Recent Changes in the Mediterranean Water Cycle: A Pathwaytoward Long-Term Regional Hydroclimatic Change?
ANNARITA MARIOTTI
Earth System Science Interdisciplinary Center, University of Maryland, College Park,
College Park, Maryland, and ENEA, Rome, Italy
(Manuscript received 19 May 2009, in final form 9 October 2009)
ABSTRACT
An observational analysis of Mediterranean Sea water cycle variability based on recently available datasets
provides new insights on the long-term changes that affected the region since the 1960s. Results indicate an
overall increase in evaporation during 1958–2006, with a decrease up until the mid-1970s and an increase
thereafter. Precipitation variability is characterized by substantial interdecadal variations and a negative long-
term trend. Evaporation increase, primarily driven by SST variability, together with precipitation decrease
resulted in a substantial increase in the loss of freshwater from the Mediterranean Sea toward the overlying
atmosphere. An increase in the freshwater deficit is consistent with observed Mediterranean Sea salinity
tendencies and has broad implications for the Mediterranean water cycle and connected systems.
These observational results are in qualitative agreement with simulated Mediterranean Sea water cycle be-
havior from a large ensemble of models from the Coupled Model Intercomparison Project Phase 3 (CMIP3).
However, simulated anomalies are about one order of magnitude smaller than those observed. This in-
consistency and the large uncertainties associated with the observational rates of change highlight the need for
more research to better characterize and understand Mediterranean water cycle variations in recent decades,
and to better simulate the crucial underlying processes in global models.
1. Introduction
The Fourth Assessment Report (AR4) of the Intergov-
ernmental Panel on Climate Change (IPCC; Solomon
et al. 2007) and related investigations project major
changes in the Mediterranean region, in particular, as
a ‘‘hot spot’’ in hydrological change with significant
impacts on both mean precipitation and variability
(Gibelin and Deque 2003; Giorgi 2006; Giorgi and
Lionello 2008; Sheffield and Wood 2008; Ulbrich et al.
2006). In a recent study, Mariotti et al. (2008) show how
in the Coupled Model Intercomparison Project Phase 3
(CMIP3) multimodel simulations, the combination of
projected twenty-first-century precipitation reduction
and warming-enhanced evaporation concurs to deter-
mine even greater alterations of Mediterranean water
cycle characteristics with significant increases in land sur-
face dryness and loss of freshwater over the Mediterra-
nean Sea. In these simulations, the ‘‘transition’’ toward
such conditions is seen to be ongoing, with an acceler-
ation around the turn of the century toward the larger
rates projected for the twenty-first century. This study
also finds consistency between these tendencies and ob-
servational evidence of twentieth-century-long negative
trends in regionally averaged precipitation, the Palmer
Drought Severity Index, and discharge from numerous
Mediterranean rivers. However, recent long-term changes
in sea evaporation (the biggest single component of the
Mediterranean water cycle) and sea surface freshwater
fluxes (potentially a combination of precipitation and
evaporation changes), as suggested by the CMIP3 sim-
ulations, have yet to be explored. In view of the semi-
enclosed nature of the Mediterranean–Black Sea system,
connected to the Atlantic Ocean via the Strait of
Gibraltar, and the semiarid and/or arid conditions in
land regions downstream of Mediterranean moisture
fluxes, the impacts of such changes could be substantial.
Increases in sea evaporation (and associated latent heat)
and freshwater loss effect the salt, water, and energy
budgets with potentially important implications for Med-
iterranean Sea salinity (note that Mediterranean Sea
Corresponding author address: Dr. Annarita Mariotti, 5825
University Research Court, Earth System Science Interdisciplinary
Center, University of Maryland, College Park, College Park, MD
20740-3823.
E-mail: [email protected]
15 MARCH 2010 M A R I O T T I 1513
DOI: 10.1175/2009JCLI3251.1
� 2010 American Meteorological Society
salinity is among the highest globally), circulation, and sea
level (e.g., Skliris et al. 2007; Tsimplis et al. 2008), and
Atlantic circulation via changes in Strait of Gibraltar
water fluxes (Lozier and Stewart 2008; Millot et al.
2006; Potter and Lozier 2004; Reid 1979). Additionally,
an increase in evaporation (i.e., the amount of moisture
the Mediterranean Sea injects into the overlying atmo-
sphere) would enhance moisture fluxes to downstream
regions, potentially affecting precipitation there [e.g., in
the Sahel (Jung et al. 2006)]. In spite of the high CMIP3
intermodel consistency regarding the Mediterranean re-
gion, substantiating the tendencies suggested by the
models with observational evidence is crucial in view
of the significant limitations in state-of the-art global
models’ representation of the Mediterranean Sea (e.g.,
Marcos and Tsimplis 2008).
A number of previous studies have indicated a long-
term increase in Western Mediterranean Deep Water
salinity and temperatures during the latter half of the
twentieth century (e.g., Bethoux et al. 1998; Krahmann
1998; Rixen et al. 2005). Several of these studies have
evidenced the linkage between this salinity increase and
the long-term decrease in Mediterranean precipitation
during the period from the mid-1970s to the early 1990s,
primarily in connection with the decadal variations of
the North Atlantic Oscillation (Hurrell 1995; Mariotti
et al. 2002). Long-term salinity increase has also been
connected to a reduction in river discharge (e.g., dam-
ming of the Nile River in the 1960s) and Black Sea
freshwater inputs (Rohling and Bryden 1992; Skliris
et al. 2007). Interannual evaporation anomalies and as-
sociated cooling have been identified as a key factor in
the eastern Mediterranean transient event of 1991–93
(Josey 2003; Roether et al. 2007). However, decadal evap-
oration changes over the Mediterranean Sea are still vir-
tually unknown, because previous attempts were limited
by data availability (e.g., Krahmann 1998; Mariotti et al.
2002). In fact, both oceanic precipitation and evaporation
estimates and their interdecadal variability have long re-
mained elusive in the absence of suitable climatic datasets.
Substantial data developments in the last decade, with
some datasets now going back 25 yr or more, have brought
major new opportunities to investigate long-term water
cycle variability in oceanic regions. For instance, Adler
et al. (2008) use Global Precipitation Climatology Pro-
ject data (GPCP; Adler et al. 2003) to investigate the
SST–precipitation relationship on interannual-to-decadal
time scales. Yu (2007) and Yu and Weller (2007) explore
global long-term oceanic evaporation changes in the ob-
jectively analyzed air–sea fluxes (OAFlux) dataset and
find widespread increases in many regions. Wentz et al.
(2007) investigate overall changes in the global water
cycle in relation to global warming.
Research presented here explores the regional mani-
festations of these long-term global water cycle changes
in the sensitive Mediterranean system, exploiting recent
progress in data availability. The focus is on the com-
bined effects of precipitation and evaporation changes
on the Mediterranean water cycle. A major question
that this work addresses is whether the behavior ob-
served during the last few decades is consistent with the
‘‘transition’’ phase suggested by the CMIP3 simulations
for the Mediterranean as a pathway toward future pro-
jected changes. The attribution of observed changes and
the analysis of related impacts is beyond the scope of this
paper. The paper is organized as follows: Section 2 pres-
ents the data used in the analyses and the basic method-
ologies. Section 3 depicts mean characteristics during
recent decades, defining a baseline for change. Follow-
ing sections describe long-term changes since 1960 and
their consistency with the CMIP3 results for evaporation
(section 4), precipitation (section 5), and freshwater flux
(section 6). Finally, summary and concluding remarks
are presented in section 7.
2. Data and methodology
A variety of different data sources are used in this study
taking advantage of recently available datasets combin-
ing in situ, satellite, and reanalyses data for global water
cycle climatic studies. We use monthly GPCP oceanic
precipitation estimates based on blended gauge–satellite
products. The data, available since 1979 on a 2.58 3 2.58
grid, are combined using microwave-based estimates
from the Special Sensor Microwave Imager (SSM/I), in-
frared rainfall estimates, and surface rain gauges. Among
the satellite-based datasets, we use monthly precipitation
estimates and other air–sea fluxes derived by Remote
Sensing Systems (REMSS) using a variety of passive
microwave radiometers, including SSM/I retrievals
[available 1987–2006, at 2.58 3 2.58 resolution (Wentz
et al. 2007)]. To extend the analysis of Mediterranean
Sea precipitation variability back in time, we analyze a
recent reconstruction of oceanic precipitation (Smith
et al. 2008), which aims at capturing the large-scale fea-
tures of global precipitation since 1900 (REOFS here-
after). The data are based on a covariance analysis of
satellite microwave estimates (SSM/I) merged with at-
mospheric reanalyses for the period of 1992–2001. Re-
sults from this analysis are used jointly with historical
land gauge precipitation to reconstruct oceanic precipi-
tation back to the presatellite era. REOFS data are
available on a 58 3 58 grid for the period of 1900–2006.
Because the Mediterranean Sea is semienclosed by land,
land gauge precipitation estimates from surrounding re-
gions are also analyzed for comparison. We analyze the
1514 J O U R N A L O F C L I M A T E VOLUME 23
monthly precipitation dataset from the National Climatic
Data Center (NCDC) Global Historical Climatology
Network (GHCN) version 2 data (Vose et al. 1992; 1900–
2007, 58 3 58 resolution), the Climatic Research Unit
(CRU) TS2.1 data (Mitchell and Jones 2005; 1901–2002,
0.58 3 0.58 resolution), and the Climate Prediction
Center (CPC) Precipitation Reconstruction over Land
(PRECL) data (Chen et al. 2002; since 1948, 2.58 3 2.58
resolution).
Evaporation is derived from the novel air–sea fluxes
dataset OAFlux, which objectively synthesizes surface
meteorology obtained from satellite products [including
retrievals from SSM/I, Quick Scatterometer (QuikSCAT),
Advanced Very High Resolution Radiometer (AVHRR),
and Tropical Rainfall Measuring Mission (TRMM)]
and model reanalyses (Yu et al. 2008). SST in OAFlux
is originally from the Reynolds et al. (2007) product.
OAFlux data are available monthly since 1958 on a 18 3 18
grid. The OAFlux dataset does not use in situ data as
a direct input. Hence, in situ evaporation estimates de-
rived from the National Oceanography Centre South-
ampton (NOCS) new air–sea flux dataset, version 2
[NOCS Flux Dataset v2.0; hereafter NOCS (Berry and
Kent 2008)] is analyzed here as an independent data
source. These data use the International Comprehensive
Ocean–Atmosphere Data Set (ICOADS) to derive
monthly oceanic air–sea fluxes (Worley et al. 2005; 1973–
2006, 18 3 18 grid). For both OAFlux and NOCS, we
compute sea surface humidity based on the Coupled
Ocean–Atmosphere Response Experiment (COARE
3.0) bulk formula (Fairall et al. 2003) using SST and
the Hadley Centre’s mean sea level pressure (HadSLP;
Allan and Ansell 2006) data. Among the satellite-based
air–sea fluxes datasets, we analyze the Goddard Satellite-
Based Surface Turbulent Fluxes version 2.0 (GSSTF 2.0)
data, which use SSM/I and National Centers for Envi-
ronmental Prediction–National Center for Atmospheric
Research (NCEP–NCAR) reanalyses (hereafter NCEP;
Chou et al. 2003; 1987–2000, 18 3 18 resolution), and the
Hamburg Ocean Atmosphere Parameters and Fluxes
from Satellite Data, version 3 (HOAPS-3; Andersson
et al. 2007), which uses SSM/I data and National Oceanic
and Atmospheric Administration’s (NOAA’s) Pathfinder
SSTs (1987–2005, 0.58 3 0.58 resolution). Meteorological
reanalyses are used only to estimate climatologies, be-
cause these data are less well suited to study long-term
variability and trends of regional water cycle charac-
teristics. NCEP (Kalnay et al. 1996) monthly data are
available on a spectral grid at about 1.98 resolution; 40-yr
European Centre for Medium-Range Weather Fore-
casts (ECMWF) Re-Analysis (ERA-40; Uppala et al.
2005) covers the period of 1957–2002, with data avail-
able at a 2.58 3 2.58 resolution. The twentieth-century
CMIP3 model simulations analyzed in this study are
coupled runs with various observed forcings (climate of
the twentieth-century experiment; details available on-
line at http://www-pcmdi.llnl.gov/ipcc/standard_output.
html#Experiments). For the twenty-first century, we ana-
lyze CMIP3 projections for the Special Report on Emis-
sions Scenarios (SRES) A1B. Ensemble means include
data from 14 different models and multiple runs from
each model [refer to Mariotti et al. (2008) for more in-
formation regarding these simulations and their valida-
tion over the Mediterranean region].
In view of data availability, we compute climatologies
over the common period of 1988–2000, although the re-
sults presented in sections 4–6 show that this was a period
of substantial regional water cycle changes. Our approach
is to use data intercomparison to evaluate the uncer-
tainties associated with the various Mediterranean wa-
ter cycle estimates (for individual dataset evaluation, see
the above-mentioned references). In addition to data
quality, resolution disparities also contribute to differ-
ences among individual estimates. Climatological esti-
mates are computed by averaging all available data
sources. Uncertainties are evaluated by computing the
standard error representing the deviation of various
estimates from the datasets’ mean. Area averages are
computed by using individual datasets’ land–sea masks
for the Mediterranean Sea and the surrounding land in
the domain 288–478N, 108W–408E. The disparities in
land–sea mask specifications also add to differences
between the estimates (observational evaporation is
provided as sea only; a fractional cover land–sea mask is
used for CMIP3 model output and for observational
precipitation, if available). Eastern Mediterranean means
are area averages over the sea in the 308–408N, 138–408E
domain; western Mediterranean sea-only means are in
the 328–448N, 58W–128E domain. Anomalies are relative
to the monthly 1988–2000 climatology. Six-year running
means are used to depict low-frequency variations (on
decadal or longer time scales). Although the focus is on
these variations, annual mean values are displayed for
selected datasets as an indicator of interannual variabil-
ity. Linear trends of annual mean values are computed in
order to quantify long-term changes; this does not imply
that the original signal is best represented by a linear
increase in time.
3. Mean characteristics
Annual mean precipitation over the Mediterranean
Sea for the period 1988–2000 is 1.12 mm day21, based
on the mean of various available datasets (Table 1).
Mean evaporation for the same period is 2.94 mm day21,
15 MARCH 2010 M A R I O T T I 1515
that is, well over twice the precipitation rate. These lead
to an estimated annual mean loss of freshwater by the
Mediterranean Sea (E 2 P) of 1.71 mm day21. Uncer-
tainties on these mean values based on the standard error
are 18% and 13% of climatology for precipitation and
evaporation, respectively. Evaporation has a sharp min-
imum in May (;1.5 mm day21); after that there is a
steady increase up until the November–December max-
imum (;4 mm day21), with evaporation remaining high
through February (over 3 mm day21) and then rapidly
decreasing into spring (Fig. 1). Among the various data-
sets, NOCS gives the lowest evaporation rates; GSSTF2.0
and NCEP are consistently above the mean. In contrast
to evaporation, precipitation is low in June–August (with
a minimum in July, ;0.3 mm day21), and increases rap-
idly during the autumn season. Precipitation is about
2 mm day21 in November–December and then starts
to gradually decline. GPCP gives the highest precipita-
tion, while ERA-40 and REMSS are among the lowest.
Overall precipitation has a smaller seasonal cycle com-
pared to evaporation (1.7 mm day21 vis-a-vis 2.5 mm
day21), and E 2 P is positive year-round, minimum in
late spring (;0.7 mm day21), and peaks in late summer
(;2.5 mm day21) when precipitation is still low, while
evaporation has already significantly increased from its
May minimum. All datasets show a roughly similar be-
havior, however monthly estimates vary significantly
and uncertainties are high. Biases tend to follow those of
evaporation, with E 2 P estimates using NCEP and
GSSTF2.0 generally higher than the mean and those
using NOCS being the lowest. Annual mean, E 2 P
fluxes are highest in the southeastern Mediterranean (to
2–3 mm day21, according to GPCP and OAFlux; not
shown) and lowest in the Adriatic and Ligurian Seas (less
than 0.5 mm day21). The E 2 P rates over the eastern
Mediterranean are roughly double those of the western
Mediterranean (see Table 1). This is due to differences
in E rates, because P rates are similar over the two
regions.
4. Long-term mean evaporation changes
We now explore long-term Mediterranean Sea evap-
oration changes relative to the 1988–2000 mean values
for the period since 1958 (Fig. 2). For the period of 1996–
2005, a variety of evaporation data sources are available.
All datasets unanimously show an increase in evapora-
tion leading up to the current period. However, the
increase over this period varies significantly among
datasets, with OAFlux giving the highest increase and
REMSS the lowest. Only the OAFlux and NOCS
datasets are available going back to the 1970s. Accord-
ing to OAFlux, the evaporation increase discussed above
is part of a long-term progressive evaporation increase,
which started in the mid-1970s, when E was about
0.1 mm day21 below the 1988–2000 mean. Similarly to
OAFlux, NOCS also shows mid-1970s evaporation rates
TABLE 1. Mean precipitation (P), evaporation (E), and E 2 P
over the Mediterranean Sea (Med) and major subbasins (eastern
Mediterranean: EMed; western Mediterranean: WMed). Reported
in each cell, annual mean averages for the period of 1988–2000
computed as the mean of available data sources (for precipitation:
GPCP, REMSS, NCEP, ERA-40; for evaporation: GSSTF,
HOAPS-3, OAFlux, NOCS, REMSS, NCEP, ERA-40) and the std
error of various estimates from the datasets’ mean. Units: mm day21.
P E E 2 P
Med 1.12 6 0.21 2.94 6 0.37 1.71 6 0.38
EMed 1.04 6 0.22 3.25 6 0.42 2.10 6 0.42
WMed 1.08 6 0.21 2.51 6 0.37 1.30 6 0.38
FIG. 1. Seasonal cycle of Mediterranean Sea mean (top) evapo-
ration (E), (middle) precipitation (P), and (bottom) E 2 P for the
period of 1988–2000. Individual data sources (various colors; see
legend) and their average (thick solid black line) are reported to-
gether with overall annual averages (dashed black horizontal
lines). Units: mm day21.
1516 J O U R N A L O F C L I M A T E VOLUME 23
lower than the 1988–2000 mean that progressively in-
crease. However, during the 1980s, before the common
evaporation increase of the 1990s, the two datasets dis-
agree: while OAFlux evaporation remains roughly level,
NOCS 6-yr means show a local maximum in 1985. The
two datasets give an overall trend for the period of 1979–
2006 of 0.1–0.2 mm day21 decade21 (4%–8% of cli-
matology per decade; these trends are significant and
statistically consistent, see Table 2). Only OAFlux is
available since the 1960s. According to this dataset,
evaporation rates in the 1960s were comparable to those
of the late 1990s, and subsequently decreased up until
the mid-1970s before the increase leading up to the
current period. Considering the entire 1958–2006 pe-
riod, these interdecadal variations have amounted to
an equivalent linear mean evaporation increase of 2%
decade21 (or 0.06 mm day21 decade21). Evaporation
that is separately area averaged over the eastern and
western Mediterranean gives similar results to those
described above for the entire Mediterranean (not shown).
CMIP3 models’ behavior since the 1960s is broadly similar
to that of OAFlux: an initial steep decrease (1965–1975)
and then a progressive increase since the mid-1970s up
until present times (CMIP3 leads OAFlux by about
5 yr). However, CMIP3 rates are about one order of
magnitude smaller than those of OAFlux.
The humidity gradient at the air–sea interface (Qs 2
Qa) and surface winds are primary factors controlling
sea evaporation. SST influences evaporation by directly
affecting humidity at the sea surface (Qs) and may also
impact the large-scale structure of air humidity (Qa).
The analysis of long-term Mediterranean SST anomalies
from OAFlux and NOCS consistently shows that the
mid-1970s–present evaporation increase described above
was associated with a progressive SST increase during this
period (see Fig. 3 and Table 3). Rates of increase for the
1979–2006 period are in the range of 0.2–0.3 K decade21
(various estimates are statistically consistent). The SST in-
crease over the 1958–2006 period is about 0.1 K decade21.
Both OAFlux and NOCS datasets show an increase in
Qs and Qa since the mid-1970s (not shown). An increase
in Qa is consistent with the increased moisture holding
capacity of the atmosphere at higher temperatures (fol-
lowing the Clausius–Clapeyron equation). During the
period of 1996–2005, both NOCS and OAFlux show an
increase in humidity gradient and evaporation. During
the 1980s, Qs 2 Qa differences between datasets reflect
those in evaporation and are in part associated with SST
differences. From 1960 to mid-1970, OAFlux shows a
steep decrease in Qs 2 Qa, consistently with decreasing
SST, which resulted in decreased evaporation during
this period. An increase in wind speed since the 1990s,
seen in both OAFlux and NOCS, also contributed to
increased evaporation during this period. Before that,
the two datasets show inconsistent behavior, so results
regarding the long-term wind change contribution are
inconclusive.
Figure 4 illustrates how the increase in evaporation
between the decades of 1979–88 and 1996–2005 (by about
TABLE 2. Linear trends of (top to bottom) annual mean evapo-
ration (E), precipitation (P), and E 2 P for the periods of 1958–
2006 and 1979–2006 (mm day21 decade21) using various data
sources. Statistically significant results are in bold.
E
1958–2006 1979–2006
OAFlux 0.063 6 20.039 0.235 6 0.073
NOCS 0.107 6 0.058
CMIP3 0.003 6 20.004 0.011 6 0.007
P
1958–2006 1970–2006
GPCP 20.046 6 0.084
REOFS 20.041 6 0.032 0.007 6 0.078
CRU 20.031 6 0.023 20.021 6 0.071
PRECL 20.036 6 0.018 20.033 6 0.044
GHCN 20.018 6 0.025 0.006 6 0.052
CMIP3 20.011 6 0.006 20.009 6 0.016
E 2 P
1958–2006 1979–2006
OAFlux/GPCP 0.276 6 0.077
OAFlux/REOFS 0.104 6 0.046 0.228 6 0.097NOCS/GPCP 0.148 6 0.090
NOCS/REOFS 0.100 6 0.095
CMIP3 0.014 6 0.006 0.019 6 0.014
FIG. 2. Decadal variations in Mediterranean Sea mean evapo-
ration over the period of 1958–2007. Shown are 6-yr running means
of Mediterranean Sea area-averaged evaporation anomalies rela-
tive to the period of 1988–2000 (lines, mm day21). Various ob-
servational sources are used (see legend; left-hand scale). For
OAFlux, annual mean values are also represented (symbols).
CMIP3 models ensemble running mean averages are also displayed
(note different scale at right).
15 MARCH 2010 M A R I O T T I 1517
0.4 mm day21 according to OAFlux) affected the mean
seasonal cycle. Evaporation increased year-round, with
most of the increase occurring in September–December.
NOCS evaporation (not shown) gives similar results,
although the anomalies are about half those of OAFlux.
Considering the pattern of evaporation change during
1979–2006 (Fig. 5), OAFlux displays increases every-
where over the Mediterranean Sea, and also extending to
the Black Sea and the neighboring Atlantic Ocean waters
[see Yu and Weller (2007) for global evaporation trends].
Annual mean increases are greatest in the Ligurian Sea,
the Adriatic, and parts of the southeastern Mediter-
ranean (up to 0.4–0.5 mm day21 decade21). Other parts
of the Mediterranean show evaporation increases of at
least 0.1–0.2 mm day21 decade21. The October–March
pattern of increase is similar to that of the annual means,
except that regions with increases of 0.4–0.5 mm day21
decade21 are widespread and most parts of the Medi-
terranean see an increase of at least 0.2–0.3 mm day21
decade21. NOCS evaporation trends (not shown) give
similar patterns of change as those described above,
except the values are lower (maximum annual trends are
up to 0.3–0.4 mm day21 decade21, with trends gen-
erally in the range of 0.1–0.2 mm day21 decade21).
The Mediterranean Sea evaporation trends described
FIG. 3. Factors controlling decadal Mediterranean Sea evapo-
ration changes. (top) Mediterranean-averaged SST (K), (middle)
specific humidity gradient (g g21), and (bottom) 10-m winds (m s21).
Anomalies are for the period of 1958–2007 relative to 1988–2000.
Data are 6-yr running means of area-averaged values from OAFlux
(solid) and NOCS (dashed) datasets. (top) SST from CMIP3 simu-
lations are also displayed (dotted).
TABLE 3. Same as in Table 2, but for SST. OAFlux SST is from
Reynolds et al. (2007), NOCS SST is from ICOADS (Worley et al.
2005). Data are in K decade21.
SST
1958–2006 1979–2006
OAFlux 0.086 6 0.057 0.304 6 0.098NOCS 0.245 6 0.098
CMIP3 0.113 6 0.022 0.211 6 0.035
FIG. 4. Changes in the seasonal cycle of Mediterranean Sea av-
eraged evaporation (E), precipitation (P), and E 2 P. Repre-
sented, 1979–88 means (solid) and 1996–2005 means (dashed).
Annual mean values for the two periods are also reported (hori-
zontal lines). Evaporation is from OAFlux and precipitation is
from GPCP (mm day21).
1518 J O U R N A L O F C L I M A T E VOLUME 23
above were primarily due to local increases in SST
and associated changes in Qs 2 Qa (see Fig. 6). The
greatest annual SST increases during 1979–2006 are
seen in the Ligurian Sea and southern parts of the
Mediterranean (up to 0.4 K decade21 according to
OAFlux). The Qs 2 Qa increases broadly follow the
SST pattern of increases, although differences exist
(e.g., in the Alboran Sea).
5. Long-term mean precipitation changes
Long-term mean precipitation changes in the Medi-
terranean region for the period since 1958 are presented
in Fig. 7. As discussed in section 2, only indirect esti-
mates of Mediterranean Sea precipitation are available
before 1979, based on REOFS reconstruction and land
gauge precipitation from regions surrounding the sea.
The analysis presented here considers together Medi-
terranean Sea estimates (GPCP, REMSS, and REOFS)
and those from land gauges (PRECL, CRU TS2.1, GHCN).
During 1979–2006, decadal variations characterize the
low-frequency variability of Mediterranean precipita-
tion. These are quite similarly represented in the various
datasets: a substantial decrease of 0.2–0.3 mm day21
(about 15%–25% of climatology) during the 1980s,
followed by a rapid increase in precipitation (about
0.2 mm day21) until the mid-1990s; after that, precipi-
tation is seen to decrease slightly up until the turn of the
twentieth century, with a recent tendency to increase.
Despite the general similarity, differences exist among
the various estimates, owing in part to the different areas
considered in the spatial averages (land-only versus sea
averages) and differing spatial resolutions. Overall,
GPCP precipitation suggests larger anomalies even com-
pared to the other sea-only averages, with an especially
large decrease during the 1980s. Between 1958 and 1979,
land gauge estimates show roughly constant precipitation
amounts, at about 0.1 mm day21 above the 1988–2000
FIG. 5. Linear trends for evaporation over the Mediterranean Sea for the period of 1979–
2006. Trends are for (top) annual means and (bottom) October–March means based on
OAFlux data. Units: mm day21 decade21.
15 MARCH 2010 M A R I O T T I 1519
means; REOFS gives higher precipitation compared to
the land gauge estimates (about 0.15 mm day21 above
the mean). Overall, for the period of 1958–2006 both
REOFS and the land gauge estimates suggest that these
decadal variations were superimposed on a long-term
negative trend (0.03–0.04 mm day21 decade21 or ;4%
of climatology per decade; various estimates are sta-
tistically consistent). In contrast, on the shorter 1979–
2006 period there is no significant precipitation trend.
Overall, CMIP3 model behavior is quite similar to
that depicted by the observational datasets; however,
the anomalies are about 5 times smaller than those
observed.
Changes in the mean seasonal cycle of precipitation
between the decades of 1979–88 and 1996–2005 are
mixed (see Fig. 4). Based on GPCP, precipitation de-
creased in January–March (;0.4 mm day21) and in-
creased during early fall. The spatial pattern of the
linear precipitation trend over the period of 1979–2006
is shown in Fig. 8. Annual mean decreases are found
only in the Adriatic Sea and parts of the southeastern
Mediterranean (0.1–0.2 mm day21 decade21). Nega-
tive trends become more substantial and widespread
October–March, with most of the Mediterranean ex-
periencing a precipitation decrease (0.1–0.2 mm day21
decade21), especially in the eastern Mediterranean (up
to 0.3 mm day21 decade21).
6. Long-term mean surface freshwater flux changes
The combination of the evaporation and precipitation
changes described in previous sections resulted in sig-
nificant long-term changes in Mediterranean Sea surface
freshwater fluxes during the period of 1958–2006 (Fig. 9).
FIG. 6. Same as Fig. 5, but for annual mean (top) Qs 2 Qa (g g21 decade21) and (bottom) SST
(K decade21).
1520 J O U R N A L O F C L I M A T E VOLUME 23
Estimates based on OAFlux/REOFS suggest a substantial
increase in E 2 P over this period (;0.5 mm day21 in
total). Considering the 1979–2006 subperiod, the E 2 P
rate of increase is estimated as 0.1–0.3 mm day21 decade21
(see Table 2; estimates are mostly statistically consis-
tent). The E 2 P increase during the 1980s is primarily
driven by the decrease in precipitation during this pe-
riod. Similarly, the ‘‘dip’’ in E 2 P during the mid-1990s
is also precipitation driven, and it is depicted quite
consistently across data sources. In contrast, the most
recent E 2 P increase is dominated by an evaporation
increase. The observational E 2 P results discussed here
are broadly consistent with those from the CMIP3 sim-
ulations, with an overall tendency for Mediterranean
E 2 P to increase during 1958–2006. However, CMIP3
E 2 P anomalies are about one order of magnitude
smaller than observed.
The comparison of the E 2 P seasonal cycle during the
decades of 1979–88 and 1996–2005, based on OAFlux/
GPCP, indicates that most of the E 2 P increase
occurred during September–March (see Fig. 4). Evapo-
ration increase was the main cause of the September–
December E 2 P increase, with the precipitation decrease
significantly contributing to the January–March increase.
Results using NOCS/GPCP are similar, although their
rates of change are smaller (not shown).
During 1979–2006, annual mean E 2 P increased
everywhere in the Mediterranean Sea, and most sub-
stantially in the Ligurian Sea, Adriatic Sea, and parts
of the southeastern Mediterranean (up to 0.4–0.5
mm day21 decade21 based on OAFlux and GCPC esti-
mates; Fig. 10). Increases of 0.2–0.3 mm day21 decade21
were widespread. October–March means shows a similar
pattern of increase but the rates of increase are much
higher (over 0.5 mm day21 decade21) in vast parts of
the Mediterranean. A similar analysis based on NOCS/
GPCP (not shown), gives E 2 P trend patterns that are
consistent with those described above, except that the
rates of change are generally more modest (maximum
annual rates are 0.3–0.4 mm day21 decade21).
7. Summary and concluding remarks
The analysis of the Mediterranean water cycle vari-
ability presented here provides new insights on the long-
term changes that have affected the region since the
1960s. Sea evaporation significantly increased since the
mid-1970s (0.1–0.2 mm day21 decade21) with a ten-
dency toward higher rates of increase during the 1990s.
According to OAFlux, this long-term increase followed
a period of evaporation decrease during 1965–75, but
overall evaporation has increased since the 1960s by
about 10% in total (0.06 mm day21 decade21). Much of
the evaporation increase since the mid-1970s has been in
early winter, especially in the Ligurian Sea, Adriatic
Sea, and southeastern Mediterranean. Increases in SSTs
have primarily driven these evaporation changes via
changes in the surface humidity gradient. Based on
OAFlux data, the estimated Mediterranean mean rate
of evaporation change in relation to the warming is
;0.7 mm day21 K21 (or 25% K21) over the period of
1958–2006 (similarly during 1979–2006). Significant in-
terdecadal variations characterized precipitation dur-
ing the period of 1979–2006. Most noticeable is the
well-know precipitation decrease during the 1980s, which
has been attributed to the regional influence of the
North Atlantic Oscillation (e.g., Hurrell 1995). Average
Mediterranean Sea precipitation decreased during the
period of 1958–2006 (0.03–0.04 mm day21 decade21)
while there is no annual precipitation trend during
1979–2006. The combination of the evaporation and
precipitation changes described above resulted in sub-
stantial long-term changes in Mediterranean Sea surface
freshwater fluxes with an overall increase estimated in
the range of 0.1–0.3 mm day21 decade21 during the
period of 1979–2006. This increase affected the Ligurian
Sea in particular, the Adriatic and parts of the south-
eastern Mediterranean, and especially the October–
March means. Concerning the longer 1958–2006 period,
OAFlux/REOFS datasets give an E 2 P decrease during
the period from 1958 to the mid-1970s, but indicate an
overall mean increase in E 2 P since 1958 of 0.1 mm
day21 decade21 (;0.5 mm day21 in total).
FIG. 7. Decadal variations in Mediterranean mean precipitation
over the period of 1958–2007. Shown are 6-yr running means of
area-averaged precipitation anomalies relative to the period 1988–
2000 (lines, mm day21) from various observational sources (see
legend). PRECL, CRU, and GHCN are land-only averages for the
region surrounding the Mediterranean Sea; REOFS, GPCP, and
REMSS are Mediterranean Sea–only averages. For GPCP, annual
mean values are also shown (symbols). Running mean averages
based on CMIP3 simulations are also displayed (note different
scale at right).
15 MARCH 2010 M A R I O T T I 1521
The dataset intercomparison highlights qualitative
agreement among the datasets on the long-term be-
havior described above, albeit large uncertainties on the
quantitative estimates of the rate of change. These un-
certainties are less well defined for the presatellite era
because fewer data sources are available for compari-
son. Among the caveats is that many of the analyzed
datasets, and especially the satellite-based ones, share
common data sources, although data are often diversely
processed and combined (see section 2). Other datasets
(e.g., OAFlux) instead use reanalyses for their deriva-
tion, and hence are not purely observational. At a global
level, the OAFlux-derived evaporation increase since
1958 (;30% K21) is quite high (Stephens and Ellis 2008;
Takahashi 2009; Vecchi and Soden 2007; Wentz et al.
2007). However, over the 1979–2006 period, OAFlux
global evaporation rates are similar to those of NOCS
evaporation and global oceanic precipitation rates by
GPCP (L. Yu 2009, personal communication). At a
regional level, the changes in Mediterranean Sea evap-
oration described here reflect the Mediterranean Sea
warming observed in recent decades. Similarly, an in-
crease in Mediterranean Sea surface freshwater loss
is broadly consistent with previous studies suggesting
an increase in the salinity of various components of
the Mediterranean Sea system (e.g., Millot et al. 2006;
Potter and Lozier 2004; Rixen et al. 2005). For instance,
Rixen et al. (2005) estimate a mean Mediterranean
salinity increase of 0.03–0.04 psu over the period of
1950–2000. A simplified form of the salt conservation
equation, neglecting all dynamical terms, indicates that
an increase in freshwater loss of 0.06–0.15 mm day21
decade21 (0–0.09 mm day21 decade21) during 1958–
2006 (1958–2000) would result in a 0.07–0.15 psu (0.-0.08
psu) increase in mean Mediterranean salinity. While
other studies have emphasized the effects of decreased
precipitation and river discharge on observed salin-
ity changes (Krahmann 1998; Rohling and Bryden
1992; Skliris et al. 2007), our results suggest an im-
portant role for evaporation increase, which needs to
be further investigated. Similarly, underlying causes
and broader implications of the water cycle changes
FIG. 8. Same as Fig. 5, but trends are for GPCP precipitation data.
1522 J O U R N A L O F C L I M A T E VOLUME 23
described here also need to be explored in future
work.
Results from a large ensemble of CMIP3 model
simulations are in qualitative agreement with the ob-
servational results. However, simulated water cycle
anomalies are about one order of magnitude smaller
than that observed despite the fact that CMIP3 SST
anomalies are comparable to the observations. A muted
response of surface evaporation and precipitation to
warming in global climate simulations compared to
observations has previously been noted (Richter and
Xie 2008; Wentz et al. 2007). Internal decadal climate
variability (e.g., related to NAO variability), not cap-
tured by the CMIP3 ensemble means, also contributes
to discrepancies between CMIP3 and observations and
needs to be further investigated. CMIP3 projections for
the twenty-first century indicate that the Mediterranean
Sea loss of freshwater will accelerate in future decades
(Mariotti et al. 2008). Confidence in these projections
will depend on our ability to reconcile differences be-
tween observations and simulations, better defining past
observed changes and improving our understanding and
simulation capability. Long-term monitoring is needed
to ensure that future long-term changes in the Medi-
terranean water cycle do not go undetected.
FIG. 9. Same as Fig. 2, but for E 2 P fluxes area-averaged over
the Mediterranean Sea. Annual E 2 P means based on OAFlux
and GPCP data are also shown (symbols).
FIG. 10. Same as Fig. 5, but trends are for E 2 P fluxes based on OAFlux and GPCP datasets.
15 MARCH 2010 M A R I O T T I 1523
Acknowledgments. The author would like to thank
Rong-Hua Zhang, Jin-Ho Yoon, Ning Zeng, Lisan Yu,
Lucrezia Ricciardulli, and Volgango Rupolo for their
valuable input and the two anonymous reviewers for
their comments; the modeling groups, the Program
for Climate Model Diagnosis and Intercomparison
(PCMDI) and the WCRP’s Working Group on Coupled
Modeling (WGCM) for their roles in making available
the WCRP CMIP3 multimodel dataset. Support of this
dataset is provided by the Office of Science, U.S. De-
partment of Energy. The author thankfully acknowl-
edges all those who provided data for this study and the
EC for financial support under CIRCE Integrated Pro-
ject (Contract 036961).
REFERENCES
Adler, R. F., and Coauthors, 2003: The version-2 Global Pre-
cipitation Climatology Project (GPCP) monthly precipitation
analysis (1979–present). J. Hydrometeor., 4, 1147–1167.
——, G. Gu, J.-J. Wang, G. J. Huffman, S. Curtis, and D. Bolvin,
2008: Relationships between global precipitation and surface
temperature on interannual and longer timescales (1979–2006).
J. Geophys. Res., 113, D22104, doi:10.1029/2008JD010536.
Allan, R., and T. Ansell, 2006: A new globally complete monthly
historical gridded mean sea level pressure dataset (HadSLP2):
1850–2004. J. Climate, 19, 5816–5842.
Andersson, A., S. Bakan, K. Fennig, H. Grassl, C.-P. Klepp, and
J. Schulz, 2007: Hamburg Ocean Atmosphere Parameters and
Fluxes from Satellite Data—HOAPS-3—Monthly mean. World
Data Center for Climate, doi:10.1594/WDCC/HOAPS3_
MONTHLY.
Berry, D. I., and E. C. Kent, 2008: A new air–sea interaction
gridded dataset from ICOADS with uncertainty estimates.
Bull. Amer. Meteor. Soc., 90, 645–656.
Bethoux, J., B. Gentili, and D. Tailliez, 1998: Warming and fresh-
water budget change in the Mediterranean since the 1940s,
their possible relation to the greenhouse effect. Geophys. Res.
Lett., 25, 1023–1026.
Chen, M. Y., P. Xie, J. E. Janowiak, and P. A. Arkin, 2002: Global
land precipitation: A 50-yr monthly analysis based on gauge
observations. J. Hydrometeor., 3, 249–266.
Chou, S.-H., E. Nelkin, J. Ardizzone, R. M. Atlas, and C.-L. Shie,
2003: Surface turbulent heat and momentum fluxes over
global oceans based on the Goddard satellite retrievals, ver-
sion 2 (GSSTF2). J. Climate, 16, 3256–3273.
Fairall, C. W., E. F. Bradley, J. E. Hare, A. A. Grachev, and
J. B. Edson, 2003: Bulk parameterization of air–sea fluxes:
Updates and verification for the COARE algorithm. J. Cli-
mate, 16, 571–591.
Gibelin, A. L., and M. Deque, 2003: Anthropogenic climate change
over the Mediterranean region simulated by a global variable
resolution model. Climate Dyn., 20, 327–339.
Giorgi, F., 2006: Climate change hot-spots. Geophys. Res. Lett., 33,
L08707, doi:10.1029/2006GL025734.
——, and P. Lionello, 2008: Climate change projections for the
Mediterranean region. Global Planet. Change, 63 (2–3), 90–104.
Hurrell, J. W., 1995: Decadal trends in the North Atlantic Oscil-
lation—Regional temperatures and precipitation. Science,
269, 676–679.
Josey, S. A., 2003: Changes in the heat and freshwater forcing of the
eastern Mediterranean and their influence on deep water for-
mation. J. Geophys. Res., 108, 3237, doi:10.1029/2003JC001778.
Jung, T., and Coauthors, 2006: Response to the summer of 2003
Mediterranean SST anomalies over Europe and Africa.
J. Climate, 19, 5439–5454.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-
analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Krahmann, G., 1998: Longterm increases in western Mediterra-
nean salinities and temperatures: Anthropogenic and climatic
sources. Geophys. Res. Lett., 25, 4209–4212.
Lozier, M. S., and N. M. Stewart, 2008: On the temporally varying
northward penetration of Mediterranean Overflow Water and
eastward penetration of Labrador Sea. J. Phys. Oceanogr., 38,
2097–2103.
Marcos, M., and M. N. Tsimplis, 2008: Comparison of results of
AOGCMs in the Mediterranean Sea during the 21st century.
J. Geophys. Res., 113, C12028, doi:10.1029/2008JC004820.
Mariotti, A., M. V. Struglia, N. Zeng, and K.-M. Lau, 2002: The
hydrological cycle in the Mediterranean region and implica-
tions for the water budget of the Mediterranean Sea. J. Cli-
mate, 15, 1674–1690.
——, N. Zeng, J.-H. Yoon, V. Artale, A. Navarra, P. Alpert, and
L. Z. X. Li, 2008: Mediterranean water cycle changes: Tran-
sition to drier 21st century conditions in observations and
CMIP3 simulations. Environ. Res. Lett., 3, 044001, doi:10.
1088/1748-9326/3/4/044001.
Millot, C., J. Candela, J.-L. Fuda, and Y. Tber, 2006: Large
warming and salinification of the Mediterranean outflow due
to changes in its composition. Deep-Sea Res. I, 53, 656–666.
Mitchell, T. D., and P. D. Jones, 2005: An improved method of
constructing a database of monthly climate observations and
associated high-resolution grids. Int. J. Climatol., 25, 693–712.
Potter, R. A., and M. S. Lozier, 2004: On the warming and salini-
fication of the Mediterranean outflow waters. Geophys. Res.
Lett., 31, L01202, doi:10.1029/2003GL018161.
Reid, J. L., 1979: On the contribution of the Mediterranean Sea
outflow to the Norwegian–Greenland Sea. Deep-Sea Res., 26,
1199–1223.
Reynolds, R. W., C. Liu, T. M. Smith, D. B. Chelton, M. G. Schlax,
and K. S. Casey, 2007: Daily high-resolution-blended analyses
for sea surface temperature. J. Climate, 20, 5473–5496.
Richter, I., and S.-P. Xie, 2008: Muted precipitation increase in
global warming simulations: A surface evaporation perspec-
tive. J. Geophys. Res., 113, D24118, doi:10.1029/2008JD010561.
Rixen, M., and Coauthors, 2005: The Western Mediterranean
Deep Water: A proxy for climate change. Geophys. Res. Lett.,
32, L12608, doi:10.1029/2005GL022702.
Roether, W., B. Klein, B. B. Manca, A. Theocharis, and
S. Kioroglou, 2007: Transient Eastern Mediterranean deep
waters in response to the massive dense-water output of the
Aegean Sea in the 1990s. Prog. Oceanogr., 74, 540–571.
Rohling, E. J., and H. L. Bryden, 1992: Man-induced salinity and
temperature increase in Western Mediterranean Deep Water.
J. Geophys. Res., 97 (C7), 11 191–11 198.
Sheffield, J., and E. F. Wood, 2008: Projected changes in
drought occurrence under future global warming from multi-
model, multi scenario, IPCC AR4 simulations. Climate Dyn.,
31, 79–105.
Skliris, N., S. Sofianos, and A. Lascaratos, 2007: Hydrological
changes in the Mediterranean Sea in relation to changes in the
freshwater budget: A numerical modelling study. J. Mar. Syst.,
65 (1–4), 400–416.
1524 J O U R N A L O F C L I M A T E VOLUME 23
Smith, T. M., M. R. P. Sapiano, and P. A. Arkin, 2008: Historical
reconstruction of monthly oceanic precipitation (1900–2006).
Atmospheres, 113, D17115, doi:10.1029/2008JD009851.
Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis,
K. B. Averyt, M. Tignor, and H. L. Miller, 2007: Climate
Change 2007: The Physical Science Basis. Cambridge Uni-
versity Press, 996 pp.
Stephens, G. L., and T. D. Ellis, 2008: Controls of global-mean
precipitation increases in global warming GCM experiments.
J. Climate, 21, 6141–6155.
Takahashi, K., 2009: Radiative constraints on the hydrological
cycle in an idealized radiative–convective equilibrium model.
J. Atmos. Sci., 66, 77–91.
Tsimplis, M. N., M. Marcos, and S. Somot, 2008: 21st century
Mediterranean sea level rise: Steric and atmospheric pressure
contributions from a regional model. Global Planet. Change,
63 (2–3), 105–111.
Ulbrich, U., and Coauthors, 2006: The Mediterranean climate
change under global warming. Mediterranean Climate Vari-
ability and Predictability, P. Lionello et al., Eds., Elsevier,
398–415.
Uppala, S. M., and Coauthors, 2005: The ERA-40 re-analysis.
Quat. J. Roy. Meteor. Soc., 131, 2961–3012.
Vecchi, G. A., and B. J. Soden, 2007: Global warming and the
weakening of the tropical circulation. J. Climate, 20, 4316–
4340.
Vose, R. S., R. L. Schmoyer, P. M. Steurer, T. C. Peterson,
R. Heim, T. R. Karl, and J. K. Eischeid, 1992: The Global
Historical Climatology Network: Long-term monthly tem-
perature, precipitation, sea level pressure, and station pressure
data. Carbon Dioxide Information Analysis Center, Oak
Ridge National Laboratory Doc. ORNL/CDIAC-53, NDP-
041, 324 pp. [Available online at http://cdiac.ornl.gov/ftp/
ndp041/ndp041.pdf.]
Wentz, F. J., L. Ricciardulli, K. Hilburn, and C. Mears, 2007: How
much more rain will global warming bring? Science, 317,
233–235.
Worley, S. J., S. D. Woodruff, R. W. Reynolds, S. J. Lubker, and
N. Lott, 2005: ICOADS release 2.1 data and products. Int.
J. Climatol., 25, 823–842.
Yu, L., 2007: Global variations in oceanic evaporation (1958–2005):
The role of the changing wind speed. J. Climate, 20, 5376–5390.
——, and R. A. Weller, 2007: Objectively analyzed air-sea heat
fluxes for the global ice-free oceans (1981–2005). Bull. Amer.
Meteor. Soc., 88, 527–539.
——, X. Jin, and R. A. Weller, 2008: Multidecade global flux
datasets from the Objectively Analyzed Air–Sea Fluxes
(OAFlux) Project: Latent and sensible heat fluxes, ocean evap-
oration, and related surface meteorological variables. Woods
Hole Oceanographic Institution OAFlux Project Tech. Rep.
OA-2008-01, 64 pp. [Available online at http://oaflux.whoi.edu/
pdfs/OAFlux_TechReport_3rd_release.pdf.]
15 MARCH 2010 M A R I O T T I 1525