Impact of moderate energetic ﬁne-scale dynamics on the phytoplankton community structure in the western Mediterranean Sea

. Model simulations and remote sensing observations show that ocean dynamics at ﬁne scales (1–100 km in space, day–weeks in time) strongly inﬂuence the distribution of phytoplankton. However, only few in situ samplings have been performed and most of them in boundary currents which may not be representative of less energetic regions. The PROTEVSMED-SWOT cruise took place in the moderately energetic waters of the western Mediterranean Sea, in the southern region of the Balearic 5 Islands. Taking advantage of near-real time satellite information, a sampling strategy was deﬁned in order to cross a frontal zone separating different water masses. Multi-parametric in situ sensors mounted on the vessel, on a towed ﬁsh and on an ocean glider were used to sample at high spatial resolution both physical and biogeochemical variables. A particular attention was put in adapting the sampling route, in order to also estimate the vertical velocities in the frontal area. Such a strategy was successful in sampling quasi-synoptically an oceanic area characterized by the presence of a narrow front with an associated 10 vertical circulation. A multiparametric statistical analysis of the collected data identiﬁes two water masses characterized by different abundances of several phytoplankton cytometric functional groups, as well as different contents in chlorophyll a and O 2 . Our study shows that the Lagrangian fronts induced by the ﬁne-scale circulation, even if much weaker than the fronts occurring in boundary current systems, maintain a strong structuring effect on phytoplankton community by segregating different taxa at the surface. we can clearly identify the separation between the two types of AW at Author contributions. RT post-processed the in situ observations, performed the analysis of the results and leaded the writing of the 410 manuscript. FD, AMD, GG, PG collected the in situ data. AAP, SB and FdO provided on land support to the sampling strategy. LI and MTh carried out the analysis of ﬂow cytometry data. AP and BBL contributed to the vertical velocity analysis. FC, NB and MTe conducted the glider deployment and the processing of its data. All the authors discussed the results and contributed to the writing of the manuscript. interests. The authors

m depth. ii) a temperature sensor SBE 38 installed at the entry of the water intake.
An automated CytoSense flow cytometer (CytoBuoy b.v) was installed on board and connected to the seawater circuit of the TSG, to perform scheduled automated sampling and analysis of phytoplankton (Thyssen et al., 2009(Thyssen et al., , 2015. The instrument contains a sheath fluid made of 0.1 µm filtered seawater which stretches the sample in order to separate, align and drive the individual particles (i.e. cells) through a light source. This light source is made of a 488 nm laser beam. When the particles cross the laser beam, they interact with the photons. Several optical signals are recorded for each single particle: the forward angle light scatter (FWS) and 90°side-ward angle scatter (SWS), related to the size and the structure (granularity) of the particles. Two signals of fluorescence induced by the light excitation were also recorded, a red fluorescence (FLR) induced by chlorophyll a and an orange fluorescence (FLO) induced by the phycoerythrin pigment. Two distinct protocols have been run sequentially every 30 min, in order to process the 1164 samples. The first protocol (FLR6) has a FLR trigger threshold 135 fixed at 6 mV and is able to analyze a volume of 1.5 cm 3 . It was dedicated to the analysis of the smaller phytoplankton.
For instance, Synechococcus were optimally resolved and counted with this protocol. The second protocol (FLR25), targeted nanophytoplankton and microphytoplankton with FLR trigger level fixed at 25 mV and an analyzed volume of 4 cm 3 . The data were acquired thanks to the USB software (Cytobuoy b.v.) but analyzed with the CytoClus software (Cytobuoy b.v.).
The combination of the various variables recorded by the flow cytometer exhibits various clusters of particles (cells), which 140 abundance (cells per cubic centimetre) and average variable intensities are provided by the CytoCLUS software. The later generates several two-dimensional cytograms (e.g. Fig. 12, see section 3.3 for explanation of the group identification) of retrieved information from the 4 pulse shapes curves (FWS, SWS, FLO, FLR) obtained for every single cell.
On 5 May 2018, a SeaExplorer glider, manufactured by Alseamar (codename: SEA003), was deployed at sea. After a short transit, it performed a route approximately parallel to the NS hippodrome (red track on Fig. 1). This glider, set to dive down 145 to 650 m depth, was equipped with a pumped conductivity-temperature-depth sensor (Seabird's GPCTD) from which the conservative temperature (Θ), the absolute salinity (S A ) and the density anomaly referenced to the surface (σ 0 ) were derived using TEOS-10 toolbox (McDougall and Barker, 2011). This GPCTD was also equipped with a dissolved oxygen (O 2 ) sensor (Seabird's SBE-43F) to measure oxygen concentrations. The glider also embarked a WET Labs ECO Puck FLBBCD for measurements of i) [Chla] fluorescence (targeting excitation and emission wavelengths at λEx/λEm: 470/695 nm), converted into Chla concentrations (in microgram per litre), ii) backscattering at 700 nm (BB700), and iii) a FDOM fluorophore, namely the humic-like fluorophore or peak C in the Coble (1996)'s classification (λEx/λEm: 370/460 nm), expressed in microgram per litre equivalent quinine sulfate units (microgram per litre QSU). Finally, the SeaExplorer was also equiped with two MiniFluo-UV fluorescence sensors (hereafter called MiniFluo) for the detection of various FDOM fluorophores (Cyr et al., 2017(Cyr et al., , 2019. In this study, the MiniFluo-1 was used for the detection of tryptophan-like fluorophore (λEx/λEm: 275/340 155 nm), while the MiniFluo-2 was used for the detection of tyrosine-like fluorophore (λEx/λEm: 260/315 nm). Tryptophan-and tyrosine-like fluorophores, referred to as peak T and peak B, respectively, in the Coble (1996)'s classification, are amino acidlike components commonly found in the marine environment, and generally associated with autochthonous biological processes (see review by Coble et al. (2014)). Here, fluorescence intensities of tryptophan-and tyrosine-like fluorophores are provided in relative unit (RU) and are not converted into mass concentration (microgram per litre) (Cyr et al., 2017(Cyr et al., , 2019. Glider 160 observations were processed with the Socib glider toolbox (Troupin et al., 2015) for cast identification and geo-referencing.

Vertical velocity estimation
The vertical velocity has been diagnosed solving the so-called quasi-geostrophic (QG) omega equation (Hoskins et al., 1978;Tintoré et al., 1991;Allen and Smeed, 1996, Eq. (1)): where w is the vertical component of the velocity field and Q is the vector determined by horizontal derivatives of water density and horizontal velocity (Hoskins et al., 1978;Giordani et al., 2006, Eq. (2)): with V g the geostrophic horizontal velocity vector, ρ the density, ρ 0 a reference density equal to 1025 kg m −3 , g the gravi- tational acceleration, f the Coriolis parameter (considered constant and computed at the mean latitude of the area), and N 2 the 170 Brunt-Väisälä frequency. The QG theory is valid for low Rossby numbers, a condition that is satisfied in this study.
High-resolution in situ data are necessary to solve Eq. (1). In this work, σ is obtained from Seasoar CTD measurements ( Fig. 2) and the geostrophic component of the horizontal velocity has been estimated from the measurements performed with the VMADCP as in the work of Barceló-Llull et al. (2017). Following Allen et al. (2001)'s suggestions to preserve as much as possible synopticity, four transects of the NS hippodrome (see Fig. 1) have been selected between 11 May and 12 May 2018, to 175 obtain a "butterfly" design as in Cotroneo et al. (2016) and Rousselet et al. (2019). The σ, u and v fields have been interpolated onto a 3D grid, using objective analysis (Le Traon, 1990;Rudnick, 1996). The horizontal grid resolution is 0.9 km × 0.9 km and the vertical resolution is 6 m (from 19 m to 253 m depth). We have followed the method described by Rudnick (1996), using the scripts freely downloadable from his web page at the address http://chowder.ucsd.edu/Rudnick/SIO_221B.html, last access: smaller scale variability and instrumental error. The fluctuation part of the field's statistics is assumed to have a decorrelation length scale of 20 km in both the x and y directions, with a structure orientation of 18.4°from north. The correlation length scale has been chosen by analyzing the auto-covariance matrix of the σ field and performing several sensitivity tests. The noise-to-signal ratio is assumed to be 0.05 as in Rudnick (1996). The interpolated fields are shown superposed to the in situ measurements in Fig. 2. The ageostrophic component of the velocity measured by the VMADCP is then removed and Eq. (1) 185 is solved with an iterative relaxation method and constrained by Dirichlet boundary conditions (w = 0) as in the case of the front studied by Rudnick (1996). To minimize the effect of the imposed boundary conditions, only data with an error on the objective mapping of σ ≤ 0.0025 have been then considered.

190
At first, we describe the hydrodynamic conditions encountered during PROTEVSMED-SWOT in order to characterize the area. For simplicity, two transects only, representative of each of the two hippodromes, are presented: a first transect, on the WE hippodrome, performed from 9 May 16:50 to 23:45 UTC is referred to as the WE transect (Fig. 3a). The second one performed on the NS hippodrome, from 11 May 02:00 to 08:40 UTC, is referred to as the NS transect (Fig. 3b) for the corresponding date. The intensity and the direction of the current vary along the transects. Around 38°N 20', a zonal fine scale feature, slowly evolving, is present in the altimetry-derived FSLE field and confirmed by the VMADCP data. At this latitude, the current direction changes drastically along the NS transect. The WE transect shows a larger current variability than the NS transect, due to its alignment with the fine scale structure. The FSLE features and the variation of current direction are likely induced by the presence of a fine scale structure, in this case a front. The intensities of these vertical motions, ranging from 2 to 8 10 −5 m s −1 (corresponding to 1.7 to 6.9 in meter per day), are stronger in the intermediate layer.

Hydrology and biogeochemistry
The typical southwestern Mediterranean water masses are observed in the Θ-S A diagrams of the Seasoar CTD data of the WE and NS transects ( Fig. 5a and Fig. 5c  of the two types of AW, using the same color code as in the Θ-S A diagrams. Interpretation of these data is the following. The surface layer is occupied by Atlantic Waters (AW) with different residence times in the Mediterranean Sea. We refer to them as "younger AW" (in light blue) and "older AW" (in dark blue). The "younger AW" corresponds to AW entered more recently in the Mediterranean basin and is characterized by a salinity between 37 g kg −1 and 38 g kg −1 , while the "older AW" is characterized by a higher salinity. Indeed, this AW has been modified during its circulation in the Mediterranean Sea: its 225 salinity has increased due to evaporation and mixing. Other authors refer to this water as "local AW" (Barceló-Llull et al., 2019) or "resident AW" (Balbín et al., 2012). The "younger AW" is located at the west and at the south of the WE and NS transects, respectively. Moreover, the separation between the two types of AW is in agreement with the localisation of the front identified by the FSLE and the change in current direction ( The SeaExplorer glider has also performed temperature and salinity measurements along a transect parallel and slightly west of the NS hippodrome (see Fig. 1). The Θ-S A diagrams of the glider data ( Fig. 6a and Fig. 6b) confirm the presence of the two surface water masses mentioned above, in particular during the outward route ( Fig. 6a). During the glider return route (Fig. 6b) the surface water masses begin to be more homogeneous than during the outward route and the NS Seasoar transect (Fig. 5c).

235
This fact can be explained because the transects completed with the Seasoar were realized within a few hours, while the glider transects lasted several days. Moreover, the deeper glider sampling has allowed to detect another thermohaline signature, with temperature values about 13°C and salinity ranging around 38.5 g kg −1 , which corresponds to the Western Mediterranean Deep Water (WMDW) as found by Balbín et al. (2012).
The vertical extension of the surface and intermediate water masses can be observed plotting the Seasoar data (conservative 240 temperature, absolute salinity and density) as vertical sections along the WE (Fig. 7) and NS transects (Fig. 8). The warm surface layer with temperature greater than 15°C extends until about 100 m. This layer is also characterized by a salinity between 37.5 g kg −1 and 38 g kg −1 and, as a consequence, by the lowest density. Below 100 m depth there is the intermediate water layer. Looking at gradients along the transects we can clearly identify the separation between the two types of AW at Picophytoplankton, are located at the western and southern parts of the front, along with the WE and the NS transects. On the other side of the front, their abundances are lower, (≤ 1 10 4 cells cm −3 for Synechococcus and ∼ 2 10 3 cells cm −3 for 280 Picophytoplankton). Microphytoplankton abundances vary between 8-18 cells cm −3 ( Fig. 13c and Fig. 14c) and present an opposite distribution with the previous groups. Indeed, higher abundances are found in the eastern and northern parts of the front. Nanophytoplankton ( Fig. 13d and Fig. 14d) and Cryptophytes ( Fig. 13e and Fig. 14e) ranged from 650-1000 cells cm −3 and 10-30 cells cm −3 respectively. However, these latter exhibited a less clear pattern between the two sides of the front. The distribution of the abundances of phytoplankton between the front (specially observed for Synechococcus, Picophytoplankton 285 and Microphytoplankton) fits well with the hydrodynamic and the hydrological observations.

Statistical analysis
Surface temperature and salinity data measured with the TSG are merged with the data of abundances of the 9 different groups of phytoplankton at each cytometry sampling point along the WE and NS hippodromes. Thus, the final data set consists of 11 variables, each variable containing 215 observations.

290
The principal component analysis (PCA) consists of summarizing the information contained in the data set by replacing the initial variables with new synthetic variables (called principal components), linear combinations of the initial variables and uncorrelated two by two. Applied to our data set, the PCA points out that the first three components explain respectively 36.7 %, 18.1 % and 13 % of the total variance of the data. The following statistical analysis treats these three components, representing 67.8 % of the total variance of the data. The K-medoid algorithm, described by Hartigan and Wong (1979); Kaufman and Rousseeuw (1987), is an other method to 300 represent the various aspects of the data structure. This algorithm divides M points in N dimensions into K groups (or clusters).
A cluster is an object for which the average dissimilarity to all the data is minimal. In our case of study, the K-medoid algorithm divides the 215 points into three clusters in two dimensions (Fig. 15b). Each point of a cluster shows a high degree of similarity with the others points of the same cluster. The three clusters are well separated. Only a few points of the black and red clusters are difficult to disentangle. Finally, the average of each variable, called local average, has been calculated for each cluster, as 305 well as the global average, to show the contribution of each variable to a cluster. The most discriminating variables for each cluster (Table 3) have also been determined with the standard deviation. Figure 15c is a spatio-temporal representation of the three main clusters obtained by the K-medoid algorithm. The WE hippodrome from 8 to 10 May is characterized by the presence of the black cluster in the east and the red cluster in the west.
The NS hippodrome starts the 11 May with the red cluster present in the south and the black cluster in the north. The latter 310 remains dominant in the north for the remaining of the sampling. In the south, the red cluster is gradually replaced by the blue cluster, except a few points on the 13 May.
In Fig. 15d  Given the ephemeral nature of these latter processes, the temporal evolution of the distribution of the red and blue cluster is 320 probably due to the displacement of the frontal structure.

Discussion and conclusion
For the case of energetic open ocean regions, like boundary current systems, model-based analysis, remote sensing and some in situ studies have all shown that the fine scales can modulate the phytoplankton growth and diversity (e.g., d 'Ovidio et al., 2010;Barton et al., 2014;Clayton et al., 2014;Mahadevan, 2016;Lévy et al., 2018). The typical mechanisms suggested 325 for this coupling effect is a barrier effect at fronts associated to intense vertical velocities. However, large part of the ocean is characterized by weaker circulation, and hence by weaker transport barriers and vertical velocities. The impact of these moderately energetic fine-scale structures on phytoplankton communities has not yet been fully evaluated. In particular, their role on phytoplankton diversity deserve closer study. In this work, we have considered a front in the south of the Balearic Islands, in order to provide a view of the physical forcing occurring in this structure and its effect on the distribution of 330 phytoplankton groups.
In Fig. 16, we have summarized the findings of this study. The strong variation of current direction observed in the horizontal velocities measured by VMADCP data and the analysis of the ocean color images combined with the altimetry-derived FSLE field have allowed to identify a frontal area located at the latitude of about 38°N 30' and between 3°E and 4°E of longitude.
The data from the CTD sensors mounted on the Seasoar towed fish and on the SeaExplorer ocean glider identify a rapid 335 shift between two different types of surface water masses on the Θ-S A diagrams ( Fig. 5 and Fig. 6). These water masses are considered AW at different stages of its circulation path in the western Mediterranean basin. Indeed, AW entering the Mediterranean through the Strait of Gibraltar is forced by the combination of the Coriolis effect and the presence of coasts to form an anticlockwise circulation along the continental slope of the western Mediterranean (Millot, 1999;Millot et al., 2006).
South of this basin, the circulation is dominated by the Algerian Current (AC). The AC can become unstable due to baroclinic 340 and barotropic instabilities (e.g., Millot, 1999). Then, meanders of AC, that can also form mesoscale eddies, spread over the basin and join the region south of Balearic Islands. There, the recently entered AW encounters the surface water coming from the north after its complete cyclonic circulation around the western Mediterranean basin. In the studied region, the presence of this water has already observed by Balbín et al. (2012)  During the cruise, the satellite-derived surface [Chla] showed contrasted values between the south-west and the north-east 360 of the studied area (Fig. 1). Thanks to the flow cytometry measurements, we identified several groups of phytoplankton that showed contrasted abundances across the front, in particular the two main groups: Synechococcus and Microphytoplankton. This confirms the results of previous modelling (e.g., Lévy and Martin, 2013;Lévy et al., 2015Lévy et al., , 2018, remote-sensing (d' Ovidio et al., 2010) and in situ studies (Clayton et al., 2014), suggesting that frontal structures can act as barriers in the surface ocean, structuring the spatial distribution of phytoplankton. We observed high abundances of Synechococcus and pi-365 cophytoplankton south of the front, while Microphytoplankton is more abundant north of it. In the Balearic Sea, Mena et al. (2016) already found higher abundances of Synechococcus in the "new" AW than in the "resident" one. In our study, thanks to In conclusion, we can say that our results, obtained thanks to an adaptive Lagrangian strategy and high-resolution coupled physical-biological sampling provide an in situ confirmation of the findings obtained by previous modelling studies (e.g., Lévy et al., 2018) and remote sensing (d 'Ovidio et al., 2010) about the structuring effect of the fine-scale ocean dynamics on 385 the community structure of the surface phytoplankton community. Furthermore, our results are consistent with the work of Clayton et al. (2014Clayton et al. ( , 2017 in an analysis of the phytoplankton accross the Kurushio front, and confirm that ubiquitous and less energetic fronts than those found in boundary currents have a similar impact on the diversity of phytoplankton. In the future, a better understanding of the biogeochemical processes generating this observed fine-scale physical-biological coupling is needed. In particular, we plan to estimate the growth rate of the phytoplankton cells thanks to the data collected in 390 Lagrangian manner and applying a method similar to the one by Marrec et al. (2018). Moreover, the role of nutrient supply and of zooplankton grazing are key factors in explaining the differences in abundances of the different phytoplankton groups.
To address these latter points, future cruises will need high-resolution (and also high-precision, considering the oligotrophy of the Mediterranean Sea) nutrient measurements coupled with zooplankton sampling and dedicated experiment about its grazing on the different phytoplankton groups. Finally, this study also shows how the satellite information is extremely useful for 395 the design of the cruise sampling strategy and, then, for the on-shore post-cruise interpretation of the data. The new satellite SWOT (https://swot.jpl.nasa.gov/, last access: 10 February 2021 ; https://swot.cnes.fr, last access: 10 February 2021) will provide ocean topography and surface current at an unprecedented resolution (Morrow et al., 2019). In particular during the few months after its launch, the period called "fast sampling phase", the satellite will be on a special orbit that will overfly a portion of the global ocean, called "cross-overs", where high spatial resolution will be associated with high temporal resolution 400 (d' Ovidio et al., 2019). Thus, the large data set collected during PROTEVSMED-SWOT represents precious new information on the cross-over area located in the south of the Balearic Islands. Furthermore, our work paves the way to future cruises, planned in this area which will be a great opportunity to study in more detailed way the physical-biological fine-scale coupling.

405
The script for objective mapping can be openly available on Rudnick's web page: http://chowder.ucsd.edu/Rudnick/SIO_221B.html, last access: 10 February 2021. B and Peak T of "older AW" are significantly higher than those of "younger AW" (t test, p inf 0.0001).