Drivers and impact of the seasonal variability of the organic carbon offshore transport in the Canary Upwelling System

. The Canary Upwelling System (CanUS) is a productive coastal region characterized by strong seasonality and an intense offshore transport of organic carbon (C org ) to the adjacent oligotrophic offshore waters. There, the respiration of this C org substantially modiﬁes net community production (NCP). While this transport and the resulting coupling of the biogeochemistry between the coastal and open ocean has been well studied in the annual mean, the temporal variability, and especially its seasonality has not yet been investigated. Here, we ﬁll this gap, and determine the seasonal variability of the offshore transport 5 of C org , its mesoscale component, latitudinal differences, and the underlying physical and biological drivers. To this end, we employ the Regional Ocean Modeling System (ROMS) coupled to a nutrient, phytoplankton, zooplankton, and detritus (NPZD) ecosystem model. Our results reveal the importance of the mesoscale ﬂuxes and of the upwelling processes (coastal upwelling and Ekman pumping) in modulating the seasonal variation of the offshore C org transport. We ﬁnd that the region surrounding Cape Blanc (21 °N) hosts the most intense C org offshore ﬂux in every season, linked to the persistent, and far reaching Cape 10 Blanc ﬁlament (cid:58)(cid:58)(cid:58) and (cid:58)(cid:58)(cid:58) its interaction with the Verde front. Coastal upwelling ﬁlaments dominate the seasonality of the total offshore ﬂux up to 100 km from the coast, contributing in every season season at least 80 % to the total ﬂux. The seasonality of the upwelling modulates the offshore C org seasonality hundreds of km from the CanUS coast via lateral redistribution of nearshore production. North of 24.5 °N, the sharp summer-fall peak of coastal upwelling results in an export of more than 30 % of the coastal C org at the 100 km offshore due to a combination of intensiﬁed nearshore production and offshore ﬂuxes. To the 15 south, the less pronounced upwelling seasonality regulates an overall larger, but farther-reaching and less seasonally varying lateral ﬂux, which exports between 60 and 90 % of the coastal production more than 100 km offshore. Overall, we show that the temporal variability of nearshore processes impacts (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) modulates the variability of C org and NCP hundreds of km offshore from the CanUS coast via the offshore transport of the nearshore production. of (cid:58)(cid:58) of (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) quantifying (cid:58)(cid:58)(cid:58) the C org (cid:58)(cid:58)(cid:58)(cid:58)(cid:58) lateral (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)

While the tropical circulation found south of Cape Blanc is relatively weak with small and highly seasonal coastal filaments (Peña-Izquierdo et al., 2012;Menna et al., 2016), high upwelled nutrient concentrations and a widespread positive windstress :::: wind : ::::: stress curl signature fuel high productivity both at the coast and offshore (Arístegui et al., 2006).
Owing to the simultaneous presence of cross-shore mass fluxes and of tracer gradients, EBUS laterally export large amounts of organic and inorganic material properties to the adjacent oligotrophic waters (Nagai et al., 2015;Lachkar and Gruber, 70 2011; Amos et al., 2019). Among these fluxes, the offshore transport of coastally-produced organic carbon (C org ) is especially important, as this might help to explain the purported net heterotrophy of some oligotrophic open waters that are located adjacent to some of these EBUS (Burd et al., 2010;Arístegui et al., 2003;Pelegrí et al., 2005). For example, the CanUS was shown to transport more than one third of its coastal net community production (NCP) toward the offshore, reaching as far as 2000 km from the coast (Lovecchio et al., 2017). This enhanced supply of C org to the North Atlantic Subtropical Gyre 75 increases respiration there, pushing vertically integrated net community production to become negative. Over a year, narrow coastal filaments are, on average, responsible for about 80 % of this offshore flux at 100 km from the coast, while coastally generated mesoscale eddies dominate this transport further offshore (Lovecchio et al., 2018).
In contrast to the relatively well studied annual mean offshore transport, substantially less is known about its seasonal nature ::: and :::: how :::: this ::::::: impacts :::::::: biological ::::::: activity :: in :::: the :::: open :::::: waters. Given the intensity of this transport, it is very likely that the 80 strong seasonal variability of coastal upwelling and production is readily transferred towards the open ocean, albeit with some temporal delay. There, and especially in the nutrient deprived subtropical gyres, the seasonality of the lateral supply may be rather critical, especially when this lateral supply occurs during the summer, when the nutrient deficiency is most acute. In this paper, we quantify the seasonal variability of the offshore flux of C org in the CanUS. In particular, we analyze the relative role of the mean and mesoscale circulation in each season, we study the role of upwelling seasonality as physical drivers of the 85 offshore fluxes and we discuss the impact of these offshore flux changes on the ecosystem of the open waters.

Model setup and output
Our analysis of the seasonality of the offshore transport in the CanUS is based on the results from a CanUS model simulation undertaken and described by Lovecchio et al. (2018). This simulation used the UCLA-ETH version of the Regional Ocean 90 Modeling System (ROMS) (Shchepetkin and McWilliams, 2005) coupled with a Nutrient Phytoplankton Zooplankton Detritus (NPZD) ecosystem model (Gruber et al., 2006). This coupled model was run with monthly climatological forcing derived from ERA-Interim (Dee et al., 2011) on an Atlantic telescopic grid. This grid combines a full Atlantic basin perspective with a mesoscale-resolving resolution in the region of study, achieved through a strong grid refinement towards the north-western African coast. We refer to Lovecchio et al. (2017) and Lovecchio et al. (2018) for further details on the model, on its forcing, 95 and for providing a careful evaluation against a large range of observational constraints. The simulation we are analyzing was run for 53 years, of which we use the last 24 years for our analyses. We use data in the form of 2-day means. On top of the necessary state variables, i.e., temperature, sea surface height (SSH), velocity fields and C org concentration, the model was setup to output also net community production (NCP) calculated at run time for each grid cell from total biological production minus all respiration fluxes. The 2D field of the lateral fluxes of C org was calculated 100 from the model output as the product of the horizontal velocities and the tracer concentration. :::::::::: Throughout ::: this :::::::::: manuscript, ::: we :::: focus ::: on ::: the ::: 100 ::: m ::::: depth, :::::: which :::::::::: corresponds :: to ::: the ::::: mean ::::: depth :: of ::: the ::::::: euphotic ::::: layer :::: (i.e. ::: the ::::::::: productive ::::: layer) :: in ::: the :::::: region :: of ::::: study.
In our employed NPZD model, C org corresponds to the sum of non-sinking zooplankton, sinking phytoplankton, and a small detritus pool that sinks slowly and a large detritus pool that sinks fast. No dissolved organic pool (DOC) is included in the 105 model : ; ::: we :::::: address :::: this :::::::::: shortcoming ::: in ::: the ::::::::: Discussion :::::::::: (subsection ::: 5.2 ::::::: "Model :::::::::: limitations" : ). For the purpose of our model evaluation and analysis, we define the four seasons as follows: spring is the mean of March, April, and May (MAM), summer is the mean of June, July, and August (JJA), fall is the mean of September, October, and November (SON), and winter is the mean of December, January, and February (DJF). Since one year of model output consists in 360 days, we define the unit "season" (seas) as 90 days of simulation. While we analyze the entire CanUS system from 17 • N and 32 • N, we focus our study on the central and northern subregions (CSR and NSR, respectively), with the border located at 24.5 • N. The central subregion thus extends from 17 • N to 24.5 • N] and the northern subregion from 24.5 • N to 32 • N] (Figure 1). This correspond :::::::::: corresponds to the two homonym subregions studied in Lovecchio et al. (2018). Our main motivation for restricting our analyses on the CSR and NSR is that the model performs 115 well in these two regions, especially compared to the southern CanUS subregion (<17 • N) where the model biases tend to be much larger (Lovecchio et al., 2017(Lovecchio et al., , 2018.
We study tracer concentrations and fluxes from the Northwest African coast out to an offshore distance of 2000 km, corresponding to the middle of the North Atlantic Gyre. We divide the offshore region of the upwelling system in five bands having the following boundaries defined as isolines of offshore distance from the

Upwelling Estimation
To diagnose the potential drivers of the flux variability, we estimate the magnitudes of the offshore coastal transport (U ) and of 125 Ekman pumping (w E ) from the forcing. The offshore coastal transport is obtained from the "classical" Ekman transport: where τ as is the alongshore wind stress, ρ is the density of sea water, and f is the Coriolis parameter. U (m 2 /s or m 3 /s * m of coast) is calculated as result of the mean τ as in the first 100km from the coast in order to be consistent with the calculated offshore fluxes. Then, it is integrated along the coast in order to obtain the vertical transport. U units are m 3 /s or Sverdrup 130 (Sv) (1 Sv = 10 6 m 3 /s). Ekman pumping is estimated from: where curl(τ ) is the wind stress curl, ρ is the density of sea water, and f is the Coriolis parameter. w E (m/s) is integrated offshore in the first 100km (to obtain m 2 /s or m 3 /s * m of coast) and along the coast in order to obtain the vertical transport (w E in m 3 /s or Sv). Following Messié et al. (2009), all negative values are set to 0 before integrating w E (m/s). This is done 135 to reflect an asymmetry in the transport of nutrients. When the Ekman pumping is positive (upward), nutrients are brought from the thermocline to the surface and stimulate biological production. In contrast, when the Ekman pumping is negative (downward), essentially no nutrients are transported downwards, since most of them have already been consumed.

Eddy and filament budgets and fluxes
The contribution of mesoscale eddies and upwelling filaments to the C org budget and fluxes is calculated using the same 140 structure identification algorithms as those employed in Lovecchio et al. (2018). Cyclonic eddies (CE) and anticyclonic eddies (AE) are separately found using the SSH-based identification algorithm developed by Faghmous et al. (2015). This algorithm retrieves the exact contour of each eddy at each time step according to the shape of SSH. At each time step, the CE and AE masks are defined as 2D fields of zeros and ones, with ones corresponding to the surfaces of the retrieved eddies on the grid. Upwelling filaments are found through the use of the sea surface temperature (SST) based identification algorithm 145 fully described and evaluated in the supplement to Lovecchio et al. (2018). In analogy with the eddy identification routine, this algorithm builds a zero-one filament mask for each time step, with the ones identifying the filaments. The upwelling filament identification algorithm reads at each time step the locations of the previously identified eddies (both CE and AE) and excludes their surfaces from the filament mask. This makes sure that the filament and eddy masks never overlap. At each time step, the area that is not covered by either the eddy field or the filament field is defined as non-filament-non-eddy 150 (NF-NE :::::::::::::::::: non-eddy-non-filament ::::::: (NE-NF) field.
Although the algorithms identify the structures at the surface only, we assume that the eddies and filaments have a vertical prismatic structure, i.e., they occupy the same i,j grid points at each depth. We consider this a good approximation as we only focus on fluxes and budgets in the euphotic layer (defined as the first 100 m depth). Fluxes and budgets for each type of structure are calculated by multiplying at each time step the eddy and filament masks by the 2D fields of concentrations or 155 fluxes integrated vertically over the euphotic layer depth.

Evaluation
We summarize here the main relevant findings of the model evaluation for the CanUS in the region [17 • N, 32 • N]. We refer the reader to Lovecchio et al. (2017) and Lovecchio et al. (2018) for additional details.
In the annual mean, the model-observation misfits (biases) of SST and sea surface salinity (SSS) are smaller than 0.75 160 • C and 0.2, respectively. The regional pattern of currents including the alongshore flowing Canary Current and Mauritanitan :::::::::: Mauritanian Current and the Cape Verde front are well reproduced by the model. The regional pattern and offshore gradient of net primary production (NPP) as well as surface POC (S-POC :::::::: particulate ::::::: organic :::::: carbon ::::: (POC) correlate well with satellite derived estimates (correlation > 0.7) even though the model is biased slightly low at latitudes < 25 • N.
The model represents the observed seasonal variability of currents and biological activity in the region of study reasonably 165 well (Lovecchio et al., 2017). Modeled horizontal velocities in each season reproduce in both magnitude and pattern those obtained from drifter data (Lumpkin and Johnson, 2013), with an intensification of the Canary Current in summer and a more intense Mauritanitan :::::::::: Mauritanian Current circulation in fall and winter. The Ekman offshore transport is maximum in summer north of Cape Blanc (21 • N) and in winter-spring south of it. Biological activity also follows such a seasonal pattern, with coastal NPP showing a peak in summer and spring north and south of 21 • N, respectively. As expected, between 21 • N and 170 25 • N, in the proximity of Cape Blanc, coastal NPP remains quite high in all seasons compared to the surrounding latitudes.
The absolute value of NPP is biased low when compared to satellite derived products such as SeaWiFS VGPM and CbPM (Behrenfeld and Falkowski, 1997;Westberry et al., 2008), reaching respectively 1/3 and 2/3 of their magnitude. However, its pattern correlates well with SeaWiFS VGPM (0.8) and CbPM (0.7) in all seasons, giving us confidence that the relative impact of the C org fluxes on the offshore biological activity is well represented in the model. Absolute C org fluxes, however, 175 may be biased low. Further discussion of these strengths and limitations are provided in the model evaluation and discussion of Lovecchio et al. (2017).
Both the modeled turbulent kinetic energy and the standard deviation of SSH are especially close in both pattern and magnitude to the ones derived from the AVISO satellite products (Maheu et al., 2014), confirming that the model provides a good representation of the highly variable and small scale flow in the region (Lovecchio et al., 2018). In the region of study, differ-180 ences between the modeled and the satellite-derived turbulent kinetic energy consist mostly in higher modeled values in the nearshore, likely due to small scale coastal filaments that are resolved by the model but are not captured by the lower resolution satellite data. The regional pattern of large eddy (> 50 km diameter) density retrieved by the identification algorithm (Faghmous et al., 2015) in model and satellite data are also especially similar in both offshore gradient and absolute value, with peaks in the eddy density in the proximity of the coast and of the Canary archipelago. The portion of time occupied by filaments for 185 each grid point according to our identification algorithm is highest in correspondence of capes that are known to be associated to recurrent upwelling filaments. We therefore expect our filament field to reproduce quite well the known regional pattern of these structures. In fall and winter, the C org offshore flux has a more limited extent at all latitudes, with the CSR maintaining the farthest 195 reaching offshore flux. The intensity of the offshore transport close to the coast remains high in fall. Winter shows the weakest lateral C org redistribution also close to the coast at all latitudes. This reduced signal is a result of a combination of weaker currents and lower near-surface C org concentrations, the latter being especially small offshore in the NSR (see Appendix: Figure B1).
The filament transport has a strong seasonal character. In the nearshore, the filament transport is maximum in summer and very intense in fall. In these two seasons, this flux component also shows its maximum offshore extension at every latitude 215 and especially in the CSR around Cape Blanc. Spring sees a smoothing of the zonal gradients in the filament flux, with a more homogeneous intensity at all latitudes, especially in the nearshore. In winter, the filament transport reaches its minimum intensity, barely exceeding that of the eddy flux. :::: with : a ::::: sharp :::: drop :: in ::: the :::::::: nearshore ::::::::: compared :: to ::: the :::: other ::::::: seasons The total (cyclonic + anticyclonic) eddy contribution to the C org offshore flux has a very similar intensity in all seasons, with only a . :: A moderate intensification in the nearshore ::: eddy :::::::: transport :: is :::::: visible in summer and fall in the NSR and in spring 220 and summer in the CSR. In terms of the offshore extension, the eddy flux reaches its maximum in spring, when also the C org distribution reaches its maximum value in the offshore waters.
Across the CanUS as a whole, the relative contributions of eddies and filaments remain similar in all seasons ( Figure 4).
The filament flux is always the dominant flux in the nearshore, often exceeding the total offshore transport between 100 km 8 Figure 3. Seasonality of the eddy (cyclonic + anticyclonic) and filament offshore transport of Corg. The first column represents the filament transport. The second column represents the eddy transport. The flux is integrated over the top 100 m depth. Subpanels: (a) and ( and 200 km offshore. This is possible when the non-filament-non-eddy :::::::::::::::::: non-eddy-non-filament : flux is directed onshore and 225 therefore has a negative contribution to the total offshore flux. The maximum relative contribution to the total flux by filaments both in terms of maximum flux share and in terms of offshore extension is found in fall, when both eddy and NF-NE fluxes are weaker. In fact, even though filaments are intense in summer, this season is also characterized by a widespread peak of the upwelling, triggered by the mean (largely NE-NF) offshore Ekman transport. : , ::::: which :::::: clearly :::::: reaches ::: its :::::::: maximum :::::::: intensity :: in ::: this :::::: season. :
Hovmöeller diagrams allow us to better assess the propagative nature of the offshore transport and quantify the time it takes for the seasonal C org variations to reach and therefore impact the open waters ( Figure 5). As expected, nearshore C org concentrations peak in summer in the NSR and in spring in the CSR, while the widespread spring offshore maximum in C org in at these latitudes (Chelton et al., 2011). These slow propagating signals are likely responsible for a lag in the response of the offshore biological activity to the nearshore seasonal dynamics of upwelling and production.

Quantification of the seasonal variability of the offshore flux by subregion and offshore distance
Our analysis highlights important differences between NSR and CSR in terms of the seasonal and spatial variability of the 250 offshore flux of C org , as well as in terms of the magnitude of the contribution of each structure to the total flux ( Figure 6).
Overall, offshore fluxes in the CSR are higher and have a larger offshore extension than in the NSR.  i.e. isolines of distance from the coast. From left to right in each group: 0 to 100 km offshore, 100 km to 500 km offshore, 500 km to 1000 km offshore, 1000 km to 1500 km offshore, 1500 km to 2000 km offshore. Colors in each bar represent the contribution by a certain type of structure: non-eddy-non-filament (NE-NF), filaments (FIL), anticyclones (AE), cyclones (CE). Positive and negative contributions to the flux through each boundary are plotted separately in order to make them clearly visible.
The offshore flux of C org in the CSR is characterized by moderate seasonal variability and intense fluxes. The offshore decrease :::::: gradient : in the intensity of the flux is less dramatic than in the NSR. The flux at 100 km peaks with nearly equal 270 values in spring and summer, but it is still intense in fall and, to a lower extent, in winter. In spring, summer and fall, this flux exceeds 70 % of the NCP produced in the first 100 km from the coast, therefore constituting an extremely large lateral redistribution of organic material produced at the coast. Filaments :: At ::: 100 :::: km, ::::::: filaments : are responsible for the large majority of the offshore fluxat 100 km, while eddies often exceed the rather small contribution of the non-eddy-non-filament flux.
Differently from what we find in the NSR, AE have a moderate but still relevant :::::::: significant ::::::: positive : contribution to the total 275 flux :::::: offshore :::::::: transport :: in ::: the :::::::: nearshore ::::: CSR, : especially in spring, when the upwelling is strongest south of Cape Blanc. This can be explained by the fact that some southern eddies may be shed by the northward-flowing MC, which likely generates C org rich AE. At 500 km offshore, the filament flux is still important, differently from what we found in the NSR. This is due to the large extension of the Cape Blanc filament, which is located just in the middle of the CSR. At distances larger than 500 km from the coast, the non-eddy-non-filament flow, likely in the form of the Cape Verde frontal circulation, is responsible for the 280 majority of the offshore flux.

Offshore flux divergence variability and its impact on the open waters
The divergence (div) of the offshore flux quantifies the net amount of C org that is added (div > 0) or removed (div < 0) by the offshore flux from each offshore domain. In steady-state, this divergence is balanced by the local balance between production and remineralization (net community production) or vertical exchange fluxes. Our results (Figure 8) show that, 285 in both subregions and in every season, div < 0 in the 0 -100 km offshore domain, the most productive coastal band. This to 100 km offshore, 100 km to 500 km offshore, 500 km to 1000 km offshore, 1000 km to 1500 km offshore, 1500 km to 2000 km offshore.
Colors in each bar represent the contribution by a certain type of structure: non-eddy-non-filament (NE-NF), filaments (FIL), anticyclones (AE), cyclones (CE). Positive and negative contributions to the flux through each boundary are plotted separately in order to make them clearly visible. means that the cross-shore flux always removes C org from this nearshore region and transports it offshore towards the open waters. However, the absolute magnitude of this C org transport out of the 0 -100 km offshore domain differs between NSR and CSR, and changes substantially between seasons in the NSR. Further away from the 100 km offshore boundary, subregional differences in the lateral relocation of C org are even more striking both in sign and in seasonality.

290
In line with all the previous findings, the NSR shows a sharp peak in the summer fluxes. In this season, the amount of C org removed from the 0 -100 km offshore domain by the offshore flux (div < 0) is at least twice as large than that in all the other seasons. Compared to the winter, when its value is minimum, the divergence is four times larger. At 100 km off the coast, the offshore flux relocates about 1/3 of the entire nearshore NCP towards the open waters in both summer and fall. The amount of C org deposited in the range of 100 -500 km (div > 0) by the offshore flux also peaks in summer and, to a lower extent, in fall,

295
confirming that this range of distances is directly impacted by the seasonality of the nearshore fluxes. This local enhancement of C org availability is especially large if compared to the very low values of summer and fall NCP in the same offshore domain.
An analysis of the alongshore fluxes of C org (see Appendix, Figure B2) show a large southward flow of organic material in the range of 100 km -500 km offshore as the southern boundary of the NSR in summer, which indicates that in this range of distances part of the C org is further displaced towards the CSR before sinking. This alongshore displacement is connected to 300 the intense southward flow of the Canary Current, which is maximum at distances larger than 100 km from the coast (Pelegrí and Peña-Izquierdo, 2015). Offshore of 500 km from the coast, the NSR shows a weakly negative summer and fall mean of NCP in the euphotic layer, meaning that the surface waters must rely on the lateral input of C org in order to sustain the excess heterotrophic activity (see also Appendix, Figure B3). This input of C org offshore results from a combination of several lateral fluxes: the delayed signal of the seasonal nearshore flux reaching the open waters in nearly a year time (Figure 5a), the spring 305 relocation of the C org produced locally in the open waters by the North Atlantic spring bloom and nevertheless transported further offshore by the currents (Figure 6a) and, potentially, any additional alongshore input of C org connected to the gyre circulation. Further, an analysis of the model generated eddy tracks (not shown) and meridional fluxes reveals that a significant northwards C org influx happens at the southern NSR boundary. Here the long and farthermost tail of the giant Cape Blanc filament oscillates and sheds C org rich eddies, which drift northwards while moving away from the coast. This is also visible in 310 the Hovmöller diagram of C org in the NSR (Figure 5a) in the form of high concentration signals that seemingly form at about 1000 km from the coast, and continue to propagate offshore. The picture is extremely different in spring and winter, when the deepening of the North Atlantic mixed layer enhances productivity offshore, therefore turning NCP positive in the open waters.
This increased local productivity of the open waters, combined with a weakening of the offshore flux, reduces the impact of the lateral C org transport on the NSR offshore ecosystem between December and May.

315
In the CSR, both the seasonal variations and the offshore gradient in the horizontal divergence of the zonal fluxes are less pronounced compared to the NSR. Spring and summer show remarkably similar patterns of the divergence, with the offshore flux transporting away from the first 100 km from the coast at least 60 % of the nearshore production. In fall, the offshore flux at 100 km offshore is as large as about 90 % of the coastal production at these latitudes. This extremely large offshore relocation of coastal-derived C org is possible also thanks to the lateral convergence of organic material into the CSR from both 320 northern and southern latitudes of the CanUS: the southward flowing CC and CUC and the northward flowing MC contribute to displacing and accumulating C org towards Cape Blanc (Lovecchio et al., 2017). The alongshore influx of C org into the CSR was demonstrated to represent on average about 30 % of the annual mean coastal production. In this sense, the CSR exports offshore not only its local production but also a fraction of the C org produced in the adjacent CanUS subregions. Most of this C org is added to the 100 -500 km offhore :::::: offshore : domain. However, differently from what we found for the northern 325 latitudes, the divergence decreases smoothly offshore, possibly due to the combined influence of the Cape Blanc filament and of the Cape Verde front, which extend far into the open Atlantic. For this reason, the horizontal divergence approaches and sometimes exceeds the magnitude of the local NCP also at the farthest offshore boundary of the CSR. Despite this lateral influx of C org , we do not find significant levels of net heterotrophy in the near-surface euphotic layer at these latitudes (see also Figure   B3). This may be due to the positive signature of the wind stress curl, especially south of the Cape Verde front, associated with 330 the anticyclonic tropical circulation. This positive wind stress curl contributes to sustaining high levels of offshore primary production.
The spatiotemporal variability of the lateral fluxes of C org in the nearshore and the associated delay in the relocation of the coastal C org at large offshore distances determines the seasonal and zonal variations in net community production integrated over the entire water-column of the adjacent offshore waters (Figure 9). While the coastal upwelling band has a positive water-335 column NCP at all latitudes, i.e., represents a net source of organic carbon, beyond 100 km from the coast, the picture becomes more complex. In the NSR, the magnitude of the seasonal fluxes is reflected in the pattern of negative water-column NCP, i.e., water-column net heterotrophy. During summer and fall, this offshore heterotrophy is widespread and far reaching, even though values remain low beyond 500 km from the coast. This reflects the limited offshore reach of the quick C org transport by upwelling filaments at these latitudes. In spring, the North Atlantic spring bloom induces high levels of offshore production in 340 the open waters of the northern CanUS. Therefore, the offshore water-column heterotrophy of the NSR is confined strictly to the first 500 km from the coast. In winter, the weakening of the offshore transport combined to a deepening of the mixed layer depth and therefore high euphotic layer production offshore (see also Figure B3) result in a neutral water-column NCP.
In the CSR, the intense C org offshore fluxes are reflected in a persistent water-column heterotrophy offshore, even though there exists substantial seasonal variations in the intensity and spatial distribution of the water-column NCP. Despite the intense 345 lateral fluxes in spring, the excess consumption of C org by remineralization is maximum in summer and fall in the offshore regions, which are reached by the coastal C org signal with a delay of at least one season. Moreover, while offshore fluxes are large in spring, so is also offshore production, especially away from the first 500 km offshore (Figure 7). Since the ratio between the organic carbon released by the lateral fluxes offshore and local offshore production determines the net heterotrophic activity, spring only shows significant negative water-column NCP between roughly 100 km and 700 km offshore, where the 350 intense filament transport is still relevant. In winter, the offshore water-column of the CSR is weakly heterotrophic, showing that at these latitudes the deepening of the mixed layer depth (and therefore increase in surface production) is not enough to compensate the intense and far-reaching offshore transport of C org .

Physical and biological drivers of the nearshore flux variability
The efficiency of the offshore transport of organic carbon is strongly dependent on the coastal production of C org , which is 355 sustained by the local upwelling of nutrient-rich deep water via Ekman Transport and Ekman Pumping. Overall, our results show that over the entire CanUS the C org offshore flux at 100km from the coast varies seasonally and in phase with both NCP and Total :::: total upwelling (sum of coastal upwelling, U and Ekman pumping, w E ) ( Figure 10).
Comparing the two sub-regions, the ratio between :: of offshore transport at 100 km offshore and :: to NCP in the first 100 km from the coast is higher in the CSR in all seasons, due to the presence of the giant Cape Blanc filament (Figure 3, see also Appendix: Figure B6). The :::::::: However, ::: the seasonal variation of the upwelling processes results more pronounced along : is ::::: more ::::::::: pronounced :: in : the NRS, similarly to the fluxes (e.g. offshore flux and NCP flux, see section 4.3). Nevertheless, the yearly mean upwelled volume in the two regions is comparable: the NSR supplies 1.2 Sverdrup (Sv) of upwelled water at surface while the CSR supplies 1.5 Sv.
In the NSR, : the total upwelling (U + w E ) peaks in summer and is still intense in spring, similarly to the other biological and Cape Juby (see Appendix: Figure B6).
The fluxes and the upwelling processes in the CSR are characterized by moderate seasonal variability (Figure 10b). The total upwelling peaks in summer (1.8Sv) and it shows nearly equal values in spring (1.7Sv). It is still intense in fall (1.3Sv) and to a lower extent, in winter (1Sv). Even thought upwelling maxima :::::: though ::: the :::::::: upwelling ::::::::: maximum is in summer, the NCP flux, 375 and in turn the C org flux, are slightly more intense in spring. This could be explained, as presented before, by the C org sinks in summer, characterized by nearly neutral or slightly heterotrophic water-column nearby the coast (e.g. north of Cape Blanc, Figure B3b). In contrast to what we find for NSR the two upwelling processes show different seasonality. Either the coastal upwelling (U ) and the Ekman pumping (w E ) peak during summer, but the latter is still strong in winter (0.22Sv) , without showing a maximum in spring :::::: ( Figure :::: 10b). The reason for this discrepancy and for the moderate seasonal variability of the 380 upwelling is embedded in the definition of the CSR sub-region. The CSR expands from 17 • N to 22 • N and it includes a large portion of ocean around Cape Blanc (21 • N). This choice was made in order to properly evaluate the offshore transport by the giant Cape Blanc filament, which accounts for a large amount of lateral export of particulate matter (Gabric et al., 1993).
Nevertheless, Cape Blanc bounds two zonal bands of the CanUS which are characterized by different seasonality of alongshore winds and, in turn, of the upwelling . Hovmöller diagrams of costal ::::: coastal : upwelling ( Figure 11b) and

385
Ekman pumping (Figure 11c) reveal pronounced upwelling north of Cape Blanc in summer and spring. By contrast, south of Cape Blanc, the upwelling processes are more robust in winter and spring. The combination of these two different upwelling peaks results in a mixed seasonal character for the CSR. In particular, the seasonality of the coastal upwelling is dominated by the strong alongshore winds ( Figure B4) during summer north of Cape Blanc, while the Ekman pumping seasonality, which depends on the gradient of the winds, is affected by the winter to spring upwelling variability south of Cape Blanc.

390
Howmöller diagrams and annual mean upwelling estimations along latitude (Figure 11a) allow the comparison between the two different upwelling processes along the coast. In particular, coastal upwelling is the dominant mechanism along the latitudes, accounting up to 80% of the total upwelling. A local exception is found south of Cape Ghir (31 • N) where Ekman pumping plays a crucial role likely due to the coastline geometry, the bay south of Cape Ghir prevents the direct effect of the along shore winds (see Appendix Figure B4).

Comparison with previous studies
Seasonality in the CanUS varies substantially with latitude, allowing to identify a set of subregions with distinct peaks in upwelling and production and distinct levels of mesoscale activity Pelegrí and Peña-Izquierdo, 2015). Our model is agrees with previous studies which distinguished between a northern region of summer upwelling intensification, a 400 central zone of semi-permanent upwelling, and the southern latitudes characterized by a late winter to spring peak of upwelling.
Ekman transport is higher next to capes (31 • N;21 • N) or just downstream due to alongshore wind acceleration, whereas Ekman 435 pumping is stronger in the lee of the capes (29 • N) where winds weaken while remaining high offshore, increasing the cyclonic curl (Bakun and Nelson, 1991;Pickett and Paduan, 2003;Koračin et al., 2004).
Our seasonal analysis of the impact of the offshore transport of C org onto the open waters of the NASG provides further insight about the net metabolic state of the near-surface open waters in low productive areas (Williams et al., 2013;Duarte et al., 2013;Ducklow and Doney, 2013). Even though in the annual mean the CanUS euphotic layer was proven to be net autotrophic 440 (Lovecchio et al., 2017), our new results show that in summer and fall the offshore waters of the NSR present weak levels of net heterotrophy in the first 100 m depth (Figure 7 and Figure B3). This excess heterotrophy is fueled by the lateral redistribution of C org in the offshore and alongshore directions combined with :::: lower :::::::: offshore :::::::::: productivity :::::: driven ::: by the shoaling of the mixed layer depth in the warm seasons, and is maximum in specific hotspots located in the offshore region adjacent to the coastal upwelling band (roughly between 100 and 200 km offshore). When compared to the net autotrophic activity of the 445 spring-winter season and to the intense autotrophy of the upwelling area, these spots of near-surface net heterotrophy are weak.
However, they may indicate that measurements made in the summer and fall in the offshore NSR could highlight weak levels of excess remineralization, i.e., negative NCP.

Model limitations
Our climatological run was forced with monthly mean winds derived from ERA-Interim (Dee et al., 2011), which, at the time 450 of testing, provided an overall better performance of the model compared to other QuickSCAT :::::::: QuikScat products (Risien and Chelton, 2008). Despite the well known biases of Era-Interim product at coast (Bonino et al., 2019b;Taboada et al., 2019), our upwelling estimations are in agreement with Messié et al. (2009)

estimations, which compute vertical transports from
QuikSCAT winds. This is likely due to the fact that we computed the coastal upwelling transport from the winds interpolated on the Atlantic telescopic grid and we considered the mean value of the alongshore winds within 100km from the coast in 455 order to be consistent with the computation of the fluxes, which are estimated at the 100km offshore boundary. Nevertheless, a higher resolution global data set of winds would allow for a more accurate assessment of the relative proportion of transport and pumping over CanCS ::::: CanUS.

485
The CanUS is characterized by the most intense seasonal fluctuations in physical and biogeochemical fluxes among all EBUS (Chavez and Messié, 2009) and by an offshore transport of C org that determines C org availability in the adjacent open waters.
Upwelling processes, driven by the surface alongshore winds, play a crucial role in determining the seasonal variations and latitudinal changes of the nearshore C org fluxes.