The CO 2 system in the Mediterranean Sea : a basin wide perspective

The Mediterranean Sea (MedSea) is considered a “laboratory basin” being an ocean in miniature, suffering dramatic changes in its oceanographic and biogeochemical conditions derived from natural and anthropogenic forces. Moreover, the MedSea is prone to absorb and store anthropogenic carbon due to the particular CO 2 chemistry and the active overturning circulation. Despite this, water column CO2 measurements covering the whole basin are scarce. This work aims to be a base-line for future studies about the CO 2 system space-time variability in the MedSea combining historic and modern CO2 cruises in the whole area. Here we provide an extensive vertical and longitudinal description of the CO2 system variables (total alkalinity – TA, dissolved inorganic carbon – DIC and pH) along an East-West transect and across the Sardinia-Sicily passage in the MedSea from two oceanographic cruises conducted in 2011 measuring CO2 variables in a coordinated fashion, the RV Meteor M84/3 and the RV Urania EuroFleets 11, respectively. In this sense, we provide full-depth and length CO 2 distributions across the MedSea, and property-property plots showing in each sub-basin post-Eastern Mediterranean Transient (EMT) situation with regard to TA, DIC and pH. The over-determined CO 2 system in 2011 allowed performing the first internal consistency analysis for the particularly warm, high salinity and alkalinity MedSea waters. The CO2 constants by Mehrbach et al. (1973) refitted by Dickson and Millero (1987) are recommended. The sensitivity of the CO2 system to the atmospheric CO 2 increase, DIC and/or TA changes is evaluated by means of the Revelle and buffer factors.


Introduction
The Mediterranean Sea (MedSea hereafter) is a particular system with respect to its physical and biogeochemical oceanography (e.g., Tanhua et al., 2013a). The Med-Sea is a land-locked sea exporting warm and salty intermediate water into the Atlantic 20 Ocean through the shallow and narrow Strait of Gibraltar affecting the global thermohaline circulation. Apart from the shallow open thermohaline cell encompassing the whole Mediterranean basin, two closed deep overturning cells are active in the western and eastern basins, respectively. Additionally, a complex upper layer circulation is present including significant permanent and quasi-permanent eddies.
sampling, analysis and quality control are given in Tanhua et al. (2013b). The bottle data is available at the Carbon Dioxide Information Analysis Center (CDIAC) (Tanhua et al., 2012).
The DIC content was determined using a SOMMA instrument. Samples were collected in borosilicate bottles according to standard operation protocol. The precision of 10 the analysis was determined to ±0.6 µmol kg −1 by titration of several bottles filled from the same Niskin bottle. The accuracy was determined to be 2.5 µmol kg −1 by analyzing a total of 42 bottles of certified reference material (CRM, Andrew Dickson, Scripps, CA, USA, batch 108); the DIC of this batch is certified at 2022.70±0.7 µmol kg −1 . Measurements of the CRMs were also used to daily correct the temporal drift in the coulometer 15 cell; this correction was never larger than 3 µmol kg −1 .
pH was measured in cylindrical optical glass 10 cm pathlength cells with a spectrophotometric procedure (Clayton and Byrne, 1993). pH is reported in the total scale at 25 • C (pH25T). The precision was estimated to ±0.0012 by measuring replicates from the same Niskin bottle. Regarding the accuracy, the theoretical pH25T value for 20 the 108 CRM batch using the dissociation constants from Mehrbach et al. (1973) refitted by Dickson and Millero (1987) is 7.8782. Replicate pH measurements along the cruise on CRMs are lower than the theoretical value by 0.0962 ± 0.0012 pH units. We will discuss this point in Sect. 3.2.
Samples for TA determination were collected in 600 mL borosilicate bottles and were 25 analyzed within one day after sampling. The TA samples were analyzed following the double end point potentiometric technique by Pérez and Fraga (1987) and Pérez et al. (2000). Measurements of CRM were performed in order to control the accuracy 1450 Introduction Using the preferred set of constants reported on in Sect. 3.2 we compared measured DIC with that calculated from pH and TA using option 4 and the CO2SYS program for MATLAB (see next section). After rejecting some outliers, the mean difference between measured and calculated DIC is −0.4 ± 10 µmol kg −1 . In summary, no adjustments were done to the EF11 DIC and pH measurements, whereas the TA was reduced 5 by 11 µmol kg −1 as a consequence of the comparison between M84/3 and EF11 data. The EF11 CO 2 data will be used to discuss the CO 2 system properties in water masses of the passage between Sardinia and Sicily, Sect. 4.1.1.

Internal consistency analysis for the CO 2 system in the MedSea
There is a wealth of literature regarding the internal consistency of CO 2 analysis in 10 seawater and its dissociation constants, see the review by Millero (2007), but to the best of our knowledge the last study about this issue was published in 2002 compiling up to 6000 measurements from NSF/JGOFS, NOAA/OACES and DOE/WOCE cruises in the Atlantic, Pacific, Southern and Indian oceans (Millero et al., 2002), but without any CO 2 data from the MedSea. 15 Despite lacking of surface or column partial pressure CO 2 (pCO 2 ) measurements, this study presents high-quality DIC, TA and pH data covering the whole MedSea that can be used to perform the first internal consistency analysis in this area for CO 2 measurements.
By using the consensus program for seawater CO 2 calculations CO2SYS, config-20 ured for Matlab by van Hueven et al. (2011) we tested the internal consistency of our measurements for the various CO 2 constants using all options, except that for pure water only (option 8). We also changed the sulfate constants (Dickson, 1990;Khoo, 1977) and the parameterization of borate as a function of salinity (Uppström, 1979;Lee et al., 2010). The inorganic nutrients concentrations were also taken into account. 25 The option yielding minimum differences between the calculated and measured variable for the three combinations (TA-pH, TA-DIC and DIC-pH) ( son and Millero, 1987), the sulphate constant from Dickson (1990) and the parameterization of borate from Uppström (1979). Using this borate function and the sulphuric constant from Khoo (1977) yields practically the same results (not shown). Using option 13 (Millero et al., 2006), the more recent borate function by Lee et al. (2010) and either sulphate constant (Khoo, 1977or Dickson, 1990 low residuals are also obtained 5 (not shown). By using our preferred set of constants (see above) our pH, TA and DIC measurements along the M84/3 cruise are internally consistent to −0.0003 ± 0.005 for pH, 0.1 ± 3.3 µmol kg −1 for TA and −0.1 ± 3 µmol kg −1 for DIC. This combination of constants is also the preferred option in global ocean CO 2 data synthesis projects such 10 as GLODAP (Key et al., 2004) and CARINA (Key et al., 2010). At the light of these results, the bias of 0.0962 ± 0.0012 in the pH25T measurements obtained after CRM measurements for pH during M84/3 was disregarded.
The effect of uncertainties in pK 1 and pK 2 using the pH-TA combination is identical to those observed using pH-DIC. As summarized in Lee et al. (1997) when pH and 15 TA (pH and DIC) are used as input parameters, calculated DIC (TA) depends on pK 2 , which is also dependent upon the DIC/TA ratio; the lower the ratio the higher the sensitivity to pK 2 . However, only a reliable pK 1 is needed to calculate pCO 2 from the pH-TA (pH-DIC) combination; in this case the lower the DIC/TA ratio the lower the sensitivity to pK 1 . Calculated pH from the TA-DIC input combination relies on a good estimation 20 of pK 1 − pK 2 (the K 1 /K 2 ratio). The DIC/TA ratio is an indicative of the concentration of carbonates; the lower the ratio the higher the [CO 3 ] 2− concentration, which is also sensitive to pK 2 . Lacking measured pCO 2 data we are not able to check the dependence on K 1 in the calculation of pCO 2 (pH-DIC or pH-TA inputs) or pH (pCO 2 -DIC or pCO 2 -TA inputs). Introduction MedSea pointing to a decreasing trend eastwards for TA (Fig. 3a) and on the contrary an increasing trend eastwards for DIC and pH ( Fig. 3b and c) which seem to be related with the eastwards salinity increase (Fig. 3 in Tanhua et al., 2013a), DIC/TA ratio decrease ( Fig. 10h) or [CO 3 ] 2− increase. The pH residuals are a function of pK 1 − pK 2 , but mostly dependent on K 2 , while the TA or DIC residuals are a function of pK 2 5 (Lee et al., 1997;Millero, 2007). The relation between pH and TA residuals is linear ( Fig. 4) but two regression lines are detected depending on salinity. Samples with salinity higher than 36.1 have a higher slope (−0.0015 ± 4.5 × 10 −6 , r 2 = 0.9937, n = 725) than fresher ones (−0.0020 ± 2.9 × 10 −5 , r 2 = 0.9945, n = 28). Including measured pH, TA and DIC data from five CARINA cruises (Key et al., 2010) between the Strait of 10 Gibraltar and 24 • W and 40 • N to 24 • N the same tendency is detected (data with salinity < 36.1, slope = −0.0022 ± 8.7 × 10 −6 , r 2 = 0.9919, n = 671; data with salinity ≥ 36.1, slope = −0.0016 ± 6.7 × 10 −6 , r 2 = 0.9977, n = 182). This result points to the need for a revision of the K 2 parameterization for waters with high salinity and low DIC/TA ratio, as in the MedSea, as the CO 2 chemistry is more sensitive to K 2 .

15
In summary, the preferred option for CO 2 constants for global studies on internal consistency (CO 2 constants from Mehrbach et al., 1973 refitted by Dickson andMillero, 1987;sulphate constant from Dickson, 1990; total borate from Uppström, 1979) also applies for MedSea waters. However, further studies are needed, with an additional fourth measurement, either pCO 2 or [CO 3 ] 2− (Byrne and Yao, 2008) to confirm the in-20 ternal consistency of K 1 and K 2 with the measurements in the peculiar MedSea waters.

Distribution of CO 2 species in the Mediterranean Sea
The general aim of this section is to provide, for the first time, a large-scale distribution of measured and derived CO 2 properties along the full length of the Mediterranean Sea and corresponding sub-basins, being complementary to the physical (temperature and Introduction by Tanhua et al. (2013a). Water mass characterization will be commented using vertical distributions and property-property plots for the different sub-basins.

15
AW also presents a low TA signal evolving from values < 2560 µmol kg −1 in the western basin to higher ones in the Levantine Basin and Ionian Sea (< 2600 µmol kg −1 ) ( Fig. 5). The same trend was seen in DIC evolving from 2250 µmol kg −1 in the western to 2270 µmol kg −1 in the Levantine Basin (Fig. 6). But the opposite was found for pH25T, which increases from 8 to 8.014 (Fig. 7).

20
The MedSea is considered one of the most oligotrophic areas in the world (Azov, 1991) with low phytoplankton biomass and primary production decreasing eastward, confirmed by in situ measurements (Ignatiades et al., 2009;Moutin and Raimbalut, 2002;Siokou-Frangou et al., 2010), satellite data (Bosc et al., 2004;Volpe et al., 2007;D'Ortenzio and Ribera d'Alcalà, 2009) and modelling studies (Crise et al., 1999(Crise et al., , 2001 of nutrients into surface waters during most of the year; this is intensified when the water column is thermally stratified during spring and late fall. Surface Chlorophyll a (Chl a) and primary production data during the M84/3 cruise are presented in Rahav et al. (2013) with a clear westward increasing gradient, and values ranging from 0 to 0.4 µg L −1 for Chl a, and 0.2 to 15 µgC L −1 d −1 for primary production. The period of the 5 M84/3 cruise, April coincides with the end of the winter-spring bloom (D'Ortenzio and Ribera d'Alcalà, 2009;Lazzari et al., 2012). The gathered biogeochemical data during the M84/3 cruise shows that the upper 50 dbars are supersaturated in oxygen (AOU < 0) with more negative values in the western basin pointing to higher primary production as confirmed by direct measure-10 ments (Rahav et al., 2013). While in the western basin inorganic nutrients are still available, in the eastern basin they are depleted, especially phosphate (see corresponding figures in Tanhua et al., 2013a). For waters with Pres < 60 dbars, there is no clear relationship between AOU and either DIC, TA or pH25T, except in the western basin where pH25T relates with AOU (pH25T = −0.0015 ± 0.00008 × AOU + 7.97 ± 0.001, r 2 = 0.83 15 p < 0.001), pointing to a modulation of pH by primary production.
(iii) Levantine Intermediate Water (LIW): this warm and high salinity water is formed in the Levantine Basin, mainly in the Rhodes Gyre area. During the M84/3 cruise LIW was clearly detected by the salinity maximum around 200 dbars in the eastern basin (Tanhua et al., 2013a), with TA > 2620 µmol kg −1 , DIC > 2310 µmol kg −1 but no 20 clear signal in pH which varies between 7. . In the western basin LIW was found around 500 dbars as a maximum in salinity, corresponding to TA (≈ 2590 µmol kg −1 ) (Fig. 5), maximum in DIC (≈ 2330 µmol kg −1 ) (Fig. 6) and minimum pH25T (≈ 7.9) (Fig. 7). In the eastern basin, the maximum layer of remineralization indicated by the oxygen minimum (see Fig. 6 in Tanhua  Levantine Basin below 1500 dbars the water column was occupied by a more saline (Tanhua et al., 2013a) water than in deep layers of the Ionian basin. In the Ionian Sea, recent EMDW was characterized by higher DIC (Fig. 6 has an Aegean Sea origin produced during the EMT (Eastern Mediterranean Transient) event while in the Ionian Sea EMDW has a more recent Adriatic Sea component produced after the EMT (Hainbucher et al., 2013). The upper panels in Figs. 5, 6 and 7 show the distribution of CO 2 species in the through the centre of the Ionian Sea to the Adriatic Pit. In this section, below 3000 dbars 10 DIC ( Fig. 6) increased towards the bottom, a feature not seen neither in salinity or TA (Fig. 5). A slight, but hard to detect, decrease in pH25T ( Fig. 7) was also noted. The tracers distributions, CFC-12 and SF 6 (Stöven, 2012), also increase towards the bottom. These results point to a recent addition of EMDW formed in the Adriatic Sea. In the centre of the Ionian Sea the broad pH minimum between 125 and 1000 dbars 15 extends trough LIW and the upper part of EMDW, also covering the layer of maximum remineralization centred at 1000 dbars as indicated by the oxygen minimum (Tanhua et al., 2013a).
(v) Western Mediterranean Deep Water (WMDW): being fresher and colder than EMDW, the deep layers below 1500 dbars in the western MedSea during M84/3 were 20 quite homogeneous in the physical (potential temperature and salinity) properties but a closer look showed two distinct WMDW, a younger one as noted by the CFC-12 distribution (Stöven, 2012) and saltier than the WMDW immediately above 2000 dbars (Hainbucher et al., 2013). Bottom (younger) WMDW also presented higher DIC values by about 7 µmol kg −1 , lower pH25T values by about 0.004, and higher TA values by 25 about 4 µmol kg −1 , i.e., within the accuracy limits of the latter two variables (Figs. 5-7). Another outstanding feature in the vertical distribution of CO 2 variables was the pH25T minimum (Fig. 7) around 500 dbars in the Levantine and Ionian basins. In the Levantine basin this minimum was coincident with the oxygen minimum ( Fig. 6 (Fig. 6). Whereas in the Ionian Sea the pH25T minimum/DIC maximum (500 dbars) was shallower than the oxygen minimum that was situated around 1000 dbars. All these extreme layers were not coincident with LIW, the salinity maximum was shallower in both basins, Levantine and Ionian. However, in the western basin, the LIW signal was coincident with the layer of highest remineralization 5 as indicated by the DIC maxima and the oxygen/pH25T minimum. During the M84/3 cruise we sampled the interior of the Tyrrhenian Sea: in the upper layer low salinity (Tanhua et al., 2013a) and TA values (Fig. 5) point to AW. Around 400 dbars the maximum in salinity, TA, DIC and pH25T (Figs. 5-7) are associated with LIW entering trough the Sicily channel; Tyrrhenian Deep Water (TDW) formed from the 10 mixing of LIW and WMDW was detected below the TA maximum associated with LIW and bottom water with lower TA and pH25T and higher DIC values . This bottom water corresponds to WMDW that has started to cascade into the Tyrrhenian Sea from the western basin in 2011, also higher in tracer concentrations (Stöven and Tanhua, 2013).

15
As a reference for future studies and calculations in Sect. 4.1.2, Table 2 shows the mean and standard deviation values for salinity and different CO 2 species in the Med-Sea sub-basins and Atlantic waters west of the Strait of Gibraltar as sampled in 2011 during the Meteor M84/3 cruise by different depth intervals.
The direct comparison of our CO 2 data with previously published CO 2 data, either 20 DIC, TA or pH, in order to detect trends and variability is out of the scope of our manuscript. We do however make a rough comparison in order to detect possible biases in the historic data set as compared to ours and vice versa. Qualitatively, our TA and DIC data in the western MedSea seems to be 5-10 and 40-70 µmol kg −1 lower, respectively, than those shown in Millero et al. (1979) from a cruise in 1976. This result is 25 counterintuitive due to the expected increase in DIC due to input of anthropogenic carbon (C ANT ). However, the data from Millero et al., (1979) originates from before the use of certified reference materials (CRMs), so that the accuracy of these early data can be questioned. Our TA and DIC are however more comparable to those from the Almofront Introduction  Schneider et al. (2010). The TA data are comparable but our DIC are roughly 10 µmol kg −1 higher, which could be a real signal considering the 10 yr between the measurements and the uptake of anthropogenic carbon. The VECTOR cruise in 2007 sampled 8 stations in the eastern and western Med-Sea for pH and TA, whereas DIC was calculated (Rivaro et al., 2010). Examining data from the western basin below 700 dbars, the TA data are comparable to each other whereas the DIC and pH values reported by Rivaro et al. (2010) are higher than ours by 10 µmol kg −1 and by 0.01, respectively. In the eastern basin, our data below 700 dbars for pH and TA are comparable but our DIC seems 20 µmol kg −1 higher. More recent CO 2 data are reported from the BOUM cruise in 2008 (Pujo-Pay et al., 2011); their DIC data 15 seem higher than ours by about 10 µmol kg −1 in both the eastern and western basins.

Sardinia Sicily Passage
The vertical distributions of physical properties in the Sardinia Sicily passage, already analysed in the last decade of last century (Astraldi et al., 2002), indicate that the passage is a crucial chokepoint where almost all MedSea waters can be intercepted (see 20 a review of the general characteristics of the transect in Fig. 5 of Astraldi et al., 2002). This section focuses on the description of the oceanographic settings in the Sardinia Sicily passage as sampled during the Urania EF11 cruise. The Sardinia Sicily transect, for its local character, was not included in the overall MedSea description done by Tanhua et al. (2013a). The main water masses crossing the passage have been 25 identified according to their physical characteristics using potential temperature and salinity ( Fig. 8a and b), AOU, (Fig. 8c) and CO 2 variables ( Fig. 8d- Fig. 8f); given its high density tEMDW cascades at greater depths to the interior of the Tyrrhenian Sea, contributing to the formation of the Tyrrhenian Deep Water, TDW, and provoking an intense mixing that dilutes its original physical properties (tEMDW is significantly warmer and saltier than TDW), in contrast the chemical properties of these two deep water masses remain very similar (Fig. 8).

Derived variables
We used measured TA, DIC and inorganic nutrients data to calculate the degree of 20 saturation (Ω) for calcite and aragonite using the MATLAB version for CO2SYS with the preferred set of constants (see Sect. 3.2). The whole MedSea is supersaturated (Ω 1) with respect to both calcite and aragonite throughout the whole water column. Figure 9 shows the vertical distribution of Ω,  Kleypas et al. (1999) only the eastern basin in the upper 200 dbars presents adequate to optimal conditions for the development of coral reefs with Ω-Ar > 3.5. This work expects to be a benchmark for future studies predicting the effects of acidification in the MedSea from the chemical and biological point of view. Several projects, finished (e.g., EPOCA) or on going (e.g., BIOACID, MedSeA), try to decipher the com-5 bined effect of acidification and warming on the highly adapted MedSea calcareous and non-calcareous organisms. Because of being too small but too complex in its circulation and biogeochemistry the MedSea is usually not included in global projections models (e.g., Ricke et al., 2013) and a specific ocean-biogeochemistry model needs to be used. Here we provide a basin wide perspective for the CO 2 chemistry in the 10 MedSea as a basis for future studies and projections.
In order to quantify the ability of MedSea waters to resist changes in the CO 2 system we calculated the buffer factors formulated by Egleston et al. (2010). The six factors proposed in their Table 1 quantify the sensitivity of aqueous CO 2 [CO 2 ] (γ i ), protons [H+] (β i ) and the carbonate saturation state Ω (ω i ) to changes in DIC and 15 TA (subindex i ) when the other parameter is kept constant. The calculations were performed with the output from CO2SYS at in situ conditions with the preferred set of constants and measured DIC and TA as inputs. The buffer factors have dimensions of mol kg −1 × 10 −3 of seawater (mmol kg −1 hereinafter). See the formulation in Table 3 as several editing errors were detected in the original equations from Egleston 20 et al. (2010). The buffer factors γ i express the fractional change in CO 2 when DIC changes at constant TA (air-sea CO 2 exchange) or when TA changes at constant DIC (additions of strong acid or base). The traditional Revelle factor, R, is directly calculated as R = DIC/γ DIC . The buffer factor β TA expresses the fractional change in hydrogen 25 ion concentration or activity when TA changes at constant DIC (addition of a strong base or acid). The factor β DIC is identical to −γ TA .   Table 3. As the DIC/TA ratio approaches one the buffer capacities are reduced since [CO 2 ] is more sensitive to temperature changes at this DIC/TA 15 ratio. MedSea waters in the eastern basin have lower DIC/TA ratios than in the western basin ( Fig. 10h) and in general, MedSea waters present lower DIC/TA ratios compared to Atlantic waters. The highest ratios, i.e. higher [CO 2 ] sensitivity to temperature, for the M84/3 cruise were reported west of the Strait of Gibraltar and associated with the Mediterranean Overflow, DIC/TA ratio > 0.9 (Fig. 10h).

20
From the vertical distributions in Fig. 10 we can conclude that: -MedSea water is generally more resistant to changes in DIC and/or TA than Atlantic waters to the west of the Strait of Gibraltar, since all buffer factors are lower in the Atlantic.
-The CO 2 system in the western basin is in general more sensitive to changes in 25 DIC and/or TA than the eastern basin, i.e., buffer factors are lower in the western basin.
OSD 10,2013 The CO That is, the sensitivity of CO 2 concentration to changes in DIC at constant TA (air-sea CO 2 exchange) is the highest, followed by the change in pH due to the air-sea CO 2 exchange (equivalent to the change in CO 2 due to a strong acid/base addition), then changes in pH due to strong acid/base addi-5 tions. Being |ω DIC | and ω TA the highest buffer values means that MedSea waters are able to buffer more easily than Atlantic waters changes in the saturation state due to additions of a strong acid or base and the air-sea CO 2 exchange.
-By examining the Revelle factor it is obvious that for a given change in CO 2 the relative change in DIC for MedSea waters is higher than in the adjacent Atlantic 10 waters. That means that MedSea waters have the ability to store more anthropogenic carbon (C ANT ) than Atlantic waters. Within the MedSea the eastern basin is more prone to absorb more C ANT for a given CO 2 increase than the western basin.
The impact of the three main anthropogenic processes altering the surface CO 2 chem-15 istry in MedSea waters is given in Table 4, which shows absolute changes in aqueous CO 2 ([CO 2 ]), pH and the aragonite saturation (Ω-Ar) from the initial surface values in Table 2; the different sub-basins are treated separately. A C ANT input through air-sea exchange (affecting DIC at constant TA) increasing DIC by 10 µmol kg −1 will lead to a increase of [CO 2 ] from 0.56 to 0.78 µmol kg −1 , a pH decrease of around 0.016 units 20 in MedSea waters, less than the pH decrease in Atlantic waters (0.020) for the same perturbation; the saturation state is reduced by about 0.1 units, but a reduction of at least 0.9 is required for deep waters to become under-saturated in aragonite. A hypothetical decrease in calcification, resulting in a DIC increase of 10 µmol kg −1 and double in TA, will compensate the pH and Ω decrease due to the input of C ANT and roughly 25 half the corresponding increase in [CO 2 ]. Within the MedSea changes in [CO 2 ] and pH increase westwards being smaller than in Atlantic waters. An increase in salinity of 0.03 will cause minor changes in the CO 2 chemistry. 10,2013 The CO  The buffer factors here provided are useful to predict changes in the CO 2 chemistry as a function of DIC and/or TA changes due to different processes, natural and/or anthropogenic. However, it is relevant to relate the increase of C ANT in the ocean to the increasing partial pressure of CO 2 (pCO 2 ) in the atmosphere instead of using a DIC increase as reference. Table 5 shows the absolute changes in DIC, pH and Ω-Ar from 5 initial states in Table 2 as a consequence of prescribed pCO 2 increases. Note that a pCO 2 increase of 25 µatm causes changes in DIC by about 10 µmol kg −1 , as already presented in Table 4. The main result in Table 5 is that as the Revelle factor in warm, salty and high TA waters is lower (Table 3) the change in DIC for a given pCO 2 increase is higher and consequently the changes in Ω-Ar and pH. Therefore, MedSea waters 10 are more sensitive to DIC changes induced from the pCO 2 increase in the atmosphere (C ANT increase) despite having higher buffer factors.

Water mass characterization
Up to our knowledge no work has presented the chemical characterization with regard to CO 2 variables for water masses in the western and eastern MedSea sub-basins.

15
Here we present property-property plots for general physical and chemical variables, typical potential temperature-salinity, AOU-salinity and silicate concentration (SiO 2 )salinity (Figs. 11 and 13) and the CO 2 variables, TA, DIC and pH25T vs. salinity, SiO 2 and AOU (Figs. 22 and 14). We separated the western and eastern basins showing with colours the different sub-basins comprised. Some special characteristics, mainly  The Aegean Sea was early recognized as a source of dense waters (Nielsen,25 1912) but not dense enough to contribute to the EMDW (e.x., Wüst, 1961). This traditional view changed after the EMT event was detected (Roether et al., 1996). 10,2013 The CO  Kasos Strait (both in the southern Aegean Sea). In the upper 100 dbars high salinity identifies LSW (Levantine Surface Water) entering the Aegean Sea through the eastern Cretan arc straits (Fig. 11a). This LSW is also identified by the highest pH25T values measured in the M84/3 cruise (> 8.01, Fig. 12h) and also high TA and DIC values (≈ 2620 and ≈ 2280 µmol kg −1 , respectively, Fig. 12a and d). At In the upper water column, < 100 dbars, AW and LSW can be distinguished using salinity (Fig. 11a), LSW presents salinity > 38.9 and less variable values than AW with salinity ranging from 38.58 to 38.9 decreasing westwards ( Fig. 11a and Fig. 3 20 in Tanhua et al., 2013a). TA, DIC and pH25T values much higher in LSW can also differentiate them (Fig. 12a, d and g). Interestingly, LSW presents pH25T values higher, and AW lower than 8.01 for waters with AOU < 0 ( Fig. 12i). recognized by minima of SiO 2 associated with a salinity maximum (Fig. 11c) and high TA (≈ 2630 µmol kg −1 , Fig. 12a) values.

OSD
In the deep layers below 1000 dbars within the Levantine basin, the salinity minimum and then a pronounced temperature-salinity inversion (temperature and salinity increase) typically detected during the EMT ( Adriatic EMDW to the west of the East Mediterranean Ridge is differentiated from Aeg-EMDW mainly with AOU and SiO 2 ( Fig. 11b and c), both being lower in the Adriatic EMDW, which also contains higher transient tracer values (Stöven and Tanhua, 2013), indicating a more recent origin. It is difficult to separate them us-15 ing TA, DIC and pH25T, both EMDW present similar values, Adriatic EMDW is slightly higher in DIC (around 5 µmol kg −1 ) and TA (around 4 µmol kg −1 ) and lower in pH25T (0.005 pH units) than Aeg-EMDW (Fig. 12c, f and i) (blue and black lines in Fig. S1).
The salinity minimum around 1000 dbars and below the temperature-salinity in-20 version previously commented provoke a sort of hook in the relationship between DIC-TA-pH25T vs. AOU or SiO 2 (red lines in Fig. S1) for deep waters in this basin. DIC and TA decrease with decreasing SiO 2 or AOU until their corresponding maxima (salinity minimum) and then towards the bottom DIC and TA increase with decreasing AOU (increasing salinity). pH is the opposite, it decreases with 25 AOU increasing until the AOU maximum (salinity minimum) and then pH increases and AOU decreases (Fig. S1). 10,2013 The CO (Fig. 11a). During M84/3 LSW in the Ionian Sea was found in the stations to the western part of the Hellenic Trench (Sts. 299, 300, 301) (Fig. 1). Below the surface waters, two intermediate water masses, LIW 5 and CIW can be distinguished. LIW with temperature and salinity maxima around 200 dbars is also associated with a TA and DIC maximum ( Fig. 12a and d). The mixing between LIW and AW produces an exponential relationship between TA and DIC vs. salinity ( Fig. 12a and d). Deviations from this relationship towards lower TA and DIC values correspond to LSW in this basin. The other intermediate 10 water detected, CIW, mixes with LIW south of Crete, in the eastern basin was found in stations 299, 300 and 301 to the west of the Antikithera strait. This water mass with also very high salinity, presents an even higher TA than LIW (Fig. 12ac). For example in the case of the TA vs. SiO 2 plot (Fig. 12b) CIW is detected as a deviation towards higher values from the typical TA-SiO 2 relationship. CIW is 15 also seen as a hump between 0 to 30 µmol kg −1 AOU over the LIW data (Fig. 12c).

OSD
Towards the bottom (please see the corresponding zoomed temperature-salinity plots in Hainbucher et al., 2013), the EMT temperature-salinity inversion is eroded in the Ionian Sea. Post-EMT EMDW occupies the layer below 1500 dbars, it presents lower AOU and SiO 2 values than EMT Aeg-EMDW ( Fig. 11b and Fig. S2). These younger bottom waters in the Ionian Sea explain the DIC increase (Fig. 6), AOU and pH25T (Fig. 7) decrease towards the bottom of the Ionian Sea (compare red 25 line with the blue and cyan lines in Fig. S2).
The hook structure in the relationships DIC-TA-pH vs. AOU for deep waters in the Levantine basin (Fig. S1)  as clearly associated with the salinity minimum as in the Levantine basin. Below 500 dbars DIC and TA decrease with increasing AOU, from the AOU maximum TA practically stays constant and DIC increases. The very recent input of bottom Ad-EMDW is noticed in the pH vs. AOU relationship, from 500 dbars pH decreases with AOU increasing, from the AOU maximum pH increases (AOU decreases) but 5 close to the bottom both pH and AOU decrease (Fig. S2).
(d) Adriatic Sea (stations 310 and 313): No CO 2 data was taken in the western part of the Otranto Strait (Station 311). Therefore only two stations were sampled in the Adriatic Sea for CO 2 variables, one in the Adriatic Pit (313) and one to the east of the Otranto Strait (310). One 10 of the densest deep waters during the M84/3 cruise was sampled in the Adriatic Pit at 1100 dbars (Fig. 13a), but this water remains confined as the Otranto Strait is only 800 m deep (Hainbucher et al., 2013). Waters in the Adriatic present high TA values despite a moderate salinity (Fig. 12a), they are out of the general linear relationship between TA and salinity due to the high TA of the rivers discharging The relationship TA vs. SiO 2 is also different for Adriatic waters (Fig. 12b), being lower than the general trend in the eastern MedSea. Additionally DIC and pH in the Adriatic have a distinct signal, high values for DIC and low values for pH in waters with salinity lower than 38.7 ( Fig. 12d and g). This basin acts as a connector between the eastern and the western MedSea, here intermediate waters entering from the east mix with deep waters entering from the west (e.g., Hopkins, 1988;Astraldi et al., 1996). This fact is noticed in OSD 10,2013 The CO Fig. 13 were the points corresponding to the Tyrrhenian Sea clearly connect the western and eastern MedSea waters.
In the surface layer, AW with salinity lower than 38 is detected in all the Tyrrhenian stations except 319 in the centre of the basin (Fig. 13a). The freshest AW is seen in station 316, the nearest to Sicily, here AW comes directly from Gibraltar through 5 the Algerian sub-basin. AW is also distinguished by low AOU and SiO 2 values close to zero (Fig. 13b and c). The minimum TA and DIC associated with AW are 2466 and 2178 µmol kg −1 , respectively ( Fig. 14a and d). In the centre of the Tyrrhenian Sea the upper 52 dbars are occupied by probably modified AW with 38.5 > salinity > 38 affected by evaporation or mixed with the salty intermediate 10 water entering through the Strait of Sicily. This water also presents a different relationship between salinity and pH25T compared to unmixed AW (Fig. 14g).
Maxima salinity values in the intermediate layer between 250 and 500 dbars point to LIW entering from the eastern basin (Fig. 13a). As LIW flows into the Tyrrhenian Sea remineralization processes increase its AOU, DIC and lower pH25T, with Deep layers in the Tyrrhenian are occupied by TDW with a typical linear relationship in the temperature-salinity plot (Fig. 13a) the Tyrrhenian is older, lower in CFC12 (Stöven and Tanhua, 2013), than WMDW found in the bottom of the western basin. In the Tyrrhenian Sea bottom WMDW stands out from TDW by the sharp decrease in AOU and pH25T (Fig. 7) and increase in DIC (Fig. 6) for example with salinity (red dots in Fig. S4).

5
The upper layer in this basin is occupied by AW very fresh (<< 38) and practically 100 % saturated in oxygen (AOU ≈ 0 µmol kg −1 ) and with no SiO 2 (Fig. 13b and   c When in the western basin, LIW is associated with a maximum in AOU (Fig. 13b), 15 DIC ( Fig. 14d) and minimum in pH25T (Fig. 14g), so it is the layer of maximum remineralization of organic matter, contrary to the eastern basin where this layer is right below LIW.
The near bottom layer in the western basin is occupied by a recently formed WMDW with higher salinity, so the temperature-salinity plots present a sort of

Conclusions
Two coordinated cruises sampled the Mediterranean Sea for CO 2 variables (TA, DIC and pH) in 2011: the RV Urania EF11 (Sardinia Sicily passage) and the RV Meteor

M84/3 (East-West full-length transect).
The over-determined and high-quality CO 2 data collected in the M84/3 cruise allowed performing the first internal consistency analysis for CO 2 data in the peculiar warm, salty and low DIC/TA ratio MedSea waters. Using the MATLAB version for the consensus CO2SYS program the preferred set of constants is option 4 (CO 2 con-15 stants from Mehrbach et al., 1973 refitted by Dickson and Millero, 1987) along with the sulphate constant from Dickson (1990) and the parameterization of borate from Uppström (1979). Using this combination our pH, TA and DIC measurements from the M84/3 cruise are internally consistent to −0.0003 ± 0.005 for pH, 0.1 ± 3.3 µmol kg −1 for TA and −0.1 ± 3 µmol kg −1 for DIC. Despite this encouraging result further studies are 20 needed including a fourth variable: pCO 2 in order to verify the internal consistency of K 1 (only a reliable K 1 is needed to calculate pCO 2 from the input of pH-TA or pH-DIC) and [CO 2− 3 ] in order to revise the K 2 parameterization in low DIC/TA ratio waters where the CO 2 system is more sensitive to K 2 . Using the combination of pH-TA-DIC only a reliable K 2 is needed in order to calculate bicarbonate, carbonate and aqueous CO 2 25 (Park, 1969 The Meteor M84/3 data allowed for the first time a full length and depth description of CO 2 variables in the MedSea. This study aims to be a benchmark for future data and model studies about the likely effects of natural (climate) and human (warming, acidification, etc. . . ) driven changes in the CO 2 species and ancillary variables associated. Our data provides a picture of the post-EMT physical and biogeochemical distributions 5 in the different MedSea sub-basins.
As a general view, the eastern MedSea presents higher TA (2560-2644 µmol kg −1 ) and pH25T (7.935-8.032) and less variable DIC (2247-2331 µmol kg −1 ) than the western basin with TA (2388-2608 µmol kg −1 ), pH25T (7.861-7.988) and DIC (2110-2336 µmol kg −1 ). Consequently, the saturation for calcite and aragonite in both basins 10 is well above 1, for the eastern basin Ω-Ca vary 2.5-5.78 and Ω-Ar 1.73-3.80 and for the western basin Ω-Ca 2.73-5. 25 and Ω-Ar 1.86-3.44. The CO 2 chemistry in the warm, salty and high TA waters in the MedSea compared to the Atlantic is in general less sensitive to temperature changes because of the low DIC/TA ratio. Therefore, MedSea waters are more resistant to direct changes in DIC 15 and/or TA due to natural and/or anthropogenic processes, since the calculated buffer factors are higher in MedSea waters, showing an increasing westwards gradient. The CO 2 chemistry in the western basin would be more sensitive than in eastern basin.
However, from the point of view of an increasing pCO 2 in the atmosphere, that increases indirectly DIC and keeps TA constant in the ocean, waters in the MedSea have 20 a lower Revelle factor than in the Atlantic, and therefore are able to store more DIC for a given pCO 2 increase than in the Atlantic and consequently the changes in the CO 2 chemistry (carbonate saturation, species concentration, pH) would be higher than in the Atlantic.
As a summary the water masses encountered according to depth are:  In the Levantine and Ionian basins LIW is found shallower (≈ 250 dbars) than in the western basin (≈ 450 dbars). In the eastern MedSea LIW is situated above the layer of maximum organic matter mineralization while in the western MedSea they coincide and therefore is associated with a maximum in DIC and AOU and minimum in pH.
(iv) Cretan Intermediate Water (CIW): formed in the Cretan Sea, this intermediate 25 (250 dbars) water is characterized by even higher TA (2630-2645 µmol kg −1 ) than LIW and detected in the Aegean Sea and the Ionian Sea west of the Antikithera strait.
(v) Cretan Deep Water (CDW): produced by convection in the Cretan Sea, is clearly identified apart of by the high density, by a distinct relation between SiO 2 -salinity, high OSD 10,2013 The CO  (vi) Pre-EMT Eastern Mediterranean Deep Water (EMDW): this water mass is still detected in the Levantine basin by a salinity minimum around 1000 dbars, it presents a minimum in tracer values as CFC-12 or SF 6 , but also a minimum in pH because is 5 the layer of organic matter remineralization as also indicated by the maximum in AOU and inorganic nutrients in the Levantine basin. Corresponding to pre-EMT conditions is has an Adriatic Sea origin.
(vii) EMT-EMDW: below the salinity minimum in the Levantine basin temperature and salinity increase producing a hook in the temperature-salinity diagram which is 10 associated with the input of EMDW produced in the Aegean Sea during the EMT. Due to a more recent origin it presents lower AOU, slightly higher DIC and pH values than Pre-EMT EMDW.
(viii) Post EMT-EMDW: bottom layers in the Hellenic Trench (Levantine basin) and deep layers (> 1000 dbars) in the Ionian Sea are occupied by EMDW with an Adriatic 15 origin produced after the EMT. The hook in the temperature-salinity diagram is no longer detected and a less salty Post EMT-EMDW occupies the deep layer in the Ionian Sea. This water is also identified by lower AOU and pH values compared to the saltier EMT-EMDW.
(ix) Waters in the Adriatic Sea: surface waters in this basin present a distinct TA vs. salinity relationship with values around 2610 µmol kg −1 for salinity < 38.7, higher than the general TA-salinity relation mainly due to the high TA input through Adriatic rivers. 25 The overflow from the Adriatic Sea is mainly detected in the pH vs. AOU relationship, for AOU values between 40-45 µmol kg −1 pH deviates to 7.94, those are waters from around 500 dbars in the Adriatic Pit spilling into the Ionian Sea. 10,2013 The CO  (x) Waters in the Tyrrhenian Sea: the property-property plots shows that this basin is a connector between the eastern and western basins, where LIW mixes deep waters from the west producing Tyrrhenian Deep Water (TDW). During 2011 no recently formed Western Mediterranean Deep Water (WMDW) was detected in the bottom of the Tyrrhenian Sea.

5
The water mass exchange between the eastern and western MedSea is highlighted using data from the Urania EF11 cruise in the Sardinia Sicily passage: AW mainly enters the Tyrrhenian on the eastern part of the passage, close to Sicily, and exits this basin on the Sardinian side, becoming saltier and also higher in TA; the same circulation scheme is done by LIW, which becomes colder and saltier after circulating The CO 2 and ancillary data collected in 2011 highlight the need for a sustained program monitoring the temporal evolution of water masses in the MedSea. This small but relevant marginal sea is as a perfect laboratory basin to detect the effect of natural 25 and anthropogenic driven changes in the physics and CO 2 chemistry. Special emphasis should be given to CO 2 variables as MedSea waters are especially sensitive to pCO 2 increases in the atmosphere: they are prone to absorb more DIC for given pCO 2 increase, this DIC increase in the surface is rapidly transported to depth with the over-

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version