This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy ARTICLE IN PRESS Deep-Sea Research I 56 (2009) 1708–1716

Contents lists available at ScienceDirect

Deep-Sea Research I journal homepage: www.elsevier.com/locate/dsri

Decadal evolution of anthropogenic CO2 in the northwestern Mediterranean Sea from the mid-1990s to the mid-2000s Franck Touratier , Catherine Goyet Laboratoire IMAGES (Institut de Mode´lisation et d’Analyse en GeoEnvironnement et Sante´), Universite´ de Perpignan Via Domitia, 52 avenue Paul Alduy, 66860 Perpignan, France

a r t i c l e i n f o

abstract

Article history: Received 18 September 2008 Received in revised form 18 May 2009 Accepted 25 May 2009 Available online 2 June 2009

Monthly observations accumulated over more than a decade at the DYFAMED timeseries station allow us to estimate the temporal evolution of anthropogenic CO2 in the western Mediterranean Sea. This objective is reached by using recognized interpolation procedures to reconstruct the incomplete distributions of measured total dissolved inorganic carbon and total alkalinity. These reconstructed fields, associated with those available for dissolved oxygen and temperature, are used to estimate the distribution of anthropogenic CO2. This is done with the recently developed Tracer combining Oxygen, inorganic Carbon, and total alkalinity (TrOCA) approach. The main results indicate that (1) the concentrations of anthropogenic CO2 are much higher than those found in the Atlantic Ocean (the minimum concentration at the DYFAMED site is 50 mmol kg1), and (2) the temporal trend for anthropogenic CO2 is decreasing, especially in the intermediate and the deep layers of the water column at the DYFAMED site. This decrease in anthropogenic CO2 is significantly correlated with a decrease in the dissolved oxygen and with an increase in both salinity and temperature. These trends are discussed in the light of recent published works that propose explanations for the observed increases in salinity and temperature that occurred in the western basin since the 1950s. We conclude that the decrease in anthropogenic CO2 probably resulted from an invasion of old water masses. Different hypotheses on the origin of these water masses are considered and several arguments indicate that the eastern Mediterranean transient (EMT) could have played an important role in the observed decrease in anthropogenic CO2 concentrations at the DYFAMED site. & 2009 Elsevier Ltd. All rights reserved.

Keywords: Anthropogenic tracers Carbon chemistry Mediterranean Sea

1. Introduction During the second half of the 20th century, the Mediterranean coastal regions experienced a strong growth of its urban population, from 94 million in 1950 to 274 million in 2000. The anthropogenic pressure on the Mediterranean marine ecosystems thus increased considerably during this period. At the same time several studies revealed important changes in the circulation or physical properties of the Mediterranean Sea (decrease in

 Corresponding author.

E-mail address: [email protected] (F. Touratier). 0967-0637/$ - see front matter & 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.dsr.2009.05.015

the sea level, Tsimplis and Baker, 2000; appearance of a new deep water formation site in the southern Aegean Sea, Roether et al., 1996; increases in salinity and temperature in intermediate and deep waters of the western Mediterranean, Be´thoux et al., 1990) and also in its chemical composition (increase in phosphate and nitrate in the deep layers of the western basin; Be´thoux et al., 2002). The causes of these major changes and the consequences of the functioning of the Mediterranean Sea ecosystems are continuously debated in the literature (Be´thoux et al., 2002; Painter and Tsimplis, 2003; Skliris et al., 2007). It is difficult to determine if these changes are directly or indirectly linked to a global or a regional increase in the anthropogenic pressure. As suggested by

Author's personal copy ARTICLE IN PRESS F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

Reid (1979), some of these Mediterranean perturbations could affect significantly the general circulation and the composition of seawaters in the North Atlantic through exchange of waters between the Atlantic Ocean and the Mediterranean Sea at the Gibraltar Strait. Among the biogeochemical cycles affected by these perturbations, the carbon cycle is of major concern since it is directly affected by the concentration of anthropogenic CO2 (Cant) that accumulates in the atmosphere since the beginning of the industrial era. Since the 1990s an intense effort has been made to quantify and understand the role of the world ocean in the sequestration of anthropogenic carbon dioxide (WOCE program). Numerous hydrographic cruises were organized in the five oceans to measure the distributions of the carbonate system properties (total dissolved inorganic carbon, CT; total alkalinity, AT; pH; and CO2 partial pressure, pCO2). Most sections were also regularly revisited in order to capture the temporal variability in the sequestration process. Despite this large research effort, little is known about the cycle of inorganic carbon in coastal areas or in regional seas like the Mediterranean Sea. It is important to understand how the Mediterranean ecosystems will react to the increase in atmospheric CO2 because (1) the anthropogenic pressure exerted by the countries of the Mediterranean Sea rim on marine ecosystems is particularly high; and (2) the residence time for the Mediterranean waters is short (16 and 50 years for the Algero-Provencal and the eastern basins, respectively) compared to 250 years for deep waters of the North Atlantic (Be´thoux et al., 2002). Consequently the ventilation of the deep Mediterranean waters is fast as is clearly indicated by the levels of chlorofluorocarbon (CFC12) measured in the eastern basin (40.4 pmol kg1; Schlitzer et al., 1991). The distribution and the concentration levels of anthropogenic CO2 in the Mediterranean Sea are unknown. Are the deep layers already contaminated by Cant? Does the seawater Cant level increases with time in parallel to that of the atmosphere? Till now the scientific community was unable to answer these questions for two main reasons: (1) very few data of good quality were available for the Mediterranean Sea to describe the carbonate system properties (CT, AT, pH, and pCO2); and (2) since Cant cannot be measured, it was

1709

difficult to apply a methodology that uses available and measurable properties. Using data of the time-series station DYnamique des Flux Atmosphe´riques en MEDiterrane´e (DYFAMED; http:// www.obs-vlfr.fr/sodyf/home.htm) located in the central part of the Ligurian Sea (43125 N, 7152 E), we estimate the distribution and the temporal evolution of Cant in order to answer the above questions. First, we will reconstruct the temporal evolution of CT and AT profiles using interpolation procedures like the one developed by Goyet and Davis (1997). Second, from the two previous properties, we estimate the distribution of Cant over more than a decade (1993–2005) at the DYFAMED site using the simple TrOCA approach (Touratier et al., 2007). Most other approaches currently used to estimate Cant in the ocean, like the DC* method of Gruber et al. (1996) or the jCt approach of Va´squez-Rodrı´guez et al. (2009), cannot be applied at the DYFAMED site since they all require the knowledge of additional properties like nutrients and/or chlorofluorocarbons.

2. Data sets and methods As part of the French JGOFS program, the DYFAMED site has been chosen to study the vertical fluxes of materials in an area where the advection is relatively small (south of the Ligurian Current; Fig. 1). Since the beginning of the 1990s, this site has been visited monthly in order to perform measurements of numerous physical, chemical, and biological properties throughout the water column from the surface to the seafloor (2350 m). In the present paper, we use the following five properties from the DYFAMED database: temperature (T; 1C), salinity (S), concentration of dissolved oxygen (O2; ml l1), total alkalinity (AT; mmol kg1), and the total dissolved inorganic carbon (CT; mmol kg1). Using the simple TrOCA approach proposed by Touratier et al. (2007), the anthropogenic CO2 concentration (Cant; mmol kg1) is estimated with the following equation: C ant

  2 2 5 O2 þ 1:279 C T  12AT  eð7:511ð1:08710 Þyð7:8110 =AT ÞÞ ¼ 1:279 (1)

Fig. 1. Map of the Mediterranean Sea. (+) DYFAMED site.

Author's personal copy ARTICLE IN PRESS 1710

F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

where y is the potential temperature (1C). According to Eq. (1), the four properties O2, CT, AT, and y are required to directly estimate Cant. However, while measurements for O2 and temperature are available from the year 1993, those for both CT and AT are available during three periods only: (1) period P1: from February 1998 to February 2000 (Be´govic and Copin-Monte´gut, 2002); (2) period P2: from July 2003 to December 2004 (data originator: N. Aı¨t-Ameur); and (3) period P3: from June to December 2005 (data originator: C. Goyet). The Ph.D. thesis of Aı¨t-Ameur (2007) gives more details concerning the data for periods P2 and P3. The first step consists in reconstructing the CT and AT profiles for the whole period from 1993 to 2005, from which we will estimate the temporal evolution of Cant in the water column using Eq. (1). 2.1. Reconstruction of the CT profiles Goyet and Davis (1997) have demonstrated that, below the wintertime mixed layer depth (MLD), CT can be parameterized using the following relationship: C T ¼ a þ by þ cAOU þ dS

pooled together and used to estimate the coefficients a–d, x¯ R and SDR are equal to 0.0, and 6.6 mmol kg1, respectively. However, the SDR value corresponding to data from period P1 or P3 is much lower (o2.5 mmol kg1) than that for the period P2 (5.0 mmol kg1). This probably indicates some inaccuracies in the measurements of CT during period P2. When data corresponding to periods P1 and P3 are pooled together, x¯ R and SDR are equal to 0.0 and 4.7 mmol kg1, respectively. Based on these calculations, we decided to use data from periods P1 and P3 only to estimate the four coefficients a–d. After selecting data from below the MLD (300 m), we estimate the following values for the coefficients: 8 a ¼ 810:26 > > > < b ¼ 4:72 (3) > c ¼ 0:12 > > : d ¼ 37:29 Using this set of coefficients, the vertical variability in the residuals R is shown in Fig. 3. 2.2. Reconstruction of the AT profiles

(2)

where AOU (mmol kg1) is the apparent oxygen utilization. The coefficients a–d are determined by multiple linear regression using discrete bottle data. The bottle data are selected to estimate the four coefficients corresponding to the periods P1, P2, and P3 where measurements of CT are available. As indicated in Fig. 2, the potential temperature y is calculated from the three measured properties S, temperature (1C), and pressure (dbar). The AOU is also a derived property since it is deduced from one measured property (S) and two calculated properties (y, and O2 in mmol kg1). The goodness of fit of the multiple linear regression is then assessed by examining the mean (¯xR ) and the standard deviation (SDR) of the residuals R (i.e. the difference between the measured CT and CT estimated from Eq. (2)). When all data from the three periods are

Previous studies in the western Mediterranean Sea have shown that AT is linearly correlated with salinity (Copin-Monte´gut, 1993; Copin-Monte´gut and Be´govic, 2002). Using all DYFAMED data available throughout the water column from periods P1 to P3, the linear relationship between S and AT is given by 1

AT ¼ 99:26 S  1238:4  4:5 mmol kg

(4)

The correlation coefficient is 0.94, and the mean (x¯ R ) and the standard deviation (SDR) of the residuals R (i.e. the difference between the measured AT and AT estimated from Eq. (4)) are 0.0 and 4.5 mmol kg1, respectively. As shown in Fig. 4, Eq. (4) is very similar to the relationships published by Copin-Monte´gut (1993) for the Alboran Sea and by Copin-Monte´gut and Be´govic (2002) for period P1 at the DYFAMED site.

Fig. 2. Computational scheme for the calculated properties, the carbonates system properties, and the concentration of anthropogenic CO2 derived from the properties measured at the DYFAMED site.

Author's personal copy ARTICLE IN PRESS F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

Fig. 3. Vertical variability of the residuals (R) for CT.

Fig. 4. Linear relationships between AT and salinity for the western Mediterranean Sea.

3. Results and discussion Very clear increases in S and y with time exist in the time-series at the DYFAMED site (Figs. 5a,b). At depths 1000, 1500, and 2000 m, we could identify significant increase in trends of S or y with time (r is 0.84, 0.88, and 0.81 for S; and r is 0.79, 0.87, and 0.95 for y, respectively). Because of a large influence of the seasonal variability in the surface layer, there are no clear trends in the mixed layer (0–200 m). As indicated by the y/S diagram in Fig. 6, the water column at the DYFAMED site is occupied by four water masses: (1) the Atlantic Water (AW), located in the surface layer and characterized by a large variability in y and S (the potential density sy for this water mass is o28.8); (2) the Winter Intermediate Water (WIW), characterized by minima in both y and S (syE28.9); (3) the Levantine Intermediate Water (LIW), centred at a sy value of 29.05 and characterized by maxima in both y and S; and (4) the Western Mediterranean Deep Water (WMDW), with a typical sy value close to 29.1. At depths Z2000 m (WMDW), Be´thoux et al. (1990) did not find any significant increase in density for the period 1959–1989. The time-series at the DYFAMED site shows, however, an

1711

uplifting of the WMDW isopycnals by several hundred metres (Fig. 5a). The DYFAMED mean annual rates of increase in y and S have been estimated for the three depths 1000, 1500, and 2000 m (Table 1). At a depth of 2000 m, the rates for y and S reach values of 5  103 1C year1 and 2  103 year1, respectively. At 1500 and 1000 m, these rates increase by a factor two and three, respectively. For comparison, other estimates available from the literature for the WMDW (Be´thoux et al., 1990; Krahmann and Schott, 1998; Rixen et al., 2005; Schroeder et al., 2008) are also shown in Table 1. The latter studies and the present one all suggest a continuous increase in both y and S since the 1950s in the deep layers of the western Mediterranean. As indicated by the recent studies of Rixen et al. (2005) and Schroeder et al. (2008), the temporal evolution of these rates experienced two periods of acceleration: the first occurred after the second half of the 1980s, and the second appeared during the years 2005 and 2006. However, these accelerations cannot be seen in the present paper since the period of analysis for the DYFAMED time-series is 1993–2005. The distributions of the reconstructed CT and AT are shown in Fig. 5c and d, respectively. The concentrations of O2 and Cant in the WMDW decrease continuously with minima appearing during the years 2004 and 2005 (Fig. 5e and f). Over a 12-year-period (from 1993 to 2005), the concentrations of both CT and AT increased by approximately 5 mmol kg1 at a depth of 500 m, while those for O2 and Cant decreased by 20 and 15 mmol kg1, respectively. The uncertainties associated with the estimations of CT, AT, and Cant and the measurement of O2 are 4.7 mmol kg1 (SDR for CT, see above), 4.5 mmol kg1 (SDR for AT, see above), 6.25 mmol kg1 (Touratier et al., 2007), and 1 mmol kg1, respectively. Comparing the increase or the decrease in the above properties to their uncertainty, we conclude that only O2 and Cant decrease significantly. In the context of global warming, which is mainly due to the increase in amounts of greenhouse gases (of which CO2 is one of the most abundant after H2O) that accumulate both in the atmosphere and in the ocean, we did expect an increase in Cant content in the water column as is the case in many other oceanic regions (Gruber et al., 1996; Goyet et al., 1999; Touratier and Goyet, 2004b). The decrease in trend of Cant at the DYFAMED site is thus very surprising; especially because during the same period of time the concentration of CT increases (Cant represents a small fraction of CT). In the following discussion, we examine the best hypotheses that could explain the distribution of the estimated Cant. The decrease in concentration of Cant below 500 m is strongly and positively correlated with the O2 content (rZ0.98). It is thus interesting to understand why the concentration of O2 is decreasing in the WMDW, with the hope that it could help us to understand the evolution of Cant at the DYFAMED site. Recently, the paper of Joos et al. (2003) pointed out that significant reductions in O2 (from a few mmol kg1 up to 30 mmol kg1) were detected in all major oceans. According to them, both the observationbased analyses and the global ocean models identify the ocean circulation changes, induced by anthropogenic

Author's personal copy ARTICLE IN PRESS 1712

F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

Fig. 5. Temporal evolution of the measured or the estimated properties at the DYFAMED site. (a) Salinity and sy (isolines); (b) potential temperature (y); (c) total dissolved inorganic carbon (CT) derived from Eq. (2); (d) total alkalinity (AT) derived from Eq. (4); (e) dissolved oxygen (O2); and (f) anthropogenic CO2 (Cant).

radiative forcings, as the main cause of the observed decreases in dissolved O2. At the DYFAMED site and for the period 1993–2005, the decrease in O2 saturation corresponding to an increase in S and y represents only 5.5% of the total O2 decrease (20 mmol/kg). Another hypothesis is the one proposed by Klein et al. (2003) to explain a decrease in O2 of 5 mmol kg1 in the

eastern Mediterranean Deep Water (EMDW) during a four-year-period (1995–1999). As a consequence of the Eastern Mediterranean Transient event (EMT; see Roether et al., 1996), which happened in the late 1980s, approximately 20% of the EMDW was replaced by dense waters originating from the Aegean Sea. Before 1995, the newly formed EMDW was not only characterized by higher

Author's personal copy ARTICLE IN PRESS F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

Fig. 6. y/S diagram for the DYFAMED site. AW: Atlantic Water; WIW: Winter Intermediate Water; LIW: Levantine Intermediate Water; and WMDW: Western Mediterranean Deep Water. All y and S data available from the DYFAMED database were used to plot the y/S diagram.

Table 1 Increases in y and S in deep waters of the western Mediterranean Sea, according to different studies.

Dy  103

DS  103 (1C year1) (year1)

Source

Period

DYFAMED, this study 1000 m 1500 m 2000 m

1993–2005

Be´thoux et al. (1990) Krahmann and Schott (1998) Rixen et al. (2005) Schroeder et al. (2008)

1959–1989 4 1955–1994 1.6 1985–2000 5.5 2004–2006 19

16 10 5

6 4 2 1 0.8 1.2 8

temperature and salinity but also by higher oxygen content because of its recent contact with the atmosphere. According to Klein et al. (2003), the decrease in dissolved oxygen that occurred in the newly formed EMDW after 1995 resulted from the utilization of large amounts of DOC (characterized by a large fraction of labile material) made available by the EMT which in turn enhanced O2 consumption. However, the two above hypotheses, i.e. a decrease in O2 due to a slowdown of ocean circulation or an increase in the consumption of DOC, are not consistent with the parallel decrease in Cant at the DYFAMED site. A slowdown of the ocean circulation may decrease the rate at which Cant is sequestrated, but it cannot decrease the stock of Cant by itself. On the other hand, the assimilation of labile DOC by bacteria is a process that requires O2 for respiration. A small fraction of the CO2 released by bacteria during respiration is of anthropogenic origin

1713

since DOC was produced initially in the surface layer (mainly by the phytoplankton). So it is expected that the consumption of DOC in the EMDW would have increased the level of Cant. Another possibility is that the lowered concentrations of O2 and Cant at the DYFAMED site could have resulted from an invasion of an older water mass, also characterized by higher S, y, CT, and AT. For instance, Roether et al. (1998) detected such an intrusion in the LIW of the eastern basin of the Mediterranean Sea. Using the anthropogenic tracer CFC-12, they estimated that the age of the LIW increased by 6 years from 1987 to 1995 (the level of CFC-12 decreased by 30%). According to them, this invasion was a direct consequence of the EMT because the EMDW (low CFC-12 concentration) was uplifted by the newly formed EMDW (characterized by a higher concentration of CFC-12 since they are much younger). The old EMDW in contact with the above LIW decreased the level of CFC-12 in the LIW by mixing. Using only the DYFAMED database, one cannot explain the decrease in Cant since it is probably linked to the important changes that occurred in the Mediterranean Sea during the last decades. A review of these changes and their consequences is thus essential to identify the possible cause of the Cant decrease. A major change that affected the whole Mediterranean Sea during the recent decades is the higher state of the North Atlantic Oscillation (NAO), which tends to reduce precipitation (Tsimplis and Josey, 2001). Some important changes were also noted in the wind stress over the eastern Mediterranean by Samuel et al. (1999). The damming of important rivers like the Nile (with the construction of the Aswan Dam in 1964), the Ebro river in the western Mediterranean (in the early 60’s), and several Russian rivers flowing to the Black Sea all contributed directly or indirectly (via the Bosphorus Strait) to a significant decrease in the freshwater load (reduction by more than 90% and 60% for the Nile and the Ebro rivers). Most of these changes led to an increase in the surface salinity of the Mediterranean Sea, which is thought to have repercussions on the deep water formation (DWF) sites. The recent work of Skliris et al. (2007), based on the results of the Princeton Ocean Model (POM), suggests that 95% of the observed salinity increase in the WMDW over the last 40 years could be due to a combined effect of the damming of rivers (to which the Nile damming represents a major contribution) and the decrease in precipitation. According to Painter and Tsimplis (2003), the changes observed in the wind stress and the modifications in the physical properties of the Black Sea outflow could explain the shift of the main DWF site in the eastern basin of the Mediterranean Sea. The southern part of the Adriatic Sea is known as the usual DWF site for the EMDW. However, it has been observed that the newly formed EMDW could originate from a different DWF site. An important shift, known as the eastern Mediterranean Transient (Roether et al., 1996; Klein et al. 1999), occurred between 1987 and 1992 with a DWF site located in the southern Aegean Sea. When compared with the Adriatic DWF site, the new deep waters were characterized by an increase in both the salinity and the temperature during the EMT (Skliris et al.,

Author's personal copy ARTICLE IN PRESS 1714

F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

2007). Using their model, Skliris et al. (2007) concluded, however, that the increase in salinity alone was probably insufficient to produce the climatic transient. Important surface cooling should also happen during winter in order to allow the propagation of the increased surface salinity signal to the deep layers. The recent papers of Gasparini et al. (2005) and Schroeder et al. (2006) have put forward the possibility that the recent increases in salinity and temperature in the deep and intermediate waters of the western basin could be attributed to the propagation of the EMT signal through the Sicily Strait (called hypothesis H1 hereafter). During the 1990s, however, two other hypotheses were proposed in the literature to explain the increases in S and y since the 1950s. The first (called H2 hereafter) is that of Be´thoux et al. (1990), Krahman and Schott (1998), and Be´thoux and Gentili (1999), who suggest that these trends resulted from air–sea exchanges and dense-water formation, two processes forced by specific atmospheric conditions prevailing on the northwestern basin. The last hypothesis (called H3 hereafter), proposed by Rohling and Bryden (1992) states that the damming of major rivers, which increased significantly the salinity of the LIW, could be responsible for the increases in S and y in the WMDW. According to Painter and Tsimplis (2003), the transfer of properties from east to west in the Mediterranean Sea via the LIW (hypothesis H3) does not represent a major contribution to changes observed in S and y of the WMDW. This point of view is shared by Skliris et al. (2007) since the results of their model suggest that the observed increase in salinity in the WMDW during the last 40 years is explained by both H2 and H3 at a comparable level of 50% each. There is no doubt, however, that hypothesis H1 of Gasparini et al. (2005) and Schroeder et al. (2006) has at least played an important role since the 1990s. Schroeder et al. (2008) describe two different effects generated by the EMT. On one hand, there is a decadal salt and heat accumulation at the intermediate levels as a consequence of the EMT signal, first in the Tyrrhenian Sea and then in the Ligurian Sea (Gasparini et al., 2005; Schroeder et al., 2006). On the other hand, when the S increase observed in the LIW reaches the DWF sites located in the northwestern Mediterranean Sea (Gulf of Lions, Catalan and Ligurian sub-basins; Salat and Font, 1987; Sparnocchia et al., 1995), there is a propagation of the EMT signal (higher S and y) to the WMDW via the formation of new deep waters. Many uncertainties still exist in the exact way and timing by which the intermediate and deep waters have been modified at the DYFAMED site since the 1990s. It is clear, however, that the distributions of presented properties (S, y, CT, AT, O2, and Cant) have been influenced by the processes that occurred as far away as the eastern Mediterranean more than 10 years ago. It is likely that the EMT has played a role, and it seems also evident that the distribution of the properties, as observed since the 1950s, results from local atmospheric forcings. The relative proportion of each process remains to be determined.

What we learn from the present study is that the decrease in O2 and Cant (Figs. 5e and f) is probably due to an invasion of older waters in the intermediate and deep layers of the DYFAMED site. This is incompatible with the idea that these waters originate directly from the DWF sites, since their recent contact with the atmosphere should have increased their levels of O2 and Cant. A more realistic scenario for the western basin is the one observed in the eastern basin during the EMT: an uplifting of the existing and the old Mediterranean Deep Water due to the sinking of denser water masses (the new Mediterranean Deep Water) produced recently by the DWF sites is generally observed. The situation at the DYFAMED site is expected to be far more complex since the effects of local atmospheric forcings and EMT originated from different regions and different timescales. For instance, the propagation of the EMT signal from east to west, up to the Ligurian Sea, is difficult to describe because some pieces of the puzzle are still missing. As an example, the central role played by the Tyrrhenian Sea in transferring the EMT signal between the eastern basin and the Algero-Provencal basin is difficult to study because of the scarcity of data in this area.

4. Conclusions The present paper is the first one to present a realistic estimate of the distribution of anthropogenic CO2 in the western Mediterranean Sea and its evolution over a decadal timescale. This could not be done earlier since (1) good quality data were too scarce, and (2) a robust and simple model to estimate Cant was unavailable. Using the interpolation procedures described above to estimate CT and AT, and the TrOCA approach to compute Cant, the two problems were solved successfully. All recent publications where the TrOCA approach is used to estimate the distribution of Cant (Touratier and Goyet, 2004a, b; Lo Monaco et al., 2005; Touratier et al., 2005; Aı¨t-Ameur and Goyet, 2006; Sandrini et al., 2007; Touratier et al., 2007; Goyet and Touratier, in press; Goyet et al., 2009; Laika et al., in press; Alvarez et al., 2009; Va´zquez-Rodrı´guez et al., 2009) show that this model provides realistic results whatever the oceanic region where the approach is applied (Atlantic Ocean, Indian Ocean, Pacific Ocean, Gulf of Cadiz, and Antarctic Ocean). Compared to other approaches, the TrOCA method is the easiest to use (one line of simple calculation taking into account only the four properties y, O2, CT, AT), and it can be applied to a single station, an ocean section, an ocean basin, or on the global scale. It can also be introduced in any 3D OGCM (Gerber et al., 2009). An expected result was that Cant in the Mediterranean Sea should increase in parallel with the level of atmospheric anthropogenic CO2. Aı¨t-Ameur and Goyet (2006) demonstrated that the Mediterranean outflow water (MOW) at the Gibraltar Strait is a significant source of Cant for the Atlantic Ocean (0.032–0.066 Pg C year1), indicating that Cant is efficiently transferred from the atmosphere to the Mediterranean Sea. At the DYFAMED site, however, the present study shows that the

Author's personal copy ARTICLE IN PRESS F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

concentration of Cant decreases with time in the intermediate and deep waters. This trend has been discussed, and we conclude that the explanation of this phenomenon probably is an overlapping of three hypotheses H1–H3. At this stage, however, we have no argument to estimate the relative impact of each hypothesis. From the Cant and O2 signatures, it is obvious that an invasion of one or several older water masses did occur during this decade. The older water masses may have originated indirectly from the eastern basin as a consequence of the EMT or from the DWF sites located in the northwestern Mediterranean Sea. These events have in common the characteristic that they produce large volumes of deep waters sufficiently dense to uplift the existing deep waters, which results in an apparent increase in the age of the overlying waters (and thus a decrease in the Cant and O2 contents). The utilization of measurable anthropogenic tracers like CFCs, 14C, and 3H are shown to be very useful to validate the estimated anthropogenic CO2 profiles in both the Indian and the Atlantic Ocean (Touratier and Goyet, 2004b; Touratier et al., 2007; Va´zquez-Rodrı´guez et al., 2009). We highly recommend that measurements of such tracers can be made parallel with those of the carbonate system properties, especially in the western Mediterranean Sea, where such data are lacking. The validation of the present results by this means is an essential step, especially because our results indicate that the seawater at DYFAMED is contaminated with Cant at different levels (the minimum concentration is 50 mmol kg1). For comparison, large volumes of Cant-free waters can still be found in the bottom layers of the southern Atlantic Ocean (Touratier and Goyet, 2004b; Va´zquez-Rodrı´guez et al., 2009). Even in the intermediate layers, the level of Cant is well above the 50 mmol kg1 level. Consequently, this study indicates that the Mediterranean Sea probably sequesters very large amounts of anthropogenic CO2, at least in the northwestern Mediterranean basin. Even if the concentration of Cant observed at the DYFAMED site has decreased with time over the last 12 years, the level remains incomparably high, especially with the Atlantic water near Gibraltar. Since acidification in the oceans results mainly from the accumulation of Cant, there has been a significant reduction in pH (0.15) in the Mediterranean Sea since the pre-industrial era (probably much larger than the potential decrease of pH in the Atlantic Ocean). Thus the western Mediterranean Sea, which plays an important role in anthropogenic CO2 sequestration, is a key area for observing the effects of anthropogenic impacts (increase in temperature, CO2, pH, nutrients, etc.) on marine chemistry and ecosystems.

Acknowledgements This work is part of the CARBOOCEAN project (Marine carbon sources and sinks assessment), funded by the European Community. We thank N. Aı¨t-Ameur who measured CT and AT from the DYFAMED samples collected during the years 2003 and 2004. We are very grateful to the scientists and technicians from the Service d’observation DYFAMED (SODYF) and the Service National d’Ana-

1715

lyse des parame`tres oce´aniques du CO2 (SNAPO-CO2) involved in the sampling and analysis of the water samples. References Aı¨t-Ameur, N., 2007. Contribution a` la quantification du CO2 anthropique en mer Me´diterrane´e et en mer de Ross. Ph.D. Thesis, University of Perpignan, France. Aı¨t-Ameur, N., Goyet, C., 2006. Distribution and transport of natural and anthropogenic CO2 in the Gulf of Cadiz. Deep-Sea Res. II 53, 1329–1343. Alvarez, M., Lo Monaco, C., Tanhua, T., Yool, A., Oschlies, A., Bullister, J.L., Goyet, C., Metzl, N., Touratier, F., McDonagh, E., Bryden, H.L., 2009. Estimating the storage of anthropogenic carbon in the subtropical Indian Ocean: a comparison of five different approaches. Biogeosciences 6, 681–703. Be´govic, M., Copin-Monte´gut, C., 2002. Processes controlling annual variations in the partial pressure of CO2 in surface waters of the central northwestern Mediterranean Sea (DYFAMED site). Deep-Sea Res. II 49, 2031–2047. Be´thoux, J.P., Gentili, B., Raunet, J., Tailliez, D., 1990. Warming trend in the western Mediterranean deep water. Nature 347, 660–662. Be´thoux, J.P., Gentili, B., 1999. Functioning of the Mediterranean Sea: past and present changes related to freshwater input and climate changes. J. Mar. Syst. 20, 33–47. Be´thoux, J.P., Morin, P., Ruiz-Pino, D.P., 2002. Temporal trends in nutrient ratios: chemical evidence of change in the Mediterranean ecosystem due to human activity. Deep-Sea Res. II 49, 2007–2016. Copin-Monte´gut, C., 1993. Alkalinity and carbon budgets in the Mediterranean. Global Biogeochem. Cycles 7, 915–925. Copin-Monte´gut, C., Be´govic, M., 2002. Distributions of carbonate properties and oxygen along the water column (0–2000 m) in the central part of the NW Mediterranean Sea (Dyfamed site): influence of winter vertical mixing on air–sea CO2 and O2 exchanges. Deep-Sea Res. II 49, 2049–2066. Gasparini, G.P., Ortona, A., Budillon, G., Astraldi, M., Sansone, E., 2005. The effect of the Eastern Mediterranean Transient on the hydrographic characteristics in the Strait of Sicily and in the Tyrrhenian Sea. Deep-Sea Res. I 52, 915–935. Gerber, M., Joos, F., Va´zquez-Rodrı´guez, M., Touratier, F., Goyet, C., 2009. Regional air-sea fluxes of anthropogenic carbon inferred with an Ensemble Kalman Filter. Global Biogeochem. Cycles 23, GB1013. Goyet, C., Davis, D.L., 1997. Estimation of TCO2 concentration throughout the water column. Deep-Sea Res. I 44, 859–877. Goyet, C., Coatanoan, C., Eischeid, G., Amaoka, T., Okuda, K., Healy, R., Tsunogai, S., 1999. Spatial variation of total CO2 and total alkalinity in the northern Indian Ocean: a novel approach for the quantification of anthropogenic CO2 in the seawater. J. Mar. Res. 57, 135–163. Goyet, C., Ito Gonc- alves, R., Touratier, F., 2009. Anthropogenic carbon distribution in the eastern South Pacific Ocean. Biogeosciences 6, 149–156. Goyet, C., Touratier, F. Challenges for present and future estimates of anthropogenic carbon in the Indian Ocean. Geophysical Monograph Series, in press. Gruber, N., Sarmiento, J.L., Stocker, T.F., 1996. An improved method for detecting anthropogenic CO2 in the oceans. Global Biogeochem. Cycles 10, 809–837. Joos, F., Plattner, G.-K., Stocker, T.F., Ko¨rtzinger, A., Wallace, D.W.R., 2003. Trends in marine dissolved oxygen: implications for ocean circulation changes and carbon budget. Eos 84, 197–204. Klein, B., Roether, W., Manca, B.B., Bregant, D., Beitzel, V., Kovacevic, V., Luchetta, A., 1999. The large deep water transient in the eastern Mediterranean. Deep-Sea Res. I 46, 371–414. Klein, B., Roether, W., Kress, N., Manca, B.B., d’Alcala, M.R., Souvermezoglou, E., Theocharis, A., Civitarese, G., Luchetta, A., 2003. Accelerated oxygen consumption in eastern Mediterranean deep waters following the recent changes in thermohaline circulation. J. Geophys. Res. 108, 8107. Krahmann, G., Schott, A.F., 1998. Long term increases in Western Mediterranean salinities and temperatures: anthropogenic and climatic sources. Geophys. Res. Lett. 25 (22), 4209–4212. Laika, H.E., Goyet, C., Vouve, F., Poisson, A., Touratier, F. Interannual properties of the CO2 system in the southern ocean south of Australia. Antarctic Science, in press. Lo Monaco, C., Goyet, C., Metzl, N., Poisson, A., Touratier, F., 2005. Distribution and inventory of anthropogenic CO2 in the Southern

Author's personal copy ARTICLE IN PRESS 1716

F. Touratier, C. Goyet / Deep-Sea Research I 56 (2009) 1708–1716

Ocean: comparison of three data-based methods. J. Geophys. Res. 110, C09S02. Painter, S.C., Tsimplis, M.N., 2003. Temperature and salinity trends in the upper waters of the Mediterranean Sea as determined from the MEDATLAS dataset. Cont. Shelf Res. 23, 1507–1522. Reid, J., 1979. On the contribution of the Mediterranean Sea outflow to the Norwegian–Greenland Sea. Deep-Sea Res. A 26, 1199–1223. Rixen, M., Beckers, J.-M., Levitus, S., Antonov, J., Boyer, T., Maillard, C., Fichaut, M., Balopoulos, E., Iona, S., Dooley, H., Garcia, M.-J., Manca, B., Giogetti, A., Manzella, G., Mikhailov, N., Pinardi, N., Zavatarelli, M., 2005. The Western Mediterranean Deep Water: a proxy for climate change. Geophys. Res. Lett. 32, L12608. Roether, W., Manca, B.B., Klein, B., Bregant, D., Georgopoulos, D., Beitzel, V., Kovacevic, V., Luchetta, A., 1996. Recent changes in the Eastern Mediterranean deep waters. Science 271, 333–335. Roether, W., Klein, B., Beitzel, V., Manca, B.B., 1998. Property distributions and transient-tracer ages in Levantine Intermediate Water in the Eastern Mediterranean. J. Mar. Syst. 18, 71–87. Rohling, E.J., Bryden, H.L., 1992. Man-induced salinity and temperature increases in the western Mediterranean deep water. J. Geophys. Res. 97, 11191–11198. Salat, J., Font, J., 1987. Water mass structure near and offshore the Catalan coast during the winters of 1982 and 1983. Ann. Geophys. 1B, 49–54. Samuel, S., Haines, K., Josey, S., Myers, P.G., 1999. Response of the Mediterranean Sea thermohaline circulation to observed changes in the winter wind stress field in the period 1980–1993. J. Geophys. Res. 104, 7771–7784. Sandrini, S., Aı¨t-Ameur, N., Rivaro, P., Massolo, S., Touratier, F., Tositti, L., Goyet, C., 2007. Anthropogenic carbon distribution in the Ross Sea, Antarctica. Antarctic Science, doi:10.1017/S0954102007000405. Schlitzer, R., Roether, W., Oster, H., Junghans, H.-G., Hausmann, M., Johannsen, H., Michelato, A., 1991. Chlorofluoromethane and oxygen in the eastern Mediterranean. Deep-Sea Res 38, 1531–1551.

Schroeder, K., Gasparini, G.P., Tangherlini, M., Astraldi, M., 2006. Deep and intermediate water in the western Mediterranean under the influence of the Eastern Mediterranean Transient. Geophys. Res. Lett. 33, L21607. Schroeder, K., Ribotti, A., Borghini, M., Sorgente, R., Perilli, A., Gasparini, G.P., 2008. An extensive Western Mediterranean Deep Water renewal between 2004 and 2006. Geophys. Res. Lett. 35, L18605. Skliris, N., Sofianos, S., Lascaratos, A., 2007. Hydrological changes in the Mediterranean Sea in relation to changes in the freshwater budget: a numerical modeling study. J. Mar. Syst. 65, 400–416. Sparnocchia, S., Picco, P., Manzella, G., Ribotti, A., Copello, S., Brasey, P., 1995. Intermediate water formation in the Ligurian Sea. Oceanol. Acta 12, 151–162. Touratier, F., Goyet, C., 2004a. Definition, properties, and Atlantic Ocean distribution of the new tracer TrOCA. J. Mar. Syst. 46, 169–179. Touratier, F., Goyet, C., 2004b. Applying the new TrOCA approach to assess the distribution of anthropogenic CO2 in the Atlantic Ocean. J. Mar. Syst. 46, 181–197. Touratier, F., Goyet, C., Coatanoan, C., Andrie´, C., 2005. Assessment of anthropogenic CO2 distribution in the tropical Atlantic Ocean. DeepSea Res. I 52, 2275–2284. Touratier, F., Azouzi, L., Goyet, C., 2007. CFC-11, D14C and 3H tracers as a means to assess anthropogenic CO2 concentrations in the ocean. Tellus 59B, 318–325. Tsimplis, M.N., Baker, T.F., 2000. Sea level drop in the Mediterranean Sea: an indicator of deep water salinity and temperature changes? Geophys. Res. Lett. 27, 1731–1734. Tsimplis, M.N., Josey, S.A., 2001. Forcing of the Mediterranean Sea by atmospheric oscillations over the North Atlantic. Geophys. Res. Lett. 28. Va´squez-Rodrı´guez, M., Touratier, F., Lo Monaco, C., Waugh, D., Padin, X.A., Bellerby, R.G.J., Goyet, C., Metzl, N., Rios, A.F., Perez, F.F., 2009. Anthropogenic carbon distributions in the Atlantic Ocean: databased estimates from the Arctic to the Antarctic. Biogeosciences 6, 439–451.