 60 Hz for 30 seconds 67 . The contents of the vials were placed in sedimentation chambers and coccospheres and coccoliths identified and counted with an inverted microscope (×1,000 magnification). The coccoliths and coccospheres observed in a known area of the bottom of the chamber were counted and the data extrapolated to the total volume of the faecal pellets 6 . The samples also were examined by scanning electron microscopy to confirm the coccolithophores species identified with light microscopy. Finally, the number of coccoliths per coccolithophore species were transformed to CaCO 3 using conversion factors from the literature 68 , and the result was extrapolated to the total biovolume of the FPa measured for each sample.

Statistical analyses.
A Kolmogorov-Smirnov test (Minitab software) showed that our data for POC and CaCO 3 fluxes mediated by FPa in the CQ and CC sites were not distributed normally. We therefore used the non-parametric Mann Whitney test for multiple statistical comparisons (confidence limits set at 95%).