The Mediterranean Sea (MS) consists of two basins, the Western Mediterranean (WMS) and the Eastern Mediterranean (EMS), connected by the Sicily Channel (SC), and is characterized by a limited water exchange with the Atlantic Ocean. In the MS the net evaporation exceeds the precipitation, driving an anti-estuarine circulation through the Strait of Gibraltar, which has important implications on the physical and biogeochemical properties of the MS. The thermohaline circulation of the MS is described as an open basin-wide cell, which involves the gradual salinification of the fresher surface Atlantic water (AW) while propagating eastward (Gačić et al., 2013), and its transformation into intermediate water (IW), called Levantine Intermediate Water (LIW) or Cretan Intermediate Water (CIW), depending on the specific formation area (Levantine or Cretan Sea, CS). IW spreads westward and its core is identifiable in the whole MS by a subsurface salinity maximum (S_max, Schroeder et al., 2017). It flows into the Ionian Sea (IS), with a significant flow northward to the Adriatic Sea (AdS), and constitutes an important preconditioning agent for the formation of both the Adriatic deep water (AdDW) and Mediterranean deep waters (eastern, EMDW and western, WMDW) (e.g. Roether et al., 1996; Gačić et al., 2013). Some authors have highlighted the presence of two main modes of circulation (cyclonic and anticyclonic) of the North Ionian Gyre (NIG), interpreted in terms of internal processes (e.g. Borzelli et al., 2009; Gačić et al., 2010; Gačić et al., 2011; Gačić et al., 2014; Theocharis et al., 2014), which, alternating on a multiannual scale, operate a different redistribution of heat and salt in the central and Levantine area of the MS. Furthermore, the quasi decadal reversal of the NIG influences the strength of the Mid-Ionian Jet (MIJ; Robinson et al., 2001) and generates cascading repercussions on the physical and chemical characteristics of the water masses in the EMS, as well as in the SC and in the WMS (Menna et al., 2019). A process to explain such reversal is a feedback mechanism, named Adriatic-Ionian Bimodal Oscillation System _BIOS_ (Civitarese et al., 2010; Gačić et al., 2010), driven by the difference in salinity between the saltier and warmer waters originating from EMS (LIW and Levantine Surface Water, _LSW_), and the less saline AW entering from the SC (Gačić et al., 2011; Gačić et al., 2014). Briefly, an anticyclonic phase (1993–1997 and 2006–2010) inhibits the AW advection (weakening the MIJ) to the southern Levantine basin and favours the production of salty and warm LIW/CIW. Conversely, a cyclonic phase (1998–2005) favoring the AW advection (strengthening the MIJ) to the southern Levantine Basin, leads to a stronger dilution in the EMS (Schroeder et al., 2017 and internal references). In particular, the anticyclonic mode represents a preconditioning mechanism for the dense-water formation processes in the Aegean Sea and eventually for Eastern Mediterranean Transient (EMT) like events (Demirov and Pinardi, 2002; Gačić et al., 2011). During these events it has been showed that, under favourable conditions, the dense water formation area can switch from the AdS to the CS, with a consequent change of the Eastern Mediterranean Deep Waters (EMDW) properties and a repercussion on the entire thermohaline cell (e.g. Gačić et al., 2013). The effects related to EMT and BIOS also show that the MS is not in a steady-state and that in the last decades it has been affected by a continuous increase in temperature and salinity (e.g. Lacombe et al., 1985; Rixen et al., 2005; Millot, 2007; Borghini et al., 2014; Schroeder et al., 2017), probably due to the effects associated with global warming. Salinity and temperature positive trends, especially in the IW, also show a specific interannual variability (e.g. Gasparini et al., 2005; Gačić et al., 2013; Bonanno et al., 2014; Schroeder et al., 2017), whose effects migrate from one to another basin and can be observed mainly in the SC with a time delay respect to the formation period of about 8 years, based on the data of the transient tracer (Roether et al., 1998), or 10-13 years on the basis of the salinity anomalies (Gačić et al., 2013). In this context, the SC can be considered as a key point for the comprehension of the phenomena underlying the thermohaline circulation at basin scale, as well as of the effects of this variability on the biogeochemical characteristics of the water masses that pass through it.

Biogeochemically, the MS is an oligotrophic basin despite being almost entirely surrounded by land with high nutrient loadings (e.g. Béthoux et al., 1998; Krom et al., 2004) derived from a large coastal population. The main source of dissolved organic phosphorus (P) and nitrogen (N) entering the WMS and EMS, primarily via the Straits of Gibraltar and Sicily, are mineralized to phosphate (PO₄) and nitrate (NO₃) and subsequently exported out of the basin by the prevailing anti-estuarine circulation (e.g. Powley et al., 2017). The very low productivity of the MS is therefore linked both to the anti-estuarine circulation (Krom et al., 2010) and to the chemical speciation of the dissolved P and N. They in fact reflects a switch from less bioavailable chemical forms of P and N entering the MS to more bioavailable forms leaving the MS (Powley et al., 2017). The anti-estuarine circulation also contributes to the higher primary productivity of the WMS respect to the EMS, as it acts on the lateral transport of dissolved inorganic nutrients in the deeper water, from the EMS to the WMS. It follows that deep waters (DW) of the WMS are characterized by higher NO₃ and PO₄ concentrations than those of the EMS (e.g. Ribera D'Alcalà et al., 2003) and carried up through upwelling phenomena in the photic zone. Another peculiarity of the MS is the higher molar NO₃:PO₄ DW ratios (21:1 on average in the WMS to 29:1 within the EMS, Ribera D'Alcalà et al., 2003) respect to the global ocean (N:P=16:1, Redfield et al., 1963). The debate on the anomaly of this peculiarity is still open, although recent analysis suggest that for the EMS, the high N:P ratio of the external inputs, together with very limited denitrification (Krom et al., 2010; Huertas et al., 2012; Van Cappellen et al., 2014), explains the very high NO₃:PO₄ DW ratio (Powley et al., 2017).

The aim of this study is to examine the interannual variability of thermohaline and biogeochemical properties of the water masses flowing through the SC in summer during the period since 2010 to 2020. We analyzed the potential mechanism controlling the long-term variability of the water masses properties and evaluated the impact on nutrient concentrations and on the productivity of the marine ecosystem.

Oceanographic data have been acquired over 10 oceanographic surveys (named Bansic) carried out in the SC during the summer period from 2010 to 2020 (except for year 2019) aboard the Urania, Minerva Uno and Dallaporta vessels (Figure 1 and Table 1). For this study, in-situ sampling design was focused on an inshore-offshore transect approximately perpendicular to the Sicilian coast, composed by 10 stations spaced about 12 NM, comparable to the Ross by deformation radius (Sorgente et al., 2011).

While the surface temperature is able to provide the almost instantaneous signature of interactions with the atmosphere, which is by its nature subjected to great variability, the analysis of the variability of the heat and salt content in the deeper water masses provided a climate index more suitable for monitoring any changes over long periods. For this reason, we have integrated the intermediate and deep layer dataset with the previous ones carried out by Bonanno et al., 2014 and Ben Ismail et al., 2014 (Figure 6). Regarding the IW layer, time series presented in this study are in agreement with comparable time series collected in the SC by Gasparini et al., 2005; Gačić et al., 2013; Ben Ismail et al., 2014; Bonanno et al., 2014 and Schroeder et al., 2017. Specifically, it is thus possible to highlight a general increasing trend and the presence of “cyclic patterns” in temperature and salinity, that could be associated with the alternation of different circulation modes (cyclonic and anticyclonic) of the NIG in addition to IW formation processes. In particular, we infer an average annual increase of dT/dt ~0.047°C yr^-1 and dS/dt ~0.01 yr^-1 from the 1998-2007 (Figure 6), comparable to those estimated by Gasparini et al. (2005) and Schroeder et al. (2017). Such a trend of T and S could be related to the anticyclonic phase of the NIG estimated from 1988 to 1997 (mean ADT available from 1993 to 1997, Figures 7A, E), according to Bonanno et al., 2014. The anticyclonic circulation inhibits the AW advection to the southern Levantine basin, favouring the production of saltier and warmer IW, whose effects were observed in the SC with a variable time lag that deserves further studies and insights. The period 1998-2007 is followed by an abrupt reversal trend up to 2010, characterized by a dT/dt of -0.12°C yr^-1 and dS/dt of -0.01 yr^-1 (Figure 6). These trends, in agreement with Gačić et al., 2013; Ben Ismail et al., 2014; Bonanno et al., 2014; Schroeder et al., 2017, could be linked to the previous cyclonic phase of the NIG (1998-2005, Figures 7B, E). It favoured the strengthening of the MIJ and thus the AW advection to the southern Levantine basin, leading to an increasing of the water masses dilution in the EMS, according to Schroeder et al., 2017. Approximately, a new salty/warm phase occurs from 2011 until 2017 (Figure 6), probably associated to the previous anticyclonic phase of the NIG (2006-2010, Figure 7C). Compared to the first average annual increase in θ and S, attributable to the anticyclonic phase 1993-1997, this increasing trend shows considerably higher rates of dθ/dt and dS/dt, despite being related to a shorter and less defined anticyclonic circulation. Indeed, some authors identified the 1993-1997 phase as “fully anticyclonic” due to the absence or weak activity of the MIJ. Differently, during the subsequent anticyclonic phase (2006-2010), the MIJ occurs with an intense, non-negligible, zonal activity from the SC to the Levantine basin, prompting authors to distinguish this period as an “anticyclonic-zonal” phase (Bessières et al., 2013). Therefore, our considerably higher rates of increase in θ and S can be only partially associated with the anticyclonic-zonal phase of the NIG, from which a lower rate of increase would be expected. A further possible explanation could be linked to the increasing of the drying process (E-P), affecting the surface waters from which LIW and CIW originate, in agreement with Schroeder et al. (2017).

The last period of the time series (2018-2020, Figure 6) showed a new decreasing trend significantly lower than the previous one (i.e. 2008-2010). This last reversal trend followed the cyclonic phase occurred in 2011-2016 (Figures 7D, E). The comparison of the ADT averages observed during the two cyclonic phases (Figures 7B, D) allowed to evidence a significant difference in term of intensity and areal extension of the NIG. In the first phase the NIG extends down to 35°N while in the second phase is confined at northern latitudes, higher than 36.5°N, probably due to the presence of an anticyclonic structure in the centre of the IS (Cardin et al., 2015; Placenti et al., 2018). This structure has weakened the AW flux toward Levantine Basin (Cardin et al., 2015), limiting the influence of colder and fresher AW toward the area where the formation of the IW take place. This interpretation is supported by recent studies that found a premature inversion of the NIG from cyclonic to anticyclonic, related to the extremely strong winter 2012 in the Adriatic (Gačić et al., 2014) leading to a strong weakening of the AW flux toward Levantine Basin between mid-2012 and the first months of 2013 (Menna et al., 2019). This phenomenon probably produced, between 2012 and 2013, warmer and saltier IW in the EMS which could have modulated the decreasing rate of θ and S in the SC.

The correlation between the NIG circulation mode and the temperature and salinity trends observed in the SC was investigated through a cross-correlation analysis that show a significant correlation for time lag of 6 (Cramer’s V = 0.456, p <0.05) and 7 years (Cramer’s V = 0.497, p <0.05), highlighting an association between cyclonic phase-negative trends and between anticyclonic phase-positive trends. However, we recognized a temporal variability between the reversal of the circulation mode and the trend inversion that suggest the presence of unadressed additional mechanisms that tune the onset of the signals and encourage the need of further studies.

Moreover, the time series of θ and S of the DW in 2010-2020 has been investigated and showed similar trends to the IW, but characterized by lower increments (Figure 3). However, it is important to note that most of the data collected in this study regarding the DW are referred to the upper part of this water mass, due to the limited depths that characterize the SC. Regarding this water mass, integrating the θ and S data discussed in Bonanno et al. (2014) and Ben Ismail et al. (2014) with our time series, it is possible to detect a general increase in the last twenty years of about 0.4°C of θ and 0.1 of S, respectively (Figure 8). However, the time series showed two strong break points in 2006 and 2009 characterized by an abrupt decrease of θ and S that were also discussed by Ben Ismail et al. (2014) regarding the 2006, and by Bonanno et al. (2014) regarding both years. In addition, these authors found a simultaneous increase of density (σ) in the DW, associating these findings to the presence of tEMDW in the SC. It is confirmed by our data also in 2010 (not shown), while in the following years (2011-2017) a new warm and salty phase seems to affect the deeper waters of the SC (Figures 6 and 8). We also hypothesize that these trends concerning the deeper layers are a direct response to the climate change, in particular in terms of net evaporation over to the eastern basin, i.e. where the deeper water is formed (Schroeder et al., 2017).

The coupling of the nitrates and silicates time series (2010-2020) with the salinity in the deeper waters of the SC showed anti-phase trends (Figure 5). These trends have been observed in IW both in the EMS (Ozer et al., 2017) and in the AdS (Borzelli et al., 2009; Civitarese et al., 2010). In agreement with Ozer et al. (2017), we think that a possible explanation could be related to the surface circulation dynamic of the NIG.

Specifically, during the anticyclonic phase (2006-2010, Figure 7C) the influx of AW, considered the main external source of P and N to the Mediterranean system (Powley et al., 2017), towards the Levantine Basin was limited, producing saltier and poorer nutrient IW. Conversely, the slight decrease in salinity and increase in nutrients observed between 2017 and 2020 (Figure 5) could be linked to the previous cyclonic phase of the NIG (2011-2016, Figure 7D).

In addition, the interannual fluctuations affecting the nutrient concentration has been compared to the variability of the Chl-a concentration, which is considered a good proxy of the phytoplankton biomass and provides useful indications on the productivity of the upper layer of the water column (e.g. Falkowski et al., 2004). Interestingly, our findings show similar trends among these time series, especially taking into account the nitrates and silicates concentration of the deeper layers. Indeed, starting from 2010, these time series were characterized by variable fluctuations and reached a simultaneous minimum peak in 2015, then followed by an increasing trend until the 2020. These observations suggest that vertical oceanographic structures, such as upwelling processes identified along the Sicilian coast (Lermusiaux and Robinson, 2001; Bignami et al., 2008; Patti et al., 2010; Torri et al., 2018), could play a significant role in the diffusion of nutrients from the lower layers, making them available for the photosynthetic activity with a cascading effect on the marine organisms that in the Central Mediterranean Sea found a suitable habitat for feeding and reproduction (e.g. Bonomo et al., 2018; Patti et al., 2020; Russo et al., 2021). Indeed, it could also be noted that the SC is a place of strong diffusion and dissipation that may favor such vertical transport (Ferron et al., 2017; Vladoiu et al., 2018). However, discordant findings regarding this relationship emerged between our study areas and the Levantine basin, in which an opposite trend has been evidenced between the nutrient concentration of the IW and the Chl-a concentration (Ozer et al., 2017), suggesting the need of further studies on a wider scale. In this context, the long-time series in the Strait of Gibraltar and Sicily, which represent key points of the entire Mediterranean thermohaline circulation, are of fundamental importance for climate studies. It is necessary to continue maintaining long-term observational points to improve our understanding of the physical mechanisms that drive climate variability and to discriminate the effects related to natural variability and the anthropogenic effect. Similarly, in these key areas, the acquisition of nutrients high-quality data on a multiannual scale are essential for understanding the relationships between oceanographic, biogeochemical and biological processes.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

FPI wrote the manuscript and analyzed the dataset. MT wrote the manuscript and prepared the figures. VT, LG participated to the fieldwork and laboratory analyses. BP, AC and GT coordinated the field work and supervised the writing of the text. RS and FPe helped with some specific aspects of the manuscript and to review the manuscript. All authors contributed to the article and approved the submitted version.

The study was supported by the research projects SSD-PESCA, coordinated by the Ministry of the Education, University and Research (MIUR) and founded by the Ministry of Economic Development (MISE), and the Flagship Project RITMARE - The Italian Research for the Sea, coordinated by the Italian National Research Council and funded by MIUR.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.