You are on page 1of 32

Progress in Oceanography 44 (1999) 5–36

Recent changes in deep water formation and


spreading in the eastern Mediterranean Sea:
a review
a,*
Alex Lascaratos , Wolfgang Roether b, Kostas Nittis c,
Birgit Klein b
a
Department of Applied Physics, University of Athens, Athens, Greece
b
Institut fuer Umweltphysik, Universitaet Bremen, Bremen, Germany
c
National Centre for Marine Research, Athens, Greece

Abstract

Observations of the last decade testify that the characteristics of the deep thermohaline
circulation in the Eastern Mediterranean Sea have changed thoroughly. The source of the most
dense waters of the basin has moved from the Adriatic to the Aegean Sea. This new source
has proved to be very efficient since the estimated formation rate for the period 1989–95 was
more than 1 Sv, about three times more than the dense water formation rate of the Adriatic
Sea. These new waters with hydrological characteristics, that are warmer and more saline,
have replaced almost 20% of the older deep waters of the basin, and have uplifted the deep
isopycnals by about 500 m. This major event can be attributed to important meteorological
anomalies in the Eastern Mediterranean and to changes of circulation patterns. The extended
dry period of 1988–93 and the exceptionally cold winters of 1987 and 1992–93 created favor-
able conditions for increased formation of dense water in the Aegean. Furthermore, changes
in the circulation patterns in the intermediate water range (Levantine Intermediate Water LIW
and Cretan Intermediate Water), themselves possibly linked to meteorological anomalies,
appear to have played an important role in the redistribution of salt. As a result of an interrup-
tion to exchanges between the Ionian and Levantine Basin, the salinities in the latter started
to rise, high salinity waters were diverted into the Aegean [Malanotte-Rizzoli, P., Manca, B.
B., Ribera dAcala, M., & Theocharis, A. (1998). The Eastern Mediterranean in the 80s and
in the 90s: the big transition emerged from the POEM-BC observational evidence. Rapport
du Commission International de la Mer Medittanee, 35, 174–175] and the westward transport
of LIW was reduced. An additional effect of the deep water discharge from the Aegean and

* Corresponding author.
E-mail address: alasc@oc.phys.uoa.gr (A. Lascaratos)

0079-6611/99/$ - see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 7 9 - 6 6 1 1 ( 9 9 ) 0 0 0 1 9 - 1
6 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

the resulting uplifting of mid-depth waters was to lower salinities in the LIW layer. This effect
is most strongly felt in the Ionian Sea. A 3-D primitive equation numerical model for the
Eastern Mediterranean with a 20 km grid size is used to simulate the observed changes and
understand the basic mechanisms which caused them. Under appropriate atmospheric forcing
the model successfully reproduces the main characteristics of the transient. These results indi-
cate that the observed changes can be, at least partially, explained as a response of the Eastern
Mediterranean, and more specifically of the Aegean, to atmospheric forcing variability. 
1999 Elsevier Science Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2. A review of intermediate and deep water formation processes in the Mediterranean 8


2.1. Western Mediterranean deep water . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2. Eastern Mediterranean deep water . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3. Levantine intermediate water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4. Role of the Aegean Sea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3. The recent water mass changes in the Eastern Mediterranean . . . . . . . . . . . . 13

4. A numerical model approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24


4.1. Model setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2. The atmospheric forcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3. Model results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5. Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

1. Introduction

The Mediterranean Sea is a semi-enclosed marginal sea that communicates with


the world ocean through the narrow Strait of Gibraltar (Fig. 1). The negative fresh-
water and heat budgets of the basin drive its lagoonal (anti-estuarine) thermohaline
circulation and determines its hydrological characteristics. Low salinity Atlantic
water (AW) that enters in the upper layer of the Gibraltar Strait is transformed to
saline Mediterranean water that subsequently exits into the Atlantic via the lower
layer. Although the surface AW progressively loses its characteristics through mixing
and evaporation during its travel to the east, (salinity rising from 36.15 at Gibraltar
Strait, to 38.6 in the eastern Levantine Basin according to Lacombe & Tchernia,
1972; Ozsoy, Hecht & Unluata, 1989), the transformation to intermediate and deep
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 7

Fig. 1. The Mediterranean Sea. Shaded areas mark the sites of dense (intermediate and deep) water
formation. The inner frame indicates the area where the numerical model is applied.

water occurs in selected areas within the basin (Wüst, 1961). In these areas, favorable
oceanic conditions and extreme air–sea interaction processes lead to the downward
mixing extending deep into the water column.
This process of dense water formation is as important for the Mediterranean Sea
as it is for the world ocean. First of all, it allows the exchange of physical and
biochemical properties (e.g. oxygen, nutrients) between the surface and deep layers.
On a larger scale, this process maintains the permanent thermocline at low and mid
latitudes. Once the dense water leaves the surface and sinks to deep layers, it main-
tains its characteristics that were determined by the prevailing atmospheric conditions
at its site of formation. Therefore, the study of deep water characteristics gives us
access to past climatic variations at time scales depending on the renewal time scales
of the basin (for the Mediterranean: decades to centuries).
In the past few years a number of research projects in the framework of the
EU/MAST project have, among others, studied water mass formation processes and
the thermohaline circulation of the basin. These studies, both experimental and
numerical, have greatly increased our understanding of the functioning of the Medit-
erranean. At the same time, there has been a significant follow-up of the large scale
coordinated data collection efforts that were undertaken during the 1980s by the
POEM and PRIMO groups (POEM Group, 1992).
The most recent observational efforts have revealed that significant changes have
taken place during the past few years in the intermediate and deep thermohaline
circulation of the Eastern Mediterranean (Roether et al., 1996). As a result of these
changes within a few years a totally new pattern of thermohaline circulation has
developed in the basin. In this paper we will reconstruct the old and the new patterns
of circulation in the Eastern Mediterranean Sea, focusing upon water mass formation
and spreading processes. In Section 2, we review the existing knowledge on water
mass formation in the Mediterranean making extended reference to results recently
acquired in the framework of MAST/MTP projects. In Section 3 we review the
8 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

observations that reveal the new picture of the Eastern Mediterranean thermohaline
circulation. In Section 4, we present the first results of a numerical study that attempts
to simulate the transition from the old to the new state. The results of the work are
summarized and discussed in the last section (5) of this paper.

2. A review of intermediate and deep water formation processes in the


Mediterranean

The circulation of the Mediterranean Sea is usually described in a schematic way


in the main as an open thermohaline cell with two closed secondary cells, one for
each sub-basin (Fig. 2). The principal cell describes the transformation of the surface
Atlantic Water (AW) to the Levantine Intermediate Water (LIW), which is the main
contributor to the Mediterranean outflow into the Atlantic. The two secondary cells
describe the transformation of surface and intermediate water to Western Mediter-
ranean Deep Water (WMDW) and to Eastern Mediterranean Deep Water (EMDW).
The existence of an intermediate depth cell is mainly controlled (Phillips, 1966) by

Fig. 2. Schematic representation of the thermohaline circulation in the Mediterranean Sea.


A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 9

the presence of the two relatively shallow sills (Gibraltar and Sicily Straits). Further-
more, the sill in the Strait of Sicily prevents a direct communication between the
EMDW and WMDW but coupling is achieved via the LIW layer.

2.1. Western Mediterranean deep water

In the Western Mediterranean Sea, dense water is formed in the Gulf of Lions
during winter time by excess cooling and evaporation. A main preconditioning is
provided by the presence of a cyclonic gyre (Gascard, 1978) as in almost all the open
ocean convection regimes (Killworth, 1983). An important factor is the presence of
the salty LIW at intermediate depths. Formation occurs during strong Mistral events
(cold and dry winds from the north) that persist for a few days, during which heat
loss can exceed 1000 W/m2. The diameter of the area of convection can vary between
a few tens to ⬎100 km and vertically extend down to 苲2000 m, approximately the
full water depth in the area. Inside this patch, convection occurs in plumes with
horizontal scales less than 1 km (Gascard, 1973; Jones & Marshall, 1993). The pro-
cess was extensively studied by the MEDOC group during a number of oceano-
graphic surveys conducted in the area during the late 1960s (MEDOC Group, 1970).
During the last few years, new methods such as acoustic Doppler profiling for the
measurement of vertical velocities have significantly improved our understanding of
how the dense water forms in the area (Schott & Leaman, 1991; Gaillard, Desaubies,
Send & Schott, 1997).

2.2. Eastern Mediterranean deep water

Deep and bottom waters of the Eastern Mediterranean Sea are mostly formed in
the south Adriatic Sea with small contributions from the dense waters formed in the
northern part (Pollak, 1951). In the shallow waters of the north Adriatic, the forma-
tion is by a typical shelf sea formation process: the water on the shelf becomes
denser during winter time as a result of cooling and evaporation associated with
local meteorological conditions (Bora events). The dense water moves from the north
to the central and southern Adriatic following the western coast of the Adriatic,
where it is mixed and stored in the Jabuka Pit depression. It seems to contribute,
together with the inflowing LIW, to the formation of dense water in the southern
Adriatic. In this area an open ocean formation process occurs that is similar to the
Western Mediterranean and is strongly preconditioned by the presence of a perma-
nent cyclonic gyre. The Adriatic Deep Water (ADW) that exits through the Otranto
Strait has a typical temperature of 13°C and salinity of 38.65 (Gacic et al., 1996).
First it fills the deepest parts of the Ionian Sea and then of the Levantine Sea. The
deep waters in these basins are very homogenous in temperature and salinity (13.3°C,
38.663) and display only small horizontal gradients (Schlitzer et al., 1991). Roether
and Schlitzer (1991) computed an annual mean formation rate of 苲0.3 Sv for the
ADW. The same estimate was made by Lascaratos (1993) for both the Adriatic and
the Gulf of Lions formation areas.
10 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

2.3. Levantine intermediate water

The intermediate layers of the basin (200–500 m) are occupied by the saline Lev-
antine Intermediate Water. This water mass is known to be formed in the permanent
Rhodes cyclonic gyre in the NW part of the Levantine Basin (Ovchinnikov, 1984;
Malanotte-Rizzoli & Hecht, 1988; Lascaratos, Williams & Tragou, 1993). During
summer, the surface layers of the Levantine Sea are occupied by a warm and salty
water mass known as Levantine Surface Water (LSW). Winter cooling increases the
density of this water that so that it sinks and mixes with underlying water inside the
Rhodes cyclone during February and March.
The localization of deep mixing by the presence of a cyclonic circulation is similar
to the other two sites of dense water formation (Gulf of Lions, southern Adriatic).
Lascaratos et al. (1993) using a 1-D Krauss-Turner mixed layer model, found that
even with a horizontal homogeneous heat loss pattern over the Levantine, intermedi-
ate water is formed only in the cyclonic area as a result of the preconditioning by
the hydrology. The same conclusion was reached by Lascaratos and Nittis (1998)
who applied a high resolution 3-D model to the Levantine Sea; they found that under
mean climatological conditions, LIW is only formed inside the Rhodes gyre where
the doming of isopycnals promotes deep mixing.
There are, nevertheless, observations of intermediate water formation in other
areas of the Levantine Sea. Sur, Ozsoy and Unluata (1992) found LIW formation
in the northern Levantine Sea with simultaneous deep water formation in the center
of the Rhodes gyre during the cold winter of 1992. The south Aegean Sea and the
south Levantine Sea have also been proposed as sites of LIW formation (Wüst,
1961; Morcos, 1972). Nittis and Lascaratos (1998) propose that this is a result of
the interannual variability of atmospheric forcing over the area. Using meteorological
data from different years, they found that both the variability of the mean buoyancy
loss and the characteristics of the synoptic scale atmospheric events modulate the
LIW formation process. During very cold winters, the formation area is extended
over the whole north Levantine Basin while deep water is formed inside the Rhodes
gyre. The spatial variability of forcing is also capable of shifting the area of inter-
mediate water formation towards the south Levantine. The interannual variability of
atmospheric forcing affects also the annual LIW formation rates; they vary between
0.6 and 1.3 Sv with a typical climatological value of 苲1.0 Sv. This value is consistent
with previous estimates, based on completely different methods (Ovchinnikov, 1984;
Tziperman & Speer, 1994; Lascaratos, 1993).
The controlling role of baroclinic eddies formed at the periphery of the convection
area was also revealed by the high resolution numerical model of Lascaratos and
Nittis (1998). Their presence in the open ocean has been known for some time
(Gascard, 1978) but their functioning in convection processes became understood
only recently as a result of laboratory experiments and schematic numerical models
(Jones & Marshall, 1993). The age distribution within the LIW, calculated from CFC
measurements, had confirmed that LIW is formed by open ocean convection
(Roether, Klein, Beitzel & Manca, 1998). Minimum apparent ages within the newly
formed LIW were found to be 5–6 years in 1987. Since open ocean convection
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 11

implies pronounced overturning with a fairly small contribution from original mixed
layer water, non-zero ages can be expected in such a process. The baroclinic eddies
were also found to be very effective in carrying the newly formed LIW away from
the source during the spreading phase (Wu & Haines, 1996). Similar results have
been presented by Madec, Chartier, Delecluse and Crepon (1991) for the Western
Mediterranean Sea.

2.4. Role of the Aegean Sea

The Aegean Sea has also been discussed as a possible source of deep water by
Nielsen (1912), Pollak (1951), Lacombe, Tchernia and Benoist (1958), Wüst (1961),
Miller (1963), Malanotte-Rizzoli and Hecht (1988), Pollak (1951), Lacombe, Tcher-
nia & Benoist (1958), Wüst (1961), Miller (1963) and Malanotte-Rizzoli & Hecht
(1988). The evidence is, however, that during the past decades the Aegean Sea has
at most been only an occasional and minor contributor to the deep and bottom waters
of the Eastern Mediterranean (see also Theocharis, Nittis, Kontoyannis, Papageor-
giou & Balopoulos, 1999). Hydrographic and tracer data obtained in 1987 (Schlitzer
et al., 1991) provided direct evidence of Aegean Sea-derived waters occurring to the
south of Crete, but these represented an intermediate-depth water mass. As confirmed
by later work (Malanotte-Rizzoli et al., 1998), the outflow from the Aegean Sea
formed a layer below the LIW, which was particularly distinct south of Crete and
along the western coast of Greece. Schlitzer et al., 1991 coined the name Cretan
Intermediate Water (CIW) for this water mass. Evidence for its presence is provided
by the enhanced concentrations of the anthropogenic tracer tritium (Roether, Beitzel,
Sueltenfuss & Putzka, 1999), which results from the replenishment of this water
mass from surface waters being distinctly faster than for the EMDW that occurs
around and below it. Along a W–E tritium section of the Eastern Mediterranean
(Fig. 3, upper panel), relatively higher tritium occurs in a layer between about 500
and 800 m depth to the south and west of Crete, compared with further east. Accord-
ing to a tritium map at 700 m depth (Fig. 3, lower panel), this signature is strongest
in the region south and west of Crete. In fact it appears from Fig. 3 that EMDW
extends up to distinctly shallower depths in the central and eastern Levantine Sea
than elsewhere.
CIW had never been reported in the literature previously, but it may have gone
unnoticed because, unlike the LIW above it, it is not accompanied by extreme values
of either temperature or salinity. Since its properties are intermediate between those
of LIW and EMDW, CIW might have been interpreted as if being a mixture of LIW
and EMDW (the so-called transition-layer water). However, T–S diagrams in the
depth range between LIW and EMDW show a trend towards higher salinity (e.g.
Wüst, 1961). Such a curvature may be indicative of CIW, since the Aegean subsur-
face waters have always been relatively more saline for a given density. Fig. 3 indi-
cates that the term transition layer is appropriate for the Levantine Sea, but somewhat
misleading for the Cretan Passage and the Ionian Sea.
The density contrast between LIW and the deep Adriatic waters has always been
quite small, amounting to approximately 0.15 kg/m3 (Wüst, 1961). This means that
12 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

Fig. 3. Tritium section along the eastern Mediterranean in 1987 (Meteor M5/6; upper panel), for stations
see inset map. Horizontal distribution of tritium at 700 m depth in 1987 (Meteor M5/6; lower panel).
Please note that isolines have been drawn for the whole basin and cover areas without supporting data.

the depth at which a deep outflow from the Aegean Sea stabilizes would show a
delicate dependence on quite minor variations in the properties of the dense waters
formed in the Adriatic and Aegean source areas. This concept is underlined by Fig.
4, which shows T–S diagrams from the south Adriatic Basin and the south Aegean
(Cretan Sea) for different years. The individual T–S relationships in both basins
show a distinct variability, which is to be expected in small basins such as these.
Still, Fig. 4 is indicative of a density excess of the Adriatic of generally as little as
0.05 kg/m3 (at approximately sill depths, i.e. 苲800 m; crosses in Fig. 4). It follows
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 13

that a quite moderate density increase of the source of CIW would be sufficient to
disturb severely the situation that apparently prevailed up to 1987.

3. The recent water mass changes in the Eastern Mediterranean

The caveat that there was a delicate density balance between the former water
masses is obvious in retrospect. But when observations revealed that the classical
thermohaline circulation in the Eastern Mediterranean as previously described had
entered an entirely new state (Roether et al., 1996) this came as a big surprise. As
Fig. 5 demonstrates, by 1995 rather salty (and warm) overflow from the Aegean Sea
had replaced the previuos bottom and deep waters over much of the deep domain.
Direct proof of this origin of the saline waters is provided by Fig. 6, showing T–S
diagrams for 1987 and 1995 in some Eastern Mediterranean key regions. The Ionian
and Levantine Sea stations exhibit inversions in temperature and salinity, and the

Fig. 4. Outflow characteristics in the Strait of Otranto and Kassos Strait for different years. Thin solid
lines represent profiles in the Strait of Otranto from 1987 and the thick solid line a profile from 1961.
The thin dashed line represents a profile from Kassos Strait in 1987, the thick dashed line a profile
from 1961. Water mass characteristics at 800 m are indicated by circles (Strait of Otranto) and asterisks
(Kassos Straits).
14 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

waters below the inversion are apparently mixtures of T–S characteristics, as


observed in 1995 at the sills of the Cretan Arc [these characteristics are indicated
as an asterisk; see Fig. 6(d)] and those of the previously residing bottom and deep
waters (Roether et al., 1996). Moreover increases in salinity between 1987 and 1995
were noted everywhere except in the western Ionian Sea [Fig. 6(a)]; in this sea there
has been a decrease in salinity for the mid-depth range while the deep waters remain
virtually unchanged. Thus it appears that the Aegean Sea has extended its role from
being a producer of just intermediate water to also generating deep and bottom waters
in large quantities. The salinity balance reveals that by 1995 as much as 20% of the

Fig. 5. Salinity section along the Eastern Mediterranean, 1987 (Meteor M5/6; upper panel) and 1995
(Meteor M31/1; lower panel), for stations see inset maps.
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 15

Fig. 6. T–S diagrams in 1987 (Meteor M5/6, thin line) and 1995 (Meteor M31/1, solid line) in the
western Ionian Basin (a), the eastern Ionian Basin (b), the Levantine basin (c) and the Aegean Sea (d).
Water mass properties at 1200 m are indicated by circles. The asterisks mark the outflow characteristics
at the Straits between the Cretan Arc at 800–1000 m depth in 1995.

deep waters of the Eastern Mediterranean had been replaced by this new Aegean
overflow (Roether et al., 1996). In contrast to the well-known, slow drifts in salinity
and temperature that have been observed over the entire Mediterranean for several
decades (Rohling & Bryden, 1992), this change represented a qualitative jump. By
this jump the Eastern Mediterranean thermohaline system has changed from a quasi-
steady state into a truly transient one. A general overview of the hydrography prior
to the event (1987) in comparison with that in 1995 has been given by Klein et al.
(1999). Recent work has started on elucidating the initiation and temporal develop-
ment of the transient. Although it is still too early to provide a fully coherent and
quantitative picture, the following description attempts to sketch out the picture that
is evolving.
Early analysis revealed that the event was related to rising salinity in the Cretan
Sea (Roether et al., 1996, which lead to an overflow of water from this sea through
the straits of the Cretan Arc that was sufficiently dense to enter the deep and bottom
waters beyond. In Fig. 7 we illustrate the time sequence of the changes in salinities,
temperatures and densities at 1000 m in the Cretan Sea (southern Aegean). The area
16 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

covered extends from 35.5°–36.5°N and 23.5°–27°E. The methodology used follows
Rohling and Bryden (1992). Initially, values at 1000 m at all stations within the area
are averaged for each cruise. Then, the cruise averages within the same year are
averaged to produce a single annual mean. The 1000 m depth has been chosen as
representative of the deeper layers of the Cretan Sea. Note also that at such depths
the seasonal variability signal is minimal.
Fig. 7 clearly shows that the waters of the Cretan Sea at 1000 m underwent a
continuous increase in density throughout the period 1987–95. This period of change
can be subdivided into two distinct periods: firstly 1987–92 when the density increase
was primarily the result of increasing salinities, the latter period 1992–95 appears
to have been caused by a lowering of temperature (by about 0.4°C). (Note that a
similar increase, although less important, of densities in the Cretan Sea occurred
during the early 1970s. However, this event differs from the present one in that the
increase of densities was linked to a synchronized increase in salinities and decrease

Fig. 7. Time series of mean values of temperature (a) salinity (b) and density (c) in 1000 m depth,
computed from all the available stations in the Cretan Sea of each cruise. The data are taken from the
MODB-MED5 data base (Brasseur et al., 1996) with the inclusion of data from ‘PACERS’ cruises conduc-
ted in the Cretan Sea by the Institute of Marine Biology in Crete (IMBC).
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 17

of temperatures). Very similar results are presented by Theocharis et al. (1999), who
show that the σθ=29.2 level in the Cretan Sea shoaled from 1988 up to 1992 when
it reached a minimum depth, wherafter it deepened again. At the same time even
denser water accumulated in the deeper layers. According to Tsimplis, Velegrakis,
Theocharis and Collins (1997) and Theocharis et al. (1999), direct measurements in
the straits of the Cretan Arc revealed a strong deep outflow from the Aegean Sea
through Kassos Strait in 1989. In the straits themselves current speeds up to about
0.5 m/s have been observed, but not far away the current flow broadens so that
speeds are more moderate. No further direct current measurements are available until
1994, but geostrophic calculations indicate that there were even stronger flows than
in 1989. Such flows are consistent with those required by the salinity balance con-
siderations. Thus, the flow measurements strongly support the theory that 1989 was
the starting year of the event.
Further insight is provided by a recent detailed analysis of the POEM-BC hydro-
graphic survey of 1991 (Malanotte-Rizzoli et al., 1998). Apparently, at this time the
transient was well underway already, in keeping with the evidence from the outflow
through the Cretan Arc. An increased volume and salinity of the Aegean-derived
waters were obvious, and the σθ=29.18 isopycnal, which in the Eastern Mediter-
ranean proper was previously found in the vicinity of the Adriatic overflow region
only (Malanotte-Rizzoli et al., 1997), was now present over most of the sea. An
intriguing observation of Malanotte-Rizzoli et al. (1998) was that in 1991 an
extended, multi-lobe anticyclone blocked an exchange of waters between the Lev-
antine and Ionian Seas. This blocking anti-cyclone in the southern Levantine was a
three lobe feature comprising the former Iera-Petra and Mersa-Matruh gyre and a
third sub-gyre centered at 29–30°E (Malanotte-Rizzoli et al., 1998). It appears that
its blocking led to salinity increasing in the Levantine upper waters as a result of a
diminution in the intrusion of lower-salinity waters from the Ionian Sea, and to a
diversion of the high-salinity waters into the Aegean. Furthermore the LIW, another
high-salinity water mass which previously had followed a westward pathway to the
south of Crete (Malanotte-Rizzoli et al., 1997), was also diverted into the Aegean
Sea. A blocking anticyclone was also present in 1995 (B. Manca, personal communi-
cation, 1998), but no observations are available from which it can be determined
when the blocking actually began, but this large anticyclone might well have been
a principal player in triggering the event.
These new findings are consistent with the observations in 1995, and particularly
offer an explanation for some of the features seen in Fig. 6. One such feature is that
between 1987 and 1995 salinity also increased in the water above the deep T–S
inversion. This increase, according to the 1991 hydrography is in keeping with the
gradual rise of density in the Cretan Sea, which can be ascribed to the beginnings
of penetration of the saline Cretan Sea overflow into the deep water during the early
stages of the event. This interpretation implies that the salinity increase at mid-depths
to a large part preceded that in the bottom layers. Furthermore, for any given density,
supply to the Levantine Sea presumably preceded that to the Ionian Sea; one cause
of this graded supply is the fact that Kassos Strait, east of Crete, being the deepest
outlet from the Aegean, receives the densest outflow from the Cretan Sea. The other
18 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

cause that is a result of the blocking, is that the overflowing waters received by the
Levantine Sea, were largely retained there. Later on, once the Aegean Sea essentially
had been filled up with the denser waters (in 1992), the preferential supply to the
Levantine Sea presumably ceased. Then Antikithera Strait, west of Crete, became a
major outlet into the Ionian Sea, perhaps even assisted by the blocking. Considering
that jagged T–S diagrams are indicative of recent intrusions of water, the concept
that the supply to the Levantine Sea precedes that into the Ionian Sea serves to
explain why the T–S diagram for the former sea is so much smoother than that for
the latter [Fig. 6(b) and (c)].
A supply of Cretan Sea overflow in the early stages predominantly into the Lev-
antine Sea, and missing the deepest layers, helps to explain an apparent discrepancy
between a start in 1989 as suggested above and the observations of Della Vedova,
Pellis, Camerlenghi and Foucher (1998) that pointed to bottom water temperatures
above the Mediterranean Ridge west of Crete increasing no earlier than about 1992.
The same study indicated how the Mediterranean Ridge acts as a barrier to the deep
overflow, which is in keeping with the deep water characteristics which have
remained almost unchanged in the western Ionian in 1995 [Figs. 8(b) and 5(b)].
Later observations showed that by 1996 the salinity signal had also penetrated the
western Ionian Sea (Fig. 9). It thus appears that the new waters have moved cyc-
lonically around the Ionian Basin, and were confined more or less to the northeastern
part of this basin during the early stages of the transient.
Fig. 8 reproduces sections of the anthropogenic tracer CFC-12 observed in 1987
and 1995 (Roether et al., 1996). It is evident that new waters (i.e. of higher CFC-
12 content) filled the water column in the vicinity of Crete between 1987 and 1995,
giving results similar to that for salinity (Fig. 5). This input of water lifted the pre-
vious deep and bottom waters upwards, and is evident from the mid-depth CFC-12
minimum shoaling by several 100 m between the two years of observation (an aver-
age from approximately 1600 to 1100 m). It was shown above that this displacement
of the minimum-layer is consistent with the overall renewal of the deep layers already
deduced from the salinity balance (Roether et al., 1996). Another feature is that the
vertical extent of the minimum layer was greatly reduced, and that the ‘minimum’
concentrations at least doubled between 1987 and 1995 which provides independent
evidence for new waters having been added at mid-depths. Moreover the vertical
displacement of the minimum layer in the Ionian Sea was larger than in the Levantine
Sea, so the addition of the new waters to the former sea was preferential.
Another effect of the uplifting was in the upper water column where Roether et
al. (1998) observed significantly higher CFC concentration ages in the LIW layer in
1995 than in 1987 (the CFC age represents the time elapsed since the CFC concen-
tration observed at depth was to be found in the mixed layer). This they interpreted
as resulting from CFC-dilution by the upwelling of low-CFC waters into the LIW
from below, and they concluded that the upwelling caused by the event was essen-
tially continuous up to the LIW layer. The dilution effect was particularly strong in
the Ionian, thus supporting the hypothesis that the uplifting occurred preferentially
in this sea.
Oxygen and silica sections for 1987 and 1995 are shown in Figs. 10 and 11. The
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 19

Fig. 8. CFC-12 section along the eastern Mediterranean, 1987 (Meteor M5/5; upper panel) and 1995
(Meteor M31/1; lower panel), for stations see inset maps.

changes between the two years are fully compatible with both Fig. 5 (salinity) and
Fig. 8 (CFC-12). The ‘new’ waters intruding from the Aegean raised oxygen and
lowered silica contents through most of the deeper waters. Note that in the upper
few hundred meters the isolines sloped down from the Levantine into the Ionian in
1987, whereas in 1995 they were approximately level; once again lending further
support to upwelling being preferential in the Ionian. According to Figs. 10 and 11
one implication of the changes is that nutrient concentrations were significantly
increased in the upper water column but reduced in the deep waters, with correspond-
20 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

Fig. 9. Selected salinity profiles in the western Ionian Sea from 1995 (thick lines, 35° 47.58⬘, 17°
27.96⬘E and 35° 47.94⬘N 16° 36.12⬘E) and 1996 (thin line 36° 28.71⬘N, 17° 7.87⬘E).

ing inverse changes in oxygen. The changes in deep silica and oxygen concentrations
were by at least 10%. For the upper waters, Klein et al. (1999) show that the nutric-
line rose to depths of 150 m over much of the Eastern Mediterranean. The implied
extra supply of nutrients close to the euphotic zone might well be a cause of enhanced
productivity. On the other hand, in future the supply of nutrients by upwelling of
deep waters will be somewhat lower, which may possibly have negative effects on
the productivity. The exact magnitude of these effects is hard to predict.
Prior to the event, below the LIW salinity decreased with depth, so that the recent
uplifting of deep waters has tended to reduce salinity at intermediate depths. This
salinity decrease is apparent in the Ionian Sea in the vicinity of the tracer minimum
layer [Fig. 8(b)], whereas elsewhere it is partly or fully masked by the mid-depth
addition of a high-salinity outflow from the Cretan Sea [Figs. 5(b) and 6(a)]. As
Klein et al. (1999) show, one consequence was that by 1995 the salinity input to
the Adriatic had decreased relative to the classical situation. This was concluded
from the observation that the deeper zone of the high-salinity layer off the Strait of
Otranto produced by LIW and CIW in 1987 had been replaced in 1995 by waters
of a lower salinity, but somewhat higher density. A similar change appears to have
occurred in the subsurface outflow through the Sicilian Passage (Klein et al., 1999;
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 21

Fig. 10. Oxygen section along the eastern Mediterranean, 1987 (Meteor M5/5; upper panel) and 1995
(Meteor M31/1; lower panel), for stations see inset map.
22 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

Fig. 11. Silicate section along the eastern Mediterranean, 1987 (Meteor M5/5; upper panel) and 1995
(Meteor M31/1; lower panel), for stations see inset map.
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 23

Astraldi, Gasparini & Sparnocchia, 1998). These changes affect the preconditioning
in, and thus also the performance of, the deep water formation areas of the Adriatic
and the Western Mediterranean. Furthermore, the decrease in mid-depth salinity in
the Ionian indicates that the volume of CIW had greatly diminished, presumably
because the Cretan Sea outflow was being largely or totally diverted to greater depths.
The replacement of as much as 20% of the waters below 1200 m by dense Aegean
overflow after 1987 (Roether et al., 1996) corresponds to a total formation of 7.3
Sv*years (1 Sv=1 Sverdrup=106 m3/s). The probable starting year of the transient,
i.e. 1989, implies that the average formation rate is approximately 1.2 Sv. To get a
feeling for the magnitude of this event, one should note that this rate about equals
that of the main Mediterranean thermohaline cell, and that it amounts to as much
as four times the rate deduced for the Adriatic source in the classical situation
(Roether & Schlitzer, 1991; Lascaratos, 1993).
The event has led to increases in salinity over most of the Eastern Mediterranean,
but the net total gain and the extent of the partial compensation by the decreases in
salinity in the Ionian upper waters are hard to quantify. The reason is that, whereas
for the classical situation up to 1987, data from various years can be combined, but
this is no longer true for the years subsequent to the transient so the available salinity
database is too limited to assess the changes of the past decade. There are two
possible explanations for a net salinity gain. Based on the heat and fresh water flux
climatology for the Eastern Mediterranean, Josey, Pascal and Taylor (1997) found
that, for the period 1988–95, the Aegean was subject to anomalous high cooling and
net evaporation. Theocharis et al. (1999) also present evidence of significant
reduction of precipitation during this period. This may have assisted in the initiation
of the event, but because of the restricted area of this sea the implied salinity changes
are insufficient to account for the net salinity gain observed. Josey et al. (1997)
also reported a changed fresh water forcing of the same sign for the entire Eastern
Mediterranean, but here the magnitude of the change appears to fall short of account-
ing for the extra salt observed in the water column. The other conceivable net sup-
pliers of salt are the exchanges of water with the Western Mediterranean and with
the Black Sea. A quantitative assessment of the salt balance for the event resolving
this complex issue is a subject of ongoing work.
Finally, in view of the delicate competition between the dense waters formed in
the Aegean and Adriatic Seas, a key question is whether this event is unique or
whether it has occurred before. A time series of the mean characteristics inside the
Cretan Sea (Fig. 7) indicates that weaker events have occurred in the past, for
example in the 1970s when density values reached 29.22. However, because of their
small magnitude their effects in the Eastern Mediterranean would probably have
been limited in duration and scope. Fig. 12 shows selected T–S diagrams from the
western and eastern basin between 1910 when the earliest observations were made
(Nielsen, 1912), and 1995. The drift in temperature and salinity is apparent through-
out the time series, although the accuracy and spatial density data in the early work
is limited. The long-term trends in the deep and intermediate waters of both the
western and Eastern Mediterranean have been known for some time (Bethoux, Gen-
tili, Raunet & Tailliez, 1990; Rohling & Bryden, 1992). However, whereas the new
24 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

Fig. 12. Selected T–S diagrams in the Western (left) and Eastern (right) Mediterranean since 1910.
Profiles from 1910 (asterisks) and 1961 are given as dashed lines. The 1987 profiles are represented by
thick dash-dotted lines and the ones from 1995 by thick solid lines.

state is characterized by a T–S inversion marking the interface between the older
waters of Adriatic origin and the recent Aegean overflow, the earlier diagrams up
to 1987 all exhibited steadily decreasing of temperature and salinity with depth. We
interpret this qualitative difference before and after 1987 to be convincing evidence
that this event has been unique as far back as can be traced with observations. Fig.
12 furthermore indicates that, in contrast to the Eastern Mediterranean, the Western
Mediterranean subsurface waters are still experiencing steady increases in salinity
and temperature.

4. A numerical model approach

In this section we present preliminary results of a numerical study that attempts


to simulate the changes in the deep thermohaline circulation described above. It is
an ongoing investigatino that is being carried out in the framework of MAST/MTP-
II MATER project. The strategy we are following is: (i) to set up a realistic model
of the Eastern Mediterranean that can reproduce the old regime of dense water forma-
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 25

tion in the Aegean and the Levantine; (ii) to use atmospheric forcing of recent years
and try to simulate the new regime of dense water formation; and (iii) to simulate
the transition from the one state to the other by carrying out a 16-year long (1979–
94) experiment forced by realistic atmospheric data. The results presented here are
from a 20 km model resolution of the basin; a more detailed model (10 km) is being
currently developed as a follow-up to this study.

4.1. Model setup

The numerical model used in our experiments is based on the Princeton Ocean
Model (POM) developed by Blumberg and Mellor (1987) for coastal and open sea
applications. In the past, the model has been successfully applied to the Mediter-
ranean Sea for studies of the general circulation (Zavatarelli & Mellor, 1995), water
mass formation (Lascaratos & Nittis, 1998) and regional modeling of the Adriatic
and the Aegean Seas. Recently, it has been used for a real time oceanographic
nowcast/forecast system of the Mediterranean (Horton, Clifford, Schmitz & Kan-
tha, 1997).
The model grid covers the whole Eastern Mediterranean excluding the Adriatic
Sea (Fig. 1). The horizontal grid resolution is 20 km, while in the vertical we use
19 sigma levels with logarithmic distribution near the surface. The model is
initialized using the MODB-MED5 climatology (Brasseur, Beckers, Brankart &
Schoenauen, 1996). At each grid point the depth has been computed using bilinear
interpolation of the DBDB5 bathymetric data of the US Navy that have a 1/12 of a
degree resolution. The resulting model topography has been filtered in order to elim-
inate any pressure gradient error associated with sigma layers over steep topography
(Mellor, Ezer & Oey, 1994). We have first used a third order Shapiro (1970) filter
applied twice, but then applied a selective filtering in order to fulfill the dxH/H⬍0.2
criterion in the whole domain.
Open boundary conditions similar to those described in Lascaratos and Nittis
(1998), are used to simulate the exchanges with the Adriatic Sea (to the north) and
the Western Mediterranean (to the west). A different boundary condition with pre-
scribed inflow and outflow transports is used for the Dardanelles Straits (north
Aegean) that control the communication with the Black Sea. The transports follow
a seasonal cycle with maximum values during June but the net transport into the
Aegean sea is always kept constant to 0.01 Sv. The surface boundary conditions
(momentum, heat and freshwater fluxes) use an interactive scheme that allows rep-
resentation of the main air–sea interaction mechanisms including one way (ocean to
atmosphere) feed back (Rosati & Miyakoda, 1988; Lascaratos & Nittis, 1998).

4.2. The atmospheric forcing

The atmospheric data used in our experiments were extracted from the 1979–94
6-hourly ECMWF re-analysis. The parameters used are the two components of sur-
face wind speed Wu, Wv, the air temperature Ta, the dew-point temperature TDP at
surface and the total cloud cover CTCC. Relative humidity is computed from Ta and
26 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

TDP using the formula for saturation vapor pressure. For the period we have calcu-
lated two different ‘perpetual year’ climatologies: climatology A is computed from
years 1979–86 and climatology B from years 1987–94. These two climatologies
represent the atmospheric conditions associated with the ‘old’ and the ‘new’ state
of dense water formation in the Aegean. A comparison between the two periods
indicates that climatology B (new state) introduces higher wind speed for most of
the months. It also introduces higher air temperature for the summer months
(attributed to the warm summers of 1987 and 1988, see also Castellari, Pinardi &
Leaman, 1998) but lower temperature during the winter (attributed to the very cold
winters of 1987, 1992 and 1993). This is an initial indication that local atmospheric
conditions have played a crucial role in the observed changes in the dense water
characteristics. The high frequency (6 h) forcing was used for the third group of
experiments when the variability during the period 1979–94 was simulated.
As already mentioned, the precipitation fields used in our study are not from the
ECMWF reanalysis but from the Jaeger (1976) monthly climatology. This means
that no signal of interannual variability has been introduced by this parameter. There
are, nevertheless, observations that rainfall over Greece during this period was highly
variable. The years 1989–90 marked the driest period recorded during this century
(Tselepidaki, Zarifis & Asimakopoulos, 1992) and the years 1992–93 were just as
dry (Theocharis et al., 1999). There was a ⬎40% reduction in rainfall during these
4 years compared with normal periods. In order to include this significant anomaly
in our forcing we used a 20% average reduction in rainfall over the Aegean for
climatology B. In the experiments with the high frequency forcing (1979–94) we
used the observed reduction of 40% for the years 1989, 1990, 1992 and 1993. This
anomaly was applied for the Aegean Sea and for all the months of the respective
years.
A number of customization experiments using climatology A were carried out for
the selection of the appropriate bulk formulas that must be used for the air–sea
flux calculations. The same procedure has been followed in previous studies in the
Mediterranean but with different data sets (Lascaratos & Nittis, 1998). Our experi-
ments with climatology A showed that the combination of the Kondo scheme for
latent and sensible heat fluxes together with the formula of Bignami for the back
radiation give realistic values of annual mean heat budget (see Castellari et al., 1998
for a detailed description of the method).

4.3. Model results

Our first experiments were carried out for the perpetual year (monthly mean) of
climatology A (1979–86) and B (1987–94). In each case, the model has been inte-
grated for 12 years until a steady state (repeating cycle) has been reached. In Fig.
13(a) and (b) we present maps of the 29.05 isopycnal depth during late February of
year 12 for experiments A and B. The depth of this isopycnal marks the areas of
dense water formation when it reaches the surface. In this case we have outcropping
of the specific isopycnal and ventilation of the corresponding layers i.e. formation
of new intermediate or deep water. The central and south Aegean Sea is clearly
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 27

Fig. 13. Depth of 29.05 isopycnal during late February for experiment A (1979–86 climatology) and B
(1987–94). Formation areas are marked in blue (outcropping of the isopycnal).

marked in both cases as area of dense water formation together with the cyclonic
Rhodes gyre area. Experiment B (new state) gives a far greater area of dense water
formation at the same sites, associated with higher densities because of lower tem-
perature and higher salinity. These are the result of the lower winter air temperatures
28 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

and higher wind speeds of climatology B together with the reduction of rainfall over
the Aegean. The maximum densities of the bottom waters inside the Cretan Sea in
experiment B, reach a maximum value of σθ29.25. They are considerably higher
than those produced by climatology A (maximum σθ 29.18) but still quite lower
than the observed densities that reached the extreme value of σθ 29.4 during 1994
(Theocharis et al., 1999).
The salinity on a selected isopycnal surface is often used as a tracer for the spread-
ing of a new water mass. In Fig. 14 we present maps of salinity on isopycnal 29.20
for a sequence of years from experiment B. At the beginning of the experiment
(initial conditions) the density of 29.20 was confined to a very few deep parts of
the basin and was characterized by low salinity values (⬍38.75) which is typical of
the EMDW of Adriatic origin. In the following years high salinity values appear in
the central and south Aegean deep basins where the newly formed dense water is
stored. As soon as these deep basins have been filled, the dense water starts to
outflow through the straits of the Cretan arc and spreads, towards the Ionian and the

Fig. 14. Experiment B (1987–94 climatology): Salinity on isopycnal 29.20 at different years.
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 29

Levantine Seas. The most significant outflow is through the eastern straits and
especially via the Kassos Strait. Hence, the newly formed water first spreads into
the Levantine Basin and thereafter into the Ionian. Both results on the outflow and
spreading are in agreement with the observational results presented in Section 3. It
should be noted that the dense water stored in the deep basin of the North Aegean
(Athos Basin) has characteristics that are different to those of the Central and south
Aegean. They are colder and less saline (Fig. 14) because of the influence of the
Black Sea waters and they do not appear to contribute to the dense water from the
Aegean Sea.
The next series of experiments was carried out using the high frequency 16 years
of atmospheric forcing (experiment C). In Fig. 15 we present the time series of
atmospheric parameters and hydrological properties in the Aegean Sea as reproduced
by the model. We also present calculated annual rates of CDW formation and trans-
ports of outflows through the straits of the Cretan Arc. The formation rates were
computed following Lascaratos and Nittis (1998) using the 29.20 as the lower limit
of CDW while the outflowing transports were computed for the intermediate and
deep layers (below 250 m). We note that these transports also include the outflows
of intermediate water (CIW) as well as part of the LIW that passes through the
Cretan Sea during its eastwards spreading. The annual mean total heat flux over the
Aegean, undergoes strong variations that exceed 30 W/m2. The maximum heat loss
is observed during 1987, 1992 and 1993. Those years together with 1989–90 are
also characterized by high E⫺P (evaporation minus precipitation) as a result of both
increased evaporation and low precipitation. The response of the Aegean Sea to these
variations is shown by its drop in temperature during the years of excess cooling
and by the continuous increase of salinity from 1988. The CDW formation rates
present high values after 1987 reaching a maximum during 1992, 1993 (above 1.0
Sv) and secondary maxima occurred in 1987, 1989 and 1990 (0.3–0.6 Sv). Increased
outflowing transports are also observed through the intermediate and deep layers of
the Cretan straits after 1987 (1–5–3.2 Sv). Note once again that these transports
do not correspond only to CDW but also include intermediate water masses (CIW
and LIW).
The areas of dense water formation during these years are shown in Fig. 16 where
the depth of isopycnal 29.10 is presented during the usual outcropping period
(February). During the mild years (1981–84–85–86) small quantities of dense water
are formed in the central Aegean Sea. Those waters lose their characteristics very
quickly and fail to give any signal outside the Aegean. During the relatively cold
winters we observe an expansion of the area of dense water formation in the Aegean
Sea. The years of 1987, 1992 and 1993 are characterized by very extensive areas
of dense water formation throughout the Aegean Sea and in the NW Levantine
(Rhodes gyre). The years between these two very cold periods i.e. 1988–89–90–91
are also characterized by increased volume of dense water as a result of the high
E⫺P values and corresponding loss of buoyancy.
The increased dense water formation rates during those years have affected the
density structure in the whole Eastern Mediterranean. The typical depth of the 29.10
isopycnal in the basin was 苲400–500 m in the beginning of the simulated period
30 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

Fig. 15. Time series of atmospheric and hydrological parameters in the Aegean Sea: wind speed (m/s),
air temperature (°C), total heat flux (W/m2), volume integrated temperature (°c) and salinity, CDW annual
formation rate (Sv), and outflowing transports (Sv) at intermediate and deep layers of the Cretan Straits.
Annual means have been used to mask the seasonal cycle. Solid line: experiment C. Dashed line: experi-
ment D (no modification of precipitation).

(1979) but by 1994 it is at 200–300 m. This indicates uplifting of the deep isopycnals
in the basin of more than 250 m. The model, therefore, reproduces the observed
uplifting of isopycnals described in the Section 3, but underestimates its extent.
In order to assess the impact of the rainfall anomaly on the evolution of the event,
we repeated experiment C excluding the precipitation reduction (experiment D). The
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 31

Fig. 16. Depth of isopycnal 29.10 for each February of the 16 years (experiment C) end.

results of this experiment confirm that the increase of salinity inside the Aegean
after 1988 and almost 50% of the CDW formation rate during that period was mainly
attributable to the reduction in precipitation (Fig. 15). In experiment D, the salinity
presents only a minor increasing trend after 1992 in response to the high evaporation
during cold winters 1992–93. Despite the absence of the precipitation anomaly in
experiment D, CDW seems to continue to be formed in the Cretan Sea during 1989–
90 albeit in small amounts (0.1–0.2 Sv). The cold years of 1992–93 produce high
CDW formation rates (0.3–0.8 Sv) but much lower than those reproduced by experi-
ment C (0.8–1.2 Sv). Similar results, i.e. reduced values, are obtained from the calcu-
lation of the outflowing transports through the straits of the Cretan arc.

5. Discussion and conclusions

Quite recently the Eastern Mediterranean thermohaline circulation has altered dra-
matically. Contrary to the situation that has prevailed since hydrographic obser-
vations began, the Adriatic and Aegean Seas began to compete for deep water pro-
32 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

duction. Previously the Adriatic had been the predominant source for the deep waters
of the Eastern Mediterranean. In retrospect the long time this situation has persisted
now seems rather surprising, given the subtle differences in the densities of the deep
outflows from the two seas.
From the observational evidence gathered on this event to date, a coherent,
although still tentative, picture of its evolution has begun to emerge. The Aegean
appears to have started to discharge large amounts of unusually dense waters in
1989. The average outflow rate up to 1995 has been about 1.2 Sv (Roether et al.,
1996), which is more than three times as large as the previous output from the
Adriatic (Schlitzer et al., 1991). As stated in Section 3, the event was initiated by
a strong increase in salinity in the Cretan Sea, which was subsequently followed by
a further decrease in temperature. Although there may be some relation to the steady
drifts in T–S properties that had been observed in the past, the indications are that
the major factors triggering the event were (a) the significant anomalies in air–sea
heat and freshwater budgets of the area that were associated with peculiar atmos-
pheric conditions (Theocharis et al., 1999); and (b) the disruption of normal com-
munication between the Ionian Basin and the Levantine Basin as a result of develop-
ment of a special flow regime in the Cretan Passage (Malanotte-Rizzoli et al., 1998).
The blocking resulted in increasing salinities in the Levantine and in the Aegean.
Various pieces of evidence, for example smoother T–S profiles in the Levantine
relative to the Ionian, suggest that in the early stages of the event outflow from the
Aegean mostly discharged into the Levantine Basin and the Ionian Sea was only
supplied later; nor did the outflow in the initial stages penetrate all the way down
into the bottom waters. We presume that this is why Della Vedova et al. (1998)
observed no increases in bottom temperatures west of Crete prior to 1992.
The enhanced production of deep water has been the cause of a substantial
uplifting of the residing water column by as much as about 500 m, which appears
to have been strongest in the Ionian Basin. One result has been that the subsurface
waters being supplied to the Adriatic (Klein et al., 1999) and to the Aegean have
become lower in salinity; a greatly diminished production of CIW (Klein et al., 1999)
may also have contributed to the lower salinities. These changes also affect the
preconditioning of the water in the two competing deep water formation areas, and
thus represent a potential feedback mechanism which may affect future changes in
the thermohaline structure of the Eastern Mediterranean.
A result of the uplifting of the water column by 500 m has been a rise of the
nutricline over large parts of the Eastern Mediterranean to depths as shallow as 150
m, i.e. far closer to the euphotic zone than previously (Klein et al., 1999). To date
there are no conclusive observations as to whether or not this rise has had any impact
on the ecosystem, but the redistribution of nutrients underlines the fact that the East-
ern Mediterranean is presently in a transient state which is of relevance for many
disciplines working in this sea.
The enhanced deep water production has implied a deposition of salt in the deep
and bottom waters with a simultaneous decrease of salinity higher up. Since the
turnover rate for the waters below 1200 m has been estimated to exceed 100 years
(Roether & Schlitzer, 1991), this extra salt will take many decades to return into the
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 33

upper waters. Its return, however, might well induce changes in the thermohaline
circulation, considering the dependence of the two potential sources of deep water
on the salinity preconditioning. It will therefore take many decades before the Eastern
Mediterranean returns to a new quasi-steady state. An interesting question in this
connection is whether the system will recover its previous mode of operation with
a single source of deep water production in the Adriatic or evolve into an entirely
different, perhaps even un-envisaged direction.
Preliminary results from numerical experiments indicate that the observed changes
can be, at least partially, explained as a response to variability in atmospheric forcing.
The sequence of (i) the cold winter of 1987; (ii) the very ‘dry’ years 1989–90 and
1992–93; and (iii) the very cold winters of 1992–93 have clearly affected the whole
Eastern Mediterranean and especially the Aegean Sea. This sequence of meteorologi-
cal events is in agreement with the evolution of water mass characteristics in the
Cretan Sea (illustrated in Fig. 7) that presents the first evidence of deep water forma-
tion during 1987 and an intensification since 1989. In response to this variability of
atmospheric forcing, the model reproduces increased dense water formation rates in
the Aegean Sea and corresponding outflow rates through the straits of the Cretan
arc. The computed formation rates are somewhat smaller than those estimated from
observations (1.0 Sv for 7 years) and hence the model underestimates the intensity
of the event. Sensitivity experiments without the precipitation anomaly, confirm that
this meteorological event was a significant contributor to the evolution of CDW
formation. In agreement with observational results, the simulated outflow is more
important during the first period at the eastern straits and thus affected the Levantine
Basin first. Finally, the observed uplifting of deep isopycnals in the whole Eastern
Mediterranean is also reproduced by the numerical simulations, although the com-
puted rise (苲250 m) is smaller than the observed change (苲500 m).
The change that has occurred in the Eastern Mediterranean represents a model
example of the potential instability of the oceanic circulation in general that has
come into focus recently. Dedicated observations to monitor the future evolution of
the transient are therefore needed. These have to be supplemented by modeling
efforts directed at producing, on the basis of observations available or yet to be
carried out, a coherent and quantitative assessment of the temporal evolution of the
event since its onset.

Acknowledgements

The work has been supported by the European Union-DG XII in the framework
of the MAST program, under contacts MAS3-CT96-0051 and MAS3-CT95-0043.

References
Astraldi, M., Gasparini, G. P., & Sparnocchia, S. (1998). Water masses and seasonal hydrographic con-
ditions in the Sardinia–Sicily–Tunisia region. Rapport du Commission International de la Mer Medit-
erranee, 35, 122–123.
34 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

Bethoux, J. P., Gentili, B., Raunet, J., & Tailliez, D. (1990). Warming trend in the Western Mediterranean
deep water. Nature (London), 347, 660–662.
Blumberg, A. F., & Mellor, G. L. (1987). A description of a three-dimensional coastal ocean circulation
model (Vol. 4). In N. S. Heaps, Coastal estuarine science (pp. 1–16). Washington, DC: AGU.
Brasseur, P., Beckers, J. M., Brankart, J. M., & Schoenauen, R. (1996). Seasonal temperature and salinity
in the Mediterranean Sea: climatological analysis of a historical data set. Deep-Sea Research, 43,
159–192.
Castellari, S., Pinardi, N., & Leaman, K. (1998). A model study of air sea interactions in the Mediterranean
Sea. Journal of Marine Systems, 18, 89–114.
Della Vedova, B., Pellis, G., Camerlenghi, A., & Foucher, J.-P. (1998). Recent warming of deep sea
sediments in the Eastern Mediterranean. Rapport du Commission International de la Mer Mediter-
ranee, 35, 66–67.
Gacic, M., Kovacevic, V., Manca, B., Papageorgiou, E., Poulain, P. M., Scarazzato, P., & Vetrano, A.
(1996). Thermohaline properties and circulation in the Strait of Otranto. In F. Briand, Dynamic of
Mediterranean straits and channels (pp. 117–145). CIESM Science Series No. 2.
Gaillard, F., Desaubies, Y., Send, U., & Schott, F. (1997). A four dimensional analysis of the thermal
structure in the Gulf of Lion. Journal Geophysical Research, 102, 13163–13180.
Gascard, J. C. (1973). Vertical motions in a region of deep water formation. Deep-Sea Research, 20,
1011–1027.
Gascard, J. C. (1978). Mediterranean deep water formation, baroclinic instability and oceanic eddies.
Oceanologica Acta, 1, 315–330.
Horton, C., Clifford, M., Schmitz, J., & Kantha, L. H. (1997). A real time oceanographic nowcast/forecast
system for the Mediterranean Sea. Journal of Geophysical Research, 102 (25), 125–156.
Jaeger, L. (1976). Monatskarten des Niedersschlags für die ganze Erde. Berichte des Deutschen wetterdi-
enstes, 18 (139), 38 pp.
Jones, H., & Marshall, J. (1993). Convection with rotation in a neutral ocean: a study of open ocean deep
convection. Journal of Physical Oceanography, 23, 1009–1039.
Josey, S., Pascal, R., & Taylor, P. (1997). Air–sea interaction in the Mediterranean Sea. In Abstract
volume: International Conference: Progress in Oceanography of the Mediterranean Sea, Rome, Nov.
17–19, pp. 253–254.
Killworth, P. D. (1983). Deep convection in the world ocean. Review in Geophysics and Space Physics,
21, 1–26.
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 Research, I 46, 371–414.
Lacombe, H., & Tchernia, P. (1972). Characters hydrologiques et circulation des eaux en Mediterranee.
In D. J. Stanley, The Mediterranean Sea: a natural sedimentation laboratory. (pp. 25–36) Stroudsburg,
PA: Dowden, Hutchinson and Ross.
Lacombe, H., Tchernia, P., & Benoist, G. (1958). Contribution a l’étude hydrologique de la mer Egée
en period d’été. Bulletin d’Information COEC, 10, 454–468.
Lascaratos, A. (1993). Estimation of deep and intermediate water mass formation rates in the Mediter-
ranean Sea. Deep-Sea Research II, 40, 1327–1332.
Lascaratos, A., & Nittis, K. (1998). A high resolution 3-D numerical study of intermediate water formation
in the Levantine Sea. Journal of Geophysical Research, 103, 18497–18511.
Lascaratos, A., Williams, R., & Tragou, E. (1993). A mixed layer study of the formation of Levantine
intermediate water. Journal of Geophysical Research, 98, 739–749.
Madec, G., Chartier, M., Delecluse, P., & Crepon, M. (1991). A three dimentional numerical study of
deep water formation in the northwestern Mediterranean Sea. Journal of Physical Oceanography, 21, 9.
Malanotte-Rizzoli, P., & Hecht, A. (1988). Large scale properties of the Eastern Mediterranean: a review.
Oceanologica Acta, 11, 323–335.
Malanotte-Rizzoli, P., Manca, B. B., Ribera DAcala, M., Theocharis, A., Bergamasco, A., Bregant, D.,
Budillon, G., Civitarese, G., Georgopoulos, D., Michelato, A., Sansone, E., Scarazzato, P., & Sou-
verrmezoglou, E. (1997). A synthesis of the Ionian hydrography, circulation and water mass pathways
during POEM—Phase I. Progress in Oceanography, 39, 153–204.
Malanotte-Rizzoli, P., Manca, B. B., Ribera DAcala, M., & Theocharis, A. (1998). The Eastern Mediter-
A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36 35

ranean in the 80s and in the 90s: the big transition emerged from the POEM-BC observational evi-
dence. Rapport du Commission International de la Mer Mediterranee, 35, 174–175.
MEDOC Group (1970). Observation of formation of deep water in the Mediterranean Sea. Nature
(London), 227, 1037–1040.
Mellor, G. L., Ezer, T., & Oey, L. Y. (1994). The pressure gradient conundrum of sigma coordinate
ocean models. Journal Atmospheric Oceanic Technology, 11, 1126–1134.
Miller, A. R. (1963). Physical oceanography of the Mediterranean Sea: a discourse. Rappport du Com-
mission International de la Mer Mediterranee, 17, 857–871.
Morcos, S. A. (1972). Sources of Mediterranean intermediate water in the Levantine Sea. In A. L. Gordon,
Studies in physical oceanography (Vol. 2) (pp. 185–206). New York: Gordon and Breach.
Nielsen, J. N. (1912). Hydrography of the Mediterranean and adjacent waters. In Report of the Danish
Oceanographic Expedition 1908–1910 to the Mediterranean and Adjacent Waters 1, 72–191, Copen-
hagen.
Nittis, K., & Lascaratos, A. (1998). Diagnostic and prognostic numerical studies of LIW formation. Jour-
nal of Marine Systems, 18, 179–195.
Ovchinnikov, I. M. (1984). The formation of Intermediate Water in the Mediterranean. Oceanology, 24,
168–173.
Ozsoy, E., Hecht, A., & Unluata, U. (1989). Circulation and hydrology of the Levantine Basin. Results
of POEM coordinated experiments 1985–1986. Progress in Oceanography, 22, 125–170.
Phillips, O. M. (1966). On turbulent convection currents and the circulation of the Red Sea. Deep-Sea
Research, 13, 1149–1160.
POEM Group (1992). General circulation of the Eastern Mediterranean. Earth Sciences Review, 32,
285–308.
Pollak, M. I. (1951). The sources of deep water of the Eastern Mediterranean Sea. Journal of Marine
Research, 10, 128–152.
Roether, W., & Schlitzer, R. (1991). Eastern Mediterranean deep water renewal on the basis of chlorofluo-
romethans and tritium. Dynamics of Atmospheres and Oceans, 15, 333–354.
Roether, W., Manca, 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. Journal of Marine Systems, 18,
71–87.
Roether, W., Beitzel, V., Sueltenfuss, J., & Putzka, A. (1999). The Eastern Mediterranean tritium distri-
bution in 1987. Journal of Marine Systems, 20, 49–61.
Rohling, E., & Bryden, H. (1992). Man induced salinity and temperature increases in the Western Mediter-
ranean deep water. Journal of Geophysical Research, 97, 11191–11198.
Rosati, A., & Miyakoda, K. (1988). A general circulation model for upper ocean simulation. Journal of
Physical Oceanography, 18, 1601–1626.
Schlitzer, R., Roether, W., Oster, H., Junghans, H., Hausmann, M., Johannsen, H., & Michelato, A. (1991).
Chlorofluoromenthane and oxygen in the Eastern Mediterranean. Deep-Sea Research, 38, 1531.
Schott, F., & Leaman, K. D. (1991). Observations with moored acoustic Doppler current profilers in the
convection regime in the Golfe du Lion. Journal of Physical Oceanography, 21, 558–574.
Shapiro, R. (1970). Smoothing, filtering and boundary effects. Review of Geophysics and Space Physics,
8, 359–387.
Sur, H. I., Ozsoy, E., & Unluata, U. (1992). Simultaneous deep and intermediate convection in the northern
Levantine Sea, winter 1992. Oceanologica Acta, 16, 33–43.
Theocharis, A., Nittis, K., Kontoyannis, H., Papageorgiou, E., & Balopoulos, E. (1999). Climatic changes
in the deep waters of the Aegean sea and their influence in the deep thermohaline circulation of the
Eastern Mediterranean (1986–1997). Geophysical Research Letters, 20, 1617–1620.
Tselepidaki, I., Zarifis, B., & Asimakopoulos, D. N. (1992). Low precipitation over Greece during 1989–
1990. Theoretical and Applied Climatology, 46, 115–121.
Tsimplis, M. N., Velegrakis, A. F., Theocharis, A., & Collins, M. B. (1997). Low-frequency current
variability at the Straits of Crete, Eastern Mediterranean. Journal of Geophysical Research, 102,
25005–25020.
36 A. Lascaratos et al. / Progress in Oceanography 44 (1999) 5–36

Tziperman, E., & Speer, K. (1994). A study of water mass transformation in the Mediterranean sea:
analysis of climatological data and a simple 3-box model. Dynamics of Atmospheres and Oceans, 21,
53–82.
Wu, P., & Haines, K. (1996). Modeling the dispersal of Levantine intermediate water and its role in
Mediterranean deep water formation. Journal of Geophysical Research, 101, 6591–6607.
Wüst, G. (1961). On the vertical circulation of the Mediterranean Sea. Journal of Geophysical Research,
66, 3261–3271.
Zavatarelli, M., & Mellor, G. L. (1995). A numerical study of the Mediterranean Sea circulation. Journal
of Physical Oceanography, 25, 1384–1414.

You might also like