Bathypelagic particle flux signatures from a suboxic eddy in the oligotrophic tropical North Atlantic : production , sedimentation and preservation

Particle fluxes at the Cape Verde Ocean Observatory (CVOO) in the eastern tropical North Atlantic for the period December 2009 until May 2011 are discussed based on bathypelagic sediment trap time-series data collected at 1290 and 3439 m water depth. The typically oligotrophic particle flux pattern with weak seasonality is modified by the appearance of a highly productive and low oxygen (minimum concentration below 2 μmol kg−1 at 40 m depth) anticyclonic modewater eddy (ACME) in winter 2010. The eddy passage was accompanied by unusually high mass fluxes of up to 151 mg m−2 d−1, lasting from December 2009 to May 2010. Distinct biogenic silica (BSi) and organic carbon flux peaks of ∼ 15 and 13.3 mg m−2 d−1, respectively, were observed in February–March 2010 when the eddy approached the CVOO. The flux of the lithogenic component, mostly mineral dust, was well correlated with that of organic carbon, in particular in the deep trap samples, suggesting a tight coupling. The lithogenic ballasting obviously resulted in high particle settling rates and, thus, a fast transfer of epi-/mesopelagic signatures to the bathypelagic traps. We suspect that the twoto three-fold increase in particle fluxes with depth as well as the tight coupling of mineral dust and organic carbon in the deep trap samples might be explained by particle focusing processes within the deeper part of the eddy. Molar C : N ratios of organic matter during the ACME passage were around 18 and 25 for the upper and lower trap samples, respectively. This suggests that some productivity under nutrient (nitrate) limitation occurred in the euphotic zone of the eddy in the beginning of 2010 or that a local nitrogen recycling took place. The δ15N record showed a decrease from 5.21 to 3.11 ‰ from January to March 2010, while the organic carbon and nitrogen fluxes increased. The causes of enhanced sedimentation from the eddy in February/March 2010 remain elusive, but nutrient depletion and/or an increased availability of dust as a ballast mineral for organic-rich aggregates might have contributed. Rapid remineralisation of sinking organic-rich particles could have contributed to oxygen depletion at shallow depth. Although the eddy formed in the West African coastal area in summer 2009, no indications of coastal flux signatures (e.g. from diatoms) were found in the sediment trap samples, confirming the assumption that the suboxia developed within the eddy en route. However, we could not detect biomarkers indicative of the presence of anammox (anaerobic ammonia oxidation) bacteria or green sulfur bacteria thriving in photic zone suboxia/hypoxia, i.e. ladderane fatty acids and isorenieratene derivatives, respectively. This could indicate that suboxic conditions in the eddy had recently developed and/or the respective bacterial stocks had not yet reached detection thresholds. Another explanation is that the fast-sinking organic-rich particles produced in the surface layer did not interact with bacteria from the suboxic zone below. CarbonPublished by Copernicus Publications on behalf of the European Geosciences Union. 3204 G. Fischer et al.: Bathypelagic particle flux signatures from a suboxic eddy ate fluxes dropped from ∼ 52 to 21.4 mg m−2 d−1 from January to February 2010, respectively, mainly due to reduced contribution of shallow-dwelling planktonic foraminifera and pteropods. The deep-dwelling foraminifera Globorotalia menardii, however, showed a major flux peak in February 2010, most probably due to the suboxia/hypoxia. The low oxygen conditions forced at least some zooplankton to reduce diel vertical migration. Reduced “flux feeding” by zooplankton in the epipelagic could have contributed to the enhanced fluxes of organic materials to the bathypelagic traps during the eddy passage. Further studies are required on eddy-induced particle production and preservation processes and particle focusing.


Introduction
Time-series particle flux studies have been performed in many ocean areas including typical oligotrophic settings in the Atlantic and the Pacific (BATS, HOT, ESTOC; 15 e.g. Karl et al., 1996Karl et al., , 2003Neuer et al., 2002Neuer et al., , 2007Lampitt and Antia, 1997;Honjo et al., 2008) and in Eastern Boundary Upwelling Ecosystems (EBUE) (e.g. Fischer et al., 2010;Romero et al., 2002). In general, seasonality is low in areas with low primary production while it increases towards coastal and open ocean high production (equatorial, polar) settings (Berger and Wefer, 1990;Romero and Armand, 2010). 20 Mass fluxes at the French oligotrophic EUMELI site (ca. 21 • N, 31 • W) located NW of the Cape Verde Ocean Observatory (CVOO) study site were rather low (mostly below 60 mg m −2 d −1 ) with a low to moderate seasonality (Bory et al., 2001).
In near coastal areas, particle fluxes can vary dramatically due to productivity events triggered by upwelling and submesoscale frontal processes such as filaments (Fis-dies, three types may be distinguished : cyclonic, anticyclonic, and anticyclonic modewater eddies (ACME). In particular, ACMEs have been reported in the past to support high productivity and chlorophyll standing stock (McGuillicuddy et al., 2007), primarily related to a very shallow mixed layer base in the eddy and the efficiency in vertical transport of nutrients into the euphotic zone. A comprehensive 15 overview to mesoscale eddies including ACMEs and their physical and biogeochemical linkages is given by Benitez-Nelson and McGullicuddy (2008). Multi-year oxygen time series data from CVOO show frequent sudden drops in oxygen concentration associated with the passage of ACMEs (Karstensen et al., 2015a). One particularly strong event lasted the entire February 2010 with lowest oxygen concentrations of only 20 1-2 µmol kg −1 at about 40 m depth (Karstensen et al., 2015a). Using satellite data, the propagation path of this particular ACME has been reconstructed and found to have formed in summer 2009, at about 18 • N at the West African coast (Fig. 1).
Here we describe particle flux signatures of the passage of this ACME crossing CVOO in February 2010. We will make use of monthly catches (29 day intervals) from 25 bathypelagic sediment traps installed at 1290 and 3439 m for the period from December 2009 to March 2011 (Table 1). The total length of the sediment trap data time series of about 16 months allows us to compare the winter 2009-2010 with an ACME passage to the winter 2010-2011 without an ACME passage in the vicinity of the moor-spring and summer at the African coast, in the area between Cape Blanc and Cape Vert, Senegal (around 15 and 20 • N), and propagate westward with about 5 km day −1 (Schütte et al., 2015a). Some of the eddies, in particular the ACMEs, develop low dissolved oxygen (DO) concentrations at very shallow depth (Karstensen et al., 2015a). It has been proposed that the low DO concentrations are created within the eddies through intense respiration, associated with high particle fluxes, and sluggish ventilation of the eddy core. During CVOO-3, one particular high productive/low oxygen ACME passed the CVOO site over a period of about one month, in February 2010 (Figs. 1,3). The ocean area off West Africa receives the highest supply of dust of the world (Schütz et al., 1981;Goudie and Middleton, 2001;Kaufman et al., 2005;Schepan-15 ski et al., 2009). Dust is not only relevant for the climate system (e.g. Ansmann et al. 2011;Moulin et al., 1997) and the addition of nitrate, phosphate and iron to the surface ocean (e.g. Jickells et al., 1998), but also for the ballasting of organic-rich particles Ploug et al., 2008;Fischer and Karakas, 2009) formed in the surface ocean ("ballast theory", Armstrong et al., 2002;Ittekkot, 1993). Mineral 20 dust has been shown to contribute about 1/3 and 1/2 of the total deep ocean mass flux off Cape Blanc and south of the Cape Verdes (CV-1-2 trap, ca. 11 • 30 N / 21 • W; Ratmeyer et al., 1999), respectively. Typically, mineral dust correlates with the satellitebased annual aerosol optical index (Fischer et al., 2010) and high dust fluxes have been found at the oligotrophic EUMELI site at around 21 • N/31 • W far north of CVOO 25 (Bory et al., 2001). We obtained a mean annual dust flux of 14 g m −2 yr −1 for the eastern North Atlantic off West Africa (Fischer et al., 2009a). This is in accordance with other studies including modelling approaches (e.g. Kaufman et al., 2005;Jickells et al., 2005). Grain size studies from Ratmeyer et al. (1999) showed dust particles in the Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | or 2% of measured speed (whatever is larger). Speed data < 1.1 cm s −1 has been set to the threshold of 1.1 cm s −1 . Compass accuracy is ±7.5 • for speed < 5 cm s −1 and 5 • above that threshold.

Sediment traps and bulk particle fluxes
Particle fluxes were aquired using two classical cone-shaped and large-aperture sedi-5 ment traps (0.5 m 2 ; Kiel type, Kremling et al., 1996) which were moored at the CVOO site (Karstensen et al., 2015a), The trap depths were in 1290 and 3439 m, respectively. We collected sinking material with bathypelagic traps to circumvent flux biases such as undersampling due to strong ocean currents and/or zooplankton activities (e.g. Buesseler et al., 2007, for a review). It is known that trapping efficiency is best at water 10 depths > 1200 m water depths (see summary in Boyd and Trull, 2007;Berelson, 2002;Yu et al., 2001). We used samples collected on roughly monthly intervals (each 29 days) during the period between December 2009 and March 2011. Detailed sampling periods are given in Table 1, together with the bulk mass fluxes and relative composition of the particles. The traps were equipped with 20 cups, which were poisoned 15 with HgCl 2 before and after deployment by addition of 1 mL of a saturated HgCl 2 solution in distilled water at 20 • C per 100 mL. Pure NaCl was used to increase the density in the cups prior to the deployments (final salinity was 40 ‰). Large swimmers were removed manually, and other swimmers were removed by filtering carefully through a 1 mm sieve. Thus, all fluxes refer to the size fraction of < 1 mm. Flux of the size frac-20 tion of particles > 1 mm was negligible. Samples were wet-split in the home laboratory using a rotating McLANE wet splitter and subsequently freeze-dried. For a detailed description of the methods see Fischer and Wefer (1991). Sediment trap samples were analyzed using freeze-dried homogenized material of 1/5 wet splits. It was weighed for total mass and analysed for organic carbon, total nitrogen, carbonate and biogenic silica. Particulate organic carbon, total nitrogen and calcium carbonate were measured by combustion with a Vario EL III Elemental Analyzer in a sequential 1 M NaOH-leaching method according to Müller and Schneider (1993). The precision of the overall method based on replicate analyses is between ±0.2 and ±0.4 %. Lithogenic fluxes were calculated from total mass flux by subtracting the flux of carbonate, biogenic opal and two times the flux of TOC to approximate organic matter.
Deep ocean sediment traps collect material from a rather large catchment area, 10 typically around 100 km in diameter or wider, depending on particle settling rates and ocean currents (Siegel and Deuser, 1997). Making use of current meter data records from the upper water column (600 and 1300 m), the progressive vector diagrams (PVD) (Fig. 2) show that the collected material before the eddy passage was under the impact of a current from the NE, while after the eddy passage the material was transported 15 more from the southwest. In general, the currents are about twice as strong in 600 m compared to the 1300 m depth and remained mostly below 10 cm s −1 .

Siliceous phytoplankton studies
For this study, 1/125 splits of the original samples were used. Samples were rinsed with distilled water and prepared for siliceous plankton studies following the method 20 proposed by Schrader and Gersonde (1978). Qualitative and quantitative analyses were done at x1000 magnifications using a Zeiss ® Axioscop with phase-contrast illu- the analytical error of the concentration estimates is ≤ 15 % (Schrader and Gersonde, 1978). The resulting counts yielded abundance of individual diatom taxa as well as fluxes of diatom valves per m −2 d −1 calculated according to Sancetta and Calvert (1988), as follows: where

Coccolithophores studies
For coccolith counts, wet split aliquots of each sample (1/25 of the < 1 mm fraction) were further split by means of a rotary sample divider (Fritsch, Laborette 27) using buffered tap water as the split medium. Studied splits ranged between 1/250 and 1/2500, which were filtered onto polycarbonate membrane filters (Schleicher and 15 Schuell TM 47 mm diameter, 0.45 µm pore size). The filters were dried at 40 • C at least for 12 h before a randomly chosen small section of the filter was cut out and fixed on an aluminium stub, sputtered with gold/palladium. The coccolith analysis was carried out using a scanning electron microscope (Zeiss DSM 940A) at 10 kV accelerating voltage. In general more than 500 coccoliths were counted on measured transects at 20 a magnification of 3000×.

Calcareous zooplankton studies
The mass flux of carbonate is mainly constituted of planktonic foraminifera, pteropods and nanofossils/coccolithophores. To determine the proportion of calcareous zooplankton, a 1/5 split of the < 1 mm fraction was used to pick planktonic foraminifera and Introduction tion/oxidation column, reduction column, and water trap work in series to insure that only N 2 and CO 2 enter the GC column and exit the EA. All of the data are expressed in the conventional delta (δ) notation, where the isotopic ratio of 15 N/ 14 N is expressed relative to air which is defined as zero. The N 2 reference gas we use is research grade and has been calibrated to air using IAEA-N1 and IAEA-N2. The internal standard used 20 was pepton with a δ 15 N value of 5.73 ± 0.07 % (1σ).

Biomarker studies
For biomarker analyses, about 70-200 mg of freeze-dried and homogenized samples were extracted three times with dichloromethane (DCM) : methanol (MeOH) 9 : 1 (v/v) in an ultrasonic bath for 10 min. Internal quantification standards (squalane, 25 500 ng/nonadecanone, 499.5 ng/C 46 -GDGT, 500 ng/erucic acid, 500.5 ng) were added prior the first extraction step. The solvent mixture was decanted and combined after each extraction step and following centrifugation. This total lipid extract (TLE) was dried under a gentle stream of nitrogen and saponified for 2 h at 80 • C using 1 mL of a 0.1 M KOH-solution in methanol : water (9 : 1). Following saponification, neutral lipids (NL) were extracted with 4 × 0.5 mL n-hexane. After acidification with HCl to pH 5 < 2, fatty acids were recovered with 4 × 0.5 mL dichloromethane and esterified with methanolic HCl (12 h, 80 • C). Silica-gel chromatography was used to further separate NL. n-Hexane was used to elute a hydrocarbon fraction, n-hexane : DCM (2 : 1) for aromatic hydrocarbons, DCM : n-hexane (2 : 1) for ketones and DCM : MeOH (1 : 1) for polar compounds. 10 Alkenones were analyzed using a 7890A gas chromatograph (Agilent Technologies) equipped with an cold on-column (COC) injection system, a DB-5MS fused silica capillary column (60 m, ID 250, 0.25 µm film) and a flame ionisation detector (FID). Helium was used as carrier gas (constant flow, 1.5 mL min −1 ) and the GC was heated using the following temperature program: 60 • C for 1 min, 20 • C min −1 to 150 • C, 6 • C min −1 to 15 320 • C and a final hold time of 35 min. Alkenone fractions were dissolved in n-hexane and the injection volume was 1 µL. Concentrations were calculated based on integrated peak areas and using the response factor of the internal standard (nonadecanone). The alkenone unsaturation index U k 37 was calculated as defined by Prahl and Wakeham (1987): and converted to SSTs using the calibration of Conte et al. (2006). The aromatic as well as the fatty acid methyl ester (FAME) fractions were analyzed by gas chromatography/mass spectrometry for the presence of isorenieratene and its

Mass fluxes
Mass fluxes increased in winter-spring 2009-2010 in both trap depths during the passage of the ACME at CVOO-3 but were rather low in winter-spring 2010-2011 ( Fig. 3; Table 1). Fluxes were well correlated between both traps (r 2 = 0.6, N = 20), suggest- . We consider this as the "normal conditions". The flux pattern of biogenic silica (BSi) showed a more discrete peak than total mass with maxima in February-March 2010 (Fig. 4a). BSi fluxes were highest in March for both traps and not in February when the ACME passed the study site. The high BSi fluxes arrived simultaneously at both trap depths without a time/cup lag between 15 sampling depths. BSi fluxes were more than 3-fold higher in the deeper trap during February-March 2010 (Fig. 4a). Very low BSi fluxes were measured in winter-spring 2011 when no larger eddy passed and they were slightly higher in the upper trap level which is the common pattern. On an annual basis, the contribution of BSi to total flux mass was 2.8 % (upper) and 5.75 % (lower trap), respectively. However, during the 20 ACME passage, the contribution increased significantly to 4.5-7.8 % (upper) and 8.3-12.3 % (lower trap) ( Table 1). The opal fraction was mainly composed of marine diatoms. Organic carbon fluxes revealed a slightly different pattern from BSi with one distinct flux peak in February 2010 (Fig. 5a). Organic carbon fluxes in the deep trap were almost twice as high as those collected in the upper trap during February 2010. 25 In contrast, during the "normal conditions" in winter-spring 2011, organic carbon fluxes showed only minor differences between the upper and lower traps. Introduction Lithogenic mass fluxes were more than twice higher in the deep trap during the period influenced by the ACME passage (Fig. 6) and followed organic carbon flux with a distinct peak in February 2010. In particular the deeper trap samples provided an almost perfect correlation between mineral dust and organic carbon fluxes (r 2 = 0.97, N = 17). This correlation was less pronounced for the upper trap samples (r 2 = 0.63, 5 N = 18). As there is no river input in the study area, we assume that all non-biogenic (= lithogenic) material was supplied via atmospheric transport. Total carbonate mass fluxes showed less seasonality than BSi and organic carbon with broad maxima in winter-spring 2009-2010, largely following total mass (Figs. 3-5 and 7). However, carbonate fluxes showed a decrease in February 2010 during the 10 passage of the ACME, in particular in the deep trap. Fluxes of the major carbonate producers revealed a decrease in pteropod fluxes at both depths during February-March 2010. Planktonic foraminifera, however, showed a clear flux peak in the deep trap during February 2010 and a rather broad increase in the entire winter-spring 2009/10 at the upper trap (Fig. 7b). Total carbonate mass flux in winter-spring 2011 15 during "normal, non-eddy conditions" was much lower than in 2010 and decreased between the upper and lower trap, which is typical for years without eddy passage.

C/N-and δ 15 N ratios
The molar C : N ratio of the organic material in both traps is rather high for deep ocean material compared to previous findings (Fischer et al., 2003(Fischer et al., , 2010. In Febru-20 ary 2010, C : N ratios were unusually high with values around 18 and 25 in the upper and lower trap, respectively (Fig. 5b). The δ 15 N ratios of the lower trap samples varied between 6.99 and 3.11 ‰ (Fig. 5c). The lowest value (3.11 ‰) was measured following the passage of the ACME in February 2010 while the highest value with almost 7 ‰ was recorded in December 2010. Distinct decreases were found from January to Introduction

Diatom fluxes
Biogenic silica flux showed a similar seasonal pattern as diatoms: a major peak occurred in the transition from late winter into early spring. The total diatom flux in the 5 upper trap ranged from 2.3 × 10 3 to 1.7 × 10 5 valves m −2 d −1 in the upper trap ( Fig. 4b; erally twice to three times higher than in the upper trap. In total, 56 coccolithophore species were identified. The coccolithophores were generally dominated by lower photic zone (LPZ) species, such as Florisphaera profunda and Gladiolithus flabellatus, together with more omnipresent species such as Emiliania huxleyi and Gephyrocapsa spp Florisphaera profunda constituted between 21.7 and 49.2% of the total assem-5 blage and cosmopolitan E. huxleyi ranged between 13.4 and 29.4 %. Coccolith fluxes as well as %-abundances of F. profunda slightly decreased in January-March 2010, although this species shows a distinct flux peak in February (Fig. 8a). In contrast, fluxes of E. huxleyi as well as their relative proportion clearly increased during the interval February-March 2010 (Fig. 8a). Other taxa that considerably contributed to the assemblage are Gephyrocapsa ericsonii (2.3-16.7%), G. oceanica (0.9-6.7%), G. muellerae (0.3-14.0%) and Umbilicosphaera sibogae (1.1-6.7%), which all show a pattern generally similar to that of E. huxleyi. In contrast, deep-dwelling G. flabellatus (1.3-7.3%) and upper zone species Umbellosphaera tenuis (1.3-5.3%) tend to show less prominent fluxes in February 2010 during ACME passage. Other, more oligotrophic species 15 (U. irregularis, R. clavigera) display a similar pattern.

Flux of planktonic foraminifera
Planktonic foraminifera showed a clear flux peak in February 2010 in the deep trap (not shown) and a rather broad increase over the entire winter-spring season in 2010 at the upper trap level ( Fig. 7b; Table 2). The surface dwellers and warm water species 20 Globigerinoides ruber white and pink and Globigerinoides sacculifer were the three dominant species to the total foraminifer flux in both the upper and the deeper trap throughout ( Fig. 8b and c). In February 2010 during the passage of the ACME, however, all three species exhibit a decrease in occurrence. During this interval, they were replaced by the subsurface dweller Globorotalia menardii, dominating the foraminiferal 25 flux at both trap levels ( Fig. 8d, only upper trap shown). The deep dwellers were generally rare at the CVOO-3 site, either they were missing almost completely (Globorotalia truncatulinoides), or they were present in low numbers. instance, showed a flux pattern with a maximum in April-May in both trap levels, following the ACME passage in February 2010.

Lipid biomarkers
A reduced sample set from the upper trap, covering the sample period from December 2009 to July 2010 (samples #1-8), was used for investigation of the organic 5 biomarker composition and the characterization of the ACME passage. Alkenonederived U k 37 values, a biomarker based proxy for SSTs, varied from 0.82 to 0.98 with the minimum value occurring in March, following the ACME passage ( Table 3). Translation of the index into absolute temperatures by using the Conte et al. (2006) global calibration for surface particulate matter resulted in SSTs from 23.6 to 28.0 • C (Fig. 9a). the "Giant Cape Blanc filament" that is characterized by high chlorophyll streaming offshore (Van Camp et al., 1991;Helmke et al., 2005). We argue that the unusual high BSi flux during the eddy passage was due to BSi and diatom production within the surface waters of the ACME. The diatom flux pattern revealed a distinct increase in February 2010 with a major peak later in early spring (Fig. 4b). The thermocline shoaled 5 from about 50-60 m before the passage of the eddy to about 20 m in February 2010 during the eddy passage at CVOO (Karstensen et al., 2015a). Elevated chlorophyll and primary production within the eddy is seen from selected chlorophyll maps ( Fig. 1) and has been discussed in the context of upward nutrient fluxes, particularly associated with ACMEs (e.g. Karstensen et al., 2015a;Benitez-Nelson and McGullicuddy, 2008). Considering the distinct BSi flux signal, this may indicate that the organic carbon is primarily fixed on the western side of the eddy where the most intense bloom is expected (Chelton et al., 2011). Sargasso Sea ACMEs, for instance, contain significant numbers of diatoms, regardless of the age of the eddy (McNeil et al., 1999;Sweeney et al., 2003;Ewart et al., 2008). The upwelling of nutrients within the eddy 15 is driven by submesoscale processes, which are highly variable in space and time. As such, pulsed nutrient/silicate injections from subsurface waters probably combined with a high regeneration of nutrients within the upper layer are likely to occur. The BSi flux data support these findings and further suggest some vertical transport of nutrients (from the silicate-richer water of the shadow zone region east of the CVFZ) into the 20 photic zone of the eddy in the beginning of 2010. In addition, protection of the organic materials in the diatom valves while sinking through the low-oxygen zone of the eddy may have contributed to elevated BSi fluxes in the deep ocean due to reduced BSi dissolution (Ragueneau et al., 2000). The molar C : N ratios of organic matter were unusually high in February 2010 for 25 both trap samples (Fig. 5b). They clearly fall far off the range of deep-ocean sediment trap samples or surface sediments with partly degraded organic marine material (C : N around 8- 10;Fischer et al., 2003Fischer et al., , 2010C : N= 5-10;Tyson, 1995;Wagner and Dupont, 1999 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | material which also contains some terrestrial organic matter with elevated C : N values (global mean = 24; Romankevich, 1984), clearly above the marine signal (e.g. Müller, 1977;Wagner and Dupont, 1999). This terrestrial organic matter is mixed with the debris of major marine primary producers (e.g. diatoms, coccolithophores) whose C : N values are around the Redfield Ratio (Redfield et al., 1963;Martiny et al., 2013). 5 The exceptionally high ratios in February 2010 (C : N= 18 (upper) and 25 (lower trap) (Fig. 5b), however, cannot be explained by mixing processes of marine and terrestrial organic materials alone, because this would suggest a preferential contribution of terrestrial organic matter. Nitrogen (nitrate) limitation in the generally oligotrophic setting north of the Cape Verde Islands must be taken into account to explain the high 10 C : N ratios of organic matter (e.g. Laws and Bannister, 1980;Martiny et al., 2013;Löscher et al., 2015a). We have to consider that production within the surface eddy was sometimes characterized by low growth rates of the primary producers (both diatoms and coccolithophores), combined with a reduced nutrient availability, e.g. in January/February 2010. This process could explain the extraordinary high C : N ratios 15 (Laws and Bannister, 1980;Martiny et al., 2013) in February 2010. Nitrogen limitation is also known to increase the C : N ratios of the alkenone producers (e.g. Löbl et al., 2010), and might result in an increase in the production and storage of alkenones (e.g., Eltgroth et al., 2005;Prahl et al., 2003). Alkenone temperature records from the Subtropical Front at the Chatham Rise, SW Pacific Ocean 20 (Sikes et al., 2005) showed that biases occurred during times of highest lipid fluxes and low nutrient conditions in the surface mixed-layer. When plotting the C : N ratios versus the alkenone fluxes of the upper trap samples, we indeed obtain a relationship (Fig. 9, r 2 = 0.77, n = 8) which could point to nutrient limitation during or shortly before the ACME passage. Our temperature record derived from the unsaturation index of the 25 alkenones revealed a stepwise decrease in SST by about 2 • C (Fig. 9a)  The scenario of nutrient limitation within the surface water of the eddy at some time is supported by elevated fluxes of BSi and organic carbon in February-March 2010 and agrees with a decrease of chlorophyll between November/December 2009 and again between January/February 2010 as shown in satellite ocean color maps (Fig. 1). However, we have to consider that our time resolution of the sampling cups is rather 5 low (29 days) and our traps thus may not capture the highly dynamic biogeochemistry within the eddy and the resulting export fluxes which may fluctuate within days and weeks. Secondly, variable settling rates of different larger particles produced in the surface and subsurface waters of the eddy make it difficult to estimate the variable time lags between the arrival of the flux signature in the bathypelagic traps and the  nitrogen species utilized by the primary producers, in most cases, nitrate (Ryabenko et al., 2012;Altabet and Deuser, 1985). More specifically, the isotope ratio reflects the degree of nitrate utilization (Mariotti et al., 1982) Ryabenko et al., 2012), both close to global averages (Liu and Kaplan, 1989). Phytoplankton preferentially takes up the lighter isotope during photosynthesis (e.g. Altabet et al., 1991), leaving the remaining nitrate pool enriched in 15 N.
In general, δ 15 N is high in temperate oceans after nitrate is depleted due to phytoplankton growth and low in more stable, oligotrophic seas (Saino and Hattori, 1987). 5 Our δ 15 N record in winter-spring 2009-2010 may reflect episodic nutrient injection into the euphotic zone of the ACME ), leading to increased particle formation and fluxes documented in February-March 2010 in the lower trap (Fig. 5c). This nutrient injection from below can be deduced from a stepwise cooling starting in January 2010 and ending in March, as seen in the record (Fig. 9a) The higher N-fluxes 10 were associated with a lowering of δ 15 N as expected from other studies. The relatively high δ 15 N value of 5.21 ‰ in January 2010 ( Fig. 5c) shortly before the ACME passage could document some depletion in nitrate which may also explain the unusually high molar C : N ratios (18-25) one month later (Fig. 5b). Under low oxygen conditions, denitrification by nitrate-reducing bacteria can affect 15 the isotopic signature of the nitrate pool, leading to a significant enrichment of 15 N in the residual nitrate pool relative to a deep water value of around 6 ‰ (Liu and Kaplan, 1989;Libes and Deuser, 1988 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | a low oxygen ACME in the Cape Verde region studied in 2014. In the Arabian Sea, for instance, source waters had values around 6-10% due to denitrification (e.g. Schäfer and Ittekkot, 1983). The vertical distribution of many coccolithophore species is often controlled by upper photic-zone temperature and water stratification (e.g. Jordan and Chamberlain, 5 1997;Hagino et al., 2000). In particular, E. huxleyi is known to preferentially thrive in more turbulent and nutrient-enriched waters as found in upwelling areas or coastal regions (e.g., Haidar and Thierstein, 2001;Hagino and Okada, 2006;Boeckel and Baumann, 2008). Thus, the increasing fluxes during February-March 2010 (Fig. 8a) correspond well to nutrient-enriched conditions during this time interval or somewhat 10 before. Alkenones, synthesized by planktonic algae such as coccolithophorids show a peak in flux during this time interval (Fig. 9). These observations correspond to nutrient measurements conducted in the low oxygen ACME in 2014 (Fiedler et al., 2015). The coccolithophore flora in the upper photic zone (UPZ) down to about 40-60 m is often composed of Umbellosphaera tenuis, U. irregularis, and Discosphaera tubifera, 15 adapted to warm temperatures and low nutrient levels (e.g., Honjo and Okada, 1974;Hagino et al., 2000;Malinverno et al., 2003;Boeckel and Baumann, 2008). The same pattern is displayed by Rhabdosphaera clavigera, R. stylifer and Syracosphaera pulchra, all of which are non-placoliths known to prefer stable stratified waters (Hagino et al., 2000). All these latter three species show a rather similar pattern with slightly 20 increased fluxes in February-March 2010 when the ACME passed. The species F. profunda, G. flabellatus are well established as species belonging to the lower-photic zone community (e.g., Honjo and Okada, 1974;Takahashi and Okada, 2000;Andruleit et al., 2003). In particular, F. profunda is known to occur exclusively in the deep photic zone (ca. 40-200 m), typically occurring at maximum abundances below the deep chloro-25 phyll maximum in relatively high abundances (Haidar and Thierstein, 2001). During the ACME passage, we observed an increase in coccolith fluxes in February-March 2010 (Fig. 8a)  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | these species thrive. However, a clear impact of the low oxygen conditons in the ACME on the photosynthetic coccolithophore community cannot be observed.

Origin of hypoxia/suboxia and organic matter preservation within the eddy
Neither the diatom nor the coccolithophore communities do show any significant coastal influence in the collected materials. Given the ocean currents to the south-5 west and the proximity to the NW African coast, it is not unreasonable to suspect that diatom blooms above the CVOO mooring may have been due to a seed population from coastal waters. The diatom assemblage, however, shows no signature of coastal upwelling and benthic diatoms, as indicators of entrained coastal waters. Low relative contributions of coastal upwelling-related resting spores of Chaetoceros (Romero et al., 10 2002) and a few benthic species, which thrive in near-shore waters above 50 m water depth (Round et al., 1990;Romero et al., 2015), suggests weak transport of plankton communities from near-shore/coastal waters into the pelagial north of the Cape Verde Islands. This east-to-west seaward transport did not carry substantial amounts of microorganisms nor did it vastly contribute to the pool of nutrients in waters over- 15 lying the CVOO site. Further evidence is provided by the coastal : pelagic ratio of the diatom assemblage of the upper trap (Fig. 4c). Compared to the values recorded at 200 nm off Cape Blanc (Mauritania, CB trap site), the coastal : pelagic ratio of 20 to 25 at CVOO-3 is lower than values recorded at the CB site. At all times, the dominance of oceanic species at the CVOO-3 site reveals in situ diatom production with minor 20 transport from the coastal realm. This indicates that the eddy at the time of its passage at CVOO-3 had significantly altered since its origin at the African coast at around 18 • N in summer 2009 (Karstensen et al., 2015). At the origin of the ACME in summer Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | data, the ACME approaching the CVOO site showed a decrease in chlorophyll between November/December 2009 and January and again between January and February 2010 (Fig. 1). In February 2010, only an unclear and ring-like structure of slightly elevated but still rather low chlorophyll of approximately the size of the ACME remained within the oligotrophic surrounding area (Fig. 1a). However, a general high cloud cover Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | detection of these compounds requires the respective bacterial stocks to be present in concentrations above a certain detection threshold and/or an effective export mechanism for them leading to incorporation into sinking particles. Although one might expect these compounds to be present during the ACME passage when low oxygen conditions prevailed in the subsurface waters (Löscher et al., 2015a), likely the populations of the organisms have not reached significant levels, because suboxic conditions just recently developed within the eddy. Another possible explanation is that the above mentioned bacteria were present in the suboxic zone of the eddy, but did not interact and were not attached to the fast sinking organic-rich particles which originated in the surface layer and later constituted the mass flux.
No signs of dissolution in sinking calcareous particles i.e. coccolithophores or foraminifera are seen, which might have occurred due to reduced pH within the suboxic/hypoxic parts of the eddy. The low oxygen ACME waters surveyed in 2014 had a pH of about 7.6 (Fiedler et al., 2015). No signs of carbonate dissolution could either point to a rapid transport of sinking carbonate particles through the suboxic/hypoxic wa- 15 ter column of the ACME as outlined above or to some protection of carbonate particles by periotrophic membranes of fecal pellets or both.

Mineral dust and particle settling rates
In general, both traps revealed similar flux signals with maxima around February to March 2010  and coinciding with the passage of the ACME with elevated 20 biomass in the surface ocean over the CVOO-3 site (Fig. 1). Considering the synchronicity of peaks in BSi and other components, a fast vertical transport of the surface particle flux signature into the meso-and bathypelagic is expected within the eddy. Given the 29 day sampling interval of the traps, the particle settling rate for the bathypelagic water column should at least 150 m d −1 , applying the methods described in Fis- the EBUEs and much closer to the coast. For the sediment trap mooring sites south of the Cape Verdes (CV 1-2), an even higher mean sinking speed of 416 m d −1 was estimated (Fischer and Karakas 2009). The authors argued that high organic carbon fluxes in the CC compared to other EBUEs are at least partly due to high particle settling rates which result in low carbon respiration rates , most 5 probably favored by a high ballast content such as mineral dust. Deep trap organic carbon fluxes plotted versus the fluxes of mineral dust provided an exceptionally good empirical relationship (r 2 = 0.97; N = 17, Fig. 7) which we never observed before off NW Africa (e.g. Fischer et al., 2010). This relationship, however, does not explain the complex processes involved in the formation of larger and fast 10 sinking settling particles in the surface and subsurface waters and the interaction of biogenic with non-biogenic particles. Lab experiments with roller tanks and ballast minerals, however, clearly indicate the importance of mineral ballast for increasing sinking rates and lower carbon degradation within marine snow aggregates (Ploug et al., 2008;. Additional evidence is provided by observations gained dur- 15 ing a field campaign in winter 2012 off Cape Blanc (eutrophic site CBi): higher organic carbon fluxes at 100 and 400 m water depths using drifting traps were recorded, matching faster particle settling rates after a 1-2 days, low altitude dust storm event (Iversen et al., 2015). Besides the question of the development of suboxia/hypoxia within the eddy dis-20 cussed above, the causes of enhanced sedimentation of biogenic detritus in February-March are unclear. From our field studies in the Cape Blanc area (e.g. Fischer and Karakas, 2009) and lab studies with in situ chlorophyll and mineral dust (e.g. Iversen, unpubl.; van der Jagt, unpubl.), we speculate that Saharan mineral dust which preferentially settles in winter in the Cape Blanc and Cape Verde ocean area (e.g. Gama et al., 25 2015) could have contributed or even initiated particle settling via ballasting of organicrich aggregates (Ploug et al., 2008;Iversen and Robert, 2015) produced within the chlorophyll enriched eddy. Some effect on particle produc- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | tion and fluxes by fertilization due to the input of macro-nutrients by dust (e.g. nitrogen; Fomba et al., 2014) via dust cannot be excluded. Fluxes of organic carbon and mineral dust co-varied (Fig. 6) which means that both components settled in close association into the bathypelagic. In the high dust region south of the Cape Verdes, Ratmeyer et al. (1999)  By comparing the fluxes in winter-early spring 2009-2010 under the influence of the ACME and the suboxia/hypoxia with winter-early spring 2011, when no larger eddy passed the CVOO site, the contribution of the ACME to annual mass flux can be esti-20 mated. This estimation does not consider interannual variability of absolute mass fluxes nor changes in seasonality/timing of maxima from year-to-year and therefore has to be regarded as a first estimation. When comparing the organic carbon fluxes of the upper trap for the first four months of both years, we roughly obtain a three-fold increase in carbon flux when the eddy passed over the CVOO site compared to an eddy-free year 25 (Fig. 5a). These estimates match rather well with data determined in the low oxygen ACME in 2014 (Löscher et al., 2015b). These authors obtained chlorophyll concentrations and carbon uptake rates within the eddy of up to three times as high as in the surrounding waters.

Differences of fluxes in the water column
There is a significant increase in mass fluxes with depth from December 2009 to May 2010, a common feature of many ocean areas, in particular at near-continental margins sites (e.g. Neuer et al., 2002;Honjo et al., 2008;Fischer et al., 2009b). At the open ocean site CVOO-3, however, the organic carbon fluxes were more than twice 5 as high in the deep trap compared to the upper trap and correlated with r 2 = 0.70. BSi flux was more than three-fold higher at greater depth (correlated coefficient r 2 = 0.91) during the eddy passage. The flux of coccoliths increased with depth by three-fold was well. For organic carbon, an overall decrease in flux with depth has to be expected (when excluding lateral advection), following an exponential equation in classical oceanic settings with sufficient oxygen in the water column (see summary in Boyd and Trull, 2007). As pointed out by Siegel and Deuser (1997), deeper traps have a larger catchment area than shallower ones and may sample additional material from lateral sources like a statistical funnel. Assuming a rather conservative settling rate of 200 m d −1 for particles with high ballast content (see Fischer and Karakas, 2009), 15 we obtain catchment areas with a length scale of around 300 km for the upper trap and 400 km for the lower one when using particle trajectories from the Sargasso Sea (Siegel and Deuser, 1997). The mean currents at the CVOO site were sluggish with monthly mean velocities between 2 to 6 cm s −1 (equivalent to 1.5 to 5.1 km day −1 ) for the upper trap and up to 20 2 cm s −1 (1.7 km day −1 ) for the lower trap, considering the velocity data from CVOO-2 (March 2008 to October 2009, not shown here) and CVOO-3. These currents will primarily add a displacement of the sinking particles that results in a difference between the particle source areas of the two sediment traps (Siegel and Deuser, 1997). The gravitational sinking speed has been found to vary over a wide spectrum but it is likely 25 that in our study area values of several hundred of meters per day are reached (e.g. Fischer and Karakas, 2009). In case of a settling rate of 100 m day −1 and sluggish lateral flux (2 km day −1 ), the setting of a particle through a 3500 m water column will take BGD 12,2015 Bathypelagic particle flux signatures from a suboxic eddy about 35 days and the material is displaced by less than 100 km. For interpreting the flux data, the origin of laterally derived particles and hence the prevailing flow direction is of particular importance. High chlorophyll is found in the coastal upwelling off West Africa approximately 300 to 700 km away from CVOO (Fig. 1). Comparing the progressive vector diagrams (PVD) from three depth at CVOO-3 for the period December 2009   5 to May 2010 it is evident that the upper trap (Fig. 2b) is primarily under the impact of meridional transport from the south and also reflected in the thermocline transport (Fig. 2a), while the lower trap is more under zonal transport from the east (Fig. 2c). However, because of the rotor failure it is unclear how far the catchment in the lower trap is extending. As expected, during the eddy passage all three RCMs show varying 10 currents dominated by the local circulation associated with the eddy. The nearest and most probable additional particle source area for the deep CVOO-3 trap to the east and northeast is the approaching ACME at a time when chlorophyll, production and export to deeper waters was higher than in February 2010, e.g. in November/December 2009 and in early January 2010 (Fig. 1). This additional material could be laterally trans- 15 ported to the deep trap by the prevailing current system (Fig. 2c), thus enhancing the deep bathypelagic fluxes by a factor of two to three in winter-spring 2010.

Zooplankton within the eddy and organic carbon degradation
Acoustic backscatter data from Karstensen et al. (2015a) suggest that at least some zooplankters stopped their diel vertical migration behavior when the suboxic part of 20 the eddy approached the CVOO site around February 2010. This seems to be typical for open ocean OMZ (Ayon et al., 2008). Mobile zooplankton such as certain copepods escape from oceanic dead zones (e.g. the ACME 2010), while certain less mobile protozoa such as planktonic foraminifera, may be encountered by the suboxia, die and settle down. In a low oxygen eddy observed in spring 2014 at CVOO, acoustic backscatter data and multinet sampling indicated a compression of zooplankters in the surface waters with a high abundance of calanoid copepods and euphausiids (Hauss , 2015). This suggests a high grazing pressure on these organisms in the surface layer during eddy passage. The flux patterns of planktonic foraminifera revealed a clear peak flux in February 2010 in the lower trap, matching the passage of the suboxic eddy. The subsurface (50-100 m water depth) dweller Globorotalia menardii largely responsible for this flux 5 peak in the upper trap in February 2010 (Fig. 8d), is a tropical to subtropical, nonspinose species with changing depths habitats (Hemleben et al., 1989). We assume that the oxygen within the ACME became too low in early 2010 and the more or less immobile G. menardii died , resulting in sedimentation and elevated fluxes in both trap levels. Foraminifera are generally assumed to settle with high rates of several hundreds 10 to a few thousand meters per day (Kucera, 2007), thus, a clear flux signal without time delay is expected in the two bathypelagic traps. The near-surface dwellers Globigerinoides ruber pink and white and Globgerinoides sacculifer, on the other side, showed a clear decline in flux in February 2010 in both trap samples (Fig. 8b and c), contributing to reduced total carbonate fluxes (Fig. 7). This pattern might be due to the shoaling 15 of the thermocline from 50-60 m to about 20 m (Karstensen et al., 2015a) and a decrease in SST (Fig. 9) during ACME passage. Foraminifera trapped in the uppermost water layer during ACME passage might have suffered from a high grazing pressure because of the low oxygen eddy core below. The foraminiferal peaks in the deeper trap in April-June 2010 were mostly due to high fluxes of G. sacculifer that followed 20 the eddy passage. The increase of foraminiferal flux at both depths in April-June may represent a return to regular (non-eddy) conditions and a recovery/deepening of the thermocline. The actively migrating pteropods (Chang and Yen, 2012) show some decrease in the fluxes in February-March 2010 at both bathypelagic depths (Fig. 7b). This can be explained by the escape from the dead zone area of the approaching eddy 25 and some sedimentation elsewhere.
Missing diel migration of a number of zooplankton groups due to the passage of the suboxic eddy (Karstensen et al., 2015a;Hauss et al., 2015) could have resulted in less organic matter degradation of sinking particles due to reduced "flux feeding" within the suboxic/hypoxic zone (around 40-170 m). This depth range is the most active zone in terms of organic carbon turnover under normal conditions with sufficient oxygen (e.g. Hedges, 1992). "Flux feeding" may account for a large part of organic carbon degradation in the uppermost few hundred meters of the water column and determine the shape of the carbon attenuation curve (Iversen 5 et al., 2010), although quantitative estimates are lacking. Under oxic conditions, overall carbon-specific respiration due to microbial degradation is estimated to be 0.13 d −1 in the uppermost ocean Ploug, 2010, 2013;, independent of particle size and type. It is likely that the severe hypoxia/suboxia reduced both oxic microbial respiration and zooplankton "flux feeding". As a result, the organic 10 carbon flux to greater depths might have increased.

Summary
The impact of a low oxygen eddy (ACME) on particle fluxes at the CVOO mooring site has been investigated from two sediment traps time series covering the period from satellite derived chlorophyll maps reveal surface signatures of high chlorophyll standing stocks within the eddy (Fig. 1). The schematic Fig. 10 summarizes the most important processes and the responses in the bathypelagic ocean to the eddy passage: -BSi, diatoms and organic carbon fluxes increased and seasonality was unusually high in winter-spring 2010 when the ACME passed, compared to 2011 during diatoms showed no signature from coastal environments but were mostly of pelagic origin, suggesting a pronounced alteration of the eddy since its origin 5 at the African coast in summer 2009, molar C : N ratios of organic matter were unusually high (18)(19)(20)(21)(22)(23)(24)(25) in February 2010, although nutrient supply is high within the ACMEs . This may indicate that some production was under nutrient (nitrate) limitation (e.g. Löscher et al., 2015a), probably combined with low growth rates in the beginning we have no indications of any carbonate dissolution due to a reduced pH (∼ 7.6, Fiedler et al., 2015) within the suboxic/hypoxic parts of the ACME through which the particles have to sink, sinking detritus and organic matter degradation might have contributed to the severe suboxia/hypoxia in February 2010 in the subsurface waters. We assume that 5 the severe suboxia began in the beginning of 2010, sedimentation from the eddy might have occurred due to nutrient exhaustion and/or deposition of mineral dust combined with enhanced chlorophyll in December 2009 and January 2010 (Fig. 10), daily migrating zooplankton is reduced in low oxygen eddies (Karstensen et al.,10 2015a; Hauss et al., 2015) which should have resulted in less organic matter degradation due to missing "flux feeding". This could have resulted in less organic carbon flux attenuation and thus, a higher bathypelagic organic carbon flux.

Conclusions and outlook
The passage of the eddy (ACME) with the suboxia in the subsurface waters which was 15 studied at CVOO may serve as a natural experiment or open-ocean "mesocosm" with respect to particle sedimentation. Oceanic oxygen levels in the future oceans might decrease significantly and develop into OMZs due to global warming and increased water column stratification (e.g. Stramma et al., 2008Stramma et al., , 2010Codispoti, 2010;Löscher et al., 2015a). These potential changes might influence the nitrogen cycle and the op-20 eration of the biological pump, e.g. via a better preservation of organic materials due to reduced or non-existing microbial respiration  combined with reduced zooplankton activities (missing "flux feeding") within the developing OMZs. Such processes could enhance marine CO 2 sequestration and operate as a negative feedback on global warming. The studies are further interesting when considering warm periods in the Earth's history (e.g. the Upper Cretaceous) when OMZs were widely expanded and black shales with high organic carbon contents have been deposited (e.g. Takashima et al., 2006;Schönfeld et al., 2015). This study may contribute to the unsolved question of production versus preservation of organic materials when trying to explain the origin of 5 black shales and oil source rocks in the Earth's history (e.g. Calvert, 1987). In addition, sedimentation signals with erratic character such as peaks in large diatoms in pelagic sediments below oligotrophic areas of the world ocean (e.g. Ethmodiscus rex in South Atlantic, Romero and Schmieder, 2006) might be explained by processes occurring within recurring eddies. For instance, the Agulhas current system with its 10 retroflection zone releases continuously large numbers of different types of eddies both into the northwest into the South Atlantic and the Southern Ocean. These eddies may serve as productive oasis within generally oligotrophic ocean deserts and become suboxic/hypoxic at some time, probably depending on the amount and type of organic-rich marine snow particles being remineralized in its subsurface waters (e.g. Löscher et al.,15 2015b; Karstensen et al., 2015a) as well as on specific eddy physics.

BGD
Among others, remaining questions concerning the processes within suboxic/hypoxic eddies are: 1. what are the sedimentation processes associated with these eddies from local and global perspectives? Are such eddies characterized by pulsed sedimentation 20 or a more continuous particle rain? Which processes might trigger pulsed sedimentation events within an eddy?
2. how frequent are these eddies and why do suboxia develop in their subsurface waters?