1538
Decadal changes in surface carbon dioxide and related variables
in the Mediterranean Sea as inferred from a coupled datadiagnostic model approach
Ferial Louanchi, Meriem Boudjakdji, and Lamri Nacef
Louanchi, F., Boudjakdji, M, and Nacef, L. 2009. Decadal changes in surface carbon dioxide and related variables in the Mediterranean Sea as
inferred from a coupled data-diagnostic model approach. – ICES Journal of Marine Science, 66: 1538 – 1546.
A coupled approach based on available datasets of temperature, salinity, oxygen, nutrients, and chlorophyll, and a surface layer box
model previously developed and modified for the present study, allowed us to reconstruct dissolved inorganic carbon (DIC), total
alkalinity, and carbon dioxide fugacity ( fCO2) mixed-layer fields for the Mediterranean Sea, from the 1960s to the 1990s. The approach
used in this study resulted in a 7% relative error on reconstructed surface fCO2 fields. The Mediterranean Sea transformed from a source
of 0.62 Tg C year21 for atmospheric CO2 in the 1960s, to a net sink of 21.98 Tg C year21 in the 1990s. The annual cycle in surface fCO2
was driven mainly by temperature variations in the Mediterranean Sea, whereas its decadal variations resulted from a balance between
primary production and the thermal effect. According to our model results, the atmospheric CO2 increase of 40 matm over the
period of our investigation induced an increase in DIC of 30 mmol l21 in surface waters. A 50% reduction in the magnitude of seasonal variations in surface temperature occurred during the 1990s relative to the earlier decades. Therefore, surface fCO2 only increased
by 24 matm from the 1960s to the 1990s. Changes in pH were not significant over this period.
Keywords: air– sea CO2 fluxes, carbon dioxide, decadal variability, exported production, Mediterranean Sea, primary production.
Received 15 August 2008; accepted 8 February 2009; advance access publication 17 April 2009.
F. Louanchi and M. Boudjakdji: Ecole Nationale Supérieure des Sciences de la Mer et de l’Aménagement du Littoral (ex-ISMAL), Bois des cars,
Dely-Brahim, 16320 Algiers, Algeria. L. Nacef: Office National de la Météorologie (ONM), Dar El Beida, Algiers, Algeria. Correspondence:
F. Louanchi: tel/fax: þ213 21 91 77 90; e-mail: [email protected].
Introduction
The global ocean is well known as a net sink for carbon dioxide,
absorbing 30 –40% of the anthropogenic atmospheric CO2
released every year (Mikaloff Fletcher et al., 2006). It has also
been demonstrated that the penetration of anthropogenic
carbon into the sea results in surface ocean acidification (Orr
et al., 2005), which can have drastic consequences for marine
ecosystems. In this general context, the Mediterranean Sea is
an interesting region for studying climate change effects on the
ocean. The general circulation in the Mediterranean Sea is one
of three zones of deep and intermediate water formation
(CIESM, 2002), which contribute to the penetration of anthropogenic carbon. The Mediterranean Sea’s 100-year water mass
turnover is faster than that of the global ocean, which implies
a rapid response to climate variability and change (Béthoux
et al., 1999). As a semi-enclosed area, the Mediterranean Sea is
also highly influenced by other anthropogenic activities, such
as eutrophication, which affects the biogeochemical cycles and
the foodweb. The in situ observations conducted in the
Mediterranean Sea over past decades have already demonstrated
a change in the marine biology and chemistry, which seem to
be linked to climate and environmental change (Béthoux,
1989; Caddy et al., 1995; Béthoux et al., 1998, 2002;
Malanotte-Rizzoli et al., 1999; CIESM, 2002; Marty and
Chiavérini, 2002).
However, little is known about the role of the Mediterranean
Sea as a source or a sink of CO2 over the past few decades, nor
whether its carbon cycle has changed over time. In fact, only a
few measurements of carbon exchange have been conducted
from the 1960s to the 1990s. Some measurements of total alkalinity (TA) and pH collected during the 1980s are available.
These data were acquired before the release of Dickson CRM (certified reference material; DOE, 1994). Several intercomparison
exercises performed during the 1980s revealed large discrepancies
in measured alkalinity among laboratories, sometimes up to
30 meq l21. Moreover, using these two variables to calculate
carbon dioxide fugacity, fCO2, result in errors of up to 15 matm
(Wanninkhof et al., 1999; Lee et al., 2000; Millero et al., 2006).
Dissolved inorganic carbon (DIC) was measured in the
Mediterranean Sea during the 1990s, but these data have not yet
been made available.
In this study, we took advantage of the MEDAR/MEDATLAS
(2002) database release to reconstruct decadal average annual
cycles of temperature, salinity, oxygen, chlorophyll, and nutrients
for the 1960–1990s. These climatologies were used to constrain
the simple box model developed by Louanchi et al. (1996) and
modified for the needs of the present study. The aim of the
study was to infer the decadal change in surface fCO2 fields, and
related parameters, to assess the role of the Mediterranean Sea as
an atmospheric source or sink of CO2.
# 2009 International Council for the Exploration of the Sea. Published by Oxford Journals. All rights reserved.
For Permissions, please email: [email protected]
1539
Decadal changes in surface carbon dioxide in the Mediterranean
Methods
Diagnostic model description
The surface box model used to extrapolate the average decadal
annual CO2 cycles is the one developed by Louanchi et al.
(1996), and described in subsequent studies (Metzl et al., 1998;
Louanchi et al., 1999a, b). The surface box model considers the
main processes controlling CO2 in the oceanic mixed layer,
namely the air–sea exchange, carbonate chemistry and thermodynamics, phytoplankton uptake and release, and the exchange
between surface and subsurface layers. The contribution of each
process is calculated using the monthly variations in the forcing
factors, which are windspeed (W), mixed-layer depth (MLD),
temperature (T ), salinity (S), and chlorophyll (Chl) concentration. In the version used here, we did not simulate the nutrients
(phosphates P, nitrates N, and silicates Si), but used them as
forcing factors as well. An expression of how fCO2 varies with
time is
ðDf =DtÞ ¼ ðDf =DtÞF þ ðDf =DtÞM þ ðDf =DtÞT þ ðDf =DtÞB ; ð1Þ
Figure 1. The eighteen zones of Mediterranean Sea used in this
study.
information about the physical properties of the area. In this
box model, it is not necessary to define initial surface properties,
because the unknown surface concentrations stabilize after
running the model for 3 –4 years. However, it is sensitive to the
subsurface properties, as pointed out by Peng et al. (1987).
Data processing
where subscripts F, M, T, and B represent air–sea exchange,
mixing, temperature, and biological effects, respectively. Each
process is constrained by one factor or a set of forcing factors.
The model’s equations are based on month-to-month budget
calculations. The contributions of the processes are added to
obtain the total CO2 fugacity variation. Each term in Equation
(1), except the thermal component, is derived from DIC and TA
variations. The conversion from DIC and TA variations to fCO2 is
done using Mehrbach et al.’s (1973) dissociation constants.
Air– sea CO2 flux depends on windspeed, water temperature,
and salinity, and it is calculated using Wanninkhof’s (1992) parameterization. It influences DIC variations in the mixed layer. An
annual cycle of atmospheric pCO2 is reconstructed for each
decade based on Global Atmospheric Watch (GAW) network
observations (Global Atmospheric Watch International programme supervised by WMO, http://www.wmo.ch). Vertical
mixing only affects DIC and TA and is a simple entrainment
equation (Peng et al., 1987), where the subsurface parameters
are set as a constant for each decade. The biological effect is
derived from the monthly changes in Chl content. Chl variation
is the result of a budget between photosynthesis (gain) and respiration and mortality (total losses). The gain depends on nutrient
(minimum Michaelis –Menten function of P or N) and light limitation. The total losses are calculated using Chl variations and the
gain previously estimated. In all, 80% of the total losses are allowed
to remineralize within the mixed layer. This number is typical of
oligotrophic conditions. Then, we calculate the effect of photosynthesis –remineralization on P and DIC changes (TA is considered
to be invariant during this process) using constant C:P and C:Chl
ratios of 106 (Anderson, 1995) and 40 (Taylor et al., 1991),
respectively. The exported production (EP) from the surface
layer is composed of 10% of CaCO3, the latter decreasing both
DIC and TA. Primary production (PP) is the total production
(new and regenerated) in the surface layer. Finally, the thermodynamic effect on fCO2 is constrained monthly by the variations of
temperature and salinity. It is calculated directly from Mehrbach
et al.’s (1973) dissociation constants.
Direct horizontal advection is not taken into account in this
model because this effect is probably weak on a monthly scale,
and MLD, S, T, P, and Chl constraints already contain some
In this study, we needed to build the forcing fields for the model.
As in all similar studies, data availability is the first limitation one
can expect. Therefore, the data processing aims at interpolating the
available information on the required time and space scales. Here,
we describe the main steps in our approach.
The extensive quality control procedures of the data already
made by the MEDAR/MEDATLAS group allow us to extract the
data for depths between the surface and 1000 m that have successfully passed the tests. We chose to extract the data already interpolated to the standard depths identified by the MEDAR/
MEDATLAS group. The extracted variables are T, S, Chl, N, Si,
P, and oxygen (O2). We did not include data from too close to
the coast, because they do not represent open-sea conditions
and could be highly influenced by local freshwater and anthropogenic inputs.
We adopted the 29 boxes defined by MEDAR/MEDATLAS as a
first step in our own quality control and to determine the consistency of the data within these zones. We applied a variance-based
analysis to the temperature and salinity data from these 29
boxes. This analysis helps to establish whether these zones are
effectively different from each other in terms of hydrological variability. The results of this analysis allowed us to identify 18 independent zones (Figure 1). This grid is used for subsequent binning of
the data and in applying the box model.
The data for each variable and depth are then averaged monthly
for all 18 zones for the period running from January 1960 to
December 1999. We note that before 1955, the data are too
scarce to be interpolated (Nacef, 2006).
For spatial interpolation, we used linear regressions between a
variable in a given zone and the same variable located in adjacent
zones to fill in missing values. Only regressions with a high correlation coefficient (r . 0.9) were applied in this process. When the
correlation was weak, the value of the variable located in a given
zone was taken as the average for the same variable in the adjacent
zones.
For temporal interpolation, we used the binned and spatially
interpolated database to construct 3- and 5-month moving
averages. For example, for January 1978, we have one binned
field representing an average of the data for the same month,
a second representing the averaged data from December 1977 to
1540
F. Louanchi et al.
Table 1. Forcing fields used to constrain the surface box model,
their sources, and relative error estimates.
Parameter
Incoming solar
radiation
Windspeed
MLD
Temperature
Salinity
Oxygen
Chlorophyll
Phosphates
Nitrates
Silicates
Sources
ERA-40 (ECMWF archived data
and products)
ERA-40 (ECMWF archived data
and products)
Climatologies based on MEDAR/
MEDATLAS (2002) data
Climatologies based on MEDAR/
MEDATLAS (2002) data
Climatologies based on MEDAR/
MEDATLAS (2002) data
Climatologies based on MEDAR/
MEDATLAS (2002) data
Climatologies based on MEDAR/
MEDATLAS (2002) data
Climatologies based on MEDAR/
MEDATLAS (2002) data
Climatologies based on MEDAR/
MEDATLAS (2002) data
Climatologies based on MEDAR/
MEDATLAS (2002) data
Relative
error (%)
n/a
n/a
n/a
7.04
0.82
6.80
62.5
53.1
51.8
41.2
February 1978, and a third representing the averaged data from
November 1977 to March 1978. When the one-binned datum is
missing, it is replaced by the three- or five-binned one.
At this stage, our database is a monthly climatology of each
variable over the 18 zones and covering the period from January
1960 to December 1999. We constructed the average T, S, O2,
Chl, P, N, and Si monthly climatologies for four decades (1960s,
1970s, 1980s, and 1990s), and from the surface to 1000 m. The
present study only requires averaged parameters in the mixed
layer and in the subsurface layer. Because the box model is suitable
only for surface waters, we defined the subsurface layer as ranging
from the MLD to 300 m. Indeed, D’Ortenzio et al. (2005) demonstrated that the typical winter MLD was 300 m in the
Mediterranean Sea. The MLD is defined as the depth at which
the temperature is 18C lower than that found at the surface for
each month.
Error estimates for the monthly fields are difficult to assess. The
data were obtained during different cruises and periods, and using
different measurement techniques. We neglect the latter, because
they must be much smaller than those that result from
interpolation processing, or are caused by interannual variability
within the same decade. We therefore used the standard deviation
in the 1-month binned fields and followed a procedure similar to
the processing of the data themselves. Table 1 summarizes
the averaged relative errors found for each variable in the
Mediterranean Sea. Although the relative errors on the biogeochemical variables seem huge, they are not important in
absolute value. In fact, the concentrations in nutrients and
chlorophyll in the Mediterranean Sea are generally considered
low relative to other parts of the world’s oceans.
Sensitivity tests were performed to assess uncertainty in the fCO2
fields relative to errors in the forcing field. The errors in
the biogeochemical variables (nutrients and chlorophyll) result
in an average error in fCO2 of 9 matm (relative error of 2.5%).
In addition, fCO2 is sensitive to the errors in the temperature and
salinity fields, which result in an average error on fCO2 of
19 matm (relative error of 5.3%), which implies that the sensitivity to hydrographic variables is more important than that
linked to the biogeochemical properties of the Mediterranean Sea.
Subsurface DIC and TA definition
Subsurface TA is a linear function of subsurface salinity for
each decade. Such a relationship has been described by several
authors for the Mediterranean Sea (Copin-Montégut, 1993;
Begovic and Copin-Montégut, 2002; Copin-Montégut and
Begovic, 2002; Aı̈t-Ameur, 2007). The relationship applied in
this study is derived from the METEOR-51 cruise data (available
from CDIAC) that took place in 2001:
TA ¼ 71:488 S 196:55 ðr2 ¼ 0:97Þ:
ð2Þ
The error linked to this regression does not exceed 10 meq l21,
even when it is applied to other datasets [e.g. DYFAMED
(DYnamique des Flux Atmosphériques en MEDiterranée) site observations (http://www.obs-vlfr.fr/jgofs2/sodyf/home.htm)].
Subsurface DIC changes from one decade to another follow
anthropogenic carbon penetration into the sea. It is therefore calculated as the sum of a natural component DICn and an anthropogenic part DICa. DICn is taken from the multilinear equation
established by Aı̈t-Ameur (2007), based on observations between
1998 and 2005 at the DYFAMED site:
DICn ¼ 1371:65 21:1841 T þ 0:8393 AOU
þ 99:8395 S;
ð3Þ
where AOU stands for apparent oxygen utilization, calculated
using an oxygen-saturation polynomial described by Garcia and
Gordon (1992).
The DICa is derived from a Revelle factor b of 10 [according to
the Mediterranean Sea temperature range and the global ocean’s
average proposed by Takahashi et at. (1993)]. We assumed that
the atmosphere and the sea were at equilibrium over a decade:
DICa ¼
DICn D pCOa2
;
b
pCOn2
ð4Þ
where DpCOa2 is the atmospheric pCO2 difference between the
decade of concern and the pre-anthropogenic condition, and
pCOn2 the pre-anthropogenic atmospheric partial pressure of
280 matm. It is difficult to evaluate the errors in subsurface DIC.
Consequently, we computed it from the errors in T, S, and
AOU. The computation gives a relative error ,10%.
A validation of this approach for the estimation of subsurface
DIC was attempted. For this, we used temperature, salinity, and
oxygen METEOR-51 data to reconstruct DICn in subsurface
waters at each station. We calculated DICa using the atmospheric
pCO2 of 372.35 matm at the Lampedusa site (GAW programme) in
the Mediterranean Sea in October 2001. The estimated subsurface
DIC was compared with measured subsurface DIC from the
METEOR-51 cruise. The differences did not exceed 15 mmol l21.
Moreover, our approach predicts an anthropogenic DIC of
70 mmol l21 for the METEOR-51 data, which matches
Aı̈t-Ameur’s (2007) results at the DYFAMED site for the period
1998–2004.
We also conducted a sensitivity analysis to assess the magnitude
of the errors in the fCO2 field attributable to errors in the subsurface
Decadal changes in surface carbon dioxide in the Mediterranean
Table 2. Subsurface (from the MLD to 300 m) DIC and TA used as
constraints for the surface box model for each decade and
averaged over the western and eastern Mediterranean Sea.
Western Mediterranean
Sea
Decade
1960 –1969
1970 –1979
1980 –1989
1990 –1999
DIC
(mmol l21)
2 198.1
2 205.8
2 217.0
2 228.2
TA
(meq l21)
2 530.6
2 537.7
2 543.4
2 531.3
Eastern Mediterranean
Sea
DIC
(mmol l21)
2 211.5
2 219.3
2 230.5
2 241.8
TA
(meq l21)
2 568.2
2 571.9
2 570.7
2 565.0
1541
properties. We found an average fCO2 error of 11 matm (relative
error of 3%).
Table 2 summarizes the subsurface constraints applied to each
decade and averaged over the western Mediterranean Sea (WMS)
and eastern Mediterranean Sea (EMS). Our calculations indicate
that subsurface DIC increased by 30 mmol l21 from the 1960s
to the 1990s, because of anthropogenic carbon penetration. On
average, subsurface TA varied over a range of 13 meq l21 in the
WMS to 7 meq l21 in the EMS, in response to decadal variability
in salinity. In the WMS, subsurface TA increased from the 1960s
to the 1980s following the salinity increase mentioned by
Béthoux et al. (1998). This trend was less clear for the EMS
Figure 2. The forcing field averages (MLD, T, S, Chl, and P) over the WMS (left panels) and the EMS (right panels). The solid grey lines
represent the 1960s, the solid black lines the 1970s, the dotted grey lines the 1980s, and the dotted black lines the 1990s.
1542
Figure 3. (a) Modelled annual DfCO2 (matm) cycle as averaged over
the four decades and the whole Mediterranean Sea. The envelope
surrounding the solid line illustrates the errors derived from the
forcing fields and subsurface properties. (b) Seasonal cumulative
effects on fCO2 (matm) as averaged for the four decades and the
whole Mediterranean Sea. In the legend, F indicates the air –sea
exchange effect, B the biological effect, M the mixing effect, and T
the thermal effect.
basin. The low subsurface TA variability seems to have been caused
by a low level of calcification there.
Results and discussion
The model was applied to the 18 zones for each of four decades,
using the forcing fields and subsurface conditions defined above.
To estimate the errors in model results, we also ran the model
applying the estimated errors in the forcing fields (Table 1),
together with those defined for the subsurface properties. On
average, we found a relative error of 7% in fCO2.
Forcing fields, annual CO2 cycle, and validation
Figure 2 shows the average forcing fields for the WMS and the
EMS. MLD peaked at 175 m in February/March in both
basins, then decreased to 10 m during summer. The annual
cycles in MLD appear not to have changed significantly over the
decades (Figure 2a and b). Annual cycles of temperature were
also similar for the two basins, except that the EMS exhibited
higher temperatures (18C) for both winter and summer
(Figure 2c and d). In the WMS, the highest temperatures were
during the 1960s, when seasonal variation of 9.968C was highest.
The lowest temperatures and seasonal variations were in the
1990s. A similar trend was prevalent in the EMS, except that
both the highest temperatures and seasonal variations were
encountered in the 1980s. The EMS was generally more saline
than the WMS, by 0.5 psu (Figure 2e and f). That is related to
the climate-induced intense evaporation from the eastern basin
F. Louanchi et al.
Figure 4. (a) Observed (solid line) and modelled (black square) fCO2
along the west– east track of the INDOMED 3 cruise in December
1977. Error bars indicate the range of the modelled values as related
to the forcing fields and subsurface properties errors. (b) Observed
(diamond, 1998; square, 1999; triangle, 2000) and 1990s modelled
(plain line) fCO2 at the DYFAMED site. Error bars indicate the
modelled fCO2 errors derived from the forcing fields and subsurface
properties.
that results in Levantine Intermediate Water formation.
Although Chl fields resulted in the highest uncertainties, the
annual cycles found in both basins are in agreement with earlier
observations (Antoine and Morel, 1995). It is worth noting that
the Mediterranean Sea has more than one productive season
(Figure 2g and h). In the WMS, Chl peaked in late winter,
summer, and autumn. These results agree with those of Moran
and Estrada (2005) from the northwest Mediterranean Sea. In
the EMS, the productive period seems to be during spring and
autumn. Phosphate concentrations were always ,0.3 mmol l21
in the WMS and 0.15 mmol l21 in the EMS, illustrating the oligotrophic nature of the Mediterranean Sea (Figure 2i and j). There
was a strong increase in P from the 1960s to the 1980s in the
WMS. It resulted either from an increase in the anthropogenic discharge (Béthoux et al., 1998) or from the dryness of those years,
which favoured continental erosion and phosphorus deposition
in the sea (Ridame et al., 2003). However, over the past decade,
P concentrations seem to have decreased to their earlier values
in the WMS. These trends were not observed in the EMS, where
the highest P values was in the 1970s, which matches the increase
in Chl (Figure 2h and j).
A further comment can be made regarding the observed
trends in the forcing fields. We did not find in our decadal climatologies a clear increase in surface temperature in the
Mediterranean Sea. On the contrary, we found that the magnitude
of the seasonal variation in temperature decreased by 50%
during the 1990s relative to the earlier decades. The summer
temperature for the 1990s was 58C lower than observed during
the 1960s, 1970s, and 1980s.
Decadal changes in surface carbon dioxide in the Mediterranean
1543
Figure 5. Averaged DfCO2 (matm) in the Mediterranean Sea over the (a) 1960s, (b) 1970s, (c) 1980s, and (d) 1990s. Red corresponds to
super-saturation and blue to under-saturation.
1544
F. Louanchi et al.
Table 3. Integrated FCO2, PP, EP, annual cumulative air– sea flux (F), biological (B), and mixing (M) effects on fCO2, and temperature
seasonal magnitude (DT ) for each decade and averaged for the whole Mediterranean Sea.
Magnitude of processes
Decade
1960 –1969
1970 –1979
1980 –1989
1990 –1999
FCO2 (range;
Tg C year21)a
+0.62 (23.74 to 4.80)
+0.13 (23.53 to 3.54)
20.14 (24.17 to 3.75)
21.98 (25.92 to 1.80)
PP (range;
Tg C year21)a
643 (141–1 341)
961 (218–1 956)
773 (183–1 540)
670 (146–1 403)
EP (range;
Tg C year21)a
71 (16 –149)
107 (24 –218)
86 (20 –171)
74 (16 –155)
F effect
(matm year21)
211.2
29.9
213.6
21.8
B effect
(matm year21)
220.9
229.6
227.1
221.1
M effect
(matm year21)
32.4
40.5
40.4
23.4
DT
(88 C)
8.90
9.51
10.02
4.65
a
1 Tg C ¼ 1012 g C.
The modelled annual cycle of DfCO2 and its driving forces are
presented as an average over the whole Mediterranean Sea and
all decades (Figure 3), because these features were quite similar
for the EMS and the WMS or for several decades. We found that
DfCO2 reached a minimum in February or March, following the
lowest temperature, and increased by 80 matm through the
summer. Disregarding the interdecadal variability, winter CO2
under-saturation seems to have been compensated for by
summer CO2 super-saturation, resulting in a near-equilibrium
state for the whole 1960– 1999 period (Figure 3a). This annual
cycle is similar to that observed in the northwestern
Mediterranean Sea (Begovic and Copin-Montégut, 2002;
Aı̈t-Ameur, 2007). According to our model results, the processes
driving DfCO2 variations are dominated by the thermal component
(Figure 3b). Air–sea CO2 fluxes act to decrease DfCO2 in summer
by outgassing, whereas this effect is negligible during the rest of the
year. Biological effects compensate for the mixing effects over the
year, which is consistent with what Begovic and Copin-Montégut
(2002) observed at the DYFAMED site. Therefore, the thermal
component appears to be the main driver of seasonal variations
in DfCO2, and it is worth noting that most of the oligotrophic
oceanic regions follow the same process (Winn et al., 1994;
Bates et al., 1996; Metzl et al., 1998). In winter, all the effects are
low, because variations in the forcing field are weak then.
Just a few fCO2 data were available to evaluate model results over
the study period. Surface fCO2 data from the DYFAMED site
(Begovic and Copin-Montégut, 2002) and underlying fCO2
measurements from the INDOMED 3– 4 cruise (R. F. Weiss, available from CDIAC) were used to evaluate the model results.
Because the DYFAMED data were acquired between 1998 and
2000, we extracted the model results from the appropriate grid
box for the 1990s. INDOMED 3– 4 took place in December
1977, and the track crossed the southern Mediterranean Sea
basin from the Gibraltar Strait to Alexandria in the EMS, so we
extracted all the model results from the southern grid boxes for
December in the 1970s. Observations and model results are compared in Figure 4. It appears that the model was able to reproduce
the order of magnitude of fCO2 in both cases. The averaged differences between the observation and the model did not exceed
10 matm. Most of the differences could be accounted for by the
differences between the in situ measured temperature and the
average decadal temperature that we used. The observed west –
east gradient in the southern part of the Mediterranean Sea was
also well reproduced (Figure 4a), together with the annual
cycle in fCO2 at DYFAMED (Figure 4b). These results demonstrate
that the forcing fields and model parameterization reproduced the
decadal variations in fCO2 in the Mediterranean Sea reasonably
well.
Decadal CO2 variations in the Mediterranean Sea
Figure 5 shows the decadal DfCO2 averages from the 1960s
(Figure 5a) to the 1990s (Figure 5d). The Mediterranean Sea was
transforming from a net source of CO2 in the 1960s to a net
sink in the 1990s, and this was particularly true for the
Levantine and the Algerian –Provincial basins. For all decades,
there was a west–east gradient in the distribution of DfCO2, with
lower values in the EMS than in the WMS. The Tyrrhenian Sea
appears to have been a net CO2 source for the atmosphere from
the 1960s to the 1980s and only transformed into a sink in the
1990s. It is worth noting that the north Adriatic Sea exhibited
strong decadal variability, with CO2 under-saturation during the
1960s and the 1980s and CO2 super-saturation during the 1970s
and the 1990s. Such decadal variations might be related to
changes in hydrological conditions (CIESM, 2002), following the
North Atlantic Oscillation patterns. Overall, DfCO2 changed from
about +2 matm during the 1960s to 213 matm during the
1990s, the magnitude of this signal being higher in the WMS
(20 matm) than in the EMS (12 matm). The decrease in
DfCO2 does not mean that sea surface fCO2 also decreased. On the
contrary, the latter increased by 24 matm over time, following
the anthropogenic carbon input to the sea, but not as fast as the
atmospheric fCO2 did.
We noted a decrease of 0.02 in pH between the 1960s and the
1990s (not presented here), and this decrease was not linear over
time, but followed the interdecadal variations in DIC, T, and
S. This pH change is not significant regarding the errors in fCO2
results.
Table 3 summarizes the decadal air–sea CO2 fluxes (FCO2), the
PP, the EP, and the magnitude of processes over the whole
Mediterranean Sea. The model results indicate that the PP and
the EP ranged from 18 to 27 mol C m22 year21 and from 2 to
3 mol C m22 year21, respectively. These values agree closely with
earlier observations in the Mediterranean Sea (Béthoux, 1989;
Marty and Chiavérini, 2002; Moran and Estrada, 2005). The PP
derived from in situ observations ranged from 3 to
60 mol C m22 year21 in the WMS, whereas the EP ranged from
0.2 to 54 mol C m22 year21. Our modelled PP ranges also
matched those derived from CZCS satellite Chl observations of
308 Tg C year21 for the entire Mediterranean Sea (Antoine and
Morel, 1995).
The value of FCO2 varied from 0.62 Tg C year21 during the
1960s to –1.98 Tg C year21 during the 1990s. The decadal FCO2
variations appear to be clearly related to the balance between the
two main processes. The magnitude of seasonal variations in
temperature was high during the 1970s and 1980s, illustrated by
the high summer values found in both basins (Figure 2c and d),
which should have increased fCO2, but this effect was partly
1545
Decadal changes in surface carbon dioxide in the Mediterranean
counterbalanced by increases in PP, which peaked during those
two decades. The strong sink during the 1990s was probably
related to the reduced seasonal variations in temperature estimated
from the data.
Conclusions
Thanks to the release of the MEDAR/MEDATLAS (2002) database, which we used to construct the forcing fields constraining
our diagnostic model, the decadal fCO2 changes in the
Mediterranean Sea from the 1960s to the 1990s could be estimated
for the first time. Our model results agreed well with the available
observations in terms of fCO2 and production (PP, EP). The processes that drive changes in fCO2 at seasonal scales also matched
earlier observations for a restricted area of the Mediterranean Sea.
The Mediterranean Sea transformed from a net CO2 source for
the atmosphere during the 1960s (0.62 Tg C year21) to a net CO2
sink during the 1990s (21.98 Tg C year21) following a balance
between decadal temperature anomalies and PP. The spatial
pattern of the sources and sinks indicates that the southern part
of the Mediterranean Sea was a stronger sink than the northern
part and that there was a west–east gradient, resulting in greater
sinks in the EMS than in the WMS.
Although DfCO2 decreased over time by 15 matm for the
whole Mediterranean Sea, sea surface fCO2 increased, following
the anthropogenic carbon input to the sea of 30 mmol l21 and
the gradual atmospheric CO2 increase that took place. However,
sea surface fCO2 only increased by 24 mtm, because of a strong
temperature anomaly in the 1990s, resulting in a 58C decrease in
mean summer temperatures. We believe this trend resulted from
a climate anomaly, and that this hypothesis needs to be investigated further.
According to our model results, the highest PP and EP were
present during the 1970s and the 1990s. This could have been
caused by the dryness of those years in this region, resulting in a
“greening” of the Mediterranean Sea through increased atmospheric input of nutrient. Our results indicate that PP ranged
from 18 to 27 mol C m22 year21, depending on the decade,
whereas EP ranged from 2 to 3 mol C m22 year21. This resulted
in an f ratio of 0.12, which is typical of the Mediterranean Sea
oligotrophic regime.
The decreasing trend in the magnitude of the seasonal variation
in temperature encountered in the MEDAR/MEDATLAS (2002)
data from the 1960s to the 1990s has to be confirmed by other
datasets and/or if more recent data become available (e.g. 2000
decade). If this trend persisted, it would most likely have had a significant effect on the Mediterranean Sea marine ecosystems, as well
as on the marine biogeochemical cycles.
A change in hydrological conditions driven by climate variability, such as observed in the Mediterranean Sea during the 1990s,
might affect the anthropogenic carbon penetration into the sea,
because cold waters favour the solubility pump. Aı̈t-Ameur and
Goyet (2006) pointed out that the Mediterranean Sea is a source
of anthropogenic carbon for the Atlantic Ocean. Therefore,
more extensive observation programmes are needed to study the
carbon penetration and the air–sea CO2 exchange in that area.
Some regions of the Mediterranean Sea exhibited a strong
decadal variability in both hydrological and biogeochemical properties, which might well be associated with the North Atlantic
Oscillation, or other climate anomalies. We were not able to
explore this hypothesis with the box model that we used in this
study. Further investigations on both data and three-dimensional
model results are needed to understand the coupling between the
sea and climate variability in this region.
Acknowledgements
We thank the MEDAR/MADATLAS group for the Mediterranean
Sea database release. We also thank the CDIAC for making available carbon parameter datasets. Pierre Pepin and two anonymous
reviewers are acknowledged for their stimulating comments,
which helped to improve the manuscript and enhanced the discussions. We also thank F. Z. Belkhiter-Louanchi for her help on an
earlier version of the manuscript. This study formed part of the
activity within the EUR-OCEANS Network of Excellence coordinated by Paul Tréguer and Louis Legendre.
References
Aı̈t-Ameur, N. 2007. Contribution à la quantification du CO2 anthropique en Méditerranée et en mer de Ross. PhD thesis, University of
Perpignan, France. 236 pp.
Aı̈t-Ameur, N., and Goyet, C. 2006. Distribution and transport of
natural and anthropogenic CO2 in the Gulf of Cadiz. Deep Sea
Research II, 53: 1329– 1343.
Anderson, L. A. 1995. On the hydrogen and oxygen content of marine
phytoplankton. Deep Sea Research I, 42: 1675– 1680.
Antoine, D., and Morel, A. 1995. Algal pigment distribution and
primary production in the eastern Mediterranean as derived
from coastal zone color scanner observations. Journal of
Geophysical Research, 100: 16193– 16209.
Bates, N. R., Michaels, A. F., and Knap, A. H. 1996. Seasonal and interannual variability of carbon dioxide species at the U. S. JGOFS
Bermuda Atlantic Time-Series Study (BATS) site. Deep Sea
Research II, 43: 347– 383.
Begovic, M., and Copin-Montégut, 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 Research II, 49: 2031– 2047.
Béthoux, J-P., Gentili, B., Morin, P., Nicolas, E., Pierre, C., and
Ruiz-Pino, D. 1999. The Mediterranean Sea: a miniature ocean
for climatic and environmental studies and a key for the climatic
functioning of the North Atlantic. Progress in Oceanography, 44:
131– 146.
Béthoux, J-P., Gentili, B., and Tailliez, D. 1998. Warming and freshwater budget change in the Mediterranean since the 1940s, their
possible relation to the greenhouse effect. Geophysical Research
Letters, 25: 1023 – 1026.
Béthoux, J-P., Morin, P., and Ruiz-Pino, D. 2002. Temporal trends in
nutrient ratios: chemical evidence of Mediterranean ecosystem
changes driven by human activities. Deep Sea Research II, 49:
2007– 2016.
Béthoux, P. 1989. Oxygen consumption, new production, vertical
advection and environmental evolution in the Mediterranean
Sea. Deep Sea Research, 36: 769– 781.
Caddy, J. F., Refk, R., and Do-Chi, T. 1995. Productivity estimates for
the Mediterranean: evidence of accelerating ecological change.
Ocean and Coastal Management, 26: 1 – 18.
CIESM. 2002. Tracking long-term hydrological change in the
Mediterranean Sea. CIESM, Workshop Series, Monaco, 16. 134 pp.
Copin-Montégut, C. 1993. Alkalinity and carbon budgets in the
Mediterranean Sea. Global Biogeochemical Cycles, 7: 915– 925.
Copin-Montégut, C., and Begovic, M. 2002. Distribution 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 vertical mixing on air– sea CO2 and O2 exchanges. Deep
Sea Research II, 49: 2049 – 2066.
D’Ortenzio, F., Iudicone, D., de Boyer-Montégut, C., Testor, P.,
Antoine, D., Marullo, S., Santoleri, R., et al. 2005. Seasonal
1546
variability of the mixed layer depth in the Mediterranean Sea as
derived from in situ profiles. Geophysical Research Letters, 32:
LI2605. doi 10.1029/2005GL022463.
DOE. 1994. Handbook of the Analysis of the Various Parameters of
the Carbon Dioxide System in Sea Water. Ed. by A. G. Dickson,
and C. Goyet. ORNL/CDIAC-74.
Garcia, H. E., and Gordon, L. I. 1992. Oxygen solubility in seawater:
better fitting equations. Limnology and Oceanography, 37:
1307– 1312.
Lee, K., Millero, F. J., Byrne, R. H., Feely, R. A., and Wanninkhof, R.
2000. The recommended dissociation constants for carbonic acid
in seawater. Geophysical Research Letters, 27: 229– 232.
Louanchi, F., Hoppema, M., Bakker, D. C. E., Poisson, A., Stoll, M., de
Baar, H. J., Ruiz-Pino, D., et al. 1999a. Modelled and observed sea
surface fCO2 in the Southern Ocean: a comparative study. Tellus,
51B: 541– 559.
Louanchi, F., Metzl, N., and Poisson, A. 1996. Modelling the monthly
sea surface fCO2 fields in the Indian Ocean. Marine Chemistry, 55:
265– 279.
Louanchi, F., Ruiz-Pino, D., and Poisson, A. 1999b. Temporal variations of mixed-layer oceanic CO2 at JGOFS-KERFIX time-series
station: physical versus biogeochemical processes. Journal of
Marine Research, 57: 165 – 187.
Malanotte-Rizzoli, P., Manca, B. B., Ribera d’Alcalà, M., Theocharis,
A., Brenner, A., Budillon, G., and Ozsoy, E. 1999. The Eastern
Mediterranean in the 80s and in the 90s: the big transition in the
intermediate and deep circulations. Dynamics of Atmosphere
and Oceans, 29: 365– 395.
Marty, J-C., and Chiavérini, J. 2002. Seasonal and internannual variations in phytoplankton production at DYFAMED time-series
station, northwestern Mediterranean Sea. Deep Sea Research II,
49: 2017– 2030.
MEDAR/MEDATLAS Group. 2002. MEDATLAS/2002 Database. 5
CD-ROM. IFREMER.
Mehrbach, C., Culberson, C. H., Hawley, J. E., and Pitkowitcz, R. M.
1973. Measurement of the apparent dissociation constants of carbonic acid in seawater at atmospheric pressure. Limnology and
Oceanography, 18: 897 –907.
Metzl, N., Louanchi, F., and Poisson, A. 1998. Seasonal and interannual variations of sea surface carbon dioxide in the subtropical
Indian Ocean. Marine Chemistry, 60: 131– 146.
Mikaloff Fletcher, S. E., Gruber, N., Jacobson, A. R., Doney, S. C.,
Dutkiewicz, S., Gerber, M., Follows, M., et al. 2006. Inverse estimates of anthropogenic CO2 uptake, transport and storage
F. Louanchi et al.
by the ocean. Global Biogeochemical Cycles, 20, GB2002.
doi: 10.1029/2005GB002530.
Millero, F. J., Graham, T. B., Huang, F., Bustos-Serrano, H., and
Pierrot, D. 2006. Dissociation constants of carbonic acid in seawater as a function of salinity and temperature. Marine
Chemistry, 100: 80– 94.
Moran, X. A. G., and Estrada, M. 2005. Winter pelagic photosynthesis
in the NW Mediterranean. Deep Sea Research I, 52: 1806– 1822.
Nacef, L. 2006. Etude des variations spatio-temporelles des flux de
chaleur à l’interface air-mer en Méditerranée Applications à la prévision climatique. Masters dissertation, ISMAL. 106 pp.
Orr, J. C., Fabry, V. J., Aumont, O., Bopp, L., Doney, S. C., Feely, R. A.,
Gnanadesikan, A., et al. 2005. Anthropogenic ocean acidification
over the twenty-first century and its impact on calcifying organisms. Nature, 437: 681 –686.
Peng, T-H., Takahashi, T., Broecker, W. S., and Olafsson, J. 1987.
Seasonal variability of carbon dioxide, nutrients and oxygen in
the northern North Atlantic surface water: observations and a
model. Tellus, 39B: 439– 458.
Ridame, C., Moutin, T., and Guieu, C. 2003. Does phosphate adsorption onto Saharan dust explain the unusual N/P ratio in the
Mediterranean Sea? Oceanologica Acta, 26: 629 – 634.
Takahashi, T., Olafsson, J., Goddard, J. G., Chipman, D. W., and
Sutherland, S. C. 1993. Seasonal variation of CO2 and nutrients
in the high-latitude surface oceans: a comparative study. Global
Biogeochemical Cycles, 7: 843 – 878.
Taylor, A. H., Watson, A. J., Ainsworth, M., Robertson, J. E., and
Turner, D. R. 1991. A modelling investigation of the role of phytoplankton in the balance of carbon at the surface of the North
Atlantic. Global Biogeochemical Cycles, 5: 151– 171.
Wanninkhof, R., Lewis, E., Feely, R. A., and Millero, F. J. 1999. The
optimal carbonate dissociation constants for determining surface
water pCO2 from alkalinity and total inorganic carbon. Marine
Chemistry, 65: 291– 301.
Wanninkhof, R. H. 1992. Relationship between wind speed and gas
exchange over the ocean. Journal of Geophysical Research, C97:
7373– 7382.
Winn, C. D., Mackenzie, F. T., Carrillo, C. J., Sabine, C. L., and Karl,
D. M. 1994. Air– sea carbon dioxide exchange in the North
Pacific Subtropical gyres: implications for the global carbon
budget. Global Biogeochemical Cycles, 8: 157– 163.
doi:10.1093/icesjms/fsp049