The NW Mediterranean is characterized by a cyclonic circulation forming a gyre-like cell in the Ligurian Sea (Millot, 1991, 1999; Millot and Taupier-Letage, 2005). The northern branch of this circulation is formed by the Northern Current (also known as the Liguro-Provençal-Catalan Current, and hereafter NC). The NC is the return branch of the cyclonic circulation that flows along the Italian, French and Spanish coasts, generally following the 1000 m isobath (The MerMex Group, 2011). This permanent boundary current is part of the thermohaline circulation of the Mediterranean Sea and is formed by the merging of two current branches flowing on both sides of Corsica (Figure 1). The core of the NC is stronger and located about 10-40 km away from the coast in winter, while it is less defined and further offshore in late summer and early fall (Birol et al., 2010; Piterbarg et al., 2014; Declerck et al., 2016).

This study focuses on the Nice-Calvi transect line (Figure 1, bottom panel). This line is regularly monitored as part of the Mediterranean Ocean Observing System for the Environment (MOOSE), a multi-scale and multi-site observation network targeting coastal/open ocean and ocean/atmosphere interactions. As part of MOOSE, monthly and annual ship-based observations are made along this transect. The transect is also regularly occupied by glider since 2007, with the aim of making it an endurance line. Located along the transect about 53 km offshore from Nice, the DYFAMED oceanographic station (red star on Figure 1) is also one of the most studied station of the basin (Marty et al., 2002). The good spatio-temporal coverage of the area makes it a perfect place for the first in situ application of the MiniFluo on the glider.

The annual cycle of the biogeochemical system functioning at DYFAMED may be described in three distinct periods: winter, spring and summer/fall (Avril, 2002; Migon et al., 2002). The winter period is dominated by deep convection and the water column is well mixed. It is followed by the spring bloom (March/April) occurring after the onset of stratification. The rest of the year, the NW Mediterranean exhibits a Deep Chlorophyll Maximum (DCM) at mean depth between 40 and 60 m that persists until being destroyed by the following winter convection (Lavigne et al., 2015). Maximum stratification is reached at the end of summer/early fall, and corresponds to a period when DOM accumulates in the surface waters (Jones et al., 2013). Smaller phytoplankton blooms may also occur in the fall.

The sensor presented here is a new version of the MiniFluo, a field-portable fluorometer introduced by Tedetti et al. (2013), which is now glider-compatible. Basic working principles are recalled below.

Glider measurements presented here were performed during three campaigns across the NC and along the Nice-Calvi transect (Figure 1). The first campaign was realized by ALSEAMAR with one of their glider (code name SEA007) while the two other campaigns were carried out with the Mediterranean Institute of Oceanography's SeaExplorer glider (code name SEA003). The fall campaign was conducted between 28 October and 10 November 2015, the spring campaign between 29 April and 20 May 2016 and the summer campaign between 25 July and 4 August 2016. For the three campaigns, the glider 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 (σ₀) are derived using TEOS-10 toolbox (McDougall and Barker, 2011). This GPCTD is also equipped with a dissolved oxygen (O₂) sensor (Seabird's SBE-43F) from which concentrations are derived. A WetLabs ECO FLBBCD was also mounted on the glider for measurements of Chlorophyll-a (Chl-a) fluorescence (λ_Ex/λ_Em: 470/695 nm), backscattering at 700 nm (BB700) and humic-like fluorophore fluorescence (λ_Ex/λ_Em: 370/460 nm) expressed in μgL⁻¹ equivalent quinine sulfate units (μgL⁻¹QSU). For this sensor, the manufacturer's calibration was used. This means that Chl-a concentrations were not calibrated on in situ measurements and are rather expressed in their equivalent of Thalassiosira weissflogii phytoplankton mono-culture. Measured Chl-a concentrations values were also shifted by an offset in order to obtain [Chl-a]=0 at 250 m (e.g., Lavigne et al., 2015). This was achieved by subtracting 0.02μgL⁻¹ (fall campaign) and 0.08μgL⁻¹ (spring and summer campaigns) from the whole dataset. In a complement to these relatively common glider measurements, the MiniFluo was used on an exploratory basis: never before such high frequency measurements of Trp and Phe were performed from a glider.

Only subsets of the glider campaigns are presented here (see Figure 1 caption for dates). They consist of about the first 100 km of the Nice-Calvi transect (red-dashed line in Figure 1), with the DYFAMED station located at x = 53.2 km. This section was chosen because of the relatively good overlap of the glider tracks during the three surveys. Note that while the glider traveled from Nice toward Calvi for the fall and spring transects, the summer transect was realized in the opposite direction. Glider observations were first processed with the Socib glider toolbox (Troupin et al., 2015) for cast identification and georeferencing. Then, in order to directly compare the three surveys, each cast was projected to its closest point on the straight dashed-line of Figure 1. Each transect is displayed in term of its distance from the origin of the transect (Figures 6–8). The horizontal resolution between these transects is variable since the depth of glider dives was different from one deployment to the other, respectively 300, 650, and 400 m. In order to focus on biogeochemical variables, only the top 250 m are considered here, a depth range where most of the variability is found.

This section recalls useful observations made during the early glider deployments with the MiniFluo. These have revealed some limitations and interesting observations regarding the sensor's in situ working principle.

Figures 6–8 show contour plots of the glider data from the three deployments, respectively fall 2015, spring 2016 and summer 2016, along the first 100 km of the Nice-Calvi transect and for the depth range 10–250 m. As it is shown below, the maximum offshore extent of the NC during these 3 transects (e.g., in Figure 6) is about 60 km from the coast, further than the DYFAMED station. In order to obtain data representative of the “offshore” region, i.e., a portion of the transect not under the direct influence of the Northern Current, we used the dashed box drawn in Figures 6–8 (70–80 km from the shelf break). Average profiles over this region are presented in Figure 9, which summarizes the seasonal differences for every variables presented in these contourplots. Physical and biogeochemical features are presented below.

One limitation to our study is the relative absence of hindsight with the MiniFluo technology. For example, fine tuning improvements were made between the different campaigns presented here (e.g., monitoring section change between MiniFluo #05 and #14 and optical cap change between spring and summer 2016 missions). Although the scale factor and blank values were adjusted between these missions, the repetition of the measurements with a single setup over many years would help identifying if one part of the signal is due to instrumental changes. While the relative coherence in concentration range observed from one campaign to the other is reassuring, discrete water samples at the time of glider deployment and/or recovery would certainly increase our confidence in the measurements. Bootstrap analysis on scale factors suggests a maximum error on measurement of 7% for the Trp and 10% for the Phe (WAF calibration) with MiniFluo#14, and respectively 50 and 33% with MiniFluo#05. It is likely that the change of monitoring section between the two MiniFluo generation has led to a decrease in the calibration error.

Trp-like fluorophore concentrations measured with the MiniFluo in the present work (0–6 μgL⁻¹) are higher than concentrations of total (free + combined) dissolved Tryptophan determined by alkaline hydrolysis and HPLC in coastal marine waters (0.2–0.4 μgL⁻¹; derived from Yamashita and Tanoue, 2003) but close to those observed in freshwaters (2–8.5 μgL⁻¹; derived from Wu and Tanoue, 2001). Two assumptions can be made to explain why the MiniFluo provides higher concentrations. First, while HPLC analyses allow the determination of dissolved Trp concentrations only, the MiniFluo also record fluorescence due to Trp bounded to particulate matter that may lead to an increase in the total Trp-like fluorescence signal in seawater. Second, as highlighted by Carstea et al. (2016), other organic compounds, such as lignins, may also fluoresce in the Trp spectral domain. Discrete molecular analyses of dissolved tryptophan combined with MiniFluo measurements would help understanding the real contribution of dissolved tryptophan compound in the MiniFluo-derived Trp-like fluorescence. Interestingly, Khamis et al. (2015) found Trp-like fluorophore concentrations in groundwater ranging from 3 to 15 μgL⁻¹, i.e., very close to the concentration found here, using UviLux (Chelsea ctg) and Cyclops 7 (Turner Designs) fluorometers.

Phe-like fluorophore concentrations measured here with the MiniFluo (0–8 ngL⁻¹) are derived from calibrations taking into consideration non-methylated and methylated Phe (WAF solutions). These are in the same range than the concentrations of dissolved Phe (non-methylated + methylated) determined by GC-MS in the open Mediterranean Sea (0.4–4 ngL⁻¹; Berrojalbiz et al., 2011) and in the coastal Mediterranean Sea (i.e., Bay of Marseille, 5–50 ngL⁻¹; Guigue et al., 2011, 2014).

It should be noted that our fluorescence measurements were not corrected for inner filtering effects, because absorption measurements were not conducted during our glider cruises. In these waters (Nice-Calvi transect, within the 0–200 m layer), DOM absorption measurements performed with a multiple path length, liquid core wave-guide UltraPath, revealed maximal absorption coefficients at 280 nm of 0.8 m⁻¹ at the seasonal level (Organelli et al., 2014). Assuming an exponential fit for absorption spectra and a high spectral slope of 0.050 nm⁻¹ over the range 280-250 nm, maximal absorption coefficients at 250 nm would be of 3.6 nm⁻¹. Therefore, since DOM absorption coefficients at the two excitation wavelengths of the MiniFluo (255 and 275 nm) remain <10m⁻¹ in this marine area, we can reasonably assume that inner filtering effects had no significant influence on our fluorescence data (Stedmon and Bro, 2008).

It is also important to note that concentrations reported here were derived from laboratory calibrations. The relatively good agreement with historical data suggests that these calibrations are sufficient to capture the bulk part of the targeted signal. However, as it is commonly done for Chlorophyll fluorescence measurements, in situ calibration on water samples would be essential to refine estimates and assess concentrations with greater confidence.

The general understanding of the NC dynamics is that it is deeper, stronger and closer to the coast during the winter, with a relative slowing down throughout the year, becoming widest and shallowest in the early fall (Millot and Taupier-Letage, 2005; Birol et al., 2010; Declerck et al., 2016). Results presented here rather suggest no significant difference in isopycnal positioning in the NC between the spring and summer, but a wider and shallower current during the fall. It is important to note that above the seasonal variability of the NC, the mesoscale activity (e.g., meanders) may also change the positioning of the NC on much shorter timescales (Millot, 1991; Flexas, 2004), making a generalization of our snapshot transects hazardous. One important aspect to note, however, is that the offshore influence of the NC may reach further than the DYFAMED station (see for example the position of the front at x≃65 km in Figure 6). This observation shows that DYFAMED is not free from the influence of the NC as it was suggested in the past (Andersen and Prieur, 2000; Marty et al., 2002).

The mesoscale activity plays an important role for the physical-biogeochemical coupling in the NW Mediterranean, since the NC acts as a physical barrier between coastal and offshore waters (André et al., 2009; Barrier et al., 2016; Ross et al., 2016). A recent study by Piterbarg et al. (2014) suggests that the use of ocean gliders is a promising tool to address the role of mesoscale activity of such a frontal system. The clear distinction of the different zones (coastal, front, offshore) visible in Figures 6–8 confirms this potential. The relative good match between the regional Aviso altimetry product used to draw current arrows in Figure 1 is also worth noting since remote sensing is a powerful tool to address mesoscale dynamics of such current systems (e.g., Birol et al., 2010). Here, the surface expression of the NC main vein seems to disappear just offshore from DYFAMED (Figure 1) consistent with position of the NC in the glider transect (frontal zone located just offshore from DYFAMED in Figure 6). Although mesoscale interactions are beyond the scope of this study, it is worth recalling that the new physical-biogeochemical data presented here may help addressing the question. One interesting feature revealed by the MiniFluo is the difference between Trp-like and Phe-like behavior across the front. While Trp-like fluorophore exhibits a clear transition at the front (Figure 6e), Phe-like fluorophore displays a relative continuity throughout the frontal zone (Figure 6f). This intriguing difference will be further discussed below.

An important and still open question concerns the mechanisms (likely submesoscale interactions) driving the horizontal and vertical exchanges at the frontal zone. The first glider measurements of physical-biogeochemical interactions across this frontal zone was presented by Niewiadomska et al. (2008) during the winter time, i.e., when NC velocities are higher and the mesoscale activity at its maximum. Their study was carried out along the first half of the transect presented here and highlighted the importance of biogeochemical tracers for understanding submesoscale frontal dynamics. Among other things, they revealed the presence of what looks like downwelling at the front edge. Such vertical re-distribution of biogeochemical tracers at the front edge is likely to trigger photosynthesis production if nutrients are brought to the euphotic zone. In our data, spring contourplots (Figure 7) may suggest some similar downwelling at the front edge. What resembles a downwelling patch is located approximately at the same position (~25–40 km), but less pronounced in terms of depth, compared to the study of Niewiadomska et al. (2008) (see S_A and O₂ inversions in the 50–150 m depth range, and the deepening of Chl-a, Trp-like, Phe-like and BB700 data).

Another interesting submesoscale feature is the subsurface low-salinity/high-oxygen anomaly found between 100–200 m during the spring campaign (see Figures 7b,c near x = 80 km). This structure, which has also a signature in the humic-like plot (Figure 7g), is associated with a local “doming” of the isopycnals that suggests a subsurface eddy. Recent observations of submesoscale coherent vortices (SCV) that resembles the structure observed here were reported for the same geographical area (Bosse et al., 2015, 2016). However, their origins are likely different since SCV reported in these studies are observed deeper (200–500 m) and are associated with high-salinity anomalies. Nevertheless, the new structure revealed here is worth exploring in existing datasets since its physical-biochemical signature suggests a key role in lateral exchanges of biochemical tracers. Again, glider biogeochemical observations, including the one with the MiniFluo, may give useful insights on the origin and the fate of such small-scale processes.

Previous observations of Trp-like fluorescence in marine environment (not under strong anthropogenic pressure) suggest that it is particularly high in region of maximum primary productivity (Stedmon and Markager, 2005). The high vertical resolution of the present dataset rather suggests that this relationship may be more complex. For example, mean profiles show that the maximum Trp-like fluorescence is always shallower than the maximum Chl-a fluorescence (Table 3). This difference may be due to the vertical mismatch between Chl-a fluorescence and biomass, or to the vertical distribution of planktonic communities. Unfortunately, no plankton measurements are available to test these hypotheses.

The seasonal dynamics also suggests that Trp-like concentration is not simply proportional to the primary productivity: while both tracers peak in the summer, their dynamics for spring (high Chl-a/low Trp) and fall (high Trp/low Chl-a) are contrasted (see Section 3.2.2). Three hypotheses are provided to explain the seasonal decoupling between Trp and Chl-a concentrations. First, it may suggest that a specific group of microalgae may be responsible for the Trp production (Romera-Castillo et al., 2010). In the studied area, the succession of phytoplankton species during transition from a mesotrophic production regime to an oligotrophic production regime is well described (Andersen and Prieur, 2000; Marty et al., 2002). The diatom bloom that develops in March is followed by a predominance of small-size phytoplankton species (Chromophyte nanoflagellates and picoplankton). Re-increase of diatom biomass during the fall may be observed due to wind events that re-inject nutrients in the productive layer. Thus, in the summer period, when we observed the maximum Trp-like fluorescence, small size nano- and pico- plankton predominated over the phytoplankton community, whereas microphytoplankton is at its lower contribution (Ramondenc et al., 2016).

The second hypothesis involves changes in bacterial activity following the spring bloom. Trp releases in the medium have already been attributed to bacterial degradation of DOM by marine microbes (Stedmon and Cory, 2014). In this scenario, the increase in bacterial degradation of phytoplankton-derived DOM after the spring bloom is followed by an increase of the Trp pool in the surface/sub-surface waters. This hypothesis is however questionable since measurements realized after the spring bloom show relatively low Trp-like concentrations.

The third hypothesis is related to preservation/refractorization processes of DOM that could lead to an accumulation of Trp in Mediterranean surface waters. DOM accumulation during summer oligotrophy has already been observed in the surface water in NW Mediterranean (Avril, 2002; Goutx et al., 2009; Xing et al., 2014). Interestingly, the co-accumulation in the summer of other DOM descriptors (e.g., humic-like and Phe-like fluorophores) suggests that preservation/refractorization processes are predominant over degradation processes at this time of the year. Both abiotic and biotic processes may account for such preservation. On one hand, refractorization of labile DOM, of which the tryptophan may be a proxy, is triggered under UV irradiation in the surface water column (Mopper et al., 2015), subtracting refractorized DOM from the bacterial attack. On the other hand, a reduced utilization of labile compounds by nutrient-limited bacteria during oligotrophic summer conditions (Thingstad et al., 1997) may also lead to such Trp-like accumulation.

It is however very likely that the distribution of Trp reflects the integration of these three processes. The accumulation of DOM in surface Mediterranean oligotrophic waters in summer is an important driver for the efficiency of carbon and nitrogen export to the deep ocean. In the NW Mediterranean, this pattern has however mostly been assessed using variations in bulk DOC and DOM (Avril, 2002; Pujo-Pay and Conan, 2003; Santinelli, 2015). Only few studies have addressed the dynamics of DOC components (Bourguet et al., 2009; Goutx et al., 2009; Organelli et al., 2014; Xing et al., 2014), and the relative importance of diagenetic processes (microbial production, microbial degradation and phototransformation) responsible for production and removal of the DOM pool remains difficult to discern. Interestingly, Trp-like fluorescence presented here provides the first high frequency measurements of nitrogenous dissolved material (DON) component, showing this accumulation in summer. In oligotrophic regions, the DON material plays a particularly vital role in oceanic biogeochemical cycles.

The measurements of Trp-like fluorophore variations reinforce the idea that FDOM may be a good proxy for processes influencing the wider DOM pool. In the present study, the relative localization of Trp-like fluorophores along Chl-a depth profile and the temporal decoupling between Trp and Chl-a accumulation suggest that the Trp-like fluorescence is not only a marker of primary production but could also give more subtle information on the microbial processes occurring.

Ocean surface deposition of atmospheric PAHs is now being recognized as a major source of carbon in the ocean (González-Gaya et al., 2016). Their role in DOM dynamics and carbon export is however uncertain. New high resolution measurements using glider and optical sensors such as the one presented here may help addressing this question.

One of the most extensive study reporting PAHs concentrations in Mediterranean waters (Berrojalbiz et al., 2011) confirms that Phe and its alkylated (methylated) homologs were the predominant PAHs in water samples (dissolved and particulate) and in plankton biomass. However, most of the water samples used in this study were taken from the surface, where concentrations are rather low compare to the sub-surface. For the few samples taken at the DCM, these authors found that dissolved PAH concentrations were systematically higher than in surface waters. The high vertical resolution of the observations presented here confirms this pattern, which suggests an accumulation of Phe-like material just below the pycnocline at all seasons (Figure 9f). In the light of these new results, a re-interpretation of physical and biogeochemical processes affecting PAH concentrations in surface waters and in plankton (Berrojalbiz et al., 2011, e.g., schematics presented in their Figure 5) can be considered.

The good spatial coverage possible with glider-based observations also reveals other interesting features such as the Phe-like fluorescence continuum observed throughout the frontal zone during the fall (Figure 6f), while all other physical and biogeochemical tracers show striking differences on both sides of the front. Such intriguing results are certainly worth investigation. Among other questions to answer, it would be interesting to decipher whether Phe-like fluorescence observed here is related to direct atmospheric input, or rather land drainage origin. The relative abundance of Phe and methylated Phe may be used to distinguish the origin (petrogenic vs. pyrogenic) of these compounds (Prahl and Carpenter, 1983; Garrigues et al., 1995). In general, the predominance of Phe among its methylated homologs reflects a pyrogenic origin (i.e., combustion products routed by atmospheric aerosols for instance). On the other hand, high methylated Phe/Phe ratios are attributed to a petrogenic origin (i.e., crude oil derived from sea-based activities). The combination of glider measurements with traditional water sample collection would thus shed some light on Phe dynamics in the NW Mediterranean Sea.

The authors declare that the affiliation of FB and LB, the manufacturer of the MiniFluo and the glider SeaExplorer, do not preclude to the scientific independence of this scientific work which meets the standard of academic research. All other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.