The Levantine Intermediate Water in the western Mediterranean and its interactions with the Algerian Gyres: insights from 60 years of observation

. The presence of two large scale cyclonic gyres in the Algerian basin inﬂuences the general and eddy circulation, but their effect on water mass transfer remain poorly characterized. Our study has conﬁrmed the presence of these gyres using the ﬁrst direct current measurements of the whole water column collected during the SOMBA-GE2014 cruise, speciﬁcally designed to investigate these gyres. Using cruise sections and a climatology from 60 years of in situ measurements, we have also shown the effect of these gyres on the distribution at intermediate depth of Levantine Intermediate Water (LIW) with 5 warmer ( ∼ 0.15 ◦ C) and saltier ( ∼ 0.02 g.kg − 1 ) characteristics in the Algerian basin than in the Provençal basin. The Algerian gyres also impact horizontal density gradients with sinking of the isopycnals at the gyres’ centres. Temporal cross-correlation of LIW potential temperature referenced to the signal observed south of Sardinia reveal timescale of transit of 4 months to get to the centre of the Algerian basin. The LIW temperature and salinity trends over various periods are estimated to: +0.0017 ± 0.0014 ◦ C.year -1 and +0.0017 10 ± 0.0003 year -1 respectively over the 1960-2017 period, and accelerating to +0.059 ± 0.072 ◦ C.year -1 and +0.013 ± 0.006 year -1 over the 2013-2017 period. the information concerning water mass distribution across the basin, an increase in salinity (and temperature),

We applied rigorous and systematic quality controls and corrections on temperature and salinity data, in particular for XBTs, as described by Houpert et al. (2015), allowing us to detect interannual variations at the basin scale with enough confidence. 90 Some additional quality controls, tailored for the Algerian basin, based on visual check have been applied: Correction of salinity offsets: Temperature profiles are generally consistent in the deep layers (below ∼1000 m depth) where the variability is small, and this provides good confidence in the upper measurements where the variability is higher.
However, salinity profiles can present a constant, sometimes large offset in the deep layers where variability is supposed to be similarly small. We assumed they were due to sensor calibration issues and considered as profiles subjective to 95 offset correction the ones with a density larger than 29.13 or lower than 28.98 kg/m 3 between 800 and 1500 m (density thresholds chosen outside ±4 standard deviations of natural variability in this layer of slow evolving characteristics over depth). In order to correct them, the deep part of the salinity profiles were aligned with the mean regional deep salinity.
To this end, we used reference salinity data not concerned by the previous criterion, in the 1100-1500 m range within a 100 small natural variability within a year (0.01) compared to the corrections applied (typically 0.1-0.2) .This step ensures a relatively consistent data set in salinity.
Removal of outliers: Based on climatological analysis previously published (Manca et al., 2004), and profile visualisations carried out with our data set, some profiles were considered as outliers and thus discarded if one of the following criteria applies to them: 105 -Salinity larger than 39 or smaller than 36 below 100 m; -Potential Temperature (θ) larger than 17 • C below 200 m, larger than 14 • C below 1000 m, or smaller than 10 • C; -Potential densities larger than 29.2 kg/m 3 between 0 and 2000 m, or smaller than 28.5 kg/m 3 between 400 to 1000 m, or smaller than 29.02 kg/m 3 bellow 1000 m.
These quality controls and corrections result from many iterations and represent a trade-off between measurements accuracy 110 and spatio-temporal coverage.
In addition to the potential temperature and salinity data, current measurements from SOMBA-GE 2014 research cruise  were used. For this cruise, two 300kHz Acoustic Doppler Current Profilers (ADCPs) were attached to the Rosette used to perform the CTD casts: namely LADCPs (Lowered Acoustic Doppler Current Profilers). The measured 115 currents were processed using the velocity inversion method of Visbeck (2002) implemented in the LDEO software version IX-12 (Thurnherr, 2010) with typical horizontal velocity uncertainty of 2-3 cm.s -1

Objective analysis of the LIW properties
To identify LIW, a density range between 28.95-29.115 kg.m -3 was considered (red shaded area in Fig. 2). The temperature and salinity maximum values were chosen to be representative of the LIW core characteristics for each profile.

120
To confirm that the water mass detected correspond to the LIW and not the base of the thermocline, we controlled the temperature maxima to make sure they actually correspond to an infexion point in the temperature profile.
One of the objectives of this study, is to describe a basin-scale mean repartition of LIW. To this end, we objectively analyzed the LIW salinity and temperature over the Algero-Provençal Basin. We first averaged the LIW salinity within 0.1 • x 0.1 • boxes, and then analyzed this mean field using the method of Boehme and Send (2005) with a covariance function conditional to the 125 topography and the planetary vorticity. We chose Φ=0.5 as the scaling parameter representing the influence of the topography, and a spatial correlation scale of 100 km which is consistent with the basin-scale variability we want to emphasize similarly as Bosse et al. (2015). 5 https://doi.org/10.5194/os-2021-120 Preprint. Discussion started: 14 December 2021 c Author(s) 2021. CC BY 4.0 License.

Comparison of basin-scale CTD transects
Three East-West basin scale transects acquired during research cruises in the Algerian basin were available in our data set: MEDCO08 in November 2008 , Venus1 in August 2010 (Borghini et al., 2019) and SOMBA-GE2014 in August-September 2014 . The comparable position and synoptic character of the cruise sam-150 pling allow for a direct comparison of this East-West section across the Algerian basin at these different dates over a period of 6 years (see Figure 3). Relatively warm and salty LIW extending far into the Algerian basin and away from the south Sardinian LIW vein can be observed. The signature of LIW fades away to the west as the distance from the source location, the Sardinia channel, increases, but one can identify a marked patch of LIW at about 400 km during each cruise (Fig. 3).
In addition to the information concerning water mass distribution across the basin, an increase in salinity (and temperature),  follow remarkably the f/H contours with a magnitude of about 5cm/s (blue arrows in Fig. 4).
Between 5 • and 6 • E, velocities are larger than 10 cm/s with a direction not matching the cyclonic circulation of the eastern Algerian Gyre (red ellipse in Fig. 4). This is due to the presence of a strong anticyclonic eddy at this location with a clear surface signature visible on SST and Ocean Color MODIS satellite images (see Fig. 5).      Gyres, particularly in the centre of the eastern Algerian Gyre (see contour 29.04 kg.m -3 in Fig. 7).

Temporal evolution of LIW characteristics
The evolution of potential temperature of the LIW as seen in Fig. 8 is showing an overall increase over the 1960-2017 period. 190 However, this increase does not appear monotonous. The general shape of the timeseries suggests a roughly stable warming from the sixties to the eighties, followed by a decrease until the late eighties, then a significant increase after 2012. This would indicate 4 different phases in the basin regime (hereafter, period 1, 2, 3 and 4) during the full 1969-2017 period of our study.
To estimate the trends in every phase, Θ is linearly fitted over time using least squares with a 95 % confidence interval, allowing estimates of dΘ/dt and its uncertainty from the slope of the regression slope and its error. We also indicated the correlation 195 coefficient R 2 for every regression.
In table 1, these potential temperature trends in each area were documented, during the different phases best fitted for each area.
To describe the salinity trends, the same method was used, the results are shown in table 2.
Here we will describe the evolution of the LIW temperature and salinity on average over the whole Algerian Basin for every   -1979-1987: A brutal decrease in temperature is observed in all the areas, with a strong regression coefficient( r 2 >0.8).

205
In four of the areas, the salinity data shows the same behaviour. This cooling signal is on average of -0.037 ±0.007 • C.year -1 and the freshening signal is on average of -0.0024 ±0.0081 year -1 .
One can identify the cooling event in Figure 8 as it appears first on the green curve (south Sardinia polygon) with an amplitude of about 0.3 • C, then the signal propagates to the other areas.

Transit time of LIW thermohaline signals
This section will be dedicated to quantify more thoroughly the transit time of the cooling signal observed in the 80s, using a cross correlation with a maximum lag considered of four years, of the signal between 1974 and 1992. In order to isolate this event on the time series, monthly averaged data, smoothed over four years were used.

225
In Fig. 9 the result of the cooling signal tracking across the Algerian basin, is presented. On the map, we can see in solid gray arrows, the along-slope circulation as shown in Millot and Taupier-Letage (2005b), the shear red polygons with the numbers, show in months, the time needed for the signal to travel from south Sardinia (SSar polygon) to the other areas in the Algerian basin. In about two and a half years, the LIW travels from its source all the way to the Alboran Sea region. It appears that the fastest 230 way goes from the Sardinia-Menorca polygon, SaMi (2 months) to the area between the Algerian Gyres, MAlg (4), then to the Alboran Sea ,Albo (29 months). Red arrows on the map represent a scheme of eddy-driven transport that could explain the transit times obtained from our analysis.
The signals arrive to the eastern Algerian Gyre, EAlg, after 23 months, to the western Algerian Gyre, WAlg, after 19 months 235 and to the Algerian Current polygon, AlgC, after 37 months. The area that has the largest transit time is the one south of Ibiza (SIbi), 47 months.
The dashed circular gray arrows inside the Algerian Gyres represent the recirculation process in the core of the gyres. In order to validate the results of the cross correlation analysis, a few pairs of time series segments have undergone the same analysis but for another time slot. Figure 11 shows the results obtained.
LIW potential temperature in South Sardinia have been cross correlated with the one in the Sardinia-Menorca region between 1990 and 2000, the result show that the signal needed 1 month to travel from SSar to SaMi, instead of 2 months in Fig. 9.

General circulation and LIW pathway
The results of the LADCP measurements presented in Sect. 3.1 show a current pattern that matches with the description of the Algerian Gyres done by Testor et al. (2005b) in terms of location and speed, however, the magnitude of the currents appear to be larger on the southern edge of both gyres (along the Algerian coast) and on the right edge of the eastern Algerian Gyre than on the remaining sides of the loops, suggesting that a forcing of these gyres is the general along boundary cyclonic circulation 255 of the Western Mediterranean as discussed by Testor et al. (2005b). This result confirms the existence of the Algerian Gyres in The cruise sections from west to east in the Algerian Basin (Fig. 3) revealed changes in the hydrological distribution of LIW 260 properties in the basin. Warm and salty LIW appeared to invade all the eastern Algerian Basin. The temperature and salinity climatologies of the LIW in the Western Mediterranean (Fig. 6) have also shown an influence of the Algerian Gyres on the LIW distribution. We can observe a good correspondence between the location of the 1.03 potential vorticity contour (a proxy of the Algerian Gyres) and the distribution of that warm and salty water extending further off-shore from the LIW vein. This hydrological repartition have previously been observed, but was attributed to a slow accumulation over time of LIW in the 265 interior of the Algerian basin, that remain unmixed, rather than a route of LIW crossing the Algerian basin (Millot, 1999).
However, our study suggests that a direct route of LIW crossing the Algerian basin, linked to the presence of the Algerian Gyres, is instead likely to produce this effect.
From the climatological map of potential density at 350 m (Fig. 7), we can see a sinking of the isopycnals in the Algerian

270
Gyres region. This may be the signature of numerous anticyclonic AEs, characterized by a deepening of isopycnals in their cores, circulating and accumulating in the basin.
In fact, in their study of coherent vortices in the Western Mediterranean using satellite altimetry, Escudier et al. (2016bEscudier et al. ( , a) (1993Escudier et al. ( -2012 and Isern-Fontanet et al. (2006) (1992 observed intense anticyclonic eddies being particularly aggregated in 275 the Algerian Gyres area, and appear to follow the gyres' cyclonic circulation. This was also confirmed by Pessini et al. (2018), which used 1993 to 2016 altimetry data. Anticyclonic eddies were described by Puillat et al. (2002) to be the most energetic ones, capable of lasting several months to years, looping around the Algerian Gyres, some at least for 3 years. Provenzale (1999) evidenced that these vortices induce regular Lagrangian motion inside their cores and are highly impermeable to inward and outward particle fluxes. Passive tracers can be trapped inside vortex cores for long times and are transported over large 280 distances.
In the potential temperature time series (Fig. 8), one particularly strong cooling signal from 14.1 to 13.7 • C could be identified. It was then tracked across the basin as it progressed from east to west, using a cross correlation analysis.

285
The transit time analysis has shown a preferential path to get to the Alboran Sea region by entering the Algerian basin, from the northern edge of the eastern Gyre, then further flow south-westward at the periphery of the Algerian Gyres, as illustated by the thick red arrows on Fig. 9, and as seen in the climatologies on Fig. 6. The cooling signal was chosen to perform this analysis because it represents a particularly strong signal that appeared in all the timeseries, but the conclusions on the circulation features are independent of this particular signal as they are governed by the internal dynamic of the basin. In fact, 290 it corresponds to the eddy track that was observed in the multiple studies referred to hereabove (Testor et al., 2005b;Isern-Fontanet et al., 2006;Escudier et al., 2016a, b;Pessini et al., 2018). the anticyclonic eddies in the Algerian Basin cross from east to west with the Algerian Gyres's flow. There is also a resemblance with the Sardinia Eddies' track observed once by Testor and Gascard (2005) and modeled by Testor et al. (2005a), these eddies were observed to detach from the southwestern corner of Sardinia, accumulate in the region here referred to as the Sardinia-Menorca polygon before being advected southward.

295
In the transit time analysis, the last area to get the signal was the south Balearic one, likely because this region receives most of its LIW from the along slope current of intermediate water and not much input from the most likely faster, eddy-driven cross-shelf transport. The intermediate water that gets to the south Balearic area is looping around the Northern Gyre first before being advected south, it has also been affected by the convection occurring in the Gulf on Lion area, thus, the thermo-

LIW trends
The overall aspect of the temperature time series is very similar to the Western Mediterranean intermediate layer temperature evolution from Rixen et al. (2005) documented from 1950 to 2000. An increase from the 60s to 80s, followed by a drop lasting until the 90s, then a slower increase until 2000. The regression of the full temperature time series presents some positive trends, however, the uncertainties are large (mean increase of 0.0017 ± 0.0014 • C.year -1 ) and the correlation coefficient is small (R 2 = 0.2 on average). Krahmann et al. (1998) and Rixen et al. (2005) reported the absence of a long term trend. However, positive trends in the intermediate water temperature from the sixties to the nineties have been shown by Béthoux et al. (1990) year -1 ) and Béthoux andGentili (1996, 1999) (Rohling and Bryden, 1992) and 0.0018 year -1 (Béthoux andGentili, 1996, 1999)

320
The cooling signal observed during the late 70s, and start of the 80s in our study was reported by Brankart and Pinardi (2001). They showed that the origin of the phenomenon started in the Cretan Arc region, and have linked it to the heat flux anomaly evidenced by COADS time series. Krahmann et al. (1998)   Overall, the long term evolution of the temperature time series have allowed to identify a slow increasing trend from the sixties to 2017, but helped confirm the rapidly increasing trend after 2010.
Our study provides additional evidence that the Algerian Gyres represent an important circulation feature in the basin. It appeared on the current measurements that those gyres have an impact on the circulation over the whole water column. The study of the hydrological characteristics of LIW, using in situ data, showed that its distribution across the basin is linked to the presence of the gyres. A westward, cross-shelf, eddy-driven transport of LIW from the south Sardinia vein toward the interior of the Algerian basin following the periphery of the Algerian Gyres is evidenced by the climatology of potential temperature 350 and confirmed with the cross-correlation of a particular signal.
The LIW temperature and salinity trends estimates over various periods contribute to document LIW evolution in the Algerian basin and confirm the results of previous studies. More importantly, the warming acceleration that is observed all over the basin from 2010 is alarming. A closer monitoring of water mass properties need to be sustained, it is crucial to maintain and reinforce existing surveillance systems as there is a direct impact on the regional climate and the marine resources.

355
Author contributions. KM carried out the analyses, prepared the figures and wrote the main manuscript. HLG performed the processing of the current measurements. LH compiled the multisource in situ temperature and salinity data in one homogeneous product. KM and FM updated and improved the quality of the product. AB helped with optimal interpolation analysis. PT and AB provided guidance and supervision. LM and FL provided funding and administrative coordination. All authors have conributed in providing ideas, discussing the results and reviewing the manuscript.