Seasonal variability in mass, nutrients and DOC lateral transports off Northwest African Upwelling System

Abstract. The Coastal-Ocean Carbon Exchange in the Canary Region Project (COCA) arises in order to analyse and get to understand the impact of lateral export of nutrients and organic matter from the highly productive Coastal Upwelling System off NW Africa in the biogeochemical cycles during two different seasons. The circulation patterns off NW African Upwelling System are examined by applying an inverse model to two hydrographic datasets gathered in fall 2002 and spring 2003. The mass transports estimated by model are consistent with the thermal wind equation and the conservation of mass in a closed volume. Besides, the Ekman transport and the freshwater flux are also considered. These estimates show a seasonal variability in the circulation patterns at central levels, particularly in the southern boundary of the domain, where the Cape Verde Frontal Zone is located. In the beginning of fall, this circulation is deeper and northward with a net transport of 6 ± 3 Sv and, in the late spring, it is shallower and southward with a similar intensity. At intermediate levels important differences are also observed between the two seasons. In fall, the Antarctic Intermediate Waters reaches higher latitudes with 2 ± 2 Sv flowing northward. During spring, there is no significant northward flow of AAIW. However, there is a moderate westward mass transport which impacts both the lateral transports of inorganic nutrients and organic matter at intermediate layers and also the shallowest lateral transports of organic matter. Seasonal variability in circulation patterns are also reflected in lateral transports of inorganic nutrients and dissolved organic carbon. Therefore, the changes in the circulation patterns between the two seasons have allowed us to assess the variability in the contributions of SiO2, NO3, PO4 and DOC from the first to the second season. In fall, the transports are mainly northward from the south with −0.80 ± 0.34, −1.11 ± 0.47 and −0.07 ± 0.03 kmol s-1 of SiO2, NO3 and PO4, respectively. In spring, however, lateral transports off-shore are favoured with 0.75 ± 0.37, 1.34 ± 0.66 and 0.08 ± 0.04 kmol s-1 of SiO2, NO3 and PO4, respectively. This westward transport stimulates in turn an intensified westward DOC transport at shallow layers, specifically 0.50 ± 0.25 x 108 mol C day-1.



INTRODUCTION
The North Atlantic Subtropical Gyre (NASG) is one of the most important components in the thermohaline circulation.
It presents a well-known intensification in its western margin, the Gulf Stream, with maximum velocities up to 2 m s −1 25 (Halkin et al., 1985). The currents observed in this western margin of the gyre occupy a small horizontal extension as compared to that of the currents in the eastern side, resulting in an asymmetric gyre (Stramma, 1984;Tomczak and Godfrey, 2003).
The low intensity of the currents at the eastern boundary made them very little studied until the 1970s, when CINECA program focused on the productive African upwelling (Ekman, 1923;Tomczak, 1979;Hughes and Barton, 1974;Hempel, 1982). Käse and Siedler (1982) found striking intense currents south of the Azores connected to the Gulf Stream and suggested that 30 part of the recirculation of the NASG occurs southward in the vicinity of the African coast. Later on, several surveys based on both in situ and remote sensing observations contributed to define the general characteristics for the average flow of the region (Käse and Siedler, 1982;Stramma, 1984;Käse et al., 1986;Stramma and Siedler, 1988;Mittelstaedt, 1991;Zenk et al., 1991;Fiekas et al., 1992;Hernández-Guerra et al., 1993).
Most of the eastward flow from the Gulf Stream is confined to a band between the Azores and Madeira Islands, recirculating 35 southward through the Canary Islands and north of the Cape Verde Islands to become into a southwestward flow (Stramma, 1984). This current system is composed by the Azores Current (AC), the Canary Current (CC), the Canary Upwelling Current (CUC), the North Equatorial Current (NEC) and the Poleward Undercurrent (PUC). The AC divides into several branches defining the boundary current system off Northwest Africa. It firstly feeds the Iberian Current (Haynes et al., 1993) while a second significant branch enters the Mediterranean Sea (Candela, 2001). Most of the AC recirculates southward splitting 40 into the main CC across the Canarian archipelago and the secondary CUC (Pelegrí et al., 2005. These currents extend southward developing the Cape Verde Frontal Zone (CVFZ), a density-compensated front with North Atlantic Central Water at its northern side and South Atlantic Central Water at its southern one (Zenk et al., 1991;Martínez-Marrero et al., 2008).
Finally, the PUC is located below the CUC flowing northward on the continental slope (Barton, 1989;Machín and Pelegrí, 2009;Machín et al., 2010;Pelegrí and Peña-Izquierdo, 2015). 45 The mesoscale activity constitutes a second main feature in the area of interest, which might be even more energetic than the average flow itself (Sangrà et al., 2009). Three mesoscale domains may be defined: the Canary Eddy Corridor (CEC, NADW in the whole domain during fall. Finally, the surface layer is thicker in fall than in spring in all the sections made with respect to γ n .
These seasonal differences may also be described transect to transect. The northern transect (Fig. 2, stations 2 to 32; Fig. 3, margenta dots) is occupied by NACW, AAIW, MW and NADW in both seasons. At intermediate levels, a higher contribution of MW is observed in spring while a slightly higher contribution of AAIW is obtained in fall. The western transect (Fig. 2,160 stations 32 to 42; Fig. 3, dark grey dots) has a similar distribution as the northern one, with a lower variability in the upper layers and a smaller influence of MW. In the southern transect (Fig. 2, stations 42 to 63 − 66; Fig. 3, blue dots), the highest spatio-temporal variability is observed. This variability at surface and central levels is associated to the position of the CVFZ and, in turn, to the meso-and submesoscale structures associated to the front. The CVFZ is located where the isohaline of 36, or equivalently S A = 36.15 g kg −1 , intersects the 150 m isobath (Zenk et al., 1991) (Fig. 4). CVFZ is found in the southern 165 transect in its westernmost position in fall, at stations 46 − 48. Hence, SACW with relatively low S A is observed above the upper limit of CW east of the CVFZ location (Fig. 4). In spring, the CVFZ shifts to a position closer to the African coast at station 52, with a water incursion of higher salinity NACW centred at station 58 (Figs. 4

and 5). At intermediate levels, MW
is registered at the northern transect while in the southern one the predominant water mass is AAIW. Regarding the seasonal variability, the contribution of MW in the northern transect is higher in spring while the contribution of AAIW in the southern and southern (50 to 56) transects in fall. Values observed are 1-5 µmol kg −1 for NO 3 and 0.1-0.4 µmol kg −1 for PO 4 higher 180 than those recorded in spring at similar places (Fig. 7). This might be related to long-lived mesoscale eddies or instabilities related to the CVFZ (Zenk et al., 1991;Sangrà et al., 2009). IN concentrations are notably high at intermediate and deep levels as compared to those at central levels ( Fig. 6) and have the same order of magnitude as those documented before in the domain Pérez-Hernández et al., 2013). The distributions of SiO 2 , NO 3 and PO 4 are similar in both cruises and their concentrations increase with depth as a result of the remineralization of organic matter (Fig. 7) maximum concentrations of SiO 2 are 28-29 µmol kg −1 . Nevertheless, and specifically in spring, maximum concentrations of NO 3 and PO 4 , 28 µmol kg −1 and 1.8-1.9 µmol kg −1 , are lower than those recorded at intermediate levels, providing a similar vertical variability as that reported by Machín et al. (2006) (Tab. 2).
DOC concentrations are higher and more widely distributed in the water column in fall than in spring, when the DOC maximum values are more confined to surface and central waters (Figs. 8 and 6,Tab. 2). This fact is especially significant 195 in the southern transect occupied by SACW (Fig. 6). This last water mass presents maximum concentrations of DOC 35 − 40 µmol L −1 lower than those found for NACW (Tab. 2). This difference is more pronounced in spring season (Tab. 2).
In addition, the fall DOC observations present a larger variability in central waters as previously seen for IN. Lower DOC concentrations are observed for stations sampled in the western transect while the highest concentrations are recorded in the stations next to the African slope with values above 100 µmol L −1 (Fig. 8). On the other hand, it is noteworthy the high

THE INVERSE MODEL
An inverse box model is applied to the hydrographic data of the two COCA cruises to provide the absolute velocity field across the three sections (Wunsch, 1978). This method has been widely applied in different areas of the Atlantic Ocean as an efficient method to obtain absolute geostrophic flows (Martel and Wunsch, 1993;Paillet and Mercier, 1997;Ganachaud, 2003a;205 Machín et al., 2006;Pérez-Hernández et al., 2013;Hernández-Guerra et al., 2017;Fu et al., 2018). Assuming geostrophy and the conservation of mass and other properties in the ocean bounded by the African coast and the hydrological sections, the velocity fields are obtained allowing an adjustment of freshwater flux and Ekman transports.

Selection of layers
The closed ocean where the inverse model is applied is divided into nine layers by means of the neutral densities defined by 210 Macdonald (1998) and modified by Ganachaud (2003a) for the North Atlantic Ocean. This distribution is slightly modified to include two layers instead of one between 26.85 and 27.162 kg m −3 by adding the isoneutral 27.035 kg m −3 as others authors have done previously in this side of the NASG (Comas-Rodríguez et al., 2011;Pérez-Hernández et al., 2013). The

The system of equations
The inverse box model takes into account mass conservation per layer and also in the whole water column. The salinity is actually introduced as a salinity anomaly, which is also conservative within individual layers and in the whole water column 220 (Ganachaud, 2003b). On the other hand, heat is introduced as a heat anomaly in the two deepest layers where it is also considered conservative. The salinity and heat are added as anomalies to improve the conditioning of the inverse model and get a higher rank in the system of equations by reducing the linear dependency between equations (Ganachaud, 2003b). Therefore, the model is composed of a set of 22 equations (10 for mass conservation, 10 for salt anomaly conservation and 2 for heat anomaly conservation). Those equations are solved for 32 and 34 unknowns, comprised of 28/30 reference level 225 velocities in fall/spring, 3 unknowns for the Ekman transport adjustments (one unknown per section), and 1 unknown for the freshwater flux. The resulting system is undetermined and a Gauss-Markov estimator is used to select a solution by adding a priori information. This a priori information consists of the uncertainties for both the unknowns (R xx ) and the noise of the equations (R nn ).

230
The geostrophic velocity field is calculated in the central position between two consecutive stations. The isoneutral selected as the reference level is the deepest common γ n for all the stations, 27.962 kg m −3 (Fig. 2). The variance of the velocity in the reference level at each location is used as a measure of the a priori information. These variances are calculated with an annual mean velocity extracted from the daily velocity provided by GLORYS. These velocities are interpolated to the reference level depth. This reference level depth is estimated from the climatological mean depth of 27.962 kg m −3 extracted from WOA13.

235
The stations closer to the coast in the northern and southern transects have the highest variability in the velocity field.
The initial Ekman transports are estimated from the wind stress for both cruises. The uncertainty associated to these Ekman transports is related to the error in their measurements and to the variability of the wind stress. A 50% uncertainty is assigned to the initial estimate of Ekman transports. The initial freshwater flux is a climatological mean of 0.0171 Sv, which is also assigned an uncertainty of 50 % as reported in similar approaches (Ganachaud, 1999;Hernández-Guerra et al., 2005;Machín et al., 240 2006).
Both the Ekman transports and freshwater flux with their uncertainties are added to the model in the conservation equations corresponding to the shallowest layer of the mass transport and salt anomaly and also in the conservation equations of total mass transport and total salt anomaly.

245
The noise of each equation depends on the density field, on the layer thickness and on the uncertainties of the unknowns (Ganachaud, 1999(Ganachaud, , 2003bMachín et al., 2006). In fact, Ganachaud (2003b) established that the largest source of uncertainty in conservation equations arises from the deviation of the baroclinic mass transport from their mean value at the time of the cruise. Thus, an analysis of the annual variability in the velocity field for the nine layers is performed. The velocity variability is examined in the mean depth between two successive isoneutral surfaces whose climatological mean depths are defined by the section involved. These a priori transport uncertainties are presented in Table 3. Furthermore, the uncertainty assigned to the conservation equation in the total mass is the sum of the uncertainties from the rest of the nine conservative mass equations.

255
The equations for salt and heat anomaly conservation depend on both the uncertainty of the mass transport and the variance of these properties (Ganachaud, 1999). In these cases, the a priori noise of each equation will not depend strictly on the water mass but on the layer considered, as shown in the following equation (Ganachaud, 1999;Machín, 2003): where R nn (Cq) is the uncertainty in the anomaly equation of the property (salt or heat anomaly); var(C q ) is the variance 260 of this property; a is a weighting factor of 4 in the heat anomaly, 1000 in the salt anomaly and 10 6 in the total salt anomaly; q is a given equation corresponding to a given layer. In fall, two eddies are linked in the S transect, an anticyclonic one between stations 48 and 52 and a cyclonic one between stations 52 and 60, both associated with the CVFZ. In spring, two anticyclonic eddies are observed, one centred at station 285 extension ( Figure 10). In fall, these structures have higher velocities at IW and DW levels and they also affect a higher extension along each transect. In spring instead, these structures are vertically shortened (Fig. 10). The SLA also shows a high variability region with more intense structures in fall than in spring (Fig. 11).

Velocity fields and mass transports
Mesoscale structures are also visible in the vertical sections of NO 3 and PO 4 in fall, when their concentrations are higher 290 than those observed in spring at similar locations ( Fig. 7). Furthermore, high concentrations of DOC in fall at CW levels are recorded in the same area where the deep anticyclonic eddy is located, between stations 8 and 18 (Fig. 8). In spring, mesoscale structures in the vertical sections of IN and DOC at CW levels are less intense than in fall (Fig. 10). Nonetheless, DOC concentrations below the two anticyclonic structures at CW levels in spring are higher than at their surroundings.
The accumulated geostrophic mass transport is integrated to group the variability at different levels, having the first shallow-295 est layer for SW, the next four layers for CW, then two layers for IW and the deepest two layers for DW ( Figure 12). The total accumulated geostrophic mass transport, integrated for all the nine layers, is also represented. The horizontal axis has the same direction as the rest of the vertical sections and the three transects are separated by two vertical dashed grey lines. Sv is used here as equivalent to 10 9 kg s −1 . The positive/negative transport values indicate outward/inward transports from/to the box.
The accumulated mass transports show a significant horizontal spatial variability, especially marked in the southern transect 300 in accordance to the geostrophic velocity distribution (Fig. 10). The presence of significant mesoscale structures might be one of the sources for the total imbalances in the accumulated mass transport. In fall, the total imbalance is -1.43 Sv and in spring 3.55 Sv (Tab. 4).
On the other hand, the geostrophic mass transport can be integrated per layer and transect together with the total imbalance inside the box and the total mass transport uncertainty per layer (black line and horizontal black bars in Fig. 13). Moreover, 305 Table 4 compiles these transports integrated for the different water levels. More than 65% of the mass transport is given at SW and CW levels (Tab. 4). In fall, these water masses mostly get into the box across the northern and southern transects with transports of −5.61 ± 1.86 Sv and −4.35 ± 1.48 Sv, respectively; the mass leaves the box by flowing westward with a value of 5.96 ± 1.75 Sv. In spring, water masses also get in the box mostly through the northern transect with −6.69 ± 1.63 Sv but they set off along the western and southern transects with transports of 4.05 ± 1.75 Sv and 5.20 ± 1.55 Sv, respectively. It is 310 remarkable how the inward transport in fall across the southern transect is reversed to an outward flow in spring at the southern transect (Fig. 13).
The position of CVFZ in both seasons could partly explain that seasonal variability in the mass transports at central levels ( Fig. 14). In fall, the CVFZ is located further from the African coast, so SACW is present at almost all stations of the south transect. This location of the CVFZ prevents a latitudinal mass transport from north to south. However, in spring the CVFZ is 315 closer to the African slope allowing an important mass transport from north to south.
Between 5 and 30% of the mass transport is given in intermediate levels (Tab. 4). In fall, the intermediate water transport directs northward in the southern transect with −1.93 ± 1.69 Sv and it leaves the box with 1.94 ± 1.85 Sv and 0.48 ± 1.71 Sv across the northern and western transects, respectively. During spring, this transport weakens and changes its direction in the transport to 1.21 ± 1.68 Sv.
The mass transport in deep water layers barely exceeds 3% (Tab. 4). An exception is the 8% given in the northern transect during fall where the estimated transport is 0.73 ± 1.71 Sv. In both cruises the transport at deep levels is nearly balanced.  . This is done because the refractory fraction renewal is thousands of years, a period much longer than the time required in the processes we are focused on (Hansell, 2002). On the other hand, it should be emphasized that DOC transports may be underestimated due to the scarcity of measurements performed. Finally, at DW during both seasons, the net transports of the three nutrients are similar to those at IW but with smaller values 355 due to the low velocities at these depths, despite their high nutrient concentrations (Figs. 15 and 16). Furthermore, the relative error in these layers is always larger than the IN transport values.

Nutrient and
In spring, DOC transports at SW and CW levels are the same order of magnitude and one order of magnitude higher than those at IW levels. In turn, these transports at IW levels are one order of magnitude higher than those at DW levels during this season. In contrast, during fall at the northern transect DOC transports have the same magnitude in both SW, CW and IW and On the other hand, the net DOC transports are outward for both SW, CW and IW levels with 0.10 ± 0.13 ×10 8 mol C day −1 at SW level, 1.34 ± 0.80 ×10 8 mol C day −1 at CW levels, and 0.12 ± 0.72 ×10 8 mol C day −1 at IW (Tab. 8 and Fig. 16).

DISCUSSION
The circulation patterns in the studied area of the Canary Basin change significantly showing a seasonal variability from fall to spring. The differences between the two seasons are reflected in the estimated mass transports for both cruises (Fig. 13 and Tab. 4).
permanent upwelling in this region north of Cape Blanc. In contrast, the developed EBUS intensity and its off-shore development change from fall to spring (Benazzouz et al., 2014). In the beginning of spring there is a strong heating that generates a sharp water stratification particularly in the interior ocean of the NASG and a very intense upwelling which makes the EBUS to develop strongly far off-shore. In early fall, the EBUS weakens and becomes a shallower front which approaches towards the coast (Pelegrí and Benazzouz, 2015a). In fact, the variability related to its location and intensity may be the cause that the at CW and IW waters. It is also deduced from DOC transport estimates that the upwelling drives the changes in the size of the high production domain and equivalently, the position for the eastern boundary of the oligotrophic region in this area (Pastor et al., 2013). Westward transport of DOC is not observed even at the shallowest layer.
It is still necessary to continue with the understanding of the physical and biogeochemical processes and the interactions between the productive EBUS and the interior ocean in its vicinity, especially in dynamically complex regions as this area where the EBUS interacts with the CVFZ. Larger and more robust hydrological and biogeochemical databases would help to   Figure 2. γn vertical sections during fall (top) and spring (bottom) cruises. White dashed isoneutrals limit the different water type layers.
The direction chosen for the representation of the transects is the course of the vessel. Distance is calculated with respect to the first station (2). The section is divided into three transects: northern transect from east to west (from station 2 to 32), western transect from north to south       Fig. 12. The net transport in the whole box is shown by the black line.