Short-Term Spatiotemporal Variability in Picoplankton Induced by a Submesoscale Front South of Gran Canaria (Canary Islands)

The distribution and variability of phytoplankton in the upper layers of the ocean are highly correlated with physical processes at different time and spatial scales. Model simulations have shown that submesoscale features play a pivotal role on plankton distribution, metabolism and carbon fluxes. However, there is a lack of observational studies that provide evidence for the complexity of short-term phytoplankton distribution and variability inferred from theoretical and modeling approaches. In the present study, the development and decay of a submesoscale front south of Gran Canaria Island is tracked at scales not considered in regular oceanographic samplings in order to analyze the picoplankton response to short-term variability. Likewise, the contribution of each scale of variability to the total variance of the picophytoplankton community has been quantified. We observe statistically different picophytoplankton assemblages across stations closer than 5 km, and between time periods shorter than 24 h, which were related to high physical spatiotemporal variability. Our results suggest that both temporal and spatial variability may equally contribute to the total variance of picoplankton community in the mixed layer, while time is the principal contributor to total variance in the deep chlorophyll maximum (DCM).


INTRODUCTION
As higher plants, unicellular marine primary producers' growth mainly depends on nutrient and light availability. Access to these resources may be limited in the highly dynamic oceanic environments, which are dominated by physical processes that generally alter resource availability. Indeed, a large amount of studies has indicated that the distribution and variability of phytoplankton and other biogeochemical parameters like nutrients and organic matter in the upper layers of the ocean are highly correlated in time and space with physical processes (Abraham, 1998;Mahadevan and Campbell, 2002;Lévy and Klein, 2004;Niewiadomska et al., 2008;Omta et al., 2008;Lehahn et al., 2017).
Mesoscale motions have commonly been assumed to be the most important factor modulating the distribution of biogeochemical properties at the upper levels of the ocean (Falkowski et al., 1991;Oschlies and Garçon, 1998;McGillicuddy et al., 2007;Johnson et al., 2010). However, recent theoretical studies (Lévy et al., 2001;Mahadevan and Campbell, 2002;Mahadevan and Tandon, 2006;Klein and Lapeyre, 2009) have highlighted the role played by smaller processes that operate below the local Rossby radius of deformation, referred to here as submesoscale. An estimated 50% of the total variance of vertical velocities in the upper layer of the ocean may be explained by submesoscale processes (Klein and Lapeyre, 2009). These smallscale motions arise from the disruption of the geostrophic balance by mesoscale straining being common in fronts and eddies edges. Vertical motions associated with ageostrophic secondary circulation (ASC) are originated at both sides of the fronts (upward on the warm side and downward on the cold side) leading to small-scale fluxes of biogeochemical properties like nutrients (Mahadevan and Tandon, 2006;Lévy et al., 2012a). Diapycnal mixing has been shown to be a dominant component of the vertical velocity in submesoscales fronts and filaments by destroying the thermal wind and driving intense ASC in the upper layers (Estrada-Allis et al., 2019). Thus, intensification of diapycnal mixing may enhance vertical transport of nutrients (Arcos-Pulido et al., 2014;Corredor-Acosta et al., 2020;Tsutsumi et al., 2020) as well as upwelling/downwelling of phytoplankton communities from sub-surface layers into the euphotic zone and vice versa. These physical cells act to restore the geostrophy by means of restratification in a process known as frontogenesis (Hoskins and Bretherton, 1972;Hoskins, 1982;Capet et al., 2008;McWilliams, 2016). The importance of submesoscale lies in that their spatiotemporal scales are similar to those in which biological process acts, i.e., from 0.1 to tens of kilometers and of the order of 0 (1-10) days. Phytoplankton productivity and growth may thus be influenced by those changes in nutrient and light availabilities (Allen et al., 2005;Lévy et al., 2009Lévy et al., , 2018Lathuiliere et al., 2011;Taylor and Ferrari, 2011;Shulman et al., 2015;Liu and Levine, 2016;Taylor, 2016;Hosegood et al., 2017). Additionally, submesoscale motions may also induce shifts on phytoplankton community structure (D'Ovidio et al., 2010;Lévy et al., 2018) affecting food web dynamics and, ultimately, the carbon cycle (Mayot et al., 2017).
The study of the mechanisms controlling frontogenesis and the associated ASC is a relevant topic due to its potential impact, not only on the short-term modulation of nutrients, organic matter or light, and hence on phytoplankton communities (Mahadevan and Campbell, 2002;Klein and Lapeyre, 2009;Lévy et al., 2009Lévy et al., , 2012a, but also because of its role in example global ocean circulation (D'Asaro et al., 2011;Taylor and Ferrari, 2011;Lévy et al., 2012b), heat transport (Siegelman et al., 2020) or fish and marine mammal distribution (Snyder et al., 2017;Siegelman et al., 2020). However, due to the inherent complexity of sampling at such high-resolution levels, only a few studies have reported in situ data of submesoscale spatial phytoplankton distribution across a frontal region yet (e.g., Martin et al., 2005;Taylor et al., 2012;Clayton et al., 2014;Cotti-Rausch et al., 2016;Mousing et al., 2016;Hernández-Hernández et al., 2020). Therefore, our knowledge about submesoscale-influenced phytoplankton distribution and variability mostly constrained to the information extracted from theoretical and modeling studies (Lévy et al., 2001(Lévy et al., , 2012aLi et al., 2012;Liu and Levine, 2016;Taylor, 2016).
In this study, we provide physical and biogeochemical observations on the development and decay of a submesoscale wind-shear front formed on the wake of Gran Canaria Island. Our overarching objective is to discuss how short-term front-generated physical variability affects the distribution and community structure of picophytoplankton organisms, which are major contributors to total phytoplankton biomass and primary production in the subtropical waters surrounding the Canary archipelago (Zubkov et al., 2000a,b;Arístegui et al., 2009). For that aim, we first study the spatiotemporal evolution of the front and biogeochemical parameters. We then statistically examine the effect of the front over the phytoplankton distribution and community structure via Metric Multidimensional Scaling Analysis. We finally compare the variance induced by the spatial and temporal variabilities to determine which source of variability has major influence over picoplankton community variability.

Hydrography, Wind and Sampling Design
Data reported in this paper were collected from 9 to 12 of May of 2011 on board R/V Atlantic Explorer from a section across a wind-shear convergent front (Figures 1a,b). In order to assess both spatial and temporal variability at submesoscale range [horizontal scale of O (1-10 km), vertical scale of O (100 m) and temporal scale of O (1 day)], a section consisting in 6-7 oceanographic stations (Figures 1a,b), separated ∼4 km (25 km in total), was entirely sampled every 24 h, during a 96 h period. Unfortunately, intense wind speed (Figure 1c) did not allow the sampling of the section at 48 h (May 11th). At each station, conductivity-temperature-depth (CTD) casts were made from surface to 300 m using a SeaBird SBE25 CTD sensor additionally equipped with a Wet Lab ECO-AFL/FL Fluorometer. The CTD was mounted onto an oceanographic rosette implemented with six Niskin bottles of 12 L. Discrete water samples were collected for chlorophyll a (Chl a), nutrients and picophytoplankton abundances at six levels, from surface to 150 m, including the deep chlorophyll maximum (DCM). TEOS-10 algorithms were used to calculate all physical derived variables. Mixed layer depth (MLD) was calculated following (de Boyer Montégut et al., 2004). Wind velocities and directions every 10 min were obtained from the Meteorological Station based on the Gando Airport, at the wind-exposed eastern coast of the Gran Canaria Island. Raw wind data was averaged every 4 h for plotting (Figure 1c).

Satellite-Derived Data
Satellite-derived wind velocities and directions components displayed in Figures 1a,b were obtained from the scatterometer mounted on the polar-orbiting satellite MetOp-A (Meteorological Operational satellite) of the European Space Agency (ESA) and provided by Copernicus Marine Environment Monitoring Services (CMEMS). Sea surface temperature (SST) and salinity (SSS) from the CMEMS's product, Atlantic -Iberian Biscay Irish-Ocean Physics Reanalysis accessible through https://resources.marine.copernicus.eu, was used to track the frontal evolution during the cruise. Both data sets present daily temporal resolution, while offering a horizontal resolution of 12.5 × 12.5 km and 8 × 8 km for wind and temperature and salinity, respectively.

Vertical Motions
Vertical velocities associated with diapycnal mixing were calculated under the assumption of negligible viscous forces and important rotation effects. In this case, the ageostrophic Coriolis forcing can be balanced by vertical mixing, and it holds by the scaling of Garret and Loder (1981), w GL hereinafter: where x is the cross-frontal direction, A v is the vertical eddy viscosity, and b is the buoyancy in terms of density (ρ), mean density (ρ o ), and gravitational acceleration (g), such as b = −g (ρ/ρ o ). Though vertical velocities from w GL must not be taken as total vertical velocity, it allows us to compare the magnitude of the diffusive flux and vertical advective flux, i.e., the magnitude of the vertical velocity due diapycnal mixing. Modeling studies have shown that w GL resembles the shape of the total vertical velocity near the surface while differs in its magnitude (Mahadevan and Tandon, 2006;Gula et al., 2014).

Chlorophyll a
For Chl a analysis, 500 mL of water were filtered through 25 mm Whatman GF/F glass-fiber filters, and then stored frozen at −20 • C until their analysis in the land-based laboratory. Pigments were extracted overnight in 10 mL of 90% cold acetone. Chl a was measured fluorometrically, before and after acidification (by adding two drops of 37% HCl) by means of a Turner Designs bench fluorometer previously calibrated with pure Chl a (Sigma Co.) following Holm- Hansen et al. (1965). Chl a data were used to calibrate the Wet Lab ECO-AFL/FL Fluorometer mounted on the oceanographic rosette and connected to the CTD probe.

Inorganic Nutrients
Seawater samples for nitrate + nitrite (NO x − ) determination were collected in 15 mL polyethylene tubes (Van Waters and Rogers Co., VWR) and preserved frozen at −20 • C until their analysis in the land-based laboratory. Nitrite was colorimetrically measured using a Bran+Luebbe Autoanalyzer AA3 model following Hansen and Koroleff (1999) protocol for automated seawater nutrient analysis. Vertical nutrient fluxes were assessed by following Fick's law: where k z is the vertical eddy diffusivity. Notice that the nature of our survey does not allow us for a direct analysis of the kinetic energy dissipation rates from microstructure profilers to obtain k z (e.g., Arcos-Pulido et al., 2014;Tsutsumi et al., 2020). Notwithstanding, the increasing interest of the impact of mixing and turbulence in the biological marine systems, prompted a series of studies that compares k z from microstructure data and fine-structure parameterizations with a reasonable degree of agreement (e.g., Inoue et al., 2007;Arcos-Pulido et al., 2014).
Here, we calculate k z based on the parameterization of Zhang et al. (1998) validated in Inoue et al. (2007 and Arcos-Pulido et al. (2014), in which both turbulence and double diffusion mixing process are combined to obtain k z . The approach of Zhang et al. (1998) is valid in a salt-fingering regime as dominance of Turner angles higher than 45 • indicates (Supplementary Figure 2). The reader could refer to Arcos-Pulido et al. (2014), for a full derivation of the parameterization used here.

Picoplankton Abundances and Biomass Conversion
Cyanobacteria-like Prochlorococcus (Pro) and Synechococcus (Syn), as well as photosynthetic picoeukaryotes (Euk), were counted with a FACSCalibur (Becton and Dickinson) flow  cytometer. Seawater samples (1.8 mL) were collected on 2 mL cryotubes (VWR) and fixed with 20% paraformaldehyde to 2% of final concentration. Fixed samples were stored at 4 • C for 20 min and then frozen and preserved in liquid nitrogen (−196 • C) until their analysis. 200 µL of sample were transferred to a flow cytometer tube and inoculated with 4 µL of yellow-green 1 µm ø latex beads suspension, as an internal standard (Polyscience Inc). Samples were run at 60 µL·min −1 for 150 s approximately. Groups were identified comparing red (FL3-H) fluorescence versus both orange (FL2-H) fluorescence and side scatter (SSC-H) in bivariate scatter plots.
Carbon biomasses were estimated using empirical conversion factors provided by M. F. Montero (Montero et al., unpublished data). They carried out more than 60 experiments of sequential filtration (through seven polycarbonate filter from 0.2 to 3 µm) with water from the surface and the DCM of the coastal waters of Gran Canaria island. Picoplankton biovolumes were calculated via sigmoidal fits of cell counts obtained by Flow Cytometry. Spherical shape was assumed for picoplankton. Abundances were then multiplied by its corresponding average carbon content (43 fg C·cell −1 for Pro; 100 fg C·cell −1 for Syn and 444 fg C cell −1 for Euk) obtaining average biomass data for each group. Biomass data were integrated from 0 to 150 m, from 0 to MLD, and in the DCM. Integrated biomass data were then used as input for statistical computations.

Statistical Analysis
In order to identify potential effects of submesoscale processes over picoplankton community structure, a Metric Multidimensional Scaling Analysis (MDS) [also referred to as Principal Coordinate Analysis (PCoA)] was carried out for every sampling day. Ecological distance matrices of integrated picoplankton biomass were calculated by means of Bray-Curtis dissimilarity and then used as inputs for MDS analysis. The two orthogonal axes (MDS1 and MDS2) obtained from the MDS analysis were used as axes for results' scatter plots.
Stations were grouped using K-means clustering method, which aims at partitioning the data into groups such that the sum of squares from data within the assigned cluster is minimized. The value of between-cluster sum square (BSS) divided by the total sum of squares (TSS) was used to decide the optimal number of clusters. The number of clusters that provides higher BSS/TSS ratio was chosen. A first approximation to the optimal number of clusters was also done following the Elbow method.
To quantify the contribution of each scale of variability to the total variance of picoplankton community, a Variance Component Analysis (VCA) was conducted. Previously, the biomass dataset was Winsorized to avoid extreme values. A random effect Linear Mixed Model (LMM) was then fitted to the whole water column, mixed layer (ML) and DCM. Variance components were extracted from fitted LMM. Finally, variances were expressed as the percentage of total variance. R software 1 was used to conduct all statistical analysis.

Spatiotemporal Evolution of the Front
The cruise took place during a highly variable wind regime according to data provided by the Gando Airport meteorological station (Figure 1c). During the first 48 h of the experiment, trade winds (northeast) increased from ∼8 m·s −1 (0 h) to more than 14 m·s −1 (48 h). 72 h on, wind shifted its direction blowing from the south, and its speed dropped down to less than 6 m·s −1 . Satellite-derived wind velocities and direction shown in Figures 1a,b support the above data. At 24 h, intense (up to ∼14 m·s −1 ) trade winds are observed at both flanks of the island. However, in the lee of the island winds dropped down to ∼6 m·s −1 . Notice that the sample section crossed the wind shear zone. Unfortunately, the studied zone was not in the satellite trajectory at 72 h. Instead, wind velocity and direction for the day after (96 h) are plotted in Figure 1b. As Figure 1c shows, 72 and 96 h wind conditions were quite similar. At 96 h, due to weak (∼5 m·s −1 ) northward winds, the windless zone in the lee of the island disappeared and consequently the wind shear front which crossed the section vanished.
Since S1 was not sampled on the first day (0), only the eastern part of the front (S2, S3, and S4) was recorded. It was characterized by doming of the isopleths inside the ML, introducing relatively colder and denser water into shallower depths (Figures 2a,d,g). Vertical isopleths at S2 (19.80 • C, 26.15 kg·m −3 , and 36.77) suggest that the downward branch of the front, which would be on S1, was also affecting this station. Downward movement of the isotherms was observed at S5 and S6 (Figure 2a). Highest surface values of T q , s q , and S occurred at S7. Vertical velocities tracked pretty well with T q , s q , and S fields in the ML (Figure 3a). Negative w GL (downward) were associated with the deepening of the isopleths at S5 and S6 while positive w GL (upward) occurred at S2, S3 and S4 where isopleths dome. Thermo and pycnocline were situated at ∼70 m remaining relatively stable along the section as well as MLD.
Wind intensification in the first 24 h (Figures 1a,c) strengthened the front that led to a reinforcement of the 19.80 • C isotherm, 26.15 kg·m −3 isopycnal and 36.77 isohaline, deepening from ∼30 to ∼90 m and spreading from S5 and S6 to S3, S4, and S5 (Figures 2b,e,h). A doming of the isopleths associated with the front affected the entire S6. Like at 0 h, w GL field was consistent with the physical structure. Vertical velocities also strengthened at 24 h (Figure 3b). Downward velocities were associated with the front-related downwelling whilst upward velocities coupled with isopleths upwelling. Thermocline (pycnocline) reshaped by the downwelling produced by the front, and the upwelling produced by the doming of the isopleths, presenting a vertical zig-zag pattern along the section.
On day 4 (72 h), neither wind intensity nor direction were favorable for front development (Figures 1b,c). Indeed,19.80 • C isotherm, 26.15 kg·m −3 isopycnal and 36.77 isohaline horizontally crossed the whole section (Figures 2c,f,i). Nevertheless, a relative weak zig-zag pattern was still recognizable in the MLD similar to 24 h scenario (upwelling at S5 and S6, downwelling at S3 and S4; Figures 2b,e,h); probably a remnant of the thermocline deformation caused by the up-and downwelling fluxes driven by the front the day before. w GL also maintained its bipolar structure between upwelling and downwelling stations above the MLD (Figure 3c). Below the isopleth doming observed at S5 and S6, counterpart bowl-shaped structure with negative w GL highlights, both conforming a bipolar lentil-like shaped structure. Thermo and pycnocline were placed at ∼ 40 m at S1, S2, S3, and S4, whilst at ∼100 m at S6 and S7. The Shallowest thermo and pycnocline occurred at S5 associated with upwelling motions.
Though satellite-derived data should be used carefully in high-resolution cruises (as is the case) due to their significantly coarse horizontal resolution, a thermo-haline frontal zone is observed crossing through approximately the middle of the sampled section (Supplementary Figure 1) supporting our in situ observations. SST and SSS data also support the front temporal evolution of the front, showing a moderate intense front is observed at 0 and 24 h (Supplementary Figures 1A,B) compared with 72 h (Supplementary Figure 1D).

Biogeochemical Features
Nutrients (NO x − ) present the typical vertical distributions of oligotrophic systems throughout the cruise (Figures 4a-c). Low values (<0.5 mmol·m −3 ) were found at surface waters down to the thermocline, where NO x − increased into deeper waters (nitracline), reaching more than 3 mmol·m −3 at 150 m. Nevertheless, this typical nutrient distribution is not consistent along the sections. At 0 h (Figure 4a), the nutricline did not coincide with the thermocline along the section, being deeper at S3 and S4 (∼120 m) compared with S5, S6, and S7 (∼80 m). Deeper nutricline at S3 and S4 coincides with downward w GL below the MLD while shallower nutricline at S5, S6, and S7 is associated with upward velocities (Figure 3a).
Highest NO x − values were found at S5, S6 and S7 below the MLD and at S3-S4 and S6-S7 above it. Notwithstanding, while high NO x − in the first stations seen to be linked to high NO x − concentrations below the MLD at S5-S6, high values at S6-S7 are not connected with NO x − maximums below the MLD. The most intense upward (positive) F NOx below the MLD ( Table 1) were observed at S5-S7 (0.039, 0.057, and 0.108 mmol·m −2 ·d −1 , respectively) as well as the nutrient gradients (g NOx ). Negative values of g NOx indicates a favorable nutrient gradients for upward fluxes. Comparing NO x − and F NOx with the w GL field, it could be observed that the higher values of NO x − and F NOx at S5 and S6 were associated with most intense w GL in the MLD as well as that w GL in the MLD at S7 were negative (downward) which may be the reason of the detached high NO x − patch observed in the surface waters of S6-S7.
The reshaping of the thermocline (pycnocline) by the reinforcement of the front at 24 h also reshaped the nutricline (Figure 4b), which shows the same zig-zag pattern observed in T q and s q Figure 2b). At S1 and S6, sloping of the isotherms introduced water with concentrations of about 2 mmol·m −3 ∼ 40 m above the main thermocline reaching the surface at S6 while deepening of the isopleths at S2-S4 sinks surface waters to ∼ 90 m depth. The both, most intense positive and negative w GL (Figure 3b) occurred associated with this isotherm sloping, respectively. The highest F NOx and g NOx were found at S6 and S1 (0.114 mmol·m −2 ·d −1 ) while lowest were found at S2-S4 and S7 (Table 1). In day four (72 h; Figure 4c), the doming of the isopleths below the MLD at S1-S2 and S7 introduced nutrientricher waters from deeper layers into the bottom of the MLD (Figure 4b). Nevertheless, large amounts of NO x − inside the ML were observed at S1-S2 and S5 (Figure 4c). The highest F NOx and g NOx values right below to the MLD supports the upwelling of nutrients in those stations (Table 1), although w GL did not completely agree with NO x − and F NOx . Beside NO x − distribution suggest that a tongue of nutrient-richer waters outcrop from ∼100 m to about 20 m at S5, upward vertical velocities only dominated on the MLD while downward velocities are presented below the MLD.
Similarly, the vertical distribution of Chl a follows the characteristic pattern of an oligotrophic system (Figures 4d-f), presenting low values (<0.1 mg·m −3 ) at surface waters, while a DCM was consistently observed over the nitracline. At 0 h (Figure 4d), the horizontal distribution of Chl a along the section revealed a discontinuity in the DCM between eastern stations (S1, S2, S3, and S4), where a deeper and more intense DCM occurred (0.4-0.5 mg·m −3 ) and western stations (S5 and S6), as seen in nutricline (Figure 4a). This discontinuity became obvious when the front intensified at 24 h (Figure 4e). An intense DCM (∼0.6 mg·m −3 ) due to the front-driven sloping of the isotherms was placed at 40 m at S1 and at ∼150 m at S3 and S4. Relatively high values of Chl a were also observed at 24 h in surface waters of S6 and S7, coinciding with nutrient upwelling (Figure 4b). In the western stations, the DCM remained centered at 60 m depth. However, a slightly increase in Chl a (∼0.45 mg·m −3 ) coincided with nutrient upwelling in S5. Weakening of the front at 72 h (Figure 4f) resulted in an overlap of the two DCM cores at S5 coinciding with the lentil-like shaped structure described in the section above (Figures 2c,d).

Picoplankton Distribution and Community Structure
Maximum values of Prochlorococcus (Pro) biomass were generally distributed between subsurface waters (>20 m) and just above the DCM throughout the cruise (Figures 5a-c). At 0 h (Figure 5a), Pro biomass was widely distributed along the section presenting higher values in the MLD of S3, S4, and S5. On the second sampling day (Figure 5b), Pro biomass decreased, and the maximum values were associated with upwelling velocities at S4, S5, and S6 (Figure 3b). At 72 h (Figure 5c), the highest Pro biomass values were found at S2 at 60 m coinciding with positive (upward) w GL and F NOx (Figure 3c and Table 1). The highest Pro biomass values were consistently placed in nutrient upwelling zones.
Synechococcus (Syn) was generally widely distributed in the well-mixed waters above the thermocline (Figures 5d-f). Like Pro, Syn presented its highest biomass at S4 in the first sampling day (Figure 5d). At 24 h (Figure 5e) the general Syn biomass distribution changed, and high Syn biomass values were found below the thermocline, at the base of the front (∼120 m). Deep high Syn biomass values were also observed on day four (72 h; Figure 5f) associated with the downwelling occurred below the MLD at S5. Nevertheless, maximum values at 72 h were found in surface waters of S1 and S2. The distribution of the Euk and the DCM resembled throughout the first 24 h (Figures 5g,h). At 72 h the relationship between Euk and DCM distribution was observed along S4-S7, while maximum values of Euk biomass were found below the DCM (MLD) in S1-S3 breaking with the observed general pattern (Figure 5i). Table 2. At 0 h, all picoplankton groups present higher integrated biomass in the eastern stations, showing Cyanobacteria group differences of up to 3-fold between both ends of the section (S2 and S7). Differences of up to 2-fold between S4 and S6 separated by ∼10 km in Cyanobacteria's biomass can be observed. Pro is the major contributor to picoplankton community biomass along the section at 0 h (56.59 ± 2.64%), while Syn and Euk show similar contributions to total biomass (22.68 ± 3.88% and 20.73 ± 1.43%, respectively; Figure 6a). With the enhancement of the front on day two (24 h) the highest Cyanobacteria biomasses are observed at the front-associated stations (S3-S5) while both ends of the section show similar values ( Table 2). Nevertheless, differences are not as high as at 0 h. Euk presents a consistent integrated biomass along the section at 24 h. Pro still is the major contributor to the total biomass (51.32 ± 7.88%) except in S2 where Euk and Pro show similar contribution rates (38.81% and 38.04%, respectively; Figure 6b). At 72 h, both Syn and Euk present similar integrated values among the stations. Pro keeps showing higher values in front-affected stations (S2-S5) as well as in the major contribution percentages to total biomass (48.16 ± 2.51%; Figure 6c).

Front Effects Over the Community Structure
Metric Multidimensional Scaling Analysis sorts stations according to differences in picoplankton community structure. Therefore, closer stations present similar picoplankton community assemblages and vice versa. At 0 h, three stations groups were obtained from K-means clustering method ( Figure 7A): (1) S2, S3, and S4 where front-driven up motions occurred; (2) S5 and S6, that were situated at the western boundary of the front; and (3) S7, the farthest from the front. At 24 h (Figure 7B), three groups were also observed: (1) S1, S2, and S7, that represent the eastern and western boundaries of the front; (2) S3, S4, and S5 situated at the upwelling front; and (3) S6, where the upwelling front occurs. The vanishing of the front at 72 h also lead to the vanishing of community structure heterogeneity and no significant differences were observed among station in community structure ( Figure 7C). The selection of three groups at 0 h and 24 h was supported by high BSS/TSS ratio (91.7% and 85.1%, respectively). At 72 h the BSS/TSS ratio is not displayed since there were no differences among stations.

Spatial vs. Temporal Variability
The contributions of every source of variability to total variance are compiled in Table 3. Depending on the phytoplankton group, two behaviors may be observed in the entire water column. Cyanobacteria-like Pro and Syn present higher space variability (i.e., among station in the same day), while Euk presents higher temporal variability. This pattern is not observed in the DCM, where all phytoplankton groups show higher temporal than spatial variability. Inside the mixed layer, phytoplankton groups present almost equally spatiotemporal variability. Chl a shows higher temporal variability in all cases.

Wind Forcing Frontogenesis
Oceanic fronts originated south of Gran Canaria in the area of eddy formation at the wind shear flanks were reported in earlier studies describing the eddy field in the region (Arístegui et al., 1994(Arístegui et al., , 1997Barton et al., 1998). Later, Barton et al. (2000) and Basterretxea et al. (2002), in more frontfocused studies, suggested a potential mechanism for their development. They observed that wind velocities dropped down up to one order of magnitude at the lee region of the island with respect to a station placed in the wind exposed region ∼2 km apart. As a consequence, net westward Ekman transport in the lee region would be practically absent, favoring the convergence (divergence) of water in the eastern (western) side of the wake and its subsequent downwelling (upwelling).
The data presented here fit the Barton et al. (2000) and Basterretxea et al. (2002)'s spatial wind field observations (Figures 1a,b) but they also indicate a positive temporal relationship between wind speed and front intensity. During the first 48 h, the increase in down-front blowing winds (Figure 1c) strengthened the front signal as seen in T q , s q , and S plots (Figures 2a,b,d,e,g,h). Conversely, the change in wind direction at 72 h (up-front winds) caused the vanishing of the front signal and the increase in water column stratification, as suggested by the shallowest MLD (Figures 2c,f,i). This agrees with the nonlinear Ekman effect theory for frontogenesis of Thomas and Lee (2005), that proposes that winds blowing in the direction of the geostrophic flow generate an Ekman flux that tends to advect colder water from one side of the front over warmer water from the other side, enhancing convective mixing and, thus, strengthening the front. ASC-related upwelling in the warm side of the front and downwelling in the cold side would be triggered as consequence of convective mixing (Nagai et al.,   Statistical analysis was carried out for integrated biomass values in the whole water column (0-150 m), for the deep chlorophyll maximum (DCM) and for the mixed layer (ML). Variances were extracted from Variance Component Analysis (VCA). -Sanz et al., 2010). By contrast, winds blowing against geostrophic flow generate the advection of warmer water over colder water favoring vertical stratification and, hence, the weakening of the front (frontolisis). This theory has been later sustained by several modeling and numerical studies (Thomas, 2005;Thomas and Lee, 2005;Thomas and Ferrari, 2008;Mahadevan et al., 2010).

2006; Pallàs
The bipolar structure (upwelling/downwelling; warm/cold) described by the nonlinear Ekman effect theory agrees with the physical structure clearly shown in Figures 2b, 3b. A steep deepening of the isotherm in S2-S4 sinks water in the cold side of the front while doming of the isotherm in S5 and S6 entrained deep waters in the warm side suggesting downwelling and upwelling motions, respectively. This is supported by downward and upward w GL coinciding with S2-S4 and S5 and S6. At 0 h, only the upwelling side of the front is observed (Figures 2a, 3a), being characterized by a less intense doming of the isotherm in the warm side (S2-S4) compared to the one observed at 24 h. Notwithstanding, positive w GL were also observed at these stations, coinciding with 24 h observations.

Effects of Frontal Dynamics Over Nutrient Distribution
Subtropical oligotrophic areas such as the Canary region are characterized by a well sun light illuminated mixed layer throughout the year, with very low inorganic nutrient concentrations (Levitus et al., 1993) due to the presence of a strong almost permanent thermocline that prevents the outcrop of deeper nutrient-rich waters into the euphotic zone (León and Braun, 1973). It has been suggested that ASC associated with submesoscale fronts may locally alleviate nutrient shortage in oligotrophic surface waters by driving vertical nutrient fluxes into the ML (Mahadevan and Tandon, 2006;Lévy et al., 2012a;Estrada-Allis et al., 2019).
In the present case, the canonical oligotrophic nutrient distribution was broken by high nutrient concentrations that outcropped trough the thermocline into the ML in S4 and S3 at 0 h; S6 at 24 h, and S5 at 72 h (Figures 4a-c). In all cases, nutrient intrusions were associated with isopleths doming driven by front-associated upwelling, where positive (upward) w GL and F NOx near the MLD occurred. Though small, the upward fluxes are consistent with other observations in areas of intense mesoscale and submesoscale activity (Arcos-Pulido et al., 2014;Corredor-Acosta et al., 2020). The overlapping of positive w GL with F NOx suggests that diapycnal mixing is acting as an important contributor to the vertical velocity (Ponte et al., 2013) and may be associated with submesoscale process (Estrada-Allis et al., 2019). In summary, upward movements at each side of the front favors the injection of NO x into the euphotic layer as well as downward motions deepens the nutricline impoverishing the ML, supporting earlier theoretical studies (Lévy et al., 2018 and reference therein).

Does Frontal Dynamics Modulate Picoplankton Distribution and Community Structure?
The picoplankton distribution presented here (see Figure 5) largely corroborates the commonly reported distribution described for each group in the region (e.g., Baltar et al., 2009). Syn was abundant in the well-mixed surface layers (Mackey et al., 2013;Grébert et al., 2018), whereas higher amounts of Pro were present in deeper layers (Bouman et al., 2006;Johnson et al., 2006;Biller et al., 2015). Euk were the main contributors to DCM laying close to the nitracline, suggesting that they require higher inorganic nutrients concentrations than prokaryotic phytoplankton for their growth (Painter et al., 2014). Nonetheless, these general distributions were eventually modified by frontal dynamics.
For instance, we observed high Syn concentrations at ∼150 m depth at 24 h coinciding with the downwelling branch of the front. This distribution agrees with the subduction of phytoplankton by submesoscale front-associated downwelling as proposed by several authors (Lévy et al., 2001(Lévy et al., , 2012a(Lévy et al., , 2018. Whether the front-driven enlargement of the mixed layer at this station was responsible for the Syn distribution observed or, by contrast, Syn cells found at these depths were dragged from surface waters due to front intensification, is difficult to discern, although front dynamics seem to be behind the distribution patterns in both cases. High Euk biomass at 72 h found below the DCM and the MLD of S1, S2, and S3, is another example of how frontal dynamics subducts phytoplankton biomass. In this case, the high Euk biomass patch appears to be a leftover from the intense DCM observed at 24 h, which has been left out the ML due to frontolisis restratification. One of the most striking exceptions to the usually reported picoplankton distribution is observed in Pro. The fact that high Pro biomass patches consistently coincided with high nutrient concentrations (Figures 3a-c, 4a-c) differs from the distribution patterns previously reported in the literature. Due to their high nutrient diffusion per unit of cell volume (Raven, 1998;Marañón, 2015) and their capacity of uptake dissolved organic matter (DOC) for growth (Berman and Bronk, 2003;Mulholland and Lee, 2009;Znachor and Nedoma, 2010;Duhamel et al., 2018;osmotrophy), Pro inorganic nutrient requirements are low and thus, they usually present higher abundances in nutrient-poor zones (Bouman et al., 2006;Johnson et al., 2006;Biller et al., 2015). Indeed, several studies have reported low Pro abundances related to eddy-driven nutrient upwelling in the region along with high Pro biomass associated with high dissolved organic matter concentrations (Baltar et al., 2009;Hernández-Hernández et al., 2020). Maximum DOC concentrations along the cruise (not shown here) also coincided with Pro biomass peaks. For these reasons, we considered that Pro and NO x − maximums resemblance seems to be a coincidence rather than a causality, and that front-driven accumulation of DOC would be the reason of high abundances of Pro at nutrient upwelling stations.
Besides the general distribution, the data presented in Figure 5 reveal high biomass patches for every picoplankton group. Several authors have reported local increases of different phytoplankton size groups across frontal zones due to the input of nutrients in a constrained zone, which usually favors the growth of large cells such as diatoms (Abraham, 1998;Rivière and Pondaven, 2006;Mahadevan et al., 2012). D'Ovidio et al. (2010) observed that phytoplankton is organized in submesoscale patches of dominant types separated by physical barriers. Our data reflect two main differences with respect to the studies mentioned above: (1) previous works observed that patches were dominated by different phytoplankton sizegroups (i.e., pico, nano, or microplankton;Abraham, 1998;Rivière and Pondaven, 2006;D'Ovidio et al., 2010;Mahadevan et al., 2012), while we observed that patchiness also occurs within the same size-group. This finding raises the question of at what level of organization patchiness actually works.
(2) They observed high biomass patches related with local nutrient injection (Abraham, 1998;Rivière and Pondaven, 2006;D'Ovidio et al., 2010;Mahadevan et al., 2012). In our study, by contrast, only Pro high abundance spots are related to high nutrient concentrations albeit this is probably not due to a causality, as we explained above. While it is true that we only report picoplankton data, the occurrence of these non-nutrient related "hotspots" of high picoplankton biomass suggests that submesoscale dynamics modulates both the hydrographic and biogeochemical fields, favoring the local growth of some groups against others. It is known that although picoplankton groups generally co-occur in subtropical oceans, they present different nutrient requirements, light harvesting, different temperature or physical forcing acclimation (Moore et al., 2002;Scanlan et al., 2009;Mella-Flores et al., 2012;Flombaum et al., 2013;Otero-Ferrer et al., 2018). Therefore, picoplankton groups' distribution across a submesoscale front would be expected to be affected by the front-generated physical and biogeochemical variability (Lévy et al., 2018 and references therein).
A relevant result of our study is the modulation of the picoplankton community structure by the front. Phytoplankton community assemblages were strongly structured in the MDS ordination space in accordance with the frontal structure (Figure 7). Few studies have reported the variability in phytoplankton community structures across a frontal systems at submesoscale level (e.g., Taylor et al., 2012;Clayton et al., 2014;Mousing et al., 2016). In all of these studies, different assemblages were observed; at each side of the front and within the front, as the result of the separation of two well-defined water masses and hence two different biomes, with two different communities. Conversely, we observed that picoplankton communities were not separated by the front but showed a mirror-like distribution with respect to the middle of the front. While observations from previously cited authors suggest that fronts work like a physical barrier for different niches, our results suggest that frontal dynamics modulates the phytoplankton community structure. However, it should be noted that while fronts reported by the authors mentioned above were permanent features that separate different water masses, we sampled an ephemeral front that is originated inside the same water mass (Lévy et al., 2018).
Due to section proximity to the coast, tidal forcing was initially considered as another potential driver for the observed picophytoplankton variability. Sangrá et al. (2001) studied the effect of internal waves on Chl a in the shelf break of the lee region of Gran Canaria during a spring and a neap tide. They reported an increase in Chl a of up to 47% during some pulses of the spring tide from a station situated over the shelf (100 m depth). Notwithstanding, depth integrated Chl a values presented little differences between samplings. Since our section was situated on the 2,000 m isobath (i.e., the island slope), the cruise took place during a full neap tide, and phytoplankton biomass increases were significantly larger than those reported by Sangrá et al. (2001), we considered that the tidal forcing effects if any, they would be negligible compare to front-related effects.

Spatiotemporal Variability
In an earlier study, Martin et al. (2005) observed higher variability in picoplankton community biomass at mesoscale ranges than in the normally used large-scale ranges, arguing that sampling should be done at smaller scales to avoid inaccurate plankton distributions. The integrated biomass data presented here (Table 2) reveal that picoplankton biomass varies between 2 and 3-fold on spatial scales of ∼ 2.5 km, and temporal scales of ∼ 24 h. This variability is comparable to the picoplankton biomass seasonality reported for the region (Zubkov et al., 2000b;Baltar et al., 2009).
In order to assess which source of variability was dominant, we compared the temporal and spatial variances observed during our study ( Table 2). We found that picoplankton biomass variance is almost equally shared between time and space in the mixed layer, while it mostly depends on time in the DCM, i.e., in the front reported here, picoplankton biomass temporal variability is just as important as, or even more important than spatial variability. Although the front is constrained to a marginal part of the Canary Current region, mesoscale processes and associated submesoscale motions, are ubiquitous around the global ocean (Chelton et al., 2007(Chelton et al., , 2011. Therefore, our results beg to question whether oceanographic samplings in regions of high mesoscale activity should be designed considering submesoscale spatiotemporal resolutions, in order to gain a more accurate approximation of the biogeochemical fields variability in the region of study.

CONCLUSION
The spatiotemporal development and decay of the convergent wind-driven submesoscale front south of Gran Canaria, as well as their effects on picoplankton community structure and distribution, is reported for the first time through in situ measurements. Like in earlier studies in the region (Barton et al., 2000;Basterretxea et al., 2002), our data shows a positive relationship between wind and front development and intensity. Upward diapycnal nutrient flux occurs near the mixed layer of the stations located on the front. This diapycnal mixing was implicated in the observed enhancement of nutrients and chlorophyll in the upper layer advected by positive vertical velocities based on the scaling of Garret and Loder (1981). Conversely, picophytoplankton biomass subduction is also reported. The present study is consistent with model outputs and past predictions, supporting that submesoscale fronts may drive nutrient fluxes into the euphotic layer and subduct picoplankton biomass below it (Mahadevan and Archer, 2000;Lévy et al., 2001Lévy et al., , 2012a.
On the other hand, our results also provide new insights in front formation and erosion, pointing out to nonlinear Ekman effects as a potential driver of front dynamics, and their effects on picoplankton community structure. The front favors the patch formation of different picoplankton groups' dominance and modulates the picoplankton community structure. Temporal variability was found to be a significant source of error in phytoplankton variability providing evidence that, at least in regions of high hydrographic variability, plankton, as well as other biogeochemical features, must be sampled at shorter spatial and temporal resolutions than regularly done in order to obtain more accurate datasets. Although daily repeated cruises are in many cases economically unviable and time-consuming, submesoscale measurements would help to get more accurate regional and long-term interpretation of biogeochemical fluxes.
It is worth mentioning that the physical results presented in this study are constrained by the spatiotemporal scales of the survey and the lack of horizontal velocities. However, validated parameterizations and solid scaling of the vertical velocity formulations, allow us to provide a first approximation of submesoscale and diapycnal mixing impact on the biological system in the leeward side of Gran Canaria Island.

DATA AVAILABILITY STATEMENT
The data sets generated for this study are available on request to the corresponding author.

AUTHOR CONTRIBUTIONS
JA conceived and designed the cruise. NH-H, YS-F, and JA carried out the sampling and data analyses. SE-A contributed to the analysis of the physical data. NH-H wrote the manuscript with inputs from all authors. All authors contributed to the article and approved the submitted version.

ACKNOWLEDGMENTS
We thank to the members of the Biological Oceanography group (GOB-IOCAG) for their help with the analysis of the samples, and Laura Marín for reviewing the English grammar.