Impact of a medicane on the oceanic surface layer from a coupled, kilometre-scale simulation

. A kilometre-scale coupled ocean–atmosphere numerical simulation is used to study the impact of the 7 November 2014 medicane on the oceanic upper layer. The processes at play are elucidated through analyses of the tendency terms for temperature and salinity in the oceanic mixed layer. Whereas comparable by its maximum wind speed to a Category 1 tropical cyclone, the medicane results in a substantially weaker cooling. As in weak to moderate tropical cyclones, the dominant contribution to the surface cooling is the surface heat ﬂuxes, with secondary effects from the turbulent mixing and lateral 5 advection. Upper-layer salinity decreases due to heavy precipitation that overcompensates the salinizing effect of evaporation and turbulent mixing. The upper-layer evolution is marked by several features believed to be typical of Mediterranean cyclones. First, strong, convective rain occurring at the beginning of the event build a marked salinity barrier layer. As a consequence, the action of surface forcing is favoured and the turbulent mixing dampened, with a net increase of the surface cooling as result. Second, due to colder surface temperature and weaker stratiﬁcation, a cyclonic eddy is marked by a weaker cooling, oppositely 10 to what is usually observed in tropical cyclones. Third, the strong dynamics of the Sicily Strait enhances the role of the lateral advection in the cooling and warming processes of the mixed layer.

instability and mixing at the base of the ML, the presence of the BL mitigates the efficiency of the mixing.

60
The upper-ocean salinity changes in the wake of a TC have motivated less studies than the surface cooling, because of their indirect effect on the cyclone intensity. However, they can modulate the subsurface stratification and enhance or weaken the mixing or heat extraction. Sea surface salinity (SSS) changes result from the competing effects of precipitation and vertical shear and mixing near the base of the OML bringing saltier water up to the surface. Rain rates in TCs are maximum 35 to 50 km away from the storm centre and their magnitude ranges 3 to 12 mm h −1 on average depending on the cyclone intensity (Lonfat et al., 2004). In a study of more than 800 TCs based on satellite observations and reanalysis of the upper-ocean characteristics, Jourdain et al. (2013) showed that, despite strong precipitation and neglecting the effect of evaporation, the wake of TCs is marked by salinization. Strong mixing brings saltier water from below the OML, overcompensating the freshening effect of rain. Note that, in this study, the SST cooling is attributed entirely to the turbulent mixing. A recent, more realistic study using in situ observations brought contrast to that, showing that, at least in a 150 km radius around the cyclone centre, the freshening 70 effect of strong precipitation dominates the saltening due to mixing and evaporation (Steffen and Bourassa, 2018).
The cooling of the oceanic upper layer by extratropical storms or hybrid cyclones like medicanes has motivated fewer studies than for TCs, probably because their magnitude is much smaller and there is supposedly no or a very small feedback on the atmosphere. Indeed, coupled simulations of midlatitude storms in the North Atlantic showed that, in contrast with TCs where the cooling can reach several degrees, the mean effect is below 1°C (Ren et al., 2004;Yao et al., 2008). A recent statistical 75 study on North Pacific storms, based on satellite observations and reanalysis over 20 years, gave also a mean SST cooling one order of magnitude smaller than what is obtained for TCs (Kobashi et al., 2019). Surface heat fluxes and turbulent mixing contributed equally to the cooling. Finally, a study of a strong storm in the Gulf of Lion, in May 2005, using a coupled ocean-atmosphere-waves simulation obtained a surface cooling of 2°C over a large area (Renault et al., 2012). During this storm, the major contributor to the OML cooling was the vertical mixing enhanced by the strong surface stress (68 %), with 80 secondary contributions from the latent and sensible heat fluxes (15 and 7.5 % of the cooling). These contributions are similar to those obtained in TCs of Category 2 or above.
In the present study, we use of a coupled ocean-atmosphere kilometre-scale simulation to investigate the impact of the short but intense medicane Qendresa (7 November 2014) on the OML. Comparing the atmospheric processes of medicanes with those of TCs is a tempting but challenging undertaking. Indeed, no systematic statistical study provides, to the best of our knowledge, 85 a comprehensive assessment of the processes at play according to the cyclone intensity, especially for the Categories 1 and 2 that could be compared with medicanes. Conversely, as detailed here above, the impact of TCs on the upper ocean and the corresponding mechanisms were the subject of systematic studies scanning the different TC categories (e.g. Vincent et al., 2012a;Jullien et al., 2014). We thus take advantage of this knowledge to contrast the coupling mechanisms and oceanic upper-layer processes in a medicane with those of TCs. 90 A first study of the lifecycle and atmospheric processes of this medicane, including the assessment of the impact of the SST cooling on the atmosphere, showed that the surface cooling is at least one order of magnitude lower than in typical TCs (Bouin and Lebeaupin Brossier, 2020). The present study aims at investigate i) how the surface cooling / salinity changes obtained in this case compare with changes observed in TCs or midlatitude storms; ii) whether the OML processes are similar to those of Mediterranean oceanic conditions prior to the medicane. Section 2 summarizes the case study and presents the oceanic conditions before the event, and the numerical tools used in this work. The results, with the evolution of the ocean and an analysis of the role of the atmospheric forcings and of the mechanisms at play, are given in Section 3. Section 4 presents the role of the oceanic conditions on the cooling and salinity changes in different areas. These results are discussed and conclusions are given Section 5.

2 Case study and simulations
We present here a summary of the November 2014 medicane, the oceanic conditions that pre exist in Central Mediterranean, and the ocean-atmosphere simulating configuration used in this study.

Synoptic situation
The medicane formed north of Lampedusa in the morning of 7 November, from the conjunction of a baroclinic disturbance at low level and of upper-level instability. Strong convection developed in the morning with heavy precipitation (more than 150 mm locally) in the Sicily area. The low-level system rapidly deepened, with a sudden drop of sea-level pressure of 8 hPa 110 in 6 hours, and evolved into the quasi-circular structure of a tropical cyclone with spiral rain bands and a cloudless eye-like centre. The maximum intensity was reached around 12 UTC on 7 November north of Lampedusa, with sustained 10 m wind speeds above 34 m s −1 . Strong winds persisted during its transit towards the Sicilian south coasts, with a landfall at Malta around 17 UTC. It reached Sicily in the evening of 7 November and continued its decay during the following night on the Ionian Sea close to the Sicily coasts, until 12 UTC on 8 November. Three distinct phases of the event are described in Bouin The simulated medicane spent most of its lifetime in the Sicily Strait area, which is significantly shallower than the surrounding basins with a mean bathymetry close to 500 m and large areas shallower than 100 m, for instance in the Gulf of Gabès. As a transition between the Western and Eastern Mediterranean, the region is characterized by large-scale gradients (Drago et al., 2010): a north-south thermal gradient and a west-east salinity gradient between Atlantic Water flowing eastwards along the Tunisian coasts (AW, 0 to 100 m depth, T = 15-17°C, S = 37.2-37.8 with a salinity minimum around 50 m) and Ionian Water 130 (from 50 to 100 m depth, T = 15-16.5°C, S = 37.8-38.4). These two water masses cap the Levantine Intermediate Water (LIW, with a core depth at 300 m, T = 13.75-13.92°C, S = 38.73-38.78) that originates from the thermohaline circulation in the Eastern Basin and flows westwards. These gradients are well represented in the surface initial conditions used in this study ( Fig. 2 and 3b). North of the Sicily Strait, the SST is below 21°C, while it reaches 24°C close to the Libyan coasts.
This marked SST contrast has been shown to largely control the surface heat fluxes during the most intense part of the event, 135 along with the surface wind speed (Bouin and Lebeaupin Brossier, 2020). In the Tyrrhenian Sea and the Sicily Strait, the SSS is below 38, with a strong impact of the AW flowing eastwards along the Tunisian coasts (the AW flow can also be seen on the SST map). When comparing the model initial SSTs (at 01 UTC on 7 November) with those provided by a satellite analysis at 00 UTC on 7 November, the model SSTs are biased low of −0.63 ± 0.49°C, but the general patterns are well represented with a correlation of the surface layer) and achieve a homogeneous density ρ: At the beginning of the simulation (Fig. 3a), the stratification is pronounced in the Sicily Strait with an influence of the AW (SI = 125 ± 14 kg m −2 ) and weaker in the Ionian Sea (92 ± 11 kg m −2 ). Minimum values are visible on the Tunisian continental 160 shelf (around the Gulf of Gabès, with depth less than 30 m, see Fig. 1). Continental shelf waters are known to be significantly more sensitive to atmospheric conditions (Béjaoui et al., 2019) with wind-induced mixing that quickly homogenizes the water column, and salinity reacting instantaneously to evaporation, runoff or heavy rain (Ismail et al., 2017).

Numerical settings and tools
The ocean-atmosphere coupled numerical simulation of the event was performed using the state-of-the-art atmospheric model

Atmospheric model
The non-hydrostatic French research model Meso-NH version 5.3.0 is used in the present study with numerical and physical packages as described in Bouin and Lebeaupin Brossier (2020). The radiative transfer is computed by solving long-wave and short-wave radiative transfers separately using the European Centre for Medium-range Weather Forecasts (ECMWF) opera-170 tional radiation code (Morcrette, 1991). The sea-surface fluxes are computed within the SURFEX module (Surface Externalisée, Masson et al. (2013)) using the iterative bulk parameterization ECUME (Belamari, 2005;Belamari and Pirani, 2007) linking the surface turbulent fluxes to the meteorological gradients and the SST through the appropriate transfer coefficients.
The Meso-NH model shares its physical representation of parameters, including the surface fluxes parameterization, with the French operational model AROME used for the Météo-France weather prediction with an horizontal resolution of 1.3 km (Se-175 ity et al., 2011). In the present study, a first atmosphere-only simulation at the horizontal resolution of 4 km has been run on a larger domain of 3200 km × 2300 km (not shown here) to provide initial and boundary conditions for the simulation at 1.33 km on a smaller domain of 900 × 1280 km (Fig. 1). This run at coarser resolution started at 18 UTC on 6 November and lasted 42 h until 12 UTC on 8 November. Its initial and boundary conditions come from the ECMWF operational analyses every 6 h. The coupled simulation on the inner domain that is used here to investigate the oceanic evolution starts at 00 UTC on 7

180
November and lasts until 12 UTC on 8 November.

Oceanic model
The ocean model used is NEMO (version 3_6) with physical parameterizations as follows. The total variance dissipation scheme is used for tracer advection in order to conserve energy and enstrophy (Barnier et al., 2006). The vertical diffusion follows the standard turbulent kinetic energy formulation of NEMO (Blanke and Delecluse, 1993). In case of unstable conditions, 185 a higher diffusivity coefficient of 10 m 2 s −1 is applied (Lazar et al., 1999). The sea surface height is a prognostic variable 6 https://doi.org/10.5194/os-2020-38 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License.
solved thanks to the filtered free-surface scheme of Roullet and Madec (2000). A no-slip lateral boundary condition is applied and the bottom friction is parameterized by a quadratic function with a coefficient depending on the 2D mean tidal energy (Lyard et al., 2006;Beuvier et al., 2012). The diffusion is applied along iso-neutral surfaces for the tracers using a laplacian operator with the horizontal eddy diffusivity value ν h of 30 m 2 s −1 . For the dynamics, a bi-Laplacian operator is used with the 190 horizontal viscosity coefficient η h of −1.10 9 m 4 s −1 .
The configuration used here is sub-regional and eddy-resolving, with a 1/36°horizontal resolution over an ORCA-grid (from 2 to 2.6 km resolution) named SICIL36 (tripolar grid with variable resolution Madec and Imbard, 1996), that was extracted from the MED36 configuration domain (Arsouze et al., 2013) and shares the same physical parameterizations with its "sister" configuration WMED36 (Lebeaupin Brossier et al., 2014;Rainaud et al., 2017). It uses 50 stretched z-levels in the vertical,    The SST cooling obtained here is lower than for TCs of Category 1 where wakes 1 to 2°C colder than the surrounding ocean are currently observed (e.g. Vincent et al., 2012a). The weak surface freshening we obtain is rather consistent with the work of Steffen and Bourassa (2018)

Oceanic processes
This part describes the simulated OML processes obtained throughout the event, with reference to the processes commonly observed in TCs. As seen in Section 2.1.2, the SFA represents a strongly dynamic region with large horizontal gradients of temperature and salinity, strong currents and the presence of the ABV cyclonic eddy. The stratification index in the SFA is high at the beginning of the simulation with values of 125 ± 14 kg m −2 except in the ABV cold eddy (88 ± 14 kg m −2 , Fig. 3, Table   265 1). The MLD is 34 ± 11 m around the eddy, 16 ± 3 m in the eddy itself. At 13 UTC on 7 November, the cyclonic circulation at 15 m within the eddy has reinforced due to strong wind stress, with a marked diverging component and maximum velocities between 0.7 and 0.8 m s −1 . As a consequence of this subsurface divergence, the sea surface height has decreased of 10 cm in the eddy centre and strong upward motions develop under the MLD, around 50 m depth. The strong shear resulting from the horizontal currents close to the bottom of the OML generates violent mixing, which brings colder water from below the 270 thermocline up. The upwelling is maximum at 15 UTC, with an ascent of the thermocline of 5 m, and a lowering of the SSH of 17 cm. Following this, vertical oscillations establish with period close to 19 h (the inertial period for this area is between 20.5 and 21 h). At 00 UTC on 8 November, the amplitude of the subsurface cyclonic circulation has returned to its initial value, but with a marked converging component. At that time, the vertical velocity under the OML is negative, and the SSH is still 10 cm depressed. The OML is deeper at the eddy centre, and shallower around the outer edge of the eddy.

275
The processes obtained are similar to those observed in TCs. To assess the respective roles of the surface heat fluxes and of the turbulent mixing in the time evolution of the OML, we performed a budget analysis of the temperature and salinity in the OML, using the equation of the tracer tendency as in Vialard and Delecluse (1998). The tendency of a given tracer X (θ or S) within the OML is given by where h is the MLD, − → U is the horizontal velocity vector, and w the vertical velocity. A h and A v are the horizontal and vertical 285 eddy diffusivity coefficients (m 2 s −1 ), F X the surface flux, and w X the turbulent flux of tracer X. Finally, X = 1 h 0 −h Xdz. From that, the different terms of the temperature and salinity tendencies can be deduced: • The surface forcing term (FOR) corresponds to the net heat flux for the potential temperature θ (shortwave, longwave radiative net fluxes plus latent and sensible heat fluxes) and to the water flux for the salinity.
• The horizontal advection h-ADV that can be decomposed into zonal (ADV-X) and meridional (ADV-Y) advections.

290
• The vertical advection across the OML v-ADV.
• The diffusion term (DIF), that can decomposed into lateral and vertical diffusions, this latter term corresponding to the turbulent mixing within the OML (TM hereafter). These tendency terms are used in the following to quantify the processes controlling the evolution of the temperature and salinity within the OML.

Temperature tendency 300
The time evolution of tendency terms for θ in the SFA is given Fig. 7, with the corresponding evolution of θ . The continuous decrease of θ throughout the simulation is mainly due to the surface forcing FOR. The evolution of this FOR term mimics the one of the median/mean values of heat fluxes in the SFA (Fig. 4b), with a maximum at 09 UTC. The TM evolution is closer to the evolution of the median value of the WEF (Fig. 4a), with a maximum between 10 and 12 UTC on 7 November.
Other significant contributions to the thermal evolution of the OML are from the lateral advection terms ADV-X and ADV-305 Y, which contribute to alternatively cool and warm the OML with a time period driven by the quasi inertial oscillation. The is observed in intense TCs, the surface forcing is key here, the turbulent mixing playing only a secondary role like the lateral advection. The time evolution of the distribution of the FOR and TM terms for temperature in the SFA (Fig. 8) shows that all the quantiles of FOR are higher than those of TM. During the development phase and the beginning of the mature phase of the medicane (between 06 and 12 UTC on 7 November), TM even contributes to warm the OML for 25 % of the SFA. This is the consequence of the ascent of the MLD under the effect of a surface salinity change. This is discussed in more details Section The spatial distribution of the cooling due to FOR and TM within the SFA is also different. A snapshot at 10 UTC on 7 November when the effect of TM is maximum (Fig. 9) shows the values of the FOR and TM terms for temperature averaged on radial bins around the medicane centre. The turbulent mixing effect is stronger than the surface forcing one within 20 km 320 around the medicane centre, but it drops rapidly and becomes small after 200 km. Conversely, the effect of FOR stays above −0.2°C within 500 km around the centre.
The lateral advection contributes to alternatively cool and warm the OML (Fig. 7a) in phase with the near-inertial oscillation.
Throughout the event, both ADV-X and ADV-Y cool the OML by −0.12°C, or 16 % of the total cooling. A compensating warming effect of ADV-X corresponds to 0.06°C.

325
These results are consistent with those of the statistical study of Vincent et al. (2012a). First, the relative part of the surface forcing is higher than its of the turbulent mixing, as expected for the weakest TCs (70 % for FOR for TCs of Category 1 or below). Second, turbulent mixing and surface fluxes contribute equally within 2 radii of maximum wind, but surface fluxes dominate outside of that and concern a wider area. A case study in the Gulf of Mexico also showed that the surface heat fluxes are at the origin of a widespread, moderate cooling affecting the whole surface of the gulf, while vertical mixing results in 330 stronger, more localized cooling within 200 km around the cyclone centre (Morey et al., 2006). Third, the lateral advection contribute significantly to the cooling (and weakly to the following warming). This latter effect is likely due to the strong dynamics and horizontal gradients of the Sicily Strait.
In the following, we document the time evolution of the salinity in the OML.

335
The time evolution of the OML salinity tendency terms in the SFA is given Fig. 10, with the corresponding evolution of the OML mean salinity and of the precipitation. The overall effect is a significant freshening until 21 UTC on 7 November, as long as the precipitation rate is above 2 mm h −1 . This freshening is thus due to the FOR term, and is partly compensated by TM and, as secondary processes, by the horizontal advection and entrainment. This compensating effect drops rapidly after 12 UTC on 7 November with the decrease of the WEF (Fig. 4) whereas the surface forcing generates freshening until 21 UTC. In 340 11 https://doi.org/10.5194/os-2020-38 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License. addition, the large amount of precipitation in the early hours of the 7 November deepen the BL by shoaling the mixed layer.
During the event, the depth of the thermocline (defined by the depth where θ is equal to the 10 m-temperature minus 0.3°C) averaged over the SFA oscillates of a few meters (Fig. 10b) due to the near inertial waves. The freshening due to heavy rain in the morning of 7 November makes the MLD defined by a density criterion shallower and governed by salinity rather than temperature. As the OML moves 2 m upwards between 04 and 07 UTC on 7 November, its mean temperature slightly increases 345 under the apparent effect of turbulent mixing -this actually corresponds to the points with warming tendency due to TM in Fig.   8. The BL thickness increasing from 5.8 m at 04 UTC to 8.1 m at 12 UTC on 7 November insulates the OML from the colder water below the thermocline. It enhances the efficiency of the surface heat extraction, and reduces the effect of the turbulent mixing.

Role of preconditioning and oceanic evolution
To quantify the role of oceanic preconditioning and precipitation in the oceanic response to the medicane, we define different zones within the SFA. The EDDY zone is defined as the area of the ABV cyclonic eddy, using a SST of 20.5°C as a threshold (1005 grid points). We also define a heavy rain zone (HR hereafter) as the area, in the Sicily Strait, where integrated water flux at 00 UTC on 8 November reaches 90 mm of water. This HR zone includes 2383 grid points and its mean integrated water flux 355 at 12 UTC on 8 November is 136 mm. Note that it is not entirely within the SFA (Fig. 6). Finally, a reference zone includes the grid points of the SFA that are not part of either the EDDY or the HR zone (REF hereafter, 4343 grid points). The objective here is to evaluate whether the characteristics of Central Mediterranean (bathymetry, dynamics) or heavy rain control the oceanic response to the event. ±0.23°C at 00 UTC on 8 November) as is the integrated surface heat flux (Fig. 11b). The weaker cooling in EDDY originates from lower heat fluxes throughout the event and especially between 09 and 16 UTC on 7 November (see also Table 3

370
The tendency terms for θ ( Fig. 12 and Table 3) confirm that different surface forcing leads to different cooling in EDDY and REF. In EDDY, the time evolution of FOR mimics the heat fluxes (red curves in Fig. 12a), with almost no forcing between November. Yet, the WEF is much weaker in EDDY than in REF, the integrated value at 12 UTC on 8 November represents one third of it. The stronger mixing is EDDY is due to a weaker stratification and a much shallower −0.3°C thermocline (around  Within EDDY, the salinity does not change significantly throughout the simulation (Fig. 13a, Table 2). Except between 09 and 11 UTC on 7 November, the rainfall amount in REF is weak (44 % of the integrated values of the SFA). However, the integrated water flux (dominated by precipitation) is even much weaker in EDDY (9.6 versus 27.7 mm at the end of the simulation; Fig.   13c, Table 3). The tendency terms confirm that the precipitation drives the salinity evolution similarly in both zones ( Fig. 14a and c), with a largely compensating turbulent mixing (bringing saltier water from under the OML). In both zones, the OML 385 deepens of a few meters at the beginning of the event due to the cooling effect (Fig. 13b).
Within the ABV eddy, colder SSTs and shallower MLD and thermocline result in less surface cooling. The turbulent mixing intensifies due to the weaker stratification, but the surface forcing is reduced, resulting in less overall cooling. Lateral advection contributes more to the cooling and warming of the OML than outside of the eddy. Low precipitation and moderate mixing bringing more saline water upwards compensate each other and do not change notably the OML salinity.

Role of heavy precipitation
We compare here the time evolution of the temperature and salinity in the OML, and the corresponding tendency terms, in the HR and REF zones. At the beginning of the simulation, the OML is shallower in HR than in REF (29 ± 10 versus 36 ± 11 m) with SI at 100 m around 125 kg m −2 in both cases. The SST is also similar (∼ 21.5 ± 0.5°C) and the surface water is slightly fresher in HR (37.61 ± 0.12 versus 37.73 ± 0.17). The time evolution of θ is rather similar, with respective coolings at 00 395 UTC on 8 November of −0.57 and −0.51°C (Fig. 11a, Table 2). Yet, the integrated heat loss in HR is 31 % higher in REF than in HR (Fig. 11b, Table 3). Thus, the surface forcing is much more efficient in cooling the OML in HR than in REF ( Fig.   12b and c). Looking at the tendency terms shows that FOR is stronger in HR than in REF, but that TM is stronger in REF than in HR (Table 3). These discrepancies in the cooling effects in HR and REF can be related to the impact of the precipitation on the upper-ocean salinity and the resulting MLD variations. In HR, the heavy rain in the morning of the 7 November produces 400 a first drop of the SSS of −0.1 (Fig. 13a) and shoals the OML of a few meters (Fig. 13b). As a result, the shallower OML is more sensitive to the atmospheric forcing and is insulated from the colder waters below the thermocline by a 13 m thick BL (Fig. 14b). As was shown for TCs, the effect of the surface heat fluxes on the SST is enhanced while the turbulent mixing is dampened (Yan et al., 2017). Indeed, TM starts to decrease at 10 UTC in HR while the WEF increases until 12 UTC on 7 November. The net result here is a stronger cooling, since the FOR term is dominant because of the weak intensity of the event.

405
After this first freshening and cooling of the OML due the heavy rain in the morning of the 7 November, lighter rain occurs 13 https://doi.org/10.5194/os-2020-38 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License. in the afternoon. Yet, because the thinned OML is very sensitive to the atmospheric forcing at that time, this leads to strong additional freshening (−0.5 in total) and shoaling of the OML (Fig. 13). The water flux controlling the OML depth explains the stronger cooling in HR, with a SST cooling of 0.8°C at 21 UTC on 7 November (Fig. 11).
This results are consistent with the role of BLs (usually pre existent) in modulating the surface cooling due to TCs. In weak 410 TCs where the cooling effect is mainly due to the surface heat fluxes, the presence of BLs makes the OML shallower and enhances the surface forcing. Its isolating effect from colder water below the thermocline does not impact much the cooling since turbulent mixing is a secondary mechanism. The novel result here is that the BL formation and deepening arise from heavy precipitation occurring at the beginning of the event. A realistic study of the net effect of precipitation versus evaporation and mixing on the surface salinity in TC wakes showed that the BL thickness increases in every cyclonic basin (Steffen 415 and Bourassa, 2018). It is not clear however whether this overall freshening happens early enough during the TC development to substantially impact the cooling. Here, because of the deep convective rainfall in the first phase of the event (11 mm h −1 on average during the first 10 h), a BL rapidly forms and strengthen the cooling effect of surface fluxes.

420
Comparing a medicane with low-intensity TCs from the viewpoint of the upper-layer oceanic processes proves to be relevant and reveals many similarities and a few differences. According to its maximum sustained wind, the medicane studied here is comparable to TCs of Category 1 on the Saffir-Simpson scale. Its duration is nevertheless much shorter than a typical TC with wind energy flux above 0.2 kg s −3 for a few hours only, between 07 and 13 UTC on 7 November, and upward heat fluxes above 400 W m −2 for 30 hours between the beginning of the event and 07 UTC on 8 November. The mean surface cooling on the 425 SFA is therefore less than −0.6°C, significantly lower than the typical cooling of weak TCs. The dominant cooling mechanism is the surface forcing (at the origin of a −0.45°C cooling) while the turbulent mixing accounts for less than −0.1°C. These two mechanisms do not act simultaneously: the surface forcing starts earlier, peaks at 09 UTC on 7 November at stays above −400 W m −2 for more than 24 h in total; the turbulent mixing increases until 12 UTC on 7 November and drops rapidly in the afternoon of the 7 November. This may explain that surface forcing has a much larger effect than turbulent mixing on the 430 OML temperature. Yet, the cooling effect of the surface fluxes is at the upper threshold of the typical range given by Vincent et al. (2012a), while the cooling effect of the turbulent mixing is always weak, even at the peak of intensity of the event. This discrepancy can be explained by the effect of heavy rainfall in the first hours of the event, which create or deepen an existing BL. Indeed, comparing the cooling of two areas forced by similar wind stress and heat fluxes, but very different water fluxes shows that the presence of the BL results in a 37 % increase of the effect of the surface forcing and 64 % decrease of the 435 effect of the turbulent mixing. A diagram based on the simulated mean profiles and MLD values summarizes the upper-layer evolution throughout the event, and the impact of heavy rain (Fig. 15). Without strong freshwater input, temperature governs the MLD evolution and the intense surface heat fluxes throughout the event continuously cool the upper-layer (Fig. 15a). This deepens the OML of a few meters (Fig. 15c) and moderate rain, partly balanced by evaporation, slightly freshen the OML (Fig. present, mainly acts on the surface temperature during the second half of the event (Fig. 15a, see also Fig. 11) even though it occurs since the beginning of the event. This delayed effect is due to the positive feedback of a salinity-induced thinning of the OML, which settles after a few hours only. Heavy convective precipitation are usual during the development phase of medicanes, before the wind reaches its maximum speed, and the preconditioning effect of the oceanic upper layer we obtain here is probably rather typical. Mediterranean cyclones in general come with heavy rain, which are susceptible to alter the 445 response of the oceanic upper layer as is observed here. On the OML salinity, both the effect of the surface forcing and of the turbulent mixing are enhanced by the BL.
Lateral advection plays a secondary but significant role in cooling or warming the OML. In particular, strong upper-layer horizontal currents develop close to the cyclone centre in the ABV eddy. Colder water is shifted southwards, reinforcing the cooling, and warmer water is brought towards the eddy centre by converging motions from the surrounding area with deeper 450 OML. In addition, due to shallower MLD and weaker 100 m stratification in the eddy, the effect of the surface heat fluxes is lower while the effect of the turbulent mixing is strengthened. As a result, the cooling in the ABV eddy is dampened with respect to the surroundings. The peculiar dynamics of the Sicily Strait, with its strong currents and the regular presence of eddies is able to modulate the impact of cyclones on the ocean.
The present study is in strong contrast with the case study of a strong storm with comparable sustained wind speed, in North-455 western Mediterranean (Renault et al., 2012). The cooling obtained by these authors was between −1.5 and −2°C over the large area of the Gulf of Lion. A precise comparison of the two events and of their impact on the oceanic upper layer is beyond the scope of the present study but several factors can partly explain the stronger cooling they obtained. First, the duration of the storm was longer than Qendresa, with strong wind stress and heat fluxes from the 4 May until the 6 May. The track of the storm was also favourable to a stronger impact on the ocean, since it looped for three days over the Gulf of Lion. Second, the 460 MLD in May in the Gulf of Lion is usually shallower than in November in the Sicily Strait, with values in the range 10-30 m (d'Ortenzio et al., 2005). More precisely, observations of the Lion buoy show that prior to the storm, the temperature at 10 m was −0.3°C below the SST (taken at 2 m, see Houpert et al., 2016). No measurements were available at that time at depths between 10 and 200 m, but observations at the same date in 2012 show a −0.6°C difference between the temperature at 15 m and the SST, and a −1°C difference at 35 m. The upper-layer was then likely strongly stratified before the storm, and this 465 favours the turbulent mixing in bringing cold water at the surface. As far as we can judge, atmospheric and oceanic conditions acted together to enhance the impact of the turbulent mixing and this resulted in a stronger cooling.
This study is based on a single case, and our results should be confirmed by simulating more events. A broader study involving a statistically significant number of Mediterranean cyclones is nevertheless still challenging. Intense cyclones like the present case occur 1 to 2 times per year (e.g. Cavicchia et al., 2014) and our results show that high-resolution oceanic models are 470 necessary to represent the fine-scale features modulating the oceanic response. The atmospheric forcing should also be well resolved to realistically reproduce the intensity and variations of the surface input. The simple comparison with the oceanic impacts obtained by Renault et al. (2012) shows that Mediterranean storms are diverse and, depending on the area and season concerned, can lead to different oceanic responses due to the different processes. Previous studies at the climatological scale Beuvier, J., Béranger, K., Lebeaupin Brossier, C., Somot, S., Sevault, F., Drillet, Y., Bourdallé-Badie, R., Ferry, N., and Lyard, F Oceanography, 23, 1363-1388, 1993. Ocean Modelling, 84, 84-103, 2014. Leipper, D. F.: Observed ocean conditions and Hurricane Hilda, 1964, Journal of the Atmospheric Sciences, 24, 182-186, 1967 Hurricane heat potential of the Gulf of Mexico, Journal of Physical Oceanography, 2, 218-224, 1972. Lellouche, J.-M., Le Galloudec, O., Drévillon, M., Régnier, C., Greiner, E., Garric, G., Ferry, N., Desportes, C., Testut, C.-E., Bricaud, C.,   35 https://doi.org/10.5194/os-2020-38 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License.