Long distance particle transport to the central Ionian Sea

In the upper layers of the Ionian Sea, young Mediterranean Atlantic Waters (MAW) flowing eastward from the Sicily channel meet old MAW. In May 2017, during the PEACETIME cruise, fluorescence and particle content sampled at high resolution revealed unexpected heterogeneity in the central Ionian. Surface salinity measurements, together with altimetry-derived and hull-mounted ADCP currents, describe a zonal pathway of AW entering the Ionian Sea, consistent with the so-called cyclonic mode in the North Ionian Gyre. The ION-Tr transect, located ~19-20°E- ~36°N turned out to be at the crossroad of three water masses, mostly coming from the west, north and from an isolated anticyclonic eddy northeast of ION-Tr. Using Lagrangian numerical simulations, we suggest that the contrast in particle loads along ION-Tr originates from particles transported from these three different water masses. Waters from the west, identified as young AW carried by a strong southwestward jet, were intermediate in particle load, probably originating from the Sicily channel. Water mass originating from the north was carrying abundant particles, probably originating from northern Ionian, or further from the south Adriatic. Waters from the eddy, depleted in particles and Chl-a may originate from south of Peloponnese, where the Pelops eddy forms.


Introduction
The Ionian Sea is a cross road where young modified Atlantic water (AW) flowing from the Sicily channel meets older MAW flowing from the East, and water coming from the Adriatic Sea (Malanotte-Rizzoli et al 1997). Circulation in the North Ionian, and thus the path of AW spreading, is variable and two alternate states have been proposed, the anticyclonic and cyclonic mode of the so called North Ionian Gyre (NIG, Gačić et al 2010). The direction of the NIG has a strong influence on the dispersal of water masses and properties in the Ionian basin, also impacting its productivity (Lavigne et al 2018)..
Although the circulation in the Ionian basin has been documented according to seasons and NIG modes (Malanotte-Rizzoli et al 1997, Gačić et al 2011, Menna et al 2019, the fine scale pathways of AW crossing Ionian Sea have only been seldom sampled. The contact zone between water masses can only be characterized through high resolution sampling, potentially revealing small scale horizontal discontinuities in properties missed by traditional coarse sampling. Ionian Sea is generally considered oligotrophic (Boldrin et al 2002), with a north-south gradient of Chl-a (d 'Ortenzio et Ribera d'Alcala 2009). However, it is not homogeneous, as distinct phytoplankton communities associated to the main water masses have been described (Casotti et al 2003). Three main phytoplankton communities were associated with water coming from Adriatic in the north west, water from the Eastern Mediterranean in the north east and AW from Sicily channel to the south. Zooplankton community was also contrasted between northwestern and eastern Ionian (Mazzocchi et al 2003).
The influence of water masses on particle distribution is less described. Presently few data are available on the horizontal distribution of particles in the Ionian Sea. These data are mostly from transmissometry, bottle derived particulate matter filtration (Rabitti et al 1994, Boldrin et al 2002, Karageorgis et al 2008 and few data are from optical devices (Ramondenc et al 2016).

Automated cytometry
Surface phytoplankton community was analysed using an automated flow cytometer (Cytosense benchtop flow cytometer from CytoBuoy b.v) installed on a dedicated continuous sampling system set up to pump surface water (at 5m depth). The Cytosense automated flow cytometer was equipped with a 120 mW, 488 nm laser beam. The volume analyzed was controlled by a calibrated peristaltic pump and each particle passed in front of the laser beam at a speed of 2 m.s -1 . The particle resolved size range varied from <1 µm up to 800 µm in width and several hundreds of µm in length for chain forming cells. The trigger to record a signal was based on pigment fluorescence. Stability of the optical unit was controlled before and after the cruise using fluorescing 2 µm Polyscience beads, and a set of silica beads of 1.0 , 2.0 ,3.0, 5.0, 7.0 µm.

Satellite data
Several remote sensing datasets were exploited using the SPASSO (Software Package for an Adaptive Satellite-based Sampling for Ocean campaigns, https://spasso.mio.osupytheas.fr/; last access: 2 Oct. 2020). Altimetry data was obtained from the AVISO Mediterranean regional product. Delayed-time L4 maps of SST Sat (Mediterranean Sea -High Resolution L4 SST Reprocessed; Nardelli et al., 2013;Pisano et al., 2016) and Chl Sat a (Mediterranean Sea Reprocessed Surface Chlorophyll Concentration from Multi Satellite observations) were retrieved for the period of the cruise, on regular grids of 0.04 × 0.04 • resolution for SST Sat and of 1 × 1 km resolution for Chl Sat a from CMEMS -Copernicus Marine Environment Monitoring Service (http://marine.copernicus.eu/). SPASSO provided several Lagrangian diagnostics of the altimetry-derived current field, such as the Finite scale Lyapunov Exponent (FSLE) computed using the algorithm of d 'Ovidio et al. (2004), in order to help in positioning the sampling transect and stations taking into account the ocean dynamics. This approach was successfully adopted during previous cruises such as LATEX , KEOPS2 (d 'Ovidio et al., 2015), OUTPACE (Moutin et al, 2017).

Lagrangian numerical simulations
In order to estimate the origins of the sampled surface water along transect ION-Tr, the ARIANE package (Blanke and Raynaud, 1997)  Environment Marine Service (CMEMS) were used, with a resolution of 1/8°, using the near real time data, validated by an accurate comparison with both the hull-mounted ADCP-data and the SVP drifters deployed during the cruise (Barrillon et al 2019). The numerical particle positions were initialized in a polygon around the ship's track on May 29 (longitude: 18-20.1, latitude: 35.4-35.9, +-0.2° in latitude). This polygon takes into account possible offset between in situ velocity and satellite derived geostrophic velocity. Given the daily time series of the surface velocity field, the ARIANE package computes numerical particle trajectories through a backward or forward integration time. The so-called "qualitative" mode is chosen: each particle trajectory can be followed step by step through the integration time. Only surface trajectories were considered. 10000 particles chosen randomly inside the launch area were advected backward for one month. This duration was chosen as a trade-off between source identification and error accumulation along the trajectories. The particle trajectories and final positions were analyzed to determine the main source areas.
The cruise data are available on LEFE-CYBER database.

Oceanographic context
Time series of satellite Chl-a showed that surface phytoplankton concentration was decaying since mid-March (figure 2).
Since the end of April, the Ionian Sea was steadily warming while Chl-a concentration remained low (0,09 mg.m -3 ) indicating the absence of significant surface production events. On average over April-May, maximum Chl-a concentrations were observed in the area South of the Strait of Messina and in the Gulf of Taranto (not shown).
At the end of May in the Ionian basin, satellite imagery (figures 3A and 3B) shows a slight large scale meridional gradient of SST and Chla, with warmer and low-Chl water to the south and colder and medium Chl-water to the north. In addition, a strong variability was observed at the mesoscale with cold and Chl-rich filaments south of the Strait of Messina and south of Sicily, extending toward the center of the basin. Several cyclonic and anti-cyclonic structures were visible in the center of the basin.
The along route salinity showed a general increase toward the east and the north. On the northern part of the route (SAV-ST7-ION), there was a transition from medium salinity to the north to low then high salinity at the easternmost end of the route close to long duration station ION. On the southern part of the route (ION to ST8), there was an alternation of low and high salinity water, sampled during ION-Tr transect. Then west of 18°E, salinity was stable lower than 38.2 up to the Sicily channel. Therefore SAV and ST7 were in medium salinity waters, ION (ST8) was in high (low) salinity waters, while ION-Tr crossed waters with contrasted salinities. The two cytometric groups showing contrasted distribution and sufficient abundance are shown (figures 3C and 3D). In the Ionian Sea, nanoeukaryotes1 and coccolithophore like showed a distribution consistent with salinity, with higher values in saltier water except southeast of Sicily. The pattern was also mostly consistent with satellite Chl-a, with high abundances southeast of Sicily and south of the Strait of Messina. However, along ION-Tr, abundance was variable while satellite Chl-a was fairly constant.
Altimetry-derived surface geostrophic current showed a general circulation with an anticyclonic flow across the basin ( figure   4A). South of Sicily, water entered the Ionian Sea around 36°N along a southeastward path, progressed eastward along 35.5°N, then veered southwestward close to 19.5°E. East of 18°E, the currents show a pattern consistent with the intense (~0,5 m/s) currents measured by the hull-mounted ADCP along the ship route, depicting an anti-cyclonic meander, more intense on its eastern flank. There is a small spatial offset between altimetry and ADCP (~0.4°, figure 4B). The southwestward branch of this meander was located approximately in the center of the ION-Tr transect, while the ION station was located on the eastern border of the meander. Altimetry-derived FSLE field (figure 3), as well as the time series of ADT (annex A) showed that this anticyclonic meander east of 18°E was stable during May. The southwestward jet sampled during ION-Tr coincided with a slightly cooler and higher Chl-a filament than surrounding waters, as shown on figures 3A and 3B.

Distribution of properties along transect ION-Tr
Based on the salinity distribution (figure 5), ION-Tr can be divided into three parts. From West to East, a western part (noted W) with salinity >39, except the fresher top layer 0-30m, a center part (noted C) with a low salinity vein (<38.7) extending to 70m deep, and an eastern part (noted E) with quasi-homogeneous salinity on the vertical, slightly lower than in W (39).
Hereafter we define three pseudo-stations along ION-Tr, named ION-W, ION-C, ION-E. Note that ION long duration station was located very close to the eastern part of ION-Tr (figure 1), thus ION-E and ION can be approximated as a single station.
This salinity distribution coincided with pattern of ADCP currents magnitude, with ION-W and ION-E being in quiescent water (current<0.2 m/s), while ION-C correspond to a strong (>0.3 m/s) southwestward jet extending from the surface to 300m depth. The total abundance of particles, as well as Chl-a fluorescence clearly contrasted between ION-W and ION-E.
In ION-W, particle abundance was low (<5 10 4 #.m -3 ), essentially concentrated within the 20-30m layer. Fluorescence was rather homogeneous with a broad maxima around 110m. In ION-E part, abundance was much higher (>2 10 4 #.m -3 ) with a deep maximum at 100m depth (values around 1 10 5 #.m -3 ) close to the depth of the Deep Chlorophyll Maximum (DCM) and also high particle abundance in the upper 20-90m layer (~ 3 10 4 #.m -3 ). ION-C was distinct from ION-E and ION-W in term of vertical distribution and abundance of particle as well as fluorescence, with the low salinity vein coinciding with higher particle abundance and the maximum fluorescence being located ca 70m, at the interface between low salinity waters and underlying waters.
The T-S distribution of ION-Tr (figure 6) mirrored the distributions of stations ST8 and ION. Water at ION-C is similar to ST8 for densities greater than 1028 (ie deeper than 25m). ION-E is similar to ION with more scatter. ION-W does not match the other stations. Northern stations ST7 and SAV had similar water masses properties, distinct from ST8, ION and the ION-Tr profiles.

Comparison of vertical distribution across stations
Carousel measurement performed at the SAV, ST7, ST8 stations together with the one from the ION transect are shown in figure 7. These vertical profiles allow a finer analysis of the contrast between stations and among parts of ION-Tr. Consistent with figure 6, Salinity increased from ST8 to ION-E. If we take apart the upper 0-25m depth corresponding to the mixed layer (ML), couples of stations were similar: ION-C and ST8, ION-E and ION-W, ST7 and SAV. A narrow (<10m thick), lower salinity layer was present at 30m on every profile except ION-E. The surface pycnocline was located at 10 to 25m, followed by a steep thermocline down to 60-80m. Surface nitrate was depleted everywhere. Nitracline depths increased from 60m at SAV up to 90m at ION, together with nitrate concentrations below the euphotic zone. DCM was always present, but its intensity was highest at ST8 and lowest at ION-W and ION-C, where fluorescence was more diluted. The depth of DCM corresponded to the top of nitracline and to the bottom of the euphotic zone (~90m at ION).
Profiles of particle abundance were strongly variable among stations. The depth-integrated abundance was highest at SAV and lowest at ION-W. Some profiles had several maxima. At ST8 and ION-E, the main peak was just above the DCM depth, and a secondary peak was present above DCM at ~20m, also present at ION-W. SAV profile was different with a large peak at 50m, above DCM, and a surface peak at 10m.
Considering particle parameters (table 2), the slopes of the NBSS were similar for stations ST8, ION-W, ION-C (0,92), and slightly lower for SAV and ION-E (0,85), indicating more abundant larger particles. On a depth average, referring to the calibration given in Espinasse et al (2018), MEP proportion indicated that at ION-C and ION-W, particles counted were 6 0% zooplankton, while at ION-E, ST8 and SAV particles were mainly detritus (~75-90%) ie non-living particles such as aggregates produced after intense primary production. AI was similar across depth layers but higher at ION-E (0.1 against 0.05 and 0.08), while MEP was more abundant in the 50-150m layer, where the particle maxima were observed.

Origin of particles
Lagrangian simulations showed that the water parcels observed along ION-Tr originate from four main regions (figure 8).
One located south, one located west and two located north of ION-Tr. As the altimetry-derived velocity was shifted ~0.4°w est with respect to in-situ velocity from ADCP (figure 4B), the initial positions of numerical particles were extended westward to 18°E. Considering the main pathways of particles, four source regions can be defined (South origin So, West origin Wo, North origin No, Eddy origin Eo). Along ION-Tr, the proportion of each origin region showed a rather clear pattern, with some overlap in the center (figure 8F). Referring to the three parts identified along ION-Tr, ION-W was in waters mainly originating from Eo. ION-C was mainly originating from Wo but also from Eo and to a lesser extent from So.
ION-E was only originating from No. Finally, west of 18.6°E, beyond the limit of the actual transect, So was the main contributor.

Surface circulation and distribution of water masses in the Ionian Sea
Surface circulation, together with the along track surface salinity distribution, describe one pathway of MAW penetrating Cyclonic circulation of the NIG implies a downwelling of nutricline at the northern border of the NIG, and an upwelling of nutricline in the central NIG , Lavigne et al, 2018. Our observations of nutricline depth (isoline 1µM at 55m for SAV and ~95m for ST8) are consistent with the shallow nutriclines during cyclonic mode shown by Lavigne et al, (2018). Cyclonic mode also favors higher Chl-a in the center NIG, and a late winter bloom (Lavigne et al, 2018). However the Chl-a evolution over 2017 only shows a weak bloom in the Northern Ionian ( figure 2). This may result from a weak winter buoyancy loss, independent of the NIG circulation, as pointed out by Lavigne et al (2018). The last increase of Chl-a occurred in the last week of April 2017, then Chl-a decreased and reached its minimum value around the end of May, during our sampling.

Origin of water masses sampled along ION-Tr
In the center of ION-Tr, a strong current is associated to a low salinity vein (down to 38.4, figure 6). The center current dynamically separates the two parts of the transect, and this separation is stable as supported by the FSLE ridge (figure 3).
While the density stratification is similar along the transect, particle and Chla distributions show a striking difference from E to W. At ION-E, there is 60% higher particle abundance, and twice higher biovolume compared to ION-W (table 2), essentially due to particles with and equivalent spherical diameter (ESD) smaller than 500 µm. At ION-E particles are filling the pycnocline from 1028 to 1028.7 kg.m -3 , with a maximum associated to a narrow DCM at 100m, while at ION-W particles are only present in the density layer 1028-1028.2, and the DCM is broad and weak, not associated to a particle peak.
The distinct Chl-a and particle distributions along ION-Tr suggest that waters sampled at ION-E and ION-W have distinct histories of biological production. One hypothesis is that these water masses originate from distinct locations. To explore this hypothesis, backtracking of particle trajectories was carried out and led to three main origins of water sampled along ION-Tr (Wo, No and Eo, as So is only marginally contributing to the transect), that correspond to different part of the transect (figure 8). When two origins overlap, we only consider the dominant origin. Water found at ION-C would be transported from the West (Wo), ie the Sicily channel, while water found at ION-E (ION-W) would be transported mostly from the Northern Ionian (No), and from an eddy located North (Eo) respectively. Referring back to the general circulation in the Ionian, the ION-Tr appears to be at the crossroad between waters from the North Ionian, waters from the MIJ coming from the West and waters trapped into an anticyclonic eddy. Stemmann et al 2004). Therefore the particle abundance in the euphotic layer generally peaks shortly after a primary production event, then decreases with a speed depending on several factors.

Biological history of water masses sampled along ION-Tr
Although the age of particles cannot be determined from their size distribution, we can make some hypothesis. At the time of sampling, the whole Ionian Sea was stratified, mixed layer was shallow (~25m) and nitrate was fully depleted over the euphotic zone except at SAV, indicating stable, post-bloom conditions. As surface Chl-a continuously decreased since the end of April (figure 2), this suggests that the last surface primary production event around ION station occurred about a month before our sampling. This is longer than the time lag observed between primary production and subsequent export in large mesocosms experiments (Stange et al., 2017). Therefore, freshly formed particles are not likely. Second, aggregates are slowed down by density gradients, because of the equilibration time of their interstitial water (MacIntyre et al 1995, Prairie et al 2015. Therefore pronounced stratification can locally increase their abundance, as reported in numerous settings (Espinasse et al 2014, Marcolin et al 2013, Ohman et al 2012. This explains that all profiles have a particle peak at the main density gradient, just below the mixed layer depth (figure 7). However, this mechanism does not explain deeper peaks observed at 60m and 90m.
Finally, zooplankton can also contribute to the particle production through fecal pellet, and transformation (sloppy feeding), but we are lacking zooplankton abundance data for ION-C and ION-W.
We now review the potential processes that led to the observed particles and Chl-a distributions along ION-Tr, from west to east.

Western part
At ION-W, below the weak particle peak (figure 7, 2 10 4 #.m -3 ) associated with the density gradient at the bottom of the mixed layer, abundances steeply decrease and reach minimum values at 80m (1 10 4 #.m -3 ). Such a low particle abundance in

Center part
At ION-C, the backtracked trajectories indicate two overlapping origins (Eo and Wo mainly). However the low salinity vein extending down to 70m is a clear footprint of AW observed at ST8. Given the distribution of surface salinity from TSG (figure 3A and 3B) this gives more support to Wo. As with ION-E, significant particles abundance above DCM suggest that particles were produced upstream, near ST8 or further west, possibly in the enriched area south of Sicily, then advected to central Ionian.

Eastern part (ION station)
At ION-E DCM, microphytoplankton was dominant (Marañón et al 2020). The DCM was associated to the bottom of the euphotic zone (~90m at ION), but not to a density gradient. Therefore the particle peak associated with DCM probably result from aggregates produced locally, from diatoms decay and fecal pellet production by zooplankton. Association of DCM with a particle peak was not observed in similar settings (see station A in Espinasse et al 2018), which suggests that the plankton community at ION-E DCM enhances aggregate production, perhaps through mucous production, as mucous was reported in sediment traps in the Northern Ionian (Boldrin et al 2002). Above DCM, significant particle abundance (2-3 10 4 #.m -3 ) was observed together with phytoplankton dominated by nanophytoplankton (Marañón et al 2020). This significant particle abundance above DCM is also observed at SAV and ST8, with much higher peak concentration at SAV than at ST8 (1.5 10 5 #.m -3 and 3-4 10 4 #.m -3 resp). When particles are suspended or sink slowly with respect to the horizontal velocity, particle abundance act as tracer for horizontal advection (e.g. Karageorgis et al 2012, Chronis et al 2000. We argue that at stations SAV, ST8 as well as at ION-E, the abundance of particles above DCM is a remainder of past surface production event that has not been exported out of the surface layer yet. In the Northern Ionian, a POM maximum above DCM (0-50m) was also reported by Rabitti et al (1994), and attributed to relict bloom. As above DCM waters are depleted in nutrients, above-DCM particles at ION-E could be transported from an area where phytoplankton production was more intense and shallower, such as SAV according to the backward trajectories and the vertical profiles. This continuity is supported by the similar slope of the particle size spectra at SAV and ION-E (table 2). It is also supported by the higher surface abundance of cytometric counts of nanoeukaryotes and coccolithophores at SAV, along Calabria and at ION-E with respect to the other parts of the route (figures 3C, 3D). During transport from SAV to ION-E, particle abundance above DCM may have decreased because of the processes mentioned earlier.
In the upper 0-90m layer, nutrients were depleted and nanophytoplankton was dominant. Nanoeukaryotes abundances were stable during the five days of occupation of ION station, according to cytometric counts (not shown). This implies an active nutrient recycling, probably linked to the particle associated microbial community, also supported by the high bacterial production in the layer 0-100m reported by Marañón et al (2020) at ION.
ION-E particle characteristics are closest to the 'continental shelf' habitat type as defined by Espinasse et al (2014), in the Gulf of Lion in early May. The integrated particle abundance is about half (18,7 10 3 #/m 2 vs 38 10 3 #/m 2 ), the average AI is lower (0,08 vs 0,2), the %MEP is higher (6 vs 1,8), and the NBSS slope is similar but steeper (0,85 vs 0,79). Zooplankton abundance was slightly higher (300 10 3 ind/m 2 vs 206 10 3 ind/m 2 , see Feliu et al 2020, but this value may be biased high due to the combination of two mesh sizes (100 and 200µm) instead of one (120µm in Espinasse et al 2018). Thus ION-E was characterized by porous aggregates, with a rather high contribution of large aggregates (%MEP).
Taking into account Chl-a, nanophytoplankton abundances, significant particle load above DCM and high bacterial production, ION station was not strongly oligotrophic, possibly under the influence of particles transported from North Ionian.

Conclusions
High resolution sampling of water masses, including fluorescence and particle contents revealed unexpected heterogeneity in the central Ionian Sea. Surface salinity measurements, together with altimetry and ADCP currents, describe a zonal pathway of AW entering into Ionian Sea, consistent with the cyclonic mode of the North Ionian Gyre. The ION-Tr appears precisely located at the crossroad of three water masses, coming from the West, North and from an isolated anticyclonic eddy. Water mass originating from the north carried abundant particles, probably originating from North Ionian, or further from the Adriatic. Waters from the eddy were depleted in particles, and probably nutrients, and may originate from the Pelops eddy Even away from the coasts, the central Ionian Sea appears as a mosaic area, where waters of contrasted biological history meet. ION station is potentially influenced by particles transported from the North Ionian Sea. This contrast is probably amplified in spring, when blooming and non-blooming areas co-occur. Long distance particle transport appears as a significant contribution to particulate matter load, together with atmospheric input.
The small scale heterogeneity of particle abundance revealed here emphasizes the spatial decoupling between particle production and particle distribution. Such decoupling added to the time lag between production and export of particles (Stange et al 2017) may have large impact on assessing the efficiency of carbon export from the surface ocean (Henson et al 2011). This also implies that neutrally buoyant particles can sustain production away, further horizontally and to deeper depths. Interpreting the complex dynamics of physical-biogeochemical coupling from discrete measurements made at isolated stations at sea is a big challenge. The combination of multiparametric in-situ measurements at high resolution with remote sensing and Lagrangian modeling appears as one proper way to address this challenge.    1). Note that ST8 and ST8-Tr are considered in the study as one single station.