Spatio-temporal patterns of C : N : P ratios in the northern Benguela upwelling system

On a global scale the ratio of fixed nitrogen (N) and phosphate (P) is characterized by a deficit of N with regard to the classical Redfield ratio of N : P = 16 : 1 reflecting the impact of N loss occurring in the oceanic oxygen minimum zones. The northern Benguela upwelling system (NBUS) is known for losses of N and the accumulation of P in suband anoxic bottom waters and sediments of the Namibian shelf resulting in low N : P ratios in the water column. To study the impact of the N : P anomalies on the regional carbon cycle and their consequences for the export of nutrients from the NBUS into the oligotrophic subtropical gyre of the South Atlantic, we measured dissolved inorganic carbon (CT), total alkalinity (AT), oxygen (O2) and nutrient concentrations in February 2011. The results indicate increased P concentrations over the Namibian shelf due to P efflux from sediments resulting in a C : N : P : -O2 ratio of 106 : 16 : 1.6 : 138. N reduction further increase C : N and reduce N : P ratios in those regions where O2 concentrations in bottom waters are < 20 μmol kg−1. However, off the shelf along the continental margin, the mean C : N : P : -O2 ratio is again close to the Redfield stoichiometry. Additional nutrient data measured during two cruises in 2008 and 2009 imply that the amount of excess P, which is created in the bottom waters on the shelf, and its export into the subtropical gyre after upwelling varies through time. The results further reveal an inter-annual variability of excess N within the South Atlantic Central Water (SACW) that flows from the north into the NBUS, with highest N values observed in 2008. It is postulated that the N excess in SACW occurred due to the impact of remineralized organic matter produced by N2 fixation and that the magnitude of excess P formation and its export is governed by inputs of excess N along with SACW flowing into the NBUS. Factors controlling N2 fixation north of the BUS need to be addressed in future studies to better understand the role of the NBUS as a P source and N sink in the coupled C : N : P cycles.


Introduction
The biological carbon pump is the term used for the production of organic carbon from dissolved carbon dioxide (CO 2 ) in the surface mixed layer of the ocean and its transport into the large CO 2 reservoir of the ocean beneath the mixed layer. It is both driven and limited by the availability of macronutrients, such as fixed nitrogen (N) and phosphate (P), as well as micronutrients such as iron (Watson et al., 2000;Behrenfeld et al., 2006b). Macronutrients are required in specific stoichiometric ratios for the photosynthetic production of organic matter, traditionally termed the Redfield ratio of C : N : P = 106 : 16 : 1 (Redfield et al., 1963). Nevertheless, on a more regional scale, phytoplankton C : N : P ratios vary with growth rate, taxonomy, ambient CO 2 concentrations and nutrient availability (e.g. Arrigo, 2005;Riebesell et al., 2007). Especially upon exhaustion of the nutrients, changes in the nutrient uptake ratios could strongly influence Published by Copernicus Publications on behalf of the European Geosciences Union.

886
A. Flohr et al.: Spatio-temporal patterns of C : N : P ratios the marine productivity and the oceans ability to sequester CO 2 from the atmosphere (McElroy, 1983;Heinze et al., 1991;Falkowski, 1997).
The ratios of remineralized C, N, P and consumed oxygen (-O 2 ) (C : N : P : -O 2 ) are useful to characterize the coupled nutrient and carbon cycling in the oceans. On a global scale the mineralization of nutrients is essentially constant with depth and in good agreement with the traditional Redfield ratio for the photosynthetic production of organic matter (Redfield et al., 1963;Takahashi et al., 1985;Anderson and Sarmiento, 1994). However, depending also on the methods used, variations of the C : N : P : -O 2 mineralization ratios, e.g. with depth have been suggested indicating elemental fractionation during the mineralization of sinking organic matter (Takahashi et al., 1985;Li and Peng, 2002;Schneider et al., 2003;Brea et al., 2004) and an impact of N loss occurring in oxygen minimum zones (OMZ) (Gruber and Sarmiento, 1997;Tyrrell and Law, 1997).
Eastern boundary upwelling systems (EBUS) are regions of high CO 2 concentrations (Boehme et al., 1998;Torres et al., 1999) and intense biological production and export of carbon (Carr, 2002). They also play an important role in supplying nutrients to the surface mixed layer of adjacent oligotrophic subtropical gyres, where nutrient supply is limited by stable thermal stratification (Behrenfeld et al., 2006a). The Benguela upwelling system (BUS) is a coastal upwelling system known for non-standard nutrient (N : P) ratios in upwelling waters (Dittmar and Birkicht, 2001;Tyrrell and Lucas, 2002) caused by N loss (anammox and/or denitrification) and P release from sediments in low-O 2 environments (Kuypers et al., 2005;Nagel et al., 2013). The C : N : P remineralization ratios in the subsurface waters are poorly constrained but are crucial in order to characterize the cycling of C in the BUS.
We analysed nutrient (NO − 3 , NO − 2 , P) and dissolved O 2 data in conjunction with data on dissolved inorganic carbon (C T ) and total alkalinity (A T ) produced during an expedition in 2011, and complemented these with nutrient data from two other expeditions staged in 2008 and 2009. Our objectives were to investigate C : N : P : -O 2 mineralization ratios of the northern Benguela upwelling system (NBUS) to study the spatial and temporal impact of N reduction and the associated consequences for the export of nutrients from the eutrophic upwelling system into the oligotrophic subtropical gyre.

Study area
The BUS spans along the southwestern coast of Africa, covering the western South African and Namibian coastline roughly from Cape Agulhas (∼ 34 • S) to the Angola Benguela Frontal Zone (ABFZ) (Hutchings et al., 2009) (Fig. 1a). At the ABFZ, which is centred between 14 and 16 • S (Meeuwis and Lutjeharms, 1990) but is highly dynamic in terms of shape and location, the cold Benguela current system converges with warm tropical waters of the Angola Current (AC). To the south, the system is bordered by the Agulhas Current, which reverses and partly converges with the South Atlantic water, resulting in the formation of eddies (Agulhas rings, AR) and filaments (Hall and Lutjeharms, 2011).
Along the coast of southwestern Africa, the interaction of southerly trade winds with coastal topography forces upwelling, which is strongest at three distinct upwelling cells (Shillington et al., 2006;Hutchings et al., 2009). The Lüderitz upwelling cell (∼ 26 • S) accounts for roughly 50 % of physical upwelling and separates the upwelling region into a northern and a southern subsystem (Shannon, 1985;Duncombe Rae, 2005). In the southern region (south of 26 • S) the trade winds are seasonal and upwelling maximizes during austral spring and summer. The northern region (from 26 • S to the ABFZ) is characterized by perennial alongshore winds and upwelling along the coast (Shannon, 1985).
The Lüderitz cell also marks the boundary of two central water regimes that cause distinct differences in biogeochemical properties of upwelling waters. Upwelling in the southern BUS entrains Eastern South Atlantic Central Water (ESACW) into the offshore Ekman drift (Duncombe Rae, 2005;Mohrholz et al., 2008). The ESACW is a central water mass that forms in the Indian Ocean and enters the South Atlantic Ocean by Agulhas Current intrusions (Poole and Tomczak, 1999;Stramma and England, 1999). Between the ABFZ and the Lüderitz cell, the ESACW mixes with an Angola Gyre subtype of the South Atlantic Central Water (SACW). This subtype of SACW originates in the subtropical Angola Gyre and enters the Namibian shelf and continental margin in a poleward undercurrent via the Angola Current (Duncombe Rae, 2005;Mohrholz et al., 2008). The SACW originally forms in the Brazil-Malvinas Confluence off South America (Gordon, 1981), flows eastward with the South Atlantic Current and is then diverted northwards along with the Benguela Current towards the equatorial current system. Here, a branch spreads eastwards and flows into the Angola Gyre (Poole and Tomczak, 1999). This Angola Gyre subtype is older and thus strongly enriched in nutrients and depleted in O 2 compared to the ESACW (Poole and Tomczak, 1999;Mohrholz et al., 2008). Increased inflow of SACW into the NBUS preconditions the development of an oxygen minimum zone (OMZ) and anoxic events over the Namibian shelf and upper slope (Weeks et al., 2002;Monteiro et al., 2006;Mohrholz et al., 2008).
On the shelf at low O 2 concentrations (< 20 µM O 2 ), fixed nitrogen (N) is reduced by denitrification and/or anammox (Lam and Kuypers, 2010;Kalvelage et al., 2011). The loss of fixed N to anammox within the OMZ on the shelf has been estimated to be ∼ 1.4 Tg N yr −1 (Kuypers et al., 2005) and the loss to denitrification to~2.5 Tg N yr −1 (Nagel et al., 2013). However, N loss exceeds estimates on N 2 fixation  (Sohm et al., 2011), suggesting that the BUS acts as net sink for fixed N. Furthermore, the BUS shelf is a region of modern phosphorite deposition (Glenn et al., 1994;Föllmi, 1996) associated with a massive organic-rich diatomaceous mud belt that roughly follows the Namibian coast between 50 and 200 m water depth (Fig. 1b) and covers an area of ∼ 18 000 km 2 (Bremner, 1980;Bremner and Willis, 1993;Emeis et al., 2004). The mud belt surface is settled by consortia of large sulfur bacteria (including Thiomargarita namibiens) that release phosphate (P) into the anoxic pore water (Nathan et al., 1993;Schulz and Schulz, 2005;Goldhammer et al., 2010) and thereby enrich pore waters to ∼ 1000 µM PO 3− 4 (van der Plas et al., 2007).

Water sampling, laboratory work and data analysis
Nutrient samples were collected during three cruises in austral summer and early autumn ( Fig. 1b Kunene, Rocky Point and Walvis Bay were sampled during all of the three cruises. Samples were collected by CTD casts using a rosette system equipped with 10 L Niskin bottles. The upper water column was sampled at fixed depth levels (5, 10, 15, 20, 30, 50, 100 and 200 m). From 200 m downwards the depth levels were extended to 100-300 m intervals depending on the bottom depth and, for example, the O 2 profile. The analysed parameters presented in this study comprise dissolved inorganic carbon (C T ), total alkalinity (A T ) and dissolved nutrients (NO x = nitrate (NO − 3 )+ nitrite (NO − 2 ) and phosphate (PO 3− 4 )). The apparent oxygen utilization (AOU) was calculated from O 2 concentrations using the equations for O 2 saturation according to Weiss (1970).

C T and A T
The C T and A T samples were taken during the MSM17/3 cruise. Transects off Kunene (17.25 • S), Rocky Point (19 • S), Terrace Bay (20 • S), Toscanini (20.80 • S) and Walvis Bay (23 • S) were sampled. For C T and A T analysis the samples were filled into 250 mL borosilicate bottles using silicone tubes (Tygon). The bottles were rinsed twice and filled from the bottom to avoid air bubbles. Duplicate samples were periodically taken. The samples were fixed with mercury (II) chloride solution (250 µL of a 35 g L −1 HgCl 2 solution) directly after collection and analysed on board using 888 A. Flohr et al.: Spatio-temporal patterns of C : N : P ratios the VINDTA 3C system (Mintrop, 2005). The A T was determined on the basis of a semi-closed cell titration principle. The samples were titrated with a fixed volume of hydrochloric acid (HCl, 0.1 N). C T was quantified by the coulometric method (Coulometer CM 5015, precision of 0.1 %) after extracting the CO 2 out of the acidified water samples. Certified reference material (CRM, batch #101 and #104, provided by A. Dickson (Scripps Institution of Oceanography, La Jolla, CA, USA)) was used to calibrate the VINDTA 3C system. Measured A T and C T values agreed within ±2.5 µmol kg −1 for the CRMs and within ±2 µmol kg −1 for the duplicate samples. In the following A T is reported as the salinity corrected value (A T = A Tmeas × 35/S meas ).

Dissolved nutrients
The nutrient samples were filtered through disposable syringe filters (0.45 µm) immediately after sampling, filled into prerinsed 50 mL PE bottles and frozen (−20 • C). Samples collected during the MSM07/2b-3 cruise were measured on board, whereas samples taken during Afr258 and MSM17/3 cruises were analysed in the shore-based laboratory subsequent to the expedition. Dissolved nutrients were measured by a continuous-flow injection system (Skalar SAN plus System) according to methods described by Grasshoff et al. (1999). The detection limits were NO x = 0.08 µM and PO 3− 4 = 0.07 µM according to DIN 32645. Ammonium (NH + 4 ) concentrations were usually < 2.5 µ mol kg −1 and are not discussed in this paper. In the following N = nitrate (NO − 3 )+ nitrite (NO − 2 ) and P = PO 3− 4 will be used throughout the paper.

Elemental stoichiometry
To calculate the deviation from the classical Redfield ratio (N : P = 16 : 1) (Redfield et al., 1963), we used the tracer N* (Gruber and Sarmiento, 1997): where [NO − 3 ] and [PO 3− 4 ] are the concentrations of nitrate and phosphate in µmol kg −1 , respectively. The constants drive the global mean N* value of −2.9 µmol kg −1 to zero. Positive and negative N* are indicative of an excess and deficit of NO − 3 relative to PO 3− 4 , respectively (Gruber and Sarmiento, 1997). To quantify the PO 3− 4 anomaly (P*) from Redfield we use the concept of Deutsch et al. (2007): N* and P* are a measure for the deviation from the Redfield ratio and are arbitrary values rather than definite concentrations.

Characterization of central water masses
The potential temperature (T pot ) and salinity (S) characteristics were used to differentiate between SACW and ESACW contributions. Their definitions were adopted from Mohrholz et al. (2008), who identified an Angola Gyre subtype of SACW. This subtype is characterized by O 2 concentrations ranging between 22 and 68 µmol L −1 in contrast to the ESACW that shows O 2 values of 249-300 µmol L −1 in the Cape Basin (Poole and Tomczak, 1999;Mohrholz et al., 2008). The SACW and ESACW are defined by a line in T pot − S space that can be described by the following equations: The above equations were transformed to calculate the respective proportions of SACW and ESACW: Data derived from water depths above 100 m were excluded from this mixing analysis due to the non-conservative behaviour of T pot and S at shallow water depths. The relative contributions are reported in percentage (%). The T pot − S range used to calculate the relative contribution of SACW and ESACW is shown in Fig. 2 for the example of the MSM17/3 cruise.

Results and discussion
On a global scale the distribution of N : P is characterized by a deficit of N towards P with regard to the Redfield ratio (Gruber and Sarmiento, 1997;Tyrrell and Law, 1997). The loss of fixed N is caused by heterotrophic denitrification and anammox occurring in the oceanic OMZs (Lam and Kuypers, 2010) but also in shallow coastal OMZs, for example on the Namibian shelf. Here, N loss (Kuypers et al., 2005;Nagel et al., 2013) along with the accumulation of P in suband anoxic bottom waters (Goldhammer et al., 2010) result in extremely low N : P ratios in the water column. In the following the C : N : P : -O 2 remineralization patterns observed in 2011 are discussed with regard to their spatial variability and are complemented by results on the temporal variability of N : P anomalies recorded in 2008 and 2009.

C : N : P : AOU stoichiometry
The N and P concentrations of the NBUS scatter to both sides of the reference Redfield slope and characterize the NBUS as a system that produces both positive and negative deviations from Redfield ratio expressed in positive and negative N* anomalies (Fig. 3). Major negative anomalies were apparent at shelf sites. Tyrrell and Lucas (2002) attributed low N : P (LNP) data (N : P < 3 and P > 1.5 µmol kg −1 ) in waters of the BUS to nutrient trapping and denitrification that leads to a relative accumulation of P. Figure 4 illustrates the relationships between apparent oxygen utilization (AOU), N, P, A T and C T of the water samples in 2011 (MSM17/3). The observed average C T : AOU ratio of 0.76 (r 2 = 0.89) is close to that expected from a mineralization C : -O 2 ratio of 106 : −138 (Fig. 4a). Exclusion of the O 2 < 20 µmol kg −1 data gave no mentionable differences between the shelf (C T : AOU = 0.77, r 2 = 0.88) and offshore sites (C T : AOU = 0.75, r 2 = 0.89). The scattering of the data in Fig. 4a is likely due to the fact that the shelf system is not truly closed and that O 2 is introduced into the subsurface water masses on the shelf through mixing (Ito et al., 2004). The increase of C T at AOU > 230 µmol kg −1 corresponding to O 2 < 20 µmol kg −1 implies C T input from anaerobic respiration, such as denitrification, and is evident Fig. 3. Composite of N versus P (µmol kg −1 ) data of the MSM07/2b-3 cruise (circles), Afr258 cruise (diamonds) and MSM17/3 cruise (squares). The data were separated into shelf and slope stations (< 500 m bottom depth), indicated by open symbols, and offshore stations (> 500 m bottom depth), indicated by blackfilled symbols. The red-filled symbols correspond to data points associated with O 2 concentrations ≤ 20 µmol kg −1 . Positive and negative deviations from the expected N : P correlation of 16 : 1 (black line) are expressed in +N* and −N* (Gruber and Sarmiento, 1997). The grey-shaded area refers to the range of low N : P (LNP) defined by Tyrrell and Lucas (2002). from decreasing N associated with increasing C T concentrations (Fig. 4b). At O 2 concentrations < 20 µmol kg −1 , both anammox and denitrification increased A T through the consumption of N (Fig. 4d). However, a decrease in C T indicating a dominance of anammox over heterotrophic denitrification is not visible in our data (Fig. 4b), likely due to the low C : N stoichiometry of anammox (C : N ∼ −0.07 : −1.3) compared to that of denitrification (C : N ∼ 106 :−104) (Koeve and Kähler, 2010). A loss of N by 20 µmol kg −1 would result in a C T decrease of −1 µmol kg −1 due to anammox and a C T increase of +21 µmol kg −1 during denitrification. At higher O 2 the overall average C T : N = 6.1 (r 2 = 0.86) is similar to the Redfield ratio of 6.6, with slightly lower values in the open ocean (5.5, r 2 = 0.89) than on the shelf (6.8, r 2 = 0.86). The slope of A T : C T ∼ −0.15 (Fig. 4d) observed in the paired data with > 20 µmol kg −1 O 2 agrees with the expected effect of aerobic organic matter remineralization (A T : C T = −16 : 106) (Broecker and Peng, 1982) and implies that carbonate dissolution hardly affected the C T and A T concentrations.
A source of P besides the mineralization of organic matter in the water column is suggested by the spread of the C T : P data in Fig. 4c. Open-ocean sites had an average C T : P ratio of 101 : 1, which is similar to the global mean C : P ratio of 106 : 1 (Redfield et al., 1963;Anderson and Sarmiento, 1994). The C T : P correlation of the shelf data splits in two groups. One group reveals a slope of the regression line of  (Redfield et al., 1963) and N loss, e.g. due to denitrification A T : C =104 : 106 ∼ 1 (Gruber and Sarmiento, 1997).
∼ 98 : 1 (r 2 = 0.82) that is slightly lower but similar to those seen in the offshore samples. The remaining samples suggest a much lower average C T : P ratio of ∼ 65 (r 2 = 0.84) and are related to the mud belt region, where even lower C : P ratios of 33-48 were measured in pore waters near the sedimentwater interface (Goldhammer et al., 2011). The low C T : P occurring exclusively at shelf sites indicate an impact of pore water P effluxes from the anoxic mud sediments mediated by consortia of sulfur bacteria (Schulz and Schulz, 2005). This is in line with previous hypotheses that the stoichiometric N deficit in waters over the Namibian shelf is, in addition to the impact of N reduction, in part caused by P fluxes across the sediment-water interface (Bailey and Chapman, 1991;Nagel et al., 2013). Accordingly, the impact of N loss and benthic P fluxes on the suboxic bottom layer should be observable by a spatial decoupling of the N and P maxima; that is, N max would be expectable outside the OMZ, while P max would be expectable inside the OMZ. This is in agreement with our observations of N and P maxima along the Namibian shelf and slope (Fig. 5, MSM17/3 cruise in February 2011). The OMZ was positioned between 300 and 400 m water depth off Kunene (17.25 • S) (Fig. 5a), stretching from the slope towards the open ocean. The maxima of N (45 µmol kg −1 ) and P (2.8 µmol kg −1 ) were observed at the same depth ranges slightly below the OMZ. In contrast, the OMZ was restricted to the shallow shelf region off Walvis Bay (23 • S) (Fig. 5b) overlying the diatomaceous mud belt, where the large sulfur bacteria occur that release P into the anoxic pore water (Goldhammer et al., 2011). In fact, P max was strongly elevated (4.8 µmol kg −1 ), and coincided with the OMZ, while N max had decreased to 35 µmol kg −1 and was observed outside the OMZ. This increase of P relative to N is reflected in a pronounced N : P deviation from Redfield, as indicated by strongly decreased N* values over the shelf and shelf break off Walvis Bay (23 • S) compared to Kunene (17.25 • S). Although the N loss likely contributes to the overall N decrease, it also reflects the gradual increase of the ESACW fraction towards the south. This water mass is characterized by lower nutrient concentrations than SACW (Poole and Tomczak, 1999;Mohrholz et al., 2008).
To summarize, C : N : P : -O 2 ratios at offshore sites are 101 : 16 : 1 : 138 and are close to Redfield ratios. Over the Namibian shelf, with particular regard to the mud belt region, increased P concentrations result in a C : N : P : -O 2 ratio of 106 : 16 : 1.6 : 138. In regions with O 2 concentrations < 20 µmol kg −1 , denitrification further increases C : N and denitrification along with anammox further lower N : P ratios.

Spatial and temporal variability
Inter-annual differences in upwelling conditions during austral summer expeditions were well reflected in the distribution of sea surface temperatures (SST). The SST patterns, measured continuously by thermo-salinographs at 5 m water depth, indicate that upwelling was most intense in December 2009 (Afr258) (Fig. 6), when minimum SST (15 • C) occurred along the entire coast and temperatures < 20 • C were measured far offshore. In contrast, most of the NBUS had SST > 20 distinct distributions of the central water masses on the shelf. Mohrholz et al. (2008) found that vigorous cross-shelf circulation during phases of strong upwelling suppresses the along-shelf poleward undercurrent of SACW from the north. Weak upwelling phases, such as that in February 2011, permit SACW to protrude far southward as indicated by the high SACW contribution (70 %) off Walvis Bay (23 • S) (Fig. 7df). SACW is characterized by much lower O 2 concentrations than ESACW (Mohrholz et al., 2008), which is also evident from our data (Fig. 2). In fact, the dominance of SACW was reflected in mid-water O 2 deficits, and thus samples with ≥ 80 % SACW were associated with ≤ 50 µmol kg −1 O 2 (Fig. 7). However, during strong and weak upwelling alike, and regardless of the SACW contribution, the O 2 concentrations on the shelf off Walvis Bay were < 20 µmol kg −1 during all cruises. This reflects the strong O 2 demand caused by the organic-rich mud belt area and is in line with previous studies showing that SACW sets the precondition for anoxia in bottom waters off Walvis Bay but that the local sedimentary O 2 demand plays a decisive role as well (Monteiro et al., 2006;van der Plas et al., 2007). The N : P deviation from Redfield varied during the different expeditions and upwelling states (Fig. 8). Pronounced negative N* concentrations were observed in coastal bottom waters, in particular off Rocky Point and Walvis Bay (Fig. 8a-c), and are comparable with reported values (Tyrrell and Lucas, 2002;Nagel et al., 2013). Elevated N* was observed at offshore sites and differed strongly in magnitude especially offshore of Kunene (17.25 • S), where SACW dominates ( Fig. 7d-f), indicating that the N* signature of SACW differs. From the temporal variability of N* in the hemipelagic OMZ offshore of Kunene, it is apparent that N* in the SACW water mass varied significantly between years (Fig. 9). During March 2008 (MSM07/2b-3), N* had increased threefold within the OMZ (N* ∼ 12 µmol kg −1 ) compared to December 2009 and February 2011 (N* ∼ 3-4 µmol kg −1 ). This positive N* anomaly was imported to the NBUS in 2008 along with SACW, and thus water masses on the shelf were less N*-deficient (Fig. 8a-c) despite similarly low O 2 concentrations in 2009 and 2011 (Fig. 7a-c). Strong local sedimentary O 2 demand caused by the organicrich mud belt area (Fig. 1b)  in March 2008 by a factor of > 2. The results suggest that the considerable variability in nutrient ratios of SACW controls the inter-annual variability of N* in the NBUS. The impact of ESACW entering the NBUS from the south is difficult to assess from our data due to a lack of ESACW samples from the southern BUS. The ESACW sampled in the NBUS is characterized by lower O 2 concentrations (∼ 150-200 µmol kg −1 ) (Fig. 2) than reported for ESACW from the southern BUS region (250-300 µmol L −1 ) (Poole and Tomczak, 1999), indicating that ESACW is modified by aerobic mineralization on its way from the SBUS to the NBUS region. Following the nutrient signature of ESACW given by Poole and Tomczak (1999), N* ranges between −5.45 and −0.762 µmol L −1 . Assuming mineralization according to the Redfield stoichiometry under these oxygenated conditions and further assuming the absence of N 2 fixation in the SBUS implies that the N* signature of ESACW should not be further altered on its way towards the NBUS. An intrusion of ESACW would therefore instead contribute to the negative N* signal observed over the NBUS shelf. However, off the western South African coast, N* was shown to vary between 0 and 5 µmol kg −1 along the σ θ = 26.50 and 27.1 kg m −3 surfaces (Gruber and Sarmiento, 1997), referring to a depth range where ESACW occurs. This suggests that, comparable to SACW, the N* signature in ESACW might vary; however this aspect is not further addressed owing to the low density of ESACW data.
Positive N* anomalies in subsurface waters are commonly attributed to the mineralization of organic matter produced by N 2 -fixing organisms (Gruber and Sarmiento, 1997) that have N : P ratios of up to 150 (Krauk et al., 2006). The impact of mineralization is reflected in minimum O 2 concentrations that coincided with maximum N* values within the SACW fraction (Fig. 9). Positive N* anomalies in the tropical and subtropical North Atlantic are ∼ 4 µmol kg −1 at σ θ = 26.5 kg m −3 (Gruber and Sarmiento, 1997;Mahaffey et al., 2005). Here, intensive blooms of Trichodesmium spp. (Carpenter, 1983;Tyrrell et al., 2003;Capone et al., 2005) and high N : P (of up to 35) have been reported (Mahaffey et al., 2003) and are consistent with N 2 fixation as a significant input source, which can raise N* to ∼ 20 µmol kg −1 (Mahaffey et al., 2003). In contrast, the South Atlantic Gyre has low positive N* (Gruber and Sarmiento, 1997), in line with low rates of N 2 fixation (Mahaffey et al., 2005). N 2 fixation is facilitated by P and the availability of micronutrients, e.g. iron (Fe) (Mills et al., 2004). The low N 2 fixation in the South Atlantic has been attributed to a lack of Fe supply rather than P availability (Moore et al., 2009). This is supported by the negative N* observed in the surface water offshore of Kunene (Fig. 9) caused by P concentrations that ranged between 0.17 and 0.28 µmol kg −1 and exceeded N concentrations, resulting in surface water N* deficits. The very low N concentrations further suggest that atmospheric deposition that is assumed to be another significant N source to the ocean (Duce et al., 2008;Baker et al., 2013) is negligible. However, Sohm et al. (2011) observed N 2 fixation within the ABFZ (∼ 13-15 • S) coinciding with decreased thermocline δ 15 N − NO 3 values and elevated dissolved Fe and cobalt surface concentrations (Noble et al., 2012), which are important micronutrients for marine diazotrophic cyanobacteria (Saito et al., 2002(Saito et al., , 2004. The studies of Sohm et al. (2011) andNoble et al. (2012) were performed from November to December 2007 and hence ∼ 4 months prior to the MSM07/2b-3 expedition in March 2008 that found the positive N* anomaly in SACW. This time period is comparable to the time lag of ∼ 2 months that was observed between the occurrence of Trichodesmium spp. and the response in N* (Singh et al., 2013). Although the N 2 fixation rates were relatively low (22-85 µmol N m −2 d −1 ) (Sohm et al., 2011), it was proven that N 2 fixation occurs north of the ABFZ, even with high nitrate concentrations (∼ 20 µmol L −1 ) in surface water. This finding seems to challenge the traditional paradigm of high N 2 fixation activity being restricted to oligotrophic regions. However, there is growing evidence based on laboratory and field studies showing that N 2 fixation occurs under nutrientrich conditions (Sohm et al., 2011;Knapp, 2012;Knapp et al., 2012;Subramaniam et al., 2013). It suggests that Fe might be the factor limiting N 2 fixation at the ABFZ. In contrast to the Sahara dust plumes influencing the North Atlantic, the eastern South Atlantic experiences much lower aeolian input (Jickells et al., 2005;Mahowald et al., 2009). Dust plumes off Namibia are channelled by dry river beds and appear to be much more restricted to near-shore regions (Eckardt and Kuring, 2005). However, dust transport from southern Africa to the South Atlantic has been reported for austral spring Sarthou et al., 2003), further supported by , who found that the main air mass transport to the South Atlantic peaks during austral spring and summer, indicating that N 2 fixation initiated by atmospheric Fe input is expectable during this time of the year, in line with the results of Sohm et al. (2011).
Although not as pronounced as in March 2008, N* was also elevated during December 2009 and February 2011 within the SACW fraction, suggesting that N 2 fixation is apparent albeit of lower magnitude. Alternatively to atmospheric Fe sources, hypoxic bottom waters of the Angolan and Namibian shelf reveal high concentrations of dissolved and particulate Fe (Bowie et al., 2002;Noble et al., 2012). It suggests that upwelling of these bottom waters along the Angolan and Namibian coast serves as an Fe source analogue to the California and Peru upwelling system (Johnson et al., 2001;Bruland et al., 2005). In view of the findings on diazotrophy under nutrient-rich conditions, we assume that N 2 fixation is a feasible input source that caused the observed N excess in the SACW in 2008. Furthermore, methodological improvements (Mohr et al., 2010;Großkopf et al., 2012) and the broadening range of identified species involved in N 2 fixation (Moisander et al., 2010;Zehr, 2011) suggest that rates and regions of N 2 fixation may have been considerably underestimated so far.

NBUS -a P* source for the South Atlantic
Upwelling of N-deficient water in 2009 and 2011 caused lowest N* at 5 m depth close to the coast (Fig. 8d-f), indicating that a relative P surplus surfaces and is advected offshore into the open ocean with modified upwelling water. This should stimulate N 2 fixation in the adjacent hemipelagic ocean (Deutsch et al., 2007), but experimentally determined rates of N 2 fixation in the NBUS were very low to not detectable (N. Wasmund, personal communication, 2012). We calculated the P excess (P*) within the euphotic zone (0-20 m) that is exported offshore for the Walvis Bay (23 • S) transect (Fig. 10). Roughly 300 km offshore, beyond the continental slope, P* was close to zero in March 2008 (P* = 0.007 ± 0.09 µmol kg −1 ), suggesting that the NBUS at that time was not a regional source of P for the oligotrophic subtropical South Atlantic. During the other cruises, the NBUS was a relative P source, with P* values of 0.3 ± 0.01 µmol kg −1 . These observations are in line with Staal et al. (2007) and Moore et al. (2009), who reported surface P* = 0.15-0.30 µmol kg −1 within areas of the subtropical gyre that are influenced by the advection of water masses transported by the Benguela Current. Along with our findings, this implies that the coastal upwelling system over the shelf is a P* source to the South Atlantic most of the time. As shown before, the bottom water off Walvis Bay was low in O 2 , independent of the upwelling situation (Fig. 8), leading to strongly elevated P and reduced N maxima off Walvis Bay (Fig. 5). This implies that especially the mud belt of the shallow central Namibian shelf is a region of continuous +P* generation via N loss and P efflux. We assume that the inter-annual variability of P* in the surface (Fig. 10) depends mainly on the magnitude of +N* in the SACW that is likely produced by N 2 fixation north of the ABFZ. We further presume that N 2 fixation in this region is in turn linked to the NBUS by the export of +P* into the South Atlantic Ocean and its advection along with the major surface currents to the Angola Gyre and ABFZ region (Fig. 11).

Conclusions
Our data measured during the cruise in March 2011 show a mean C : N : P : -O 2 ratio that is close to the Redfield stoichiometry. Over the mud belt of the Namibian shelf, pore water fluxes lowered the C : N : P : -O 2 ratio to 106 : 16 : 1.6 : 138. N losses further increased C : N and decreased N : P ratios in restricted regions where O 2 concentra- tions dropped below 20 µmol kg −1 . Additional nutrient data measured during two cruises in 2008 and 2009 reveal an inter-annual variability of N excess within the SACW that flows from the north into the NBUS with highest N* values observed in 2008. The degree to which the N loss on the shelf is balanced by enhanced inputs of N along with the SACW controls the amount of P* that is exported from the NBUS into the subtropical South Atlantic. To better understand the role of the northern Benguela upwelling system as P source and N sink, factors controlling the occurrence of N 2 fixation at the Angola Benguela Frontal Zone need to be addressed in future studies.