The impact of wave physics in the CMEMS-IBI ocean system Part A : Wave forcing validation

The Iberian Biscay Ireland (IBI) wave system has the challenge to improve wave forecast and the coupling with ocean circulation model dedicated to western european coast. The momentum and heat fluxes at the sea surface are strongly controlled by the waves and there is a need of using accurate sea state from wave model. This work describes the more recent version of the IBI wave system and highlight the performance of system in comparison with satellite altimeters and buoys wave data. The validation process has been performed for 1-year run of the wave model MFWAM with boundary conditions provided by the global wave system. The results show on the one hand a slightly improvement on significant wave height and peak period, and on the other hand a better surface stress for high wind conditions. This latter is a consequence of using a tail wave spectrum shaped as the Philipps wave spectrum for high frequency waves.


Introduction
Waves constitute the interface between ocean and atmosphere and have an important role in term of exchanges through this interface (Cavaleri et al., 2012).Their representation is necessary to compute with accuracy the different air-sea fluxes of heat and momentum (Janssen et al., 2004).
However, waves are generally parameterized from 10-m local wind.While there is a correlation between wind and waves, their relationship is not exclusive.Indeed, waves are also present without wind and for a given local wind speed, the local wave field is variable (Hanley et al., 2010).Moreover, it is generally accepted that wind directly generates surface currents because about 90% of the wind momentum input to waves is immediately passed to the ocean (Cavaleri et al., 2012).In fact, waves absorb energy and momentum from the wind during their formation and growth, and return it when they break (Breivik et al., 2015).This explains why it is necessary to introduce an accurate sea state description, from wave model (or database as for example Rascle et al., 2007), which control exchanges between ocean and atmosphere.
Waves affect the ocean surface layer through different processes (Breivik et al., 2015) : -Waves induce current in surface via the Stokes drift, rapidly attenuated with depth.The Stokes drift velocity associated with the wave fields adds a term on the Coriolis effect in the momentum equation.This process is called Stokes-Coriolis forcing.
-A part of the atmospheric wind stress is used by waves to grow and is not provided to the ocean.This energy quantity must be subtracted from the oceanic wind stress which drives the ocean model.
-During wave breaking, turbulent kinetic energy is produced and induces to the ocean surface layer an enhanced turbulent mixing.A more accurate description of these processes will be given in the part B of this study.
Recent studies investigated the impact of the wave effect on the representation of the ocean surface layer at different scales of time and space.One of the major impact is the improvement of the Mixed-Layer-Depth (MLD) using wave-induced MLD parametrization (Fan et al., 2014) which lead to an important impact on the atmospheric surface temperature, pressure and precipitation (Babanin et al., 2009(Babanin et al., , 2012)).In a climate scale, this can affect global sea-surface pressure patterns and atmospheric circulation.Breivik et al., (2015) showed that the use of wave forcing in oceanic surface induces a reduction of the global annual SST bias amplitude on a period from 1979 to 2010.Ardhuin et al. (2010) which is called ST4.The MFWAM model takes also into account a swell damping term related to to the air friction at the sea surface.Recently the model has been upgraded with adjustment of the dissipation source terms and also improvement on drag limitation by using a tail shape from the Philipps wave spectrum (Janssen et al. 2014).Table 1 gives the tuned coefficients of ST4 physics for the old and new version of the model referred to as V3 and V4, respectively.In this study the model MFWAM is set for a IBI domain (25°N to 64.6°N in latitude) with a grid size of 0.10°.The model uses a bathymetry from ETOPO2 and is driven by 6-hourly analyzed winds from the IFS-ECMWF atmospheric model.The wave spectrum is discretized in 24 directions and 30 frequencies starting from 0.035 to 0.57 Hz with increasing step of 1.1.Th boundary conditions are provided from the global model MFWAM run with a time step of 3 hours.The data assimilation is not activated for this study.

Satellite data
Two different SST satellite products, OSTIA and L3S have been used in this study.The OSTIA daily product of SST is a level 4 multi-sensor product at a resolution of 0.02° built by using optimal interpolation from several satellite missions such as AVHRR_METOP_B, SEVIRI*VIIRS_NPP, AVHRRL_19, AVHRRL_18, MODIS_A, MODIS_T, AMSR2.The hierarchy can be changed in time depending on the health of each sensor.The L3S product consists in a fusion of daily SST observations from multiple satellite sensors, over a 0.1° resolution grid.It

Validation of the wave output fields
The validation of the wave model MFWAM is based on a statistical analysis of significant wave heights from the model compared to altimeter wave data (JASON-2 and Saral/Altika), as mentioned in previous section.Metrics used in this study are normalized scatter index and bias.Figure 3 shows the scatter plots of SWH from altimeters and models for the year 2014.This reveals a slightly better slope for V4, but overall good performance of the MFWAM-IBI wave model for V3 and V4 versions.The scatter index of SWH is roughly of 12.5% and the bias is negative and roughly 15 cm.The model MFWAM underestimates SWH which is mostly induced by the underestimation of the surface wind speed from the atmospheric model in the IBI domain.Table 3 highlights the very good and close performance of both versions during 2014.Figure 4 and 5 show maps of average scatter index and bias on the IBI domain, respectively.Most of scatter index are ranged between 8 to 15% depending on the ocean area.However, we can mention that wave model V3 and V4 have a better skill on the Atlantic ocean than on North and Mediterranean seas.The maps of bias also show a general underestimation of SWH except on Gulf of Biscay where the model MFWAM induced an overestimation of SWH for both V3 and V4.A detail analysis is given by the monthly variation of the model performance, as illustrated in figure 6.For V3 and V4 versions, the scatter index is ranging between 11 and 14% during all the year, which indicates that the good performance of the MFWAM-IBI wave model is for the entire year.Monthly scatter index are slightly better for MFWAM-V4, which confirms the better performance of this version on the IBI domain.
Performances of the model MFWAM are also investigated depending on regional domains.
Three domains have been selected depending on latitudes as indicated in figure 8 ranging between 12% and 13% and between -10 cm and -20 cm, respectively.One can indicate that the scatter index of SWH is better for zones 1 and 3 while the bias is better for zone 2 during 2014.
Table 4 shows also a slightly better performance for MFWAM-V4, in particular for the scatter index, for the three zones.This confirms what was previously mentioned concerning differences between the two versions.Figure 8 shows monthly scatter index and bias during 2014 for the three ocean zones.The scatter index of SWH is ranging between 10% and 15% and the best performance is obtained during summer for months june and july.While during winter with intense storms in North Atlantic ocean the scatter index of SWH is larger for zones 2 and 3.
The significant wave height from MFWAM runs has also been evaluated with buoys measurements.Histograms of Figure 9 show the monthly scatter index during 2014.At BelmA buoy, the scatter index is lower than 15% during the entire year except on June and November.
Moreover, the performance of the two runs is globally the same during all the year despite a slightly increase of MFWAM-V4.However, at BI buoy, the performance of MFWAM-V4 is generally better MFWAM-V3 during the entire year.At this location, the scatter index oscillate between 10% and 20% because of storm occurrence in this area.
The validation of the MFWAM V3 and V4 has indicated good skills during the year 2014.This opens the use of accurate description of the sea state and consequently good wave forcing to drive the ocean model NEMO.Moreover, in term of differences between MFWAM-V3 and MFWAM-V4, the validation showed an general increase of the quality thanks to the physical settings of MFWAM-V4.

5-Conclusions
The wave model MFWAM V4 version has been upgraded with physical adjustments that induced a more consistent surface stress dependency with 10-meter wind speed.The use of tail shape of Philipps spectrum aslo induces a better high frequency waves and consequently a better estimate of Stokes drift.The validation with altimeters wave data has been shown a slight improvement of bias and scatter index over IBI domain.The model MFWAM induces a negative bias that indicate an underestimation of significant wave height mostly because of uncertainties related to strong winds during winter and fall season from the ECMWF atmospheric system.The bias of significant wave height in summer season is small roughly less than 5 cm.The regional statistical analysis has revealed the best improvement of the model MFWAM V4 on the zone 2 which includes portuguese, spanish and french coasts.The validation with buoys has indicated a significant improvement of SWH at Belle-ile buoys located in coastal area of brittany french coast.
They used a coarse 1° horizontal grid resolution of NEMO ocean model and wave forcing from ECWAM wave model.An important reduction of the amplitude of the diurnal cycle of SST and 2. Description of the wave Model MFWAM The wave model of Météo-France MFWAM provides the mean wave parameters for the Copernicus Marine Environment and Monitoring Service (CMEMS) for Iberian-Biscay-Ireland seas domain.The model is based on the ECWAM-IFS38R2 computing code (ECMWF 2013) with two major changes related to the dissipation source terms.The model MFWAM uses the physics developed in Figures 2a and 2bshow the variation of the surface drag coefficient with 10-meter wind speed.It is clearly observed that the MFWAM V4 reduces significantly the drag coefficient for high wind 5 . Zone 1 concerns latitudes between 25°N-35°N (called Canarias domain), while Zone 2 considers latitudes 35°N-49°N (called Spain-France domain).Zone 3 accounts for northern latitudes ranging between 49°N-64.5°N(called GB-Ireland domain).For all zones during 2014, the scatter index and bias of SWH is 6 Ocean Sci.Discuss., https://doi.org/10.5194/os-2018-165Manuscript under review for journal Ocean Sci. Discussion started: 26 February 2019 c Author(s) 2019.CC BY 4.0 License.

Figure 6 :
Figure 6 : Monthly evolution of the normalized scatter index of SWH from model MFWAM during 2014.Blue and red histogram bars indicate MFWAM-V3 and MFWAM-V4, respectively.

Figure 7 :Figure 8 :Figure 9 :
Figure 7 : Regional domains for the validation of IBI wave heights.

Table 1 :
Values of the model settings of MFWAM V3 and V4 versions.

Table 3 :
Scatter index and bias for year 2014 and for the two versions of MFWAM.