Organic carbon budget for the eastern boundary of the North Atlantic subtropical gyre: major role of DOC in mesopelagic respiration

Transports of suspended particulate (POCsusp) and dissolved (DOC) organic carbon are inferred from a box-model covering the eastern boundary of the North Atlantic subtropical gyre. Corresponding net respiration rates (R) are obtained from a net organic carbon budget that is based on the transport estimates, and includes both vertical and lateral fluxes. The overall R in the mesopelagic layer (100–1500 m) is 1.6 ± 0.4 mmol C m−2 d−1. DOC accounts for up to 53% of R as a result of drawdown of organic carbon within Eastern North Atlantic Central Water (ENACW) that is entrained into sinking Mediterranean Overflow Water (MOW) that leads to formation of Mediterranean water (MW) at intermediate depths (~900 m). DOC represents 90% of the respired non-sinking organic carbon. When converted into oxygen units, the computed net respiration rate represents less than half the oxygen utilization rates (OUR) reported for the mesopelagic waters of the subtropical North Atlantic. Mesoscale processes in the area, not quantified with our approach, could account in part for the OUR differences observed between our carbon budget and other published studies from the North Atlantic, although seasonal or interannual variability could also be responsible for the difference in the estimates.

In the present study, following the approach used by Alonso-González and coworkers in the southern Canary Current 20 , we estimate the contribution of DOC and POC susp fluxes to net respiration in the eastern boundary of the North Atlantic subtropical gyre (Fig. 1). Our overarching hypothesis is that, contrary to the southern Canary Current where POC susp is an important component of the carbon budget 20,23 , the DOC contribution to total respiration in the study area is larger than that of POC susp . The two areas are oceanographically distinct. On the one hand, the coastline orientation in our region is unfavorable to coastal upwelling, which is absent or limited just to the summer months 24, 25 . On the other hand, our region is characterized by the density-driven exchange of water masses between the Mediterranean Sea and the eastern North Atlantic 26 . As a consequence, Eastern North Atlantic Central Water (ENACW) is entrained into the dense fast-flowing Mediterranean Overflow Water (MOW) that exits through the Strait of Gibraltar [27][28][29] and is subducted along the slope into the Gulf of Cádiz (ca. 33-37°N) [30][31][32][33] to below 1000 m depth, where it spreads into the Atlantic as Mediterranean Water (MW) [27][28][29][34][35][36][37] . This entrainment of ENACW into MOW may promote the downward flux of DOC to intermediate layers in this particular region, as has already been shown for anthropogenic carbon dioxide 38 .
In a previous study 39 , based on a rough estimation of organic carbon fluxes in a box region that included the Strait of Gibraltar, it has been suggested that DOC remineralization could explain about 90% of the inorganic carbon produced in the water column. Our study tests this hypothesis, providing direct measurements of POC susp and DOC fluxes for integration within a box-model analysis. We aim to contribute to the understanding of the organic carbon dynamics in this complex region, which is characterized by the confluence of water masses of different origin, water mass entrainment and subduction, and intermittent upwelling 24, [40][41][42][43] .

Results
Organic carbon distributions. The DOC and POC susp data used in this work were collected during a research cruise in autumn 2009. The cruise consisted of a set of hydrographic stations distributed regularly along three transects (northern, western and southern) that delimit the box presented in Fig. 1

(see Materials and
Methods for further details on sampling and analysis). POC susp and DOC distributions from 0 to 4000 m (Fig. 2) show that surface POC susp concentrations along the northern transect decrease offshore from 4-6 µmol C L −1 in the coastal area to about 2 µmol C L −1 in the open ocean. Below 500 m POC susp concentrations are 1 µmol C L −1 or less, except around a seamount at ~14° W. Waters with relatively high POC susp deepen at station 30. Surface DOC concentrations are higher than 60 µmol C L −1 for the entire transect. A deepening of DOC occurs at stations 30-37. Surface POC susp concentrations along the western transect are higher than 1-2 µmol C L −1 . A deepening of POC susp -rich waters at stations 56-58 coincides with the shoaling of isoneutral surfaces between two mesoscale   44 . POC susp concentrations remain around 1 µmol C L −1 along the rest of the section with little variation. DOC is about 60-70 µmol C L −1 near the surface, and decreases to less than 50 µmol C L −1 with depth. Central waters with high DOC concentrations descend along sloping isoneutrals at stations 53-56. The deepening of DOC north of 35°N coincides with the location of the Azores Current system 44 .
The southern transect shows surface POC susp concentrations of 2-3 µmol C L −1 . A westward decrease is observed below 250 m. Surface DOC concentrations show values above 70-80 µmol C L −1 . A patch of relatively high POC susp below 1000 m depth coincides with cool and fresh waters 44 . DOC shows values above 70-80 µmol C L −1 at the surface, with maximum concentrations above 100 µmol C L −1 at subsurface depths (~200 m) in the eastern stations close to the Lanzarote Passage. Intermediate waters (below the 27.38 isoneutral) show low DOC concentrations, except at stations 64-65 where a front between a cyclonic eddy and the southward Canary Current is periodically observed 44 .

Mass transports.
Mass transports across the northern, western, and southern transects are taken from a previous study by Pérez-Hernandez and co-workers who developed an inverse box model with data from the same cruise 44 (see Material and Methods for further details about the inverse model and its reliability). A summary of the integrated lateral mass transport at each transect is presented in Fig. 3. Note that fluxes into/out of the box are indicated by positive/negative signs, respectively.
A predominant inward (southward) transport is observed across the northern transect (~37°N; blue line), interrupted only by the outward (northward) transport of intermediate waters (isoneutrals 27.38 to 27.92). The mass transport across the western transect (~24°W; red line) is strongly inwards (eastward) in the surface and

Discussion
To evaluate the organic carbon sources within the box we build a mass balance (Fig. 5a) and an associated organic carbon budget (Fig. 5b). For the surface waters (0-100 m), 1.2 ± 0.1 Sv enter the box through the northern, southern and western transects, and −0.78 ± 0.1 Sv leave the box as the Atlantic inflow into the Mediterranean Sea (see Materials and Methods). As a result, the mass balance of inputs minus outputs to the box would be 0.4 ± 0.2 Sv. We propose that this volume is transported from the surface to the central waters to partially balance the loss of ENACW to intermediate waters when entrained into the MOW (see below the discussion about the mass balance of the central waters). The balance of organic carbon, i.e. POC susp and e-DOC, transported across the northern, western and southern transects indicates that the surface waters of the box receive an external input from the Furthermore, the non-respired organic carbon in surface waters may be transported to the dark ocean by turbulent diffusion, passive sinking, and active transport 49,50 . We estimated first the vertical turbulent diffusion at 100 m depth as F = −ε × R/(N 2 × (1 − R)) × (DOC z2 − DOC z1 )/(z 2 − z 1 ) 51 . We considered (1) 53 . In the northeast Atlantic (39-43°N, 17-19°W), the minimum vertical POC fluxes were estimated to be 0.36 ± 0.09 mmol C m − ² d − ¹ from drifting traps at 200 m depth during late summer 54 . Although the use of drifting traps has been frequently criticized for underestimating export fluxes [55][56][57] , comparison with POC fluxes at the Bermuda Time Series Station (BATS) derived from both surface-tethered traps and Thorium-234 56 shows reasonable agreement during low productivity periods. At the ESTOC site, the lowest productivity values have been obtained during autumn (~10 mmol C m − ² d − ¹) 57 ; considering an e-ratio (export/production ratio) of 5-10% 55 , the vertical POC flux during the less productive season would be 0.5-1.0 mmol C m − ² d − ¹. Therefore, an average value of the export rate from drifting traps of 0.36 to 1.0 mmol C m − ² d − ¹ is representative of the sinking flux from surface waters within the box at that time of the year. Aggregation and disaggregation of sinking particles by biotic and abiotic mechanisms could modify the sinking rates and magnitude of POC fluxes with depth across the dark ocean [58][59][60][61][62] , although our box model approach does not allow identification of these processes.
In addition, the estimated averaged vertical flux of POC mediated by migrant zooplankton (active flux) was 0.22 mmol C m − ² d − ¹ south of the island of Gran Canaria 63 , accounting for 25% of the sinking POC flux (passive flux) in the region. More recently, a mean active flux of 0.28 mmol C m − ² d − ¹ (29% of the passive flux) has been estimated north of Gran Canaria 64 ; these estimates agree with those observed at BATS (0.17 mmol C m − ² d − ¹) 65 and the Hawaiian Ocean Time Series (HOT; 0.26 mmol C m − ² d − ¹) 66 . However, the active flux accounted for only 8% and 11-44% of the passive flux at their respective regions 65,66 . If we assume that the active flux represents at best 20% of the passive flux in our region, the passive + migratory POC flux from the surface to the central waters would be around 0.82 mmol C m − ² d − ¹ which, multiplied by the area of our box, yields a total POC flux from the surface to the central waters of 11.4 × 10 8 mol C d − ¹.
The central waters (100-700 m) within the box receive an overall lateral input of POC susp and e-DOC of 10.7 ± 1.9 × 10 8 mol C d − ¹ from the surrounding Atlantic Ocean and 2.4 ± 1.5 × 10 8 mol C d − ¹ from the Mediterranean outflow through the Strait of Gibraltar. In addition, it is well known that MOW is subducted within the study box to more than 1000 m and that the central waters are entrained into the MOW as it sinks 29,38 . In fact, the mass balance of central waters within the box (Fig. 5a), considering the net water flows across the northern, southern, and western transects (2.1 ± 0.1 Sv), the Mediterranean outflow into the Atlantic (0.67 ± 0.1 Sv; see Materials and Methods), and the input of 0.4 ± 0.2 Sv from the surface waters, is 3.1 ± 0.4 Sv. Therefore, in order to maintain conservation of volume in the central waters, 3.1 ± 0.4 Sv should enter the intermediate layer by entrainment of the central waters into the subducting Mediterranean outflow. Our volume calculation is in agreement with earlier estimates (2-3 Sv 67 ; 2.6 Sv 68 ; 1.3 Sv 28 ; 2.3 Sv 38 ). To obtain the organic carbon transported by this volume, we multiply the average organic carbon concentration at the bottom of the central layer by 3.1 ± 0.4 Sv to obtain a downward POC + e-DOC flux of 14.8 ± 3.7 × 10 8 mol C d − ¹, 90% of which is DOC. Similarly, we also calculate the organic carbon transport associated with the 0.4 ± 0.2 Sv of surface waters that enter the central waters yielding 6.6 ± 2.2 × 10 8 mol C d − ¹, 10% as POC susp and 90% as DOC (Fig. 5b).
For the intermediate waters (700-1500 m), aside from the organic carbon that enters this layer by entrainment of ENACW into the subducting MOW (14.8 ± 3.7 × 10 8 mol C d − ¹), the other organic carbon transports associated with the water mass fluxes leave this layer. The net balance of POC susp and e-DOC transport across the northern, western and southern transects is −8.5 ± 0.6 × 10 8 mol C d − ¹. This indicates an export from the box to the adjacent Atlantic Ocean, and is especially intense across the western transect (see Table 1), contrary to the surface and central waters where organic carbon was imported from the Atlantic Ocean. Furthermore, the mass balance of the intermediate waters (Fig. 5a) indicates that there should be an entrainment of 1.3 ± 0.5 Sv into the deep waters for volume conservation, that translates into a drawdown of organic carbon of 3.3 ± 1.4 × 10 8 mol C d − ¹, 76% of which is DOC (Fig. 5b). In addition, an average particle flux below 700 m of 0.46 mmol C m − ² d − ¹ has been obtained from three sites located within the Canary Current region 69 . Assuming this value for our region of study, we obtain a vertical POC supply to intermediate waters of 6 The net R within the mesopelagic layer, assumed to be equivalent to the sum of the central and the intermediate waters, is 22.7 ± 5.2 × 10 8 mol C d − ¹ (1.6 ± 0.4 mmol C m − ² d − ¹). This value is lower than those obtained with a similar box model west of the Canary Islands 20 : 2.4-5.1 mmol C m − ² d − ¹. According to these authors, POC susp would support 28-59% of the total mesopelagic R in that region. However, their estimates were based solely on POC susp fluxes, assuming two scenarios in which the contribution of DOC was fixed at 15 and 30%. In the present study, we have measured DOC and calculated that it contributes about 53% to total mesopelagic R within the box. It has been hypothesized that DOC represents about 90% of the respiration of non-sinking organic carbon in the mesopelagic layer from an empirical carbon mass balance within a box bounded by the Strait of Gibraltar, 24-41° N and 22° W 39 . Our results indicate that DOC represents that fraction when only the non-sinking organic carbon is considered, in agreement with the high DOC consumption rates obtained in areas undergoing water mass transformation 6,71 .
The estimated mesopelagic R is much lower than the oxygen utilization rates (OUR) calculated for the North Atlantic subtropical gyre 72-74 (9-11 mmol C m − ² d − ¹ as converted using a -O 2 /C molar ratio of 1.4 75 ). In addition, a carbon budget north of our box region during summer 2009 76 estimated a total input of organic carbon to the twilight zone (ca. 50-1000 m) of ~7.7 mmol C m − ² d − ¹, due mainly to sinking POC. OUR values ranging from 4.7 to 16.6 mmol C m − ² d − ¹ between 100 to 750 m depth were obtained using tritium and 228 Ra methodologies in the North Atlantic subtropical gyre 77 . Similar integrated values (4.9 to 13.0 mmol C m − ² d − ¹) of carbon respiration for the 100-1000 depth zone were obtained both by 7 Be and POC flux attenuation methodologies in the subtropical North Atlantic 78 . A compilation of data from the Pacific ocean 79 shows OUR values ranging from 0.04 to 18.1 mmol C m − ² d − ¹. In the North Pacific, OUR values integrated over the first 600 m depth 80 averaged 4.6 mmol C m − ² d − ¹, while higher OUR values of 9.8 to 11.7 mmol C m − ² d − ¹ have been recently reported for a monthly time-series at HOT 12 . Values obtained during the present study might, however, underestimate the annual mean since our study took place during the unproductive autumn when POC susp concentrations are particularly low 56 . Additionally, the region within the box is strongly influenced by mesoscale features such as filaments and eddies, whose effects on carbon sequestration remain to be quantified. The enhancement of carbon fluxes to the ocean interior driven by mesoscale eddies has been reported in a number of studies [81][82][83][84][85][86][87][88] . Indeed, injection of organic carbon from surface waters was especially intense in several areas, such as in the anticyclonic eddy at the northwestern corner of the box (stations 33-35), or the Azores Front (stations 52-55 44 ). Moreover, upwelling filaments may play a key role in the coastal-ocean transport of organic matter 89,90 . The Cape Ghir filament lies within the region of study, and may export to the open ocean 29 to 63% of the annual averaged primary production associated with the coastal upwelling 91,92 . All of these features occur episodically within the box, and could enhance the supply of organic carbon to mesopelagic waters.
Near the ESTOC site, net R rates at the mesopelagic zone (ca. 150-700 m) of 2.0-3.1 mol C m − ² y − ¹ (5.5-8.5 mmol C m − ² d − ¹) were obtained by using a tracer conservation model applied to climatological data 93 . They also obtain consistent R rates based on ETS (electron transport system) respiratory activity for the same region at the end of March. Differences between these estimates and our values could be due to higher lateral advection of DOC and POC susp from the coastal upwelling region at that site 56 , and higher productivity at the end of the late winter bloom.
In summary, this study provides an estimate of net respiration in the mesopelagic waters within a box enclosing the eastern boundary of the North Atlantic subtropical gyre. This estimate is based on the assumption that the box is in steady state, i.e., there is neither accumulation of organic carbon entering the box nor consumption of previously accumulated organic carbon. This assumption is acceptable for the budget of the central, intermediate, and deep waters as their decadal renewal times preclude seasonality. However, seasonal variability is pronounced in the surface layer and, hence, the steady state assumption could lead to biased values of the annual net fluxes. Furthermore, although the contribution of mesoscale phenomena needs to be addressed in future studies, our results show that a major fraction of the mesopelagic organic carbon demand in this area is fuelled by DOC. Indeed, the formation of MW has been found to be the main mechanism for the export of anthropogenic carbon in this region 38 . Despite the inaccuracy of the steady-state assumption the carbon budget presented here shows that the vertical entrainment of ENACW into subducting MOW contributes to the export of DOC to mesopelagic waters, where it may constitute the main substrate supporting dark ocean respiration. Likewise, due to the westward advection of intermediate waters, this process may also represent a significant supply of organic carbon to the eastern North Atlantic outside the box.

Materials and Methods
Hydrography and seawater sampling. Between October 15 and November 11, 2009 the RV Hespérides carried out an intensive hydrographic survey in the eastern boundary of the North Atlantic subtropical gyre at the confluence region between the Azores Current System, the Portugal Current and the Canary Current. A grid of 81 conductivity-temperature-depth (CTD) stations was distributed along a box region defined by 28.7-37.0°N and 24.5°W. At 47 of these stations seawater samples were taken for analysis of dissolved (DOC) and particulate organic carbon (POC; Fig. 1, numbered dots).
Readers interested in a detailed description of the hydrography and water mass characterization are referred to Pérez-Hernández and coworkers 44 . Organic carbon. Discrete samples of POC were obtained at selected depths from the surface up to the bottom (5, 25, 50, 100 200, 400, 600, 800, 1000, 1500, 2000, 3000, 4000 and 5000 m) by means of a rosette sampler equipped with twenty-four 10 L Niskin bottles. Although we are aware that particles sink at different rates, we assume that all the POC collected in oceanographic bottles corresponds to the slowly-settling or suspended pool (POC susp ) 20 . Thus, this pool represents an upper threshold for the particulate organic material that is susceptible to lateral transport. Water samples (4 L) for POC susp were collected in polypropylene bottles and filtered through precombusted (450 °C, 12 hours) 25 mm Whatman GF/F-filters (pore size 0.7 µm). The filters were wrapped in precombusted aluminium foil and frozen at −20 °C. In the laboratory, the filters were thawed and dried overnight at 55 °C, then placed overnight in a desiccator saturated with HCl fumes to remove inorganic carbon, and dried again in a second desiccator with silica gel for at least 24 hours. Finally, the filters were packed in tin sleeves before being analysed with a Perkin-Elmer 2400 CHN elemental analyser following standard protocols 94 . Unused pre-combusted GF/F-filters were treated in the same way and used as handling blanks. Blanks ranged from 0.06 to 0.7 µmol C L −1 . DOC adsorption onto the filters was estimated at several random stations, at three different depth levels, as the amount of carbon retained on a backing filter placed underneath the main filter. The values ranged from 0.17 to 1.73 µmol C per 25 mm-diameter GF/F-filter, which are similar to previous estimates 20, 95,96 .
At the same stations and depth levels, samples for the analyses of DOC were collected in 250 mL acid-cleaned all-glass flasks. Samples from the upper 100 m were immediately filtered through precombusted (450 °C, 4 hours) 47 mm Whatman GF/F filters in an acid-cleaned all-glass filtration system. Filtered surface and unfiltered deeper water samples were collected in 10 mL precombusted glass ampules (450 °C, 12 hours). After acidification with H 3 PO 4 , the ampoules were heat-sealed and preserved in the dark at 4 °C until analysis in the laboratory with a Shimadzu TOC-V organic carbon analyser by high temperature catalytic oxidation (HTCO). The system was calibrated daily with potassium hydrogen phthalate (99.95-100.05%, p.a., Merck). The precision of the DOC calibration was ± 1 μmol L −1 . The performance of the instrument was tested with the carbon reference materials (CRM) provided by D. A. Hansell (University of Miami, USA). Measured concentrations of the CRM were 45.5 ± 1.7 µmol C (n = 10); the certified value is 44-46 µmol C L −1 (lot#09-06 from the Florida Strait at 700 m).
Organic carbon budget. The water column was divided into discrete neutral density layers 97 . The upper four layers coincide roughly with the main thermocline waters (below 27.38 neutral density layers; 0-700 m), the following three layers with intermediate waters (27.38-27.92; 700-1500 m), and the lowest seven layers with deep waters (27.92-28.10; below 1500 m). Our estimates were based on the inverse box model used by Pérez-Hernández and coworkers 44 to obtain mass transports (M) for each layer along the northern, western, and southern transects. The model uses the thermal wind equation after calculating the reference level velocities and their uncertainties. These velocities are calculated assuming geostrophy, and mass and property conservation. The inverse box model used has been extended from a previous one 98 to include the approximate conservation of mass and anomalies of salinity and heat, and to allow transfer between layers. This model also considers adjustment of fresh water fluxes and Ekman transports in each section. The inverse problem consists of 37 equations and 130 unknowns: 99 for reference velocities, 26 for vertical velocities and vertical diffusion, 4 for Ekman transports and 1 for the freshwater flux. To solve the inverse problem, the Gauss-Markov method, which produces a minimum error variance solution from initial estimates of the unknowns, was used. M at each layer is multiplied by the averaged suspended and dissolved organic carbon concentrations (C) at the same layer to obtain the corresponding lateral carbon fluxes (M × C). To report these carbon fluxes, the layers were grouped into surface, central, intermediate, and deep waters. The lateral carbon fluxes were summed to obtain the corresponding carbon budgets for each layer.
The robustness of the carbon fluxes and budgets calculated in this study was tested by means of the following perturbation test (see supplementary material): carbon fluxes were calculated as (M ± erM) × (C ± erC), where erM and erC represent the uncertainty of the estimation of M and C, respectively. Therefore, the values of M and C are perturbed within the limits of the uncertainty of their respective estimations. A total of 100 perturbations were performed for each carbon flux and budget. The average of these 100 values is considered the optimum solution and the corresponding standard deviation (SD) an estimate of the uncertainty of the fluxes and budgets 99 . See Pérez-Hernández and coworkers 44 for a thorough analysis of the values of M and erM used in the present calculations. Concerning the uncertainty of POC susp and DOC measurements, these were obtained by calculating the SD of all suspended and dissolved organic carbon measured in each layer at each transect.
To obtain the organic carbon fluxes into/out of the Mediterranean Sea, mass transports across the Strait of Gibraltar were set at −0.78 ± 0.1 Sv for the surface outflow to the Mediterranean and 0.68 ± 0.1 Sv for the bottom inflow to the eastern North Atlantic 100 (1 Sv = 10 9 kg s − ¹). These mass transports, which are within the range of other estimates in the literature [101][102][103][104][105][106][107]  Data Availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.