4.1. Coastal Ocean Gradients in Suspended POM
 Particles in the ocean exist in a continuum of sizes, but two classes are frequently recognized: suspended and sinking matter. They are distinguished operationally by the sampling method used to collect them: suspended matter with oceanographic bottles or large-volume in situ pumps [e.g., Turnewitsch et al., 2007], and sinking matter with sediment traps. Nevertheless, bottles may also trap particles with moderate to fast sedimentation rates [Gardner, 1977], whose contribution to the overall POM will depend on the relative importance of the sinking pool versus the suspended pool. The majority of POM in the water column is thought, however, to be formed by fine suspended material or particles with almost negligible sedimentation rates compared with horizontal fluxes [McCave, 1984; Kepkay, 2000; Verdugo et al., 2004]. These particles, referred to here as POCsusp, would be characterized by a conservative behavior in the water column undergoing lateral transport [Brun-Cottan, 1976].
 In this study we have assumed that all POM collected with bottles represents suspended material or particles with very low sedimentation rates, susceptible of being transported horizontally. Strong evidence supports this hypothesis. Recent studies south of the Canary Islands, based on particle settling velocities using IRSC (Indented Rotating Sphere Carousel) traps [Peterson et al., 1993] during a 6 month period (summer to autumn 2005), indicate that about 60% of the POM collected in the traps have low sinking rates (<5 m d−1), allowing particles to be laterally transported I. J. Alonso-González, unpublished data, 2007). Comparatively, the average POC sinking velocity recorded using the same sediment traps at the DYFAMED station in the Mediterranean Sea, is of 350 m d−1 [Peterson et al., 2005; Armstrong et al., 2009]. The reason for the low sedimentation rates in the Canary region is still a matter of debate [e.g., Arístegui et al., 2003], but would explain the low export ratio with respect to the f ratio, reported by Neuer et al. [2002b] at the ESTOC station, north of the Canary Islands, compared to the Bermuda Time Series Station (BATS). Paradoxically, the POM concentrations collected with bottles in the water column are much higher in the Canary region. Indeed, POC at the BATS [Steinberg et al., 2001], like in other open ocean regions [e.g., Menzel and Goering, 1966], decrease exponentially with depth, reaching values typically <1 μmol L−1 at depths >200 m. In our study, POC concentrations in the deep ocean (down to 3000 m) were >2 μmol L−1 in the eastern sector of the box, decreasing toward the open ocean. An intermediate situation (range 1–3 μmol L−1 POC at depths >200 m) is found in the Azores region [Vezzulli et al., 2002], a transition zone between the Canary Current and the central waters of the North Atlantic subtropical Gyre (NASG).
 Our results agree with the autumn values reported by Neuer et al.  for the 200–1000 m layer at the ESTOC, which were the lowest recorded during the annual cycle. The highest average POM values in the upper 1000 m (as high as 8 μmol L−1) were reported for the spring months. No measurements were, however, carried out during summer [Neuer et al., 2007], when the upwelling activity, and hence the potential offshore export, is higher. Indeed, POC concentrations measured between the Canary Islands (28°N) and Cape Blanc (21°N), along four cruises carried out during spring (2003) and summer (2002), reached typical average concentrations of about 6 μmol L−1 down to at least 2000 m J. C. Vilas, personal communication, 2007), without any indication of an exponential decrease with depth. Thus, lateral transport of organic matter generated over the NW Africa continental shelf may largely contribute to the POMsusp observed in the Canary Current region. The transport would be more pronounced during the periods of higher upwelling intensity, decreasing the concentrations toward the open ocean because of remineralization processes. Vilas et al.  reached the same conclusion after studying the distribution of POM around Seine, a seamount placed east of Madeira Island, in the Canary Current. These authors observed peaks of POC in the 200–1000 m layer at the stations closer to the continental shelf and coinciding with the highest activity of the upwelling system.
4.2. Offshore Suspended Organic Carbon Pumping to the Interior of the North Atlantic Ocean
 To estimate the net lateral flux of POCsusp in the mesopelagic and bathypelagic waters, we first subtracted the open ocean “baseline” POC concentration, obtained from averaging the monthly POC profiles in each depth layer at the Bermuda Atlantic Time-series (BATS) station, from the POC concentrations at the east and west transects. The BATS station, placed at the center of the NASG, presented a typical exponential decrease of POC with depth, reaching values <1μM below 200 m. In a comparative study of three subtropical time series stations (BATS, HOT and ESTOC), Neuer et al. [2002b] concluded that the annually integrated net primary production was similar at all three sites, with slightly lower values at ESTOC. However, the POC sinking flux (measured with drifting traps at about 150–200 m) was approximately 2–3 mmol C m−2 d−1 at BATS (Bermuda) and HOT (north of Hawaii), while at ESTOC (north of the Canary Islands) was only about 20% of these values (0.55 mmol C m−2 d−1). Recent results J. Arístegui, unpublished data, 2007; Alonso-González, unpublished data, 2007) of sinking POC, collected with similar traps at 150–200 m, from 30N to 21N, outside the influence of the intense eddy field, indicate that the POC flux varies on average from 0.7 to 2 mmol C m−2 d−1 from spring to autumn, increasing to 3–4 mmol C m−2 d−1 during the late winter bloom. Since sinking fluxes in our region of study are not higher than in BATS, we consider that the BATS POC would characterize the typical POC profile in the western Canary region, assuming only vertical flux of POC.
 Lateral fluxes were particularly intense along the east and west transects. Thus, we focused our analysis on the net balance considering the fluxes through these two transects. We calculated a lateral flux of suspended POC at the central and intermediate waters (100–1700 m depth layer) of 8.77 × 108 mol C d−1 and 4.8 × 108 mol C d−1 for the east and west transects respectively, extending along 1.01 × 106 m in length (Figure 5). In any case, a simple sensitivity analysis was added to assess the effect of baseline subtraction to the magnitude of the lateral POC fluxes. First, the BATS POC data were fitted to a power law function, and then a “fake baseline” was generated by applying the same equation observed at BATS to higher surface POC values (Figure 6). Subtracting the fake baseline from the POC concentrations at the east and west transects, the lateral fluxes of suspended POC at Central and intermediate waters were 5.78 and 4.26 × 108 mol C d−1 for the east and west transects respectively. The POC increase over the BATS baseline (≈1.5 times higher) results in a decrease in the lateral POC fluxes of 34% for the east transect and 11% for the west transect.
Figure 5. Offshore suspended organic carbon fluxes (108 mol C d−1) in the different water masses. The values were obtained by subtracting the open ocean baseline POC concentration calculated from the Bermuda Atlantic Time-series Study (BATS). Note that the highest POC fluxes are in central waters of the more coastal section.
Download figure to PowerPoint
Figure 6. Comparison of the average POC profile measured during this study (CORICA) with the baselines used in the sensitive analysis to compute lateral POC fluxes. The BATS baseline was generated by fitting a power law equation to the average BATS POC profiles (black line). A fake baseline (dashed line) was derived by applying the same equation observed at BATS to higher surface POC values (see text for details).
Download figure to PowerPoint
 Table 2 compares the vertical and lateral POC transport in the Canary region. The lateral fluxes correspond to this study, whereas the vertical fluxes were obtained from surface-tethered and deep-moored traps [Neuer et al., 1997, 2002b, 2007; Arístegui et al., unpublished data, 2003]. As observed, the lateral POC fluxes are 2 or 3 orders of magnitude higher than vertical fluxes depending on water masses, confirming our hypothesis of a more relevant horizontal versus vertical flux of POM per unit area. This conclusion is partly in agreement with the work of Bauer and Druffel  who suggested that suspended POC inputs from ocean margins to the open ocean interior might be more than an order of magnitude greater than inputs of recently produced organic carbon derived from the surface ocean. However, we should keep in mind that the overall differential effect of the POC transport will depend on the aerial extension of the vertical versus lateral fluxes under consideration.
Table 2. Estimates of Vertical and Lateral POC Fluxes in the Different Water Massesa
|Water Masses||Range Depth (m)||Vertical Fluxes (mmol C m−2 d−1)||Lateral Fluxes (mmol C m−2 d−1)|
|Surface waters||0–100|| ||9245|
4.3. Carbon Budget and Mesopelagic Respiration
 The surface waters (0–100 m) inside the box received an overall higher external input (17.48 × 108 mol C d−1; sum of north and east transects) of POCsusp than exported (12.39 × 108 mol C d−1; sum of south and west transects) outside the box (Table 1). The resultant balance is 5.09 × 108 mol C d−1 (0.85 mmol C m−2 d−1). This value is in the range of the sinking POC flux reported above for the region of study, but somewhat higher than that estimated by Neuer et al.  for the ESTOC (European Station for Time series in the Ocean, Canary Islands) north of the Canaries (Table 2). Sedimentation rates measured downstream the Canary Islands are however 2–4 times higher [Arístegui et al., 2004], because of the enhanced production and eddy-filament exchange processes along the intense mesoscale field [Arístegui et al., 1994, 1997; Barton et al., 1998, 2004]. Island eddies are known to increase productivity leading to positive net community production [e.g., Arístegui and Montero, 2005]. A fraction of this excess production not sunk down may be laterally advected to the open ocean, contributing to the surface carbon budget inside our box. Assuming nonsignificant atmospheric inputs, the amount of carbon sedimented into the dark ocean (see above) would match the external inputs into the box, leading to a carbon balance in the surface waters. Nevertheless, Dachs et al.  have reported high average net gaseous diffusive air-water fluxes of organic carbon (25–30 mmol C m−2 d−1) in the subtropical northeast Atlantic, which in case they occur during our study, would shift the balance toward a strong heterotrophy. Independently of whether the surface waters were or not in metabolic balance (see discussions by Duarte et al. , Dachs et al. , and Neuer et al. ), the vertical export flux of POC to the dark ocean, which is the flux in which we are interested in our study, seems to be quite constant from spring to autumn, as derived from the sediment traps records.
 In order to calculate the respiration rate that could support the total carbon supply below the surface waters, we considered all the carbon fluxes inside the box. Sinking POC collected with surface-tethered traps range from 0.55 to 2 mmol C m−2 d−1 (see the above section). Drifting sediment traps have been frequently reported to underestimate the export flux [Michaels et al., 1994; Buesseler, 1998; Buesseler et al., 2000; Neuer et al., 2007]. However, results comparing POC flux in BATS, both derived from surface-tethered traps and 234Th [Buesseler, 1998], show that during low-productivity (PP) periods (average PP from March to October: 36 mmol C m−2 d−1) the traps and 234Th estimates reasonably agree, yielding average POC sedimentation rates of 2.6 mmol C m−2 d−1, which corresponded to an export/production ratio (e ratio) of 5–10%. At the ESTOC station (which is closer to the upwelling region than our sampling area), the average annual PP is about 30 mmol C m−2 d−1, with the lowest values (<10 mmol C m−2 d−1) recorded during autumn [Neuer et al., 2007]. Similarly, low PP was measured by Basterretxea and Arístegui  in an offshore station west of the Canary Islands, during summer. An e ratio of 5–10% would therefore lead to a POC flux of 0.5–1 mmol C m−2 d−1 at PP = 10 mmol C m−2 d−1, and 1.5–3 mmol C m−2 d−1 at PP = 30 mmol C m−2 d−1. These calculations provide confidence to the measured range in sedimentation rates with drifting traps (0.55–2 mmol C m−2 d−1), which would represent a reasonable lower/upper scenario for passive sinking of POC in our region of study.
 Diel migrating zooplankton contributes also to the vertical flux of POC (the so-called “active flux”) by feeding in surface during the night and defecating unassimilated POC at depth during the day. Hernández-León et al.  estimated an active flux of 0.22 mmol C m−2 d−1 in the eddy field downstream the Canaries, similar to the average value (0.17 mmol C m−2 d−1) reported by Steinberg et al.  for BATS. These fluxes represented 25% and 8% of the passive POC fluxes in their respective regions. If we assume that the active flux represents at best a 20% of the passive flux, the vertical POC flux (passive + active) in our box would range from 0.7 to 2.4 mmol C m−2 d−1. Multiplying these values by the box area (6 × 1011 m2) yield a total POC supply of 4.2–14.4 × 108 mol C d−1.
 The central waters (approximately 100–700 m; hereafter named mesopelagic zone) inside our box received a net lateral POC supply of 8.5 × 108 mol C d−1 (Table 1). If we assume that about 90% of the vertical POC flux is respired in the mesopelagic zone [Arístegui et al., 2005b], and adding the net lateral carbon supply we obtained a total mesopelagic POC reservoir (POCmeso) of 12.3–21.5 × 108 mol C d−1.
 How this POC flux compares with the DOC flux? The relative contribution of dissolved organic carbon (DOC) to total mesopelagic respiration (R) was estimated by Arístegui et al. , in a section spanning the coastal transition zone (CTZ) in the Canary Current. In their study, DOC contributed 30% to the total mesopelagic oxygen consumption, a value 2 times higher than the average (15%) calculated by Arístegui et al.  for the global ocean. Unlike the eastern Canaries-CTZ our region of study was characterized by a stable surface thermocline, and was outside the influence of the eddy field region, which might enhance vertical mixing. Thus, we would expect to have a lower contribution of DOC to total mesopelagic R. In a best case scenario of a 30% contribution of DOC, the total carbon fluxes (lateral and vertical fluxes of DOC and POC) would support an integrated (100–700 m) mesopelagic respiration rate of 2.9–5.1 mmol C m−2 d−1 (computed by dividing POCmeso by both the box area and 0.7). If DOC contributed 15% (a more reasonable contribution) the total R would be 2.4–4.2 mmol C m−2 d−1 (POCmeso/box area × 0.85). From these calculations we can infer that the lateral POC would account for 28–49% of total mesopelagic R in the 30% DOC scenario, and 34–59% in the 15% DOC scenario.
 Total R is about an order of magnitude lower than that estimated by Arístegui et al. [2005a] for the Canary Current region, during summer time, combining actual oxygen consumption measurements and enzymatic activities (ETS activity). The discrepancy in the magnitude of the rates may be explained by the fact that, during the summer sampling, the POCsusp concentrations were about 4–6 times higher in the water column than during this study (not shown) and that the summer rates were averaged including near-coastal stations downward the Canary Islands region, where vertical POC flux is several times higher.
 Our calculated R estimates are also about half of the mesopelagic oxygen consumption rates (9 mmol C m−2 d−1) reported by Jenkins and Goldman  for the NASG, inferred from changes in the apparent oxygen utilization (AOU) and the use of tracers to calculate the apparent age of the water mass. A similar twofold imbalance was observed in BATS between the AOU/tracers approach when compared with estimations of POC and DOC 1-D fluxes in the mesopelagic zone [Carlson et al., 1994; Michaels et al., 1994]. Although there are known pitfalls in the accurate determination of carbon fluxes [e.g., Buesseler, 1998; Hansell, 2002; Arístegui et al., 2005b], we must not exclude the uncertainty associated with the inference of mesopelagic respiration from changes in the oxygen field, because of eddy diffusivity [Jenkins and Wallace, 1992].
 The sources of the lateral mesopelagic carbon fluxes to our region of study may be variable. Upwelling filaments have been identified as playing a key role in coastal ocean export of organic matter [Álvarez-Salgado et al., 2001]. According to Barth et al. , the exported organic matter from filaments may be forced downward along sloping density surface through conservation of potential vorticity along the meandering jet path. These authors found a mesopelagic chlorophyll maximum in the California Current System over 300 Km offshore and between 150 and 250 m, reporting a carbon injection into the adjacent deep ocean of 2.4 × 106 Kg C per event. They suggested that the entire benthic mineralization rate could be supplied by five of these events per year. Eddies downstream the islands [Arístegui et al., 1997; Barton et al., 1998, 2004] are another potential source for deep-water transport of organic matter. Barton et al.  estimated that island eddies may contribute to the nitrogen flux to the Canary region as much as coastal upwelling. McGillicuddy et al.  calculated that carbon export inferred from oxygen anomalies in eddies in the Sargasso Sea accounted for one to three times as much as annual new production in the region. On the other hand, Arístegui et al.  and Arístegui and Montero  showed that anticyclonic eddies may entrain high-chlorophyll water from upwelling filaments with which they interact. Recent unpublished studies (e.g., Arístegui and Alonso-González, 2007) show that eddy entrainment is particularly strong in the frontal regions between eddy pairs and filaments, enhancing sinking of organic matter to depths >800 m. This organic matter could be laterally transported, accounting for some of the deep-water maxima observed in POMsusp. Additionally, meddies, frequently reported for the Canary Basin [Richardson et al., 1991; Shapiro and Meschanov, 1996; Richardson et al., 2000], may largely contribute to the deep-water transport of organic matter. During this study the core of a meddy was sampled (29°N, 20°W; 700–1500 m) yielding an average POC concentration of ≈7 μM. If we consider a southwestward flow of 1.2 Sv (as calculated by Hernández-Guerra et al. ), the POC flux transported by the meddy would be 6.58 × 108 mol C d−1. This value is only slightly lower than the net POC consumption in the Central waters. All these features (filaments, eddies and meddies) are limited in spatial extent and occur episodically, hence the difficulty in observing such events in hydrographic surveys or sediment traps. Therefore, the mesopelagic carbon deficit could be supplied by a few of these mesoscale events per year.
 In conclusion, our results suggest that a significant fraction of the mesopelagic carbon budget in the western Canary Current may be fuelled by lateral suspended carbon advection from continental margins or mesoscale activity in the eastern boundary region. A large part of this mesopelagic carbon would be consumed in the boundary current, instead being transported to the open ocean. Our observations indicate that the influence of the lateral advected particulate carbon from the NW African coast on the oligotrophic subtropical gyre region can reach more than 1000 km offshore. If the lateral suspended POC fluxes estimated in this study are confirmed for other boundary regions, the coastal-open ocean POC transport would play a key, but unaccounted, role in the global carbon cycle of the oceans.