3.2. Nutrient Distribution in the Strait of Gibraltar
 The spatial distribution of nitrate and phosphate along with the distinctive thermohaline properties of the water masses present in the SG is displayed in Figure 3. Vertical profiles were obtained through data interpolation. At the western entrance of the Strait, the presence of MOW is evident between 260 and 358 m (Figure 3b) as seen by its salinity (>38 in Figure 3a) and thermal signature (∼13°C). Here, the upper layer corresponds to the AI, with a characteristic salinity of about 36.46 (Figure 3a) and a temperature range between 18°C and 15°C, situated between 0 and 200 m (Figure 3b). The AMI is also distinguishable as a prominent halocline between both layers (Figure 3a) and displays a mean vertical temperature variation of 2°C (15–13°C in Figure 3b). Due to topography, its position slopes upward toward the eastern side of the Strait (Figure 3a), as the AI accelerates in, and especially east of CS as it entrains the Mediterranean water. The Mediterranean layer thus occupies a larger volume, extending to about 120 m below the surface of the water column (Figures 3a and 3b).
Figure 3. Distribution of (a) salinity, (b) potential temperature (θ, °C), (c) nitrate (μmol kg−1), and (d) phosphate (μmol kg−1) in the Strait of Gibraltar. Contouring was done through the DIVA Gridding interpolation.
Download figure to PowerPoint
 Nitrate and phosphate distributions were characterized by an increasing vertical gradient everywhere in the channel, which reflects the confluence of water masses (Figures 3c and 3d). In particular, nitrate exhibited minimum values (∼1.2 μmol kg−1) within the upper 50 m of the water column in the westernmost part of the Strait and increased toward the east to reach concentrations of about 3 μmol kg−1 (Figure 3c). Overall, the average NO3− concentration in the AI penetrating into the Mediterranean was 3.1 ± 0.3 μmol kg−1 (n = 54 in waters with S < 37.8 at station 8; see Table 2). Much higher levels of nitrate (∼9 μmol kg−1) were detected in the MOW, with the maximum concentration being observed above the seafloor in the easternmost part of the Strait (Figure 3c). The average NO3− concentration in the MOW was 8.8 ± 0.6 μmol kg−1 (n = 38 in waters with S > 37 at station 6; see Table 2).
 In contrast, phosphate concentrations were constant in surface waters at a value of about 0.3 μmol kg−1 (Figure 3d). Throughout the whole channel, the AI contained lower phosphate levels than those found in the MOW and average PO43− concentrations of 0.25 ± 0.04 μmol kg−1 and 0.49 ± 0.04 μmol kg−1 were found within each layer, respectively (Table 2). A slight reduction in phosphate was observed toward the east in the upper 100 m of the water column (Figure 3d). This may be related to the different PO43− concentrations found in the two basins connected by the SG; phosphate levels are much higher in surface waters of the GoC [Navarro et al., 2006] than in the Alboran Sea [Karafistan et al., 2002]. Conversely, nitrate is generally higher in the Alboran [Ramírez et al., 2005; Prieto et al., 2009], which explains the higher concentrations detected in surface waters at the eastern entrance of the Strait (Figure 3c). These observations are also supported by recent modeling results [Skliris and Beckers, 2009]. The low variability of the average nutrients concentration (Table 2) reflects the insignificance of seasonality on the nutrient pattern in the SG.
3.4. Nutrient Balance in the Mediterranean Sea
 Most nutrients budgets available for the Mediterranean Sea imply that its oligotrophy is due to the anti estuarine circulation in the SG [Béthoux et al., 2002; Ibello et al., 2010; Ribera d'Alcalà et al., 2003]. This assumption is also applied to nutrient balances in the western Mediterranean (WM) [Schroeder et al., 2010]. Similarly, the ultraoligotrophy of the Levantine basin has been attributed to the anti estuarine regime in the Straits of Sicily [Krom et al., 2004, 2010]. In contrast, numerical models suggest that the circulation pattern in the Straits is not sufficient to explain the trophic status of the EM [Crispi et al., 2001].
 Still, estimates of nutrient exchange through the SG are largely discrepant [Coste et al., 1988; Béthoux et al., 2002; Dafner et al., 2003; Ribera d'Alcalà et al., 2003; Macías et al., 2007]. This is especially true for phosphate transport, which was last assessed a decade ago [Dafner et al., 2003]. The analysis performed here helps put constraints on these estimates, as it is based on extensive measurements recently collected in the area, combined with the in situ monitoring of water mass transport. This approach reduced the uncertainty that arises when disperse measurements are combined for computation of the exchange rate.
 The resulting transport rates of nitrate and phosphate from the Mediterranean to the Atlantic (Table 2) fall within the range of others reported in the region. Nonetheless, our results also show that Atlantic waters crossing the SG are not nutrient depleted (Figures 3c and 3d and Table 2). In fact, the AI has a nutrient content that is high enough to sustain phytoplankton growth, exceeding phytoplankton semisaturation constants for nitrate (Ks = 0.5 μM) [Eppley et al., 1969] and phosphate (Ks = 0.05 μM) [Davies and Sleep, 1989]. Furthermore, phosphate is in excess with respect to nitrate within this layer, as indicated by the N:P ratio found on both sides of the Strait (11:1 and 12:1 at stations 6 and 8 respectively). This molar ratio is well below the Redfield stoichiometry, which is only surpassed inside the MOW (17.5:1 at station 6).
 This suggests that processes occurring within the WM impact the rate of nutrient transport from the Strait of Sicily to Gibraltar. Specifically, Karafistan et al.  and Ribera d'Alcalà et al.  report that surface waters passing through the Straits of Sicily (0–50 m) are depleted in nitrate and phosphate; very different from the pattern described in the SG.
 However, interpretation of deviations in the Redfield ratio is difficult, as this coefficient is highly nonlinear, especially at low nutrient concentrations, and therefore not conservative. The amount of available fixed nitrogen can be strongly affected by biological processes, such as denitrification and nitrogen fixation by diazotrophic organisms, whereas phosphorus concentrations are controlled by the balance between river inputs and loss to sediments. Hence, the quasi-conservative parameter N* has been suggested as an alternative to explain nutrient relationships in certain ocean regions as it depicts the net influence of the aforementioned processes on nitrate distribution in a consistent manner [Gruber and Sarmiento, 1997]. N* is defined as the linear combination of nitrate and phosphate that eliminates most of the effect of nitrification of organic matter, with the remaining variability being primarily caused by the combined effect of denitrification and nitrogen fixation plus atmospheric deposition and river inflow.
 The spatial pattern of N* was analyzed in two representative regions of the WM and EM using data collected in September 2008 and gathered by the SESAME project (Figure 1). N* distribution along a longitudinal transect connecting the GoC and the South WM and sampled during the SESAMEII cruise (Figures 1 and 4a), shows that surface waters are characterized by negative N* values. In particular, N* remains negative from the surface to a depth of 700 m in the GoC, which is an indication of a deficiency in nitrate relative to phosphate in a vast portion of the mesopelagic zone [Navarro et al., 2006]. The zero N* contour, which represents the Redfield stoichiometry, rises toward the surface in the SG, with a minimum (∼−3 μmol kg−1) appearing in surface waters at its western entrance where the lowest nitrate concentrations are always detected (Figure 3c). A secondary N* low is found within the upper 250 m of the Alboran Sea, where substantial rates of primary production are regularly found [Skliris and Beckers, 2009] and a lower than Redfield N:P ratio is observed [Ramírez et al., 2005]. This distribution confirms that the entire photic zone of the transition area between the Atlantic and the WM is nitrate-deficient with respect to phosphate. In contrast, the MOW exhibits positive N* values close to 4 μmol kg−1, which is in agreement with the N* levels of the LIW that is the main component of this water layer [Minas et al., 1991]. The Mediterranean tongue carrying the high N* signal spreads into the GoC (Figure 4a) and, near 8°W, it appears in two well-reported veins of MOW flowing at depths of around 700 m and 1100 m [Ambar and Howe, 1979]. The near zero N* values observed in deeper waters of the Alboran Sea (Figure 4a) can be attributed to the influence of mesoscale processes, such as the intense cyclonic vorticity generated by the entry of the AI that causes the upwelling of nutrient enriched deep waters [Skliris and Beckers, 2009]. As a result, the vertical flux of organic matter produced by the enhanced productivity triggers bacterial growth below the photic zone, which generates nutrients with a Redfield stoichiometry [Minas et al., 1991]. It should be noted that denitrification in subsurface waters in upwelling regions cannot be ignored [Deutsch et al., 2001], which would also contribute to the decrease in N* (Figure 4a). Using all of the measurements collected in the SG during more than 3 years, the net export of N* from the Mediterranean to the Atlantic is 61.2 Gmol yr−1 (±0.8) (Table 2).
Figure 4. Spatial distribution of N* (μmol kg−1) in two regions of the (a) western and (b) eastern Mediterranean in September 2008. Data were collected during the campaigns SESAMEII and SESIL02 and are available in the SESAME project database (http://isramar.ocean.org.il/SESAMECruises/map.asp). The different water masses present in both regions are indicated by their acronyms (AI, MOW, MAW, LIW, and EMDW). Horizontal dashed gray lines in Figure 4b also denote their position in the water column in the eastern basin.
Download figure to PowerPoint
 Altogether, these results indicate that nitrogen-impoverished surface waters are transformed into phosphate-deficient bottom waters in the Mediterranean, but the deficit of phosphate is not caused by the entry of a phosphorus-depleted AI through the SG.
 This conclusion is further supported by the N* distribution in the EM (Figure 4b). In this case, data were acquired during the SESIL02 cruise carried out in the Levantine Basin (Figure 1). The upper 200 m of the water column in the region has negative N* values, confirming nitrate deficiency relative to phosphate in surface waters. This layer corresponds to the Modified Atlantic Water (MAW) that flows at the surface through the Straits of Sicily [Ribera d'Alcalà et al., 2003] and still conserves its original nitrogen deprivation. In contrast, the water column below 250 m formed by the LIW (200–500 m) and the Eastern Mediterranean Deep Water (EMDW ∼800 m), exhibits positive N* values (Figure 4b).
 Because the Mediterranean basin is a region of deep water mass formation, nutrient inputs from the bottom are unlikely and vertical modifications in the nutrient ratios must be related to external sources. Figure 5 presents a nutrient balance box model for the Mediterranean in which the quantitative contribution of sources and sinks of nutrients has been indicated. It is clear that the net nitrate export through SG (Table 2) is substantially higher than that recently measured through the Straits of Sicily, equivalent to 92 Gmol yr−1 [Schroeder et al., 2010]. Therefore, an enrichment of 47 Gmol yr−1 of NO3− must occur within the WM. A recent basin-wide study on N2 fixation has demonstrated that diazotrophy is an insignificant nitrate source in the Mediterranean [Ibello et al., 2010]. In contrast, recent studies suggest that river inputs and atmospheric deposition supply 26 and 28 Gmol yr−1 of NO3−, respectively [Ludwig et al., 2009; Markaki et al., 2010]. This implies that in the WM, 7 Gmol yr−1 of nitrate must be sequestered by sediment burial and/or denitrification. Both mechanisms are likely to occur, particularly in poorly ventilated regions of the water column (sapropel layers [Filippelli et al., 2003]) or in seafloor sediments adjacent to major rivers that allow for oxygen consumption and heterotrophic denitrification, as it has been found in certain areas of the EM [Krom et al., 2010].
Figure 5. A box model of the Mediterranean Sea nutrient balance. Average concentrations of nitrate and phosphate (in μM ± SD) in the Atlantic inflow and the Mediterranean outflow in the Strait of Gibraltar are indicated. Both layers are separated by an interface (AMI), which is identified by salinity values of 37.0 and 37.8 on the western and eastern sides of the strait, respectively. Entrainment and mixing in the AMI is given in Sv. Transports of MOW (Q2) and AI (Q1) through the strait (in Sv) along with the average salinity in each layer are provided within solid arrows that mark direction of flows. Dashed arrows denote next fluxes of nitrate (FN NO3−) and phosphate (FN PO43−) across the Straits of Sicily (data from Schroeder et al. ) and Gibraltar (this work). Fluxes and nutrient inputs and outputs are expressed in Gmol yr−1 (references given in the text).
Download figure to PowerPoint
 Also in the EM, rivers and atmospheric inputs supply 51 and 61 Gmol NO3− yr−1 [Ludwig et al., 2009; Markaki et al., 2010], which are accompanied by an input of 8 Gmol NO3− yr−1 from the Black Sea [Çolpan Polat and Tugrul, 1995]. Therefore, assuming steady state, these additional inputs into the EM suggests that there is a further removal of 28 Gmol yr−1 by both sediment burial and denitrification. This sink is lower than the previous estimate reported in the literature (37 Gmol yr−1 [Krom et al., 2010]), which was deduced by assuming an export through the Strait of Sicily of 142 Gmol NO3− yr−1.
 On the contrary, phosphate remains quite stable in the WM (Figure 5) and the export through the Straits of Sicily (4.1 Gmol yr−1 [Schroeder et al., 2010]) and Gibraltar (Table 2 and Figure 5) is almost identical. Phosphate inputs by river discharges and atmospheric deposition only represent 0.5 [Ludwig et al., 2009] and 0.55 Gmol yr−1 [Markaki et al., 2010] respectively, which results in very low rates of removal by sediments (∼0.3 Gmol yr−1). In the EM, an organic phosphorus sediment burial flux of 1 Gmol yr−1 has been indirectly inferred [Krom et al., 2004], which would not be compensated by the recent estimates of fluvial and atmospheric inputs of 1.4 and 0.6 Gmol PO43− yr−1, respectively [Ludwig et al., 2009; Markaki et al., 2010], leading to unbalanced export through the Straits of Sicily, which is not observed.
 This budget highlights the fact that phosphate inputs are not all accounted for in the EM. In the Levantine basin, some phosphorus chemical species may have been underestimated, as it is very difficult to separate labile P from nonlabile P [Krom et al., 2010]. Moreover, the contribution of the Black Sea certainly needs to be updated.
 Nevertheless, the role of the Strait of Gibraltar in the nutrient cycle of the Mediterranean is now clearer. Our data show that the Atlantic fuels the Mediterranean with phosphate. External nitrogen-enriched inputs determine the non-Redfield stoichiometry of the Mediterranean, creating a nutrient unbalance, particularly in the EM. Therefore, the input of nutrient-depleted Atlantic waters does not induce the phosphate driven oligotrophy of the Mediterranean.