We employ the UCLA-ETH version of the ROMS , coupled with the NPZD biogeochemical ecosystem module of . ROMS solves the 3-D hydrostatic primitive equations of flow on a discretized curvilinear grid, using terrain following vertical coordinates (sigma levels). Surface elevation, barotropic and baroclinic horizontal velocity components, potential temperature, and salinity are its prognostic variables. Vertical mixing is parameterized by the K profile parameterization (KPP) scheme .

The NPZD ecosystem module is a nitrogen-based model with two limiting nutrients, i.e., nitrate and ammonium, one class of phytoplankton, one class of zooplankton, and two detritus pools, i.e., a small one that sinks very slowly and a large one that is subject to more rapid sinking . These components plus a dynamic chlorophyll-to-carbon ratio form the seven prognostic variables of the nitrogen component of the model. An additional four state variables have been added to reflect the cycling of carbon and oxygen, namely dissolved inorganic carbon (DIC), alkalinity, mineral CaCO3, and dissolved oxygen (O2) . The carbon and oxygen cycles are linked to the nitrogen cycle by fixed stoichiometric ratios. Thus, the fluxes of organic carbon diagnosed in our model are actually the fluxes of organic nitrogen multiplied by the C : N ratio assumed in the model, i.e., 117 : 16.

POC in the model, i.e., the sum of phytoplankton, zooplankton, and the two detritus pools, is subject to advection and diffusion in vertical and horizontal directions, as well as explicit vertical sinking (the latter does not apply to zooplankton, however). Sinking velocities for phytoplankton, small detritus, and large detritus are 0.5, 1, and 10 m day-1, respectively (see , for a complete set of parameters). POC is formed through phytoplankton growth and is lost through zooplankton respiration and the remineralization of the two detrital pools. Within the organic carbon pool, phytoplankton mortality feeds into the small detritus pool, while zooplankton mortality and phytoplankton excretion is routed to both the small and large detritus pools with constant proportions. Coagulation of phytoplankton with small detritus as well as coagulation of small detritus with small detritus also forms large detritus, while no disaggregation is considered in our NPZD model, i.e., large detritus cannot disaggregate back into small detritus. Bacterial remineralization of organic matter is modeled as an implicit process through the definition of constant remineralization rates and takes place both in the water column and in the sediments. The sediments act as a temporal buffer in our model, receiving the organic matter from the water column and then slowly remineralizing it back to its inorganic constituents, which are immediately released back to the overlying water column. Neither burial nor sediment resuspension is considered. As the model does not include an explicit DOC pool, our modeled total organic carbon corresponds to POC only. However, the small detritus, given its very small sinking speed, behaves essentially as a suspended POC pool, i.e., it shares many similarities to DOC. To assess the possible implications of neglecting DOC, we run a sensitivity experiment where we turn the small detritus pool into essentially DOC by setting its sinking velocity to zero and by reducing it coagulation rate.

In order to optimally model the long-range transport of organic matter from the CanUS to the open North Atlantic, we employ a newly developed Atlantic telescopic grid (Fig. ) that covers the full Atlantic basin (-60 to 70∘ N) while having a strong focus in resolution toward the northwestern African coast. This was achieved using a conformal mapping that moves one pole of the ROMS grid over northwestern Africa (4∘ W, 20∘ N) and the other over central Asia (75∘ E, 37∘ N). The grid has a dimension of 813 × 397 grid points, with a resolution that goes from a full eddy-resolving 4.7 km near the African coast to a relatively coarse 50 km in the western South Atlantic and in the Caribbean (Fig. ). Within the CanUS region as defined in our analysis, the resolution ranges between 4.7 km at the coast and 19.5 km. In the vertical, we are considering 42 terrain-following (σ) levels with a surface refinement that allows to have a better vertical resolution in the euphotic layer. We use a new vertically stretched grid that transitions more rapidly from a pure terrain-following orientation to a more horizontal orientation in order to reduce spurious mixing in the stratified open ocean.

The coupled ROMS + NPZD model was run with monthly mean climatological forcing at the surface, including the fluxes of heat and freshwater, solar shortwave radiation, wind stress, and atmospheric pCO2. All forcings were derived from ERA-Interim reanalysis , with the exception of atmospheric pCO2, which was computed from the GLOBALVIEW marine boundary layer product (; see , for details). A detailed description of the data sources used for the forcing is provided in Appendix A: Table A1a and b. We next describe some corrections we had to apply to the forcing.

The shortwave radiation and total heat flux fields based on ERA-Interim have been shown to be biased high in regions of persistent cover with stratocumulus clouds . This is particularly relevant in the southern subregion of the CanUS, where stratocumulus are very pervasive and not well represented in the ERA-Interim reanalysis. We thus apply corrections to the original ERA-Interim reanalysis fields, taking advantage of the work by the Drakkar community . They have already developed corrected forcing fields, i.e., the Drakkar forcing sets (DFSs) on the basis of the ERA-Interim reanalysis , providing the corrected fields for the downwelling surface shortwave radiation (DSWR) and downwelling surface longwave radiation (DLWR). We thus compute correction masks ourselves, using the difference between the DFS and the uncorrected ERA-Interim data sets as the basis. Concretely, we first computed monthly means of the DFS daily climatology DSWR and DLWR. The monthly climatological means of the same two variables from ERA-Interim for the period were then used as a reference to calculate correction masks (C) for each month by simply differencing, i.e., Cdswr=ERAdswr-DFSdswr and Cdlwr=ERAdlwr-DFSdlwr. These correction masks were then regridded to our grid and applied to our ERA-Interim-derived monthly climatological mean forcing solar radiation (S) and total heat flux (TH) so that S′=S-Cdswr and TH′=TH-(Cdlwr+Cdswr).

Another correction to the forcing regards the regions with sea ice cover at the northern boundary of the domain. Here we account for freshwater and latent heat fluxes associated with sea ice formation and melting by correcting the surface fluxes of the model forcing. An offline correction for the forcing was calculated from the ERA-Interim sea ice fraction (c) and National Snow and Ice Data Center (NSIDC) sea ice drift (u) monthly climatologies. Using these two data sets, the corrections to the freshwater flux (Ffw) were calculated according to but simplified by using monthly mean climatologies, a constant sea ice thickness h0=1.5 m, and a sea ice density ρice=910 kg m-3. An analogous equation was used to calculate the correction to the heat flux (Fh), so that Fh=ρiceL⋅h0(∂(Ac)/∂t+∇(Acu)), where A is the grid box area, and L the latent heat of fusion of water. The heat flux is constrained throughout the model simulation, so that the surface temperature cannot drop below the freezing temperature , which prevents strong heat loss in areas covered by sea ice.

River runoff in the form of monthly climatological data was also added to the freshwater fluxes. River sources were regridded to the closest ocean grid point and spread to a number of adjacent ocean grid points that depends the order of magnitude of the incoming flux to avoid numerical problems.

The model was run with open lateral boundaries at all grid boundaries confined by water, including at the Strait of Gibraltar. These climatological monthly lateral boundary conditions were prepared the same way as in previous studies . A detailed description of the data sets used for the boundary conditions can be found in Appendix A: Table A2.

The model was initialized to be at rest with temperature, salinity, nitrate, and the inorganic carbon parameters corresponding to the mean climatological value of December and January. The remaining biogeochemical variables were initialized to small non-zero values. The model was then run forward in time for a total of 35 years, using the monthly climatological forcing described above. We use the first 29 years as spin-up, and undertake our analysis using the last 6 years (years 30 to 35) of the simulation. This permits us to obtain a good representation of the climatological mean state of the CanUS, i.e., to average out the substantial intrinsic variability present in the setup.

To quantify the offshore transport of organic carbon including all its biogeochemical transformations, we undertake a full budget analysis, calculating all fluxes of organic carbon within each grid box and between all adjacent boxes. The organic carbon budget analysis fluxes include physical fluxes through the boundaries of the boxes and the integrated biological flux within each box. Physical fluxes include vertical and horizontal advective fluxes in the three directions, vertical mixing fluxes associated with the eddy diffusivity, and the vertical sinking flux. The net biological flux of organic carbon within each box is equivalent to the net community production (NCP), i.e., the net amount of carbon added or removed by biological activity, computed by summing all organic carbon production processes (phytoplankton growth) minus the sum of all processes that convert organic carbon back to inorganic forms (respiration and remineralization). In our analysis we disregarded the contribution of the horizontal and vertical mixing associated with the background diffusivity. We also used a fixed depth for the sigma layers that define the box boundaries, disregarding their vertical oscillations. Both approximations can result in small residuals in the budget analysis.

For the whole CanUS, defined as the region between 9.5 and 32∘ N, we have defined two layers of three-dimensional large-scale boxes. Each depth layer has a constant thickness of 100 m, very close to the mean depth of the euphotic layer in the CanUS region (98.7 m), so that our analysis spans in total the first 200 m of depth. Each depth layer is subdivided into the same five offshore boxes up to 2000 km distance from the northwest African coast according to the following ranges of distances (1) 0 to 100 km from the coast, narrow coastal box directly influenced by the coastal upwelling; (2) from 100 to 500 km offshore; (3) from 500 to 1000 km offshore; (4) from 1000 to 1500 km offshore; and (5) from 1500 to 2000 km offshore.

To highlight the different roles of the three fundamental zonal bands in the CanUS, we divided the EBUS into three subregions (southern, central, northern), maintaining for each subregion the same five offshore domains as for the previous analysis and considering only the euphotic layer (0–100 m). Subregional boundaries were placed at 17 and 24.5∘ N. This allows us to distinguish between three regimes of circulation and production of the CanUS: a northern subregion dominated by coastal upwelling and coastal filaments, a southern tropical subregion, and a central subregion where the Canary and Mauritanian currents (CC and MC, respectively) converge to form the Cabo Verde front . The lateral extension of the full CanUS boxes and of the subregional boxes is presented in Fig. 3 together with the pattern of the modeled currents.

The simulation reveals in the long-term mean a strong onshore–offshore difference in the vertically integrated NCP, here ∫NCP, i.e., primary production minus respiration and remineralization integrated from the bottom of the ocean up to the surface, including remineralization in the sediments (Fig. a).

The full water column ∫NCP is negative across nearly the entire eastern subtropical North Atlantic, while only a narrow strip of less than 100 km along the northwestern African coast and a few offshore regions in the southern part of the domain have positive values. This implies that the majority of the region is net heterotrophic, as within each column of water including the sediments more organic matter is being consumed than what is being produced locally. In contrast, the shelf regions of the CanUS are characterized by high levels of organic carbon production that exceed local consumption in the total water column and sediments, leading to a positive ∫NCP and therefore an excess of organic carbon, which is available for lateral export.

A very different offshore gradient exists if NCP is just integrated over the top 100 m (Fig. b). Despite the full water column heterotrophy of the offshore waters, the top 100 m of the CanUS has a positive NCP at every latitude and distance from the coast, i.e., it is, on average, a net source of organic carbon. In contrast, NCP integrated from 100 m depth downward (Fig. c) is negative everywhere, i.e., this part of the water column and the underlying sediments are net heterotrophic. Thus, the negative ∫NCP of the offshore waters (Fig. a) arises from the excess of respiration at depth over the net production in the overlying surface ocean.

This switch between positive and negative NCP in the CanUS happens on average at ∼ 60 m in the nearshore regions, deepening to ∼ 100 m in the offshore region, separating a layer of high net production from a layer of intense net respiration (Fig. a). This depth corresponds very closely to the euphotic zone depth, here defined by the level at which the light intensity at the surface is attenuated to 1 %. Furthermore, this is also just below the depth of the maximum organic carbon concentration (Fig. b). In the upper 100 m, the majority of this organic carbon stems from small detritus and phytoplankton, while below that depth and particularly in the nearshore areas, the large particles dominate. The contribution of zooplankton to the total organic carbon pool is substantial, but never dominant (see Appendix B, Fig. ).

Since there is no substantial accumulation of organic carbon in the long-term mean in any of the reservoirs, this onshore–offshore gradient in ∫NCP requires a considerable amount of organic carbon that is transported from the shelf region into the open subtropical North Atlantic. To understand this complex spatial pattern of autotrophic and heterotrophic activity in the region, we next quantify the lateral and vertical fluxes of organic carbon in the CanUS.

The dominant nature of the offshore transport of organic carbon from the northwestern African shelf becomes clear by inspecting the annual mean and meridionally averaged section of the zonal flux of organic carbon (Fig. a). This transport is negative nearly everywhere, indicating a westward, i.e., offshore transport, with the exception of the very nearshore region, where the narrow upwelling cell recirculates the organic carbon back onshore. This offshore flux spans the entire 2000 km range of distances from the coast in the first 200 m of depth, resulting in a continuous displacement of the organic carbon from the nearshore waters to the open sea. Only in the very nearshore region, does the narrow upwelling cell cause the zonal flux to recirculate the organic carbon onshore.

The intensity of the offshore flux is maximum at the surface and generally decreases with depth, except for the offshore regions, where a secondary maximum of zonal offshore transport occurs at around 100 m. Below that depth, the transport decreases rapidly and tapers off to very low values below 200 m, with the exception of the first 500 km from the coast. While the surface maximum of the offshore flux is mainly driven by the intense mean zonal velocity, the intensification of the flux around 100 m in the offshore waters is strongly influenced by the pattern of the organic carbon concentration (Fig. b).

The lateral meridional flux (Fig. b) shows a complex alternation of northward and southward fluxes, emerging from the integration of the meridional transport across a wide meridional band. Even though this flux is weaker than the zonal flux, this does not imply the absence of substantial alongshore currents within the domain. In fact, many of these currents get averaged out by the meridional integration. Despite this, the intense southward flowing CC is still visible as a negative signature of the mean meridional flux near the coast. Northward fluxes, probably linked to an influx from the organic carbon-rich near-equatorial region, are dominant further offshore.

The vertical advective and mixing (eddy-diffusive) fluxes (Fig. c) are overall much weaker than the vertical sinking flux (Fig. d). The latter, as expected from the fact that we employ constant sinking speeds, has a pattern that directly reflects that of the organic carbon concentration (Fig. b). In contrast, the fact that the vertical advective and mixing fluxes depend on the mean circulation results in a more complex pattern. These fluxes are positive both near the coast in response to the strong upwelling and in the upper first tens of meters where the vertical mixing redistributes the organic carbon against its vertical gradient. Below this shallow layer, subduction and downward mixing are dominant and contribute to the export of organic carbon to depth.

Reflecting the relative contribution of the different pools to the total organic carbon, the fluxes below the first 200 m as well as the vertical sinking flux are dominated by the contribution of the large detritus that reaches deep into the water column, declining in concentration in the offshore direction. In the first 200 m, abundant small detritus, phytoplankton, and to a smaller extent zooplankton shape the organic carbon fluxes up to the farthest boundary of the analysis domain.

The annual mean budget of organic carbon for the upper waters of the whole CanUS highlights the key contribution of the offshore flux to the enhancement of the organic carbon pool and the maintenance of the heterotrophic activity in the open waters (Figs. a and a). In the upper 100 m, corresponding roughly to the euphotic layer, the offshore flux is the dominant lateral flux at all distances, with a magnitude that always exceeds 10 % of the integrated NCP within the box (Fig. a). More specifically, at 100 km from the coast the offshore flux of organic carbon transports as much as 1.6 Tmol C yr-1 (18.7 Tg C yr-1), a quantity that amounts to more than a third of the integrated NCP in the 0–100 km range, i.e., the first coastal box, and to 18 % of the NPP, NPP100km=8.6 Tmol C yr-1.

A good measure for the magnitude and impact of the lateral redistribution of organic carbon is the difference between the organic carbon that is produced locally through NCP and the amount of organic carbon that is exported vertically out of the euphotic zone (here 100 m). In the absence of any lateral redistribution, this difference, termed excess export, i.e., ΔE = Vertical Export - NCP is zero, while in the case of a strong lateral export of the organic matter, ΔE is negative since less carbon is available for the vertical export at depth. Conversely, if a particular region imports a large amount of organic carbon through lateral transport and then exports this carbon to depth, then the excess export ΔE is positive.

The analysis of ΔE as a function of offshore distance reveals that all regions have a positive ΔE, with the exception of the nearshore one, whose ΔE is instead negative (Fig. a). Thus, this supports the notion that the net heterotrophic activity over the whole water column in the offshore direction is fueled by a strong net growth in the very nearshore region of the CanUS. The magnitude of the excess export at depth ΔE ranges between 41 % of NCP in the 100 to 500 km range and 13 % of NCP in the most distant region (Fig. a, orange solid line), accounting for hundreds to thousands of gigamoles of organic carbon per year. This excess export of organic carbon below 100 m is explained by the divergence of the lateral fluxes. In particular, the divergence of the offshore flux (Fig. a, purple solid line) releases an amount of organic carbon that constitutes between 8 and 34 % of the local NCP in each region, with the highest value in the 100 to 500 km offshore range, and this explains the organic carbon accumulation from 62 to 80 % of the excess export at depth in the first 1500 km offshore. In the most distant analysis region (1500 to 2000 km range) the offshore flux divergence drops, resulting in a significant export of organic carbon through the 2000 km offshore boundary and little accumulation; this flux may impact the biological activity even farther in the North Atlantic Gyre.

The alongshore lateral fluxes also positively contribute to the total budget, with a net influx of organic carbon in the euphotic layer (Fig. a, blue solid line). However, the divergence of the alongshore flux exceeds 10 % of the local NCP only in the most distant analysis region and represents therefore a minor contribution to the excess export at depth.

The fate of the vertically exported carbon in the very biologically active 100–200 m depth layer is still strongly influenced by the offshore transport (Fig. b), except for the nearshore, where the upwelling cell recirculates the organic carbon onshore. The offshore flux intensifies away from the coast, reaching its maximum at 1500 km from the coast, where its intensity becomes comparable to that of the top 100 m. Given the negative contribution of NCP to the organic carbon budget in this layer, the magnitude and divergence of the offshore flux at this depth can be compared to the non-respired influx of organic carbon, i.e., the amount of incoming carbon that is available in each box after remineralization. The main sources of organic carbon at these depths are the incoming vertical flux from the euphotic layer and to a very small extent the divergence of the lateral alongshore flux (dotted yellow and blue lines, respectively, in Fig. b). Therefore, the non-respired influx is defined as the sum of the incoming vertical flux, the divergence of the alongshore flux, and the negative NCP.

If we compare the offshore flux at the boundary of the boxes to the non-respired influx in each box, we see that the offshore flux moves a substantial amount of the organic carbon available, reaching 26 % at 2000 km distance from the shore (solid purple line with circles of Fig. b). At the same time, due to its intensification in the direction of the open sea, the offshore flux does not release carbon in the boxes, as confirmed by its negative divergence up to 1500 km offshore (solid purple line with stars in Fig. b and percentages in orange in Fig. b), but the offshore flux traps carbon and transports it even farther toward the open waters. The 100–200 m depth layer of the CanUS is therefore still characterized by a significant offshore transport that moves the organic carbon towards the oligotrophic center of the North Atlantic Gyre, furthering water column heterotrophy there.

As the offshore fluxes are small below 200 m (Fig. a), we omitted them from the plot. In fact, the maximum contribution to the total offshore transport of the water column below 200 m, accounting for a few kilometers of depth, is 12 %, reached at 500 km from the shore. This fraction quickly declines offshore to a minimum of only 0.4 % at 2000 km from the shore. Thus, the vast majority of the transport occurs in the top 200 m of the water column in our model.

Substantial meridional differences in both biological activity and circulation characterize the CanUS and influence the pattern of vertically integrated NCP (see Fig. ) and the implied lateral organic carbon fluxes. In the euphotic layer the pattern of production changes with latitude transitioning from a sharp offshore NCP gradient in the northern CanUS to a wide offshore extent of high NCP in the southern CanUS. These gradients can be explained by the pattern of the nutrient fluxes (see Appendix B, Fig. ). In the northern CanUS, nutrients are in fact mostly provided by coastal upwelling, while the positive signature of the wind stress curl in the southern CanUS favors Ekman pumping of nutrients also offshore (Fig. c). Intense production in the surroundings of Cape Blanc is likely due to the convergence of the alongshore nutrient fluxes (Fig. b), in agreement with and .

Below 100 m, the northern CanUS is characterized by a weak offshore gradient of deep respiration, which, combined with a sharp offshore gradient of production in the layer above, results in an extended net water column heterotrophy in the open waters. In contrast, the southern CanUS is characterized by a widespread vertical correspondence between shallow sources and deep sinks of organic carbon that result in a vertically integrated NCP of nearly zero (∫NCP ∼ 0), with negative values of ∫NCP confined only between the African shelf and the Cabo Verde archipelago. Between these two zonal bands with distinct ∫NCP signatures, the central CanUS located in the surroundings of Cape Blanc (21∘ N) and the whole Cabo Verde frontal zone are hot spots for the respiration of the organic carbon. Here, the region of deep intense remineralization extends farther offshore than the area of intense near-surface productivity, resulting in a vast peak of negative ∫NCP that reaches far into the North Atlantic Gyre. To identify what processes drive the organic carbon redistribution that gives rise to these ∫NCP gradients and quantifies the contribution of the different zonal bands to the total transport, we analyze the spatial patterns of the physical fluxes of organic carbon in detail.

Subregional differences in the organic carbon transport are visible in all of the components of the physical fluxes (Fig. ). Both zonal and meridional fluxes integrated over the top 100 m are clearly influenced by the regional pattern of currents (see also Fig. b) and change sign in the proximity of the Cabo Verde frontal zone, the crucial boundary between the northern anticyclonic and the southern cyclonic circulation. North of the Cabo Verde front, the zonal flux is mostly offshore and intensifies towards Cape Blanc likely due to both the intense coastal mesoscale activity that culminates at 21∘ N with the giant Cape Blanc filament and due to the formation of the Cabo Verde front . South of Cape Blanc, ocean striations appear in the form of alternate onshore and offshore flux bands. The meridional transport converges around Cape Blanc, again with a sharp inversion of sign that reflects the direction of flow of the CC and the MC.

The 100 m horizontal section of the vertical advective transport of organic carbon reflects the signature of the wind stress curl that is negative north of the Cabo Verde front and positive to the south. As a consequence, the highest values of advective export at depth are found in the northern regions of low offshore production, while vertical advective export of organic carbon is not favored in the very productive southern subregion. As for the offshore transport, the vertical advective export in the northern CanUS sector and in the Cabo Verde frontal zone is also likely enhanced in the first few hundreds of kilometers by the abundant coastal filaments that quickly channel and advect down the coastally produced organic carbon. The vertical mixing fluxes of organic carbon across 100 m show that this component is important only in the northern subregion characterized by a much deeper MLD and declines offshore with the decrease in the organic carbon concentration. Sinking fluxes through the 100 m depth are not shown as their pattern is mostly proportional to the 100 m integrated NCP (Fig. b) with high export in regions of high production and do not add substantial complexity to the discussion. However, it is worth remarking that their intensity is about 1 order of magnitude higher than that of the vertical advective and vertical mixing fluxes, reaching very intense peaks (∼ -40 mol C m-2 day-1) in the surroundings of the most productive region of Cape Blanc and high values of about -10 mol C m-2 day-1 in the Mauritanian–Senegalese subregion. The southern sector of the CanUS is therefore dominated by intense sinking fluxes of organic carbon, while both non-sinking vertical components of the export have a very limited role in the organic carbon export in the region south of Cape Blanc.

The analysis of the spatial pattern of the organic carbon fluxes also takes into account the special role of the central zonal band located between the Canary Islands archipelago and the Cabo Verde islands, characterized by the most intense biological and physical fluxes. This central zonal band is characterized by the strongest heterotrophic activity offshore, a persistent and intense offshore transport, a convergence of the lateral alongshore fluxes in the shelf, strong vertical advective export at depth, intense vertical mixing, and a peak of the sinking flux. To highlight how differently the physical fluxes impact the organic carbon budget at different latitudes and to study the interaction between significant zonal bands in the CanUS, we divided the region into southern, central, and northern subregions (defined as in Fig. b) and carried out a subregional box budget analysis.

Among the CanUS subregions, the southern subregion (pink line in Fig. ) is characterized by very high levels of NCP in the euphotic layer also in the open waters (Fig. a), accompanied by a relatively low impact of the physical fluxes of carbon on the local budget. In fact, despite both the offshore transport and the vertical advective + mixing export below 100 m having high intensities in absolute terms, their divergences are low when compared to the high values of NCP in each box and therefore they do not have a substantial impact on biology (Fig. b and d). On the one side, the little accumulation of organic carbon due to the offshore flux explains the large portion of biologically neutral water column in this region (∫NCP ∼ 0, see Fig. a). On the other side, even though the sinking fluxes (Fig. e) still export substantial amounts of carbon in this subregion, the low efficiency of the advective + mixing export points to a potentially limited capacity of this subregion to export dissolved and suspended material at depth (not modeled in our study). The comparatively higher impact of the alongshore flux is explained by the strong coupling of the southern subregion with the equatorial carbon-rich area that allows a net influx of organic carbon through the southern boundary.

The impact of the lateral offshore transport in the northern subregion (blue line in Fig. ) is particularly high in the first 500 km offshore (Fig. b). This is the result of a combination of both strong export fluxes on the shelf and the fast decline of NCP in the offshore direction, with a consequently high ratio of the offshore flux divergence to NCP and an important influence of the flux on the local budget. The intense mesoscale activity in this northern subregion, especially in the form of persistent filaments that detach from the coast and quickly channel water and tracers offshore typically for some hundreds of kilometers, has an important role in this intense nearshore export. The northern subregion is also the most efficient in the combined vertical advective and mixing export up to 1000 km offshore as a consequence of the deep MLD and the abundant mesoscale coastal filaments, with a consequently high capacity to export light organic carbon species below the euphotic layer. However, both the lateral and the vertical export efficiency decline quickly offshore in the northern subregion. This decline is likely due to the low organic carbon concentration of the offshore waters, to the limited offshore extension of the filaments, and to the incoming flux of the Azores Current counteracting the offshore transport at the northern edge of the domain.

As anticipated, the most active area in terms of the organic carbon transport and export is the central CanUS subregion (green line in Fig. ), which includes the Cabo Verde frontal zone. The central subregion collects the lateral alongshore fluxes from the northern and southern subregions in the nearshore area, with a net increase of the carbon availability in the first coastal box (0–100 km range) of more than one-third of the local NCP (Fig. c). The southward-flowing CC and northward-flowing MC that converge in this zonal band are the main contributors to this organic carbon influx on the coast around Cape Blanc. The carbon collected on the shelf is likely exported offshore together with the locally produced carbon by a very intense zonal flux also characterized by a large divergence that exceeds the values of the northern subregion in the offshore waters (Fig. b). The carbon accumulation due to the divergence of the offshore transport over the central zonal band is on average as high as 57 % of the local NCP, reaching peaks of more than 70 % of NCP in the 500–1500 km range and still accounting for 37 % of NCP in the farthest offshore box (1500–2000 km range). Both the mean circulation characterized by the westward-flowing currents along the Cabo Verde front and the mesoscale activity in the form of the giant Cape Blanc filament are expected to contribute to this intense offshore transport. As a consequence of this increased carbon availability, the central subregion allows very high values of the total vertical export of carbon below the euphotic layer (advective + mixing + sinking transport), which almost doubles the local production. Among these vertical components, the advective + mixing vertical export becomes particularly important in the offshore waters. The alongshore convergence of the organic carbon on the shelf and the high lateral mobility of the organic carbon in the offshore direction of the central subregion not only explain the peak of net water column heterotrophy in the offshore waters around Cape Blanc (see Fig. a) but they also characterize the Cabo Verde frontal zone as a key region of the CanUS for the collection and export of the coastally produced organic carbon far into the North Atlantic Gyre.

Further insights into the zonal differences of the organic carbon transport below the euphotic layer are given by the zonally averaged mean vertical profiles of the offshore and vertical advective+mixing fluxes for each subregion (Fig. ). The southern subregion is confirmed to be the least efficient in the offshore transport and vertical advective + mixing export also at depth, despite an intensification of the offshore flux below the surface in the farthest region of analysis. This offshore intensification is however probably connected to the intersection of our southern zonal band with the Cabo Verde frontal region, which crosses the northern boundary of this subregion at about 1000 km offshore. The vertical advective + mixing export in the southern subregion is also remarkably different from the other subregions: not only the shallow MLD limits vertical mixing but the positive signature of the wind stress curl also favors upwelling in this subregion. The mean vertical profiles of the zonal flux for the northern subregion clearly shows a weak onshore flux in the offshore waters below the surface that confirms the important influence of the incoming Azores Current in the limitation of the offshore transport away from the coast. The deep extension of both the offshore flux and of the vertical advective+mixing downwelling in the first few kilometers from the coast of the northern subregion again suggests the link with the recurrent coastal filaments that characterize this sector of the CanUS and are known to enhance the fluxes through a depth of several hundreds of meters . The central subregion also presents the most intense and persistent offshore flux at depth that reaches 2000 km from the coast in the whole first 200 m layer, showing the important role of this subregion in the offshore redistribution of the organic carbon. Both offshore and advective + mixing fluxes extend particularly deep into the water column in the first several hundreds of kilometers from the coast in the central subregion, likely due to both the intense mean circulation and the powerful giant Cape Blanc filament renowned for its remarkable offshore extension . In both the northern and central subregions the offshore gradient of the zonal flux is the cause of the important accumulation of organic carbon that allows an enhanced respiration at depth.

The results of our subregional analysis show how physical forcing and mean and mesoscale circulation drive the lateral and vertical redistribution of the organic carbon in the CanUS, giving rise to a persistent offshore transport of organic carbon that shapes the ∫NCP pattern and can reach as far as 2000 km into the North Atlantic Gyre. Further insights into the special role of mesoscale activity in the lateral redistribution of organic carbon in the CanUS and a quantification of this component of the transport will be provided in detail in a dedicated publication.

Our results highlight the importance of the lateral transport of organic carbon from coastal regions of intense production to the oligotrophic open waters and its key role in fueling the offshore heterotrophic activity. We thus confirm the predictions of several in situ observations and estimates from multiple independent local surveys . Our modeled fraction of the coastal production of organic carbon that is transported offshore beyond 100 km amounts to 18 % of NPP (36 % of NCP) and lies in the range of previous estimates . Our results also reveal that the offshore transport extends far into the open waters, having important consequences for biological activity: even at offshore distances of 1500 km and more, the additional carbon supplied by the offshore transport can still be as much as 37 % of the local NCP. The offshore transport is particularly relevant below the euphotic layer, especially in the 100 to 200 m layer, where the lateral transport of the organic carbon can extend even farther into the North Atlantic Gyre, i.e., beyond 2000 km offshore.

This long-range offshore transport of organic matter involves two possible, not mutually exclusive, pathways. In the first direct pathway, the offshore transport is sustained entirely by the excess organic carbon produced on the African shelf, which then gets advected offshore without the addition of new, i.e., locally formed, organic carbon along the way. In the second recycling pathway, the offshore flux of organic carbon is sustained by the production of organic carbon that occurs along the way to the open sea: new and regenerated production replaces the incoming organic carbon that gets remineralized and sinks to depth, creating a continuous recycling loop of the organic carbon that is advected offshore. The nutrients required to fuel this production stem from both the upwelling along the coast and subsequent offshore transport (leading to new production) and from the local remineralization of the incoming organic matter (leading to regenerated production). In contrast, the vertical pumping of the nutrients from the deeper waters is thought to be a comparatively minor contribution, with the exception of the southern CanUS, where the wind stress curl is positive, leading to upwelling. A simple analysis of the residence times of the modeled organic carbon pools favors a dominance of the second recycling pathway. In fact, a small detritus particle resides, on average, only about 200 days in the top 200 m depth layer due to its sinking speed, if we disregard downwelling and other loss terms such as remineralization. Given typical lateral transport velocities of less than 0.05 m s-1 when averaged over the top 200 m, this particle would be able to travel only about 800 km offshore at best. This distance is only 80 km for large detritus and 1600 km in the case of phytoplankton. With the zooplankton not sinking in our model, this pool could travel farther, but its contribution to the offshore fluxes is very modest compared to those of phytoplankton and of the abundant small detritus. If we further include the loss terms (coagulation and remineralization) of these organic carbon pools, which substantially reduce their lifetimes in the upper 200 m, it is highly unlikely for any coastally produced organic carbon to reach as far as 2000 km from the coast. Furthermore, the inorganic nutrient fluxes (see Appendix B, Fig. ) are of sufficient magnitude to refuel new growth of organic matter to replace that part that is lost by sinking. Further analyses including Lagrangian experiments are necessary to gain a quantitative understanding of the succession of transformations happening along the way to the open waters, but it is clear that the most likely scenario is one where the organic carbon pool is recycled several times over along its many-months-long journey from the coast to the offshore.

The meridional alongshore transport also contributes to the redistribution of the organic carbon, especially on the shelf, where the alongshore coastal currents are the strongest. In line with the results of and , we find that the area around Cape Blanc, corresponding to the region of convergence of the coastal flows and formation of the Cabo Verde frontal zone, is a key region of the CanUS. The relevance of the central Cabo Verde frontal zone in the CanUS was discussed in and in terms of chlorophyll and nutrient convergence on the shelf and of their subsequent offshore advection, visible as a persistent bloom in the region offshore of Cape Blanc. Here we confirm and strengthen these results, affirming that this sector of the CanUS has a central role in the collection and subsequent transport offshore of the organic carbon produced over wide areas of the shelf. This offshore flux of carbon against the gradient of productivity extends far away from the coast and feeds the heterotrophic activity of the deep open waters for at least 2000 km offshore, generating a long tail of net heterotrophy. Deep offshore transport and subduction in this region are likely enhanced by the persistent Cape Blanc filament, which is known from local surveys for being able to transport an estimated 50 % of the coastally produced carbon both in the surface and at depth, extending several hundreds of kilometers offshore .

The partition of the CanUS into a northern anticyclonic and a southern cyclonic circulation regime and the differences in mesoscale activity and wind stress curl are well reflected by the differences in the transport and cycling of the organic carbon north and south of the Cabo Verde frontal zone. The portion of CanUS located north of the Cabo Verde front can be regarded as the very eastern edge of the North Atlantic Gyre . Several studies highlight the way in which the abundant coastal filaments of this northern CanUS sector substantially enhance the offshore transport and the downwelling of organic carbon in the first hundreds of kilometers from the coast . These structures together with a strong mean offshore flow explain the intense zonal transport and the vertical downwelling in the shelf region of the northern CanUS and the sharp decline of these two fluxes in the open waters. The range of influence of the offshore transport in this zonal band is further enhanced by the eddies spun off by the filaments , while the negative signature of the wind stress curl maintains the vertical downwelling in the offshore. The strength of the vertical transport by downwelling and mixing suggests that the northern subregion is potentially efficient in exporting the dissolved, suspended, and slowly settling material to depth, i.e., the organic carbon species that are difficult if not impossible to measure with sediment traps, but may still constitute a key component for the closure of the organic carbon budget at depth .

The portion of CanUS located south of the Cabo Verde front is instead mostly coupled to the southern equatorial circulation and to a small extent to the North Atlantic Gyre. Here, the net water column biological activity shows a dominantly neutral water column and little water column heterotrophy, the latter mostly confined to a region between the African coast and the Cabo Verde archipelago. The intense near-surface production, the much smaller offshore gradient in productivity , and to some extent the transitory nature of the filaments that form on the shelf at these latitudes result in a small impact of the organic carbon lateral fluxes. Sinking dominates the vertical export at these latitudes, while mixing and vertical advection are impeded by a shallow MLD and the positive wind stress curl leading to upwelling.

Overall, in a large part of the CanUS the lateral redistribution of organic carbon from the shelf to the open waters results in a very substantial lateral decoupling of the region of remineralization from the region of production. This contrasts sharply with the representation of the organic carbon pump as a pure vertical process and highlights the fundamental importance of the lateral transport of organic carbon for the maintenance of the biological activity.

However, despite the very large lateral input of organic carbon in the upper 100 m across much of the offshore region of the CanUS, our model does not show evidence for net heterotrophic conditions in the near-surface waters of these regions. Thus, the shallow open sea is a net source of organic carbon everywhere for the deeper layers. This is the case irrespective of whether the vertical integration is performed over the mean euphotic layer depth (100 m), over the local euphotic layer depth, or over the local MLD (generally shallower than the euphotic layer). Thus, our model provides strong support for the net autotrophic surface ocean hypothesis . The spatial pattern of modeled near-surface autotrophy (Fig. b) agrees with the calculated global distribution of NCP Fig. 1 once the net heterotrophic regions are substituted for by weakly autotrophic low-productivity waters. The depth at which vertically integrated sinks and sources from top to bottom compensate for each other in the model is located at more than 200 m depth and can be deeper than 1000 m almost everywhere in regions of nearly neutral water column, confirming the importance of the respiration in deep waters .

Both the mean Ekman transport and the turbulent mesoscale activity contribute to the total lateral fluxes of organic carbon connecting coastal sources to deep offshore sinks. These two processes also concur in determining the vertical downwelling and mixing that increase the organic carbon transport to depth. The magnitude of the relative contribution of these two terms to the organic carbon fluxes and their different roles in fueling the heterotrophic activity offshore must be detangled through further analysis.

We discuss here how our quantification of the offshore transport of organic carbon may be affected by the shortcomings of our model. The first set of shortcomings involves our modeling of the organic matter pool, especially our lack of consideration of an explicitly modeled DOC pool and the representation of POC at depth. The second set of shortcomings involves a few biases in our modeled physical and biogeochemical fields. We discuss the potential impact of these shortcomings in turn.

Regarding DOC, the pool that matters is that of semi-labile DOC as it has a lifetime of beyond a few days, implying that it can be transported substantial distances before it gets remineralized. As a result, it has the potential to enhance our modeled lateral export of organic carbon. This is especially the case since DOC is readily produced in the surface ocean and also contributes substantially to the export of organic matter from the near-surface ocean , in particular in subtropical regions such as the North Atlantic Gyre . Even though DOC is not explicitly modeled, the small detritus, with its sinking speed of wSD=1 m day-1, essentially represents a suspended POC pool with some similarity to semi-labile DOC, particularly regarding its susceptibility to lateral transport. However, differing from DOC, the small detritus coagulates to large detritus, resulting in a shorter lifetime than DOC in the surface ocean. At the same time, the rate of production of DOC is likely smaller than that of the small detritus. Together, this implies that the small detrital pool likely behaves in a comparable manner to the semi-labile DOC. Thus, we would argue that the impact of our lack of consideration of DOC in our model is smaller than possibly inferred at first.

In order to more quantitatively explore the potential impacts of our lack of explicit consideration of DOC, we ran a sensitivity study in which we set the vertical sinking of the small detritus, wSD, to zero and reduce the coagulation timescale for small detritus to 40 % of its baseline value. No adjustments were made to the parameterization of the large detritus. This sensitivity study needs to be considered as an extreme scenario, i.e., it is meant to explore the potential contribution of DOC rather than an attempt to quantify it in detail. We spun up the model with the new biological parameters from year 24 of the baseline run (6 years of spinup) and used years 30–35 for the analysis, as for the baseline run. The results show, as expected, an intensification of the lateral fluxes of organic carbon in the euphotic layer. The standing stock of suspended POC increases about twofold, largely due to its longer average lifetime in the surface ocean, stimulating the local recycling of organic matter. This increases both primary production and heterotrophic activity in the near-surface layer, leaving the ∫NCP pattern basically unchanged and preserving the net autotrophy of the near-surface waters. In fact, even though the lateral transport of small detritus is much larger in this sensitivity study and reaching farther out into the open North Atlantic, the net horizontal divergence of the lateral flux remains roughly the same. Thus, for the key question at hand, i.e., whether the offshore transport can fuel net heterotrophic conditions in the offshore regions of the CanUS, the answer essentially remains the same.

Another potential caveat of our study regards the lateral redistribution of the organic carbon at depths larger than the first few hundred meters. On average, our modeled offshore transport below 200 m is very small and never larger than 12 % of the total transport. However, model limitations in the representation of the offshore transport below this depth should be discussed, taking into account three potential and partially contrasting caveats. First, the model does not include the process of sediment resuspension, therefore impeding the formation of high-POC-concentration spikes near the shelves and limiting the bottom transport along the slopes . Second, and with a similar effect, the dynamics of our particulate pools only allow the aggregation of small particles into bigger and heavier ones, while it does not consider disaggregation of heavy particles into lighter ones as a consequence of degradation or partial grazing . This results in a one-way path to fast sinking that cannot be reversed. Due to these two factors that preclude the existence of deep local maxima of suspended POC, our study may underestimate the lateral transport of organic carbon at depth. Third, sinking velocities in our model are fixed at every depth to moderate values (maximum of 10 m day-1 for large detritus), while sinking velocities have been observed to be able to reach relatively high values, increasing by roughly 1 order of magnitude between the mesopelagic and bathypelagic regions , with a consequent fast vertical export at depth of the particles by sinking. Heavy particles at depths below 1000 m have been shown to have mean sinking velocities of 100–300 m day-1 and to often be accompanied by a pool of slow sinking material with mean sinking velocities of 1 to a few meters per day, resulting in a bimodal distribution of the sinking speeds .

Viewing these three caveats together, we have two missing processes that would cause our model to represent a lower-bound estimate and one process that would cause the correct offshore transport to be smaller. We cannot assess the implication of this finding in full, but submit that at least with regard to the offshore transport in the upper waters, i.e., upper 200 m, our model is likely in the right range, perhaps on the lower side. We have much less confidence in the offshore transport below 200 m, where the observed concentrations of organic carbon can be quite high, although the offshore velocities are substantially smaller.

We also need to assess the potential impact of the physical and biogeochemical biases that we diagnosed in the evaluation section. In the northern CanUS our model overestimates the MLD depth; however, our modeled MLD has the same meridional gradient as observed, with an extremely shallow mixed layer in the region located south of the Cabo Verde front and a deeper mixed layer in the north. This suggests that, even though we may potentially overestimate vertical mixing in the northern CanUS, this subregion would still be expected to be the only one in which this process is relevant. In the southern CanUS, our model shows a circulation that is weaker than observed and chlorophyll and NPP maximums that are deeper than observed, which may lead to an underestimation of the lateral transport and therefore of the net heterotrophy of the water column. Both a shoaling of the biological production towards the surface, characterized by more intense currents, and an intensification of the circulation can result in the strengthening of the lateral zonal and meridional organic carbon fluxes. However, an increase in the offshore zonal fluxes in the southern subregion could favor a more heterotrophic water column only if accompanied by an increase in the divergence of the flux, resulting in a substantial accumulation of organic carbon compared to the local production. In the meridional direction, an intensification of the alongshore MC may instead increase the influx of organic carbon from the south into the Cabo Verde frontal zone, fueling even further the deep respiration in the already strongly heterotrophic central CanUS.

To summarize, we believe that the caveats discussed above do not substantially affect our main findings, while they possibly strengthen our conclusion regarding the importance and long-range nature of the offshore transport of organic carbon. In fact our model may, if anything, underestimate the total lateral transport of organic carbon both at the surface and at depth. For this reason we believe that it is of fundamental importance to take into account the three-dimensionality of the marine organic carbon cycle and the essential role of the productive coastal ocean in the global biogeochemical cycles.