Seasonal Variability of the Tyrrhenian Sea Surface Geostrophic Circulation as Assessed by Altimeter Data

R. Iacono ENEA, C. R. Casaccia, Rome, Italy

Search for other papers by R. Iacono in
Current site
Google Scholar
PubMed
Close
,
E. Napolitano ENEA, C. R. Casaccia, Rome, Italy

Search for other papers by E. Napolitano in
Current site
Google Scholar
PubMed
Close
,
S. Marullo ENEA, C. R. Casaccia, Rome, Italy

Search for other papers by S. Marullo in
Current site
Google Scholar
PubMed
Close
,
V. Artale ENEA, C. R. Casaccia, Rome, Italy

Search for other papers by V. Artale in
Current site
Google Scholar
PubMed
Close
, and
A. Vetrano CNR-ISMAR, La Spezia, Italy

Search for other papers by A. Vetrano in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

New insights into the structure and variability of the Tyrrhenian Sea's surface circulation are obtained through the analysis of a very long series of altimetric observations (1993–2010). In late winter and part of spring, a consistent mean flow is individuated in the eastern Tyrrhenian Sea, formed by a stream of Atlantic water that meanders around four anticyclonic structures located along the Italian coast, which have smaller cyclonic companions offshore. The signatures of these vortices are also found in images of chlorophyll and sea surface temperature, as well as in modeling results, both from a high-resolution operational model of the Tyrrhenian Sea's circulation and from a dedicated numerical simulation. Analysis of the energy exchange between eddies and mean flow, together with numerical evidence, suggests that this winter–spring circulation pattern may result from basin-scale instability of the Atlantic stream. In summer, the dynamic is dominated by a well-known dipole located to the east of the Bonifacio Strait. However, in the eastern part of the basin, an anticyclonic cell is also found, probably driven by the negative wind stress curl present in summer in this region. The cell encompasses two anticyclonic vortices located in the areas of the Vavilov and Marsili Seamounts. A multichannel singular spectral analysis of the altimetric time series reveals that, besides the expected, dominant seasonal mode, a significant low-frequency mode of variability is also present. This mode has a period of about six years and is mostly localized in the western part of the basin.

Corresponding author address: R. Iacono, ENEA, C. R. Casaccia, Via Anguillarese 301, 00123 Rome, Italy. E-mail: roberto.iacono@enea.it

Abstract

New insights into the structure and variability of the Tyrrhenian Sea's surface circulation are obtained through the analysis of a very long series of altimetric observations (1993–2010). In late winter and part of spring, a consistent mean flow is individuated in the eastern Tyrrhenian Sea, formed by a stream of Atlantic water that meanders around four anticyclonic structures located along the Italian coast, which have smaller cyclonic companions offshore. The signatures of these vortices are also found in images of chlorophyll and sea surface temperature, as well as in modeling results, both from a high-resolution operational model of the Tyrrhenian Sea's circulation and from a dedicated numerical simulation. Analysis of the energy exchange between eddies and mean flow, together with numerical evidence, suggests that this winter–spring circulation pattern may result from basin-scale instability of the Atlantic stream. In summer, the dynamic is dominated by a well-known dipole located to the east of the Bonifacio Strait. However, in the eastern part of the basin, an anticyclonic cell is also found, probably driven by the negative wind stress curl present in summer in this region. The cell encompasses two anticyclonic vortices located in the areas of the Vavilov and Marsili Seamounts. A multichannel singular spectral analysis of the altimetric time series reveals that, besides the expected, dominant seasonal mode, a significant low-frequency mode of variability is also present. This mode has a period of about six years and is mostly localized in the western part of the basin.

Corresponding author address: R. Iacono, ENEA, C. R. Casaccia, Via Anguillarese 301, 00123 Rome, Italy. E-mail: roberto.iacono@enea.it

1. Introduction

The Tyrrhenian Sea (TYS hereafter, see Fig. 1), the main Italian sea, is a deep basin with complex bathymetry in which surface waters of Atlantic origin [Atlantic Waters (AW)] and salty intermediate waters coming from the eastern Mediterranean Sea [Levantine Intermediate Waters (LIW)] get transformed and mixed. From late autumn to early spring, these transformed water masses outflow vigorously into the Liguro–Provençal Sea (see, e.g., Astraldi et al. 1999), affecting the dynamics of this subbasin, where one of the main Mediterranean sites of deep-water formation is located (see, e.g., MEDOC Group 1970). The study of the TYS dynamics has therefore direct bearing on the understanding of the global Mediterranean Sea thermohaline circulation.

Fig. 1.
Fig. 1.

The Tyrrhenian Sea, with its three openings: the Sardinia Channel, the Corsica Channel, and the Sicily Strait. Shading indicates bathymetry and arrows indicate the main patterns of surface (AW; red) and intermediate (LIW; yellow) waters. Ovals in the western part of the basin schematize the Bonifacio dipole and, below, the cyclonic circulation between Sardinia and Sicily—both quasi-permanent features of the TYS circulation.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

The TYS has three main openings: the Sardinia and Corsica Channels, connecting the basin to the rest of the western Mediterranean Sea, and the Sicily Strait, which is the entrance door of the eastern Mediterranean Sea (Levantine Basin). Over the years, observations of the exchanges through these openings and of the basin hydrology have led to a description of the TYS large-scale circulation that has been summarized in classical works by Krivosheya (1983), Astraldi and Gasparini (1994), and Millot (1999). The picture is that of a mainly wind-driven circulation, with some quasi-permanent features in the western part of the basin: a wide cyclonic area to the southeast of Sardinia, and a cyclonic center to the east of Corsica (the Bonifacio gyre) having an anticyclonic companion to the south (see sketch in Fig. 1). In winter, these structures are embedded into a basin-scale cyclonic cell that brings the AW from the south up to the northern end of the basin. In summer, the AW circulation weakens and appears more confined, the cyclonic centers intensify, and a few other eddies appear in the eastern part of the TYS.

The typical wind forcing over the basin can be inferred from Fig. 2, which shows the relative vorticity at 1000 mb [1990–2010 averages from the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Re-Analysis (ERA-Interim) dataset] for February and August, representative of winter and summer conditions, respectively. In both seasons, the dominant feature in the northwestern quadrant is a dipole formed by the winds blowing in the area of the Bonifacio Strait (the narrow passage between Corsica and Sardinia), which feeds the Bonifacio couple of gyres (see, e.g., Artale et al. 1994). To the south, another positive pole is present, associated with the previously mentioned cyclonic region between Sardinia and Sicily. However, the overall structure of the wind curl is different in the two seasons, with a global cyclonic cell present in winter that breaks down in summer over the central and eastern TYS, where the forcing becomes prevalently anticyclonic. Both works using simplified dynamical models (Artale et al. 1994; Pierini and Simioli 1998) and the first—relatively coarse-resolution—GCM simulations of the Mediterranean Sea circulation (Roussenov et al. 1995; Zavatarelli and Mellor 1995; Korres et al. 2000) have indicated that this could produce a reversal of the circulation in summer in this part of the basin.

Fig. 2.
Fig. 2.

Relative vorticity of the wind at 1000 mb for (left) February and (right) August, from the ERA-Interim dataset (1990–2010 averages). Red boxes in the February and August maps delimit the regions on which the averages of Figs. 3 and 11 (both figures are described in greater detail below) are computed, respectively.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

In the 1990s, systematic satellite observations revealed rich mesoscale dynamics that could not be adequately resolved by the just-mentioned GCMs [for the TYS, see Marullo et al. (1994)]. This prompted the development of a new generation of GCMs, with a horizontal resolution of . However, works using these new models (e.g., Béranger et al. 2005; Jordi and Wang 2009; Tonani et al. 2008) have not addressed the TYS dynamics in detail.

Two recent studies dedicated to the TYS have provided contrasting indications. From their analysis of drifter trajectories and altimeter anomalies from the period 2001–04, Rinaldi et al. (2010) concluded that, except for a semipermanent central gyre, the circulation in the central and eastern TYS appears dominated by transient features, and that it is consequently difficult to identify a consistent mean flow. On the other hand, in the reconstruction of the spring 2004 circulation by Vetrano et al. (2010), based both on observations and on numerical simulations, the central and eastern TYS were filled with large cyclonic and anticyclonic structures. These structures were also found in typical spring altimeter data, and could therefore be robust features of spring dynamics around which the flow and the transport organize.

To discriminate between these two descriptions, and to gain further insight into the structure of the mean TYS circulation in the different seasons, we have analyzed the long time series of altimeter data now available, which covers almost two decades. The main purpose of this work is to present the results of this analysis, which confirm and extend in many ways the picture sketched in Vetrano et al. (2010).

The paper is organized as follows. In section 2, the seasonal circulation patterns are characterized, and several robust structures are individuated. The winter–spring and summer circulations are further analyzed in section 3, where additional, independent evidence of the presence of the main vortices found in the altimeter maps is provided. Modeling results are also discussed, which shed light on the evolution of the mean flow in the two seasons. Then, in section 4, a multichannel singular spectral analysis of the time series of altimeter maps is performed, to characterize the main modes of variability of the TYS circulation. Results are summarized in section 5, where some directions of future work are also indicated.

2. Analysis of the AVISO data over the TYS

Starting from the end of 1992, high-quality altimeter data have continuously been collected over the Mediterranean Sea. These data have often been used to investigate specific aspects of the surface circulation, usually over limited time spans in which other types of measurements were also available. More recently, however, some studies have been trying to take advantage of the full extent of the dataset (see, e.g., Amitai et al. 2010; Ioannone et al. 2011; Poulain et al. 2012). Here we shall examine Archiving, Validation, and Interpretation of Satellite Oceanographic (AVISO) data over the TYS, for the period 1993–2010.

Before looking at the data, a few points should be stressed. First of all, we should keep in mind that the altimeter data do not provide information on the circulation at the sea surface, which is dominated by high-frequency wind forcing. The altimeter signal is an integrated one, which reflects the motion of the main pycnocline (Wunsch 1997). In the western Mediterranean Sea, however, the main pycnocline, which is in the region of transition between the AW and the LIW, is quite shallow, and one may expect its motion (first baroclinic mode) to dominate the geostrophic dynamics of the surface layer.

The second point concerns the size of the structures we look for. We shall focus on the weekly updated delayed maps of sea level anomaly (SLA), which result from an offline elaboration of the tracks of all the available satellites: two of them until 1999, and at least three in the following 11 years, with a peak of four between October 2002 and September 2005. These maps have a horizontal resolution of ⅛°, comparable to the typical Rossby radius of deformation (10–15 km for the Mediterranean Sea), but we cannot hope to find in them well-resolved eddies with sizes of few tens of kilometers. This is because of the subsampling and filtering used to reduce the noise along the tracks, and of the subsequent objective interpolation between tracks, whose accuracy depends both on the geometry of the tracks and on the number of altimeters available. A detailed assessment of the “effective” resolution of the maps is not easy, but the analysis by Pascual et al. (2007) indicates that three altimeters are sufficient for a correct monitoring of the mesoscale circulation of the Mediterranean Sea, and that four may resolve circulation features with sizes significantly smaller than 1°. We can therefore expect that eddies with a size of about 100 km, or slightly smaller, could be sufficiently resolved over most of the dataset.

Another caveat concerns the quality of the reconstruction in coastal zones (see, e.g., Vignudelli et al. 2005), where the vicinity of land may corrupt raw along-track data. This is an open problem in the interpretation of altimeter data that recent works (see, e.g., Melet et al. 2010; Dussurget et al. 2011) have been trying to address, using ad hoc reprocessing techniques. We shall not attempt a similar approach; when needed, we shall seek additional independent evidence, to decide whether coastal structures are real, or are artifacts due to the treatment of the data, or to data inaccuracies.

Finally, it should be noted that the SLA—the time-varying signal extracted from the satellite measurements—is not always sufficient to draw a correct picture of the main flow patterns. Sometimes, structures appearing in the SLA maps cannot be univocally associated with specific circulation features, and to remove the ambiguity one needs information on the height change associated with the mean circulation present in the area, the so-called mean dynamic topography (MDT). Adding the MDT to the SLA yields the absolute dynamic topography (ADT), maps of which are also provided by AVISO. When looking at the ADT, however, one has to be aware that the reconstruction of the MDT has its own limitations and uncertainties, which may locally obscure the information contained in the SLA. This may be a reason of concern, particularly in areas, such as the Mediterranean, in which the circulation displays a strong seasonal cycle, which is not necessarily well resolved in the data used for the reconstruction of the MDT. The only way out of this conundrum is through carefully comparing the SLA and the ADT maps, and, in case of doubt, looking for independent observational and/or modeling evidence of the individuated circulation structures.

a. The mean dynamic topography

The MDT for the whole Mediterranean Sea, at a resolution of ⅛°, was computed by Rio et al. (2007) from seven years of altimeter anomalies (1993–99), in situ measurements, buoy data, and outputs of a general circulation model, the latter being used only to construct the initial guess for an iterative procedure. The MDT for the TYS (their Fig. 10) is displayed in the central panel of Fig. 4 (described in greater detail below). Commenting on this map, Rio et al. (2007) noted the presence of a basinwide cyclonic circulation and some large recirculation cells. Among these are the expected, permanent cyclonic vortices residing in the western TYS, but also two strong anticyclonic structures, to the north of Sicily and just off Naples, not mentioned in previous descriptions of the TYS mean circulation.1 The presence of two weaker and smaller cyclonic cells, between the two anticyclones, and of another weak anticyclonic signal near the coast of Calabria can also be noted. Thus, there appear to be five candidate vortices in the area, which could be robust features of the eastern TYS circulation. This gives us a first indication about what to look for when examining the SLA maps.

It is natural to ask whether the typical surface forcing in the area may provide a plausible driving mechanism for at least some of these structures. Figure 3 displays two time series, obtained by averaging the surface wind curl from the ERA-Interim dataset (0.75° resolution) over the two boxes indicated in the February map of Fig. 2, which include the two wide anticyclones. The first time series (Fig. 3a) has a regular structure, with small values of the wind curl in winter, strong negative values in summer, and a negative mean, suggesting that the Sicilian anticyclone is a recurrent feature of the TYS circulation, in the establishment of which wind forcing plays an important role. By contrast, in the other time series (Fig. 3b), the average wind curl is mostly cyclonic, with recurrent positive peaks in winter. Thus, other driving mechanisms must be invoked to explain the presence of the anticyclonic cell off the Campania coast.

Fig. 3.
Fig. 3.

Time series of average surface wind curl from the ERA-Interim dataset: (a) to the north of Sicily (38°–39°N, 13°–15.5°E) and (b) off the coast of Campania (39.5°–41°N, 12°–15°E). Boxes showing the regions over which the averages are performed are also indicated in Fig. 2 (left).

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

b. ADT versus SLA

We now examine the SLA and ADT maps, looking for robust circulation features. The average (1993–2010) winter (January–March) and summer (July–September) maps are displayed in Fig. 4, with the panels on the left showing the SLA for the two seasons, indicated by shading and contours, and those on the right displaying the corresponding ADT maps, with geostrophic reconstructions of the circulation superposed, also provided by AVISO. In the middle is the MDT of Rio et al. (2007). Note that the ranges of variation of the MDT and of the ADT are comparable, and significantly larger than those of the SLA, indicating that the MDT plays a strong role in determining the ADT in both seasons. However, the inclusion of the SLA is sufficient to yield quite different winter and summer circulations. In winter, the basinwide cyclonic circulation present in the MDT map gets reinforced by the analogous signal present in the SLA map, yielding a vigorous AW stream in the ADT panel, which circles all around the Italian coasts, from the Sardinia Channel to the northern end of the basin. Such a stream is not present in the summer ADT map, in which the eastern TYS is filled with vortices, without a clear mean flow. This is due to the presence of a wide anticyclonic cell in the summer SLA, which almost completely offsets the cyclonic signal present in the MDT.

Fig. 4.
Fig. 4.

Maps of average (1993–2010) (left) SLA and (right) ADT for (top) winter and (bottom) summer, with (middle) the MDT by Rio et al. (2007). A geostrophic reconstruction of the surface flow is superposed on the ADT maps. Contour intervals are 0.5 cm for the SLA and 0.75 cm for MDT.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

Let us now look more closely at the main vortices appearing in the ADT maps. In the western part of the basin, the main structures are those already mentioned in the introduction, that is, the Bonifacio dipole, and the wide cyclonic area between Sardinia and Sicily, which appears to encompass several smaller gyres. They are present in both seasons, even if with some differences; for example, the Bonifacio cyclone is more elongated zonally in summer, virtually occupying the whole space between Corsica and the Italian coast. Note, however, that while these signals are also prominent in the summer SLA map, they cannot be identified in the winter SLA, in which the Bonifacio dipole seems inverted, with a cyclonic anomaly to the east of the Sardinian coast. This provides a vivid illustration of the crucial role that the MDT can play in the construction of the ADT.

The situation in the eastern TYS is more complicated. The two wide anticyclonic vortices discussed in the previous subsection are present in the ADT in both seasons. In winter, these vortices appear inserted into a more complex, global circulation pattern: when going from Sicily to Tuscany, four anticyclones are found on the shore side of the meandering AW stream, having cyclonic companions—even if less well defined—offshore. The same pattern is also clearly visible in the winter SLA, even if the anticyclonic signals are closer to the coast. On the other hand, one of the strongest signals in the SLA, a cyclonic anomaly centered around 39°N and 13°E, does not have a counterpart in the ADT map, because in that area the MDT reconstruction places a wide anticyclonic meander, just to the north of the Sicilian anticyclone.

In summer, there are differences between the ADT and the SLA. The Sicilian anticyclone is absent in the SLA, but this could just be an indication of the fact that this structure is quasi-permanent; as such, it mostly shows up in the MDT. After the Bonifacio anticyclone, the strongest anticyclonic signals in the SLA are in a central area, quite far from the Campania coasts, and off the Calabria coasts, but more offshore than a corresponding signal present in the ADT map. It may be noted that the central anticyclonic area includes the region in which both Rinaldi et al. (2010) and Budillon et al. (2009) locate a persistent anticyclone.

c. The SLA in the eastern TYS

Let us now examine more closely the evolution of the SLA in the eastern TYS (to the right of 11°E). Figures 5a,b show 12 monthly SLA maps (averaged over 1993–2010); the basin average has been subtracted for each month to keep the same scale (−4, 4) over most of the year. From December to March, there is clear evidence in the maps of a cyclonic stream that circles around the Italian coast and becomes increasingly meandering throughout the season. The four previously noted anticyclonic signals are present along the coast, to the north of Sicily, to the west of Calabria, and then approximately off of Naples and Rome. On the March map, we have numbered these possible anticyclones from 1 to 4 and the four corresponding cyclonic signals on the offshore side of the stream from 5 to 8 [two of which (5 and 8) are better defined]. It may be noted that 2, 3, and 4, confined near coast in December and January, expand toward the inside of the basin and become better defined toward the end of the winter, when the cyclonic wind forcing weakens.

Fig. 5.
Fig. 5.

(a) Monthly SLA maps (November–April), averaged over 1993–2010, for the central and eastern TYS. Numbers from 1 to 8 in the March map denote anomalies corresponding to anticyclonic (1–4) and cyclonic (5–8) structures on the two sides of the AW stream that in winter circles cyclonically along the coasts of the basin. (b) As in (a), but for the months from May to October. The 2000- and 3000-m-depth contours have been superposed to the July map, to show that two of the main summer anticyclonic anomalies are near the Vavilov (39.85°N, 12.58°E) and the Marsili (39.28°N, 14.4°E) Seamounts, the two main submerged mountains of the basin.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

Approaching the summer season, the anticyclonic signals apparently move toward the west, while the winter flow pattern breaks down. In June, there is no signature left of the cyclonic AW stream and no clear anomaly off Naples. Between 40° and 41°N and 12° and 14°E, a weak and broad anticyclonic signal is present instead, which collapses in July into a stronger and more localized anticyclonic signal, previously also seen in the average summer SLA map. We have assigned number 9 to this anticyclonic area (see the August panel of Fig. 5b). As shown by the thick lines in the July map, indicating the 2000- and 3000-m depth contours, this signal encompasses the Vavilov Seamount, to the east of which Budillon et al. (2009) observed the signature of a strong anticyclonic eddy during two hydrographic cruises (July and December 2005). Note that the anticyclonic number 2 now leans against the Marsili Seamount, the main submerged mountain of the TYS, with a summit at about 450 m of depth. Structures number 2 and 9 are apparently inserted into an anticyclonic cell that occupies a good portion of the eastern TYS. This cell persists in August and September, developing some meanders, with associated weak cyclonic anomalies along the coast, and then breaks down at the end of the summer. We summarize as follows:

  • We have identified nine vortices in the eastern TYS that appear to be robust features of the circulation. Eight of them (four cyclone–anticyclone couples) are embedded into a meandering, cyclonic AW stream that occupies the eastern TYS in winter and part of the spring. This pattern is seen both on the SLA and ADT maps, even though the signatures of the cyclones offshore are different, with numbers 5 and 8 appearing more clearly in the former, while numbers 6 and 7 are better defined in the latter, because they also appear in the MDT reconstruction.

  • At the end of spring, a reorganization of the flow takes place, with some of the winter structures surviving, even if displaced. The reorganization is likely guided by the inversion of the wind stress curl occurring in that period that leads to the formation of an anticyclonic cell occupying the entire eastern TYS, which apparently encompasses at least two strong anticyclones.

d. Mean and eddy kinetic energy

We have also computed the winter and summer mean and eddy kinetic energy averages (MKE and EKE, respectively) from the geostrophic fields associated to the altimeter maps. Results are not shown, because they are similar to those obtained by Rinaldi et al. (2010) for a shorter period (2001–04). In winter, a large part of the MKE is along the AW stream and in the adjacent vortices, while in summer most of the MKE resides in the structures present in the western part of the basin, which intensify, and in the two anticyclonic vortices off Naples and to the north of Sicily. On the other hand, the highest values of EKE are found along the AW stream and in the northern part of the basin in winter and are concentrated in a wide area around the Bonifacio anticyclone in summer, with another sharp maximum in the area of the Calabrian anticyclone. Overall, the eddy energy is higher in summer. This is both due to an underestimation of the winter EKE (the dynamics induced by the strong, barotropic high-frequency components present in the winter winds cannot be resolved by the altimeter) and to the fact that in summer the weakening of the wind forcing and the relative isolation allow the basin to express more freely its internal dynamics in response to local forcing.

It is interesting to look at the energy exchanges between eddies and mean flow occurring in the different seasons. To do that, we restrict our focus to 2003–05, the period with the best satellite coverage. Figure 6 shows maps of the energy exchange term for February and September, with the corresponding fields superposed. Here, is the 2003–05 winter or summer average of the geostrophic velocity obtained from the ADT and is the geostrophic velocity anomaly associated to the SLA and the average (angle brackets) is over the corresponding months. Negative values of the exchange term, indicating that eddies extract energy from the mean flow, are found all along the AW stream in February and minima are associated with all four main anticyclonic meanders. This suggests the possibility, to be further explored, that the circulation pattern observed in late winter and in the first part of the spring represents a basinwide, coherent structure resulting from instability of the AW stream. In the September map, the main exchanges are seen to occur in the area occupied by the Bonifacio dipole.

Fig. 6.
Fig. 6.

Energy exchange term between mean flow and eddies for (left) February and (right) September (over the period 2003–05). Negative values indicate that eddies are extracting energy from the mean flow. In February, negative minima are distributed along the meandering AW stream, whereas the strongest signals in September are in the region between the Bonifacio cyclone and its anticyclonic companion. Note that the zero levels correspond to different colors in the two maps.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

3. Further analysis of the winter and summer circulation patterns

The previous analysis of altimeter data has allowed us to individuate interesting new patterns and structures in the winter and summer circulation of the TYS, but the evidence accumulated, while certainly suggestive, is not yet conclusive. The robustness of some of the vortices present in the winter–spring circulation pattern [the anticyclones (numbers 1–4)] may still be questioned, because they are partially located in a coastal band where the accuracy of the altimeter reconstruction is doubtful. Moreover, one would like to have a better understanding of how such a pattern can form. Likewise, it is not completely clear how many vortices are present in the eastern TYS in summer and how the circulation organizes around them. We have therefore sought additional evidence, both from independent observations and from modeling results.

a. The winter–spring case

We first looked at other remotely sensed fields, such as the sea surface temperature (SST) and the ocean color. We found very few SST images in the period of interest in which the vortices we are focusing on could be clearly individuated (an interesting example will be presented later on). This is both because of data voids due to clouds or other environmental factors and because of the very weak horizontal SST gradients (the range of variation of the SST is typically less than 1°C in March in the TYS), which make it difficult to separate dynamical signatures from the background noise.

Ocean color data appeared more promising, because in March and April there is a bloom of chlorophyll in the Mediterranean Sea. Therefore, we have examined full-resolution (1 km) Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua level-3 data, selecting as a tracer for the surface circulation the MODIS surface concentration of chlorophyll-a, computed using the Mediterranean Ocean Color 3 bands (MedOC3) algorithm of Santoleri et al. (2008). A particularly nice chlorophyll image is that of 5 March 2003, excerpts from which are presented in Fig. 7. In the center of this figure is the SLA map for the week starting on the same day, with the corresponding geostrophic circulation superposed [the SLA range is approximately from −8 (blue) to 6 cm (red), having subtracted the basin average]. The coastal anticyclones 1–4 are well defined in the SLA map, and can also be seen quite clearly in the zooms B, C, and D, extracted from the chlorophyll image. In zoom D, the signature of anticyclone number 1 is strong, and the quasi-triangular shape of the chlorophyll filament that defines its northern boundary is quite similar to that of the geostrophic current in the area, deduced from the SLA. The chlorophyll stream then moves eastward, making a cyclonic loop around 14°E, where the SLA map locates a deep cyclonic depression. Below the cyclonic area, there seems to be a small anticyclone, which is very near the coast and is apparently not resolved in the SLA map. Farther east, the signature of the wide anticyclone off the Calabrian coast is also clear and to the west of this structure there is a sharp cyclonic meander that is also present in the SLA map. Moving northward, in zoom C, we find clear evidence of two anticyclones, one off of Naples and the other just to the southeast, both in the region in which the SLA map places a positive anomaly. Farther north, in the lower part of zoom B, just above 41°N, and between 12° and 13°E, an anticyclonic structure is present, delimited to the east by a tongue that detaches from the coast, moves southward, and then turns toward the west. This appears to be a southern core in the wide anticyclonic area number 4, present in the SLA map. To the northwest of this structure, the signature of another anticyclone is present, whose eastern boundary can be seen quite clearly by following the path of the Tiber plume, which is first anticyclonic, when leaving the coast, and then makes a sharp cyclonic turn, indicating the encounter with a stream of water heading toward south. We note that the geostrophic flow deduced from the SLA appears consistent with a two-lobed structure of the anticyclone number 4. Finally, in zoom A, the signature of a small anticyclone (size of about ½°) is present exactly in the same place in which a sharp and localized positive anomaly appears in the SLA map. This is just an example supporting the conclusions of Pascual et al. (2007), that is, that relatively small mesoscale structures can be well resolved when four satellite tracks are available.

Fig. 7.
Fig. 7.

The central map shows the SLA of the week 5–11 Mar 2003 in the eastern TYS, with the corresponding geostrophic velocities superposed. The main structures present in the SLA map can also be individuated in the chlorophyll image of 5 Mar 2003 (see the zooms in the boxes A, B, C, and D). Details are given in the text.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

Further evidence of the presence of the coastal anticyclones, together with some indications on their genesis, is provided by Fig. 8, which makes use of the outputs of Tyrrhenian Regional Model (TYREM), a high-resolution operational model of the TYS circulation recently developed [see Napolitano et al. (2013) for a description of the model, and of the TYS surface and intermediate circulation of the year 2009]. The figure follows the evolution of the surface flow (10-m depth) in the eastern TYS, from 24 February to 10 March 2009. The evolution is quite dramatic and suggests instability of the AW stream with the production of anticyclones along the coast and of somewhat smaller cyclones on the offshore side of the stream. In the mature phase (10 March, and for a few weeks after) the main flow features are very similar to those of Fig. 7: note, for example, the two-lobed structure of anticyclones number 3 and 4, and the sharp, triangular meanders above the Sicilian coast. The last panel of Fig. 8 shows the altimetric SLA for the week from 4 to 10 March (indicated by contours), which is in very good agreement with the circulation map of 10 March. All the main structures are found in both maps and in approximately the same locations: the two anticyclonic cores above Sicily, centered at about 12.5° and 14°E, with a cyclone in between more offshore; the anticyclonic core off the Calabria coast with a cyclonic companion offshore; the anticyclonic area off Naples with two cores; the anticyclone–cyclone couple around 41°N, 12°E; and finally the anticyclone along the coast farther north. Along the Italian coast, the signature of some of these structures is also clearly seen in the SST field for the same week, shown by shading in the same panel, which has been obtained by combining MODIS data from the Terra and Aqua satellites, from all available night passes.

Fig. 8.
Fig. 8.

Outputs from a high-resolution operational model of the TYS circulation, illustrating the evolution of the surface flow (10-m depth) in the eastern TYS, from 24 Feb to 10 Mar 2009. The bottom right panel shows the SLA for the week 4–10 Mar 2009 (contours) with the SST for the same week superposed (colors). The SST image has been obtained by combining MODIS data from the Terra and Aqua satellites, from all night passes available for the week in consideration.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

Playing the devil's advocate, one could say that, although the results from the operational model are suggestive, it is still not totally clear whether the evolution of the circulation results from a global instability of the AW stream, or reflects different, more local dynamics, eventually constrained by the fact that altimeter tracks are assimilated in the “father” model that provides initial and boundary conditions to TYREM. To further explore the instability hypothesis, we have performed an idealized experiment with the Princeton Ocean Model (POM; Mellor 2004), with the same grid—25 sigma levels and 4-km horizontal grid spacing—and realistic bathymetry used for the experiments discussed in Vetrano et al. (2010). There is no surface forcing, and we start with zero velocity and typical winter stratification, with no horizontal gradients. At the open boundaries, the net barotropic transports are prescribed [+0.7 Sv (1 Sv ≡ 106 m3 s−1) at the Sardinia Channel, −0.7 Sv at the Corsica Channel, and 0 Sv at the Sicily Strait], together with realistic temperature and salinity sections. Thus, the simulation describes the baroclinic adjustment of the flow to the boundary conditions and to the bathymetry. Figure 9 shows the surface (10-m depth) flow patterns produced after 10, 80, 110, and 150 days. After 10 days, a cyclonic stream along the Italian coast has already formed, driven by the density gradient at the Sardinia section and by the imposed transports, and after 80 days the stream starts to develop meanders all along, which strengthen in the later panels. By day 150, several well-defined vortices have formed, anticyclonic near shore and cyclonic off shore of the stream, in most of the places in which the SLA anomalies that we have analyzed are located: anticyclones number 1–4 are present as is cyclone number 5. On the other hand, the Bonifacio couple and the cyclonic region between Sardinia and Sicily have not developed, and this just confirms that wind forcing plays a crucial role in the formation of these structures. It may be noted how closely the circulation pattern of day 150 in the eastern TYS resembles those in the March panels of Fig. 8. The results of this simulation definitely support the idea that the winter–spring circulation pattern we have analyzed results from instability of the AW stream circling around the Italian coasts.

Fig. 9.
Fig. 9.

Surface velocity fields at day (top left) 10, (top right) 80, (bottom left) 110, and (bottom right) 150 of a winter simulation of TYS dynamics made with POM in the absence of surface forcing (see details in the text). The simulation shows that the baroclinic adjustment with the boundary conditions and with the topography is sufficient to produce, in a few months, a circulation pattern in the southern and eastern TYS that is quite similar to that obtained from the analysis of the altimeter data.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

b. The summer case

It is more difficult to find remotely sensed fields providing useful images of the region of interest in summer, because the chlorophyll signal is very weak and the strong heat fluxes tend to obscure the signatures of the circulation in the SST maps. Therefore, we shall only rely on the outputs of the operational model to get further information on the circulation. The 2009 summer average surface circulation (10-m depth) produced by TYREM is shown in Fig. 10b, while Fig. 10a displays the 1993–2010 average of the summer ADT, with the corresponding geostrophic circulation superposed. The comparison between the two is very good, because most of the main vortices are present in both, even if sometimes slightly displaced. The signature of the Sicilian anticyclone is strong in both panels, while the Calabrian one is much more developed in the model, and displaced toward west, as the SLA maps of Figs. 4 and 5 would have suggested. In both panels there is a wide anticyclonic area in front of the Campania coasts, which in the model output is two lobed and reaches deeper inside the basin. To the southwest of this structure, the anticyclonic structure 9 is present in both panels, consistent with the SLA summer maps. Cyclone 5 is clearly visible, and another cyclone is to the east, in the area in which the MDT also displays a cyclonic signal.

Fig. 10.
Fig. 10.

Summer circulation in the eastern TYS: (a) 1993–2010 summer average ADT, with a geostrophic reconstruction of the circulation superposed; (b) the 2009 summer average of the surface (10-m depth) velocity from TYREM. Main vortices are present in both panels, but the model results highlight the presence of a wide anticyclonic circulation cell in the area.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

Finally, we note that, consistent with what we had deduced form the SLA maps, a wide anticyclonic cell occupies most of the eastern TYS, encompassing most of the vortices mentioned, except the cyclone centered at 39.5°N, 14.6°E. We have previously observed that the presence of such a cell could be associated with a change of forcing over the area. Evidence supporting this idea is given in Fig. 11, showing the average wind curl over the area (38°–40°N, 12°–15.5°E) from the ERA-Interim dataset, which regularly oscillates between positive values in winter and negative ones in summer.

Fig. 11.
Fig. 11.

Time series of average surface wind curl (s−1) from the ERA-Interim dataset (the average is on the box 38°–40°N, 12°–15.5°E; drawn in the August panel of Fig. 2). The wind curl average oscillates between positive values in winter and negative ones in summer.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

4. Beyond seasonal variability

Based on what is known about the Mediterranean Sea, the seasonal cycle may be expected to represent the dominant mode of variability of the TYS circulation. Having a pretty long time series at hand, we can quantitatively assess the relevance of this contribution with respect to the variability on other temporal scales. To do so, we have applied the multichannel singular spectrum analysis (M-SSA) to the time series of weekly SLA maps.

We recall that the SSA is a nonparametric spectral estimation method for extracting information from short and noisy time series that was developed in the mid-1980s (Broomhead and King 1986; Fraedrich 1986; Vautard and Ghil 1989), building on important progress in the theory of dynamical systems and the analysis of nonlinear time series made in the two preceding decades [designated as the “nonlinear revolution” in the nice review paper by Ghil et al. (2002), to which we refer for a detailed exposition of the SSA and M-SSA methods].

The basic idea was that, under certain conditions, a single time series could be used to reconstruct the attractor of a forced dissipative system, or, at least a “skeleton” of this attractor formed by a few robust periodic orbits. This can be done by embedding the discrete time series of length N into a vector space of dimension MN/2 generated by taking N′ lagged copies of the original data (N= N − M + 1). Then, with appropriate filters, one can separate the components that are statistically independent, at zero lag, in this augmented vector space, which consist of trends, oscillatory patterns—in which we are particularly interested—and noise. Details of how this is done in practice can be found in Ghil et al. (2002), but the procedure basically involves the construction of a symmetric M × M covariance matrix and the computation of its eigenvalues and eigenvectors, which can be efficiently performed using the singular value decomposition. The eigenvectors are designated as empirical orthogonal functions (EOF), by analogy with the meteorological literature, and each eigenvalue accounts for the partial variance in the direction of the associated eigenvector (the sum of the eigenvalues gives the total variance of the original time series).

In our case, the situation is more complicated, because we do not have a single time series of SLA values, but as many series as the number of sea points on which the SLA is available. As a consequence, a multivariate approach is required, the M-SSA, which is also quite standard, but involves some technical complications associated with the dimensions of the computational problem. These complications can be handled using the “reduced” approach explained in the appendix A2 of Ghil et al. (2002).

Because N = 936 weeks (18 years), we have initially chosen the widest possible time window, given by M = N/2 = 468, which could enable us to capture oscillations with periods up to 9 years. The left panel of Fig. 12 shows the resulting spectrum, where a couple of eigenvalues with frequency of about 1/52 cycles per week stand out, representing the only significant mode for periods longer than five weeks. This annual pair (EOFs 1–2) captures about 58% of the variance of the input data and accounts both for the steric oscillation and for the seasonal signal due to the variability of the circulation. To get rid of the steric signal (or at least a large part of it), we have performed a second identical test, having first subtracted the basin average of the SLA from each input map. In the new test, the two leading annual EOFs explain 20% of the total variance, and two new significant EOFs appear in the SSA spectrum, with periods of about 18 and 6 years, explaining together 4.2% of the variance. The first of these modes can be ascribed to a trend or to some oscillation that cannot be resolved given the length of our time series. The other mode is interesting, however, and we have made another test, setting M = 312 weeks (6 years), to increase the statistical significance of this component, while keeping a window large enough to resolve it. This gives the singular spectrum shown in the right panel of Fig. 12. The leading pair of modes now explains about 21% of the total variance, and the corresponding EOFs are shown Fig. 13a. They are in quadrature and correspond to two eigenvalues that are very close (in the peak at 0.019 cycles per week in Fig. 12 they are nearly indistinguishable). This means that two eigenvalues correspond to a not necessarily harmonic but certainly periodic oscillation; as suggested by Vautard and Ghil (1989), such a pair can be interpreted as the nonlinear equivalent of a pair of sine and cosine in a Fourier analysis for linear problems.

Fig. 12.
Fig. 12.

Singular spectra of the SLA map time series. Black points connected by blue lines indicate the eigenvalue amplitude. The lower and upper ticks on the error bar (red) indicate the 5th and 95th (noise) percentile from the χ2 test. For each EOF, a characteristic frequency was estimated by maximizing its correlation with a sinusoid. In (a), the whole SLA was used and M = 468 weeks for the SSA analysis. In (b), the basin average was first subtracted from the SLA maps and M = 312 weeks. The inset in (b) is a zoom around the eigenvalues corresponding to EOFs 3 and 4.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

Fig. 13.
Fig. 13.

Principal EOFs of the SLA time series after removing the basin average (and hence most of the steric effect). EOFs 1 and 3 (2 and 4) are indicated by solid (dashed) lines.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

A significant couple of smaller eigenvalues is now present (see the inset in Fig. 12). The corresponding EOFs (3–4), shown in Fig. 13b, are also in quadrature and explain about 6% of the variance. This pair is associated to a period of about 6 years and represents the signature of the interannual variability.

In the range of periods between approximately 3 and 7 weeks, several eigenvalues pass the significance test, indicating the presence of an intraseasonal component, which may be associated with the variability of some of the circulation structures or to short-living eddy features. On the other hand, for periods below three weeks the spectrum of Fig. 12 collapses below the noise line. This was to be expected, because the fastest repetition cycle of the satellite used to produce the SLA maps is about 10 days. Summarizing, after removing the spatial averages, 21% of the variance of the time series is given by the annual cycle of the circulation, 6% is due to the interannual variability, and 20% is due to the other high-frequency significant components, while the remaining 53% is noise.

Figure 14 shows the time evolution of the Tyrrhenian annual mode reconstructed using EOFs 1 and 2. The patterns in this figure are quite similar to those previously found in the monthly averages of the SLA of Fig. 5. Together with the consistent portion of variance explained (almost half of that associated to significant components), this provides a clear indication of the robustness of the seasonal cycle as we have described it. Yet, looking at the ADT maps, interesting deviations can sometimes be noted, which should be further investigated, because they could provide additional information on the mechanisms driving the circulation.

Fig. 14.
Fig. 14.

Reconstruction of the annual cycle as given by EOFs 1 and 2 with M = 312 weeks. Contour intervals of 0.5 cm. The thick line represents the zero level.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

The time evolution of the low-frequency mode, reconstructed using EOFs 3 and 4, is shown in Fig. 15. The sequence of the 18 yearly maps illustrates the progression of an oscillation with a period of about 6 years, mainly marked by the evolution of the strength of the Bonifacio anticyclone (between 40° and 41°N and 10° and 12°E). The intensity of this anticyclone displays maxima in 1995, 2002, and 2009, and minima in 1998 and 2004. Cyclonic signals are present both to the north and to the south of the anticyclone, with variable intensities, dimensions and shapes in the different years. The amplitude of the mode appears weaker in the eastern part of the basin.

Fig. 15.
Fig. 15.

Time evolution of the low-frequency mode reconstructed using EOFs 3 and 4.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

To give a quantitative measure of the strength of the oscillation in different parts of the basin, we computed the kinetic energy of the geostrophic flow associated to the SLA of Fig. 15, and averaged it in 2° × 2° boxes. The resulting kinetic energy time series in the seven boxes covering the basin is shown in Fig. 16. Looking at the scales in these plots, reaching 12 cm2 s−2 for the box including the Bonifacio dipole, and only 4 cm2 s−2 for the other six boxes, is sufficient to see that most of the energy associated to the low-frequency mode in consideration resides in the Bonifacio system. In the associated plot, three nearly complete oscillations are observed, with amplitudes approximately three times larger than those observed in the other boxes. Three oscillations are also clearly seen in the two boxes below, while the signals in the eastern boxes are more irregular. This indicates that the low-frequency mode is mostly localized in the western part of the basin.

Fig. 16.
Fig. 16.

Time series of the kinetic energy of the geostrophic flow associated to the 6 yr mode of variability of the SLA. The kinetic energy has been averaged over seven 2° × 2° boxes, covering the TYS. The signature of the mode is clear in the three western boxes, and the energy is three times larger in the northernmost of them, in which the Bonifacio couple of gyres resides.

Citation: Journal of Physical Oceanography 43, 8; 10.1175/JPO-D-12-0112.1

5. Summary and discussion

We have presented a novel assessment of the TYS surface dynamics based on the analysis of 18 years of AVISO altimeter data, focusing on the seasonal variability of the circulation. The main results can be summarized as follows.

  1. While the dynamics in the western TYS are as expected, with the Bonifacio dipole in the northern part and a wide cyclonic recirculation to the south, in late winter and part of the spring the average circulation in the eastern TYS is found to consist of a stream of AW that, while progressing cyclonically toward the north, forms wide meanders around several (at least four) anticyclones located along the Italian coast. The signatures of these vortices are found in the maps of average SLA, and for two of them, to the north of Sicily, and off Naples, in the MDT reconstruction of Rio et al. (2007). We have also found evidence of the presence of these vortices in maps of chlorophyll and of SST, and the same vortices, with cyclonic companions offshore, appear in the outputs of the operational model TYREM. All this combined evidence allows us to state that these vortices represent robust features of the winter–spring TYS circulation. The anticyclones have cyclonic companions offshore, nested in the cyclonic meanders, two of which (5 and 8) are better defined in the average SLA maps, while the others appear more clearly in those of the ADT, because they are present in the MDT reconstruction. The latter, however, can also be found in SLA maps of individual years, such as those presented in Figs. 7 and 8. Because the structures on the nearshore (offshore) side of the meandering stream are anticyclonic (cyclonic), they cannot be vortices shed from the meanders, as it happens in other well-known meandering streams (think, for example, of the ring formation in the Gulf Stream); this kind of process, in our geometry, would tend to produce cyclones near coast, and anticyclones offshore. Instead, the overall structure suggests instability of the AW stream. The evolution of the monthly SLA maps of Fig. 5, the February energy exchanges of Fig. 6, and the numerical results of Figs. 8 and 9, support this idea.

  2. The average summer circulation is different. There is no clear flow of AW into the basin and a very weak flow appears to leave from the Corsica Channel, supporting the common picture of the TYS as an isolated basin in summer [see, e.g., the review by Astraldi and Gasparini (1994)]. The circulation in the northern part of the basin is dominated by the Bonifacio dipole, which extends zonally, almost reaching the Italian coast, and intensifies. However, the SLA also indicates the presence of an anticyclonic cell in the central and eastern part of the basin, which has also been found in the summer 2009 outputs of the operational model TYREM. The cell encompasses two wide anticyclonic structures, in the areas in which the Vavilov and Marsili Seamounts are located. These structures appear to be the result of the evolution of the anticyclones present in winter off Naples and off the Calabrian coast, after the breakdown of the cyclonic AW cell and the inversion of the wind stress curl over the region.

  3. Although there is considerable variability in the SLA signal from year to year, the previously mentioned vortices are recurrent features of the circulation. Yet, not all of them appear clearly in the MDT of Rio et al. (2007), and the persistent cyclone number 5 we have identified to the north of Sicily is absent in that reconstruction. This is something to be further investigated. We believe that the main reasons for that lie in the difficulty of the reconstruction in the eastern part of the basin, where the average circulation switches sign from summer to winter, and in the nonnegligible changes introduced by the major reprocessing the AVISO dataset underwent in 2010. Perhaps, a new effort should be made to improve our knowledge of the MDT in the TYS.

  4. The winter–spring circulation pattern appears to be the result of basin-scale instability of the AW stream. This is intriguing, even if not too surprising, because the AW stream is known to develop instabilities both along the African coast (Millot 1985) and in the Ligurian Sea (André et al. 2009). Similar flow patterns have also been found in numerical and laboratory experiments, both in barotropic and baroclinic contexts (see, among others, Poulin and Flierl 2005; Jacobs et al. 1999; Flexas et al. 2004). The idealized, unforced simulation of section 3 (Fig. 9) shows that this kind of pattern can indeed result from the baroclinic adjustment of the AW stream to the basin topography, but a thorough understanding of the processes taking place in the TYS will require the inclusion of realistic forcing, and dedicated numerical work.

Our results highlight the importance of the AW inflow in the south in determining the TYS dynamics. A key point is that where the Algerian current, flowing along the shallow and narrow Skerki Bank, is “shot” into the deep basin. This is a point where the AW flow bifurcates in winter (see Fig. 4), with a branch exiting through the Sicily Channel, while the other enters the TYS bordering the Sicilian coast (see, e.g., Vetrano et al. 2004). This could also be a place where instability may be triggered by potential vorticity variations due to the strongly varying bathymetry. Although we have not shown it here, it is also a point where in early summer the AW stream often veers toward north, bordering the cyclonic structure present in the area. This calls for a detailed analysis of the dynamics in the region, in the different seasons, and particularly during the transitions between the winter and summer circulation regimes.

There are other directions for future work. Here, we only started to explore the interannual variability. The existence of significant modes from the SSA analysis with periods of about six years is interesting, because variability at close time scales results from the analysis of tide gauge station data, both in the Mediterranean Sea [see Fig. 8c of Jevrejeva et al. (2006)] and in the North Atlantic (Unal and Ghil 1995). The correlation between these modes of variability of the TYS sea level and the local and/or remote forcings acting on the basin should be further explored.

Finally, it can be said that the present work is another demonstration of the wealth of information stored in the last two decades of satellite observations, despite the limitations that have been reminded all along. Clearly, removing these limitations will be important. One may hope that new missions that are being planned, such as the Surface Water and Ocean Topography (SWOT) mission (see the website swot.jpl.nasa.gov), with its ambitious objective of characterizing the ocean meso- and submesoscale circulation at a spatial resolution of a few kilometers, will open new perspectives in the investigation of the ocean dynamics, both at a global scale and in regional and coastal applications.

Acknowledgments

It is a pleasure to acknowledge the expert and patient assistance by the handling Editor, Dr. William Kessler, as well as the useful comments and suggestions by two anonymous reviewers, which helped us to improve the paper. The altimeter product were produced by Ssalto/Duacs and distributed by AVISO, with support from CNES (the French Space Agency). Ocean color maps were generated using MyOcean products provided by the Mediterranean Ocean Colour Thematic Assembling Centre. This work was partly funded by the RITMARE project.

REFERENCES

  • Amitai, Y., Y. Lehahn, A. Lazar, and E. Heifetz, 2010: Surface circulation of the eastern Mediterranean Levantine basin: Insights from analysing 14 years of satellite altimetry data. J. Geophys. Res.,115, C10058, doi:10.1029/2010JC006147.

  • André, G., P. Garreau, and P. Fraunie, 2009: Mesoscale slope current variability in the Gulf of Lions. Interpretation of in situ measurements using a three-dimensional model. Cont. Shelf Res., 29, 407423.

    • Search Google Scholar
    • Export Citation
  • Artale, V., M. Astraldi, G. Buffoni, and G. P. Gasparini, 1994: Seasonal variability of gyre-scale circulation in the Northern Tyrrhenian Sea. J. Geophys. Res., 99 (C7), 14 12714 137.

    • Search Google Scholar
    • Export Citation
  • Astraldi, M., and G. P. Gasparini, 1994: The seasonal characteristics of the circulation in the Tyrrhenian Sea. Seasonal and Interannual Variability of the Western Mediterranean Sea, Coastal and Estuarine Studies, Geophys. Monogr., Vol. 46, Amer. Geophys. Union, 115–134.

  • Astraldi, M., and Coauthors, 1999: The role of straits and channels in understanding the characteristics of Mediterranean circulation. Prog. Oceanogr., 44, 65108.

    • Search Google Scholar
    • Export Citation
  • Béranger, K., L. Mortier, and M. Crépon, 2005: Seasonal variability of water transport through the Straits of Gibraltar, Sicily and Corsica, derived from a high-resolution model of the Mediterranean circulation. Prog. Oceanogr., 66, 341364.

    • Search Google Scholar
    • Export Citation
  • Broomhead, D. S., and G. P. King, 1986: Extracting qualitative dynamics from experimental data. Physica D, 20, 217236.

  • Budillon, G., G. P. Gasparini, and K. Schroeder, 2009: Persistence of an eddy signature in the central Tyrrhenian basin. Deep-Sea Res. II, 56, 713724.

    • Search Google Scholar
    • Export Citation
  • Dussurget, R., F. Birol, R. Morrow, and P. De Mey, 2011: Fine resolution altimetry data for a regional application in the Bay of Biscay. Mar. Geod., 34, 447476.

    • Search Google Scholar
    • Export Citation
  • Flexas, M. M., G. J. F. van Heijst, G. Jordà, and A. Sanchez-Arcilla, 2004: Numerical simulation of barotropic jets over a sloping bottom: Comparison to a laboratory model of the Northern Current. J. Geophys. Res.,109, C12039, doi:10.1029/2004JC002286.

  • Fraedrich, K., 1986: Estimating the dimension of weather and climate attractors. J. Atmos. Sci., 43, 419432.

  • Ghil M., and Coauthors, 2002: Advanced spectral methods for climatic time series. Rev. Geophys.,40, 1003, doi:10.1029/2000RG000092.

  • Ioannone, A., A. Catucci, M. Grasso, and G. L. Eusebi Borzelli, 2011: Decadal variability and scales of the sea surface structure in the northern Ionian. Cont. Shelf Res., 31, 3746.

    • Search Google Scholar
    • Export Citation
  • Jacobs, P., Y. Guo, and P. A. Davies, 1999: Boundary currents over shelf and slope topography. J. Mar. Syst., 19, 137158.

  • Jevrejeva, S., A. Grinsted, J. C. Moore, and S. Holgate, 2006: Nonlinear trends and multiyear cycles in sea level records. J. Geophys. Res.,111, C09012, doi:10.1029/2005JC003229.

  • Jordi, A., and D.-P. Wang, 2009: Mean dynamic topography and eddy kinetic energy in the Mediterranean Sea: Comparison between altimetry and a 1/16 degree ocean circulation model. Ocean Modell., 29, 137146.

    • Search Google Scholar
    • Export Citation
  • Korres, G., N. Pinardi, and A. Lascaratos, 2000: The ocean response to low-frequency interannual atmospheric variability in the Mediterranean Sea. Part I: Sensitivity experiments and energy analysis. J. Climate, 13, 705731.

    • Search Google Scholar
    • Export Citation
  • Krivosheya, V. G., 1983: Water circulation and structure in the Tyrrhenian Sea. Oceanology (Moscow), 23, 166171.

  • Marullo, S., R. Santoleri, and F. Bignami, 1994: The surface characteristics of the Tyrrhenian Sea: Historical satellite data analysis. Seasonal and Interannual Variability of the Western Mediterranean Sea, Coastal and Estuarine Studies, Geophys. Monogr., Vol. 46, Amer. Geophys. Union, 135–154.

  • MEDOC Group, 1970: Observation of formation of deep water in the Mediterranean Sea, 1969. Nature, 227, 10371040.

  • Melet, A., L. Gourdeau, and J. Verron, 2010: Variability in Solomon Sea circulation derived from altimeter data. Ocean Dyn., 60, 883900.

    • Search Google Scholar
    • Export Citation
  • Mellor, G. L., 2004: User's guide for a three-dimensional, primitive equation numerical ocean model. Program in Atmospheric and Ocean Science Rep., 11–35.

  • Millot, C., 1985: Some features of the Algerian Current. J. Geophys. Res., 90, 71697175.

  • Millot, C., 1999: Circulation in the western Mediterranean Sea. J. Mar. Syst., 20, 423442.

  • Napolitano, E., R. Iacono, and S. Marullo, 2013: The 2009 surface and intermediate circulation of the Tyrrhenian Sea as assessed by an operational model. The Mediterranean Sea: Temporal Variability and Spatial Patterns, G. Borzelli et al., Eds., Amer. Geophys. Union, in press.

  • Pascual, A., M. I. Pujol, G. Larnicol, P. Y. Le Traon, and M. H. Rio, 2007: Mesoscale mapping capabilities of multisatellite altimeter missions: First results with real data in the Mediterranean Sea. J. Mar. Syst., 18, 161178.

    • Search Google Scholar
    • Export Citation
  • Pierini, S., and A. Simioli, 1998: A wind-driven circulation model for the Tyrrhenian sea area. J. Mar. Syst., 18, 161178.

  • Poulain, P. M., M. Menna, and E. Mauri, 2012: Surface geostrophic circulation of the Mediterranean Sea derived from drifters and satellite altimeter data. J. Phys. Oceanogr., 42, 973990.

    • Search Google Scholar
    • Export Citation
  • Poulin, F. J., and G. R. Flierl, 2005: The influence of topography on the stability of jets. J. Phys. Oceanogr., 35, 811825.

  • Rinaldi, E., B. Buongiorno Nardelli, E. Zambianchi, R. Santoleri, and P. M. Poulain, 2010: Lagrangian and Eulerian observations of the surface circulation in the Tyrrhenian Sea. J. Geophys. Res., 115, C04024, doi:10.1029/2009JC005535.

    • Search Google Scholar
    • Export Citation
  • Rio, M. H., P. M. Poulain, A. Pascual, E. Mauri, G. Larnicol, and R. Santoleri, 2007: A mean dynamic topography of the Mediterranean Sea computed from altimetric data, in-situ measurements and a general circulation model. J. Mar. Syst., 65, 484508.

    • Search Google Scholar
    • Export Citation
  • Roussenov, V., E. Stanev, V. Artale, and N. Pinardi, 1995: A seasonal model of the Mediterranean Sea general circulation. J. Geophys. Res., 100, 13 51513 538.

    • Search Google Scholar
    • Export Citation
  • Santoleri, R., G. Volpe, S. Marullo, and B. Buongiorno Nardelli, 2008: Open waters optical remote sensing of the Mediterranean Sea. Remote Sensing of the European Seas, V. Barale and M. Gade, Eds., Springer, 103–116.

  • Tonani, M., N. Pinardi, S. Dobricic, I. Pujol, and C. Fratianni, 2008: A high-resolution free surface model of the Mediterranean Sea. Ocean Sci., 4, 114.

    • Search Google Scholar
    • Export Citation
  • Unal, Y. S., and M. Ghil, 1995: Interannual and interdecadal oscillation patterns in sea level. Climate Dyn., 11, 255278.

  • Vautard, R., and M. Ghil, 1989: Singular spectrum analysis in nonlinear dynamics, with applications to paleoclimatic time series. Physica D, 35, 395424.

    • Search Google Scholar
    • Export Citation
  • Vetrano, A., G. P. Gasparini, R. Molcard, and M. Astraldi, 2004: Water flux estimates in the central Mediterranean Sea from an inverse box model. J. Geophys. Res.,109, C01019, doi:10.1029/2003JC001903.

  • Vetrano, A., E. Napolitano, R. Iacono, K. Schroeder, and G. P. Gasparini, 2010: Tyrrhenian Sea circulation and water mass fluxes in spring 2004: Observations and model results. J. Geophys. Res.,115, C06023, doi:10.1029/2009JC005680.

  • Vignudelli, S., P. Cipollini, L. Roblou, F. Lyard, G. P. Gasparini, G. Manzella, and M. Astraldi, 2005: Improved satellite altimetry in coastal systems: Case study of the Corsica Channel (Mediterranean Sea). Geophys. Res. Lett., 32, L07608, doi:10.1029/2005GL022602.

    • Search Google Scholar
    • Export Citation
  • Wunsch, C., 1997: The vertical partition of oceanic horizontal kinetic energy. J. Phys. Oceanogr., 27, 17701794.

  • Zavatarelli, M., and G. L. Mellor, 1995: A numerical study of the Mediterranean Sea circulation. J. Phys. Oceanogr., 25, 13841414.

1

An anticyclone to the north of Sicily only appears in the summer surface circulation map of Krivosheya (1983).

Save
  • Amitai, Y., Y. Lehahn, A. Lazar, and E. Heifetz, 2010: Surface circulation of the eastern Mediterranean Levantine basin: Insights from analysing 14 years of satellite altimetry data. J. Geophys. Res.,115, C10058, doi:10.1029/2010JC006147.

  • André, G., P. Garreau, and P. Fraunie, 2009: Mesoscale slope current variability in the Gulf of Lions. Interpretation of in situ measurements using a three-dimensional model. Cont. Shelf Res., 29, 407423.

    • Search Google Scholar
    • Export Citation
  • Artale, V., M. Astraldi, G. Buffoni, and G. P. Gasparini, 1994: Seasonal variability of gyre-scale circulation in the Northern Tyrrhenian Sea. J. Geophys. Res., 99 (C7), 14 12714 137.

    • Search Google Scholar
    • Export Citation
  • Astraldi, M., and G. P. Gasparini, 1994: The seasonal characteristics of the circulation in the Tyrrhenian Sea. Seasonal and Interannual Variability of the Western Mediterranean Sea, Coastal and Estuarine Studies, Geophys. Monogr., Vol. 46, Amer. Geophys. Union, 115–134.

  • Astraldi, M., and Coauthors, 1999: The role of straits and channels in understanding the characteristics of Mediterranean circulation. Prog. Oceanogr., 44, 65108.

    • Search Google Scholar
    • Export Citation
  • Béranger, K., L. Mortier, and M. Crépon, 2005: Seasonal variability of water transport through the Straits of Gibraltar, Sicily and Corsica, derived from a high-resolution model of the Mediterranean circulation. Prog. Oceanogr., 66, 341364.

    • Search Google Scholar
    • Export Citation
  • Broomhead, D. S., and G. P. King, 1986: Extracting qualitative dynamics from experimental data. Physica D, 20, 217236.

  • Budillon, G., G. P. Gasparini, and K. Schroeder, 2009: Persistence of an eddy signature in the central Tyrrhenian basin. Deep-Sea Res. II, 56, 713724.

    • Search Google Scholar
    • Export Citation
  • Dussurget, R., F. Birol, R. Morrow, and P. De Mey, 2011: Fine resolution altimetry data for a regional application in the Bay of Biscay. Mar. Geod., 34, 447476.

    • Search Google Scholar
    • Export Citation
  • Flexas, M. M., G. J. F. van Heijst, G. Jordà, and A. Sanchez-Arcilla, 2004: Numerical simulation of barotropic jets over a sloping bottom: Comparison to a laboratory model of the Northern Current. J. Geophys. Res.,109, C12039, doi:10.1029/2004JC002286.

  • Fraedrich, K., 1986: Estimating the dimension of weather and climate attractors. J. Atmos. Sci., 43, 419432.

  • Ghil M., and Coauthors, 2002: Advanced spectral methods for climatic time series. Rev. Geophys.,40, 1003, doi:10.1029/2000RG000092.

  • Ioannone, A., A. Catucci, M. Grasso, and G. L. Eusebi Borzelli, 2011: Decadal variability and scales of the sea surface structure in the northern Ionian. Cont. Shelf Res., 31, 3746.

    • Search Google Scholar
    • Export Citation
  • Jacobs, P., Y. Guo, and P. A. Davies, 1999: Boundary currents over shelf and slope topography. J. Mar. Syst., 19, 137158.

  • Jevrejeva, S., A. Grinsted, J. C. Moore, and S. Holgate, 2006: Nonlinear trends and multiyear cycles in sea level records. J. Geophys. Res.,111, C09012, doi:10.1029/2005JC003229.

  • Jordi, A., and D.-P. Wang, 2009: Mean dynamic topography and eddy kinetic energy in the Mediterranean Sea: Comparison between altimetry and a 1/16 degree ocean circulation model. Ocean Modell., 29, 137146.

    • Search Google Scholar
    • Export Citation
  • Korres, G., N. Pinardi, and A. Lascaratos, 2000: The ocean response to low-frequency interannual atmospheric variability in the Mediterranean Sea. Part I: Sensitivity experiments and energy analysis. J. Climate, 13, 705731.

    • Search Google Scholar
    • Export Citation
  • Krivosheya, V. G., 1983: Water circulation and structure in the Tyrrhenian Sea. Oceanology (Moscow), 23, 166171.

  • Marullo, S., R. Santoleri, and F. Bignami, 1994: The surface characteristics of the Tyrrhenian Sea: Historical satellite data analysis. Seasonal and Interannual Variability of the Western Mediterranean Sea, Coastal and Estuarine Studies, Geophys. Monogr., Vol. 46, Amer. Geophys. Union, 135–154.

  • MEDOC Group, 1970: Observation of formation of deep water in the Mediterranean Sea, 1969. Nature, 227, 10371040.

  • Melet, A., L. Gourdeau, and J. Verron, 2010: Variability in Solomon Sea circulation derived from altimeter data. Ocean Dyn., 60, 883900.

    • Search Google Scholar
    • Export Citation
  • Mellor, G. L., 2004: User's guide for a three-dimensional, primitive equation numerical ocean model. Program in Atmospheric and Ocean Science Rep., 11–35.

  • Millot, C., 1985: Some features of the Algerian Current. J. Geophys. Res., 90, 71697175.

  • Millot, C., 1999: Circulation in the western Mediterranean Sea. J. Mar. Syst., 20, 423442.

  • Napolitano, E., R. Iacono, and S. Marullo, 2013: The 2009 surface and intermediate circulation of the Tyrrhenian Sea as assessed by an operational model. The Mediterranean Sea: Temporal Variability and Spatial Patterns, G. Borzelli et al., Eds., Amer. Geophys. Union, in press.

  • Pascual, A., M. I. Pujol, G. Larnicol, P. Y. Le Traon, and M. H. Rio, 2007: Mesoscale mapping capabilities of multisatellite altimeter missions: First results with real data in the Mediterranean Sea. J. Mar. Syst., 18, 161178.

    • Search Google Scholar
    • Export Citation
  • Pierini, S., and A. Simioli, 1998: A wind-driven circulation model for the Tyrrhenian sea area. J. Mar. Syst., 18, 161178.

  • Poulain, P. M., M. Menna, and E. Mauri, 2012: Surface geostrophic circulation of the Mediterranean Sea derived from drifters and satellite altimeter data. J. Phys. Oceanogr., 42, 973990.

    • Search Google Scholar
    • Export Citation
  • Poulin, F. J., and G. R. Flierl, 2005: The influence of topography on the stability of jets. J. Phys. Oceanogr., 35, 811825.

  • Rinaldi, E., B. Buongiorno Nardelli, E. Zambianchi, R. Santoleri, and P. M. Poulain, 2010: Lagrangian and Eulerian observations of the surface circulation in the Tyrrhenian Sea. J. Geophys. Res., 115, C04024, doi:10.1029/2009JC005535.

    • Search Google Scholar
    • Export Citation
  • Rio, M. H., P. M. Poulain, A. Pascual, E. Mauri, G. Larnicol, and R. Santoleri, 2007: A mean dynamic topography of the Mediterranean Sea computed from altimetric data, in-situ measurements and a general circulation model. J. Mar. Syst., 65, 484508.

    • Search Google Scholar
    • Export Citation
  • Roussenov, V., E. Stanev, V. Artale, and N. Pinardi, 1995: A seasonal model of the Mediterranean Sea general circulation. J. Geophys. Res., 100, 13 51513 538.

    • Search Google Scholar
    • Export Citation
  • Santoleri, R., G. Volpe, S. Marullo, and B. Buongiorno Nardelli, 2008: Open waters optical remote sensing of the Mediterranean Sea. Remote Sensing of the European Seas, V. Barale and M. Gade, Eds., Springer, 103–116.

  • Tonani, M., N. Pinardi, S. Dobricic, I. Pujol, and C. Fratianni, 2008: A high-resolution free surface model of the Mediterranean Sea. Ocean Sci., 4, 114.

    • Search Google Scholar
    • Export Citation
  • Unal, Y. S., and M. Ghil, 1995: Interannual and interdecadal oscillation patterns in sea level. Climate Dyn., 11, 255278.

  • Vautard, R., and M. Ghil, 1989: Singular spectrum analysis in nonlinear dynamics, with applications to paleoclimatic time series. Physica D, 35, 395424.

    • Search Google Scholar
    • Export Citation
  • Vetrano, A., G. P. Gasparini, R. Molcard, and M. Astraldi, 2004: Water flux estimates in the central Mediterranean Sea from an inverse box model. J. Geophys. Res.,109, C01019, doi:10.1029/2003JC001903.

  • Vetrano, A., E. Napolitano, R. Iacono, K. Schroeder, and G. P. Gasparini, 2010: Tyrrhenian Sea circulation and water mass fluxes in spring 2004: Observations and model results. J. Geophys. Res.,115, C06023, doi:10.1029/2009JC005680.

  • Vignudelli, S., P. Cipollini, L. Roblou, F. Lyard, G. P. Gasparini, G. Manzella, and M. Astraldi, 2005: Improved satellite altimetry in coastal systems: Case study of the Corsica Channel (Mediterranean Sea). Geophys. Res. Lett., 32, L07608, doi:10.1029/2005GL022602.

    • Search Google Scholar
    • Export Citation
  • Wunsch, C., 1997: The vertical partition of oceanic horizontal kinetic energy. J. Phys. Oceanogr., 27, 17701794.

  • Zavatarelli, M., and G. L. Mellor, 1995: A numerical study of the Mediterranean Sea circulation. J. Phys. Oceanogr., 25, 13841414.

  • Fig. 1.

    The Tyrrhenian Sea, with its three openings: the Sardinia Channel, the Corsica Channel, and the Sicily Strait. Shading indicates bathymetry and arrows indicate the main patterns of surface (AW; red) and intermediate (LIW; yellow) waters. Ovals in the western part of the basin schematize the Bonifacio dipole and, below, the cyclonic circulation between Sardinia and Sicily—both quasi-permanent features of the TYS circulation.

  • Fig. 2.

    Relative vorticity of the wind at 1000 mb for (left) February and (right) August, from the ERA-Interim dataset (1990–2010 averages). Red boxes in the February and August maps delimit the regions on which the averages of Figs. 3 and 11 (both figures are described in greater detail below) are computed, respectively.

  • Fig. 3.

    Time series of average surface wind curl from the ERA-Interim dataset: (a) to the north of Sicily (38°–39°N, 13°–15.5°E) and (b) off the coast of Campania (39.5°–41°N, 12°–15°E). Boxes showing the regions over which the averages are performed are also indicated in Fig. 2 (left).

  • Fig. 4.

    Maps of average (1993–2010) (left) SLA and (right) ADT for (top) winter and (bottom) summer, with (middle) the MDT by Rio et al. (2007). A geostrophic reconstruction of the surface flow is superposed on the ADT maps. Contour intervals are 0.5 cm for the SLA and 0.75 cm for MDT.

  • Fig. 5.

    (a) Monthly SLA maps (November–April), averaged over 1993–2010, for the central and eastern TYS. Numbers from 1 to 8 in the March map denote anomalies corresponding to anticyclonic (1–4) and cyclonic (5–8) structures on the two sides of the AW stream that in winter circles cyclonically along the coasts of the basin. (b) As in (a), but for the months from May to October. The 2000- and 3000-m-depth contours have been superposed to the July map, to show that two of the main summer anticyclonic anomalies are near the Vavilov (39.85°N, 12.58°E) and the Marsili (39.28°N, 14.4°E) Seamounts, the two main submerged mountains of the basin.

  • Fig. 6.

    Energy exchange term between mean flow and eddies for (left) February and (right) September (over the period 2003–05). Negative values indicate that eddies are extracting energy from the mean flow. In February, negative minima are distributed along the meandering AW stream, whereas the strongest signals in September are in the region between the Bonifacio cyclone and its anticyclonic companion. Note that the zero levels correspond to different colors in the two maps.

  • Fig. 7.

    The central map shows the SLA of the week 5–11 Mar 2003 in the eastern TYS, with the corresponding geostrophic velocities superposed. The main structures present in the SLA map can also be individuated in the chlorophyll image of 5 Mar 2003 (see the zooms in the boxes A, B, C, and D). Details are given in the text.

  • Fig. 8.

    Outputs from a high-resolution operational model of the TYS circulation, illustrating the evolution of the surface flow (10-m depth) in the eastern TYS, from 24 Feb to 10 Mar 2009. The bottom right panel shows the SLA for the week 4–10 Mar 2009 (contours) with the SST for the same week superposed (colors). The SST image has been obtained by combining MODIS data from the Terra and Aqua satellites, from all night passes available for the week in consideration.

  • Fig. 9.

    Surface velocity fields at day (top left) 10, (top right) 80, (bottom left) 110, and (bottom right) 150 of a winter simulation of TYS dynamics made with POM in the absence of surface forcing (see details in the text). The simulation shows that the baroclinic adjustment with the boundary conditions and with the topography is sufficient to produce, in a few months, a circulation pattern in the southern and eastern TYS that is quite similar to that obtained from the analysis of the altimeter data.

  • Fig. 10.

    Summer circulation in the eastern TYS: (a) 1993–2010 summer average ADT, with a geostrophic reconstruction of the circulation superposed; (b) the 2009 summer average of the surface (10-m depth) velocity from TYREM. Main vortices are present in both panels, but the model results highlight the presence of a wide anticyclonic circulation cell in the area.

  • Fig. 11.

    Time series of average surface wind curl (s−1) from the ERA-Interim dataset (the average is on the box 38°–40°N, 12°–15.5°E; drawn in the August panel of Fig. 2). The wind curl average oscillates between positive values in winter and negative ones in summer.

  • Fig. 12.

    Singular spectra of the SLA map time series. Black points connected by blue lines indicate the eigenvalue amplitude. The lower and upper ticks on the error bar (red) indicate the 5th and 95th (noise) percentile from the χ2 test. For each EOF, a characteristic frequency was estimated by maximizing its correlation with a sinusoid. In (a), the whole SLA was used and M = 468 weeks for the SSA analysis. In (b), the basin average was first subtracted from the SLA maps and M = 312 weeks. The inset in (b) is a zoom around the eigenvalues corresponding to EOFs 3 and 4.

  • Fig. 13.

    Principal EOFs of the SLA time series after removing the basin average (and hence most of the steric effect). EOFs 1 and 3 (2 and 4) are indicated by solid (dashed) lines.

  • Fig. 14.

    Reconstruction of the annual cycle as given by EOFs 1 and 2 with M = 312 weeks. Contour intervals of 0.5 cm. The thick line represents the zero level.

  • Fig. 15.

    Time evolution of the low-frequency mode reconstructed using EOFs 3 and 4.

  • Fig. 16.

    Time series of the kinetic energy of the geostrophic flow associated to the 6 yr mode of variability of the SLA. The kinetic energy has been averaged over seven 2° × 2° boxes, covering the TYS. The signature of the mode is clear in the three western boxes, and the energy is three times larger in the northernmost of them, in which the Bonifacio couple of gyres resides.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 833 126 8
PDF Downloads 461 62 7