Estuarine, Coastal and Shelf Science 233 (2020) 106529

Contents lists available at ScienceDirect

Estuarine, Coastal and Shelf Science

journal homepage: http://www.elsevier.com/locate/ecss

Impact of wind-driven circulation on contaminant dispersion in a

semi-enclosed bay
Camille Mazoyer a, b, *, Heleen Vanneste a, Christiane Dufresne c, Yann Ourmi�eres a,
Marcello G. Magaldi d, e, Anne Molcard a
a
Univ Toulon, Aix Marseille Univ., CNRS/INSU, IRD, MIO UM 110, Mediterranean Institute of Oceanography, La Garde, France
b
Universit�e de Toulon, IMATH, EA 2134, Avenue de l’Universit�e, 83957, La Garde, France
c
Institut des Sciences de la mer, Universit�e du Qu�ebec a
� Rimouski, 310 all�ee des Ursulines, C.P., 3300, Rimouski, Qu�ebec, G5L 3A1, Canada
d
Istituto di Scienze Marine, S.S. di La Spezia, Consiglio Nazionale delle Ricerche, Forte Santa Teresa, Pozzuolo di Lerici (SP), I-19032, Italy
e
Department of Earth and Planetary Sciences, The Johns Hopkins University, Olin Hall, 34th and North Charles Streets, Baltimore, MD 21218, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Mediterranean semi-enclosed bays are often exposed to high levels of contaminants originating from anthro­
Semi-enclosed bay pogenic activities in the bay. To assess their fate and impact on the environment, it is essential to investigate
Wind-driven circulation coastal circulation regimes which may play an important role in the dispersion of contaminants across the bay
Contaminant dispersion
and beyond. In this study, a high resolution coupled hydrodynamic - passive tracer model was combined with
MITgcm
Toulon Bay
ADCP observations, to identify major circulation patterns and associated dissolved contaminant dispersion
pathways in the contaminated semi-enclosed bay of Toulon (South of France, NW Mediterranean Sea). Two
dominant circulation patterns and two derived ones could be identified, driven by winds (Mistral and easterly
winds) and offshore water intrusions. Medium to strong Mistral events (> 6 m s 1) with a WNW direction cause a
bi-layer pattern with surface waters flowing out of the bay and marine waters entering at depth. Less frequently,
west Mistral winds of medium to strong strength (> 6 m s 1) may generate an anticyclonic circulation. During
easterly wind conditions (> 6 m s 1), an inward flow can be observed which is sometimes reinforced by offshore
water intrusions, probably from the local boundary current, the Northern Current (NC). Furthermore, dissolved
contaminant dispersion pathways were simulated under typical wind forcing conditions with three point sources
of copper (Cu) that were identified based on surface Cu observations. While most of the WNW Mistral wind
events transport dissolved copper plumes across and out of the bay, contaminant dispersion can remain confined
to the bay under certain west mistral conditions. Conversely, during easterly wind events, contaminants are
exiting the bay as a narrow vein along the Saint-Mandrier peninsula, before probably converging on the NC
offshore. Accordingly, this study demonstrates the important impact of hydrodynamic-driven processes on the
dispersion of contaminants within a semi-enclosed bay.

1. Introduction largest French naval base while located at less than 20 km from the
Marine National Park of Port-Cros (Fig. 1). The core of this marine
Today, marine coastal areas suffer from an extremely high anthro­ protected area includes two islands within the Hyeres Bay, namely
pogenic pressure, which is the result of a continuous development of Porquerolles and Port-Cros Islands. The surrounding waters, hosting a
human activities in these areas for centuries. These activities have lead unique marine ecosystem, are highly monitored to maintain its quality,
to pollution hot spots, endangering the water quality of ports, beaches, protect its biodiversity and promote a sustainable development of
aquaculture zones and marine protected areas, challenging local gov­ tourism and local communities. Its proximity to the highly industrialised
ernments to find a balance between economical and environmental in­ Toulon city requires a better understanding of contaminant distribution
terests. To this end, Toulon Bay (South of France, NW Mediterranean within the bay and its surroundings, to assess its potential impact on the
Sea) is an area of interest as it is a semi-enclosed bay harbouring the water quality of the national park. The long history of naval, industrial

* Corresponding author. Univ Toulon, Aix Marseille Univ., CNRS/INSU, IRD, MIO UM 110, Mediterranean Institute of Oceanography, La Garde, France.
E-mail address: [email protected] (C. Mazoyer).

https://doi.org/10.1016/j.ecss.2019.106529
Received 14 October 2018; Received in revised form 25 November 2019; Accepted 4 December 2019
Available online 15 December 2019
0272-7714/© 2019 Published by Elsevier Ltd.
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

and commercial activities in the Toulon Bay has important environ­ between the small (SB) and the large bay (LB), while Duffa et al. (2011),
mental consequences for the bay. High concentrations of trace metals (e. Dufresne et al. (2014) and Dufresne et al. (2018) recently constructed a
g. lead, copper, zinc, …) have been measured in sediments (Tessier et al., hydro-sedimentary model of the Toulon Bay, to create a post-accidental
2011; Pougnet et al., 2014) and surface waters (Dang et al., 2015; Jean management tool in the event of radionuclide releases to the environ­
et al., 2012; Coclet et al., 2018). For both matrices, the contamination is ment. These studies showed that the circulation is strongly dependent on
the strongest in the enclosed north-western part of the bay (i.e. the small wind conditions, causing for instance reversals in the water column in
Bay (SB)) where surface sediments contain up to ca. 1000 times more the channel connecting the SB with the LB (Dufresne et al., 2014). While
mercury compared to background values, while up to 80, 40 and 31 these studies focused on the quantification of the exchanges between the
times more copper, zinc and lead, respectively (Tessier et al., 2011). SB and the LB, here we investigate the circulation in the LB under
Surface waters on the other hand, are ca. 40 and 70-fold enriched in various wind conditions and the resulting exchanges with offshore
respectively dissolved lead and copper (Coclet et al., 2018), compared to waters.
trace metal concentrations observed in the open Mediterranean Sea The aim of this study is two-fold. The first objective is to identify
(Tovar-Sanchez et al., 2014). While Toulon Bay might be considered as dominant circulation patterns inside the bay, using an ocean circulation
one of the most polluted marine areas of the northern Mediterranean Sea model (MITgcm) and ADCP observations in the Toulon Bay. The second
(Tessier et al., 2011), contamination might be exported to offshore objective is to investigate contaminant transport, by defining dispersion
waters through hydrodynamic processes and reach sensitive areas as the patterns and assess the potential impact on key areas within the bay and
Marine National Park of Port-Cros. beyond.
Very few studies have focused on the hydrodynamic processes within After a description of the studied site, the model configuration and in
the Toulon Bay. Millot et al. (1981) and Tine (1981) were the first to use situ observations, we describe the wind-driven circulation within the
current profiler data to quantify water exchanges through the channel bay. An analysis of the contaminant distribution as observed in March

Fig. 1. A) Map of Toulon Bay and its surrounding area (SB: small bay, LB: large bay, S.-M. P.: Saint-Mandrier peninsula, G. P.: Giens peninsula, Porq.:Porquerolles, P.-
Cros: Port-Cros). The black box indicates the geographical extent of the model domain of TBAY100. B) Zoom of Toulon Bay with bathymetry (m). The white line
between the SB and the LB indicates the impervious seawall on the whole water column. Orange dots indicate ADCP locations: Cape Cepet and Carqueiranne. The red
line indicates the position of the vertical cross-section presented in Figs. 6 and 8. Black dots indicate the positions of the three contaminant sources: Toulon civil port
(source 1), la Seyne-sur-Mer civil port (source 2), and Saint-Mandrier civil port (source 3). Red boxes: locations for which temporal contaminant dispersion is
discussed in section 4.0.2. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

2
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

2017 is discussed and the model is used as a tool to simulate the Free-slip boundary conditions are applied at the bottom and on the
dispersion of dissolved contaminants based on wind conditions. lateral boundaries.

2. Material and methods

2.3. MITgcm high-resolution configuration TBAY100

2.1. Study area

The Toulon Bay domain spans a longitudinal range from 5:88 ∘ to
6:13 ∘ E and a latitudinal range from 43:04 ∘ to 43:13 ∘ N (Fig. 1). The
The water circulation of the southeast French Mediterranean coast, is
horizontal mesh has 200 � 100 grid points with 100 m resolution. There
characterized by a strong geostrophic boundary current flowing along
are 75 depth levels with a 50 cm minimum layer thickness at the surface,
the North Western Mediterranean Sea coast counter-clockwise, the
increasing to about 30 m near the seabed. Minimum depth is 3 m and
Northern Current (NC). This current has been largely studied especially
maximum depth is 602 m in the canyon at the south boundary (Fig. 1).
focusing on its behaviour upstream in the Ligurian basin (Taupier-Let­
The timestep is 5 s. Outputs are saved every hour to catch most of the
age and Millot, 1986; Alberola et al., 1995a; Sammari et al., 1995;
dynamics variability. The TBAY100 configuration has two open
Bethoux et al., 1982), downstream at the eastern side of the Gulf of Lion
boundaries (south and west) and is forced with a NEMO-based config­
(Alberola and Millot, 2003; Conan and Millot, 1995; Flexas et al., 2002),
uration (Declerck et al., 2016) which is a 1/192� configuration of the
and along its shelf (Auclair et al., 2001; Petrenko, 2003; Rubio et al.,
Var coast (hereafter NIDOR192), including the Hyeres Bay and the “^Iles
2009; Lapouyade and Durrieu de madron, 2001). However few studies
d’Or” islands (i.e. Porquerolles, Port-Cros and Levant; Fig. 1). Indeed,
on the NC circulation were conducted in the Var region, which can be
TBAY100 is the third level of nested coarser models (GLAZUR64,
seen as an area of transition (Ourmi�eres et al., 2011; Guihou et al., 2013;
NIDOR192). GLAZUR64 and NIDOR192 have been validated during
Declerck et al., 2016). These studies showed that in this area the NC has
previous studies (Ourmi� eres et al., 2011; Guihou et al., 2013). The
a meso and sub-mesoscale activity which can create exchanges between
rigorous downscaling set-up guarantees good confidence in the last and
the shelves and offshore waters. For instance Declerck et al. (2016)
highest resolution configuration TBAY100. The horizontal velocities and
modelled intrusions into the Hyeres Bay, nearby the Toulon Bay. They
mass fields open boundary conditions are uploaded daily. The opera­
showed that the NC, through meanders and vortices dynamics, can
tional global model ARPEGE from Meteo-France is used as atmospheric
either act as a barrier or favour in- and offshore water mixing into
forcing: it has a 1/10� spatial resolution (about 9.5 km at our latitude)
semi-enclosed bays as Hyeres Bay. Here, we will focus on the Toulon
and a 3 h temporal resolution.
Bay, analyze its inner wind-driven circulation and the possible offshore
The starting date is the 1st of February 2016, after a 1 month
water intrusions that may be driven by the Northern Current that flows
simulation (spin-up evaluated to 15 days). Initial conditions are
off the bay.
computed from NIDOR192 output.
Toulon Bay is a shallow semi-closed area (about 10 km � 15 km),
divided into two bays, separated by a seawall of 1.4 km long modelled
impervious on the whole water column: a small bay (SB; � 10 km2) 2.4. MITgcm ptracers package
which hosts mostly all the anthropogenic activities and a large bay (LB;
� 42 km2) connected to the offshore sea (Fig. 1). The LB is delimited by We used in this study the ptracers package from the MITgcm model.
Cape Carqueiranne to the east, and by the Saint-Mandrier peninsula to This is a passive tracer package, in which the dissolved tracers do not
the west. At the south of the LB, the bathymetry is very steep with a affect actively the physics of the ocean as they do not affect the water
canyon of ca. 600 m deep. density. Tracers are initialized with precise quantities (for details see
Since there is no significant tide motion (tidal amplitude of Section 4), and are released every time step.
approximately 20 cm) (Millot et al., 1981; Alberola et al., 1995b) in the
Toulon Bay, its circulation is supposed to be mostly wind-driven with
2.5. ADCP
possible boundary current intrusions. The wind forcing has an important
spatial and temporal variability with strong events (Millot et al., 1981;
Horizontal velocity data from Acoustic Doppler Current Profilers
Dufresne et al., 2014). Three wind situations can be distinguished:
(ADCP, 600 kHz, RDInstruments) were also used. Sampling frequency
Mistral, easterly wind and calm weather. Toulon is surrounded by
was 10 min. Two ADCPs were moored in the south of the LB between
mountains that deflect these winds: the Mistral blows from the
January 2012 and April 2012 (see Fig. 1B). Their location were chosen
west/north-west, especially in winter and spring. It is a cold, dry and
to measure the exchanges between the LB and offshore. ADCP 1 was
strong wind unlike the easterly wind (from north-east to south-east) that
located near Cape Cepet and ADCP 2 was located near Cape Carqueir­
usually brings clouds, rain and waves.
anne. The ADCPs were bottom-moored at 32 m, and 29 m respectively.
The water column was divided into 35 1-m thick bins with the first bin at
2.2. Model description
2 m above the bottom.
The M.I.T General Circulation Model (MITgcm) is a primitive equa­
tion ocean circulation model (Marshall et al., 1997). It is a z-coordinate 2.6. Diagnostics - composite pictures
finite-volume model that solves the incompressible Navier-Stokes
equations under Boussinesq approximation, on an Arakawa-C grid. In Previous studies on the area (Millot et al., 1981; Dufresne et al.,
this model, the free surface equation is based on a pressure correction 2014) concluded that atmospheric forcings strongly influence hydro­
method. For this study, we use the hydrostatic approximation. The time dynamics into the Toulon Bay. Strong weather conditions cause highly
scheme is a quasi second order Adams Bashford and the horizontal variable currents (due to various processes such as offshore swell, wind
advection scheme is the third-order DST (direct spacetime) flux limiter. gusts, …) combined with offshore waters intrusions and temporal re­
The horizontal subgrid-scale mixing is parametrised with the bihar­ versals. The wind rose depicted in Fig. 2 is computed from the
monic Leith viscosity (see Leith, 1968, 1996). Vertical background vis­ Meteo-France ARPEGE model dataset and used as external atmospheric
cosity is set with Az ¼ 10 5 m2 s 1, a standard value. We use the forcings for the oceanic model (TBAY100). Two major wind regimes can
turbulent closure scheme K-profile parameterisation (KPP) from Large be defined accordingly:
et al. (1994), to model vertical sub-grid scale mixing. For atmospheric
forcing, we use the bulk formulation from Large and Yeager (2004), and 1. A medium to strong Mistral (258.75� –326.25� ): a west to north-west
open boundary conditions are prescribed and relaxed by a sponge layer. wind with an intensity between 6 and 24 m s 1 (24.9% of the time);

3
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

will allow the assessment of long-term temporal variability in contam­

inant dispersion, by running a continuous contaminant release from 3
sources representing the main civil ports in the SB (black dots on
Fig. 1B).

3. Circulation regimes

3.1. Surface composite pictures

Composite pictures of surface currents (0–5 m) for Mistral and

easterly wind, obtained with model results, show opposite current di­
rections induced by the two wind regimes (Fig. 3a,c). Overall, Mistral
drags the surface water southeastwards (Fig. 3a) enhancing the surface
current intensity near Cape Cepet and Cape Carqueiranne, whereas
easterly wind pushes it towards the north-west (Fig. 3c). While Mistral
events drive the surface circulation across the entire bay, the influence
Fig. 2. Windrose for winds used as external forcings from Meteo-France of easterly wind in the LB is limited (indicated by the small arrows on
ARPEGE model, from February 2016 to May 2017. Directions are divided Fig. 3c).
into 16 slices of 22.5� each. Mistral wind is composed of west-northwest winds However in the southern part of the model domain, current in­
with three main directions: d12, d13, d14 from 258.75� to 326.25� . Easterly tensities are higher and have a stronger gradient compared to Mistral
wind is composed of east-northeast winds with three main directions: d3, d4 conditions. This may be due to intrusions of offshore waters that amplify
and d5 from 56.25� to 123.75� . Note the windrose shows the direction from the local wind-driven current. To evaluate the relative importance of
where the wind blows. these intrusions, we computed a transport index across two transects
(Fig. 3c and d), inspired from the method used by Barrier et al. (2016)
2. A medium to strong easterly wind (56.25� –101.25� ): a south-east to a for the quantification of intrusions in the Gulf of Lion. While T1 (Fig. 4)
north-east wind with an intensity between 6 and 24 m s 1 (9% of the allows us to trace the occurrence of surface offshore water intrusions
time); near the southern boundary of the model domain, T2 identifies water
intrusions in the LB. The northward meridional current velocity was
When the wind intensity is lower than 6 m s 1, we consider to be used to calculate the positive transport Tþ across each transect over the
under calm conditions (i.e. no wind). The circulation within the bay is surface (2–10 m depth to focus on the core of the water intrusions). The
therefore analyzed for each wind condition. We averaged the horizontal transect locations were chosen where the current intensities are large
velocity fields associated with the specific wind regime, for the surface combined with a standard deviation greater than 0.1 m s 1 for T1 and
layer (0–5 m) and across the capes section (red line on Fig. 1) defining 0.05 m s 1 for T2 (Fig. 3c and d). The T1 transect ends westward at the
the LB limit. These composite pictures are a powerful tool to highlight an canyon location which is a natural bathymetric guidance of offshore
average circulation and its variability under specific conditions. water intrusions. Transport is normalised by its temporal mean < Tþ >:

TðtÞþ
IðtÞ ¼ (1)
2.7. Datasets < Tþ >
We consider intrusions to occur when index IðtÞ � 1, i.e. when the
Table 1 gives an overview of all data presented, and their date cross-section positive transport exceeds its mean. Fig. 4 represents the
ranges. As one of our main goal is to study contaminant dispersion, we time evolution between February 2016 and April 2017 of the wind (a),
have decided to simulate the year 2016–2017 to coincide with dissolved T1 (b) and T2(c). As expected, most (94%) of the easterly wind episodes
copper observations available for March 2017. Simulations are validated (grey boxes) are associated with T1 intrusions. More than 13 periods
by a comparison with ADCP observations obtained in 2012 (see section reveal strong intrusions with a high T1 index (up to 6–7).
2.5). In order to focus on the wind-driven circulation, the velocities Hence, during medium to strong easterly wind periods, offshore
simulated by the model and measured by the ADCP have been filtered. surface waters are pushed toward the LB, bending in a cyclonic circu­
Data have been processed with the PL66 filter (Limeburner et al., 1985), lation at the bay entrance. Whether or not the offshore waters enter the
in order to filter all motions under 25 h, including inertial oscillations, bay can be evaluated by the T2 index. During three episodes (episodes 1,
diurnal and semidiurnal tidal constituents and eigen motions. 2 and 6 in Fig. 4), a T2 index higher than 3 highlights the presence of an
In the next sections, we will first identify the main circulation pat­ intrusion within the LB. At other times (episodes 3,4 and 5), the T2 index
terns induced by regional wind characteristics and provide a quantita­ remains below the index value 3, indicating that the current mainly
tive validation using ADCP observations. Subsequently, results of 2 days bypasses the LB. T2 intrusions have been estimated to occur in the LB
simulations with idealized wind conditions will be presented showing 78% of the time during easterly wind episodes.
the principal contaminant dispersion pathways in the Toulon Bay. These To evaluate the influence of the wind on the intrusion into the LB, a
simulations benefit from the 1-month spin-up of the longer simulation scatter plot in Fig. 5 was used to represent the T1 and T2 indexes ac­
TBAY100-PHYS. Eventually, the TBAY100-TRAC 14-month simulation cording to the wind intensity and direction.
Very high values of T1 occur 16% of the time for ENE strong winds
Table 1 (all points circled by the black line in Fig. 5) corresponding to a non
The data ranges of the presented data. intrusion circulation (T2 small), while high values of T2 (and T1 larger
Data (observations or simulations) Dates than 1) occur with medium (around 8 m s 1) northeasterly wind
(90–120� ) (all points circled by the red line in Fig. 5b). The wind di­
ADCP Cepet January–April 2012
ADCP Carqueiranne January–April 2012 rection combined with its intensity seem to be the main forcings, trig­
Seawater samples 20 March 2017 gering intrusions of offshore water into the LB. The remaining 6%
Simulation TBAY100-PHYS (physics) February 2016–April 2017 corresponds to situations without intrusions near the southern boundary
Simulation TBAY100-TRAC (tracers) February 2016–March 2017 nor into the LB.
Idealized simulations /

4
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

Fig. 3. Left panels (a,c) are composite pictures of surface currents (m s 1). Right panels (b,d) are standard deviations of surface currents for the composite pictures (m
s 1). Top panels (a,b) are for medium to strong Mistral (6–24 m s 1), bottom panels (c,d) are for medium to strong easterly wind (6–24 m s 1). The black lines T1,T2
are the cross-sections for the meridional transport estimations.

Fig. 4. Model results during February 2016–April 2017 period including (a) wind intensity (m s 1) and direction (from which the wind blew) from Meteo-France
ARPEGE model; (b) index intrusion for transect T1; (c) index intrusion for transect T2. Locations of T1 and T2 transects are black lines in Fig. 3. Grey boxes are
medium to strong easterly wind periods. The horizontal dotted line corresponds to the detection threshold of intrusions.

3.2. Cross-section composite pictures flows southward (blue) towards the open sea with a maximum mean
value of 0.18 m s 1 near Cape Carqueiranne, whereas an incoming flow
After having investigated surface circulation, this section aims at enters (red) the LB at the bottom of the water column (10–40 m). The
studying the circulation at depth. To study the impact of wind conditions average inversion depth is at ca. 10 m depth at the center of the channel,
on the entire water column, circulation has been studied along a vertical and it can deepen for strong winds.
cross-section between Cape Cepet and Cape Carqueiranne (red line in In order to evaluate the occurrence of the bi-layer circulation, the
Fig. 1B). meridional velocities were integrated at the surface (0–10 m) (blue box
For a medium to strong Mistral wind (6–24 m s 1), results show a bi- in Fig. 8a) and below the inversion depth (20–40 m) (red box in Fig. 8a)
layer circulation into the bay (Fig. 6a), as it was shown for certain time in order to avoid its variability. Results show that the bi-layer pattern,
periods by Duffa et al. (2011). We clearly see here that the surface layer defined by an inward surface flux, and outward deep flux, and according

5
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

Fig. 5. Scatter plot of wind speed (m s 1) versus wind direction (deg). The colorbar indicates the value of T1 index (a) or T2 index (b). Red circle corresponds to
winds with an entrance in the domain and in the LB. Black circle corresponds to winds with an entrance in the domain but no entrance in the LB. Red arrow
corresponds to a wind of 8 m s 1 and black arrow corresponds to a wind of 16 m s 1. (For interpretation of the references to color in this figure legend, the reader is
referred to the Web version of this article.)

Fig. 6. Left panels (a,c) are composite pictures of meridional velocity on a vertical cross-section between Cape Cepet and Cape Carqueiranne (red line in Fig. 1B).
Right panels (b,d) are standard deviations of the meridional velocities for the composite pictures. Top panels (a,b) are for medium to strong intensity Mistral (6–24 m
s 1), bottom panels (c, d) are for medium to strong easterly wind (6–24 m s 1). (For interpretation of the references to color in this figure legend, the reader is
referred to the Web version of this article.)

to our criteria, occurs for 81% of the Mistral events. The standard de­ corresponding to the Mistral regime, is globally in the direction of the
viation on the outcoming current along Cape Carqueiranne (Fig. 6b) wind near the capes and upwind in the middle of the section. The same
suggests that velocity variations are more important in this outflow observation was first shown for an idealized long narrow lake by Csa­
compared to Cape Cepet. A bathymetric effect is a probable explanation nady (1973), where the transport is downwind in shallow parts of the
for this current acceleration near this cape induced by stronger wind. lake and upwind in deeper parts. This observation has also been made in
Overall the standard deviation of the meridional current across the the Gulf of Trieste a Mediterranean semi-enclosed basin (Mala�ci�c et al.,
vertical section has the same order of magnitude of the current itself. 2012), in coastal lagoons (Fiandrino et al., 2017) and in a shallow
This can suggest that the bi-layer system may not be the only possible estuarine lake (Schoen et al., 2014) for wind-driven circulation.
scheme. The Toulon Bay bi-layer pattern typical of the classic “wind driven
The transport across this section (Fig. 7b) for all the events channel circulation” is the dominant pattern during Mistral wind

6
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

Fig. 7. Left panel (a) is a depth integration of meridional velocity on a cross-section between Cape Cepet and Cape Carqueiranne (red line in Fig. 1B) for a number of
Mistral events (6–24 m s 1). Right panel (b) is the standard deviation of the meridional velocity through time. Red and blue boxes are locations where STD � 1:5. (For
interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

conditions, occurring 81% of the Mistral-driven circulations. The com­ inward flow into the LB on the east side (Fig. 3c), with a core positioned
posite picture resulting in this BLP is depicted in Fig. 8a. The events are at the surface near Cape Carqueiranne (Fig. 6c). The outflowing current
selected according to the transport value in the upper/lower layers, follows the bathymetry at the westside of the cross section, near Cape
imposed negative at the surface and positive at depth (blue/red boxes in Cepet, with its core at ca. 50 m depth. A small confined inward current
the figure). appears to occur at the edge of the cross-section (Cape Cepet), although
Looking closer around the capes (boxes in Fig. 7b), the standard standard deviations (Fig. 6d) show that its behaviour is very variable.
deviations of the integrated meridional velocity have very high values, Indeed, this current can also disappear, for instance when the surface
exceeding 1.5 m2 s 1, revealing inversions of the currents along the current along the entire section enters the LB (not shown). Note that a
capes. In particular, another possible circulation pattern can result from significant fraction of the surface current bypasses the LB.
the Mistral wind regime, when the inflow (upwind) at the western Cape The analysis of the composite pictures described above, not only
Cepet emerges and fills the whole water column. By extracting the corroborates the previously identified highly variable nature of the hy­
remaining events, having an inflow at the western cape, and an outflow drodynamics in the Toulon Bay (Duffa et al., 2011; Dufresne et al.,
at the eastern cape (Fig. 8b, red boxes), an anticyclonic circulation 2014), but it also clearly allows to associate circulation patterns with
within the LB is revealed during 13% of the Mistral events. typical wind characteristics.
Medium to strong easterly winds (6–24 m s 1) appear to generate an
3.3. Main circulation patterns

A qualitative and quantitative analysis of the model results, using the

composite pictures described in section 3.1 and 3.2, allows the identi­
fication of four distinct circulation patterns, depending on the wind
regimes and the offshore water intrusions:

1. The Bi-layer pattern (BLP; Fig. 9a): the predominant circulation

pattern during medium to strong Mistral events (81%). It occurs
mainly with a west-northwest wind, with a median intensity of 8.88
m s 1. It is characterised by a unidirectional surface outflow to the
southeast towards the open sea inducing a bottom current entering
the Toulon Bay (Fig. 8a). The maximum velocities observed near
Cape Carqueiranne are due to a cape effect;
2. The Anticyclonic pattern (ACP, a derived circulation from the BLP
pattern; Fig. 9b): a barotropic inward flow near Saint-Mandrier
peninsula that follows the bathymetry in an anticyclonic way and
leaves the LB near Cape Carqueiranne. This circulation pattern oc­
curs for 13% of medium to strong Mistral wind events. It occurs with
a west to west-northwest wind, with a median intensity of 8.22 m s 1
and mostly during winter periods, meaning that the absence of
stratification can play a role for its occurrence. In most cases, a small
confined current flows southeastwards along the Saint-Mandrier
peninsula (Fig. 8b);
3. The Intrusion pattern (IP; Fig. 9c): an inward flow near Cape Car­
queiranne, caused by an offshore water intrusion into the LB during
easterly wind events (with median intensities of 9.15 m s 1). This
Fig. 8. Composite pictures of meridional velocity on a vertical cross-section
pattern has been highlighted by a transport analysis to detect
between Cape Cepet and Cape Carqueiranne (red line in Fig. 1B). Top panel
(a) is for a bi-layer circulation (Mistral 6–24 m s 1)), bottom (b) is for an an­ offshore water intrusions into the LB. It has been identified to occur
ticyclonic circulation (Mistral 6–24 m s 1). Red and blue boxes are used for for 78% of the time when easterly winds exceed 6 m s 1 and for wind
transport analysis (red for positive and blue for negative). (For interpretation of from 80 to 120� most of the time. South of the open boundary line,
the references to color in this figure legend, the reader is referred to the Web this circulation pattern has a cyclonic character, probably mainly
version of this article.) driven by the along-coast pathway of the NC. These offshore water

7
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

3.4. Validation with in-situ observations

The Toulon Bay model TBAY100 is the third level of nested coarser
models (GLAZUR, NIDOR) (see section 2.3), with accurate forcings at
the southern and western boundaries. Nevertheless, validations with in
situ observations is necessary and will emphasise our confidence in the
previous results. In situ current observations in such coastal areas are
scarse. ADCP data were only available from January to April 2012. The
model results are compared to these ADCP observations by sorting and
averaging the velocities obtained during the two typical wind conditions
of the area. Surface (top 2.5 m) and bottom currents (28 m) resulting
composites, computed by the model and recorded by both ADCPs at
Cape Carqueiranne and Cepet, are presented in the current rose dia­
grams in Fig. 10 as average percentages throughout the ADCP deploy­
ment period and the total period of the model simulation. The figure is
split in 4 boxes: left ones correspond to Cape Cepet ADCP analysis, right
ones correspond to Cape Carqueiranne; upper blue boxes correspond to
Mistral wind regime, and lower orange ones to easterly wind regime. In
each box 4 current roses are designed for the observation, the model, the
surface and the bottom.
At Cape Cepet, during medium and strong Mistral winds, (upper left
box) surface modelled and observed currents flow towards the same
direction, i.e. southeast (Fig. 10a and b), although modelled intensities
are stronger (0.35 m s 1 versus 0.20 m s 1). At the bottom, current in­
tensities and directions are overall the same, though observations show
more variability in direction (Fig. 10c and d). For medium and strong
easterly wind events (lower left box), differences appear between
modelled and observed surface currents (Fig. 10i and j). While the ob­
servations show a north-westwards and south-southeastward flow with
a maximum intensity of ca. 0.20 m s 1, modelled currents flow towards
the west to southwest at a significant higher velocity. At the bottom,
currents have similar directions, flowing towards the south-east, with
slightly more intense modelled currents (Fig. 10k and l). A possible
explanation for the mismatches between the ADCP and model data
(mostly situations in Fig. 10i and j), is the bathymetry approximation in
the TBAY100 model, in particular near the capes, where the bathymetry
is shallow and steep. We observe in the model outputs that current
variability is very important near the Cape Cepet location and that this
location is near the area where the current bypasses the LB in the model.
At Cap Carqueiranne, during medium and strong Mistral winds
(upper right box), surface currents are stronger than at Cape Cepet and
can reach up to 0.25 m s 1 during this period (Fig. 10e and f). Surface
Fig. 9. Composite pictures of surface velocities (m s 1) showing the 4 circu­ modelled and observed currents have similar directions and are flowing
lation patterns: a) BLP pattern, b) ACP pattern, c) IP pattern, d) NIP pattern. A to the southeast out of the large bay. At the bottom, modelled and
wind rose is presented for each pattern. observed currents are weaker than at the surface (a maximum of ca.
0.15 m s 1) but they have almost the same directions, i.e. south-
intrusions depend on the upstream circulation features, and more southeast for both model and observations (Fig. 10g and h). Accord­
specifically on whether a NC intrusion into the Hyeres Bay has ingly, model and observations are very coherent at Cape Carqueiranne
occurred (Declerck et al., 2016). for Mistral conditions both in intensity and direction. During medium
4. The Non Intrusion pattern (NIP; Fig. 9d): this derivated regime from and strong easterly wind (lower right box), surface currents flow
the IP pattern has also an anti-clockwise circulation south of the open northwestwards with a maximum intensity of 0.3 m s 1 for both model
boundary line, induced by medium to strong easterly wind events outputs and observations (Fig. 10m and n), which may correspond to an
with a predominant northeast direction (60–90� ) with a median in­ intrusion of offshore waters into the LB. Overall, the observed surface
tensity of 15.18 m s 1, which is greater than for the IP pattern. It currents at the Cape Carqueiranne location seems to be more responsive
occurs 16% of the time when easterly winds exceed 6 m s 1. This to wind, as they appear to be stronger than at Cape Cepet, and are
pattern is characterized by strong current intensities located mainly approximately downwind for both wind conditions. Finally, bottom
south of the LB and by no water intrusions in the LB. currents during medium and strong easterly wind are weaker compared
to the surface currents and have the same maximum intensity of 0.15 m
It is worth mentioning that the criteria used to identify the four s 1 for modelled data and observations. While the observed currents do
distinct circulation patterns can be differently chosen, and may give not have a dominant direction, the modelled currents flow northwest­
slightly different percentage of occurrence. However, the four patterns wards and southeastwards (Fig. 10o and p).
are clearly present in the simulations, and this will be confirmed with To summarize, the model results are coherent with most of the
the distribution of contaminants (section 4). observed velocities direction at the ADCP locations. In terms of in­
tensity, the model overestimates most of the time the occurrence and the
intensity. The most important discrepancies appear at the western side
of the LB boundary, at Cape Cepet, during easterly wind regimes, in

8
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

Fig. 10. Current roses (m s 1) for ADCP observations and model output at location Cepet (left-hand side), and Carqueiranne (right-hand-side). Blue boxes (top) are
for medium to strong Mistral conditions, orange boxes (bottom) for medium to strong easterly wind condition. For each box, observations (left) and model output
(right) are represented for surface (top) and bottom (bottom) currents. Note current roses show the direction towards which the current flows. (For interpretation of
the references to color in this figure legend, the reader is referred to the Web version of this article.)

particular at the surface. However the current rose of the ADCP Toulon Bay: the preceding conditions, offshore water intrusions, wind
(Fig. 10i) and of the model, do not show clear dominant currents, intensity and direction. The important impact of wind on hydrody­
indicating that the circulation is very variable in this particular location. namics within small semi-enclosed bays, was also observed in the Gulf of
This variability is confirmed by the model results suggesting (Figs.3a & Fos, where wind events have a stronger influence on hydrodynamics
8b) that for the anticyclonic circulation, offshore waters may enter or than the shelf offshore current (Ulses et al., 2005). Even in small basins
not into the LB, leading to a different response at the ADCP location. under strong tidal influence, wind events may interfere on local circu­
Moreover we should recall that the ADCP experiment lasted 4 months in lation, such as in Southern Australia (Black et al., 1993). The circulation
2012 and the model simulation from February 2016 to April 2017. patterns identified in this section highlight the possible export paths for
Interannual and seasonal variability could also explain part of the contaminants which are discussed in the next section.
discrepancies.
If we compare the schematic circulations suggested by the model 4. Contaminant distribution
outputs and presented in the previous sections, during Mistral regime
the results are very coherent with both the BLP and the ACP. During Now that the circulation has been validated, the second objective is
easterly wind regime, the NIP appear at the bottom but not clearly at the to better understand the distribution of contaminants in the area.
surface. More ADCP moorings would be necessary to validate this cir­ Chemical analysis of surface seawater samples from the Toulon Bay,
culation suggested by the simulations. obtained within the projects METFLUX, PREVENT and the MERMEX
In conclusion, several factors influence the hydrodynamics of the program by N. Layglon (MIO, University of Toulon) following an

9
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

established method (Jean et al., 2012), have revealed the presence of s 1), which was derived from in situ measurements in San Diego Bay
elevated copper concentrations in March 2017, varying from 3 nM in the (Earley et al., 2014). Accordingly, AR provides a minimum baseline for
LB to 113 nM in the SB (Fig. 11). While the distribution of dissolved leaching rates of Cu from antifouling paints into the marine
copper is relatively uniform in LB (and close to background concentra­ environment.
tions for the Mediterranean Sea, 1.6 nM; Tovar-Sanchez et al., 2014),
steep concentration gradients were observed in the SB with maximum 4.1. Wind dependent contaminant distribution patterns
values measured in the docks of the civil (113 nM) and navy (67 nM)
ports of Toulon and in the civil ports of La Seyne-sur-Mer (40 nM) and As shown in section 3.3 several circulation patterns could be corre­
Saint-Mandrier (82 nM). This is in agreement with previous observa­ lated with two main wind directions: the Mistral and the easterly wind.
tions (e.g. Coclet et al., 2018), hence these ports can be considered as The circulation patterns were depending on the wind intensity and di­
point sources of dissolved copper to the bay. Common sources of copper rection for the former and the existence of offshore waters intrusions for
in harbours are antifouling paints which are applied to the hulls of boats the latter. The influence of these four circulation patterns on the
to prevent growth of fouling organisms (e.g. Schiff et al., 2007). Since dispersion of dissolved contaminants in the Toulon Bay was examined
the ban of triorganotin formulations, most modern marine antifouling by running idealized simulations, imposing a constant wind during 2
paints contain a Cu-based biocidal pigment (Turner, 2010) which has days instead of using the ARPEGE atmospheric model as external wind
led to increased levels of dissolved copper in coastal waters, especially in forcings. Care was taken to select two days with wind conditions similar
and near ports (e.g. Biggs and D’Anna, 2012; Warnken et al., 2004). to the imposed one, to avoid too important numerical instabilities, and a
Accordingly, antifouling paints are hypothesised to be a major source of possible model blow-up. The same patterns were obtained for multiple
dissolved copper to the Toulon Bay. dates for each of the four wind case scenarios and associated circulation
To understand and evaluate the impact of the elevated levels of patterns: 1) Mistral 8 m s 1 (ACP), 2) Mistral 16 m s 1 (BLP), 3) easterly
dissolved copper on the water quality of the Toulon Bay, distribution wind with offshore waters intrusion (IP) and 4) easterly wind without
patterns were modelled under the influence of different wind conditions offshore waters intrusion (NIP), underlining the representativeness of
and seasonal variability was assessed via a 14-month simulation. Copper these contaminant distribution patterns for each scenario.
has been shown to have a high affinity for organic matter in the marine At medium wind intensity (8 m s 1), the plume of dissolved con­
environment, forming dissolved organic copper species (e.g. Buck et al., taminants released in the SB spreads into almost the entire LB (Fig. 12-
2007). Its distribution can thus be modelled as a passive tracer. Based on A1). Overall concentration maxima occur at the surface in the entire bay
the surface distribution map of dissolved copper (Fig. 11), copper was with exception of a small eddy, which is formed at the exit of the SB. This
released as a passive tracer at three sources in the TBAY100 model: the contaminant plume is originated from source 2 and 3 that are evacuated
civil ports of (1) Toulon, (2) la Seyne-sur-Mer and (3) Saint-Mandrier from the SB during the time span of 2 days. The waters exiting the SB
(see Fig. 1B for the precise location in the TBAY100 model). As the encounter an incoming vein flowing westward into the LB and mix
majority of the berths in the ports are occupied year-round, copper because of this front, preventing them from spreading in the whole LB.
leaching from antifouling paints is considered to be a constant process The SB waters bypass this front by following an anticyclonic motion into
and was therefore modelled as a continuous source. The quantity of the LB, spreading the contaminant at the surface of the LB (Fig. 12-A2).
copper released per time step at each source (Cusource ; nmol dt 1) was The dispersion of the contaminants released at the Toulon civil port
calculated based on the capacity of each port (i.e. number of berths) (source 1) remains confined to the northeast sector of the SB. Mistral
using the following equation: winds of high intensity (16 m s 1) export dissolved contaminants to the
LB and beyond, until the Giens peninsula, in 2 days time (Fig. 12-B1)
Cusource ¼ Nberths � Smean � AR � dt (2)
demonstrating the wind-driven circulation in the Toulon Bay area. The
where Nberths is the number of berths per port, S is the standard estimate majority of the contaminants remain at the surface (Fig. 12-B2).
for wetted hull surface area (i.e. 41 m2) of a typical recreational boat (12 In the case of easterly winds, simulations have demonstrated that the
m long with a beam of 4 m) used by paint manufacturers (i.e. the length presence or absence of offshore water intrusions is the determining
by the beam width by 0.85, a standard conversion factor; Earley et al. factor defining the contaminant distribution patterns. It is correlated
(2014)) and AR is an average release rate of antifouling paint (6.9 nmol with wind characteristics (intensity and direction), so that a medium SE
m 2 s 1). The latter is based on estimates for cumulative passive copper wind induces an entrance of offshore waters into the LB whereas a strong
leaching (i.e. without considering cleaning activities and boat usage) NE wind tends to prevent an intrusion of waters into the LB. During
during a typical paint life cycle of three years for 2 types of paint (4164 easterly wind conditions with an intrusion into the LB (corresponding to
μg cm 2 yr 1 or 1388 μg cm 2 yr 1 which corresponds to 6.9 nmol m 2 the intrusion pattern and a wind of medium intensity), the contaminant
plume is exiting the Toulon Bay at depth along a narrow route around

Fig. 11. Sea surface distribution of copper (nM) measured on 20 March 2017 (av. wind speed and direction: 2.4 m 1 SE). Positions of the three contaminant sources
in the TBAY100 model: Toulon civil port (source 1), la Seyne-sur-Mer civil port (source 2), and Saint-Mandrier civil port (source 3).

10
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

Fig. 12. On the left (A1, B1, C1, D1), test cases of contaminant release for the 3 sources source 1, source 2, source 3. Integration of tracers on the whole column
water, logarithmic scale (nmol m 2). On the right (A2, B2, C2, D2), the depth where the tracer concentration is the maximum (m). Different circulations with
constant wind conditions are simulated: ACP with medium mistral (A), BLP with strong mistral (B), IP (offshore water intrusion into the LB) with medium easterly
wind (C) and NIP (without intrusion) with strong easterly wind (D).

the Saint-Mandrier peninsula (Fig. 12-C1), following the bathymetry. originating from Toulon Bay, may reach the Marine National Park of
Passed this cape, pollutants resurface on the south coast of the peninsula Port-Cros, under mainly strong Mistral winds driving a bi-layer circu­
while remaining at mid-water depths further offshore. Once at this lation pattern which is estimated to occur ca. 20% of the time, given that
point, pollutants are likely taken up in the NC and quickly exported medium to strong Mistral correspond to 24.9% of the total time.
westwards along the coast. Without offshore water intrusions in the LB, Accordingly, the impact of Toulon Bay’s contaminated waters on the
with a strong easterly wind, contaminants are distributed over the entire National Park appears possible, which warrants further investigation
LB and are evacuated following the passageway along the Saint- using a model with a domain that extends to the National Park.
Mandrier peninsula (Fig. 12-D1). While the dissolved contaminants
remain on the surface, in the SB they sink to the bottom once passed the
seawall. Passing the peninsula, the contaminant plume mainly follows 4.2. Seasonal variability of contaminant dispersion
the bathymetry.
In conclusion, the model shows that, outside of the Toulon Bay, A 14-month simulation was conducted to assess the seasonal vari­
especially the south coast of the Saint-Mandrier peninsula is directly ability of dissolved contaminant distributions in the Toulon Bay.
subjected to frequent plumes of dissolved contaminants (for 2 out of 4 Therefore the Cu distribution map in Fig. 11 (based on 42 observations)
scenarios, which corresponds to medium and strong easterly wind pe­ was used as initial condition and copper was released at a continuous
riods or 9% of the total time). As the contaminants show a tendency of rate (Table 2) from the three point sources. The main interest of this
staying in surface layers, extra vigilance may be required for the quality simulation is to evaluate the potential impact of contaminant release on
of surface waters in this area. Furthermore, dissolved contaminants, surface waters (top 5 m of the water column) in recreational zones such
as near beaches and in areas exploited for aquaculture, where water

11
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

Table 2 nM; Tovar-Sanchez et al. (2014)) after one year. Although seasonal
Estimates of copper release rates at the three identified point sources in Toulon variability was the focus of the 14-month simulation, a validation of the
Bay. Nberths is the number of berths per port, and Cu is the quantity of copper tracer model was performed, comparing the Cu concentration map with
released per time step (nmol/5s). the model output for the same day, 20 March 2017. This comparison
Nberths Cu shows that the surface Cu concentrations are underestimated compared
(nmol/5s) to the measured values (Fig. S1 and S2 in Supplementary Material),
while the concentration gradient across the Toulon Bay is reproduced.
Source 1 (Toulon civil port) 982 1.5x106
The estimates are especially low in the small bay which is most likely
Source 2 (la Seyne-sur-Mer) 300 4.5x105
due to the minimum Cu leaching rates used in the model (section 4). The
Source 3 (Saint-Mandrier) 800 1x106 latter do not take into account boat cleaning activities and boat traffic,
two activities that are particularly concentrated in the small bay (but
difficult to quantify) and are known to have an impact on Cu leaching
quality may impact public health. Accordingly, Fig. 13 shows the tem­
from antifouling paints. Accordingly more efforts are needed in quan­
poral variation in dissolved copper at the highly frequented beaches of
tifying Cu leaching rates to improve copper distribution modelling
Mourillon (Large Bay), the aquaculture site Lazaret (Small Bay) and, for
studies in coastal areas, where antifouling paints are likely a major
comparison, the northern SB close to sources 1 and 2 (Fig. 1B). The
source of Cu.
model results show that at the three locations, the surface waters are
more concentrated in dissolved copper (up to 5 times the background
5. Conclusions with discussion
signal) during spring and early summer. This may be due to the fact that
this period is generally characterized by winds with a lower intensity
The aim of this study was to investigate circulation patterns and
(Fig. 13), which implies weaker surface currents and longer residence
scenarios of contaminant dispersion depending on weather conditions,
time (Dufresne et al., 2014). As dissolved contaminant dispersion is
thanks to the very high resolution configuration TBAY100 (configura­
wind-driven in the SB, the dispersion of dissolved copper is reduced and
tion set up with the MITgcm ocean model). As previous studies focused
an accumulation can be observed in the entire SB. On the other hand,
on the exchange between SB and LB, this study focused on the circula­
medium Mistral wind (6–10 m s 1) with an anticyclonic pattern
tion occurring in the LB and the exchanges with offshore waters, to
(Fig. 9b), is probably responsible for the observed elevated copper
assess the potential impact of the heavily contaminated waters of Toulon
concentrations in the waters of Mourillon (Fig. 12-A2). Accordingly, the
Bay on the surrounding environment, including the Marine Port-Cros
observed pattern of a degraded water quality in spring-summer in the
National Park.
Toulon Bay, merits to be tested in the future with observations to vali­
Model simulations have highlighted that the circulation in Toulon
date the coupled hydrodynamic and passive tracer model as a useful tool
Bay is mainly wind-driven. However, offshore water intrusions do occur
for pollution management. The low dissolved copper concentrations in
and can induce strong currents inside the bay. These intrusions may be
winter, are most likely due to the stronger wind intensities (>15 m s 1)
NC intrusions as in the Hyeres Bay upstream and an assessment of this
and the frequent alternation between Mistral and easterly winds, pro­
hypothesis would be part of a future work. The tides contribute only for
moting water exchanges between SB and LB as demonstrated by
3–4 cm on average of the free surface elevation, with a maximum of 20
Dufresne et al. (2014).
cm in our zone, as confirmed by local observations (Rey et al., 2019).
The 1-year simulation is not sufficient for the contaminants that were
Therefore, we can say that the tide has a minor impact compared to
continuously released at the point sources to impact the far east side of
other forcings. Overall, four circulation patterns could be identified.
the bay. Therefore, the temporal variability of contaminant dispersion
Medium to strong Mistral events do most frequently (81%) generate a
will not be discussed for the beaches located at the far east side of the
Bi-layer pattern and less often (13%) an Anticyclonic pattern. In both
TBAY model domain. Nonetheless, copper concentrations in these areas
patterns, the plume of dissolved contaminants spreads at the surface
remain close to background values for the Mediterranean Sea (i.e. 1.6
over the entire LB, but in summer period, the contaminant plume flows

Fig. 13. Surface distribution of Copper (nmol l 1) during a 14 months simulation with the model TBAY100, in three studied sites, corresponding to the boxes Fig. 1.
Wind intensity (m s 1) for medium and strong wind events: black dots for Mistral events, grey dots for easterly wind events.

12
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

towards the Giens peninsula. The anticyclonic circulation appears References

mostly during winter periods suggesting that the absence of stratifica­
tion can play a role for its occurrence. The impact of stratification on this Alberola, C., Millot, C., 2003. Circulation in the French mediterranean coastal zone near
Marseilles: the influence of wind and the Northern Current. Cont. Shelf Res. 23,
schematic circulation will be addressed in future work. When the wind 587–610 (lU).
episode is long enough and considering the frequent occurrence of the Alberola, C., Millot, C., Font, J., Jan. 1995a. On the seasonal and mesoscale variabilities
Bi-layer pattern, there could be exchanges of contaminants with the of the Northern Current during the PRIMO-0 experiment in the western
Mediterranean-sea. Oceanol. Acta 18 (2), 163–192.
Hyeres Bay and the Marine Port-Cros National Park. Medium to strong Alberola, C., Millot, C., Font, J., Jan. 1995b. On the seasonal and mesoscale variabilities
easterly winds induce offshore water intrusions into the LB (78%) of the Northern Current during the PRIMO-0 experiment in the western
associated with a cyclonic circulation (Intrusion Pattern). Under these Mediterranean-sea. Oceanol. Acta 18 (2), 163–192.
Auclair, F., Marsaleix, P., Estournel, C., 2001. The penetration of the northern current
conditions, contaminants are spreading into a narrow vein above the over the gulf of lions (mediterranean) as a downscaling problem. Oceanol. Acta 24
bottom, flowing around Saint-Mandrier peninsula before probably (6), 529–544.
meeting the NC offshore. At the same time, it can also spread inside the Barrier, N., Petrenko, A.A., Ourmi� eres, Y., Mar. 2016. Strong intrusions of the Northern
Mediterranean Current on the eastern Gulf of Lion: insights from in-situ observations
LB before being carried offshore Saint-Mandrier peninsula if there is no
and high resolution numerical modelling. Ocean Dyn. 66 (3), 313–327.
offshore waters intrusion into the LB (Non Intrusion Pattern). Bethoux, J.P., Prieur, L., Nyffeler, F., Jan. 1982. The water circulation in The north-
In addition, medium to strong Mistral events typically create up­ western Mediterranean sea, its relations with wind and atmospheric pressure. In:
welling events in the northwestern Mediterranean. These events also Nihoul, J.C.J. (Ed.), Elsevier Oceanography Series. Vol. 34 of Hydrodynamics of
Semi-enclosed Seas. Elsevier, pp. 129–142.
occur downstream in the Gulf of Lion (Millot, 1979; Barrier et al., 2016) Biggs, T.W., D’Anna, H., Mar. 2012. Rapid increase in copper concentrations in a new
and the Bay of Marseille (Pairaud et al., 2011). They are known to marina, San Diego Bay. Mar. Pollut. Bull. 64 (3), 627–635.
strongly impact the biogeochemistry of surrounding waters (Ross et al., Black, K., Hatton, D., Rosenberg, M., 1993. Locally and externally-driven dynamics of a
large semi-enclosed bay in southern Australia. J. Coast. Res. 9 (2), 509–538.
2016). To our knowledge, there is no evidence in the scientific literature Buck, K.N., Ross, J.R.M., Russell Flegal, A., Bruland, K.W., Sep. 2007. A review of total
for these upwellings in the area near Toulon, but recent observational dissolved copper and its chemical speciation in San Francisco Bay, California.
data is confirming such phenomena (Rey et al., 2019). Some upwelling Environ. Res. 105 (1), 5–19.
Coclet, C., Garnier, C., Delpy, F., Jamet, D., Durrieu, G., Le Poupon, C., Mayer, M.,
episodes could be identified in our simulations, but a specific study on Misson, B., Apr. 2018. Trace metal contamination as a toxic and structuring factor
these phenomena will have to be carried out, focusing on vertical impacting ultraphytoplankton communities in a multicontaminated Mediterranean
movements, at very short time and space scale. coastal area. Prog. Oceanogr. 163, 196–213.
Conan, P., Millot, C., Jan. 1995. Variability of the northern current off Marseilles,
As to optimise predictions of contaminant distribution in this western mediterranean-sea, from February to June 1992. Oceanol. Acta 18 (2),
exceptionally polluted area, an increment and adjustment of locations of 193–205.
moored ADCP locations is envisaged in the near future, as well as an Csanady, G.T., Oct. 1973. Wind-induced barotropic motions in long lakes. J. Phys.
Oceanogr. 3 (4), 429–438.
assessment of non hydrostatic effects on the circulation in Toulon Bay.
Dang, D.H., Sch€ afer, J., Brach-Papa, C., Lenoble, V., Durrieu, G., Dutruch, L.,
Moreover, other suggestions for future work are: evaluating the Chiffoleau, J.-F., Gonzalez, J.-L., Blanc, G., Mullot, J.-U., Mounier, S., Garnier, C.,
dispersion of contaminants toward the Marine National Park Port-Cros Oct. 2015. Evidencing the impact of coastal contaminated sediments on mussels
using a model domain incorporating the park and implementing a through Pb stable isotopes composition. Environ. Sci. Technol. 49 (19),
11438–11448.
sedimentary package into the model as to model also dispersion patterns Declerck, A., Ourmi�eres, Y., Molcard, A., Nov. 2016. Assessment of the coastal dynamics
for contaminants with a high particle affinity such as lead. in a nested zoom and feedback on the boundary current: the North-Western
Mediterranean Sea case. Ocean Dyn. 66 (11), 1529–1542.
Duffa, C., Dufois, F., Coudray, S., Nov. 2011. An operational model to simulate post-
Declaration of competing interest accidental radionuclide transfers in Toulon marine area: preliminary development.
Ocean Dyn. 61 (11), 1811–1821.
The authors declare that they have no known competing financial Dufresne, C., Duffa, C., Rey, V., 2014. Wind-forced circulation model and water
exchanges through the channel in the Bay of Toulon. Ocean Dyn. 64 (2), 209–224.
interests or personal relationships that could have appeared to influence Dufresne, C., Duffa, C., Rey, V., Verney, R., Mar. 2018. Hydro-sedimentary model as a
the work reported in this paper. post-accidental management tool: application to radionuclide marine dispersion in
the Bay of Toulon (France). Ocean Coast Manag. 153, 176–192.
Earley, P.J., Swope, B.L., Barbeau, K., Bundy, R., McDonald, J.A., Rivera-Duarte, I., Jan.
Acknowledgements 2014. Life cycle contributions of copper from vessel painting and maintenance
activities. Biofouling 30 (1), 51–68.
This work was granted access to the HPC resources of Aix-Marseille Fiandrino, A., Ouisse, V., Dumas, F., Lagarde, F., Pete, R., Malet, N., Le Noc, S., de
Wit, R., Jun. 2017. Spatial patterns in coastal lagoons related to the hydrodynamics
Universit�e financed by the project Equip@Meso (ANR-10-EQPX-29-01)
of seawater intrusion. Mar. Pollut. Bull. 119 (1), 132–144.
of the program “Investissements d’Avenir” supervised by the Agence Flexas, M.M., Durrieu de Madron, X., Garcia, M.A., Canals, M., Arnau, P., Jun. 2002.
Nationale de la Recherche. The TBAY100 simulations were also per­ Flow variability in the gulf of lions during the MATER HFF experiment (March–May
formed using GENCI-IDRIS (Grant A0030101707). Mazoyer C. PhD 1997). J. Mar. Syst. 33–34, 197–214.
Guihou, K., Marmain, J., Ourmi� eres, Y., Molcard, A., Zakardjian, B., Forget, P., Jun.
grants is supported by the French Ministry of Research. This work was 2013. A case study of the mesoscale dynamics in the North-Western Mediterranean
supported by the EU-funded Project IMPACT (PC IFM 2014–2020, Prot. Sea: a combined data–model approach. Ocean Dyn. 63 (7), 793–808.
ISMAR n. 0002269). The authors would like to thank C�eline Duffa at Jean, N., Dumont, E., Durrieu, G., Balliau, T., Jamet, J.-L., Personnic, S., Garnier, C., Sep.
2012. Protein expression from zooplankton communities in a metal contaminated
IRSN (Institut de Radioprotection et de Sûret�e Nucl�eaire) for kindly NW mediterranean coastal ecosystem. Mar. Environ. Res. 80 (Suppl. C), 12–26.
providing ADCP data. We would like to thank Nicolas Layglon for Lapouyade, A., Durrieu de madron, X., May 2001. Seasonal variability of the advective
providing the copper concentration results which were obtained during transport of particulate matter and organic carbon in the Gulf of Lion (NW
Mediterranean). Oceanol. Acta 24 (3), 295–312.
his Master 2 training at MIO (University of Toulon) as part of the Large, G., Mcwilliams, C., Doney, S., 1994. Oceanic vertical mixing: a review and a
METFLUX and PREVENT projects and the MERMEX program. We also model with a nonlocal boundary-layer parameterization. Rev. Geophys. 32 (4),
thank the SHOM for providing the bathymetry of the Toulon Bay. We 363–403.
Large, G., Yeager, S., 2004. Diurnal to Decadal Global Forcing for Ocean and Sea-Ice
dedicated this study to our colleague and friend Dr. C�edric Garnier. Models: the Data Sets and Flux Climatologies.
Leith, C.E., Mar. 1968. Diffusion approximation for two-dimensional turbulence. Phys.
Appendix A. Supplementary data Fluids 11 (3), 671–672.
Leith, C.E., Nov. 1996. Stochastic models of chaotic systems. Phys. D Nonlinear Phenom.
98 (2), 481–491.
Supplementary data to this article can be found online at https://doi. Limeburner, R., Lrish, J.D., Brown, W.S., Halliwell, G.R., Allen, J.S., Winant, C.D.,
org/10.1016/j.ecss.2019.106529. Send, E., Lentz, S.J., Rosenfeld, L.K., Beardsley, R.C., et al., 1985. CODE-2: Moored
Array and Large-Scale Data Report. Woods Hole Oceanographic Institution.
Mala�ci�c, V., Petelin, B., Vodopivec, M., 2012. Topographic control of wind-driven
circulation in the northern Adriatic. J. Geophys. Res.: Oceans 117 (C6).

13
C. Mazoyer et al. Estuarine, Coastal and Shelf Science 233 (2020) 106529

Marshall, J., Adcroft, A., Hill, C., Perelman, L., Heisey, C., Mar. 1997. A finite-volume, Sammari, C., Millot, C., Prieur, L., Jun. 1995. Aspects of the seasonal and mesoscale
incompressible Navier Stokes model for studies of the ocean on parallel computers. variabilities of the Northern Current in the western Mediterranean Sea inferred from
J. Geophys. Res.: Oceans 102 (C3), 5753–5766. the PROLIG-2 and PROS-6 experiments. Deep Sea Res. Oceanogr. Res. Pap. 42 (6),
Millot, C., Jan. 1979. Wind induced upwellings in the gulf of lions. Oceanol. Acta 2 (3), 893–917.
261–274. Schiff, K., Brown, J., Diehl, D., Greenstein, D., Mar. 2007. Extent and magnitude of
Millot, C., Broyard, R., Metais, O., Tine, J., Jan. 1981. Les oscillations propres de la rade copper contamination in marinas of the San Diego region, California, USA. Mar.
de Toulon. Oceanol. Acta 4 (3), 259–262. Pollut. Bull. 54 (3), 322–328.
Ourmi� eres, Y., Zakardjian, B., B�eranger, K., Langlais, C., 2011. Assessment of a NEMO- Schoen, J.H., Stretch, D.D., Tirok, K., Jun. 2014. Wind-driven circulation patterns in a
based downscaling experiment for the North-Western Mediterranean region: impacts shallow estuarine lake: st Lucia, South Africa. Estuarine. Coastal Shelf Sci. 146,
on the Northern Current and comparison with ADCP data and altimetry products. 49–59.
Ocean Model. 39 (3–4), 386–404. Taupier-Letage, I., Millot, C., Apr. 1986. General hydrodynamical features in the ligurian
Pairaud, I.L., Gatti, J., Bensoussan, N., Verney, R., Garreau, P., Oct. 2011. Hydrology and sea inferred from the dyome experiment. Oceanol. Acta 9 (2), 119–131.
circulation in a coastal area off Marseille: validation of a nested 3d model with Tessier, E., Garnier, C., Mullot, J.-U., Lenoble, V., Arnaud, M., Raynaud, M., Mounier, S.,
observations. J. Mar. Syst. 88 (1), 20–33. Oct. 2011. Study of the spatial and historical distribution of sediment inorganic
Petrenko, A., Sep. 2003. Variability of circulation features in the gulf of lion NW contamination in the Toulon bay (France). Mar. Pollut. Bull. 62 (10), 2075–2086.
Mediterranean sea. Importance of inertial currents. Oceanol. Acta 26 (4), 323–338. Tine, 1981. Hydrodynamisme en Rade Abri de Toulon.
Pougnet, F., Sch€ afer, J., Dutruch, L., Garnier, C., Tessier, E., Dang, D.H., Lanceleur, L., Tovar-Sanchez, A., Arrieta, J.M., Duarte, C.M., Sanudo-Wilhelmy, S.A., 2014. Spatial
Mullot, J.-U., Lenoble, V., Blanc, G., May, 2014. Sources and historical record of tin gradients in trace metal concentrations in the surface microlayer of the
and butyl-tin species in a Mediterranean bay (Toulon Bay, France). Environ. Sci. Mediterranean Sea. Front. Marine Sci. 1.
Pollut. Control Ser. 21 (10), 6640–6651. Turner, A., Feb. 2010. Marine pollution from antifouling paint particles. Mar. Pollut.
Rey, V., Dufresne, C., Fuda, J., Mallarino, D., Missamou, T., Paugam, C., Rougier, G., Bull. 60 (2), 159–171.
Taupier-Letage, I., 2019. On the Use of Long Term Observation of Water Level and Ulses, C., Grenz, C., Marsaleix, P., Schaaff, E., Estournel, C., Meul�
e, S., Pinazo, C., May
Temperature along the Shore for a Better Understanding of the Dynamics: Example 2005. Circulation in a semi-enclosed bay under influence of strong freshwater input.
of Toulon Area, France (submitted for publication). J. Mar. Syst. 56 (1), 113–132.
Ross, O.N., Fraysse, M., Pinazo, C., Pairaud, I., Mar. 2016. Impact of an intrusion by the Warnken, J., Dunn, R.J.K., Teasdale, P.R., Nov. 2004. Investigation of recreational boats
northern current on the biogeochemistry in the eastern Gulf of Lion, NW as a source of copper at anchorage sites using time-integrated diffusive gradients in
Mediterranean. Estuarine. Coastal Shelf Sci. 170, 1–9. thin film and sediment measurements. Mar. Pollut. Bull. 49 (9), 833–843.
Rubio, A., Taillandier, V., Garreau, P., Nov. 2009. Reconstruction of the Mediterranean
northern current variability and associated cross-shelf transport in the Gulf of Lions
from satellite-tracked drifters and model outputs. J. Mar. Syst. 78, S63–S78.

14