**Sea Surface Circulation Structures in the Malta-Sicily Channel from Remote Sensing Data**

**Nydia C. Reyes Suarez 1,2,\*, Michael S. Cook 3, Miroslav Gaˇci´c 2, Je**ff**rey D. Paduan 3, Aldo Drago <sup>4</sup> and Vanessa Cardin <sup>1</sup>**


Received: 20 June 2019; Accepted: 25 July 2019; Published: 31 July 2019

**Abstract:** The Malta-Sicily Channel is part of the Sicily Channel system where water and thermohaline properties between the Eastern and Western Mediterranean basins take place. Several mesoscales features are detached from the main circulation due to wind and bathymetric forcing. In this paper, surface circulation structures are studied using different remotely sensed datasets: satellite data (absolute dynamic topography, Cross-Calibrated Multi-Platform wind vector analysis, satellite chlorophyll and sea surface temperature) and high frequency radar data. We identified high frequency motions (at short time scales—hours to days), as well as mesoscale structures fundamental for the understanding of the Malta-Sicily Channel circulation dynamics. One of those is the Malta-Sicily Gyre; an anticyclonic structure trapped between the Sicilian and Maltese coasts, which is poorly studied in the literature and often confused with the Malta Channel Crest and the Ionian Shelf Break Vortex. In order to characterize this gyre, we calculated its kinetic properties taking advantage of the fine-scale temporal and spatial resolution of the high frequency radar data, and thus confirming its presence with an updated version of the surface circulation patterns in the area.

**Keywords:** remote sensing; high frequency radar; Malta-Sicily Channel; Malta Sicily Gyre; surface circulation

#### **1. Introduction**

#### *Study Area*

Although smaller than the synoptic scale, mesoscale to sub-mesoscale structures influence the stratification and contribute to the vertical and horizontal advection of sea water properties. Due to the small Rossby radius of deformation in the Mediterranean Sea (15–20 km) the typical mesoscale features are characterized by scales of the order of 10–100 km [1,2]. The circulation in the Sicily Channel (SC; defined as 11–16.5◦ E and 33—38◦ N in Figure 1 for our studies) is mainly driven by the Mediterranean thermohaline circulation together with its mesoscale and seasonal variability showing intra-annual and inter-annual variability [3–5]. This area is characterized by a complex bathymetry with a two-sill system at the Sicily Strait (SS) [6]. On the Italian side, the sill is very narrow with a maximum depth of ~430 m, while on the Tunisian side it is wider and shallower with a sill depth of ~360 m [7]. The exit to the Ionian Sea is bounded by the Malta Plateau and the Medina Bank (MP and MB in Figure 1). In the center of the SC three important depressions are present; the Pantelleria trough (PT; ~1300 m depth), the Linosa trough (LT; ~1500 m depth) and the Malta trough (MT; ~1700 m depth). These bathymetric features can strongly influence the current system in the SC.

Contrary to the SC, the Malta-Sicily Channel (hereafter referred as the channel) is a shelf in the mid Mediterranean Sea that separates the Malta island from the southern tip of Sicily. Topographically this shelf is characterized by a plateau in the middle (MP) with an average depth of 150 m as shown in Figure 1. On its northwestern side the shelf is bounded by the submarine ridges aforementioned with depressions reaching 1700 m deep where the Maltese islands represent the emerged part of the ridge [8]. On its southeastern side, the channel abruptly deepens into the Ionian Sea (~3000 m deep) due to the presence of the Sicily-Malta Escarpment (~2500 m deep) which is one of the largest and least explored underwater cliffs in the Mediterranean Sea [8]. The circulation on longer time scales (lower frequency) is mainly driven by the Atlantic Ionian Stream (AIS) and the Malta-Sicily Gyre (MSG), where the latter has been identified as a quasi-permanent structure with highest incidence in the winter period and related to the variability of the AIS [9,10]. In addition, the Malta-Sicily Gyre (MSG) has been poorly described in the literature, and is often confused with the Malta Channel Crest (MCC) and the Ionian Shelf Break Vortex (ISV) [9]. For the purpose of this work the channel has been defined as the area comprising 14–16◦ E and 35–37◦ N.

**Figure 1.** Bathymetric map of the Sicily Channel (SC) (dashed yellow box) and the channel (solid yellow box) and the surrounding area. Geographical places are indicated with red letters: Adventure Bank (AB), Gela Basin (GB), Lampedusa Island (La), Linosa Island (Li), Linosa Trough (LT), Maltese Islands (M), Malta Plateau (MP), Malta Trough (MT), Medina Bank (MB) Pantelleria Island (P), Pantelleria Trough (PT), Sicily Island (S), Sicily Strait (SS), Sicily-Malta Escarpment (SME), and Tunisia Plateau (TP). The colorbar represents depth in meters. Bathymetric data was obtained from https: //topex.ucsd.edu/cgi-bin/get\_data.cgi [11].

The AIS is a well-studied sea surface structure mainly composed of Atlantic Water (AW). During summer it strengthens and remains constrained towards Sicily where the flow behaves like a jet stream gaining positive vorticity due to westerly winds and the SC complex bathymetry [12]. As it reaches the shelf break east of Malta, it maintains an upwelling front along the southern coast of Sicily showing a seasonal behavior [13–16]. Its path starts at Adventure Bank where the AW enters the SC via the SS where complex circulation patterns are induced by the bathymetry and the shape of the coastline [1,17,18]. During summer, at the SS, the jet stream flows towards Malta following the bathymetry and ends its journey in the channel by emptying into the Ionian Sea. Depending on the phase of the North Ionian Gyre (NIG; [19–22]) the AIS can flow towards the North Ionian Sea or move as a zonal current crossing the Central Ionian Sea. In the latter case this flow is known as the Mid Ionian Jet, (MIJ; [1,4]) and connects the SC with the Cretan passage [4,19–21,23]. On its eastward propagation the stretching and squeezing of the AIS results in the formation of important mesoscales structures e.g., eddies, gyres or plumes like the MSG [14,24,25]. Additionally, local winds, tides and the Coriolis effect can introduce high frequency motions in the area.

The main focus of this paper is to study the surface circulation patterns in the channel with particular emphasis on a quantitative characterization of the MSG. This mesoscale feature appears to be important to and diagnostic of the circulation in the area under study but it has been poorly studied and often confused with other mesoscale structures. Our aim is to give an updated version of the sea surface circulation structures in the channel. This paper is divided in four parts besides the introduction. In Section 2, we introduce the data and methods used for this study. The data were selected due to their spatial and temporal availability and capability of being compared with high frequency radar data (HFR). A summary of the datasets used and their principal characteristics can be found in Table 1. Section 3 is devoted to results and discussion. We first describe the principal features that comprise the surface circulation in the channel and in the SC (solid and dashed yellow boxes in Figure 1) using surface geostrophic velocities (SGV) and absolute dynamic topography (ADT). Secondly, we identify the short time scales in the channel by analyzing the sea surface circulation and the dynamical processes in the area with available remote sensing data i.e., ADT, SGV, HFR velocities, wind stress fields, sea surface temperature (SST), and chlorophyll satellite images (CHL). Finally, we focus on the characterization of the MSG (identified using the aforementioned data) by calculating its kinematic properties. Section 4 is dedicated to conclusions.


**Table 1.** Main characteristics of the remote sensing datasets.

#### **2. Data and Methods**

To describe the channel's main surface structures, we have analyzed the available remote sensing data summarized in Table 1.

#### *2.1. Remote Sensing Data*

• HFR data from three CODAR SeaSondes stations installed in Ta'Sopu (Gozo, Malta), Ta'Barkat (Malta) and Pozzallo Harbor (Sicily, Italy) shown in Figure 2, provided surface current maps in the channel from the period August 2012 to January 2015. These data correspond to a setup of two HFR stations initially installed at Ta'Sopu and Ta'Barkat in August 2012. The third station was added in August 2013 in Pozzallo improving the spatial coverage of the channel [26]. The data are organized in time series of hourly velocity vectors with *u* (zonal, East-West) and *v* (meridional, North-South) components of the total velocity. The datasets were based on CALYPSO HFR, compiled and processed by Dr. Simone Cosoli from the University of Western Australia, Perth [27]. The CALYPSO system operating set-up and resolution (13.5 MHz frequency, angular resolution 5◦, range resolution 1.6 km, for more details see Drago et al. [10]) provides radar measurements that are representative of the first meter of the ocean with grid sizes from 0.3 to 8.3 km2. The radars share the same transmit frequency using a GPS-synchronization module and operate with both the ideal and measured antenna beam patterns. Hourly sea surface current maps were derived on a Cartesian grid with 3 <sup>×</sup> 3 km<sup>2</sup> horizontal resolution by least-squares fitting of the radial components of the ocean currents from two or more radar stations in areas of common overlap. Grid points were included in the analysis only if they satisfied a minimum data return of 50% using an interpolation technique described in Cosoli et al. [27]. Validation of this array has been carried out in different studies since the installation of the system making this dataset a reliable product [27–29].

**Figure 2.** High frequency radar stations of three CODAR SeaSondes stations at Ta'Sopu (Gozo, Malta), Ta'Barkat (Malta) and Pozzallo Harbor (Sicily, Italy) from [30] showing the area of maximum coverage of the array.

• Ssalto/Duacs multi-mission L4 altimeter products in the period January 1993 to December 2015 containing daily multi-mission ADT on a 1/8◦ × 1/8◦ mercator projection grid, and distributed by the Copernicus Marine and Environment Monitoring Service (CMEMS) [4,5,31] were used to calculate SGV where,

$$
\mu\_{\mathcal{S}}' = -\frac{g}{f} \frac{\partial \zeta}{\partial y},
\tag{1}
$$

$$
v\_{\mathcal{S}}' = \frac{\mathcal{g}}{f} \frac{\partial \zeta}{\partial \mathbf{x}}.\tag{2}$$

are the zonal and meridional geostrophic velocities, and ζ, denotes the dynamic topography provided by the altimeter.


period spanning August 2013 to January 2015 to analyze wind patterns in the channel [32]. The CCMP dataset combines cross-calibrated satellite microwave winds from scatterometers and radiometers with instrument observations using a variational analysis method to produce 1/4◦ gridded data [33]. Both radiometer and scatterometer data are validated against ocean moored buoys (in agreement within 0.8 m/s), where wind observations are referenced to a height of 10 m. For a complete description of the dataset see [32].

#### *2.2. Complex Correlation and Veering Estimates*

Average veering between two 2D time series can be obtained from the phase angle of the complex correlation coefficient [34]. Additionally, we used this method to determine if the wind and HFR velocities are in Ekman Balance.

Let,

$$w(t) = u(t) + i\upsilon(t),\tag{3}$$

be the complex representation of the velocity time series at time *t*. The complex correlation coefficient between the two vector time series 1 and 2 in terms of the east and west components is:

$$p = \frac{\langle \boldsymbol{u}\_1 \, \boldsymbol{u}\_2 + \boldsymbol{v}\_1 \, \boldsymbol{v}\_2 \rangle}{\langle \boldsymbol{u}\_1^2 + \boldsymbol{v}\_1^2 \rangle^{\frac{1}{2}} \, \langle \boldsymbol{u}\_2^2 + \boldsymbol{v}\_2^2 \rangle^{\frac{1}{2}}} + i \frac{\langle \boldsymbol{u}\_1 \, \boldsymbol{v}\_2 - \boldsymbol{u}\_2 \, \boldsymbol{v}\_1 \rangle}{\langle \boldsymbol{u}\_1^2 + \boldsymbol{v}\_1^2 \rangle^{\frac{1}{2}} \, \langle \boldsymbol{u}\_2^2 + \boldsymbol{v}\_2^2 \rangle^{\frac{1}{2}}} \,\tag{4}$$

where the phase angle or average veering is,

$$
\alpha\_{uv} = \frac{\langle \mu\_1 \,\, \upsilon\_2 - \mu\_2 \,\, \upsilon\_1 \rangle}{\langle \mu\_1 \,\, \upsilon\_2 + \upsilon\_1 \,\, \upsilon\_2 \rangle}. \tag{5}
$$

#### *2.3. Kinematic Properties of an Eddy*

In order to describe the channel in terms of its kinematic properties, we applied the method described by Sanderson [35] using the lstranslate routine developed by the HFR group at the Naval Postgraduate School.

Assuming there is an eddy linearly translated in time with spatially uniform velocity gradients as follows:

$$\mathbf{g}\_{11} = \frac{\partial \mathbf{u}}{\partial \mathbf{x}}\tag{6}$$

$$\mathbf{g}\_{12} = \frac{\partial \mathbf{u}}{\partial y} \tag{7}$$

$$\mathbf{g}\_{21} = \frac{\partial \mathbf{v}}{\partial \mathbf{x}}\tag{8}$$

$$\text{g}\_{22} = \frac{\partial v}{\partial y} \tag{9}$$

From Equations (6)–(9), the divergence is defined as *d* = *g*<sup>11</sup> + *g*22, stretching deformation *a* = *g*<sup>11</sup> − *g*22, vorticity *c* = *g*<sup>21</sup> − *g*<sup>12</sup> and the shearing deformation *b* = *g*<sup>21</sup> + *g*12, which are functions of the velocity gradients. The method assumes that the center of the eddy is located at *X*, *Y* and moves with constant velocity (*U*, *V*), whereas, at some distance from the center of the eddy, the flow velocity has an added component due to the velocity gradients. *X*0, *Y*<sup>0</sup> is defined as the initial position of the eddy. Then the flow center position can be written as,

$$X = X\_0 + \mathcal{U}(t),\tag{10}$$

$$Y = Y\_0 + V(t). \tag{11}$$

with,

$$
\alpha = \mathcal{U} - \mathcal{g}\_{11}\mathcal{X}\_0 - \mathcal{g}\_{12}\mathcal{Y}\_0. \tag{12}
$$

$$
\beta = V - \text{g} \mathbf{z} \mathbf{1} \mathbf{X}\_0 - \text{g} \mathbf{z} \mathbf{2} \mathbf{Y}\_0 \tag{13}
$$

$$
\alpha\_1 = -\mathcal{g}\_{11}\mathcal{U} - \mathcal{g}\_{12}V\_{\prime} \tag{14}
$$

$$
\beta\_1 = -\mathbb{g}\_{21} \mathbb{U} - \mathbb{g}\_{22} \mathbb{V}.\tag{15}
$$

and solving Equations (12)–(15) for *U*, *V*, *X*0, *Y*<sup>0</sup> yields,

$$
\Delta U = \frac{\mathcal{g}\_{22}\alpha\_1 - \mathcal{g}\_{12}\beta\_1}{\mathcal{g}\_{12}\mathcal{g}\_{21} - \mathcal{g}\_{11}\mathcal{g}\_{22}}\tag{16}
$$

$$V = \frac{\mathcal{g}\_{11}\mathcal{\beta}\_1 - \mathcal{g}\_{21}\alpha\_1}{\mathcal{g}\_{12}\mathcal{\mathcal{g}}\_{21} - \mathcal{g}\_{11}\mathcal{g}\_{22}}\tag{17}$$

$$X\_0 = \frac{\mathfrak{g}\_{22}(\alpha - lI) - \mathfrak{g}\_{12}(\beta - V)}{\mathfrak{g}\_{12}\mathfrak{g}\_{21} - \mathfrak{g}\_{11}\mathfrak{g}\_{22}}\tag{18}$$

$$Y\_0 = \frac{\mathcal{g}\_{11}(\mathcal{J} - V) - \mathcal{g}\_{21}(\alpha - lI)}{\mathcal{g}\_{12}\mathcal{g}\_{21} - \mathcal{g}\_{11}\mathcal{g}\_{22}}\tag{19}$$

For our studies, we focused on the vorticity component applied to the HFR data in order to quantitatively identify the mesoscale features within the study area.

#### **3. Results and Discussion**

#### *3.1. Mean Surface Circulation, the Sicily Channel*

Due to baroclinic instabilities, the Atlantic current (AC) regularly forms meanders that eventually detach from the current and become either cyclonic or anticyclonic eddies [36] like the ones observed by Jouini et al. [3] and Jebri et al. [37,38]. Figure 3 shows the SC inter-annual geostrophic circulation derived from the ADT data depicting some of the AC-born structures described in Jouini et al. [3] and Menna et al. [5] such as the cyclonic Medina Gyre (MG), the cyclonic Messina Rise Vortex (MRV), the anticyclonic Pantelleria Vortex (PV, which in the literature is mentioned as cyclonic), the anticyclonic MSG, the AIS, the Bifurcation Atlantic Current (BAC) and the Atlantic Tunisian Current (ATC). From ADT and SGV, the MSG shows a seasonal semi-permanent behavior and contributes to the channel's circulation along with the AIS. Due to its poor description in the literature its characterization is an important result of our studies. The complex bathymetry in the SC, with its longitudinal subdivision into two sills, influences the distribution of ADT leading to a difference in level between the two sides of the channel as seen in Figure 3. Towards the Italian side the sea level is lower than on the Tunisian side with a dipole like sea level shape, positive ADT close to Tunisia and negative towards Sicily.

The seasonal variability of the circulation in Figure 4 was obtained by averaging the ADT maps using the following groupings: Winter (December, January, February), Spring (March, April, May), Summer (June, July, August) and Autumn (September, October, November). In winter (Figure 4a) the mean geostrophic circulation shows the presence of the AC bifurcating at the SS in two branches due to the two-sill bathymetry of the strait. The BAC flows toward the northern tip of Sicily while the ATC to the Tunisian side. The MSG is observed as well with a radius of ~50 km and centered at 14.5◦E–36.1◦N. It is also prominent during the winter/spring season as seen in Figure 4b, dominating the circulation in the channel when the AIS is not present. This can be interpreted as a result of the conservation of potential vorticity [13–15] since the water depth abruptly changes from ~700 m at the Malta graben to shallow (~100 m) in the channel (see Figure 1). For this season, the MSG velocity ranges between 5–10 cm s<sup>−</sup>1. Additionally, other structures observed for this season are the anticyclonic MCC, the cyclonic MG and MRV.

**Figure 3.** Inter-annual map of absolute dynamic topography (ADT) and mean surface geostrophic velocities (SGV) in the SC. Permanent structures in white, summer/autumn in red and winter/spring in blue. The colorbar represents the ADT variations over the study area while gray arrows are the SGV vectors.

In spring we identify some of the winter structures indicated in Jouini et al. [3] and Menna et al. [4,5] like the ATC, BAC, MCC, MG, MSG, MRV and the PV as shown in Figure 4b. The ATC flows towards Sicily where two gyres are detached, the cyclonic MG and the anticyclonic MCC circling around Linosa island, most likely due to the influence of the LT at 36◦N. In addition, at the northern tip of Sicily there is a re-circulation of AW that enters into the SC where the PV is detached and the MSG is reinforced.

For the summer period we observe the strong influence of the AIS (Figure 4c) which behaves like a jet stream in the SC and the channel [13]. It is characterized by a velocity of ~25 cm s−<sup>1</sup> and occasionally reaching values higher than 50 cm s<sup>−</sup>1. This current consists mainly of AW and changes properties as it passes through the SS deepening and bifurcating into the Ionian Sea northwards or to the Central Ionian Sea depending on the NIG phase [19,21,23,38]. Additionally, the ADT and SGV also show another yet unidentified mesoscale structure close to the coast of Tunisia detached from the ATC, and a shift of the MG eastwards.

Figure 4d shows autumn mean geostrophic circulation which is driven by the AIS and reinforced by AW coming from the BAC and the ATC. The ATC bifurcates at the SS and then rejoins south of Pantelleria. We observed other mesoscale structures, like the cyclonic MRV and the anticyclonic MCC, as well as the formation of the anticyclonic MSG which has been undocumented in the literature.

Other permanent structures were found in the SC in disagreement with the current literature. For example, the anticyclonic PV at 37.5◦N–11.5◦E is in disagreement with the studies of Robinson et al. [12], Lermusiaux and Robinson. [15] and Jouini et al. [3]. In their case the PV is cyclonic and positioned at 37◦N–12◦E. Additionally, the MCC is described as a summer only pattern in the literature

but in our studies is also present in winter and spring. The cyclonic Adventure Bank Vortex (ABV) and the ISV were summer only structures in the literature but were not found in our seasonal or inter-annual geostrophic decomposition. Furthermore, over the monthly mean analysis we were able to identify some of these structures, but they seemed to be temporary and appeared just in some years/months where they could have been triggered by rare and strong atmospheric events. An updated picture of the surface circulation structures in the whole SC where the discrepancies and the reasons why they are enhanced should be clarified and studied with more extent and detail. A new study of all these structures and the forcing mechanisms behind them can be found in Menna et al. [5], but again the MSG is not mentioned.

**Figure 4.** Mean seasonal maps of ADT and SGV in the SC. (**a**) winter, (**b**) spring, (**c**) summer and (**d**) Autumn. Yellow arrows represent the sea surface patterns found. Gray arrows represent the SGV in cm s<sup>−</sup>1. The colorbar shows ADT in centimeters.

Wind stress curl and deseasonalized geostrophic eddy kinetic energy (EKE) studies in Menna et al. [4] revealed the influence of the wind and highlights the regions with high EKE in the Central Mediterranean Sea. Wind stress data (Figure 3c,d in Menna et al. [4]) shows that towards the southwestern side of the Sicilian coast northwesterly winds favor upwelling (+ wind stress) revealing persistent coastal Ekman pumping events [39,40]. Wind forcing imparts an indirect influence in the current field at longer time scales (weekly, monthly, and inter-annual), where even if neglected in the geostrophic approximation, helps to build up the sea level to generate a sufficiently large pressure gradient needed to balance the Coriolis force. On the other hand, deseasonalized geostrophic EKE studies (Figure 6 in Menna et al. [4]) highlight areas in the central Mediterranean Sea where the inter-annual variability is stronger with values of ~80 cm<sup>2</sup> s−<sup>2</sup> in the channel. A broader EKE variability using altimetry and model datasets in the Mediterranean Sea can be found in Pujol and Larnicol [41] and Jordi and Wang [42].

#### *3.2. Short Time Scales in the Malta Sicily Channel*

#### 3.2.1. Comparison among Available Spatial Data

Ocean currents can be described as a combination of geostrophic and ageostrophic terms, the latter being associated with wind driven features. HFRs offer good resolution in both time and space, and have the capability to measure near real-time complete currents [43,44], measuring both wind-driven and geostrophic components influencing the sea surface. Weekly, monthly, inter-annual and seasonal averages of SGV and HFR velocity field were compared. Figure 5 shows monthly means of both components correlate well geographically. This strong correlation was found at all-time scales greater than a day and thus identified the channel to be in geostrophic balance at these time scales. As expected, the stretching and squeezing of the anticyclonic MSG and the AIS jet stream were also observed from the HFR data as did the SGV and ADT. In addition, the MSG can occasionally have velocities as high as 40 cm s−<sup>1</sup> as seen in Figure 5b.

**Figure 5.** SGV calculated from ADT showing the pathway of Atlantic Ionian Stream (AIS) on September 2012 (left-hand panel) and the Malta-Sicily Gyre (MSG) identified on February 2014 (right-hand panel). Black lines represent monthly averaged current fields derived from high frequency radar data (HFR) data in cm s<sup>−</sup>1. Green arrows represent SGV in cm s−1. The colorbar gives the magnitude of each velocity vector in cm s<sup>−</sup>1.

Other spatial observations available for this area include monthly satellite derived SST and CHL data, which were compared with HFR monthly averages for the period from January to December 2013. Since the prevailing signal in the HFR data at monthly time scales is geostrophic, it is expected that the advection of properties follows this flow. However, one should remember that SST and CHL are not conservative tracers but can be used in general to track currents, circulation, and water mass mixing among other properties. Figure 6a,c show SST and CHL maps in January 2013 respectively, revealing relatively higher temperatures and lower CHL occupying the channel and lower temperatures and higher CHL constrained towards the Sicilian coast. This behavior can be associated with the presence of the MSG, which for the winter season is the predominant pattern. Figure 6b,d illustrate that as the AIS flows towards the Ionian Sea it generates a well-defined front which advects properties in the channel along its path [12]. Additionally, the SST field varies from 17 ◦C to 26 ◦C while CHL from 0.5

to 0.05 mg m−<sup>3</sup> through winter to the end of summer. A detailed statistical analysis of CHL and SST data is presented in Capodici et al. [26].

**Figure 6.** Sea surface temperature (SST) and chlorophyll satellite images (CHL) fields from Aqua-MODIS satellite mission showing the MSG on January 2013 (panel (**a**) and (**c**), respectively) and the AIS pathway August 2013 (panel (**b**) and (**d**), respectively). Gray lines represent monthly averaged surface currents from the HFR, whereas the colorbar represents the SST field in ◦C (upper panels) and chlorophyll concentration in mg pigment m−<sup>3</sup> (lower panels).

#### 3.2.2. Complex Correlation

We spatially averaged wind (1/4◦), SGV (1/8◦) and HFR (1 km) daily gridded data to match the 1/4◦ wind grid. Figure 7 shows the spatially averaged time series over the 1/4◦ grid for each dataset. The comparison of the three parameters in panels a and c of Figure 7 show that the geostrophic components (*u*,*v*SGV) behave like a low-pass filter of the HFR time series because the geostrophic adjustment takes place on a longer time scale. Wind (*u*,*v*wind) and HFR (*u*,*v*radar) time series show most of the high frequency motions included at short time scales, where the wind plays an important role on the *channel* circulation. Additionally, panels b and d compare the wind speed time series with the residual velocity obtained by removing the geostrophic component from the HFR data, *u*,*v*residual = *u*,*v*radar − *u*,*v*SGV, where the residual is expected to be better correlated to the wind components than to the geostrophic current. Here the *u* component (panel b) shows good agreement with the residual as expected, except for some time intervals such as January to March 2013 where the wind is positive and the residual is negative and higher in magnitude. The residual meridional *v* component of the velocity (panel d) behaves differently from the wind signal in the first few months (until September 2013). This is probably due to data availability since for some grid cells the lack of HFR data could play an important role in the magnitude of the vectors and thus the averaging process.

**Figure 7.** *u* and *v* components of daily wind, SGV and HFR speed time series spatially averaged over a 1/4◦ grid. Panels (**a**,**c**) compare *u* (zonal, East-West) and *v* (meridional, North-South) component of the wind (green), SGV (red) and HFR (blue) velocities respectively. Panels (**b**,**d**) compare the residual *u* and *v* speed (*u*,*v*residual = *u*,*v*radar − *u*,*v*geostrophic) with *u* and *v* components of the wind speed.

In order to look for the relationship between different variables, we compute the vector correlation between the interpolated time series in Figure 7 over the grid represented in Figure 8 where the boxes represent the 1/4◦ grid defined previously. HFR to the wind complex correlation (as seen in Table 2) shows the veering 43.5781 ± 8.238 degrees to the right, in a good agreement with theory. The angles of the velocity to the wind vector varies between 23 to 53 degrees with a range correlation between 0.24 to 0.35, showing the importance of the wind forcing at shorter time scales and in the set-up of the Ekman layer. The correlation between the residual current and the wind vector is higher than the radar to wind correlation. The angles between residual and wind are closer to 45 degrees, as a result of the removal of the geostrophic components from the time series (see Table 2). Finally, as expected the complex correlation between the geostrophic flow and the wind is very low since processes in both time series evolve at different time scales; the wind component at short time scales generate rapid and variable motions (i.e., inertial oscillations) in the ocean whereas the geostrophic field evolves at larger scales accounting for low frequency phenomena.

**Figure 8.** Grid cells where complex correlation is computed. The subscripts represent the corresponding grid cell for each set of parameters, e.g., *p*1, θ<sup>1</sup> are the parameters corresponding to cell 1.



#### *3.3. The Malta Sicily Gyre*

Evidence of mesoscale activity in the channel was highlighted by four datasets considered in this study: ADT, HFR, CHL and SST. To identify this mesoscale activity, we computed the vorticity using the method described in Sanderson [35] considering only a subset of the HFR time series, using the high resolution and continuity criteria. For the continuity criteria we defined the area in Figure 9e where data was available since the start of acquisition. This area also coincides with the place where the MSG was identified with the HFR velocities and geostrophic current analysis.

**Figure 9.** Subset time series of the original HFR grid. Panel (**a**) represents the *u* (east-west) while panel (**c**) the *v* (north-south) components of the HFR time series respectively. Blue and gray represent the time series respective to the dots on panel (**e**). Panels (**b**,**d**) represent the East-West and North-South differences calculated from the blue time series in panel (**a**,**c**) respectively. Vorticity is depicted in the right-hand side of panels (**b**,**d**) and was calculated using Sanderson's method [35]. Violet shaded areas show the months when the MSG was found. Pink shaded areas show the months where the Atlantic Ionian Stream (AIS) and Malta-Sicily Gyre (MSG) co-existed. Green shaded areas show the cyclonic mesoscale in Figure 10c.

**Figure 10.** Daily HFR current maps showing anticyclonic (**a**,**b,d**) and cyclonic (**c**) mesoscale gyres in agreement with the vorticity calculated through Sanderson's method. The colorbar gives the magnitude of each velocity vector in cm s<sup>−</sup>1. Please note that panel **c** depicts one of the cyclonic gyres corresponding to the green shaded areas in Figure 9. Panel **a** and **d** depict some of the anticyclonic gyres in the pink shaded areas of Figure 9 occurring alongside the AIS. Panel **b** shows a snapshot of the MSG in one of shaded violet areas of panels **b** and **d** of Figure 9.

Figure 9 (panel b and d) shows the presence of the anticyclonic MSG in winter between December-March 2013 and December-August 2014, corroborated by a vorticity of <sup>c</sup> <sup>=</sup> <sup>−</sup><sup>2</sup> <sup>×</sup> <sup>10</sup>−<sup>5</sup> <sup>s</sup>−<sup>1</sup> using Sanderson's method (panel b and c). These results are in agreement with the winter/spring recurrence of the MSG identified with the ADT and SGV analysis. Here we also note that the MSG co-existed with the AIS from April to August 2014 (shaded areas in Figure 9). This provides a quantitative description of the gyre. Finally, with this method we were able to identify other mesoscales gyres that were either cyclonic (+vorticity) or anticyclonic (−vorticity). Some examples are depicted in Figure 10.

#### **4. Summary and Conclusions**

The main focus of this paper has been to characterize the channel circulation, which was done using satellite-derived geostrophic currents HFR-derived currents, and wind vector analyses on daily time scales in addition to SST and CHL on monthly time scales. Analyzing the available data, with particular emphasis on the higher spatial and temporal resolution of the HFR observations, the MSG is studied in more detail. This is an anticyclonic structure occupying a substantial portion of the channel that has been poorly studied and often confused with the ISV or the MCC. We believe that the standardization of the names and characteristics for these structures will benefit researchers and resource managers working in the region. Menna et al. [5] provides a good description of the main surface circulation structures in the SC but characterizes the MSG (or MCC as they called it) as a seasonal feature, occurring in the summer only. This creates some discrepancies since based on the lower frequency qualitative studies (i.e., SGV, SST and CHL maps), that agree with [10], the mesoscale feature is present mostly in the winter-spring period. This however is not always the case since in 2014 both the AIS and the MSG were present from January to August. Tracking the anticyclonic MSG was particularly well done using the vorticity analysis performed on the HFR observations. While the channel-scale MSG was dominant, other cyclones and anticyclones were identified at daily time scales in the HFR observations making Sanderson's method a good tool for identifying mesoscale features. A longer HFR time series with better spatial coverage can be useful to study the mesoscale.

Geostrophic balance was found to be a good approximation to describe the dynamics in the channel at weekly, monthly, seasonal and inter-annual time scales. At these time scales, wind forcing is important for creating the density gradients and the subsequent geostrophic velocities, but wind data are not essential in diagnosing the circulation patterns (wind stress curl as in Menna et al. [4], shows persistent coastal Ekman upwelling events along the coast of Sicily.) The geostrophic velocity field in the channel is reinforced by the wind, which is mostly northwest and is responsible for the advection of different properties in the area. At short time scales (hours to days) we were able to use the combination of HFR and geostrophic velocities to show that direct wind forcing must be accounted for. The vector complex correlation between the HFR and the wind time series allowed us to show that the high frequency components included an Ekman balance between wind and surface current. The analysis showed a veering angle for the currents of 43.5781 ± 8.238 degrees to the right of the wind forcing, which is in good agreement with Ekman theory. The agreement was even better when the geostrophic currents were subtracted from the HFR observations to produce a residual current. Hence, we conclude that HFR observations or a wind-based correction to the geostrophic currents, are recommended at daily time scales. This set of analyses using observations at larger time and space scales from satellites and observations of higher frequency, higher resolution features from HFR suggests that this suite of observations can provide effective monitoring of the circulation state for important regions, such as the area studied here.

**Author Contributions:** Writing—original draft, N.C.R.S.; Writing—review & editing, N.C.R.S., M.S.C., M.G., J.D.P., A.D. and V.C.; Conceptualization, N.C.R.S., M.S.C., M.G. and J.D.P.; Data curation, N.C.R.S. and M.S.C.; Formal analysis, N.C.R.S., M.S.C. and J.D.P.; Funding acquisition, A.D., M.G. and V.C.; Investigation, N.C.R.S. and M.G.; Methodology, N.C.R.S., M.G. and J.D.P.; Resources, M.G., A.D. and V.C.; Software, M.S.C. and J.D.P.; Supervision, M.S.C., M.G. and J.D.P; Visualization, N.C.R.S.

**Funding:** This research was financially supported through the award of the scholarship to N.C.R.S. by the Istituto Nazionale di Oceanografia e Geofisica Sperimentale—OGS, the University of Trieste—UNITS and the TRIL program of the Abdus Salam International Center for Theoretical Physics—ICTP. HF radar data were collected within the Italia-Malta program-Cohesion Policy 2007–2013, European Union regional Development Funds (ERDF) through the CALYPSO project.

**Acknowledgments:** The authors would like to thank Aldo Drago for providing the HFR data used in this study. We also acknowledged Simone Cosoli for providing the processed and QC dataset from the HFR array. Last but not least, we also thank to the two anonymous reviewers for their constructive observations.

**Conflicts of Interest:** The authors declare no conflict of interest.

*Water* **2019**, *11*, 1589

#### **Abbreviations**

The following abbreviations are used in this manuscript:


#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **New Insights of the Sicily Channel and Southern Tyrrhenian Sea Variability**

#### **Milena Menna 1,\*, Pierre-Marie Poulain 1,2, Daniele Ciani 3, Andrea Doglioli 4, Giulio Notarstefano 1, Riccardo Gerin 1, Marie-Helene Rio 5, Rosalia Santoleri 3, Adam Gauci <sup>6</sup> and Aldo Drago <sup>6</sup>**


Received: 31 May 2019; Accepted: 27 June 2019; Published: 29 June 2019

**Abstract:** The dynamics of the Sicily Channel and the southern Tyrrhenian Sea are highly influenced by the seasonal variability of the Mediterranean basin-wide circulation, by the interannual variability of the numerous mesoscale structures present in the Channel, and by the decadal variability of the adjacent Ionian Sea. In the present study, all these aspects are investigated using in-situ (Lagrangian drifter trajectories and Argo float profiles) and satellite data (Absolute Dynamic Topography, Sea Level Anomaly, Sea Surface Temperature, wind products) over the period from 1993 to 2018. The availability of long time series of data and high-resolution multi-sensor surface currents allow us to add new details on the circulation features and on their driving mechanisms and to detect new permanent eddies not yet described in literature. The structures prevailing in winter are mainly driven by wind, whereas those prevailing in summer are regulated by topographical forcing on surface currents. The strength of the surface structures located at the western entrance of the Ionian Sea and of the mesoscale activity along the northern Sicily coast is modulated by the large-scale internal variability. The vertical hydrological characteristics of these mesoscale eddies are delineated using the Argo float profiles inside these structures.

**Keywords:** Sicily Channel; spatial and temporal variability; mesoscale eddies

#### **1. Introduction**

Thanks to its location in the centre of the Mediterranean Basin, the Sicily Channel (SC) plays a crucial role in connecting the western and eastern Mediterranean basins and modulating their exchange of surface and intermediate waters [1,2]. The SC is characterized by a complex bottom topography, with submarine ridges and shallow banks, and it is delimited to the north by the Tyrrhenian Sea and the Sicily coast, to the south by the Libyan coast, to the west by the Tunisia coast, and to the east by the Ionian Sea (Figure 1a). Its circulation can be schematized on the vertical as a two-layer exchange with an eastward flow of Atlantic Water (AW) superposed to a westward flow of intermediate water, dominated by the Levantine Intermediate Water (LIW) [3]. Microstructure measurements taken in the SC show that it is a hotspot for turbulent mixing [4,5]. Consequently, the SC is a key area for the regulation of salt exchanges between Eastern and Western basins, with an impact on deep-water formation processes [6].

**Figure 1.** (**a**) Bathymetry of the SC (100 m, 200 m, 400 m, 600 m, 1000 m, 2000 m isobaths) and geographical references. (**b**) Low-pass filtered drifter trajectories in the SC between 1993 and 2018, superimposed with the schematic surface circulation adapted from [1,3]. Acronyms are listed in Table 1.

Circulation in the upper layer of the SC and in the southern Tyrrhenian Sea is mainly dictated by the large-scale Mediterranean thermohaline circulation, the wind-driven currents along the shelf, the upwelling events off Sicily, the sub-basin scale, and mesoscale permanent and quasi-permanent structures [1,3,7–9]. The recent results of AW circulation schemes, derived by numerical model simulation [1,3], are summarized in Figure 1b (both permanent and seasonal circulation structures are depicted with the same color in Figure 1b; more details on the time scales and variability of these structures are available in Section 3; acronyms are defined in Table 1). The sub-basin scale structures are characterized by a prominent seasonal variability [9–12] associated with large wind stress fluctuations [13]. The numerous mesoscale structures located in the SC are mainly driven by the instability of the large-scale circulation, by the interactions between currents and bathymetry, and by the direct wind forcing [1].

Presently, the availability of long time series of in-situ and satellite data and of sophisticated statistical techniques allow us to add new details on the mesoscale features and on their driving mechanisms. In this study, Lagrangian drifter trajectories, Argo float profiles, and satellite data (Absolute Dynamic Topography, Sea Level Anomaly, Sea Surface Temperature, wind products) are used to describe the surface circulation of the SC and the southern Tyrrhenian Sea over the period from 1993 to 2018. The simultaneous use of all these datasets leads to overcoming the intrinsic limitations of each of them, e.g., the accurate but discontinuous spatial and temporal drifter sampling and the low accuracy of altimetry gridded data in the identification of the mesoscale field [14]. Moreover, the knowledge of the mesoscale field is ameliorated using the multi-sensor currents (defined hereafter as 'optimal currents') derived from merging the satellite altimetry data and the Sea Surface Temperature (SST) fields [15]. This product enables the improvement of the description of eddy dynamics and non-geostrophic dynamical features [15]. The vertical structures and the hydrological characteristics of the mesoscale eddies are delineated using the Argo float profiles inside these structures. All these data and products enhance the knowledge on the temporal variability of the mesoscale structures, with detection of new features not yet described in the literature, and the addition of new insights on the formation mechanisms of these structures.


**Table 1.** List of acronyms used in this paper.

#### **2. Materials and Methods**

The datasets used for this study are as follows:


(see the MedArgo program in Reference [19]). When a float drifts in a shallow area and touches the ground, it can increase its buoyancy to get away from bottom, or can stay there until it is time to ascent (depending on how it is programmed). Information about grounding events is contained in the Argo float trajectory file. Among all the data available in the Mediterranean Sea, we selected from the part of the Argo floats trajectories which correspond to a float entrapped in the mesoscale structures of the SC and southern Tyrrhenian Sea. These data were used to define the vertical hydrographic peculiarities of the mesoscale features. Details about the missions of the seven floats selected for this work are listed in Table 2.



**Table 2.** List of selected Argo float profiles with dates and positions of the first and the last profile considered in this work, parking and profiling depths, and the cycle period of each instrument.

Drifter velocities were divided in bins of 0.25◦ × 0.25◦ and pseudo-Eulerian statistics were computed over the period from 1993 to 2018 and qualitatively compared with the ADT derived from altimetry. The mean current field was also estimated in the period from 2012 to 2016, using the optimal currents. The seasonal variability of the drifter, altimetry, and optimal current fields was estimated by dividing the dataset in two extended seasons, the extended summer corresponding to May–October, and the extended winter to November–April, as suggested by Reference [9].

The CCMP six-hourly gridded analyses were used to quantify the wind stress and the vertical component of the wind stress curl, [*curl* τ]z, over the study area:

$$[\text{curl}\,\tau]\_z = \frac{\partial \tau\_y}{\partial x} - \frac{\partial \tau\_x}{\partial y}; \{\tau\_y, \tau\_x\} = \rho \mathbb{C}\_D(u\_{w\prime}, v\_w) \mathcal{U}\_{10} \tag{1}$$

where (τ*x*, τ*y*) are the wind stress components, ρ (1.22 Kg/m3) is the density of air, (*uw*, *vw*) and *U*<sup>10</sup> are the components and the magnitude of the wind speed at 10 m, respectively, and *CD* is the drag coefficient already used in the Mediterranean Sea by References [23,24], as follows:

$$\begin{array}{ll} \mathbb{C}\_{D} = 10^{-3} & |\mathcal{U}\_{10}| \le 3 \,\frac{m}{s} \\ \mathbb{C}\_{D} = \left(0.29 + \frac{3.1}{\mathcal{U}\_{10}} + \frac{7.7}{\mathcal{U}\_{10}^{2}}\right) \times 10^{-3} & 3 \frac{m}{s} \le |\mathcal{U}\_{10}| \le 6 \frac{m}{s} \\ \mathbb{C}\_{D} = \left(0.6 + 0.07 \mathcal{U}\_{10}\right) \times 10^{-3} & 6 \frac{m}{s} \le |\mathcal{U}\_{10}| \le 26 \frac{m}{s} \end{array} \tag{2}$$

Wind stress and wind stress vorticity fields were used to speculate on the link between the wind variations and the seasonal and/or interannual variability of mesoscale structures.

The monthly means of the AGV and optimal current fields were used to estimate the relative vorticity (ζ), defined as the vertical component of the velocity field curl, as follows:

$$
\zeta = \frac{\partial V}{\partial \mathbf{x}} - \frac{\partial \mathcal{U}}{\partial y}; \tag{3}
$$

where *U* and *V* are the velocity components. The resulting current vorticity fields were spatially averaged in the regions of the main mesoscale structures listed in Table 1 and filtered (13 month moving average) in order to remove the seasonal and intra-annual variations.

#### **3. Results**

#### *3.1. Mean Currents and Wind Fields*

The qualitative comparison between the ADT and drifter data shows that the two datasets fit rather well (Figure 2a) and allow us to update the pseudo-Eulerian current maps described in Reference [9] and to enhance the schematic circulation maps published by References [1,3] summarized in Figure 1b. The averages were made over different periods in accordance with the availability of data, 1993–2018 for the drifter and altimetry data (Figure 2a), 2012–2016 for the optimal currents (Figure 2b), and 1993–2016 for the wind (Figure 2c,d). The time periods are rather long and the statistics are rather robust to consider the average in Figure 2a comparable with those in Figure 2c,d and representative of the mean conditions in the SC. The optimal currents are available over a reduced period and are, therefore, not directly compared with the other datasets. Rather, they are used to bring out some aspects of the current field that are not obvious, using only drifters and altimetry data. The structures emphasized in white in Figure 2a,b, are described here for the first time or with different shapes and positions with respect to those schematized in Figure 1b.

The Atlantic Tunisian Current (ATC) originates from the branch of the Algerian Current (AC) that enters in the SC and flows southward between Pantelleria Island and the Tunisian coast [9]. It shows a complex pattern only partially described by the previous model studies. Indeed, the ATC splits in two branches at about 36.5◦ N (Figure 2a). One branch continues to move southward along the Tunisian coast (shown in white color), whereas another branch moves eastward south of Pantelleria Island. At about 35.5◦ N and 13◦ E, the ATC splits another time. A part of the current forms the Bifurcation Atlantic Tunisian Current (BATC) and the other part turns southward towards the Libyan coast at ~13◦ E. The latter branch describes the Atlantic Libyan Current (ALC [1,3]), which moves westward towards the Gulf of Gabes and eastward along the Libyan coast (Figures 1b and 2a).

Drifter and altimetry data confirm the well-known meandering pattern of the Atlantic Ionian Stream (AIS) and outline the edge of the Maltese Channel Crest (MCC), located on the Maltese Plateau. The Medina Gyre (MG) is located on the west and/or southwest side of the Malta Island in agreement with the scheme of Reference [3]. The region where the ATC splits and forms the BATC is characterized by a sudden reduction of depth due to the shelf extension (see the location of MG in Figure 1a), which probably facilitates the cyclonic rotation of the surface currents around 35◦–36◦ N and 13◦–14.5◦ E. It is interesting to note that Reference [1] located the MG in a different position southeast of Malta (see Figure 2 of [1]). South of the MG, altimetry and drifter data detect another permanent cyclonic mesoscale structure that has never been described before in the literature; hereafter we will refer to this structure as the Southern Medina Gyre (SMG) (emphasized in white Figure 2a). The Libyan Shelf Break Vortex (LSBV) is well described only by altimetry data (Figure 2a) because of a scarce quantity of drifter tracks along the Libyan coast. This structure appears meridionally elongated and squeezed along the Libyan coast, showing a different shape and location with respect to Reference [1].

**Figure 2.** Mean drifter currents (**a**) in spatial bins of 0.25◦ × 0.25◦ (blue vectors) superimposed on mean maps of absolute dynamic topography (colors) during the period from 1993 to 2018. Mean optimal currents (**b**) in spatial bins of 1/12◦ (vectors; one vector every two grid points is plotted) superimposed on mean maps of absolute dynamic topography (colors) during the period from 2012 to 2016. The structures emphasized with white arrows in panels (**a**,**b**) are new or with different shapes and positions with respect to those described in Figure 1b. The structures already known are highlighted with pink arrows. (**c**) Mean map of the wind stress amplitude (colors) and direction (vectors) and (**d**) wind stress curl over the period from 1993 to 2016.

The higher spatial resolution of the mean circulation derived by optimal currents (2012–2016; Figure 2b) permits a more detailed description of the mesoscale structure of the SC and the southern

Tyrrhenian Sea. The anticyclonic structure, clearly visible along the northern coast of Sicily and located between the Aeolian Islands and Cape Gallo (38.1–38.5◦ N; 13.5–15◦ E), was observed by Reference [25] in September 2012, but it was not described by these authors; hereafter it will be defined as the Northern Sicily Anticyclone (NSA). Another mesoscale anticyclone is located at the entrance of the SC (37.5–38◦ N; 11.5–12◦ E) and represents a kind of watershed between the waters entering the Tyrrhenian Sea (Bifurcation Tyrrhenian Current—BTC) and those entering the SC; hereafter we will define it as the Sicily Channel Anticyclone (SCA). A third mesoscale anticyclone is squeezed between the MG and Malta Island (35.5◦–36◦ N; 14◦–14.5◦ E) and we will define it hereafter as the Southern Maltese Anticyclone (SMA). Similarly, southeast of the Ionian Shelf break Vortex (ISV), there is another steady cyclonic structure hereafter defined as the Southern Ionian Shelf break Vortex (SISV). In addition, the optimal currents reveal new information on the shape of the Sidra Gyre (SG), which appears as a large anticyclone that involves two smaller anticyclonic structures (Figure 2b) and confirms the existence of the SMG (Figure 2b). The cyclonic circulation in the region of the Adventure Bank Vortex (ABV; 37–37.5◦ N; 12–13◦ E) is detected by the three datasets (drifter, altimetry, and optimal currents) but the vortex is not well resolved by any of them. It appears more like a cyclonic meander rather than a vortex (Figure 2, upper panels).

In the study area the mean wind stress is oriented to the east in the southern Tyrrhenian Sea and to the southeast in the SC (Figure 2c) with amplitude (range of values between 0 and 0.04 Nm<sup>−</sup>2) and directions in agreement with the results of Reference [12]. The regions mostly impacted by the wind stress are located in the band north of 35◦ N and south of the southern Sicily coast (Figure 2c). The rotating motion induced by the wind (wind stress curl Figure 2d) is cyclonic (positive) along the southern and eastern Sicily coasts and on the Malta plateau, whereas it is anticyclonic (negative) along the Tunisia coast and in the southern Tyrrhenian Sea (along the northern Sicily coast). We can speculate that the wind field plays an important role in shaping the sub-basin circulation (e.g., the branch of the ATC that moves southward along the Tunisia coast) and in defining the sense of rotation of the mesoscale structures located around the Sicily coast (e.g., the ABV, the Messina Rise Vortex (MRV), the ISV, and the NSA).

#### *3.2. Seasonal Variability of Currents and Wind Fields*

The seasonal variability of the drifter-derived and optimal current fields is shown in Figure 3, together with the altimetry data. Results substantially confirm the paths described in Reference [9], but add new insights. The BTC is a permanent feature, as shown by Reference [1], and it is stronger in winter (Figure 3b,d; winter speeds larger than 20 cm/s; mean summer speeds of ~10 cm/s), whereas the NSA is much more intense in summer (Figure 3a,c; mean winter speeds of ~5 cm/s; mean summer speeds of ~10 cm/s). Along the northern coast of Sicily, the drifter data describe a westward current during summer, with the consequent inflow of surface water in the SC (Figure 3a), whereas the coastal currents move eastward during winter (Figure 3b). This summer westward current (mean speeds of ~10 cm/s) was already described in Reference [3] and defined as the Tyrrhenian Sicilian Current (TSC). The TSC is not identified by the optimal currents (Figure 3c,d).

The ATC is part of the permanent pattern of the SC, in agreement with Reference [3], and it is more intense and more meandering in nature during the extended winter (Figure 3b,d; maxima winter speeds of ~30 cm/s; maximum summer speeds of ~20 cm/s). It is interesting to note that Reference [1] described it as a winter structure. The cyclonic Pantelleria Vortex (PV) is observed only during the extended winter (Figure 3b,d; speeds of 10–15 cm/s), in agreement with the optimal currents and with Reference [3] and in disagreement with Reference [1], which describes this as part of the permanent pattern. The MCC is stronger during the extended summer (speeds larger than 15 cm/s), as well as the AIS (Figure 3a,c; mean winter speed of ~5 cm/s; mean summer speeds of ~15 cm/s). The SG is stronger and larger in size during summer (Figure 3a,c; mean winter speeds of ~7 cm/s; mean summer speeds of ~12 cm/s; winter longitudinal extension of ~100 km; summer longitudinal extension larger than 250 km), in agreement with the results of Reference [26], which report the SG seasonal expansion in the summer and contraction in the winter. The ISV and the MRV are permanent structures, in agreement with Reference [1], and they are stronger in the winter (Figure 3b,d; mean winter speeds of ~9 cm/s; mean summer speeds of ~6 cm/s). In addition, the SISV is a permanent structure more intense in the winter, according to the optimal currents (Figure 3d; mean winter speeds of ~10 cm/s; mean summer speeds of ~5 cm/s). The BATC is predominant during winter south of the MG (Figure 3b,d; maxima winter speeds of ~20 cm/s; maxima summer speeds of ~10 cm/s), in agreement with Reference [1]. The ALC is stronger in the winter (Figure 3b,d; speeds larger than 15 cm/s), whereas the LSBV is stronger in the summer (Figure 3a,c; speeds larger than 25 cm/s).

**Figure 3.** Pseudo-Eulerian drifter statistics (blue vectors) superimposed on mean maps of the absolute dynamic topography (colors) for the (**a**) extended summer and (**b**) extended winter. The schematic circulation structures superimposed on the current fields are colored with red arrows (**a**) when they are most intense in summer and with light blue arrows (**b**) when they are most intense in winter. Mean optimal currents in spatial bins of 1/12◦ (blue vectors: one vector every two grid points is plotted) superimposed on mean maps of absolute dynamic topography (colors) during the period from 20 12to 2016 for the (**c**) extended summer and (**d**) extended winter.

The wind stress and the wind stress curl are more intense in winter (Figure 4), influencing the seasonal variability of some sub-basin currents, such as the BTC, the ALC, the BATC. The winter intensification of the BATC and its interaction with the topography lead to an intensification of the MG. The wind stress does not impact the circulation structures, which are more intense in summer (MCC, AIS, NSA, SG), when the wind stress and the wind stress curl are weakened. The seasonal variability of these structures is presumably related to other forcings, such as the instability of the surface currents and the interaction with the topography. The behaviour of the AIS and of the MCC confirms the following insight: The AIS is stronger in the summer (Figure 3a) when the wind stress is weaker (Figure 4a) and its meander on the Malta Plateau (the MCC) is anticyclonic, although the vorticity induced by the wind in this region is substantially cyclonic (Figure 4c). Along the northern coast of the Sicily, the amplitude of the wind stress and the anticyclonic vorticity induced by the wind are stronger in winter (Figure 4), whereas the strengthening of the NSA is observed in the summer (Figure 3a,c). This result suggests that the wind influences the sense of rotation of the surface circulation in the NSA, but other forcings modulate the strength of the seasonal and interannual variability of this circulation structure. The summer intensification of the LSBV appears to be instead related to the intensification of the cyclonic wind stress curl along the western Libyan coast (Figure 4c). The SG shows a pronounced longitudinal extension in summer (Figure 3a), when the wind stress is weaker but the vorticity induced by the wind is essentially anticyclonic in the southeastern region of the SC (Figure 4a). Its southern margin is oriented parallel to the Libyan coast following the 400 m isobath.

**Figure 4.** Mean map of the wind stress (upper panels) amplitude (colours) and direction (vectors) and (**d**) wind stress curl (lower panels) over the extended summer (**a**,**c**) and extended winter (**b**,**d**).

#### *3.3. Decadal Variations*

Decadal variations are emphasized by removing the mean ADT and AGV (1993–2018) fields from the interannual composite mean over the time periods characterized by the anticyclonic (1993–1996, 2006–2010, and 2016–2017) and cyclonic (1997–2005 and 2011–2016) circulation modes in the northern Ionian (Figure 5). In the region located between Pantelleria and Malta islands (35.5◦ N–37◦ N, 11◦ E–15◦ E), the surface currents are smaller than the mean field during the anticyclonic mode (Figure 5a; current anomalies are oriented in an opposite direction with respect to the mean field) and larger during the cyclonic mode (Figure 5b). Along the northern coast of the Sicily, the NSA is reduced in intensity, with respect to the mean field, during the anticyclonic mode (Figure 5a; the current anomalies are oriented cyclonically) and increased during the cyclonic mode.

**Figure 5.** Mean maps of absolute geostrophic velocity anomalies (vectors) superimposed to the absolute dynamic topography anomalies (colours) during (**a**) anticyclonic (1993–1996, 2006–2010, 2017–2018) and (**b**) cyclonic (1997–2005 and 2011–2016) circulation modes in the northern Ionian.

The largest variations are observed east of 15◦ E, in agreement with the results of Reference [24]. The MRV and the SISV are less intense than the mean currents during the anticyclonic mode (Figure 5a) and slightly more intense during the cyclonic mode (Figure 5b). The AIS tends to be deflected towards the northern Ionian during the anticyclonic circulation mode (northeastward currents along the Sicily eastern coast) and the MIJ is reduced in intensity with respect to the mean currents (Figure 5a). During the cyclonic circulation mode, the AIS feeds the MIJ, which shows larger intensities with respect to the mean, and the currents are mainly directed southwestward along the western coasts of the Ionian Sea (Figure 5b).

#### *3.4. Interannual Variability and Vertical Structure of the Quasi-Permanent Mesoscale Eddies in the Sicily Channel and Southern Tyrrhenian Sea*

The main quasi-permanent mesoscale eddies of the SC and southern Tyrrhenian are analyzed in terms of their interannual variability, using the time series of spatially averaged vorticity fields derived both from altimetry and optimal currents data. In the period in which the optimal currents are available, the accuracy of the vorticity derived by the AGV is generally improved, showing larger complexity in the temporal variability of the signal (see Figure 6, Figure 8, and Figure 11). The thermohaline

properties and the vertical extension in the water column of these mesoscale eddies is studied using the Argo float profiles.

**Figure 6.** Time series of the spatially averaged, low pass filtered (13 month) vorticity field over the regions of the NSA and SCA. Dashed-dotted lines refer to the vorticity field derived from the AGV. The continuous lines are related to the vorticity field derived from optimal currents from 2012 to 2016. The dashed black lines show the average values of the vorticity over each anticyclonic/cyclonic period of the Northern Ionian Gyre.

#### 3.4.1. Southern Tyrrhenian Sea and Sicily Channel Entrance

The analysis of the vorticity field in the areas of the NSA (38.1–38.5◦ N; 13.5–15◦ E) and SCA (37.5–38◦ N; 11.5–12◦ E) confirms that the anticyclonic nature of these regions persists with time (Figure 6). A more accurate analysis of the temporal evolution of the vorticity in the NSA shows quasi-decadal variations of the intensity of the vorticity field that coincide with the inversions of the surface circulation in the northern Ionian. The black dashed lines in Figure 6 give an indication of the mean vorticity values during each anticyclonic/cyclonic circulation mode. The anticyclonic vorticity of the NSA is reduced during the anticyclonic circulation modes of the northern Ionian (1993–1996, 2006–2010, 2017–2018), whereas it is enhanced during the cyclonic circulation modes (1997–2005, 2011–2016). This result supports the relationship between the large–scale interior ocean variability in the central Mediterranean Sea and the local dynamics, suggested by Reference [27]. More specifically, these authors suggest a link between the inversions of the surface circulation in the northern Ionian and the local tidal observations in the area of the Strait of Messina. The present work shows that not only the Strait of Messina, but all the coastal areas adjacent to the northern Sicily coast can be influenced by the variability attributed to the large-scale dynamics of the central Mediterranean.

The vertical structure of the NSA is defined by the profiles of the float WMO 6900981, which circulated on the border of the NSA between late April 2012 and early January 2013 (Figure 7a,b). This float shows that the NSA extends about 50 m in depth and confirms its anticyclonic nature with a reduction of density and the deepening of the isopycnal surfaces, in particular between the end of June 2012 and November 2012, when the float profiles were close to the core of the eddy. The trajectory of the float WMO 6900981 gives an indication of the intermediate current displacements at the parking depth (350 m; see Table 2). It is interesting to note that, during the period covered by the float WMO 6900981, the intermediate currents in the region of the NSA flowed in an opposite direction (cyclonic displacements; Figure 7a), with respect to the mean surface currents (Figure 2b). The diagram in Figure 7b shows a gap of the float profiles between 1 September 2012 and 20 October 2012. Despite this gap, the float remains confined to the eastern part of the NSA, from which it moves away only in December 2012. Unfortunately, we have no floats entrapped in the SCA.

**Figure 7.** Map of the trajectory (black line) and profile positions (First profile: Magenta dot; Other profiles: Black dots) of the float WMO 6900981 superimposed on the mean map of the ADT (between 18 April 2012 and 3 January 2013) (**a**) and contour diagram of the potential density versus depth and time (**b**).

#### 3.4.2. Malta Plateau

The surface circulation in the region of the Malta Plateau is strongly influenced by the flow of the AIS that forms a large anticyclonic meander defined as MCC. The permanent anticyclonic vorticity in this region (36.1–36.8◦ N; 14–15◦ E) is confirmed by the time series of the vorticity field (Figure 8). At local scale, this meander can sporadically create an anticyclonic gyre on the Malta plateau [28].

**Figure 8.** Time series of the spatially averaged, low pass filtered (13 month) vorticity field over the regions of the Malta plateau. Dashed-dotted lines are referred to the vorticity field derived from the AGV; continuous lines are related to the vorticity field derived from optimal currents from 2012 to 2016.

In the framework of the Italia-Malta Calypso Project [28], about 38 drifters were deployed on the Malta plateau between December 2012 and September 2016. These drifters were captured by the anticyclonic gyre on the plateau in two specific deployments, December 2012 and March 2014 (Figure 9). From December 2012 through January 2013, drifters were captured by the anticyclonic gyre for about four weeks (approximately between 14 December 2012 and 10 January 2013) before being transported out of the plateau (Figure 9a). This gyre is also confirmed by the trajectory of the float WMO 6901044 (Figure 10a) and by HF radar measurements [29,30]. On 22 March 2014, six drifters were deployed in the area and were trapped in the gyre for about 2 weeks (Figure 9b). Drifters allow for estimating the radius and the rotation period of the structure considering the centroids computed from all the closed loops of the drifter trajectories. The radius spanned between 11 and 27 km and the period increased from 4.1 to 8.4 days, coinciding with an increased distance from the centre.

**Figure 9.** Trajectories and deployment positions (black dots) of the drifters deployed on the Malta Plateau in December 2012 (**a**) and March 2014 (**b**).

The float WMO 6901044 was entrapped in the MCC between 15 December 2012 and 17 February 2013, then it joined a cyclonic structure located north-west of the Malta Plateau (Figure 10a). It had a cycling period of 1 day, and its trajectory gives an indication of the 350 m displacements (see Table 2). In the first part of its tracks, the float sampled the interior of the anticyclonic MCC, showing a deepening of the isopycnal surfaces (December 2012–January 2013) and reduced densities. In February 2013, the float moved along the border of the MCC, then it was entrapped in the cyclonic structure located in the north-western proximities of the Malta plateau (Figure 10b). Both the MCC and the cyclonic structure show a similar vertical structure, extending to a depth of about 200–250 m (Figure 10b).

**Figure 10.** Map of the trajectory (black line) and profile positions (First profile: Magenta dot; Other profiles: Black dots) of the float WMO 6901044, superimposed on the mean map of the ADT (between 16 December 2012 and 30 April 2013) (**a**) and contour diagram of the potential density versus depth and time (**b**).

#### 3.4.3. South of Malta

The time series of the vorticity fields obtained at the locations of the mesoscale eddies south and south-west of Malta are shown in Figure 11. The MG (35◦–35.5◦ N; 13◦–14◦ E) and the SMA (35.5◦–36◦ N; 14◦–14.5◦ E), two adjacent structures located south-west of the Maltese Islands, show larger variabilities of the vorticity field, with respect to the SMG (34◦–35◦ N; 13◦–14◦ E) and LSBV (33◦–33.5◦ N; 12.5◦–14◦ E). This behaviour is probably related to the wind-stress, which is more intense in the Malta region than in the southern SC (Figure 2c). The vorticity of SMA increased with time over the considered period.

**Figure 11.** Time series of the spatially averaged, low pass filtered (13 month) vorticity field over the regions of MG, SMA, SMG, and LSBV. Dashed-dotted lines refer the vorticity field derived from the AGV. The continuous lines are related to the vorticity field derived from optimal currents from 2012 to 2016.

The float WMO 6903242, became entrapped in the anticyclonic SMA in mid-September 2018 and described five loops around the eddy core before being captured by the eastward BATC (Figure 12a). It had a short cycling period (see Table 2) and its trajectory indicates the mean near surface displacements (0–180 m). The subsurface density distribution clearly shows the net differences between the water masses located east (potential density smaller than 24.8 kg/m3) and south (potential density larger than 26 kg/m3) of the of SMA (Figure 12b). The SMA extends down to a depth of about 40 m (Figure 12c). At the end of October 2018, the float left the cyclonic structures and moved eastward, encountering surface waters of eastern origin and denser than 1026 kg/m3.

The float WMO 1900629, coming from the Libyan coast, was entrapped in the cyclonic SMG at the end of 2007 (Figure 13a). This structure is denser than the surrounding waters (Figure 13b). Since, in the region of the SMG, the maximum depth of the sea bottom is about 400 m (Figure 1a), and since the cyclonic trajectory of the float WMO 1900629 (Figure 13a) represents the displacements of the currents at the parking depth of 350 m (see Table 2), we can conclude that the cyclonic structure affects the entire water column in this area.

The cyclonic LSBV is sampled by the float WMO 1900948 during the period from October 2015 to February 2016 is shown in Figure 14a. The entrance of this float in the LSBV is emphasized in Figure 14b by a higher density, compared to the surrounding waters, and by changes in the shape of the isopycnal surfaces. Even if float WMO 1900948 has a parking depth of 1000 m (see Table 2), its displacements represent the current at 350 m depth, due to the bathymetry of the LSBV region (see Figures 1a and 14b). We can conclude that the cyclonic structure affects the entire water column in this area (Figure 14b).

**Figure 12.** Map of the trajectory (black line) and profile positions (First profile: Magenta dot; Other profiles: Black dots) of the float WMO 6903242, superimposed on the mean map of the ADT (between 12 September 2018 and 11 November 2018) (**a**). Map of the potential density measured at 20 m depth (**b**) and contour diagram of the potential density versus depth and time (**c**).

**Figure 13.** Map of the trajectory (black line) and profile positions (First profile: Magenta dot; Other profiles: Black dots) of the float WMO 1900629, superimposed on the mean map of the ADT (between 27 August 2007 and 23 February 2008) (**a**) and contour diagram of the potential density versus depth and time (**b**).

**Figure 14.** Map of the trajectory (black line) and profile positions (First profile: Magenta dot; Other profiles: Black dots) of the float WMO 1900948 superimposed on the mean map of the ADT (between 19 July 2015 and 28 February 2016) (**a**), and contour diagram of the potential density versus depth and time (**b**).

#### 3.4.4. Ionian Cyclones

Figure 15a shows the time series of the vorticity field along the eastern coast of Sicily (MRV and ISV area—36.5◦–38◦ N; 15◦–16◦ E) and in the region of the SISV. Figure 15a shows an inconsistency between the vorticity derived from AGV and those derived from the optimal currents in the region of the SISV. From the optimal current validation carried out by Reference [15], this structure lies in an area where the method degrades the quality of the surface currents (in particular of the meridional

component), when compared with the AGV (see Figure 10 of [15]). Therefore, the optimal currents could not be consistent in the description of SISV.

**Figure 15.** (**a**) Time series of the spatially averaged, low pass filtered (13 month) vorticity field over the regions of the MRV-ISV and SISV. Dashed-dotted lines refer to the vorticity field derived from the AGV. The continuous lines are related to the vorticity field derived from optimal currents from 2012 to 2016; (**b**) time series of the monthly, spatially averaged low-pass filtered (13 month) wind-stress and (**c**) wind stress vorticity in the MRV region.

The MRV and ISV are wind-driven structures [7,24] and their interannual variability is related to the wind-stress along the eastern coast of Sicily (Figure 15b,c), e.g., lower values of current vorticity in the MRV-ISV regions are related to lower wind-stress and lower wind stress curl. The SISV (35.5◦–36.6◦ N; 15◦–16◦ E) is not influenced by the wind-stress vorticity, whereas it appears to be influenced by the quasi-decadal reversal of the northern Ionian (see black dashed lines in Figure 15a that give an indication of the mean vorticity value during each anticyclonic/cyclonic circulation mode) in the period from 1993 to 2010. After 2010, the vorticity of the currents was no longer consistent with the decadal variability and, rather, seems to be linked to some other phenomena that are not currently detectable from our datasets.

The float WMO 1900954, coming from the eastern Ionian, was entrapped in the SISV in December 2017 (Figure 16a) and showed an increase of density with respect to the water located north of this mesoscale structure (Figure 16b). Even if the float WMO 1900954 has a parking depth of 1000 m (see Table 2), its displacement represents the current at about 500 m depth, due to the bathymetry of the SISV region (see Figures 1a and 16b). The vertical extension of the SISV is about 100 m (Figure 16b).

**Figure 16.** Map of the trajectory (black line) and profile positions (First profile: Magenta dot; Other profiles: Black dots) of the float WMO 1900954, superimposed on the mean map of the ADT (between 20 October 2016 and 7 March 2017) (**a**) and contour diagram of the potential density versus depth and time (**b**).

#### **4. Discussion and Conclusions**

The main findings of this work are summarized in Figure 17, where a schematic diagram of the surface circulation, based on the mean circulation map depicted in Figure 2b, is presented. In Figure 17a the main circulation structures are classified according to their seasonal variability, whereas in Figure 17b they are identified based on the main forcing factor that determines them.

**Figure 17.** Schematized representation of the mean surface circulation in the Sicily Channel and Southern Tyrrhenian Sea (black arrows), based on the mean optimal current circulation map depicted in Figure 2b (bright grey vectors). (**a**) Black arrows emphasize the permanent sub-basin and mesoscale structures; dashed red/blue lines emphasize the seasonal summer/winter structures, respectively; red/blue arrows are superimposed on the black arrows when the structures are permanent but most intense in summer/winter, respectively. (**b**) Sub-basin and mesoscale structures are classified according to the mechanisms that drive them, wind forcing (orange arrows), the interaction between currents and topographical forcings (light green arrows), and large-scale internal processes (light purple arrows).

The basin scale circulation is essentially oriented according to the wind stress direction (northwest–southeast; Figure 2a,c and Figure 17). Most of the exclusively main sub-basin scale current systems and mesoscale structures are permanent but affected by a strong seasonal variability (Figure 17a). Only the TSC and the PV show a seasonal incidence and occur in summer and winter, respectively. The BTC, ATC, BATC, and ALC show a winter intensification (Figure 17a) concurrent with the intensification of the wind stress (Figures 3 and 4). For this reason, they are classified as wind-driven features in Figure 17b. The relationship between the sub-basin scale structures and the wind stress variability was already suggested by References [13,31]. Other sub-basin scale and mesoscale structures, e.g., the meandering AIS, the SG, and the MCC, are stronger in summer, when the wind stress is weaker. Hence, we exclude the direct impact of the wind on the variability of these structures. Rather, we suggest that they are influenced by other forcings, such as the instability of the surface currents and/or their interaction with the complex and relative shallow bottom topography (Figure 17b), characterized by continental shelves and channels.

The MG and SMA are located in a highly dynamic region (high level of Eddy Kinetic Energy, see Figure 6 of Reference [24] and Figure 9a of Reference [11]) characterized by the split of the ATC, whose eastward branch forms the BATC, and by a sudden reduction of depth (Figure 1a). These

factors probably interact in facilitating the eddies formation and retention. From these considerations, the MG and SMA are influenced by the surface current dynamics and/or topography (Figure 17b). They are stronger in winter because the sub-basin currents, involved in their formation (ATC and BATC), are most intense and energetic in this season (Figure 17a).

The ABV, the MRV, and the ISV are located along the southern and the eastern coast of Sicily where the wind stress curl is steadily cyclonic (Figures 2d and 4c,d). Their strength and their interannual variability are influenced by the temporal evolution of the wind stress amplitude and vorticity (Figures 15 and 17b).

The SMG is located on the margins of the African shelf break (Figures 1a and 2a,b) and its vertical extension affects the entire water column (maximum depth of 400 m), as documented by float WMO 1900629 (Figure 13). The wind stress is weak and dimly cyclonic in this region (Figure 2 lower panels). Therefore, we define the SMG in Figure 17b as strongly influenced by the interaction between the surface currents and the topography.

The mesoscale structures located in regions where the influence of the wind stress is lower can be forced by the large-scale internal variability of the ocean. An example of this interaction is the SISV, whose interannual variability is connected to the decadal variability of the surface circulation in the northern Ionian (Figure 15a). However, the most representative example of the influence of internal forcing on the variability of a mesoscale structure is the NSA. This anticyclone, although located in the southern Tyrrhenian Sea, is affected by the decadal variability induced by the adjacent Ionian Sea (Figure 6). This result, as suggested by Reference [27], opens to a new interpretation of the link between different Mediterranean sub-basins and underlines the importance of internal processes on the variability of the mesoscale structures.

In summary, the surface circulation in the SC and southern Tyrrhenian Sea is characterized by multi-scale spatial and temporal variability. The main spatial scales involved are the basin, sub-basin, and mesoscale. The main temporal scales involved are the seasonal, interannual, and decadal scales. In this work, the complexity of the SC current system was investigated by combining different in-situ data and satellite products. Results provide an updated picture of the surface circulation, detecting new mesoscale features and describing their temporal variability and strength in relation to the main external and/or internal forcings. The winter strengthening of the wind stress directly influences most of the structures stronger in this season. The structures stronger in summer and/or located in high dynamical regions are mainly driven by the instability of the surface current and/or by their interaction with the bottom topography. In the regions where the influence of the external forcings is weaker, the large-scale internal variability of the adjacent Mediterranean basins can influence the local dynamics.

**Author Contributions:** Writing—original draft preparation, M.M.; writing—review and editing, P.-M.P., D.C., A.D. (Andrea Doglioli), G.N., and R.G.; investigation, M.M. and P.-M.P.; data curation, M.M., G.N., R.G., D.C., and M.-H.R.; formal analysis, M.M., G.N., and R.G.; resources, D.C., M.-H.R., A.G., and R.S.; funding acquisition, P.-M.P. and A.D. (Aldo Drago).

**Funding:** This research was mainly funded by the Italian Ministry of Education, University and Research as part of the Argo-Italy program, and partly funded by the Italia-Malta Programme—Cohesion Policy 2007–2013, European Union Regional Development Funds (ERDF) through the CALYPSO and CALYPSO FO projects. The program MISTRALS of CNRS funded the drifters deployed during the PEACETIME cruise (Guieu C., Desboeufs K., 2017, RV Pourquoi pas? https://doi.org/10.17600/17000300).

**Acknowledgments:** The authors would like to thanks all the people who have deployed drifters and made their data available in the Mediterranean Sea in the period 1993–2018. We acknowledge Antonio Bussani for his technical support and his work in the production of the Mediterranean drifter dataset, and Elena Mauri for her constructive comments. We thank the two anonymous reviewers for their constructive comments on the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
